Hypoxia-regulated carbonic anhydrase IX (CAIX) protein is an independent prognostic indicator in triple negative breast cancer

Triple-negative breast cancer (TNBC) is an aggressive subtype of breast cancer with high five year mortality which is partly due to the lack of therapeutic target specificity on common breast cancer receptors such as oestrogen receptor (ER), progesterone receptor (PR) or the human epidermal growth factor receptor 2 (HER2) [1]. Further classification of TNBCs can be grouped into four molecular subgroups, driving many studies focusing on immunotherapy and new development in endocrine targeted treatments to identify potential targeted therapies [2]. Hypoxic microenvironment in tumour cells occurs in most solid malignancies, evolving tumours into an aggressive oncogenic metabolism, increasing metastasis and enhancing resistance to clinical therapies [3‐5]. Studies have also shown that hypoxia markers such as hypoxia inducible factor 1 (HIF-1) and hypoxia-driving factors are associated poorly in TNBC outcomes [6‐8].

HIF-1 is a heterodimeric protein composed of a constitutively expressed HIF-1ß subunit and an O₂- regulated HIF-1α subunit [9, 10]. Increased HIFα activates target genes involved in tumour proliferation, angiogenesis, metabolism, apoptosis and metastasis [4]. Additionally, HIFα and its regulated proteins including carbonic anhydrase nine (CAIX) and glucose transporter 1 (GLUT1) are highly expressed in several type of cancers and are associated with dismal prognosis [11‐14]. HIF-1 regulates key aspects of cancer biology, including pH regulation in glycolysis, through CAIX [15]. Over-expression of CAIX was observed in several solid tumours, and its link with invasiveness has given rise to the hypothesis that CAIX expression may contribute to advanced disease and tumour progression [11, 15]. Increased CAIX expression has been shown to be more common in TNBC compared to other subtypes of breast cancer and a marker of poor prognosis [11, 16]. Therefore, we investigated the impact of hypoxia-dependent CAIX in both protein and transcriptional expression on TNBC biology and outcome in order to elucidate its potential role as a therapeutic target in a subset of TNBC patients.

In the present study, we investigated the role of two important hypoxia-regulated markers (HIF-1α and CAIX) and found that increased expression in both CAIX protein and mRNA transcriptional levels are indicators of poorer survival in TNBC. However, HIF-1α protein expression failed to demonstrate any such association with either survival or clinicopathological factors. Interestingly, our results showed that HIF-1α protein expression is not a confounding factor in prognosis of patients expressing CAIX protein. However, co-expression of CAIX and HIF-1α protein in TNBC patients had the poorest prognosis. Furthermore, our study also identified seven CAIX-linked hypoxia genes with prognostic value in our TNBC cohort: DDIT4, ARL1, WAS, SETX, HK2, TUBA4α and CAIX which have been known to be hypoxia-regulated in vitro.

Our results were in agreement with CAIX protein in breast cancer studies, where 50% of basal-like breast cancers usually have high grade tumours expressing CAIX [22, 23]. Previous clinical studies in invasive breast cancer have also demonstrated the association of CAIX with poor outcome, suggesting that CAIX expression is linked to an aggressive phenotype [11, 16, 24, 25]. Over-expression of CAIX and carbonic anhydrase XII (CAXII) has also been associated with poor DFS in invasive breast cancer. However, the role of CAXII remains unclear and there have been conflicting reports about its role in TNBC. Chen et al. have shown that CAIX correlated with CAXII (R = 0.376, P = 0.0001) in a cohort of invasive breast cancer [26]. However, our study did not include CAXII and thus, unable show any correlation findings.

Furthermore, our study did not manage to find any prognostic value in HIF-1α protein expression, suggesting that HIF-1α may not be a reliable marker for hypoxia in TNBC. Although there are many markers to assess hypoxia in tumours, such as HIF-1α, X-Box Binding Protein 1 (XBP1), GLUT1 and Vascular endothelial growth factor (VEGF) [7, 8], the results however have been conflicting in various studies. Drawbacks associated with the modification of these hypoxia-responsive protein markers are their potential regulation by non-hypoxia-related factors such as stress, growth factor application, oncogene activation, cell culture densities, local pH, and metabolite concentrations [27]. Therefore, generating hypoxia signatures from in vivo tissue, despite the presence of contaminating stromal tissue, seem to be more robust than those generated from in vitro experiments [28]. Yehia et al. assessed the relative expression of HIF-1α among three breast cancer groups (TNBC, HER2+, ER+/PR+), with TNBC expression results differed only slightly and with little to no statistical significance from the other subgroups, and that HER2 positive tumours showed the highest levels of expression for all studied parameters [29]. This further supports that HIF-1α may not be an exclusive candidate marker for TNBC. Previous findings have demonstrated that HIF-1α was undetectable within minutes after re-oxygenation [30], suggesting that CAIX possibly activates hypoxic condition independently of HIF-1α, as CAIX protein persists longer than HIF-1α. Thus, CAIX as a biomarker for hypoxia could be more suitable as it is more stable and persists longer than HIF-1α.

Moreover, previous findings show that CAIX in high density cultures is induced via the phosphatidylinositol-3-kinase (PI3K) pathway [31] and by the mitogen-activated protein kinase (MAPK) pathway during both normoxia and hypoxia conditions [32]. Taken together, these observations suggest that CAIX expression may also be driven by other HIF-1α-independent signalling pathways to induce hypoxic conditions in the cells. Therefore, CAIX may be a better biomarker for cancer hypoxia.

The seven CAIX-linked hypoxia genes identified in our study have been linked to modulate key functions in tumourigenesis such as DNA repair, metastasis, innate immunity and metabolism in Additional file 1: Table S5. Notably, three of the genes (DDIT4, WAS, SETX) are linked to DNA repair functions. DNA damage inducible transcript 4 (DDIT4) acts as an independent prognostic factor for TNBC resistant to neoadjuvant chemotherapy [33]. DDIT4 activity supposedly enhances cancer cell resistance to mTOR inhibitors, thereby increasing cancer cells chemoresistance. Our results further support the notion of significant association between high DDIT4 mRNA level with poor survival, and reported upregulation in DDIT4 expression in our CAIX-positive TNBC cohort. Induced DDIT4 expression under cellular stressors and other chemical molecules (e.g. glucocorticoids, endoplasmic reticulum stress inducers, etc.) suggests its role in DNA repair under hypoxic conditions [34].

In the other two genes (WAS, SETX) linked to DNA repair functions, both downregulated WAS and SETX mRNA expression is associated with poorer overall-survival. Similarly, a subset of TNBC with increased expression of WAS and SETX mRNA showed better survival in other studies [35, 36]. Gene SETX role in tumourigenesis has been linked to its function in maintaining genome integrity via the coordination of transcription, DNA replication and DNA damage response [35], whereas gene WAS encodes for the cytoskeletal regulator, Wiskott-Aldrich syndrome protein (WASP), which plays a key role in tumourigenesis via binding to double strand breaks, regulating RNA Polymerase II activity and facilitating actin polymerization [37]. Its influence on actin filament dynamics and facilitation of actin reorganization, such as branching and crosslinking, are inherent in metastasis and invasion [37, 38]. Moreover, WASP and Arp2/3 complex have been reported to be recruited to damaged DNA double-strand breaks sites to promote double-strand breaks clustering and homology-directed repair [38, 39].

Thus, these further supports that the integrity of DNA-repair mechanism may be essential for protection against hypoxia-mediated DNA damage [36, 40, 41]. These biological categories have known functional relationships on breast cancer development and the aforementioned genes’ value as diagnostic markers and therapeutic targets deserves further investigation.

Within our seven gene DEG signature, TUBA4α is linked to metastasis, HK2 and CAIX is linked to promoting tumourigenesis, while the remaining ARL1 is linked to innate immunity [42]. Our results showed that these four genes were upregulated within the CAIX-positive group and associated with poorer survival outcomes in this subset of TNBC patients. Upregulation of TUBA4α disrupts the optimal tubulin isotype compositions in cell [43] and the dynamics of microtubule polymerisation and depolymerisation are of key importance in spindle formation during mitosis [44]. Moreover, upregulation of HK2 drives glucose metabolism and promotes sufficient number of metabolic intermediates to support anabolic processes (such as nucleic acid, lipid and protein synthesis), which is characteristic of rapidly dividing cancer cells [45]. While upregulation of CAIX disrupts pH balance [46], resulting in a hypoxic environment, which is also regulated under hypoxic condition through the hypoxia inducible factor (HIF1) cascade, promoting tumorigenesis. Thus, these genes are associated with aggressive cancer features and proliferation within the tumour microenvironment, reflecting the poorer survival outcome in our study.

Our study has several limitations. Since the FFPE blocks used in TMA construction were dated from 2003 to 2013, the tissue quality may be considered a limitation of this study. Tissue quality may contribute to the reduction of antigenicity and decrease in the sensitivity of the IHC reaction, leading to reduced protein detection. Furthermore, the FFPE tissue quality may also affect the amount of viable RNA for NanoString extraction and experiments. Although this study was conducted on a limited number of patient samples, the data indicates that quantification of hypoxia-related genes in TNBC can have potential prognostic value regardless of treatment type. Moreover, it is imperative that the clinical relevance of the seven hypoxia-linked gene signatures to be validated in independent studies with larger patient cohorts. Protein expression of the aforementioned genes showing significant association with survival is being studied in ongoing follow-up studies.

In conclusion, our study demonstrated that CAIX expression is independently associated with a poorer clinical and survival outcome in TNBC. Since hypoxia is increasingly being studied for being responsible for resistance against radiotherapy and emerging immunotherapy [47], the identification of the seven-genes associated with CAIX could be a step forward to test for hypoxia in TNBCs and possibly improve patients’ treatment regimen and prognosis. Thus, further studies on the seven-gene hypoxia panel are warranted.