Glomeruloid microvascular proliferation is associated with p53 expression, germline BRCA1 mutations and an adverse outcome following breast cancer

Glomeruloid microvascular proliferation (GMP) in breast cancer independently adversely affected survival (relative risk 1.9, 95% CI: 1.2–3.0), particularly among women who received adjuvant chemotherapy (10-year survival 27 vs 69%, P=0.0003), and was significantly associated with p53 overexpression and BRCA1 germline mutations. The presence of GMP may influence treatment decisions.

Angiogenesis is under intense study both for its prognostic value and for the potential therapeutic value of interfering with angiogenic pathways. In breast cancer, both higher vascular endothelial growth factor (VEGF) levels (Blackwood and Weber, 1998;Linderholm et al, 2000) and increased microvessel density (MVD) (Weidner et al, 1991(Weidner et al, , 1992de Jong et al, 2000) are associated with a poorer prognosis. Recent microarray studies have also identified an association between the expression of genes involved in angiogenesis, such as VEGF, and poor prognosis following breast cancer (van't Veer et al, 2002). Furthermore, the proangiogenic genes COL4A1 and ECGF1 are overexpressed in breast cancers found in BRCA1 mutation carriers, although their prognostic value in such women is unknown (van't Veer et al, 2002). Here, we describe our results with a morphological marker of prognosis of potential importance in the management of women who carry germline BRCA1 mutations. Glomeruloid microvascular proliferations (GMPs) (Figure 1) are focal proliferative buddings of vascular endothelial cells resembling a renal glomerulus. Glomerular microvascular proliferation has been predominantly associated with glioblastoma multiforme (Wesseling et al, 1998), the most aggressive form of glioma. They can be produced in athymic mice by overexpressing VEGF using an adenovirus vector (Sundberg et al, 2001). Recent evidence suggests that these structures may be present and prognostically useful in other tumour types, including breast cancer (Straume et al, 2002). The work presented here is an expansion of the breast cancer data presented by Straume et al, and focuses on two new aspects: (1) the inter-relationship between GMP, p53 overexpression and BRCA1/2 status and (2) the effect of GMP on prognosis in the presence or absence of adjuvant chemotherapy.

MATERIALS AND METHODS
A total of 292 consecutive Ashkenazi Jewish women aged 65 years or less with primary nonmetastatic breast cancer diagnosed at one Montreal institution between 1980 and 1995 were assessed. Sufficient follow-up and tissue were available for 251 subjects. Following ethics committee approval, specimens were evaluated by one pathologist (LR Bégin) using conventional methods. Accumulation of p53 protein was detected by immunohistochemistry as previously described (Yuan et al, 1999). Pathology blocks from all women were tested for founder BRCA1 mutations (185delAG, n ¼ 18; 5382insC, n ¼ 10) and BRCA2 mutation (6174delT, n ¼ 8) that are common in this population, using established techniques (Foulkes et al, 1997). embedded archival material as previously published (Straume and Akslen, 2001). The presence of GMP was recorded by the finding of focal glomerulus-like aggregates of closely associated and multilayered factor-VIII positive endothelial cells. Glomerular microvascular proliferations consisted of 15 -100 cells. Lumen formation was not necessary for the aggregates to be counted as GMPs. Tangentially sectioned normal vessels, or nonspecific Factor-VIII positivity in stromal components, were excluded. Glomerular microvascular proliferations were categorised as being absent (group 0), rare (not more than one per high-power field (HPF), group 1), or greater than 1 per HPF (group 2). Microvascular density was calculated as the mean number of stained vessels in 10 high-power fields ( Â 400).
Molecular, pathological and clinical assessments were collected in a mutually blinded manner in a retrospective cohort approach. Subject characteristics were compared using Wilcoxon, t-test, and Fisher's exact testing, with trends in increasing odds ratios being assessed by Cochran -Armitage's test. Differences in breast cancerspecific survival were calculated using the method of Kaplan and Meier. The Cox proportional hazards model was used to assess prognostic factors.
There were 65 breast cancer deaths in this series of women at 10 years follow-up. Kaplan -Meier survival analysis showed that 50.3% of women with GMP died of breast cancer over this period, whereas the mortality was 25.7% for those with no identified GMP (P ¼ 0.0003). Microvascular density was not significantly associated with a worse prognosis (P ¼ 0.47). In a Cox proportional hazards model, the presence of GMP (defined continuously) was associated with a poor prognosis (relative risk (RR) 1.9, 95% CI: 1.2 -3.0) as was positive lymph node status (RR 2.3). Nuclear grade (RR 1.6) and negative ER status (RR 1.7) were of borderline significance, while tumour size, p53 positivity and carrier status did not achieve significance (Table 2). Among women treated with adjuvant chemotherapy, the presence of GMP was an indicator of poor prognosis (10-year survival 27 vs 69%, P ¼ 0.0003), while among women not treated with chemotherapy, no statistically significant difference in survival was seen on the basis of GMP status (10-year survival 75 vs 79%, P ¼ 0.4).

DISCUSSION
This study is the first to demonstrate that GMP is associated with p53 expression and the presence of germline BRCA1 mutations and it suggests that the presence of GMP is an independent risk factor for death from breast cancer comparable in magnitude to conventional prognostic factors (RR 1.9).
Notably, GMP was not associated with a higher MVD, and the latter was not prognostic for poor survival in our cohort of patients. Vascular endothelial growth factor is implicated in the   (Sundberg et al, 2001) and increased MVD (De Paola et al, 2002), but the lack of association between GMP and MVD suggests that their developmental pathways may differ. In our cohort, p53 expression was associated with the presence of GMP, but not with increased MVD (P ¼ 0.8), the latter being consistent with the literature (Tas et al, 2000). Functional p53 impedes angiogenesis through the regulation of VEGF transcriptional factors Sp1 (Mandlekar and Kong, 2001) and the HIF-1a subunit (Ravi et al, 2000), as well as by upregulating thrombospondin-1 expression (Dameron et al, 1994). Mutated p53 may be one pathway by which a neovascular phenotype associated with GMP (but not MVD) formation is promoted. Our data suggest that GMP is associated with p53 expression and BRCA1 germline mutations and that all three of these factors may be associated with a worse survival (for details of the p53-BRCA1 relationship, see Goffin et al, 2003). Interestingly, patients who were treated with adjuvant chemotherapy had a poorer outcome if their tumours demonstrated GMP. Glomerular microvascular proliferation was highly significantly associated with p53 expression (P ¼ 0.0001), a protein partly responsible for inducing apoptosis in chemotherapy-treated cells and thus potentially responsible for diminished responsiveness to chemotherapy (Fisher, 2001). Opposed to this is the association of GMP with BRCA1 mutations and higher nuclear grade, both of which appear to increase tumour responsiveness to chemotherapy Wang et al, 2002). In the present study, higher nuclear grade (RR 1.9, P ¼ 0.02) and age o 50 years (RR 5.6, P ¼ 0.0001) were associated with an increased likelihood of a woman receiving adjuvant chemotherapy, while other factors were not significant on multivariate analysis. The apparent contradiction in associations and chemoresponsiveness is likely a product of the interplay of several response mitigating pathways and the incomplete association between measured factors.
There is evidence that BRCA1-related breast cancers have a distinct profile on microarray analysis (van't Veer et al, 2002) and that these cancers have a distinctive spectrum of TP53 mutations (Greenblatt et al, 2001). Along with evidence that BRCA1 is important in global nucleotide excision repair (Hartman and Ford, 2002), these data hint that BRCA1 mutations induce a genetic profile of which p53 expression and GMP are but two manifestations, with several factors influencing both prognosis and response to treatment. However, the role of BRCA1 mutations in the genesis of such a phenotype requires further investigation.
Angiogenesis is a complex process and its full understanding will require analysis at the level of morphology and gene expression. Here, we describe the poor prognosis associated with GMP, which is a highly characteristic lesion resulting from a gene expression profile that is as yet undefined. As antiangiogenic therapy is currently under intense investigation, it will be important to establish whether the presence of GMP alters the effectiveness of such therapies. Norwegian Cancer Society and the Norwegian Research Council are acknowledged. JRG is a recipient of the Canadian Association of Medical Oncologists/Canadian Institutes of Health Research Fellowship. WDF is a Chercheur Clinicien Boursier of the FRSQ.