Abstract
Purpose
Inherited variants in the cancer susceptibility genes, BRCA1 and BRCA2 account for up to 5% of breast cancers. Multiple gene expression studies have analysed gene expression patterns that maybe associated with BRCA12 pathogenic variant status; however, results from these studies lack consensus. These studies have focused on the differences in population means to identified genes associated with BRCA1/2-carriers with little consideration for gene expression variability, which is also under genetic control and is a feature of cellular function.
Methods
We measured differential gene expression variability in three of the largest familial breast cancer datasets and a 2116 breast cancer meta-cohort. Additionally, we used RNA in situ hybridisation to confirm expression variability of EN1 in an independent cohort of more than 500 breast tumours.
Results
BRCA1-associated breast tumours exhibited a 22.8% (95% CI 22.3–23.2) increase in transcriptome-wide gene expression variability compared to BRCAx tumours. Additionally, 40 genes were associated with BRCA1-related breast cancers that had ChIP-seq data suggestive of enriched EZH2 binding. Of these, two genes (EN1 and IGF2BP3) were significantly variable in both BRCA1-associated and basal-like breast tumours. RNA in situ analysis of EN1 supported a significant (p = 6.3 × 10−04) increase in expression variability in BRCA1-associated breast tumours.
Conclusion
Our novel results describe a state of increased gene expression variability in BRCA1-related and basal-like breast tumours. Furthermore, genes with increased variability may be driven by changes in DNA occupancy of epigenetic effectors. The variation in gene expression is replicable and led to the identification of novel associations between genes and disease phenotypes.
Similar content being viewed by others
Introduction
Gene expression profiles have been used extensively in the study of cancer development, treatment response and prognosis. In particular, gene expression signatures have been developed to classify tumour subtypes [1, 2], predict response to endocrine treatment [3, 4], indicate prognosis [3, 5, 6] and predict tumour recurrence [4]. The development of these signatures has relied on capturing a change in average expression between biological groups (e.g. poor responders versus good responders) and subsequently validating these results in independent datasets. The variability in gene expression levels (e.g. across samples) is often ignored and is generally only considered in terms of the impact it has on statistical power. However, gene expression variability has been shown to be under genetic control and important to cellular function [7,8,9,10,11]. Gene expression variability has been used to evaluate transcriptomic data in human disease and development [12,13,14,15]. By isolating individual embryonic cells, researchers have shown that gene expression variability provides insights into gene regulation that is essential throughout embryonic development [12]. Differences in gene expression variability have also been investigated to improve our understanding of cancer biology. Although a few studies have been reported to date, this approach has led to the identification of a pan-cancer gene set [16], a classifier for chronic lymphocytic leukaemia [17] and synthetic lethal genes in BRCA2-associated ovarian tumours [18]. These studies measured gene expression variability of whole-transcriptome data generated from microarray or RNA-sequencing platforms. Two of these studies highlighted the utility of measuring global gene expression variability from microarray data to classify leukaemia subtypes [17] and to identify a 48 gene set as a predictive marker of cancer metastatic potential and patient survival [16]. One study identified genes from RNA-sequencing data with potential synthetic lethal interactions with BRCA2 in ovarian cancer [18]. Genetic variants in BRCA1 and BRCA2 predispose women to breast and ovarian cancers and are believed to contribute to 5–10% of all breast cancers and 20–40% of familial breast cancers [19, 20]. Despite representing distinct tumour types, there has been limited success in identifying gene expression profiles related to BRCA1 and BRCA2 pathogenic variants in either tumour [3, 21,22,23,24] or normal tissue [25,26,27,28]. Across all studies, consensus on altered genes and pathways has been poor, with gene expression profiles influenced by study design rather than variant classification [29]. For example, early studies were confounded by differences in oestrogen receptor (ER) status of BRCA1-associated tumours compared to sporadic [21]. The ER status of breast tumours was later identified to be a major driver of gene expression changes [3]. These studies have also overlooked the variability in gene expression and whether these phenotypes are associated with the presence of a pathogenic variant.
In this study, we investigate transcriptomic data across multiple breast tumour datasets using differential variability analysis to identify genes that are associated with BRCA1 and BRCA2 pathogenic variant status and with different tumour subtypes. We also utilised RNA in situ hybridisation as an orthogonal approach to validate inter-tumour expression variability of a candidate gene in an independent cohort of breast tumours.
Methods
Data collection
For gene expression variability analyses, we included any microarray dataset containing gene expression profiles on at least 50 breast tumours, of which > 25 samples were either BRCA1- or BRCA2-associated breast tumours (Table 1). Raw data were acquired through GEO (https://www.ncbi.nlm.nih.gov/geo/) for three datasets (GSE19177, GSE27830 and GSE49481) generated on Illumina, Affymetrix and Agilent microarray platforms, respectively [23, 30, 31]. Raw intensities collected on the Illumina and Agilent arrays were normalised using quantile normalisation. Additionally, for the Agilent array only the intensity values (CY5 channel) corresponding to breast tumour RNA were considered. Raw intensities obtained on the Affymetrix arrays were normalised using the RMA algorithm [32].
In addition, a retrospective microarray dataset of well-curated breast cancer gene expression profiles [33] was utilised for a subtype-centric analysis [33]. Briefly, all samples were profiled on one of the Affymetrix U133A, U133A2 or U133plus2 GeneChip arrays. For consistency, only probes common to all arrays were retained. Individual arrays were normalised using the RMA method and batch effects were corrected using the COMBAT method [34]. The resulting dataset consisted of 2116 breast tumours and 22,268 probes.
Lastly, we accessed the METABRIC dataset through cbioportal (www.cbioportal.org/), in order to investigate DV genes relationship with molecular features.
Transcriptome-wide gene expression variability analysis
Transcriptome-wide gene expression variability was compared between BRCA1-associated tumours and non-BRCA1/2 (BRCAx) hereditary breast tumours, BRCA2-associated tumours and BRCAx and Basal-like and non-basal (luminal A, luminal B, HER2 and normal-like) tumours. To assess transcriptome-wide changes, tumour type-specific standard deviation (SD), coefficient of variance (CV) and median absolute deviation (MAD) were calculated for each gene. Pairwise linear regression models were calculated between tumour groups for each gene-specific statistics (i.e. SD, CV, MAD and mean) and the resulting β (slope coefficient) was compared to a model of tumour equity (β = 1). We defined the difference in expression variability between tumour groups as the percentage change in the slope coefficient compared to model of tumour equity (1 – β × 100). Additionally, pairwise polynomial regression was performed to investigate non-linear relationships between tumour-specific statistics.
Differential expression variability analysis
Genes were considered to exhibit differential expression variability (hereafter referred to as differentially variable—DV-genes) if the spread of expression values differed significantly between two tumour groups. For these analyses, the same tumour comparison was made as in the transcriptome-wide analyses (i.e. BRCA1 versus BRCAx, BRCA2 versus BRCAx and basal-like versus non-basal). Robust measures of spread were used to avoid spuriously large differences caused by outlying expression values. Spread was quantified by the median absolute deviation (MAD) and groups were compared by the Brown–Forsythe method [35], essentially an ANOVA on expression values about their medians, with p-values adjusted for multiple testing [36] was used to determine significance. Ratios of MADs with 95% confidence intervals [37] were used to quantify changes in spread between groups.
Breast cancer tissue microarrays
Women diagnosed with breast cancer were identified from 1800 families recruited into the Kathleen Cunningham Consortium for Research into Familial Breast Cancer (kConFab) [38]. For inclusion in this consortium, families must have a strong family history of breast and/or ovarian cancer, or be known to be segregating a germline variant in genes such as BRCA1 and BRCA2 (see www.kconfab.org for recruitment criteria). Breast cancer cases on tissue microarrays (TMAs) were verified from pathology reports. Ethics approval was obtained from the HREC at the Peter MacCallum Cancer Centre (97/27) and through the University of Otago Human Ethics Committee (H14/088). Informed consent at study entry was obtained from all participants, allowing access to medical/treatment reports, blood collection and tumour tissue collections. For deceased participants proxy consent was obtained from the next of kin. Where applicable, cause of death was verified from a death certificate, doctor or hospital medical records. Treatment and medical notes were accessed through physicians, hospitals, laboratories and State Cancer Registries.
Confirmation of a participant’s germline mutation status was performed using a variety of sequencing platforms in the Molecular Pathology NATA-accredited clinical laboratory. Variants were assigned a class C4–C5 (pathogenic) mutation status according to a 5-tier clinical classification introduced by ENIGMA (http://www.enigmaconsortiumorg).
RNA in situ hybridisation (RNAscope)
The mRNA expression level for EN1 was investigated using the RNAscope 2.0 BROWN assay (Advanced Cell Diagnostics, Inc.) following manufacturers’ instructions. Additionally, mRNA levels of PPIB and the bacterial gene dapB were used as positive and negative controls, respectively. Briefly, TMA sections were deparaffinised in a series of xylene and 100% ethanol steps. Sections were subjected to a series of pre-treatments before each section was incubated with target or control probes for 2 h at 40 °C in a HybEZ™ Oven. Probe signal was amplified through a series of amplification steps and colour development was done using diaminobenzidine (DAB).
Scoring/counting signals
All TMA cores were scored for abundance of PPIB and DapB signals where a positive signal was assessed as a brown punctuate dot within a cell. Supplementary Table 1 describes the criteria for each score. Briefly, a score of ‘0’ described an absence of signal, ‘0.5’ was a weak stain with < 30% of cells having a signal, ‘1’ was a modest stain with > 30% of cells with a positive signal, ‘2’ was considered moderate staining with a greater number (4–9) of signals per cell, ‘3’ was strong staining with the presence of signal clustering in < 10% of cells and ‘4’ was intense staining with signal clustering in > 10% of cells.
All TMA sections were scanned with the Aperio Scanner and digital scans were used to quantify mRNA signals using the LEICA RNA ISH algorithm (Leica Microsystems GmbH, Germany). Briefly, the algorithm was trained on a selection of cores from each TMA and across a range of genotypes. These settings were applied to all the cores stained with probes targeting EN1 RNA.
Transcription regulator enrichment analysis
To identify DNA binding elements that were overrepresented in a set of DV genes, the TFEA.ChIP package in R was used [39]. ChIP-Seq datasets were obtained through the TFEA.ChIP github page (https://github.com/LauraPS1/TFEA.ChIP_downloads). Briefly, datasets from the ENCODE Consortium, DeMap and individual GEO database were included in the analysis, and gene sequences were annotated based on transcription regulator binding within 1 kb of each gene.
Statistical analysis
R (version 3.6.1) was used to normalise microarrays and perform all statistical analyses. For microarrays on the Affymetrix platform, the RMA normalisation [32] from the affy package was applied, whilst for each of the Agilent and Illumina arrays, the quantile normalise method from the limma package was used [40]. The lawstat package was used to calculate Brown–Forsythe test. To account for multiple testing, p-values were adjusted using the Benjamini–Hochberg procedure [36].
Results
Transcriptome-wide gene expression variability
Differences in transcriptome-wide gene expression variability between familial breast cancer groups (BRCA1, BRCA2 and BRCAx) were quantified by comparing tumour-specific measurements of variability (SD, CV and MAD). BRCA1-related breast tumours share multiple biological properties with sporadic basal-like breast cancer [41], therefore we also assessed gene expression variability differences between the basal and non-basal breast tumour subtypes. BRCA1-associated and basal-like breast tumours displayed greater transcriptome-wide variability compared to non-BRCA1 and non-basal-like (luminal A, luminal B, HER2 and normal-like) tumours (Fig. 1, Supplementary Fig. 1). Using gene-specific standard deviations, BRCA1-associated breast tumours had a 22.8% (95% CI 22.3–23.2) increase in gene expression variability. Increased variability in BRCA1-associated breast tumours was also observed in linear models of gene-specific CVs (25.0%, 95% CI 24.5–25.6) and MADs (32.4%, 95% CI 31.9–33.0). Similarly, basal-like tumours had an average 28.2% (95% CI 27.7–28.7) greater transcriptome-wide expression variability compared to non-basal tumours. Gene expression variability in BRCA2-associated tumours was inconsistent, with the Waddell dataset showing comparable variability with BRCAx tumours (Supplementary Fig. 2), but the Larsen dataset showing only a modest 11.1% (95% CI 10.6–11.6) increase in variability in BRCA2-associated tumours compared to BRCAx tumours. The Nagel dataset had too few BRCA2-associated tumours to reliably estimate transcriptome-wide expression variability. In contrast to gene expression variability, no changes were observed in transcriptome-wide mean expression for any tumour groups tested.
Differential expression variability in breast tumours
As BRCA1-associated and basal-like breast tumours displayed greater transcriptome-wide expression variability, it was of interest to identify specific genes that had altered variability between tumour types. After adjusting for multiple comparisons, 503 and 337 genes were found to be significantly differentially variable between BRCA1- and BRCAx-associated breast tumours in the Nagel and Larsen datasets, respectively. There were 40 DV genes that were consistent in both the Nagel and Larsen datasets (Table 2). The directionality of the gene expression variability was consistent for all 40 DV genes. Additionally, 36 of the 40 BRCA1-associated DV genes were also DV between BRCA1-associated and sporadic breast tumours. A total of 185 basal-like-associated DV genes were identified across the four breast tumour datasets, with 184/185 (99.5%) consistent in direction. Analysis of the meta-cohort data, which comprised 2116 samples, found 58% of all genes to be DV. To identify candidate genes for in situ expression analysis, we compared the top 1.5% most significant DV genes across all datasets and between subtype and genotype analyses (Fig. 2A). A total of 10 BRCA1-associated and 22 basal-like-associated variable genes were identified in all datasets at 1.5% threshold. EN1 and IGF2BP3 were variably expressed in both BRCA1-associated and basal-like breast tumours (Fig. 2B).
EN1 gene expression variability
We used the publically available METABRIC dataset [42] to further interrogate EN1 gene expression in 1904 breast tumours (Fig. 2C, D). Consistent with analysis of the four breast cancer microarray datasets, basal-like breast tumours had significantly greater EN1 expression variability (Fig. 2C, p = 3.7 × 10−146). Additionally, as BRCA1-associated breast tumours are typically oestrogen receptor (ER) negative we explored the correlation with ESR1. ER negative tumours had significantly greater expression variability (Fig. 2D, Brown–Forsythe test p = 2.19 × 10−106) and tumours with low ESR1 expression had a greater range of expression. Lastly, to determine if gene dosage was the cause of variable expression we measured the correlation of copy number with EN1 expression. There was no correlation between copy number and gene expression (r2 = 0.026) or expression variability (r2 = 0.023, Supplementary Fig. 3).
To assess intra-tumoural expression of EN1, breast tumour tissue microarray cores from 503 patients were assayed using RNA in situ hybridisation. These samples included tumour tissue from 151 BRCA1-associated and 124 BRCA2-associated breast cancer cases (Supplementary Tables 2, 4). Tissue microarrays were stained and scored for the abundantly expressed PPIB and a negative control targeting dapB. Tumour cores with scores > 2 for the PPIB and 0 for dapB were considered high quality and used to investigate EN1 expression. Two-hundred-and-thirteen tumours had positive PPIB mRNA signals with 141/213 tumours being scored 2 or greater for PPIB staining. No dapB signal was detected in any tumours.
The LEICA RNA ISH Algorithm was used to estimate the abundance of RNA signals and the percentage of positively stained cells. Consistent with our transcriptome analysis, in situ expression analysis of BRCA1-associated tumours showed significantly greater EN1 expression variability (Fig. 3, p = 6.7 × 10−04).
Transcription regulators enrichment analysis
To identify transcription regulators that may be responsible for increased BRCA1-associated expression variability, we used the TFEA.ChIP package in R to identify transcription regulators overrepresented in the 40 BRCA1-associated DV genes (Table 2). Firstly, we identified individual ChIP-seq datasets associated with BRCA1-associated DV genes (Fig. 4) and then summarised these results at the transcription regulator level in order to identify transcription regulators that are enriched or depleted within BRCA1-associated DV genes (Supplementary Table 3). The top ranked ChIP-seq dataset in human mammary epithelial cells (HMEC) was associated with EZH2 as shown in (Supplementary Table 3). Further, EZH2 was the top ranked transcription regulator associated with BRCA1-related DV genes (Supplementary Table 3). Interestingly, EZH2 was significantly overexpressed in BRCA1-associated and basal-like breast tumours (Fig. 5).
Discussion
The robust and replicable nature of gene expression measurements provides an excellent opportunity to investigate biological variability [11]. We explored publicly available familial breast cancer microarray datasets for phenotypes associated with BRCA1- and BRCA2-related breast tumours. In addition, a cohort of 2116 breast tumours profiled on Affymetrix microarray platforms was included to identify gene expression variability in basal-like tumours. In this study, we identified that BRCA1-associated and basal-like breast tumours displayed greater gene expression variability compared to non-BRCA1 and non-basal tumours, respectively. These observations were consistent across all microarray datasets explored and across three different measures of variability. Furthermore, no significant transcriptome-wide changes in mean expression were observed between any tumour groups. Together these results suggest that globally expression variability associates with phenotypic features of BRCA1-associated cancer to a stronger degree than mean expression and at an individual gene level, associations may not be statistically significant if only mean expression level is examined.
A number of studies have used transcriptome variability in humans to described differences in disease and development states [12, 14, 16,17,18]. One study by Bueno and Mar [18] has explored gene expression variability to identify synthetic lethal genes associated with BRCA2 loss-of-function ovarian tumours. The authors proposed that 54 stably expressed (low variable) genes in BRCA2-related tumours may be essential in BRCA2-related tumour viability. However, there was no formal statistical test that estimated the association of gene expression variability. Our analysis in BRCA2-associated breast tumours suggested that no genes were differentially variable. However, it is possible that the small number of BRCA2 samples in the microarray datasets hindered the identification of BRCA2-associated DV genes.
Transcriptome-wide gene expression variability may be driven by tumour heterogeneity, aggressiveness and cellular content. For example, more aggressive lymphocytic leukaemias are associated with greater gene expression variability [17], an observation that is consistent with the greater gene expression variability seen in BRCA1-associated and basal-like breast tumours. Furthermore, the epigenetic status of cells is a heritable trait that also has significant variability [43]. Specifically, DNA methylation can contribute to gene expression variability [44] and the methylation status of tumour cells may influence gene expression variability described in this study. Consistent with this, the top ranked transcription regulator associated with 40 BRCA1 DV genes was EZH2, a component of the Polycomb repressive complex 2 (PRC2). PRC2 is important for H3 lysine 27 trimethylation (H3K27me3) and the stable repression of transcription [45]. EZH2 has been shown to be overexpressed in a number of tumours including BRCA1-deficient breast cancers [45, 46]. In addition, loss of BRCA1 function and the decrease in BRCA1 expression can alter the occupancy of EZH2 on DNA and increase H3K27me3 levels [47]. These BRCA1-related changes in epigenetic regulation are potential mechanisms that may alter transcriptional control, ultimately leading to increased cellular gene expression variability.
Technical variation is expected to contribute randomly to each sample; however, as there was no standardised protocol between and within the datasets used, it is plausible that processes such as sample collection and RNA extraction may influence variability. By considering only genes that were variably expressed across all microarray datasets we are able to reduce false-positive associations induced by these types of experimental artefacts. Our approach identified two genes (EN1 and IGF2BP3) that had increased variability in BRCA1-associated breast tumours. EN1 encodes the transcription factor Engrailed Homoeobox 1 and has been extensively investigated in neuronal development. Ectopic EN1 expression improves neurons’ survival and protects against apoptosis [48]. Conversely, knockdown of expression in dopaminergic neurons has been shown to induce apoptosis [49]. EN1 has been observed to be overexpressed in triple-negative breast cancers and basal-like breast cancers [50,51,52]. SDs and inter-quartile ranges reported by these studies were suggestive of increased gene expression variability similar to that seen in this study; however, these were not formally tested. Additionally, EN1 expression has been associated with poorer survival and a greater probability of brain metastases. Interestingly, reduced expression of EN1 in basal, but not luminal, breast cancer cell lines decreases viability, which can be partially rescued by overexpression of EN1 [52]. The evaluation of EN1 protein by immunohistochemistry has been perplexing, with Kim and colleagues previously describing EN1 protein expression as being associated with improved survival in triple-negative breast cancer, opposite to that of EN1 RNA [51]. The discrepancy may in part be due to poor antibodies targeting EN1 [52] or post-transcriptional processes. In this study, we have investigated the utility of RNAscope to measure gene expression in situ and we were able to recapture the inter-tumour variability observed in the microarray analysis. The implementation of RNA in situ hybridisation can overcome the issue of antibody specificity and may help facilitate replication of survival trends. However, the lack of survival data limited our ability to formally test for associations. Currently, measurement of intra-tumoural variability remains challenging, particularly due to the discrimination of signal in dark-stained chromatin and granulated nuclei. The development of more sophisticated algorithms in the future may provide greater power to assess variability within individual sections and allow the ability to test for associations with clinicopathological data. Furthermore, to fully appreciate tumour variability, complete tumour sections may be required for future studies.
Conclusion
BRCA1-associated and basal-like breast tumours displayed a phenotype of greater gene expression variability, with no change in global RNA abundance. EN1 had greater expression variability in BRCA1-associated breast tumours, and this was captured in transcriptomic and RNA ISH analyses. We conclude that the expression variability of a gene is replicable, thereby laying a foundation for future studies aiming to better understand the molecular mechanisms underlying the development of basal-like breast tumours.
Data availability
Original mircoarray datasets (GSE19177, GSE27830 and GSE49481) can be found in https://www.ncbi.nlm.nih.gov/geo/. The METABRIC dataset [42] was accessed through cbioportal (https://www.cbioportal.org/). The meta-cohort has been previously described [33, 34].
Abbreviations
- SD:
-
Standard deviations
- CV:
-
Coefficient of variance
- MAD:
-
Median absolute deviation
- RNA ISH:
-
RNA in situ hybridisation
- ER:
-
Oestrogen receptor
- DV:
-
Differentially variable
References
Parker JS, Mullins M, Cheang MCU et al (2009) Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol 27:1160–1167. https://doi.org/10.1200/JCO.2008.18.1370
Prat A, Parker JS, Karginova O et al (2010) Phenotypic and molecular characterization of the claudin-low intrinsic subtype of breast cancer. Breast Cancer Res 12:R68. https://doi.org/10.1186/bcr2635
van’t Veer LJ, Dai H, van de Vijver MJ et al (2002) Gene expression profiling predicts clinical outcome of breast cancer. Nature 415:530–536. https://doi.org/10.1038/415530a
Paik S, Shak S, Tang G et al (2004) A Multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 351:2817–2826. https://doi.org/10.1056/NEJMoa041588
Xie L, Cai L, Wang F et al (2020) Systematic review of prognostic gene signature in gastric cancer patients. Front Bioeng Biotechnol 8:805. https://doi.org/10.3389/fbioe.2020.00805
Wilson CS, Davidson GS, Martin SB et al (2006) Gene expression profiling of adult acute myeloid leukemia identifies novel biologic clusters for risk classification and outcome prediction. Blood 108:685–696. https://doi.org/10.1182/blood-2004-12-4633
Spielman RS, Bastone LA, Burdick JT et al (2007) Common genetic variants account for differences in gene expression among ethnic groups. Nat Genet 39:226–231. https://doi.org/10.1038/ng1955
Raj A, Rifkin SA, Andersen E, van Oudenaarden A (2010) Variability in gene expression underlies incomplete penetrance. Nature 463:913–918. https://doi.org/10.1038/nature08781
Zhang Z, Qian W, Zhang J (2009) Positive selection for elevated gene expression noise in yeast. Mol Syst Biol 5:1–12. https://doi.org/10.1038/msb.2009.58
Eraly SA (2014) Striking differences between knockout and wild-type mice in global gene expression variability. PLoS One 9:e97734. https://doi.org/10.1371/journal.pone.0097734
Osorio D, Yu X, Zhong Y et al (2019) Single-cell expression variability implies cell function. Cells 9:14. https://doi.org/10.3390/cells9010014
Hasegawa Y, Taylor D, Ovchinnikov DA et al (2015) Variability of gene expression identifies transcriptional regulators of early human embryonic development. PLOS Genet 11:e1005428. https://doi.org/10.1371/journal.pgen.1005428
Li J, Liu Y, Kim T et al (2010) Gene expression variability within and between human populations and implications toward disease susceptibility. PLoS Comput Biol 6:e1000910. https://doi.org/10.1371/journal.pcbi.1000910
Zhang F, Shugart YY, Yue W et al (2015) Increased variability of genomic transcription in schizophrenia. Sci Rep. https://doi.org/10.1038/srep17995
Wiggins GAR, Black MA, Dunbier A et al (2021) Variable expression quantitative trait loci analysis of breast cancer risk variants. Sci Rep 11:1–10
Yu K, Ganesan K, Tan LK et al (2008) A precisely regulated gene expression cassette potently modulates metastasis and survival in multiple solid cancers. PLoS Genet 4:e1000129. https://doi.org/10.1371/journal.pgen.1000129
Ecker S, Pancaldi V, Rico D, Valencia A (2015) Higher gene expression variability in the more aggressive subtype of chronic lymphocytic leukemia. Genome Med 7:8. https://doi.org/10.1186/s13073-014-0125-z
Bueno R, Mar JC (2017) Changes in gene expression variability reveal a stable synthetic lethal interaction network in BRCA2-ovarian cancers. Methods. https://doi.org/10.1016/j.ymeth.2017.07.021
Ripperger T, Gadzicki D, Meindl A, Schlegelberger B (2009) Breast cancer susceptibility: current knowledge and implications for genetic counselling. Eur J Hum Genet 17:722–731. https://doi.org/10.1038/ejhg.2008.212
Turnbull C, Rahman N (2008) Genetic predisposition to breast cancer: past, present, and future. Annu Rev Genomics Hum Genet 9:321–345. https://doi.org/10.1146/annurev.genom.9.081307.164339
Hedenfalk I, Duggan D, Chen Y et al (2001) Gene-expression profiles in hereditary breast cancer. N Engl J Med 344:539–548. https://doi.org/10.1056/NEJM200102223440801
Larsen MJ, Kruse TA, Tan Q et al (2013) Classifications within molecular subtypes enables identification of BRCA1/BRCA2 mutation carriers by RNA tumor profiling. PLoS One 8:e64268. https://doi.org/10.1371/journal.pone.0064268
Waddell N, Arnold J, Cocciardi S et al (2010) Subtypes of familial breast tumours revealed by expression and copy number profiling. Breast Cancer Res Treat 123:661–677. https://doi.org/10.1007/s10549-009-0653-1
Jazaeri AA, Yee CJ, Sotiriou C et al (2002) Gene expression profiles of BRCA1-linked, BRCA2-linked, and sporadic ovarian cancers. JNCI J Natl Cancer Inst 94:990–1000. https://doi.org/10.1093/jnci/94.13.990
Walker LC, Thompson BA, Waddell N et al (2010) Use of DNA–damaging agents and RNA pooling to assess expression profiles associated with BRCA1 and BRCA2 mutation status in familial breast cancer patients. PLoS Genet 6:e1000850. https://doi.org/10.1371/journal.pgen.1000850
Pouliot M-C, Kothari C, Joly-Beauparlant C et al (2017) Transcriptional signature of lymphoblastoid cell lines of BRCA1, BRCA2 and non-BRCA1/2 high risk breast cancer families. Oncotarget 8:78691–78712. https://doi.org/10.18632/oncotarget.20219
Kote-Jarai Z, Matthews L, Osorio A et al (2006) Accurate prediction of BRCA1 and BRCA2 heterozygous genotype using expression profiling after induced DNA damage. Clin Cancer Res 12:3896–3901
Kote-Jarai Z, Williams RD, Cattini N et al (2004) Gene expression profiling after radiation-induced DNA damage is strongly predictive of BRCA1 mutation carrier status. Clin Cancer Res 10:958–963
Wiggins GAR, Walker LC, Pearson JF (2020) Genome-wide gene expression analyses of BRCA1- and BRCA2-associated breast and ovarian tumours. Cancers (Basel) 12:3015. https://doi.org/10.3390/cancers12103015
Nagel JHA, Peeters JK, Smid M et al (2012) Gene expression profiling assigns CHEK2 1100delC breast cancers to the luminal intrinsic subtypes. Breast Cancer Res Treat 132:439–448. https://doi.org/10.1007/s10549-011-1588-x
Larsen MJ, Thomassen M, Tan Q et al (2014) RNA profiling reveals familial aggregation of molecular subtypes in non-BRCA1/2 breast cancer families. BMC Med Genomics. https://doi.org/10.1186/1755-8794-7-9
Irizarry RA, Hobbs B, Collin F et al (2003) Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4:249–264. https://doi.org/10.1093/biostatistics/4.2.249
Soon WW, Miller LD, Black MA et al (2011) Combined genomic and phenotype screening reveals secretory factor SPINK1 as an invasion and survival factor associated with patient prognosis in breast cancer. EMBO Mol Med 3:451–464. https://doi.org/10.1002/emmm.201100150
Johnson WE, Li C, Rabinovic A (2007) Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8:118–127. https://doi.org/10.1093/biostatistics/kxj037
Brown MB, Forsythe AB (1974) Robust tests for the equality of variances. J Am Stat Assoc 69:364–367. https://doi.org/10.1080/01621459.1974.10482955
Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B 57:289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x
Arachchige CNPG, Prendergast LA (2019) Confidence intervals for median absolute deviations. Preprint
Thorne H, Mitchell G, Fox S (2011) Kconfab: a familial breast cancer consortium facilitating research and translational oncology. J Natl Cancer Inst Monogr 2011:79–81. https://doi.org/10.1093/jncimonographs/lgr042
Puente-Santamaria L, Wasserman WW, del Peso L (2019) TFEA.ChIP: a tool kit for transcription factor binding site enrichment analysis capitalizing on ChIP-seq datasets. Bioinformatics 35:5339–5340. https://doi.org/10.1093/bioinformatics/btz573
Smyth GK (2004) Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3:1–25. https://doi.org/10.2202/1544-6115.1027
Turner NC, Reis-Filho JS (2006) Basal-like breast cancer and the BRCA1 phenotype. Oncogene 25:5846–5853. https://doi.org/10.1038/sj.onc.1209876
Curtis C, Shah SP, Chin SF et al (2012) The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 486:346–352. https://doi.org/10.1038/nature10983
Palumbo D, Affinito O, Monticelli A, Cocozza S (2018) DNA methylation variability among individuals is related to CpGs cluster density and evolutionary signatures. BMC Genomics 19:229. https://doi.org/10.1186/s12864-018-4618-9
Bashkeel N, Perkins TJ, Kærn M, Lee JM (2019) Human gene expression variability and its dependence on methylation and aging. BMC Genomics 20:1–19. https://doi.org/10.1186/s12864-019-6308-7
Laugesen A, Højfeldt JW, Helin K (2016) Role of the polycomb repressive complex 2 (PRC2) in transcriptional regulation and cancer. Cold Spring Harb Perspect Med. https://doi.org/10.1101/cshperspect.a026575
Puppe J, Opdam M, Schouten PC et al (2019) EZH2 is overexpressed in BRCA1-like breast tumors and predictive for sensitivity to high-dose platinum-based chemotherapy. Clin Cancer Res 25:4351–4362. https://doi.org/10.1158/1078-0432.CCR-18-4024
Wang L, Zeng X, Chen S et al (2013) BRCA1 is a negative modulator of the PRC2 complex. EMBO J 32:1584–1597. https://doi.org/10.1038/emboj.2013.95
Alvarez-Fischer D, Fuchs J, Castagner F et al (2011) Engrailed protects mouse midbrain dopaminergic neurons against mitochondrial complex i insults. Nat Neurosci 14:1260–1266. https://doi.org/10.1038/nn.2916
Albéri L, Sgadò P, Simon HH (2004) Engrailed genes are cell-autonomously required to prevent apoptosis in mesencephalic dopaminergic neurons. Development 131:3229–3236. https://doi.org/10.1242/dev.01128
Beltran AS, Graves LM, Blancafort P (2014) Novel role of Engrailed 1 as a prosurvival transcription factor in basal-like breast cancer and engineering of interference peptides block its oncogenic function. Oncogene 33:4767–4777. https://doi.org/10.1038/onc.2013.422
Kim YJ, Sung M, Oh E et al (2018) Engrailed 1 overexpression as a potential prognostic marker in quintuple-negative breast cancer. Cancer Biol Ther 19:335–345. https://doi.org/10.1080/15384047.2018.1423913
Peluffo G, Subedee A, Harper NW et al (2019) EN1 is a transcriptional dependency in triple-negative breast cancer associated with brain metastasis. Cancer Res. https://doi.org/10.1158/0008-5472.CAN-18-3264
Acknowledgements
We thank Associate Professor Lance Miller (Wake Forest School of Medicine) for his detailed curation of the clinicopathological information associated with the breast cancer meta-cohort used in this study. We wish to thank Heather Thorne, Eveline Niedermayr, all the kConFab research nurses and staff, the heads and staff of the Family Cancer Clinics and the Clinical Follow-Up Study (which has received funding from the NHMRC, the National Breast Cancer Foundation, Cancer Australia and the National Institute of Health (USA)) for their contributions to this resource, and the many families who contributed to kConFab. kConFab is supported by a grant from the National Breast Cancer Foundation and previously by the National Health and Medical Research Council (NHMRC), the Queensland Cancer Fund, the Cancer Councils of New South Wales, Victoria, Tasmania and South Australia and the Cancer Foundation of Western Australia.
Funding
LCW was supported by the Royal Society of New Zealand Rutherford Discovery Fellowships (RDF-UOO1501). GARW was supported by the Health Research Council, Breast Cancer Cure and Breast Cancer Foundation New Zealand (Breast Cancer Research in New Zealand Partnership Grant—15/639).
Author information
Authors and Affiliations
Consortia
Contributions
LCW, JFP, MAB and AD conceived the study. JFP and LW designed the study. GARW and AEMB developed the methodology. GARW and JFP analysed and interpreted the data. kConFab provided breast tumour TMAs. GARW wrote, reviewed and revised the manuscript. All authors have read, reviewed and approved the manuscript.
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Ethical approval
The studies involving human participants were reviewed and approved by the University of Otago Human Ethics Committee (Health)—H14/088. The patients/participants provided their written informed consent to participate in this study. All methods were performed in accordance with the relevant guidelines and regulations.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Wiggins, G.A.R., Black, M.A., Dunbier, A. et al. Increased gene expression variability in BRCA1-associated and basal-like breast tumours. Breast Cancer Res Treat 189, 363–375 (2021). https://doi.org/10.1007/s10549-021-06328-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10549-021-06328-y