CA IX is an Independent Prognostic Marker in Premenopausal Breast Cancer Patients with One to Three Positive Lymph Nodes and a Putative Marker of Radiation Resistance

Purpose: Hypoxia in breast cancer is associated with poor prognosis and down-regulation of the estrogen receptor. Carbonic anhydrase IX (CA IX) is a hypoxia-inducible gene that has been associated with poor outcome in many epithelial cancers. Previous studies of CA IX in breast cancer have been carried out on mixed cohorts of premenopausal and postmenopausal patients with locally advanced disease and varying treatment regimens. We examined the potential prognostic and predictive role of CA IX in premenopausal breast cancer patients.

Experimental Design: Using tissue microarrays, we analyzed CA IX expression in 400 stage II breast cancers from premenopausal women. The patients had previously participated in a randomized control trial comparing 2 years of tamoxifen to no systemic adjuvant treatment. Median follow-up was 13.9 years.

Results: CA IX expression correlated positively with tumor size, grade, hypoxia-inducible factor 1α, Ki-67, cyclin E, and cyclin A2 expression. CA IX expression correlated negatively with cyclin D1, estrogen receptor, and progesterone receptor. CA IX expression was associated with a reduced relapse-free survival (P = 0.032), overall survival (P = 0.022), and breast cancer–specific survival (P = 0.005). Multivariate analysis revealed that CA IX was an independent prognostic marker in untreated patients with one to three positive lymph nodes (hazard ratio, 3.2; 95% confidence interval, 1.15-9.13; P = 0.027).

Conclusion: CA IX is marker of poor prognosis in premenopausal breast cancer patients and it is an independent predictor of survival in patients with one to three positive lymph nodes. As all these patients received locoregional radiation therapy, CA IX may be associated with resistance to radiotherapy.

Hypoxic regions are common in solid organ tumors, including breast tumors. Tumor hypoxia is primarily a pathophysiologic consequence of a disturbed microcirculation as aggressive tumors rapidly outgrow their blood supply (1) and has been associated with several aspects of aggressive tumor biology including increased proliferation, invasion, and metastasis (2, 3). There is also a growing body of evidence suggesting that tumor hypoxia directly or indirectly contributes to resistance to standard radiotherapy and chemotherapy, and subsequently to poorer clinical outcome (4). Therefore, assessment of tumor hypoxia may provide useful prognostic and predictive information.

Tumor cells adapt to hypoxic conditions by stabilizing the hypoxia-inducible factor 1α (HIF-1α) protein, which is degraded in cells at normoxia via the ubiquitin-proteasome pathway, mediated by the ubiquitin ligase von Hippel-Lindau protein (5). Under hypoxic conditions, von Hippel-Lindau can no longer bind to HIF-1α, leading to the stabilization and accumulation of HIF-1α protein (6). HIF-1α up-regulates a wide range of genes including glucose transporters, glycolytic enzymes, angiogenic growth factors, and carbonic anhydrases (7). The increasing number of HIF-1α target genes being identified should provide us with further insights into the molecular mechanisms of tumor hypoxia and HIF-1α activation (8).

Whereas HIF-1α is the master regulator under hypoxic conditions, there is growing evidence that other factors are necessary for a complete hypoxic response (9). In fact, it is becoming apparent that HIF-1α activation is only one facet of the hypoxic response and that other transcription factors or cofactors are activated under hypoxic conditions that may contribute to the hypoxic phenotype (9).

Carbonic anhydrase IX (CA IX) is a membrane-bound glycoprotein the expression of which has been shown to be up-regulated by HIF-1α (8) and by high cell density via the phosphatidylinositol 3-kinase pathway (10). CA IX belongs to a group of zinc-containing metalloenzymes that catalyze the reversible hydration of carbon dioxide to carbonic acid. To date, 14 distinct carbonic anhydrase isozymes involved in gaseous exchange, ion transport, and pH balance have been identified with varying subcellular localizations and tissue distributions. Carbonic anhydrases are predominantly associated with differentiated cells and fulfill specialized roles in various tissues and organs (11). In normal tissue, CA IX expression is restricted to gastric mucosa (12, 13). However, CA IX expression has been shown to be expressed in several tumors of epithelial origin including carcinomas of the uterine cervix (12), renal cell carcinoma (14), esophageal cancer (15), non–small-cell lung carcinoma (16), colorectal cancer (17), breast cancer (18), and head and neck cancer (19).

In vitro studies have indicated that CA IX plays a role in the growth and survival of tumor cells; however, it does not seem to play a role in invasion (20). Retrospective studies have shown that CA IX has prognostic value in renal cell carcinoma, cervical cancer (21), nasopharyngeal cancer (22), lung cancer (23), and breast cancer (24–26). In addition, CA IX has been implicated in resistance to radiotherapy and is a therapeutic target for several novel inhibitors (27, 28).

Following reanalysis of the seminal DNA microarray data published by van't Veer et al. in 2002 (29), we identified CA IX as a gene that was specifically up-regulated in breast tumors with poor prognosis.⁶

We decided to further investigate these findings at the protein level. Breast cancer studies to date examining CA IX protein expression have been confined to small cohorts of premenopausal and postmenopausal patients with early and advanced disease and receiving various forms of treatment (24–26). Here, we examined the prognostic and predictive value of CA IX expression in tumors from premenopausal patients with stage II (pT₂ N₀ M₀, pT_1-2 N₁ M₀) invasive breast cancer. These patients had participated in a prospective randomized trial of 2 years of adjuvant tamoxifen versus no systemic treatment. Locoregional radiotherapy was given to lymph node–positive patients. We therefore investigated the predictive value of CA IX expression in relation to tamoxifen response and resistance to radiotherapy.

Study design. Between 1984 and 1991, 564 premenopausal women with primary breast cancer in the South and Southeast regions of Sweden were enrolled in a multicenter clinical trial (30) and randomly assigned to either 2 years of adjuvant tamoxifen (n = 276) or a control group (n = 288). The aim of this study was to examine the effect of tamoxifen on recurrence-free survival (RFS) and the study has been described in detail elsewhere (31). RFS considered local, regional, distant recurrences and breast cancer–specific death, but not contralateral breast cancer. The inclusion criteria were premenopausal patients or patients younger than 50 years, with stage II (pT₂ N₀ M₀, pT_1-2 N₁ M₀) invasive breast cancer treated by modified mastectomy or breast-conserving surgery with axillary lymph node dissection. Postoperative radiotherapy (50 Gy) was administered after breast-conserving surgery and all lymph node–positive patients received locoregional radiotherapy. Less than 2% of the patients received adjuvant systemic chemotherapy. The median follow-up time for patients without breast cancer events was 13.9 years. The study was approved by the ethical committees at Lund and Linköping Universities.

The results of the trial have previously been described and the trial has been has been included in the Oxford meta-analysis (32). Briefly, 2 years of adjuvant tamoxifen increased RFS in treated patients [hazard ratio (HR), 0.77; 95% confidence interval (95% CI), 0.60 to 0.98; P = 0.03]. The beneficial effect of 2 years of adjuvant tamoxifen in terms of RFS was restricted to 385 patients with hormone receptor–positive [estrogen receptor (ER) and/or progesterone receptor (PR) positive] tumors. Cox multivariate analysis revealed PR status as the most significant predictor of tamoxifen response.

Cell lines. All cell lines (MDA-MB-231, MCF-7, T47D, SKBR3, Hs578T, and BT474, and HeLa) were obtained from the European Collection of Cell Cultures (Salisbury, Wiltshire, United Kingdom). All cell lines, except for SKBR3, were grown in DMEM (Sigma, St Louis, MO) supplemented with 10% FCS (Invitrogen, Carlsbad, CA), l-glutamine (2 μmol/L), penicillin (50 IU/mL), and streptomycin sulfate (50 μg/mL). SKBR3 cells were grown in McCoy's 5A Medium supplemented with 10% FCS (Invitrogen), l-glutamine (2 μmol/L), penicillin (50 IU/mL), and streptomycin sulfate (50 μg/mL). Cells were maintained in humidified air with 5% CO₂. Hypoxic conditions were generated in a hypoxic chamber (Coy Labs, Grass Lake, MI) with 0.1% O₂, 5% CO₂, and balance N₂. All cell lines were free of Mycoplasma contamination.

Western blot analysis. Protein was extracted from cells grown to 80% confluence with radioimmunoprecipitation assay buffer [20 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 1% NP40, 0.5% sodium deoxycholate, 1 mmol/L EDTA, 0.1% SDS]. Protein levels were determined using the bicinchoninic acid method (Pierce, Rockford, IL). Samples containing 50-μg aliquots of protein were separated on a 10% Bis-Tris gel (Invitrogen) by SDS-PAGE under reducing conditions. After electrophoresis, proteins were transferred to a polyvinylidene fluoride membrane. Membranes were blocked in 5% nonfat milk for 1 hour. CA IX was detected using the mouse monoclonal anti-human CA IX antibody M75 (obtained as a generous gift from Dr. Oosterwijk, Department of Urology, University Medical Center, Nijmegen, the Netherlands) at a dilution of 1:50. Membranes were washed and incubated for 1 hour with horseradish peroxidase–conjugated antimouse immunoglobulin. Enhanced chemiluminescence plus (Amersham Biosciences, Buckinghamshire, United Kingdom) was used for detection. Membranes were stripped and reprobed with an anti-β-actin antibody (Abcam, Cambridge, United Kingdom) at a dilution of 1:5,000 to provide a loading control. All steps were carried out at room temperature. Western blotting experiments were done in triplicate to confirm reproducibility of results.

Cell pellet arrays. Cell lines were fixed in 4% formalin and processed in gradient alcohols. Cell pellets were cleared in xylene and washed multiple times in molten paraffin. Once processed, cell lines were arrayed in duplicate 0.6-mm cores using a manual tissue arrayer (Beecher Instruments, Inc., Sun Prairie, WI) and immunohistochemistry was done on 5-μm sections.

Tissue microarrays. Five hundred paraffin-embedded tumor specimens were used for tissue microarray (TMA) construction. Areas representative of invasive cancer were marked on H&E-stained slides and the TMA was constructed using an automated tissue arrayer (ATA-27, Beecher). Two 0.6-mm cores were extracted from each donor block and assembled in a recipient block. Recipient blocks were limited to ∼200 cores each. In general, cores were taken from the peripheral part of the tumor in cases where the tumor had well-defined borders. In more diffusely growing tumors, areas with the highest tumor cell density were primarily targeted. Necrotic tissue was avoided. For immunohistochemistical analysis (see below), 4-μm paraffin sections were used.

Immunohistochemical analysis. Sections of cell pellet arrays or TMAs were deparaffinized in xylene and rehydrated through descending concentrations of ethanol. Heat-mediated antigen retrieval was not required before immunostaining with the CA IX antibody. For all other antibodies, heat-mediated antigen retrieval was done by microwave treatment for 2 × 5 minutes in a citrate buffer before being processed either in the Ventana Benchmark system (Ventana Medical Systems, Inc., Tucson, AZ) using prediluted antibodies to ER (anti-ER, clone 6F11), PR (anti-PR, clone 16), and Her2 (pathway CB-USA, 760-2694) or in the Dako Techmate 500 system (Dako, Glostrup, Denmark) for CA IX (1:2,000, M75), HIF-1α (1:500, NB100-123H2; Abcam), Ki-67 (1:200, M7240; Dako), vascular endothelial growth factor (VEGF)-A (1:400, A20; Santa Cruz Biotechnology, Santa Cruz, CA), cyclin E (1:50, HE12; Santa Cruz Biotechnology), and cyclin D1 (1:100, M7155; Dako). The CA IX primary antibody was replaced with an antimouse immunoglobulin G isotype control to act as a negative control.

Before construction of the TMA, 500 primary tumors were reevaluated and graded according to Elston Ellis scoring system. All breast carcinomas were reclassified according to the WHO criteria as one of the following: ductal (n = 411), lobular (n = 43), mucinous (n = 3), tubular (n = 5), medullary (n = 25), lobular and ductal (n = 1), ductal carcinoma in situ/microinvasive (n = 5), other or not classified (n = 7). The study has been approved by the ethical committees at Linköping and Lund Universities.

Evaluation of immunohistochemical staining. CA IX expression levels in tumor specimens were evaluated by two investigators and scored as positive or negative, with any membranous staining around tumor cells classified as positive. All discordant cases were reevaluated and a consensus was reached. Of note, stromal expression of CA IX was not included as a variable in this study. HIF-1α protein was evaluated according to the fraction of positively stained nuclei; cytoplasmic staining was not included in the analysis. Cores expressing >2% positive nuclei staining were considered positive.

ER and PR were scored as positive if the fraction of positive tumor cells was >10%, which is in line with current clinical recommendations. Cyclin D1 and cyclin E were scored according to the nuclear fraction using the following categories: 0%, 1% to 25%, 26% to 50%, 51% to 75%, and 76% to 100%. Ki-67 and cyclin A2 staining was scored according to nuclear fraction using the following categories: 0% to 1%, 2% to 10%, 11% to 25%, 26% to 50%, and 51% to 100%. VEGF-A and VEGF receptor 2 staining was divided into two categories of low to intermediate staining (0-2+) versus high staining intensity (3+). Her2 staining was evaluated according to a standard protocol (HercepTest) and scored as four intensities [i.e., 0-3; these scorings were divided into two groups with normal/weak (0-2) Her2 expression and overexpression (3+)]. Her2 gene amplification was determined using fluorescence in situ hybridization according to standard protocols described elsewhere (30).

Statistical analysis. Differences in distribution of clinical data and tumor characteristics between CA IX–negative and CA IX–positive tumors were evaluated using the χ² test. Kaplan-Meier analysis and the log-rank test were used to illustrate differences between RFS, overall survival (OS), breast cancer–specific survival (BCSS), and metastases-free survival according to CA IX expression. Cox regression proportional hazard models were used to estimate the relationship to BCSS of CA IX, tumor size, Her2, lymph node status, Nottingham Harris Grade, PR, and Ki-67 in the entire cohort; the analysis was also done on untreated patients with one to three positive lymph nodes. ER was omitted from this model due to a previously shown link between hypoxia and hormone receptor down-regulation. All calculations except Fisher's exact test were done using SPSS version 11.0 (SPSS, Inc., Chicago, IL). Fisher's exact test was done using SAS v.8 (SAS, Inc., Cary, NC). P < 0.05 was considered statistically significant.

Specificity of CA IX antibody. The specificity of the anti–CA IX antibody was confirmed by Western blot analysis. HeLa cells grown under hypoxic conditions were used as a positive control. The anti–CA IX antibody recognized two strong bands at 50 and 54 kDa in hypoxic HeLa cells and normoxic MDA-MB-231 cells, but did not detect CA IX protein expression in any of the other breast cancer cell lines (Fig. 1). To confirm our findings, we did immunohistochemistry on the same cell lines. This revealed complementary results with MDA-MB-231 showing CA IX protein expression whereas all the other breast cancer cell lines did not express the protein (Fig. 1).

Distribution of CA IX. Tumor samples were available from 500 of 564 (89%) patients included in the randomized study, as previously discussed (31). Following antibody optimization and staining, it was possible to evaluate the expression of CA IX protein in 400 of the 500 (70.9%) tumors represented on the TMA. One hundred tumors were not available for analysis due to core loss during sectioning and staining. To evaluate our study for any potential selection bias, baseline clinicopathologic characteristics from both the evaluated or “CA IX known” cohort (n = 400) and the unevaluated or “CA IX unknown” cohort (n = 100) are presented in Table 1. This illustrates that the clinicopathologic characteristics were well matched in both cohorts except that there was a larger proportion of grade 3 tumors in the cohort evaluated for CA IX. There was also a larger proportion of PR-negative and Her2-overexpressing tumors in the evaluated cohort. We also saw a larger proportion of missing or unknown cyclin E (156 versus 110) and cyclin A2 (124 versus 83) tumors in the unevaluated cohort. Supplementary Table S1 shows that the distribution of patient and tumor characteristics was similar in both arms of the original randomized study and the present study, verifying the absence of a selection bias in the randomization process, as well as in the selection of the present study cohort.

Membranous staining of CA IX was found in 42 (11%) of the evaluated breast carcinomas, these tumors being considered CA IX positive. Regarding histologic subtype, CA IX expression was seen exclusively in invasive ductal and medullary carcinoma (Table 1). CA IX expression was absent in lobular, tubular, and mucinous carcinoma. CA IX expression correlated positively with tumor size and grade, as well as with HIF-1α, Ki-67, cyclin A2, and cyclin E expression, and negatively with ER, PR, and cyclin D1 expression. There was no significant association between CA IX expression and age of patient, nodal status, HER2 overexpression, and VEGF-A or VEGF receptor-2 expression (Table 1).

CA IX expression and survival within the evaluated cohort (n = 400). In the evaluated cohort (n = 400), CA IX expression was associated with a significantly worse RFS (P = 0.032), OS (P = 0.022), and BCSS (P = 0.005; Fig 2A-C).

Univariate analysis revealed that the most significant (P = 0.005) association between CA IX and survival was seen when using BCSS as an end point; we therefore conducted a Cox multivariate regression analysis of BCSS in the evaluated cohort. This model included all factors that displayed a significant association with BCSS in univariate analysis (i.e., CA IX, tumor size, cyclin E, Ki-67, Her2, PR, grade, and lymph node status; Table 2). Multivariate analysis revealed that CA IX was a significant predictor of BCSS in this cohort (HR, 1.75; 95% CI, 1.03-3.00; P = 0.04) along with tumor size (HR, 1.02; 95% CI, 1.01-1.04; P = 0.012), lymph node status (HR, 2.33; 95% CI, 1.41-3.84; P = 0.001), and grade (HR, 1.67; 95% CI, 1.03-2.70; P = 0.037). ER status was omitted from this multivariate model due to an association previously documented relating hypoxia to down-regulation of the ER (33–35) and also a proposed role for CA IX in this pathway (discussed later).

CA IX expression in the tamoxifen arm of the cohort (n = 199). One of the aims of this study was to investigate the role of CA IX expression in predicting response to tamoxifen. As shown in Table 1, there was a similar distribution of CA IX–positive tumors in the tamoxifen-treated arm (n = 199) and the control arm (n = 201) of the evaluated tumors (P = 0.345). Twenty-four (57%) of the tumors that expressed CA IX were in the control group and 18 (43%) were in the tamoxifen-treated group.

In an effort to examine the role of CA IX as a predictor of tamoxifen response, we examined RFS in tamoxifen-treated patients (n =199). CA IX expression was associated with a significantly worse RFS (P = 0.007) in treated patients (Fig. 2D); however, when we restricted our analysis to ER-positive tumors, this effect was lost (Fig. 2E). This could be attributed to the fact that CA IX was almost exclusively seen in ER-negative tumors, and there were only three ER/CA IX–positive tumors in the treated cohort. Therefore, our data set lacks sufficient power to determine whether CA IX is a predictor of tamoxifen response.

CA IX expression in the untreated evaluated cohort. The prognostic value of CA IX expression was then analyzed in the untreated cohort (n = 201). As previously stated, 24 (12%) of the tumors in this subset exhibited expression of CA IX protein. There was no significant difference in RFS, OS, or BCSS (Fig. 3A) in the untreated cohort. These findings differed from our findings in relation to the treated cohort (n = 199) where we saw a difference in RFS (P = 0.007) regarding CA IX expression (Fig. 2E). This difference could be attributed to the fact that there was a significantly higher proportion of pT₂ tumors (69% versus 58%) in the treated arms in both the original randomized trial and the current study (Supplementary Table S1).

Subset analysis of the untreated cohort (n = 201) revealed a significant difference in BCSS (P = 0.035) in lymph node–positive patients (Fig. 3B). Furthermore, the significance of CA IX expression in relation to BCSS increased in patients with one to three positive lymph nodes (P = 0.013; Fig. 3C). This was not evident in patients with advanced nodal disease (Fig. 3D).

To further explore the prognostic significance of CA IX expression in patients with one to three positive lymph nodes, a Cox proportional hazard multivariate model was constructed from known prognostic factors including tumor size, tumor grade, PR, Ki-67, and cyclin E expression. Multivariate Cox regression analysis revealed that CA IX was an independent predictor of BCSS (HR, 3.2 95% CI, 1.15-9.13; P = 0.027; Table 2).

CA IX is associated with a reduced RFS at the mRNA level in an independent cohort. To confirm our findings, we examined CA IX mRNA expression in an independent but analogous data set (36). This data set has been described in detail elsewhere (36). Briefly, breast tumors from 295 consecutive women with breast cancer were selected from the fresh-frozen tissue bank of the Netherlands Cancer Institute. Inclusion criteria for this study included 52 years of age or younger at primary diagnosis, invasive breast carcinomas that were <5 cm in diameter (pT₁ or pT₂), and negative apical axillary lymph nodes. The tumors used were diagnosed between 1984 and 1995. All patients were treated by modified radical mastectomy or breast-conserving surgery, including dissection of the axillary lymph nodes, followed by radiotherapy if indicated. One hundred fifty-one patients had lymph node–negative disease and 144 had lymph node–positive disease. Ten of the 151 patients who had lymph node–negative disease and 120 of the 144 who had lymph node–positive disease had received adjuvant systemic therapy consisting of chemotherapy (90 patients), hormonal therapy (20), or both (20). The median follow-up was 6.7 years (range, 0.05-18.3 years). Thus, this data set was very similar to the patient cohort used to construct the TMA in our study.

The data set was downloaded from a publicly available website.⁷

The log ratio of gene expression values was used without further transformation. We classified tumor samples as high or low CA IX expressers based on array hybridization levels. CA IX expression was increased at the mRNA level in 50 of the 295 (17%) tumors. High levels of CA IX expression were associated with a significantly worse probability of remaining free of distant metastases (P = 0.008) and significantly worse OS (P = 0.002; Fig. 3E and F).

We also examined the relationship between CA IX expression in this data set and a number of clinicopathologic variables (Table 3). CA IX expression was significantly associated with pT₂, high grade, and ER-negative tumors, thus confirming our previous findings. We also classified tumors into molecular subtypes using the criteria described by Sorlie et al. (37). We found that 48% of CA IX–expressing tumors were of the basal subtype (Table 3), which has previously been associated with poor prognosis (33). Finally, we found that CA IX expression was significantly associated with the 70-gene prognostic signature derived by van't Veer et al. (29). Taken together, these findings provide further evidence that CA IX is indeed a poor prognostic marker.

The aim of this study was to evaluate CA IX as a prognostic marker in stage II premenopausal breast cancer patients in relation to RFS, OS, and BCSS using a randomized study cohort. The prognostic effect of CA IX in breast cancer has not previously been studied in a randomized cohort of patients with stage II disease that had not receive systemic adjuvant therapy. Colpaert et al. (26), however, examined the prognostic significance of CA IX in a small group of patients with stage I breast cancer who did not receive adjuvant therapy (discussed below). In addition, the predictive power of CA IX in relation to tamoxifen response has not been investigated in a randomized trial.

Within the entire cohort, 11% (n = 42) of tumors showed membranous CA IX staining. Although some studies have reported a higher proportion of CA IX–positive tumors, their cohorts consisted of locally advanced tumors, with a higher proportion of hypoxic and necrotic regions (24, 26, 38). Other studies have reported similar levels of CA IX expression to ours (39). In our study, use of the TMA platform could lead to underestimation of certain markers, especially hypoxic markers, as the TMA was constructed using cores from specifically nonnecrotic areas. Therefore, to confirm our findings, we examined the expression of CA IX in a comparable cohort of 295 early-stage tumors used in a DNA microarray study by van de Vijver et al. (36). In that study, 17% (n = 50) of tumors in this cohort expressed high levels of CA IX mRNA. This finding supports our TMA data and suggests that the prevalence of CA IX expression is in the 10% to 20% range in early-stage breast cancer.

Consistent with CA IX being a known HIF-1α target gene (8), we found that CA IX expression correlated with HIF-1α expression (P = < 0.0001). However, our results indicate that HIF-1α expression alone may not be sufficient for CA IX gene activation as many HIF-1α-positive tumors (28 of 79) did not express CA IX. This finding has also been observed in other studies, especially when examining HIF-1α expression in nonnecrotic tumors (39). As our TMA was constructed from nonnecrotic areas, our results are in agreement with these previous findings. Similarly, 15 of 36 tumors were CA IX positive but HIF-1α negative, indicating that CA IX may be regulated by other factors, independent of HIF-1a activation. This is in agreement with previous findings showing that CA IX is regulated by high cell density via the phosphatidylinositol 3-kinase pathway (10) and by the mitogen-activated protein kinase pathway during both normoxia and hypoxia (40). Indeed, we found constitutive expression of CA IX in the breast cancer cell line MDA-MB-231 under normoxic conditions, indicating a hypoxic-independent mechanism of CA IX expression (Fig. 1). Taken together, these observations suggest that our study is examining not only hypoxia-regulated CA IX expression but also pathway-driven CA IX expression.

Resistance to endocrine therapy with tamoxifen is a major clinical problem and hormone receptor status remains the only accurate marker of therapeutic response. In an attempt to understand this problem, researchers have developed many in vitro models of tamoxifen resistance and evaluated numerous predictive markers. Hypoxia has been linked to ER down-regulation in breast cancer cell lines, as well as in primary breast cancer (33–35). However, the mechanism of this down-regulation is unknown. Hypoxia has been shown to activate the RAS/RAF/mitogen-activated protein kinase signal transduction pathway, and hyperactivated mitogen-activated protein kinase has been shown to down-regulate ER in both a nuclear factor κB–dependent manner (41) and via activation of extracellular signal–regulated kinase 1/2 (42). These findings would suggest that hypoxic markers may be useful predictors of tamoxifen resistance. However, it has recently been shown that HIF-1α expression is not associated with tamoxifen response (43). Similarly, we did not observe any association between CA IX and tamoxifen response (Fig. 2). This may be due to the fact that CA IX is associated with ER-negative tumors. We have shown this in two similar but independent data sets (Tables 1 and 3). In addition, whereas HIF-1α has a short half-life under normoxic conditions (44) and has been shown to be expressed in ER-positive tumors (43), CA IX expression is evident up to 96 hours after reoxygenation, thus indicating that CA IX is a stable marker of current or previous chronic or intermittent hypoxia (45). We therefore hypothesize that chronic hypoxia, as opposed to acute hypoxia, may play an important role in hypoxia-modulated ER down-regulation. We also believe that CA IX may be an intervening variable in ER down-regulation and, thus, we did not include ER in our multivariate Cox regression model.

In our study, we found no relationship between CA IX expression and VEGF-A or VEGF receptor 2. These results are in agreement with recent findings showing no association between HIF-1α and VEGF-A expression (43), indicating that VEGF is a marker of angiogenesis but not of hypoxia. In addition, VEGF expression has been shown to be regulated by estrogen in breast cancer (46). VEGF receptor 2 has also been shown to be an independent predictor of impaired tamoxifen response (30), further supporting our finding that CA IX expression is not associated with tamoxifen response.

CA IX is considered to play an important role in the growth and survival of tumor cells under normoxic and hypoxic conditions (20). In vitro studies have shown that CA IX expression is cell density dependent, which also seems to be pathway driven, as opposed to being regulated by hypoxia (8). The significant association between CA IX and tumor size further supports this finding. CA IX expression was also correlated with proteins associated with S and G₂-M phases of the cell cycle. CA IX expression was positively correlated with Ki-67, cyclin E, and cyclin A2 expression and was negatively associated with cyclin D1, which mainly functions in early G₁-phase (Table 1). These findings, combined with the association between CA IX and high grade, indicate a proliferative phenotype with an inherently poor prognosis. In contrast to previous findings, we found no relationship between CA IX and Her2 expression (34). However, small numbers of samples, many of which contained necrotic regions, were previously used, which may account for this discrepancy. Our results are further supported by the finding that in 295 independent breast tumor samples, CA IX was predominantly expressed in the basal molecular subtype, which is also associated with ER-negative tumors (37) and not with the Her2 subtype (Table 3).

As expected, we found that CA IX expression was associated with a poor prognosis. When we examined the entire cohort (n = 400), CA IX was associated with a reduced RFS, OS, and BCSS (Fig. 2). Multivariate analysis confirmed this finding (Table 2). These data confirm and strengthen the findings of others, suggesting that CA IX expression is linked to an aggressive phenotype (18, 24–26). In contrast to our findings, previous studies of CA IX in breast cancer have been carried out on mixed cohorts of premenopausal and postmenopausal women with varying disease stages and treatment regimens (24, 25). One study examined CA IX expression in a small nonrandomized cohort of patients with stage I disease and found that it was associated with a poor outcome (26). To our knowledge, ours is the first study of CA IX expression in premenopausal women and of exclusively stage II tumors. It is also important to emphasize that the median tumor size in this study was 24 mm. Previous studies of CA IX have involved large locally advanced tumors with associated necrosis, which may have given somewhat skewed results as these patients would have had an inherently poor prognosis and overexpression of a hypoxia-regulated gene such as CA IX would not be surprising.

Therefore, we believe that, until now, the true prognostic value of CA IX had not been fully investigated. Half of the patients in our study received no adjuvant systemic therapy, and thus treatment bias is eliminated. Therefore, by examining the expression of CA IX in an untreated cohort of patients, we were able to investigate the pure prognostic power of CA IX. We initially saw no difference in survival related to CA IX expression in the control cohort. Although CA IX expression did not correlate with lymph node status, we found that CA IX expression was associated with a significantly worse RFS, OS, and BCSS in patients with one to three positive lymph nodes (Fig. 3C). Cox multivariate analysis of the patients in the control arm of the trial revealed CA IX expression to be an independent prognostic factor of breast cancer–specific death in this subgroup.

It should be noted that all lymph node–positive patients in this cohort received locoregional radiotherapy. The involvement of four or more nodes, as indicated by the tumor-node-metastasis system, is currently used to define patients who will require axillary radiotherapy and to determine the type and aggressiveness of adjuvant systemic therapy (47). Treatment of patients with one to three positive nodes, which accounts for ∼20% of all breast cancer cases (48), remains controversial. The data presented in this study suggest that CA IX is not only a marker of poor prognosis in this subgroup of patients but can also be a marker of resistance to external beam radiotherapy as this is the only treatment these patients received. However, to reliably measure radiation resistance, we would have to use RFS as an end point and compare our findings to a similar control cohort who did not receive radiotherapy. We, however, found a trend towards reduced RFS in patients but this did not reach significance (data not shown).

Hypoxia and anemia are well-recognized causes of resistance to ionizing radiation. A number of approaches have been employed to attempt to overcome this resistance, ranging from the use of hyperbaric oxygen to synthetic sensitizers such as misonidazole, to the development of strategies such as accelerated radiotherapy with carbogen and nicotinamide (ARCON; ref. 49). Whereas ARCON-based approaches are currently in phase III trials, synthetic compounds, especially the nitroimidazoles, have not gained clinical acceptance, initially because of their toxicity and later because of doubts about the effectiveness of the newer generation of less toxic drugs, which have produced disappointing results (50).

As these approaches are based on reversing acute hypoxia and increasing tumor perfusion, it may be that they will not improve radiosensitivity, as chronic hypoxia will induce genes such as CA IX that will not be affected by an acute return to normoxia. Thus, another approach to improve the radiosensitivity of such tumors may be to use a selective CA IX inhibitor that may improve tumor radiosensitivity.

In summary, we have shown that CA IX is associated with a worse RFS, OS, and BCSS in premenopausal patients with stage II tumors. Furthermore, CA IX was associated with markers of aggressive disease including histologic grade, tumor size, Ki-67, cyclin E and cyclin A2, and loss of ER and PR. We were unable to show an association between CA IX expression and tamoxifen treatment response, as our study lacks the necessary power to assess the predictive value of CA IX in tamoxifen-treated patients. CA IX was the only independent prognostic marker in patients with one to three positive lymph nodes. Finally, given its subcellular localization and the generation of specific inhibitors, it is possible to speculate that CA IX is a new therapeutic target, the inhibition of which may improve the efficiency of radiotherapy in certain breast tumors. However, this will require further investigation.

We thank Dr. Oosterwijk for his generous donation of the anti–CA IX antibody, G250/M75; Elise Nilsson for excellent technical assistance; and Sten Thorstenson and Monika Haglund for Her2 expression and Her2 fluorescence in situ hybridization analyses and Nottingham Harris Grade evaluation. We also thank Professor Des Higgins, Dr. Aedhin Culhane, and Ms. Ailis Fagan, UCD Conway Institute for their assistance.