Hypoxia inducible factor 1α gene (HIF-1α) splice variants: potential prognostic biomarkers in breast cancer

Hypoxia-inducible factor 1 (HIF-1) is a ubiquitously expressed master transcriptional regulator of many genes regulating mammalian oxygen homeostasis [1, 2]. HIF-1 alters transcription of genes mainly involved in energy metabolism, neovascularisation, survival, internal pH and cell migration, and has become recognised as a strong promoter of tumour growth [3, 4]. HIF-1 is a heterodimeric protein composed of two HIF-1α and HIF-1β/aryl hydrocarbon receptor nuclear translocator (ARNT) subunits [5]. These subunits are members of the basic helix-loop-helix (bHLH) transcription factor superfamily containing a PAS [PER (Period Clock) -ARNT-SIM (Single-minded)] domain [2]. The HIF-1α subunit is specific to HIF-1 and is the unique oxygen-regulated subunit that determines HIF-1 activity [6, 7] whereas HIF-1β is a common subunit for several transcription factors. In human tumours, HIF-1α protein is overexpressed because of intratumoural hypoxia and genetic alterations affecting key oncogenes and tumour suppressor genes [8, 9]. This overexpression correlates with tumour angiogenesis, treatment failure and patient mortality [3].

Our group and others have previously reported that HIF-1α protein overexpression is a marker of poor prognosis in primary breast cancer patients [10, 11]. HIF-1α is known to be mainly post-transcriptionally regulated by protein ubiquitination and interaction with the Von Hippel-Lindau tumour suppressor protein, and then degraded by the proteasome. Hypoxia stabilises HIF-1α protein by relaxing its ubiquitin-proteasome degradation [12] and affects subcellular localisation, DNA binding capacity and transcriptional activation function of the HIF-1 complex. HIF-1α protein is composed of 826 amino acids [5]. Its N-terminal region contains the bHLH and the PAS domains, which are essential for DNA binding and dimerisation [13]. Its C-terminal region contains two transactivation domains and a nuclear localisation signal [14]. In the middle section of HIF-1α lies a Pro-Ser-Thr oxygen-dependent degradation domain (ODDD, amino acids 401 to 603), which is responsible for the stability of the HIF-1α protein [12].

Although HIF-1α protein regulation has been well studied, HIF-1α messenger regulation has only recently begun to be documented [15]. HIF-1α is encoded by the HIF1A locus, which is located on human chromosome 14q21-q24 and encodes for 15 exons, resulting in a principal transcript of about 8 kb [16]. Different functional domains of the HIF-1α protein are encoded by separate exons. In addition to the originally described wild type of HIF-1α (HIF-1α^WT) [1], eight alternative splice variants of HIF-1α (HIF-1α⁸²⁷, HIF-1α⁷³⁶, HIF-1α⁵⁵⁷, HIF-1α⁵¹⁶, HIF-1α⁷⁸⁵, HIF-1α⁴¹⁷, HIF-1α^TE and an isoform with alternative exon 1 (I.2 isoform)) with different activity have been reported in human cell lines [17–23]. All HIF-1α splice variants have been shown to be translated, to dimerise with HIF-1β and to be active in human cells. Only one isoform (HIF-1α ⁷³⁶ ) has been documented in human breast cancer [24]. The aim of this study was to analyse HIF-1α splice variant expression in human breast cancer and to evaluate its clinical relevance in terms of prognosis. In this study we quantified the mRNA levels of 3 isoforms (HIF-1α ⁷³⁶ , HIF-1α ⁵⁵⁷ and HIF-1α ⁵¹⁶ ) in breast specimens of 53 primary cancers and 29 normal tissues or benign lesions by real-time quantitative reverse transcription PCR (RT-qPCR) assays. HIF-1α ⁸²⁷ is similar to wild type except for an additional three base pairs TAG insertion at the exon 1 to 2 junction involving a difference of two amino acids upstream of the bHLH domain [17]. HIF-1α ⁷³⁶ is also characterised by an additional three base pairs TAG insertion at the exon 1 to 2 junction, and the lack of exon 14, which produces a frame shift and introduces a stop codon in the coding sequence after the Ile⁷³⁵ [17]. HIF-1α ⁵⁵⁷ loses exon 12 and HIF-1α ⁵¹⁶ lacks exons 11 and 12 [18, 19]. HIF-1α ⁵⁵⁷ has been shown to be induced by zinc ions. HIF-1α ⁵⁵⁷ and HIF-1α ⁵¹⁶ have been shown to function as dominant negative isoforms of HIF-1α in vitro. HIF-1α ⁷⁸⁵ is deprived of exon 11 and was described as being induced in vitro by phorbol-12-myristate-13-acetate [20]. HIF-1α ⁴¹⁷ isoform is deprived of exon 10 [21]. The HIF-1α ^TE variant with alternative exon 1 (I.1 isoform) was previously found to be expressed in the testis [22]. For these reasons, both these isoforms were thought to be out of the scope of this study on breast tissues samples. We also developed tools to quantify expression of HIF-1α splice transcripts characterised by insertion of a three base pairs TAG at the exon 1 to 2 junction (which we designated HIF-1α ^TAG ) involving a difference of two amino acids upstream of the bHLH domain (HIF-1α ⁸²⁷ and HIF-1α ⁷³⁶ isoforms) [17]. The last known isoform was not quantified because it was described in humans after the end of our study [23]. Expression levels of these isoforms were also compared to standard prognostic clinicopathological parameters, microvessel density of tumours and clinical outcome.

Several isoforms of HIF-1α resulting from alternative splicing have been described in various human tumour cell lines or tissues. Only one isoform (HIF-1α ⁷³⁶ ) has been documented in human breast cancer [24]. Moreover, only a few studies have been performed to link HIF-1α mRNA expression and tumour-specific parameters [32, 33]. Specific antibodies allowing HIF-1α splice variant detection at the protein level are not currently available. We used a real-time quantitative reverse transcription PCR assay to measure mRNA expression levels of 4 alternatively spliced transcripts of HIF-1α in breast specimens of 53 primary cancers and 29 normal tissues or benign lesions.

Interestingly, this work shows for the first time higher expression levels of two HIF-1α splice variants (HIF-1α ^TAG and HIF-1α ⁷³⁶ ) in OR-negative carcinomas compared to normal/benign tissues. We also investigated the prognostic value of HIF-1α transcript expression levels in breast cancer and found a significant relationship between either clinicopathological characteristics or patient metastasis-free survival. First, we found that HIF-1α ^TAG mRNAs levels were substantially higher in high grade and steroid hormone receptor-negative tumours. The second and most striking observation was that HIF-1α ^TAG mRNA levels were indicative of shorter metastasis-free survival, and that this correlated with lymph node status. Contrary to the Cayre et al. report [24], we did not find significant correlation between total HIF-1α mRNA expression and lymph node status but we observed significant association with tumour grade. This could be because our series was smaller than series of Cayre et al. or because the technology used in our study was more sensitive. In our series, total HIF-1α mRNA expression negatively correlated with OR status. Similar to Cayre et al. we did not find any correlation between total HIF-1α mRNA expression and outcome. Our results showing that HIF-1α ⁷³⁶ mRNA expression does not correlate with clinicopathological characteristics of tumours also concur with earlier findings of the same group [24]. In this study, effect of adjuvant (none, chemotherapy or hormone) treatment on survival and interactions with expression levels of HIF-1α splice variants were not checked because of the limited number of patients in each subgroup. Further experiments in a larger series are required to answer this question.

Alternative splicing is known to play an important role in gene expression regulation by modulating the functional properties of transcription factors [34]. In this regard, alternative splicing can modify DNA-binding properties of transcription factors [35], introduce or eliminate activating domains or increase the in vivo stability of a given isoform [36]. Moreover, the abundance of specific isoforms is likely to result from differential expression, RNA stability and selective splicing process leading to an increase of some mRNA species. Recent evidence indicates that in several cancers the ratio of splice variants is dramatically altered and that differential expression of alternatively spliced isoforms in cancer patients can have severe implications for clinical outcome [37]. Remarkably, statistical association has previously been reported between HIF-1β splicing variant expression, oestrogen receptors and breast cancer survival [38]. It should be noted that the primers designed on exons 1 and 2 in our study allowed quantification of the HIF-1α ^TAG sequence present in both HIF-1α ⁸²⁷ and HIF-1α ⁷³⁶ splice variants [17]. The HIF-1α ^TAG transcript is characterised by insertion of a three base pairs TAG insertion that may be generated by the use of two potential splice acceptor dinucleotides (AG) of intron 1 at a splice junction site as previously described [39, 1]. This splicing results in the replacement of Lys¹² by Asn¹² and the addition of Arg¹³ residue located upstream from the bHLH domain of the protein. This replacement may modify the DNA binding affinity of the protein complex as previously shown for Arg¹⁴ and Arg¹⁵ residues in aryl hydrocarbon receptor (AHR)/aryl hydrocarbon receptor nuclear translocator (ARNT) heterodimer [40]. Further biochemical and structural studies are required for a better understanding of the functional properties of this variant.

To our knowledge, our results suggest for the first time that at least one HIF-1α splice α variant may be a marker for the advanced clinical and oestrogen-resistant stage of breast cancer. Based on their correlation with survival, HIF-1α ^TAG mRNA levels may be a potential useful prognostic indicator whose value should be further validated in prospective studies.