The Effects of Soy Supplementation on Gene Expression in Breast Cancer: A Randomized Placebo-controlled Study

Moshe Shike; Ashley S. Doane; Lianne Russo; Rafael Cabal; Jorge S. Reis-Filho; William Gerald; Hiram Cody; Raya Khanin; Jacqueline Bromberg; Larry Norton


J Natl Cancer Inst. 2014;106(9) 

In This Article


This study demonstrates that soy supplementation alters BC-related gene expression. Using multiple molecular approaches and bioinformatics techniques, we identified a large number of cell-cycle and proliferation-associated genes that were overexpressed in BCs from the soy group in patients with elevated plasma genistein. Expression of 21 genes measured by NanoString was altered from pretreatment levels as a consequence of treatment with soy or placebo. Two of these genes, UGT2A1 and FANCC, were upregulated in the soy group, suggesting a treatment effect. DE genes that increased in the soy group included the protumorigenic growth factor receptor FGFR2. To our knowledge, this is the first report to analyze gene expression in patient-matched tumors before and after soy intake.

We found FANCC and UGT2A1 to be altered by soy consumption. While the consequences of their increased expression in human BC are unknown, both genes have the potential to influence BC biology. FANCC encodes a DNA repair protein, and mutations are responsible for the autosomal recessive disorder Fanconi Anemia and may be implicated in BC development.[36] UGT2A1 functions in metabolism, including 17β-estradiol and its enantiomers, and has been implicated in tobacco-related carcinogenesis.[37,38]

Computational analysis of genome-wide microarray data from the patients with high and low levels of genistein (subsets of the soy and placebo groups) revealed overrepresentation of several cell cycle gene categories in the high-genistein gene signature and higher levels of expression of cell cycle gene sets, including targets of E2F-family transcription factors. Gene Set Analysis limited to ER(+) samples yielded similar results as above, suggesting that enrichment of gene sets in tumors of the high-genistein subset could not be explained solely on the basis of differences associated with ER status (Supplementary Table 10, available online Analysis of publically available gene expression data in MCF-7 BC cells exposed to genistein (GSE5705) revealed upregulation of many of the same top-ranked gene sets, as identified in our data (cell cycle categories, targets of E2F and RB1; unpublished observations), which supports our findings.[39]

A subset of tumors from the soy group was notable for increased FGFR2 expression as demonstrated by multiple gene expression techniques. There is extensive evidence that FGFR2 drives cancer growth through its role as a potent oncogene, and increased FGFR2 expression is a marker of poor prognosis in BC.[40–42] We found statistically significant overexpression of FGFR2 by microarray and qPCR in tumors of patients taking soy compared with placebo and increased expression in two paired tumor samples before and after soy supplementation. In one sample, expression was increased from already elevated pretreatment levels, and one could speculate that the initial molecular alteration was reinforced by soy.

Soy intake did not result in statistically significant changes in cell proliferation and apoptosis indices compared with the placebo group. A similar observation was made in healthy breast tissue.[43] In BCs from patients with elevated serum genistein, we observed increased expression of genes and gene sets associated with increased cell proliferation and cell cycle progression. A potential explanation for the discrepancy between Ki67 and gene expression results is that a nutritional intervention such as soy intake may take longer periods of time to influence a phenotype measured by immunohistochemical analysis. A second possibility is related to limitations of the Ki67 IHC. As a consequence of tissue heterogeneity and small amounts of available tissue, the Ki67 assessments in pretreatment core biopsies may not have represented the whole tumor.

Identifying gene expression effects associated with a nutritional intervention such as soy presents numerous challenges. Expression changes from diet intervention are expected to be subtle, and detection of alterations is complicated by the molecular heterogeneity of BCs and the need for large data sets. One solution, as implemented in this study, was to take an exploratory approach in which false discovery rate and corrections for multiple hypothesis testing are often withheld in favor of limiting type-II errors, but with the possibility of increasing type-I errors. Additionally, Gene Set Analysis provides an analytical strategy to detect modest but coordinated changes in the expression of biological pathways and sets of functionally related genes. A second possible solution is analysis of paired samples before and after soy intake. Until recently, gene expression analysis required use of snap-frozen tissue, rarely available from diagnostic core biopsies. NanoString technology allows measurement of gene expression from limited FFPE tissue samples, such as those obtained from diagnostic core biopsies. This technology allowed measurement of gene expression from the same tumor before and after soy consumption. Despite the small sample size, this approach identified consistent yet subtle patterns of altered gene expression associated with soy consumption.

This study has a number of limitations. This was a short-term study utilizing a large daily supplement of soy. The implications of our findings in patients consuming smaller amounts of soy over prolonged periods are unclear. Also, the clinical impact of the subtle changes in gene expression has not been examined. Nevertheless, these data raise concern that soy may exert a stimulating effect on BC in a subset of women.