Brain metastases (BM) from breast cancer constitute an important part of therapeutic failures and are associated with severe morbidity and mortality.
The risk of BM is particularly high in HER2 positive, advanced breast cancer patients.
A greater understanding of the molecular and genetic signatures predisposing to the development of brain metastases may help guide individualized, patient-specific therapy.
Using RNA-microarray technology, the authors previously developed a 13-gene signature for HER2 positive breast cancer patients, which strongly predicts for rapid development of BM (ASCO oral presentation 2008).
In this study, they used quantitative reverse-transcriptase PCR (qRT-PCR) technology to validate their previous results in an independent group of patients and on a culture model system.
The initial discovery group included 87 samples analyzed using cDNA synthesis, annealing, selection, extension, ligation and array hybridization (DASL). RNA was extracted from primary tumor samples using the HighPure RNA Paraffin Kit, and DASL assay of 502 known cancer genes was performed using the Sentrix Universal Array.
Most patients in the initial sample of advanced breast cancer cases were premenopausal (47%), and had ductal pathology (69%), visceral disease (78%), negative estrogen receptors (55%), and grade 3 disease (48%). 72% had trastuzumab therapy and 47% developed brain metastases with a median time of 55 months to BM development.
The 13-gene signature included four HER2 related genes (CKD4, CCNC, PTK2, MYC), three DNA double stranded break repair genes (BARD1, RAD51, FANCG), and six other genes (PCNA, PRCC, TPR, EMS1, DSP, HDGF).
In the present study, the independent validation group included 75 samples analyzed using qRT-PCR.
The patients had very similar demographic profiles to the initial sample of patients. Median time to BM development was 65 months.
qRT-PCR was performed for 13 genes plus two control genes.
The Delta cycle threshold (DCT) was generated using biosystems software, and were standardized for use in univariate Cox regression models.
An "expression score" was calculated as the weighted DCT values for all 13 genes.
The median of data was used to divide patients into high vs. low expression groups.
A 3D culture validation model system was developed using immortal, non-tumorigenic human breast epithelial cells with and without ectopic expression of HER-2 and RAD5.
The number and morphology of breast acini were scored using indirect immunoflourescence and confocal microscopy to determine if overexpression of RAD51 or BARD1 predicts for development of brain metastases.
Median brain metastasis-free survival (BMFS) in the initial discovery group for 'high' vs. 'low' expression signature tumors was 36 months vs. 66 months (p=0.0068), and in the validation group 54 vs. 86 months (p=0.032).
Short BMFS was also associated with ER-negativity.
BMFS in the cohort of 'high' 13-gene signature and ER- tumors was 31 months compared with 66 months in discovery group (p<0.0001), and 41 vs. 77 months in the validation group (P=0.02).
Over-expression of RAD51 in MCF-10A breast cells altered their three-dimensional morphology and increased the percentage of invasive structures by 6.5 fold (p<0.0001), both in the presence and absence of HER2 over-expression.
Over-expression of BARD1 resulted in an increased percentage of invasion by 2.5 fold in vitro.
This 13-gene signature and ER-negativity predict for rapid development of BM in HER+ advanced breast cancer pts.
RAD51 may promote aggressiveness and invasion in breast epithelial cells.
These data may be useful in the design of BM preventive trials and may promote investigation of new treatment strategies.
This very interesting study represents a validation of the authors' prior work, and begins to address the genetic basis of cancer colonization of distant organs.
The strengths of the study include:
Discussion of a timely and clinically relevant question - this data could be used to develop tools for screening, and prophylactic strategies, as well as inform choice of HER2 targeted therapy.
Validation of prior work on an independent dataset lends credibility and robustness to the authors' original work
The initial study group and the validation group were well-matched in terms of patient and tumor characteristics
The weaknesses of the study include:
Specificity of the 13 gene signature – based on this data, the signature appears to be sensitive to predicting metastatic disease, but we cannot ascertain whether or not it is predictive specifically for brain metastases or for distant disease at any location.
Validation using qRT-PCR was not performed on the initial study group.
RAD51 was found to have prognostic significance – an analysis comparing RAD51 to the entire 13 gene signature would be helpful and informative.
The impact of the gene signature on BMFS was tested in a univariate, but not multivariate model.
This study promotes progress towards understanding how certain genes are predictive of more aggressive disease and earlier onset of metastatic disease.
Furthermore, because a large proportion of patients failed trastuzumab therapy, this model incorporates a gene signature predicting for trastuzumab resistance.
Nonetheless, this gene signature assay in not yet ready for routine clinical use. Validity in multivariate models and specificity for brain metastases need to be verified. Additionally, implications for clinical use need to be more fully explored.
Oct 21, 2014 - Several abstracts involving potential biomarkers of prognosis in cancer treatment were presented at a press briefing Nov. 18 at the American Association for Cancer Research -- National Cancer Institute -- European Organisation for Research and Treatment of Cancer International Conference, "Molecular Targets and Cancer Therapeutics," held from Nov. 15 to 19 in Boston.