BNIP3 (nineteen kilodalton interacting protein) is a pre-apoptotic mitochondrial protein isolated through its interaction with bcl-2 and the adenovirus E1B19kDa proteins.1 2 BNIP3 mRNA is induced under hypoxic stress and the BNIP3 gene is under the direct regulation of the hypoxia-inducible factor (HIF)-1 transcription factor.3^–5 Induction of BNIP3 leads to caspase-independent necrosis-like cell death by opening the mitochondrial permeability transition pore.6 7 Overexpression of BNIP3 has been linked to an aggressive tumour phenotype in lung, colon and breast cancer,8^–10

In this study, we examined the immunohistochemical patterns of BNIP3 expression in a series of endometrial adenocarcinomas of the endometrioid cell type and assessed its association with HIF-1α overepression, the expression of other HIF-regulated molecules (vascular endothelial growth factor (VEGF) and lactate dehydrogenase (LDH)-5, and that of p53 and bcl-2 proteins. In addition, we studied the prognostic relevance of BNIP3 expression in patients with stage I malignant endometrial disease.

METHODS

Formalin-fixed paraffin-embedded tissues from 72 patients with stage I endometrial adenocarcinomas of the endometrioid cell type, and 20 samples from normally cycling endometrium of both proliferative and secretory phase, were retrieved from the archives of the Department of Pathology, Democritus University of Thrace Medical School, Alexandroupolis, Greece. The series of tissue blocks studied were sequential according to the dates when specimens had been received from surgery to the department. The study was approved by the Institute research committee. All patients had been treated surgically with total abdominal hysterectomy and bilateral salpingo-oophorectomy. No lymph node sampling of the iliac nodes was performed and N staging was based on pelvic and abdominal CT scan.

Histological typing and grading of the endometrial tumours (grade 1 vs grades 2 and 3) and the depth of myometrial invasion (<1/2 vs >1/2) were assessed on H&E sections, using standard criteria.

The follow-up of 62 patients alive at the time of analysis ranged from 22 to 176 months, with a median of 70 months. Ten out of 72 patients were dead from disease at the time of analysis.

Assessment of BNIP3 protein expression

The BNIP3 protein was assessed using a rabbit polyclonal antibody. The antibodies were made from a glutathione-S-transferase fusion protein containing amino acids 1–149 of the BNIP3 protein. The rabbit serum was tested by western blotting with recombinant protein for BNIP3 and the reactive band was removed by incubating the rabbit antiserum with the glutathione-S-transferase fusion protein. The antibody was also tested on skeletal muscle that showed high levels of BNIP3 mRNA. Details of the anti-BNIP3 antibody used have been previously published.5 10

Immunohistochemistry was performed at the Department of Pathology, Democritus University of Thrace Medical School, Alexandroupolis, Greece. Sections were deparaffinised, and peroxidase was quenched with methanol and 3% H₂O₂ for 15 min. Thereafter, slides were placed in antigen unmasking buffer pH 6.0 (code no. TAR001; ILEM, Cortemaggiore, Italy) and microwaving followed (3×4 min). The primary antibody (1:400) was applied overnight, at room temperature. Following washing with Tris-buffered saline (TBS), sections were incubated with a secondary mouse anti-rabbit antibody (Kwik Biotinylated Secondary, 0.69A; Shandon-Upshaw, Pittsburgh, Pennsylvania, USA) for 15 min and washed in TBS. Kwik Streptavidin peroxidase reagent (039A; Shandon-Upshaw) was applied for 15 min and sections were again washed in TBS. The colour was developed by 15 min incubation with diaminobenzidine solution and sections were weakly counterstained with haematoxylin. Normal rabbit immunoglobulin G was substituted for the primary antibody as the negative control, at a concentration where immunostaining of control slides gave a faint cytoplasmic staining.

The percentage of cancer cells with cytoplasmic BNIP3 reactivity was recorded after inspection of all fields in the tissue sample. The percentage of positive cells was recorded in each individual field and the final score for each case was the median value obtained.

Other immunohistochemistry

Table 1 shows details of the antibodies and the immunohistochemical procedures used to detect the expression of various oncoproteins and growth factors/receptors. Extensive report of the methods used has been previously published in referenced papers.10^–15

The percentage of cells with strong cytoplasmic and/or nuclear expression of HIF-1α and HIF-2α, and LDH-5, was recorded, and two groups of low versus high reactivity were defined according to a previously described grading system.11 14 Briefly, cases with nuclear expression in >10% of cancer cells and/or with strong cytoplasmic expression in >50% were considered to bear high reactivity. The median value of the percentage of cells with cytoplasmic reactivity was used to define two groups of low versus high VEGF.12 A 10% cancer cell positivity was required to score a case as positive for bcl-2 protein cytoplasmic expression and for p53 protein nuclear accumulation; this is the general reported cut-off point for these antibodies.15

Vascular counting was used for angiogenesis assessment. The areas of the highest vascularisation were chosen at low power (×100) and vessel counting followed on three chosen ×200 fields of the highest density, at the invading tumour front. The vascular density was the mean of the vessel counts obtained in these three fields. Vessels with a clearly defined lumen or well-defined linear vessel shape were taken into account for counting, while single endothelial cells were ignored; this is a method used by our group as previously reported.16 The median value (25 vessels) was used to group cases into low and high vascular density.

Statistical analysis

Statistical analysis and graphic presentation were performed using the GraphPad Prism 2.01 package (GraphPad, San Diego, California, USA; http://www.graphpad.com, accessed November 2007). Fisher’s exact test and χ² t test were used for testing relationships between categorical variables as appropriate. Survival curves were plotted using the method of Kaplan–Meier, and the log-rank test was used to determine statistical differences between life tables. A Cox proportional hazard model was used to assess the effects of patient and tumour variables on overall survival. A p value ⩽0.05 was considered significant.

RESULTS

BNIP3 was expressed in the cytoplasm of glandular endometrium throughout the normal menstrual cycle, without any differences noted in terms of intensity of staining. No nuclear staining was noted. Stromal cells remained unreactive.

BNIP3 was expressed in the cytoplasm of cancer cells with a varying extent and intensity. In 23/72 (31.9%) cases, BNIP3 was expressed strongly in more than 50% of the cancer cells (assessed in all optical fields available at ×200). These cases were considered as having a high BNIP3 reactivity (fig 1). The remaining cases, showed either strong expression in <50% of cancer cells (23/72 cases) or weak cytoplasmic expression regardless of the extent of staining (26/72 cases); these were considered as being of low BNIP3 reactivity. No nuclear staining was noted in this series of tumours.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Cytoplasmic BNIP3 expression in glandular endometrial tumour cells.

Analysis of BNIP3 expression according to the depth of myometrial invasion, histological grade or steroid receptor expression did not reveal any significant association (data not shown). A significant association of BNIP3 reactivity with HIF-1α (p = 0.04), VEGF (p = 0.04) and, particularly, LDH-5 expression (p = 0.0001) was noted (table 2). As necrosis was almost non-existent in our cases, it was not possible to correlate BNIP3 expression with this histological feature.

Overall survival analysis showed a significant association of high BNIP3 expression with poor prognosis (fig 2; p = 0.05). The 5-year overall survival was 95% in cases with low BNIP3 expression versus 76% in the remaining (hazard ratio 3.24). In multivariate models (table 3) BNIP3 expression was revealed as an independent prognostic variable.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Kaplan–Meier overall survival curves in patients with stage I endometrioid carcinoma of the uterus, stratified according to the BNIP3 status.

DISCUSSION

BNIP3 protein is pro-apoptotic protein involved in necrosis-like cell death.17 In situ hybridisation analysis of RNA expression in human tumours has shown BNIP3 upregulation in peri-necrotic regions, although BNIP3 overexpression has also been noted in some well-vascularised tumours.5 Although hypoxia is causatively related to BNIP3 gene transcription,3^–5 additional stimuli may be also involved in BNIP3 gene regulation including the p53 pathway and methylation of the BNIP3 gene.18 19

In this study, BNIP3 cytoplasmic expression was noted in almost one-third of cases. High cytoplasmic BNIP3 expression was significantly linked with the expression of HIF-1α and two HIF-regulated proteins, namely VEGF and LDH-5. This finding has also been shown earlier in lung and colon carcinomas.8 9 Experimental studies have confirmed that HIF-1α regulates the transcription of the BNIP3 gene.5

In several human tumour types examined to date, including lung, breast and colon cancer, high BNIP3 expression has been linked with aggressive tumour behaviour, with increased invasive and metastatic ability9 10 and with a poor prognosis for patients.8 This finding is also confirmed in the present study. Despite the excellent postoperative outcome of stage I endometrial carcinoma, 25% of patients were dead within 5 years of follow-up when tumour overexpressed the BNIP3 protein. In multivariate analysis, when LDH-5 was excluded from the models, BNIP3 maintained an independent prognostic relevance. LDH-5 has been previously reported to be strongly related with prognosis14 and the close association between BNIP3 and LDH-5 expression noted herein masks the prognostic relevance of BNIP3 in models that include LDH-5. It is possible that the association with LDH-5 is the explanation for its prognostic effect, but hypoxia induces many genes that have independent effects relevant to clinical therapy.

An explanation as to why BNIP3 overexpression is linked with poor survival of cancer patients has been offered by Graeber et al who suggested that hypoxia contributes to the selection of cancer cell clones resistant to apoptosis that are able to survive under unfavourable microenvironmental conditions.20 The herein confirmed strong association of BNIP3 with HIF-1α expression and the expression of LDH-5 enzyme, indicative of a strong anaerobic cancer cell metabolism, support the above hypothesis. The BNIP3 pathway may therefore be one of those downstream pathways by which hypoxia selects an aggressive tumour phenotype. Recently autophagy has been linked to BNIP3 and this can contribute to cell survival under stress conditions.21

On the other hand, silencing of the BNIP3 gene, frequently occurring in colon and pancreatic cancer,19 22 may lead to the defective ability of tumours to undergo cell death under cytotoxic therapy and thus resistance to chemotherapy and radiotherapy.22 23 BNIP3 therefore may be a marker of tumour aggressiveness as well as a prerequisite for tumour response to cytotoxic stimuli; this may prove useful for the identification of patients who would benefit from postoperative therapeutic manipulations.

It is concluded that BNIP3 protein is highly expressed in one-third of endometrioid carcinomas of the uterus and is linked with the expression of the HIF-1α and downstream proteins VEGF and LDH-5. Its strong association with poor postoperative prognosis, considered together with a potential favourable response to cytotoxic therapy (as previously indicated in experimental studies), favours the option of an adjuvant chemoradiotherapy trial stratified by BNIP3 expression in stage I endometrial carcinomas.