Expression of hypoxia-inducible factor 1 alpha and its downstream targets in fibroepithelial tumors of the breast

After reaching a critical volume of several cubic millimeters, a growing tumor becomes increasingly depleted of oxygen and nutrients, and needs to adapt to its changing microenvironment. In order to survive, tumor cells must develop a vascular system and adapt their metabolism. A key regulator in this process is the transcription factor hypoxia-inducible factor 1 (HIF-1) [1], which controls the expression of several target genes. The protein product of HIF-1 is a heterodimer and consists of two subunits, HIF-1α and HIF-1β. Under normoxic conditions the HIF-1α protein is rapidly degraded. O₂-dependent hydroxylation of proline residues in HIF-1α causes binding of the von Hippel–Lindau tumor suppressor protein, which leads to ubiquitilation and subsequent degradation by the proteasome [2, 3]. Hypoxia inhibits this process, resulting in upregulation of HIF-1α and its downstream target genes [4]. On the other hand, hypoxia-independent upregulation of HIF-1α may be accomplished by loss of tumor suppressor genes such as the phosphate and tensin homolog deleted on chromosome 10 (PTEN) [5] and the von Hippel–Lindau tumor suppressor gene VHL [6], by activation of oncogenes like v-Src [7] and by stimulation by growth factors such as insulin-like growth factors [8, 9] and epidermal growth factor [10].

It has recently been demonstrated that HIF-1α expression is of prognostic value in several types of cancer, including breast cancer [11]. A well-known target of HIF-1α is the vascular endothelial growth factor (VEGF) gene [12]. VEGF is a potent endothelial cell-specific mitogen and is a major participant in the process of angiogenesis, resulting in the formation of microvessels. Both VEGF expression [13, 14] and microvessel density [15, 16] are established prognosticators in many types of cancer. Upregulation of carbonic anhydrase 9 (CA9) gene expression was found to be dependent on HIF-1α [17]. The protein product of the CA9 gene, carbonic anhydrase IX (CAIX), catalyzes the hydration of carbon dioxide to carbonic acid and contributes to acidification of the surrounding microenvironment. CAIX is constitutively expressed in cells lining the alimentary tract [18]. Because of its correlation with lowered pO₂ in carcinoma of the cervix, CAIX expression is regarded as an intrinsic marker of hypoxia [19]. Furthermore, CAIX expression has recently been related to poor outcome in breast cancer [20]. It therefore seems that HIF-1α and its downstream targets play pivotal roles in the development and progression of cancer.

Breast fibroadenoma and phyllodes tumor are both fibroepithelial tumors; that is, they are composed of an epithelial component and a stromal component. The distinction between both tumors may be difficult [21]. The behavior of fibroadenomas is benign, however, in contrast to phyllodes tumors that can recur and can even metastasize. In a small number of cases, fibroadenoma may progress to phyllodes tumor [22, 23]. Little is known of the mechanisms by which these tumors maintain a steady supply of nutrients as they grow. Few studies have addressed expression patterns of angiogenic growth factors in fibroepithelial breast tumors. Expression of basic fibroblast growth factor (basic FGF), FGF receptor and VEGF was found at higher levels in stroma of phyllodes tumors in comparison with fibroadenomas [24]. Unfortunately, no information on the epithelial component was provided in this work. Expression patterns of platelet-derived growth factor (PDGF) and the PDGF receptor suggest the presence of autocrine loops in stroma and paracrine stimulation of stroma by the epithelium [25]. A similar autocrine and paracrine loop has been described for acidic FGF and FGF receptor 4 in fibroadenomas [26]. These studies suggest that stromal proliferation may be stimulated by secretion of mitogens by the epithelial compartment. Interestingly, hypoxia may stimulate both FGF and PDGF expression [27, 28]. It would therefore be interesting to evaluate the role of HIF-1α and its downstream effectors in the tumorigenesis of biphasic breast tumors.

We therefore studied expression of HIF-1α and its downstream targets VEGF and CAIX in breast phyllodes tumors of various grades and in fibroadenomas. Furthermore, since HIF-1α seems to play a major role in the process of angiogenesis, we evaluated the microvascular network by counting of CD31-positive microvessels. Proliferation, as an important functional end point of various carcinogenic processes, was assessed by mitotic counts and Ki67 expression. Since HIF-1α degradation may be promoted by wild-type p53 [29], we evaluated the relation between p53 and HIF-1α expression in phyllodes tumors. Finally, the prognostic value of HIF-1α, its downstream effectors and microvessel counts was also assessed for phyllodes tumors.

Evidence is mounting that HIF-1α and its downstream targets play pivotal roles in the development and progression of many types of human cancers. This is the first study evaluating HIF-1α and CAIX expression in breast phyllodes tumors. We found that stromal HIF-1α overexpression predicts the prognosis of, and may play an important role in, the stromal progression of phyllodes tumors. Surprisingly, we regularly found concerted HIF-1α and CAIX expression in normal appearing epithelium of phyllodes tumors. Since only two fibroadenomas displayed low-level stromal HIF-1α immunoreactivity and none expressed CAIX, HIF-1α and CAIX seem to be of little relevance in these tumors.

HIF-1α overexpression is induced by hypoxia and by oxygen-independent mechanisms. In our study, stromal HIF-1α overexpression was related to the grade of phyllodes tumors, with a marked increase from borderline to malignant grade. Benign and borderline tumors showed comparable levels of HIF-1α overexpression. We only found two fibroadenomas with low-level expression of HIF-1α in the stroma, in agreement with Zhong and colleagues, who found all their fibroadenomas to be negative [35]. By clonality analysis we previously demonstrated that fibroadenomas may progress to phyllodes tumors by clonal expansion of stroma [22]. Perhaps positive stromal HIF-1α staining in a fibroadenoma may reflect an increased intrinsic capacity to progress to phyllodes tumor. The fibroadenomas with stromal HIF-1α positivity were microscopically inconspicuous, however.

In several types of cancer, hypoxia-induced HIF-1α expression seems to be characterized by a peri-necrotic distribution, whereas oxygen-independent overexpression results in a diffuse pattern of HIF-1α immunoreactivity [11, 36, 37]. Necrosis was found in the stromal component of only one malignant phyllodes tumor. Furthermore, stromal HIF-1α overexpression was not accompanied by an increase in CAIX expression, which is regarded as a marker of hypoxia [19]. Indeed, our group recently showed that, in contrast to homogeneous/diffuse HIF-1α expression, necrosis-related/focal HIF-1α expression is accompanied by CAIX expression [38]. All this suggests that HIF-1α upregulation in stroma of phyllodes tumors is normoxic and may be caused by changes in the stromal expression of oncogenes, tumor suppressor genes or growth factors. Wild-type p53 has been shown to promote MDM2-mediated ubiquitination of HIF-1α [29]. Therefore, p53 inactivation by gene mutation has been implicated in increased HIF-1α expression. We showed that aberrant expression of cell-cycle protein p53 only occurred in the stromal component. The transition from borderline to malignant phyllodes tumors was accompanied by a strong increase in stromal p53 expression, similar to stromal HIF-1α expression. Furthermore, we found a significant relation between stromal p53 and HIF-1α overexpression.

Another candidate for hypoxia-independent upregulation of HIF-1α in phyllodes tumors is PDGF. Overexpression of PDGF was found in 24% of phyllodes tumors, but not in fibroadenomas [25]. Interestingly, PDGF may induce HIF-1α expression [39]. Stromal HIF-1α and p53 expression in phyllodes tumors were predictive of disease-free survival, underlining the importance of the p53-HIF-1α axis in it's progression and clinical behavior.

Epithelial HIF-1α overexpression in phyllodes tumors was associated with CAIX expression. This suggests a causative role for epithelial hypoxia. Fibroadenomas were negative for CAIX, confirming the results of Bartosova and colleagues [40]. The mean distance from microvessels to the epithelium was significantly higher for phyllodes tumors as compared with fibroadenomas, suggesting that the hypoxic microenvironment in the epithelial component of phyllodes tumors is caused by relatively distant microvessels. Most vessels in phyllodes tumors were within 150 μm of the epithelial basal membrane, however, which is a critical distance for necrosis [41]. Still, it is conceivable that rapidly growing stroma stretches the two-layered epithelium in such way that it's oxygen supply does not keep up, resulting in a state of mild hypoxia. Indeed, a previous report detected CAIX expression 80 μm from the nearest blood vessel [42], which is comparable with the distance we found for phyllodes tumors. It is possible that the decreasing the oxygen gradient between 80 and 150 μm from the nearest vessel to the epithelium is sufficient to induce HIF-1α and CAIX expression.

The scattered foci of positive staining for HIF-1α and CAIX suggest the presence of focal mild hypoxia in the epithelial component. Microenvironmental disturbance of normal tissue by adjacent malignant disease has been described previously for HIF-1α and CAIX [40, 43, 44]. On the other hand, several other possible causes exist for epithelial HIF-1α overexpression. Epithelial PDGF expression, which is found in most phyllodes tumors [25], may cause HIF-1α overexpression [39]. Shpitz and colleagues found expression of HER-2/neu in the epithelium of 61% of phyllodes tumors [45]. It was recently demonstrated that enhanced HER-2/neu signaling induces HIF-1α protein expression [46]. Numerous factors interact with HIF-1 and other as yet unidentified changes in the epithelium of phyllodes tumors may cause HIF-1α overexpression.

The tumor biological significance of increased epithelial HIF-1α and CAIX expression is unclear. Morphologically, the epithelial component was mostly two-layered epithelium without atypia. Furthermore, clonality studies mostly found the epithelial component of phyllodes tumors to be polyclonal [22, 23]. Moreover, phyllodes tumor metastases are composed of stroma, with only one case described in which the epithelial component also disseminated, leading to a biphasic metastasis [47]. Finally, epithelial HIF-1α and CAIX expression are not predictive of prognosis. It seems therefore that HIF-1α and CAIX expression in the epithelium of phyllodes tumors merely reflects a physiological adaptation to microenvironmental disturbance by rapidly proliferating stroma with lagging peri-epithelial angiogenesis. However, possible upregulation of growth factors in the epithelium by HIF-1α may exert an additional stimulatory force on the stromal component. Indeed, it has been suggested that in the early stages of phyllodes tumor development the epithelium secretes mitogens stimulating the stromal component [48, 49]. The presence of such autocrine and paracrine loops has been described for PDGF/PDGF receptor [25]. Furthermore, stimulation of the stromal component by the epithelium was suggested previously by studying endothelin-1 expression in epithelium of phyllodes tumors [50]. Interestingly, endothelin-1 turned out to be a target of HIF-1α [51].

An abrupt increase in number of microvessels was observed in malignant phyllodes tumors, coinciding with a strong increase in stromal overexpression of HIF-1α from borderline grade to malignant grade. In contrast with a previous report [52], we found no difference in microvessel counts between benign phyllodes tumors and borderline phyllodes tumors. In addition, when regarding hot-spot and global counts we found that there was no difference between fibroadenomas, benign phyllodes tumors and borderline phyllodes tumors. This is in contrast with a previous report claiming a difference between fibroadenomas and phyllodes tumors [24]. On the other hand, Weind and colleagues found a large overlap in microvessel counts between fibroadenomas and invasive breast cancers, demonstrating that fibroadenomas are capable of producing large numbers of microvessels [53]. We found large numbers of small peri-epithelial microvessels in fibroadenomas responsible for its high numbers of microvessels. It seems that microvessel counts are not helpful in the differentiation between fibroadenomas and benign phyllodes tumors, which poses the biggest diagnostic problem.

VEGF expression in fibroadenomas did not differ from that in phyllodes tumors. Several growth factors such as FGF-4 [54], PDGF [55] and transforming growth factor beta [56] may stimulate VEGF expression. Expression of FGF [26], PDGF [25] and transforming growth factor beta [57] have been described in fibroadenomas and may contribute to HIF-1α-independent expression of VEGF. In phyllodes tumors, the relation between VEGF expression and HIF-1α reached borderline significance. Although HIF-1α most probably contributes to VEGF expression in phyllodes tumors, VEGF expression seems, at least in part, to be independent from HIF-1α.

In view of its annual incidence of 2.1 cases per 1 million women [58], we feel that our group of phyllodes tumors is reasonably sized and well composed with all grades present. Still, future investigations confirming our results in larger series are warranted. Furthermore, experiments covering a variety of molecular elements such as DNA microarray techniques may unravel the complex mechanisms underlying the presumed nonhypoxic upregulation of HIF-1α in the stromal component of phyllodes tumors and its subsequent influence on tumor progression.

This is the first report on HIF-1α and CAIX expression in breast phyllodes tumors. Our results show that HIF-1α is related to diminished disease-free survival and may play an important role in stromal progression of phyllodes tumors. The significant relation between tumor grade and stromal HIF-1α overexpression underlines its importance, with all malignant tumors showing HIF-1α overexpression. Stromal HIF-1α overexpression in phyllodes tumors most probably arises from hypoxia-independent pathways, with p53 inactivation as one possible cause. Surprisingly, the normal-appearing epithelium in phyllodes tumors frequently displayed HIF-1α and CAIX expression. The distance from the nearest vessels to the epithelium was higher in phyllodes tumors as compared with fibroadenomas and a hypoxic effect seems plausible here, although the effect is of doubtful biological significance. In contrast to phyllodes tumors, HIF-1α seems of minor relevance in the tumorigenesis of fibroadenomas. Considering it's possible role in progression of the stromal component of phyllodes tumors and the fact that metastases are composed of stroma, novel therapies targeting HIF-1α [59] may contribute to the treatment of disseminated phyllodes tumor, which is poorly responsive to conventional chemotherapy and radiotherapy.