STARD3: A New Biomarker in HER2-Positive Breast Cancer

Abstract

Pathological complete response (pCR) after neoadjuvant systemic treatment (NST) is an important prognostic factor in HER2-positive breast cancer. The majority of HER2-positive breast cancers are amplified at the HER2 gene locus, several genes are co-amplified with HER2, and a subset of them are co-expressed. The STARD3 gene belongs to the HER2 amplicon, and its role as a predictive marker was never addressed. The objective of this study was to investigate the predictive value of STARD3 protein expression on NST pathological response in HER2-positive breast cancer. In addition, we studied the prognostic value of this marker.

Methods: We conducted a retrospective study between 2007 and 2020 on 112 patients with non-metastatic HER2-positive breast cancer treated by NST and then by surgery. We developed an immunohistochemistry assay for STARD3 expression and subcellular localization and determined a score for STARD3-positivity. As STARD3 is an endosomal protein, its expression was considered positive if the intracellular signal pattern was granular.

Results: In this series, pCR was achieved in half of the patients. STARD3 was positive in 86.6% of cases and was significantly associated with pCR in univariate analysis (p = 0.013) and after adjustment on other known pathological parameters (p = 0.044). Performances on pCR prediction showed high sensitivity (96%) and negative predictive value (87%), while specificity was 23% and positive predictive value was 56%. Overall, specific, relapse-free, and distant metastasis-free survivals were similar among STARD3 positive and negative groups, independently of other prognosis factors.

Conclusion: NST is an opportunity for HER2-positive cancers. In this series of over a hundred HER2-positive and non-metastatic patients, a STARD3-negative score was associated with the absence of pathological complete response. This study suggests that determining STARD3 overexpression status on initial biopsies of HER2-positive tumors is an added value for the management of a subset of patients with high probability of no pathological response.

Keywords: STARD3; breast cancer; human epidermal growth-factor receptor 2 (HER2); neo-adjuvant systemic treatment; pathological complete response.

Figure 1

Flow chart of patient selection. ^a The initial query was all breast specimens HER2+ treated by surgery; ^b one patient was excluded as it was a second pathology opinion after a surgery in another center; ^c some initial core biopsies were stored in other pathology laboratories. Requests for stored material in other laboratories were made 3 times, some were recovered but others were not available or did not have residual material to perform immunohistochemistry assays; ^d either HER2 score 0–1 or 2+ with non-amplified in situ hybridization; ^e without anthracyclines and/or with pertuzumab.

Figure 2

STARD3 and HER2 expression are correlated in breast cancers. (A) Pearson correlation coefficient and regression line showing the correlation between STARD3 DNA, HER2 RNA, HER2 DNA, GRB7 RNA, TOP2A DNA, TOP2A RNA, and STARD3 RNA. RNA and DNA data were obtained from the Cancer Cell Line Encyclopedia database. RNA expressions are Log2 values, DNA represents copy number variation. (B) Pearson correlation coefficient (r) matrix and its statistical significance (*** = p < 0.001) between STARD3, HER2, GRB7, TOP2A DNA, and RNA. STARD3 is correlated with HER2 and GRB7 but not with TOP2A. (C) Correlation between STARD3 and HER2 mRNA expression (microarray, log2 values) in a cohort of breast cancers from The Cancer Genome Atlas (TCGA) database (Cancer Genome Atlas Network, 2012).

Figure 3

Generation of a specific anti-STARD3 antibody for immunohistochemistry. (A) Western blot analysis of STARD3 expression using the anti-STARD3 antibody 3G11 in HCC1954 cells (WT and transfected with control (siCtrl) and STARD3-targeting (siSTARD3)-siRNAs). Anti-Rab7 antibody was used as a loading control. (B) Western blot analysis of HeLa cells transiently expressing STARD1, STARD3, STARD4, STARD5, and STARD6 fused to GFP (GFP-STARD) using anti-STARD3 3G11 antibody. Anti-Rab7 antibody was used as a loading control. (C) HCC1954 cells (WT) and HCC1954 cells transfected with control siRNAs (siCtrl), and siRNAs targeting STARD3 (siSTARD3) were labeled with anti-STARD3 3G11 antibodies (red). Nuclei are stained blue. Scale bar: 10 μm. (D) Immunohistochemistry on normal breast (a) and HER2+ breast tumor (b) sections using anti-STARD3 3G11 antibody. N marks the position of a normal duct. S marks the tumor stroma. The subpanel (c) is a higher magnification image (4.5×) of the area outlined in black in (b). Scale bar: 50 μm (a,b) and 10 μm (c). All the whole western blot figures can be found in the Supplementary Materials.

Figure 4

STARD3 staining patterns in breast cancer. (A) Representative immunohistochemistry signals for STARD3 and HER2 on three biopsies. STARD3 protein is expressed at high levels in the majority of HER2-positive tumors (a,b). STARD3 expression is absent in some HER2-positive tumors (c). (B) Staining patterns for STARD3 in breast tumors. STARD3 is predominantly detected with a granular scattered pattern in the cytoplasm of cancer cells (a,c). In some cases, the granular STARD3 staining is polarized (b). In a few cases, the staining is weak and diffuse in the cytoplasm (d). Bars are 50 μm, nuclei are counterstained in blue with hematoxylin.

Figure 5

STARD3 gene expression in Gene Expression Omnibus (GEO) repository studies (density plots). Legend: panel (A) shows scaled STARD3 gene expression (RNA sequencing) in a study of n = 24 initial biopsies [26]. Panel (B) shows scaled STARD3 gene expression (RNA sequencing) in a study of n = 57 initial biopsies [27].

Figure 6

STARD3 prognosis. Association of STARD3 mRNA expression with survival (A): overall [OS]; (B): distant-metastasis free [DFMS]; (C): relapse-free [RFS]) in a cohort of breast cancers from the Kaplan–Meier plotter (KM plotter database) in the upper row, with multivariate analysis on HER2 below. The last row shows survivals in the studied cohort. Of interest, there are cross-overs in the survival curves, which indicate a non-proportional hazard. Consequently, in this survival analysis, p-values (log-rank tests) must be interpreted with caution, given the greater risk of wrongly concluding a statistically significant difference.