Overexpression of hypoxia-inducible factor HIF-1α predicts early relapse in breast cancer: Retrospective study in a series of 745 patients

Authors


Abstract

Hypoxia-inducible factor-1α (HIF-1α) is a transcription factor that is involved in tumour growth and metastasis by regulating genes involved in response to hypoxia. HIF-1α protein overexpression has been shown in a variety of human cancers, but only 2 studies have documented the prognostic relevance of HIF-1α expression in breast cancer. The aim of our study was to determine accurately the impact of HIF-1α expression on prognosis in a large series (n = 745) of unselected patients with invasive breast cancer in terms of overall survival, local recurrence and distant metastasis risk. HIF-1α expression was investigated using immunohistochemical assays on frozen sections, and correlated with patients' outcome (median follow-up = 13.5 years). Univariate (Kaplan-Meier) analysis showed that high levels of HIF-1α expression (cutoff = 10%) significantly correlated with poor overall survival (p = 0.019). HIF-1α expression correlated with high metastasis risk among the whole group of patients (p = 0.008). Multivariate analysis (Cox model) showed that the HIF-1α predictive value was independent of other current prognostic indicators. Moreover among node negative ones, HIF-1α expression was also significantly predictive of metastasis risk (p = 0.03) and of relapse (p = 0.035). All the data suggest that HIF-1α is associated with a worse prognosis in patients with invasive breast carcinoma. Furthermore HIF-1α immunodetection may be considered as a potential indicator for selecting patients who could benefit from specific therapies interfering with HIF-1α pathway. © 2005 Wiley-Liss, Inc.

Neoangiogenesis is essential for tumour growth and the development of metastases. Cancer cell proliferation may outpace the rate of angiogenesis, resulting in tissue hypoxia1 and cellular adaptation to hypoxia is a key step in tumour progression.2 This adaptation is regulated mainly by the hypoxia-inducible factor-1 (HIF-1) that is known to play an essential role in oxygen homeostasis.3, 4, 5, 6, 7 HIF-1 is a ubiquitously expressed heterodimeric transcription factor comprising an α and a β subunit. Three isoforms of the α subunit have been identified: HIF-1α, HIF-2α (also referred to as EPAS-1, MOP2, HLF and HRF) and HIF-3α. HIF-1α is the best characterized isoform that is regulated by oxygen levels and forms a heterodimer with a constitutive HIF-1β subunit that is identical to aryl hydrocarbon nuclear receptor translocator (ARNT).8, 9 The amino-terminal half of each subunit contains basic helix-loop-helix (bHLH) and PAS (Per-ARNT-/AhR-Sim) motifs that are required for heterodimerization and DNA-binding.5, 10

HIF-1α is the sole oxygen-regulated subunit that determines HIF-1 activity.4, 11, 12 Hypoxic conditions lead to HIF-1α protein stabilization,3 and thus the protein's intracellular levels increase. Once stabilized, HIF-1α translocates to the nucleus guided by a nuclear localization signal present in C-terminus.13 After translocation HIF-1α heterodimerizes with HIF-1β, and the resulting HIF-1 complex binds to an enhancer element called the hypoxia-response element (HRE) in oxygen-regulated target genes.3 The amount of HIF-1α protein in the nucleus determines the functional activity of the HIF-1 complex. HIF-1α alters the transcription of a spectrum of genes mainly involved in erythropoiesis, angiogenesis and glucose metabolism including erythropoietin, transferrin, endothelin-1, inducible nitric oxide synthetase, heme oxygenase-1, vascular endothelial growth factor (VEGF), vascular endothelial growth factor receptor-1 (VEGF-R1), insulin-like growth factor-2, insulin-like growth factor-binding proteins and different glucose transporters and glycolytic enzymes.14, 15, 16, 17 Protein products of these downstream genes function to increase oxygen delivery or to activate alternate metabolic pathways that do not require oxygen. Interestingly, most of these proteins are implicated in tumour progression.18

There is increasing evidence that HIF-1 is one of the key factors in the progression of human malignant disease.19, 20, 21 Several studies have showed that HIF-1α protein was overexpressed in a variety of human cancers22 including lung, prostate, breast, colon cancer and direct correlations between HIF-1α and tumour angiogenesis have been demonstrated.23 Only a few reports, however, have provided evidence concerning the impact of HIF-1α expression on the prognosis of human breast carcinoma. Increased levels of HIF-1α have been reported during breast carcinogenesis, especially in poorly differentiated lesions,24 and one more recent study has shown the influence of HIF-1α protein expression on the behaviour of node positive invasive breast carcinoma.25 In contrast another recent study has demonstrated that HIF-1α expression correlated with a worse prognosis in node negative but not in node positive patients.26 These conflicting and preliminary data in human breast cancer may result from either too short follow-up or too small series. The practical relevance of HIF-1α as prognostic indicator and potential target for specific therapies in node negative patients then seemed to deserve further investigation.

The aim of our study was to accurately document the variations of HIF-1α expression in a large series of breast carcinomas (n = 745), and to correlate the immunohistochemical expression of this marker on frozen samples with patients' outcome in terms of overall survival and metastasis- and recurrence-free survival (long-term follow-up, median 13.5 years).

Material and methods

Materials

Seven hundred and forty-five patients from 25–79 years of age (mean = 56.1 years, SD = 13.3) with breast carcinoma underwent surgery from 1986–95. They did not receive chemotherapy or hormone therapy before surgery. The patients underwent axillary node excision combined with wide local excision with margins clearance or mastectomy in the Department of Oncologic Gynaecology in Conception Hospital, Marseilles. All the specimens were examined in the same Department of Pathology by experienced pathologists.

The patients' follow-up ranged from 8–17 years. The 2003 records showed that 277 (37.2%) patients relapsed, among whom 191 (25.6%) died and 468 (62.8%) were disease-free. Overall survival was calculated as the period from surgery until date of death. Metastasis-free survival was calculated as the period from surgery until date of metastasis.

Mean tumour size was 20.7 mm (SD = 13.8) and 23% tumours were ≤10 mm large, 40% were >10 mm and ≤20 mm large, 22% were >20 mm and ≤30 mm large, and 15% larger than 30 mm. Histological examination of surgical specimens was carried out on paraffin embedded sections stained with hematoxylin, eosin and saffronin.

Tumours corresponded to ductal carcinomas (n = 507, 68%), lobular carcinomas (n = 134, 18%), and to other types including mucinous, medullary, papillary, apocrine or mixed carcinomas (n = 104, 14%). Tumours were Grade 1 in 24% of these cases (n = 179), Grade 2 in 51% (n = 380) and Grade 3 in 25% (n = 186). Tumor grading, initially assessed by using the grading methods of Scarff et al.,27 was re-evaluated according to Elston and Ellis.28

A mean of 14.7 (SD ± 4.3) lymph node was found in axillary node excision and 372 (49.9%) patients were node negative.

Immunostaining procedure and quantification of HIF-1α expression

Fresh tissue fragments were sampled by pathologists immediately after intraoperative diagnosis. The fragment size varied according to the tumour size (average = 5 mm long, 4 mm wide, 3 mm thick). Fragments were obtained from dense tumour areas that lacked grossly visible adipose tissue. They were then dipped promptly in liquid nitrogen and stored at −80°C in the laboratory tumour library. Immunodetection was carried out on 5-μm thick sections (cryostat Leica CM 3050, Rueil Malmaison, France).

Immunoperoxidase procedure was realized using polyclonal rabbit (1:400 dilution) antihuman HIF-1α (H-206) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) and Ventana Gene II device with Ventana kits. Immunoreactivity of HIF-1α in breast cancer tissue was determined by assessing semiquantitatively the percentage of decorated tumour cells by experienced pathologists using a Zeiss Axioplan microscope as reported previously29, 30, 31 on the whole tissue section by examining all optical fields. Staining intensity was not incorporated in the scoring method because it was more or less constant.

Statistical analysis

The Kaplan-Meier method was used to analyze disease-free and overall survival rates. The difference between curves was evaluated with the Mantel Cox test (or log-rank test) for observations regarding censored survival or events. All computations were done with NCSS 2000 statistical software (Kaysville, UT). HIF-1α expression was stratified and correlated with major events during the course of the disease (distant metastasis or local recurrence) and with the overall survival rate to define immunohistochemical thresholds of prognostic significance. The optimal HIF-1α cutoff point of positive staining endowed with prognostic significance was determined after statistical validation.32 The effect of multiple factors on survival was tested with a Cox multivariate proportional hazards mode (NCSS 2000). The assumptions of proportional hazards were evaluated by examining data on the log cumulative hazard that were stratified by the histoprognostic factors used in the model (tumour size, histologic grade, nodal status) and by examining residual data vs. survival time. All p-values were based on 2-sided testing.

Results

HIF-1α distribution in tissue sections

The staining pattern exhibited a clear-cut delineation, with discriminative cell labeling and no background reactivity. HIF-1α expression was observed in all samples although the proportion of cells expressing HIF-1α varied considerably between tumours. The spatial arrangement of HIF-1α expression indicated a heterogeneous distribution across the tumour area. HIF-1α expression was essentially observed in the carcinomatous cells. HIF-1α showed cytoplasmic reactivity with very weak nuclear reactivity in some tumour cells (Fig. 1a,b). In some cases, HIF-1α immunostaining was located in stromal fibroblasts, endothelial cells and macrophages (Fig. 1c,d). The staining intensity varied within a given section and between sections. HIF-1α expression was semiquantitatively evaluated (percentage of decorated tumour cells). The distribution of the HIF-1α levels is shown in Figure 2 (mean = 16.32%, SD = 7.98, median = 14%).

Figure 1.

Immunoperoxidase staining of cryostat tissue sections using HIF-1α polyclonal antibody in invasive breast carcinomas. (a) Strong expression of HIF-1α is observed in a vast majority of tumour cells and in the tumour/stromal margin. (b) HIF-1α is sometimes expressed in a heterogeneous pattern in the cytoplasm of tumour cells. (c,d) HIF-1α protein expression may be observed in a subset of stromal cells.

Figure 2.

Distribution of HIF-1α expression evaluated (%) in frozen sections from 745 patients with breast carcinoma.

Univariate (Kaplan-Meier/log-rank) analysis and HIF-1α prognostic significance

HIF-1α levels (cutoff point = 10%) correlated (p = 0.019) with overall survival (Fig. 3, Table I). Tumours with HIF-1α expression levels higher than 10% were associated with a poorer survival as compared to those that exhibited lower HIF-1α levels. The validation of the optimal cutoff point for HIF-1α levels is shown in Figures 4 and 5, from p-values curve.32 In node negative patients subset, however, HIF-1α immuno expression did not retain a prognostic significance. Among the total group, HIF-1α levels >10% correlated with early and widespread metastasis (p = 0.008) (Fig. 6). A similar correlation was also observed (p = 0.03) in node negative patients (Fig. 7, Table II). HIF-1α levels >10% were correlated (p = 0.015) with higher relapse risk (metastasis and local recurrence) (Fig. 8). A close correlation was also observed (p = 0.035) in node negative patients subset (Fig. 9, Table III).

Figure 3.

Survivorship plot (Kaplan-Meier log-rank test) showing overall survival (all patients) of 745 patients with breast carcinoma with a higher (p = 0.019) risk of death for tumours with high expression levels of HIF-1α (≥10%).

Figure 4.

p-Values curve (log-rank test) showing optimal cutoff points for HIF-1α immunostaining for overall survival and recurrence in all patients according to Altman et al.32

Figure 5.

p-Values curve (log-rank test) showing optimal cutoff points for HIF-1α immunostaining for recurrence in node negative patients according to Altman et al.32

Figure 6.

Kaplan-Meier univariate analysis showing higher risk of metastasis (p = 0.008) in patients (all patients) with tumours with high expression levels of HIF-1α (≥10%).

Figure 7.

Kaplan-Meier univariate analysis showing higher risk of metastasis (p = 0.03) in node negative (N−) patients with tumours with high expression levels of HIF-1α (≥10%).

Figure 8.

Kaplan-Meier univariate analysis showing higher risk of metastasis or recurrence (p = 0.015) in patients (all patients) with tumours with high expression levels of HIF-1α (≥10%).

Figure 9.

Kaplan-Meier univariate analysis showing higher risk of metastasis or recurrence (p = 0.035) in node negative (N−) patients with tumours with high expression levels of HIF-1α (≥10%).

Table I. Overall Survival of Patients Correlates with HIF-1α Expression on Frozen Sections
All patients (n = 745)HIF-1α expressionp-value
<10%>10%
Deceased391520.019
Alive163391
Table II. HIF-1α Expression Correlates with Metastasis Event in All Patients and in Node Negative Ones
 HIF-1α expressionp-value
<10%>10%
All patients (n = 745)   
 Metastasis441820.008
 No metastasis158361
Node negative patients (n = 372)   
 Metastasis21830.03
 No metastasis83185
Table III. HIF-1α Expression Correlates with Relapse Risk (Metastasis and Local Recurrence) in All Patients and in Node Negative Ones
 HIF-1α expressionp-value
<10%>10%
All patients (n = 745)   
 Relapse622150.015
 No relapse140328
Node negative patients (n = 372)   
 Relapse29990.035
 No relapse75169

Multivariate (Cox Model) analysis and HIF-1α prognostic significance

In multivariate analysis, HIF-1α expression proved to be a prognostic indicator exhibiting a predictive value independent of the tumour size and tumour grade in terms of overall survival and metastasis free survival in all patients. HIF-1α expression did not remain a significant independent prognostic variable in node negative patients (Table IV).

Table IV. Proportional Hazard Regression, Cox Model
 Probability level
Overall survival (all patients) 
 HIF-1α0.030
 Histological grade0.033
 Tumor size0.760
Metastasis free survival (all patients) 
 HIF-1α0.023
 Histological grade0.002
 Tumor size0.653
Metastasis free survival (node negative patients) 
 HIF-1α0.106
 Histological grade0.003
 Tumor size0.945
Disease free survival (all patients) 
 HIF-1α0.158
 Histological grade0.004
 Tumor size0.793
Disease free survival (node negative patients) 
 HIF-1α0.317
 Histological grade0.005
 Tumor size0.652

Discussion

Tumour hypoxia is known to correlate with increased malignancy, potential for metastasis and poor patient prognosis in a number of tumour types including breast cancer.33 The transcriptional complex hypoxia-inducible factor-1 (HIF-1) plays a crucial role in physiological adaptation to hypoxia and is frequently activated in tumours. activation of HIF-1 is considered to support tumour growth through activation of anaerobic metabolism and induction of angiogenesis that is due in part to increased VEGF gene transcription.6, 18, 34 Immunostaining for the α subunit of HIF-1 (HIF-1α) can be used to identify the extent of HIF-1 activation in tumour tissues. A significant association between HIF-1α overexpression and patient mortality has been shown in tumours of the brain,35 cervix,36 ovary,37 and in nonsmall cell lung,38 head and neck,39 oropharynx,40 oesophageal41 and nasopharyngeal carcinomas.42

Only 2 clinicopathological studies that focus particularly on the prognostic relevance of HIF-1α expression in human breast carcinoma are available in the literature.25, 26 One study found that HIF-1α protein overexpression was associated with significantly shorter overall survival in a series of 206 patients in advanced-stage breast cancer only (5-year follow-up) evidenced by positive lymph nodes.25 In contrast, the study published by Bos et al.26 in a series of 150 patients showed that increased expression of HIF-1α correlated with overall survival only in node negative tumours. In this regard, these discrepant results deserve a deeper insight into HIF-1α prognostic significance in human breast cancer that would more accurately determine HIF-1α clinical relevance not only in terms of prognosis but also for further development of specific antiangiogenic therapy targeting HIF1. The aim of our study was to determine more accurately, in a series significantly larger (n = 745) than those reported previously, the impact of HIF-1α protein expression on the prognosis of unselected patients with invasive breast cancer with long term follow-up (median = 13.5 years). We investigated HIF-1α expression on frozen sections (Leica 3050) with automated immunodetection (Ventana Gene II), which provides optimal conditions for antigen preservation and for procedure standardization. Our results show that in univariate analysis (Kaplan-Meier), greater immunocytochemical expression of HIF-1α significantly correlated with a poor overall survival. This result corroborates the study of Bos et al.26 and Schindl et al.25 in which marked HIF-1α expression correlated with overall survival of all patients. Our study failed, however, to identify HIF-1α prognostic significance in terms of overall survival in a node-negative subset of patients of particular interest for therapy monitoring. These results contrast with those published by Bos et al.26 These discrepant observations may be explained in part by the different method of tissue preparation (paraffin vs. frozen sections). Moreover, our results show that HIF-1α overexpression correlated with early relapse (local relapse and distant metastasis) in all patients but also in a node-negative subset, data not previously reported suggesting that HIF-1 pathway is also implicated in local tumour progression. Recent studies have related the expression of HIF-1α with resistance to radiotherapy in carcinomas of oropharynx,40 oesophagus43 and the head and neck.44 From these observations it can be hypothesized that HIF-1 activation induces angiogenic activity that confers proliferation advantage in breast cancer cells during postsurgical treatments such as radiotherapy. We also observed that HIF-1α expression was predictive of metastasis risk in all patients and in the node negative subgroup. This infers a sensitivity of HIF-1α expression to identify a subset of node negative patients who might benefit from more aggressive postsurgical therapies.

Given the major role of HIF-1 activity in compensating for loss of oxygen by increasing its availability or providing metabolic adaptation of tumour cells to oxygen deprivation, the inhibition of this particular activity might provide a basis for the development of future therapeutic agents targeting HIF-1. The clinical relevance of targeting HIF-1 is suggested by mouse xenograft experiments in which the loss of HIF-1α activity through pharmacological or gene-therapy means resulted in decreased tumour growth and vascular density in tumours derived from breast carcinoma cells.18, 45, 46, 47 We have shown that a marked immunohistochemical expression of HIF-1α on frozen sections can predict prognosis, in terms of overall survival of breast cancer patients. Our study also shows that HIF-1α expression has a weak predictive significance in terms of metastatic risk and local recurrence in node-negative patients. Immunodetection of HIF-1α might further serve as an indicator for future adjuvant therapies specifically aiming at HIF-1 activation and downstream transcriptional targets.

Ancillary