Cancer Cell Biology
HIF-1α and CA IX staining in invasive breast carcinomas: Prognosis and treatment outcome
Article first published online: 23 JAN 2007
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 120, Issue 7, pages 1451–1458, 1 April 2007
How to Cite
Trastour, C., Benizri, E., Ettore, F., Ramaioli, A., Chamorey, E., Pouysségur, J. and Berra, E. (2007), HIF-1α and CA IX staining in invasive breast carcinomas: Prognosis and treatment outcome. Int. J. Cancer, 120: 1451–1458. doi: 10.1002/ijc.22436
- Issue published online: 30 JAN 2007
- Article first published online: 23 JAN 2007
- Manuscript Accepted: 22 SEP 2006
- Manuscript Received: 1 MAY 2006
- Centre National de la Recherche Scientifique (CNRS)
- Ministère de l'Education
- de la Recherche et de la Technologie
- Ligue Nationale Contre le Cancer (Equipe labellisée)
- Association pour la Recherche contre le Cancer (ARC)
- CA IX;
- breast cancer;
- predictive factor;
Hypoxia stabilizes HIF-1α (Hypoxia Inducible Factor-1α), which then triggers the expression of several genes involved in many aspects of cancer progression, including metabolic adaptation, cell survival and angiogenesis. The aim of our study was to evaluate the impact of HIF-1α and CA IX (carbonic anhydrase IX) (one of its target genes) expression on prognosis and treatment outcome of patients with breast cancer. Because of the extreme O2-dependent instability of the protein, we first validated HIF-1α staining using xenograft tumours that were subjected to experimental conditions mimicking surgical clamping or sitting at room temperature under normoxic conditions after surgical excision but before fixation. Afterwards, the immunohistochemical staining of HIF-1α and CA IX was evaluated in 132 invasive breast carcinomas with a 10-year follow-up, and correlated to classical clinicopathological parameters and response to adjuvant therapy. No significant correlation was found between tumour size or nodal status and the expression of HIF-1α or CA IX. Statistically significant association was found between HIF-1α or CA IX staining and the grade, hormonal receptors loss and the presence of carcinoma in situ. Overexpression of HIF-1α and CA IX correlates with a poor prognosis in breast cancer. We show that HIF-1α is an independent prognostic factor for distant metastasis-free survival and disease-free survival in multivariate analysis. Furthermore, overexpression of HIF-1α or CA IX correlates with a poor outcome after conventional adjuvant therapy. CA IX is, however, a weaker prognostic and predictive factor than HIF-1α, and its association with HIF-1α does not modify the survival curve neither response to therapy, compared to HIF-1α alone. © 2006 Wiley-Liss, Inc.
Breast cancer accounts for the most frequent malignant tumour in women, and its incidence keeps increasing. Because of the multiplicity of oncogenic pathways, some tumours may behave extremely aggressive. Therefore, new early markers detecting this aggressivity are urgently needed.
Hypoxia is present in most of solid tumours and is associated with resistance to chemotherapy and radiotherapy as well as a more malignant phenotype.1 Stabilization of the α subunit of the Hypoxia Inducible Factor (HIF-1) is a primary response to hypoxia.2 HIF-1 is a heterodimer that consists of two subunits: the constitutively expressed HIF-1β and the rate limiting HIF-1α.3 In well-oxygenated cells, HIF-1α is a very short-lived protein (half-life less than 5 min at 21% O2) and steady-state levels are very low. In contrast, reduced oxygen availability leads to HIF-1α accumulation by relaxing its ubiquitin-proteasome degradation.4, 5, 6 Proteasome targeting is indeed triggered by the hydroxylation of two proline residues (Pro402 and Pro564) that reside within the HIF-1α Oxygen-Dependent Degradation Domain.7, 8 In addition to regulating HIF-1α stability, hypoxia affects subcellular localization, DNA binding capacity and transcriptional activation function of the HIF-1 complex. Although HIF-1 was initially identified in 1995 as a regulator of erythropoiesis by the group of G. Semenza, we know currently that HIF-1 regulates many other genes that are involved in many aspects of tumour progression, including metabolic adaptation, apoptosis resistance, angiogenesis and metastasis.9 Among these genes, there is the carbonic anhydrase 9 (ca9), which is a transmembrane glycoprotein that catalyses the transformation of carbon dioxide to carbonic acid.10 Thus, CA IX (carbonic anhydrase IX) may contribute to tumour growth and invasion via acidification of tumour microenvironment.11 Moreover, it has been recently proposed that CA IX might play an additional role, enzyme-activity-independent, in cell–cell communication.12
Because of the extreme O2-dependent instability of the HIF-1α protein, we explored the expression of this protein in xenograft tumours fixed either immediately or after storage ex vivo for varying periods of time under normoxia or hypoxia in order to validate the clinical immunohistological staining of HIF-1α. Interestingly, we showed a rather unexpected reproducible pattern of HIF-1α expression largely independent of those experimental conditions. Thereafter, we analyzed a retrospective cohort of 132 human breast carcinomas. The aim of our study was to evaluate the impact of immunohistochemically detected HIF-1α and CA IX expression on the outcome of these patients. Previous reports have provided controversial evidence concerning the impact of HIF-1α or CA IX overexpression on the behavior of human breast carcinoma.13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 Furthermore, only one study evaluated the response of tamoxifen according to HIF-1 expression and an independent report analyzed the predictive value of CA IX measured by RT-PCR.21, 24 However, no study addressed the impact on survival and treatment outcome of both together, HIF-1α and CA IX, in breast cancer. Because of the controversial results and also because we speculated that combination of HIF-1α and CA IX, one of its downstream target, might provide additional information, we performed the present study.
Material and methods
Xenograft tumours induction
Athymic mice (males aged of 4 weeks) were injected subcutaneously with 1 × 106 CAL51 cells (derived from a human breast carcinoma and cultured as previously described in Ref. 25). The animal study protocols were conducted according to approved institutional guidelines for animal use. Tumours were dissected 6 weeks after injection, fixed in formalin for immunohistochemical studies either immediately or after storage ex vivo for different times in normoxia or under hypoxia and then embedded into paraffin. Hypoxic conditions were performed by incubation of xenograft tumours in a sealed ‘Bug-Box’ anaerobic workstation (Ruskin Technologies, Biotrace). The oxygen rate in this workstation was maintained at 1–2%, with a residual gas mixture being 93–94% nitrogen and 5% carbon dioxide.
Archival surgical specimens of all patients operated on for primary invasive breast cancer at the ‘Centre Antoine Lacassagne’ in 1993 were included in this retrospective study. Thus, we studied an unselected cohort of 132 patients. The median age of patients at time of primary surgery was 62 years (range, 28–87 years). They had undergone segmental or radical mastectomy with axillary lymph node dissection. Histological types included ductal (n = 107 tumours), lobular (n = 15), colloid (n = 3), tubular (n = 2) and medullary (n = 5) adenocarcinomas. The median size of tumours was 15 mm (range, 2–100 mm). The nodal status included 48 patients with positive lymph nodes (N+) and 84 with negative lymph nodes (N−). The histological grade was based on the Scarff-Bloom and Richardson grading system modified by Elston and Ellis as grade 1 (n = 57 tumours), grade 2 (n = 44 tumours) and grade 3 (n = 10 tumours). It was not possible to score a histological grade for 21 tumours. The threshold for estrogen receptors (ER) and progesterone receptors (PR) positivity was 10 and 15 fmol/mg, respectively. None of the patients received any treatment before surgery. Nevertheless, the majority of our patients (91%) received radio-, chemo- and hormono-therapy as adjuvant treatments. The median follow-up time was 138 months (range, 11–160 months). During this period, 22 patients developed local recurrences, 26 distant metastases and there were 20 disease-related deaths. Twenty-one additional patients died from causes other than breast carcinoma.
Sections (3 μm thick) of formalin-fixed, paraffin-embedded tumour tissues were transferred to slides (X-Tra, Surgipath) and were air-dried overnight at 57°C. They were dewaxed in xylene (3 changes of 3 min) and rehydrated in alcohol (2 changes of 3 min in 100% alcohol and 2 changes of 3 min in 90% alcohol) and in distilled water (2 changes of 5 min). Then, slides were boiled in a citrate buffer (pH = 7.3) for 3 min in a pressure cooker for antigen retrieval. Immunostaining was performed manually according to the manufacturer's specifications (DAKO EnVision+System, which uses a streptavidin–biotin–horseradish peroxidase complex). Following washing in TBST (50 mM Tris pH 7.6, 300 mM NaCl, 0.1% Tween 20), endogenous peroxydase was blocked with H2O2 for 10 min. Thereafter, the slides were washed in TBST and exposed to the first antibody for 1 h 30 min at room temperature. For immunohistochemical detection of HIF-1α, we used a rabbit polyclonal antibody against HIF-1α (antiserum 2087, previously generated in our laboratory26) at a dilution of 1:500 and a mouse monoclonal anti-CA IX antibody (clone MN75, Bayer) at a dilution of 1:10,000. After 2 washes in TBST, the second antibody was incubated for 30 min at room temperature, and then rinsed with TBST. The reaction product was visualized by exposing sections to the chromogen 3,3′-diaminobenzidine (DAB) for 10 min. Sections were counterstained with hematoxylin and dehydrated in alcohol and xylene. Slides were finally mounted using the Eukitt system.
Tumour cell immunoreactivity was determined by 2 independent observers (C.T. and F.E.), who were blinded to clinical outcome. Conflicting scores were resolved at a discussion microscope. The nuclear staining of HIF-1α and the membrane staining of CA IX were noted as negative or positive (>1% of cells). The CA IX staining of stromal fibroblast-like cells was also noted (negative or positive).
We used the database management system PIGAS in association with the BMDP Statistical Software and R7.1.1. Two groups of patients were studied with regard to HIF-1α and CA IX expression (negative and positive immunoreactivity). Correlations involving HIF-1α and CA IX expression and clinicopathological parameters were performed using χ2 test (for nodal status, histological grade, hormonal receptors, and the presence of carcinoma in situ associated with the invasive tumour) or Fisher's test (in the case of age of patients and tumour size). Response to adjuvant treatments according to HIF-1α and CA IX expression was also assessed using the χ2 test. We examined specific overall survival (OS), distant metastasis-free survival (DMFS) and disease-free survival (DFS; including local or distant tumour progression). Survival was defined from the time of surgery until death of the patient or until the last follow-up visit. Deaths from a cause other than breast cancer were considered censoring events. For univariate survival analyses, we plotted curves using the Kaplan–Meier method, and differences between the curves were analyzed by the Log-rank test. Cox model's analyses including tumour size, nodal status, histological grade, HIF-1α or CA IX stainings were performed for OS, DMFS and DFS. The level of significance was 5% (p-value < 0.05).
Validation of HIF-1α immunostaining
It is well known that HIF-1α is rapidly degraded in oxygenated cells (half life less than 5 min at 21% O2), whereas hypoxia strongly and rapidly leads to the accumulation of the protein in cell cultures as well as in vivo.2, 27 Because HIF-1α stability is so tightly regulated by oxygen availability, we were concerned about the relevance of HIF-1α staining in clinical tumour samples. Indeed, artery clamping during surgical interventions (by reducing oxygen availability) might itself induce the accumulation of the protein; on the other hand, uncontrolled and inevitable variations in time until tumour fixation (by allowing reoxygenation of the tumour) might have an impact on HIF-1α protein staining. To address this question, we used xenograft tumours before starting our retrospective study. Immunohistochemical staining was performed using the anti-HIF-1α polyclonal antibody (antiserum 2087) we had previously generated and made use in our laboratory for western-blot and immunofluorescence,26 and more recently for immunohistochemical staining (Fig. 1a).
CAL51 cells (derived from a human breast carcinoma) were injected subcutaneously into nude mice and 6 weeks after injection tumours were dissected. To evaluate changes in the pattern of HIF-1α protein expression with respect to time until fixation, similar-sized tumours (100 mm3) were immediately fixed or kept at room temperature for different time periods (from 5 min and up to 3 h) before fixation in formalin and further inclusion in paraffin. In addition, to 'mimic' surgical clamping, we incubated some tumour samples under hypoxic conditions (1–2% O2) for up to 3 h at room temperature. Surprisingly, we found that the levels of HIF-1α staining in all the tumours remained unchanged (Fig. 1b). Three tumours were used for each of these experimental conditions. The same results were obtained when we kept the tumours at 37°C and when we carried out similar experiments with two independent types of xenograft tumours (data not shown). In addition, we performed an independent type of experiment using normal tissues by clamping the right kidney (the left one was the control) of 6 mice incubated at 8% O2 for 1 h. Again, the levels of HIF-1α staining in both kidneys remained unchanged. We, therefore, concluded that the storage of tissue samples either in normoxia or under hypoxic conditions has no impact on the level of HIF-1α staining and thus HIF-1α staining is not artifactual. Furthermore, validation of HIF-1α immunostaining prompted us to go on with the study of our unselected cohort of the 132 patients operated on for primary invasive breast cancer in the ‘Centre Antoine Lacassagne’ in 1993.
Expression pattern of HIF-1α and CA IX immunostainings
As expected, HIF-1α staining appeared clearly into the nucleus, whereas staining of CA IX was at the plasma membrane (Fig. 2). The topography of these stainings was diffused or confined around fibrotic or necrotic areas though necrosis was not always associated with the positivity of markers. We noticed 2 tumours showing HIF-1α and CAIX immunoreactivity in nearly all tumour cells. HIF-1α staining was detected in 59 of 132 specimens (45%). Among those, 33 tumours (56%) were HIF-1α and CA IX positive, and 26 (44%) were HIF-1α positive and CA IX negative. Thirty-eight of 132 patients (29%) showed CA IX positive immunoreactivity (including 5 CA IX positive and HIF-1α negative tumours). HIF-1α and CAIX expressions were significantly correlated (p < 0.0001). Only 1 specimen showed positive staining for both markers within the surrounding normal breast tissue.
Clinical correlations and survival analysis
No significant correlation was found between the age of patients, the size or the nodal status of the tumours and the expression of HIF-1α or CAIX (Table I, and data not shown).
|Negative (%)||Positive (%)||Negative (%)||Positive (%)|
|Grade (n = 111)||1||44 (77)||13 (23)||<0.0001||49 (86)||8 (14)||0.0004|
|2||15 (34)||29 (66)||22 (50)||22 (50)|
|3||1 (10)||9 (90)||6 (60)||4 (40)|
|Differentiation (n = 117)||Well||16 (84)||3 (16)||<0.0001||16 (84)||3 (16)||0.017|
|Moderate||33 (67)||16 (33)||38 (78)||11 (22)|
|Poor||14 (29)||35 (71)||27 (55)||22 (45)|
|Nuclear pleomorphism (n = 115)||Discrete||20 (87)||3 (13)||<0.0001||20 (87)||2 (13)||0.003|
|Moderate||41 (55)||33 (45)||53 (72)||21 (28)|
|Severe||2 (11)||16 (89)||7 (39)||11 (61)|
|Mitoses (n = 115)||Rare||58 (67)||29 (33)||<0.0001||68 (79)||18 (21)||<0.0001|
|Moderate||3 (18)||14 (82)||4 (24)||13 (76)|
|Numerous||2 (18)||9 (82)||7 (64)||4 (36)|
|Nodal status (n = 132)||N− (n = 84)||49 (58)||35 (42)||NS||60 (71)||24 (29)||NS|
|N+ (n = 48)||24 (50)||24 (50)||34 (71)||14 (29)|
|ER (n = 124)||Absent||1 (7)||13 (93)||<0.0001||2 (14)||12 (86)||<0.0001|
|Present||69 (63)||41 (37)||22 (20)||88 (80)|
|PR (n = 124)||Absent||17 (40)||25 (60)||0.01||25 (60)||17 (40)||0.02|
|Present||53 (64)||29 (36)||65 (79)||17 (21)|
|CIS (n = 132)||Absent||43 (69)||19 (31)||0.002||48 (77)||14 (23)||NS|
|Present||30 (43)||40 (57)||46 (66)||24 (34)|
Statistically significant association was found between HIF-1α or CAIX staining and the grade (p < 0.0001 and p = 0.0004, respectively) (Table I). Indeed, HIF-1α was mainly expressed in poorly differentiated tumours (71%), showing severe nuclear pleomorphism (89%) and numerous mitoses (82%), which account for poor prognostic high-grade tumours (90% of grade 3 tumours; p < 0.0001). In contrast, HIF-1α immunoreactivity was generally negative in good prognostic tumours (77% of grade 1 tumours). CA IX staining was mainly negative in low-grade tumours (85% of grade 1 tumours; p = 0.0004), well differentiated (84%), with low pleomorphism (87%) and few mitoses (79%).
Moreover, we found a significant negative correlation between HIF-1α positive staining and hormonal receptors expression: HIF-1α was mainly overexpressed in tumours without ER (p < 0.0001) or PR (p = 0.01), whereas immunoreactivity was preferentially negative when ER or PR was present (63% and 64%, respectively). CA IX staining displayed the same profile with regard to ER expression (positive in 86% of ER negative tumours, and negative in 80% of ER positive tumours; p < 0.0001), but was correlated with PR expression only for its negativity (p = 0.02) (Table I).
Finally, we also noticed a significant correlation between HIF-1α immunoreactivity (but not the one of CA IX) and the presence of carcinoma in situ associated with the invasive tumour (p = 0.002) (Table I).
When survival of patients was analyzed, we found that overexpression of HIF-1α was associated with worse OS (p = 0.0005), DMFS (p = 0.002), and DFS (p = 0.0001). Overexpression of CA IX was only significantly correlated with worse DFS (p = 0.004) (Fig. 3, and data not shown). The DFS curve of patients whose tumours were positive for HIF-1α was not significantly different to the DFS curve of patients whose tumours were positive for both markers: HIF-1α and CA IX (data not shown). CA IX positive immunoreactivity within the stroma correlated with worse OS (71% of survival versus 88% of survival, p = 0.014) (data not shown).
Finally, multivariate analyses correlating tumour size, nodal status, histological grade and HIF-1α or CA IX staining showed that (i) grade and nodal status were independent prognostic factors for OS; (ii) nodal status, HIF-1α or CA IX staining and tumour size for DMFS; (iii) HIF-1α staining, tumour size and nodal status for DFS or tumour size, nodal status, grade and marginally CA IX for DFS (Table II).
|HR||95% CI||p-value||HR||95% CI||p-value||HR||95% CI||p-value|
Association with treatment outcome
Tumour hypoxia has long been known to compromise therapy success.28 The evidence so far of an association of hypoxia-associated markers with resistance to adjuvant therapies, however, is scarce.21, 24 Thus, we correlated HIF-1α and CA IX stainings to the success of different adjuvant therapies (Tables III and IV). The occurrence of local recurrence for radiotherapy or metastasis in the case of medical treatments (chemotherapy and hormonotherapy) set out to for patients that they did not respond to adjuvant treatment. Within the ‘hormonotherapy’ group, we excluded patients who had also received chemotherapy.
|Adjuvant treatment||Total||HIF-1α||CA IX|
|Surgery + radiotherapy||57||36||21||44||13|
|Surgery + radiotherapy + chemotherapy||17||5||12||8||9|
|Surgery + hormonotherapy + chemotherapy||1||0||1||1||0|
|Surgery + radiotherapy + hormonotherapy||21||13||8||17||4|
|Surgery + radiotherapy + chemotherapy + hormonotherapy||14||10||4||11||3|
|Adjuvant treatment||HIF-1α||p-value||CA IX||p-value|
|Radiotherapy , , ,||6/64 (9%)||15/45 (33%)||0.002||13/80 (16%)||8/29 (28%)||NS|
|Chemotherapy , ,||4/15 (27%)||11/17 (65%)||0.03||9/20 (45%)||6/12 (50%)||NS|
|Hormonotherapy ,||1/17 (6%)||4/14 (29%)||NS||2/24 (8%)||3/7 (43%)||0.03|
|Medical treatment , , , ,||5/32 (16%)||15/31 (48%)||0.005||11/44 (25%)||9/19 (47%)||NS|
In our study, HIF-1α staining correlated to lack of response to radiotherapy or medical adjuvant therapy (p = 0.002 and p = 0.005, respectively) (Table IV). The same correlation (but not statistically significant) was the trend for CA IX staining.
Since breast cancer is the most frequent cancer in women, many studies have tried to identify early markers, which would be helpful to detect its aggressivity. The presence of hypoxic regions within tumours has long been reported to be associated with a poor survival.28 Because hypoxia stabilizes HIF-1α, which then triggers the expression of target genes, we decided to focus our study on HIF-1α and CA IX (one of its downstream product). Previous studies have already addressed this issue but results have been largely controversial.13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 However, this is the first time that the impact of both markers is evaluated on treatment outcome.
As HIF-1α protein level is tightly regulated by oxygen availability,2 we wanted to exclude the possibility to measure an ‘artifactual’ HIF-1α staining in tumour samples prior to start of our retrospective study. We hypothesized that time until tumour fixation might have an impact on tissue oxygen availability, and change the levels of HIF-1α staining. Wiesener et al. have previously addressed similar question but they used renal carcinomas in which pVHL is inactivated, and thus HIF-1α is constitutively expressed irrespectively of oxygen.29 To definitely validate HIF-1α immunostaining, we used xenograft tumours derived from CAL51 cells in which HIF-1α expression depends on oxygen availability. Similar-sized tumours were fixed following different situations mimicking surgical clamping or reoxygenation before fixation. In contrast to our expectation, these different experimental conditions had no effect on the levels of HIF-1α staining. We presume that the absence of irrigation ‘freezes’ exchanges within the tumoural tissues, and oxygen diffusion must be limited because of the tumour thickness. Further work will be necessary to understand the mechanisms underlying this phenomenon. Nevertheless, for the first time, our findings certify immunohistochemically detected HIF-1α staining.
In our retrospective study, the topography of HIF-1α and CA IX stainings showed 2 different patterns: (i) focal and mainly restricted around necrotic or fibrotic areas, (ii) diffuse throughout the tumour (probably controlled by mechanisms other than hypoxia). During the preparation of this paper, Vleugel et al. also reported these two types of topographies and showed a differential prognostic impact for HIF-1α.13 Indeed, perinecrotic HIF-1α overexpression was associated with a poor prognosis whereas diffuse HIF-1α overexpression resulted in a more favorable prognosis. As shown by Bos et al.,14 normal breast tissues are negative for both markers, except for one specimen in our study. This expression of HIF-1α in the normal parenchyma adjacent to the invasive carcinoma (also reported in 13, and 15) might be explained by a fast growth of the cancer, compressing the surrounding tissue and creating distant hypoxia.
Accordingly to the data reported by Vleugel et al.,13 an additional important finding of our study is the lack of strict colocalisation between the areas of HIF-1α and CA IX stainings within the tumour. It has been demonstrated, nevertheless, that ca9 is a HIF-1-dependent gene in cell cultures as well as in tumour models.30
HIF-1α and CA IX stainings were not correlated with the nodal status and the tumour size, as previously reported.13, 15-18 Only in a cohort of premenopausal stage II breast cancers HIF-1α correlated positively with tumour size and negatively with lymph-node status.24 HIF-1α overexpression was correlated with the 3 parameters setting the grade of a tumour: mitotic index, nuclear pleomorphism and differentiation status as previously reported.13, 16, 24 Accordingly, Bos et al. have shown that the positivity of HIF-1α increases during breast cancer progression and is higher in poorly differentiated lesions than in well-differentiated lesions.14 Concerning CA IX, we found that the absence of CA IX staining correlates with a low histological grade, as previously reported.18
In a preliminary study, Bos et al. found a positive correlation between ER expression and HIF-1α immunoreactivity, whereas in a larger study, a negative correlation (but not statistically significant) was the trend.14, 16 Similarly to Kronblad et al.,24 we found a significant correlation between positive HIF-1α staining and the absence of ER or PR. These results are in agreement with a recent report by Kurebayashi et al. showing that hypoxia reduces hormonotherapy responsiveness mostly by decreasing the expression of hormonal receptors.31 The loss of hormonal receptors, which is a characteristic of the most aggressive tumours, may be, indeed, a consequence of the expression of HIF-1α. Alternatively, the loss of hormonal receptors, being associated with tumour growth, may favor the emergence of hypoxic areas and therefore the expression of HIF-1α. In the case of CA IX immunostaining, we also found a negative correlation with ER and PR such as Chia et al.18
We also analysed the effect of positivity in HIF-1α and CA IX stainings on survival (specific overall, metastasis-free and disease free survivals) of patients in our cohort. We found a strongly significant correlation between the expression of HIF-1α and the 3 types of survival with a median follow-up of 138 months. In contrast, CA IX was significantly correlated to worse DFS only. Between the two markers we have evaluated in this study, HIF-1α positive immunoreactivity was more significant than CA IX staining. Furthermore, HIF-1α is an independent prognostic factor for DMFS and DFS, but not for OS. Gruber et al. reported similar results in a series of 77 N+ patients (36 months follow-up) in the subgroup of the 55 smallest tumours (T1/T2 according to the TNM classification).17 In contrast, Schindl et al.,15 in a series of 206 N+ patients, and Bos et al.,16 in the case of a subgroup of the 81 N− patients among the 153 patients included, showed HIF-1α as an independent prognostic factor for OS and DFS. More recently, Kronblad et al. showed HIF-1α as an independent prognostic factor for DFS when analyzing lymph node-positive patients as well as when excluding grade 3 tumours.24 Concerning CA IX staining in tumour cells, Chia et al., in a series of 103 patients with a mean follow-up of 74 months, found significant differences on the OS and DFS.18 In contrast, Colpaert et al. (94 patients, 91 months of follow-up) showed that the expression of CA IX in the stroma was associated with worse survivals,19 as in our study, whereas Tomes et al., in a series of 53 patients (50 months follow-up), found the expression of CA IX in the stromal fibroblasts to be associated with a better survival.20 All these discrepancies about the role of HIF-1α and CA IX on survival are probably linked to the lack of power of statistical studies, particularly when the population is divided into too small subgroups. Differences could also be explained by the use of antibodies showing different sensitivity and specificity. We speculated the assessment of HIF-1α in combination with downstream regulated genes, such as ca9, might provide additional information in terms of prognosis. In contrast with our hypothesis, CA IX association with HIF-1α does not modify the DFS curves compared to HIF-1α alone.
Finally, we are the first to show that both markers, and particularly HIF-1α, are associated with the lack of response to radiotherapy and medical therapy. Since only 2 papers have analyzed the potential value of CA IX (using RT-PCR)21 and HIF-1α (using tissue microarray)24 for treatment outcome in breast cancer, we evaluated the predictive value of both markers in our cohort. Indeed, Kronblad et al.24 evaluated the response to Tamoxifen in premenopausal women according to HIF-1α, which did not appear to be a predictive marker. In our study, HIF-1α seems to predict local recurrence and metastases occurrence even if radiotherapy and chemotherapy were used as adjuvant treatments (we can not conclude about hormonotherapy because there were not enough cases). Therefore, HIF-1α could be used to discriminate patients who are more likely to be resistant to conventional adjuvant therapy and benefit with specific therapeutics targeting HIF-1α. It is, however, difficult to draw conclusions on the predictive value of a marker from retrospective studies. Larger prospective randomized studies are necessary to evaluate definitively the impact of CA IX and above all HIF-1α, which is a stronger prognostic and predictive factor, in breast cancer progression and treatment outcome.
We thank Dr. Jan Zavada, Dr. Silvia Pastorekova and Dr. Jaro Pastorek for kindly providing the mouse monoclonal anti-CA IX antibody. We thank Ms. Marie-Catherine Pellegrin, Ms. Jacqueline Bof and all the laboratory members for their help and support.
- 24Hypoxia inducible factor-1α is a prognostic marker in premenopausal patients with intermediate to highly differentiated breast cancer but not a predictive marker for tamoxifen response. Int J Cancer 2006; 118: 2609–16., , , , .