The application of nonsense-mediated mRNA decay inhibition to the identification of breast cancer susceptibility genes

The Kathleen Cuningham Foundation Consortium for Research into Familial Breast Cancer (kConFab; http://​www.​kconfab.​org) provides a comprehensive resource upon which researchers can draw data and biospecimens for peer-reviewed, ethically- approved funded research projects on familial aspects of breast cancer. We selected three families (Family A, B and C) for use in GINI (Table 1), in which no BRCA1 or BRCA2 mutations had been found, and which had Manchester scores >19 [30, 31] suggesting a high probability of having a mutation in a breast cancer susceptibility gene. For each family, there were at least three LCLs from women affected with breast cancer at ages 34–56, and at least three from women who were unaffected at ages 26–72. Ethical approvals were obtained from the Human Research Ethics Committees of the Queensland Institute of Medical Research and the Peter MacCallum Cancer Centre. Written informed consent was obtained from each participant.

We used the colon cancer cell line, HT29, which contains a truncating mutation (c.931C > T p.Q311X) in one allele of the SMAD4 gene and a deletion of the other allele, as a positive control, as well as two positive control LCLs (BRCA1 c.2681_2682delAA and BRCA2 c.539_541insAT). These truncating mutations occur >55 nucleotides upstream from the final exon of each gene and are thus expected to undergo nonsense-mediated mRNA decay. To identify putative novel, breast cancer susceptibility genes by GINI, we used 24 LCLs established from patients affected with breast cancer and unaffected individuals from three non-BRCA1/2 breast cancer families.

Twenty-four hours prior to caffeine treatment, we seeded HT29 cells into two 75cm² cell culture flasks, each containing 1x10⁶ cells in 10mls of tissue culture medium (RPMI-1640 + 10% FBS). We then added fresh medium containing 10mM caffeine (Sigma-Aldrich, St. Louis, MO, USA) to one flask and fresh medium without caffeine to the control flask. After four hours of incubation at 37°C, we removed the medium from both flasks and washed the cells twice with phosphate-buffered saline (PBS). We then added normal medium to the untreated control flask and treated the other flask for a further four hours with 10mM caffeine medium. We then removed the media from both flasks and washed the cells twice with PBS before storing the cells at −80°C until RNA extraction. We repeated this process ten times to ensure repeatability of results. We then performed semi-quantitative real-time reverse transcriptase (RTPCR) of SMAD4 to confirm stabilisation of the target gene after caffeine treatment, and cRNA from four randomly selected replicates were then hybridised to the Illumina HumanHT-12 v3 gene expression arrays.

In order to determine the optimal caffeine concentration for treating LCLs, we treated 3.5x10⁶ LCLs with 2.5mM, 5mM, 7.5mM, 10mM or 15mM caffeine (Sigma-Aldrich) for two lots of four-hour incubations at 37°C in the same GINI process described above for HT29 cells. We determined the optimal concentration by semi-quantitative real-time RT-PCR of BRCA1 and BRCA2 (Additional File 1) and then performed three technical replicates of GINI on the positive control LCLs on different days, and hybridised the cRNA to Illumina HumanHT-12 v3 gene expression arrays.

Having optimised the GINI method on positive control LCLs with BRCA1 or BRCA2 mutations, we applied it to LCLs from 11 affected and 13 unaffected individuals from three non-BRCA1/2 breast cancer families, in parallel with positive control HT29 cells. We performed three technical replicates for each sample on different days and then hybridised cRNA from caffeine-treated and untreated control samples to Illumina HumanHT-12 v3 gene expression arrays.

We extracted total RNA from frozen cell pellets using the RNeasy RNA Extraction Kit (Qiagen, Hilden, Germany) as per the manufacturer’s instructions. We used semi- quantitative real-time reverse transcriptase PCR (RT-PCR) to validate the expression of the target genes in the positive controls and of the candidate genes found by the GINI method. Primer sequences are provided in Additional File 2. We synthesised cDNA from 1μg of total RNA in a reaction volume of 20μl using oligodT and Superscript III (Invitrogen) according to manufacturer’s instructions. The cDNA reactions were then diluted to 200μl (1:10) in RNase/DNase-free water. We performed semi-quantitative real-time RT-PCR amplifications in quadruplet on the LightCycler480 (Roche) machine using 3.75μl SYBR Green PCR Master Mix (Invitrogen) and 0.33μM forward and reverse primers in a total reaction volume of 7.5μl, using the glyceraldehydes-3-phosphate dehydrogenase (GAPDH) gene as a reference. We normalised the comparative threshold-cycle (Ct) of signal intensities of amplified product to that of GAPDH calibrating to the matched untreated control of each sample using the LightCycler480 software program (Roche Diagnostics Corp., Indianapolis, IN, USA).

For the expression profiling, we prepared biotinylated cRNA from 450ng of total RNA using the IlluminaTotalPrep RNA Amplification Kit (Ambion, Austin, TX, USA) and hybridised 750ng cRNA per sample to HumanHT-12 v3 Expression BeadChips (Illumina Inc., San Diego, CA, USA) as per the Whole-Genome Gene Expression Direct Hybridisation Assay protocol. We collated expression data using BeadStudio Version 1.5.1.3 (Illumina Inc.) and then quantile normalised the raw data using the R-Bioconductor LUMI package [32]. Microarray data were submitted to Gene Expression Omnibus (GEO) [33] and are accessible through GEO (http://​www.​ncbi.​nlm.​nih.​gov/​geo/​query/​acc.​cgi?​acc=​GSE37210). To identify differentially expressed genes, between each caffeine treated sample to its own untreated control over three technical replicates, we implemented a linear model and empirical Bayes method using the R-Bioconductor LIMMA package [34]. We considered genes with Benjamini and Hochberg adjusted P-values less than 5% that also have a greater than 50% chance of being differentially expressed (positive B- statistic) as having statistically significant changes in their expression profiles. We imported differentially expressed genes into GeneSpring v10.0 (Agilent Technologies) for visualisation and further analysis.

We defined genes that were differentially expressed between all untreated and all caffeine treated samples as “global caffeine response” genes. These genes are unlikely to contain truncating mutations in breast cancer susceptibility genes but are more likely to be expressed in response to caffeine or be naturally occurring NMD targets.

For each family, we selected candidate genes for semi-quantitative real-time RT-PCR validation and mutations screening that were upregulated after caffeine treatment in the individuals affected with breast cancer but were not differentially expressed after caffeine treatment in their unaffected family members, and therefore not part of the “global caffeine response” gene list.

We sequenced the exonic and flanking splice sites of the one candidate gene we identified by GINI in the relevant family. We amplified 500ng DNA by polymerase chain reaction (PCR) with 1X PCR Buffer, 2.5mM MgCl₂, 0.2mM dNTP, 0.25μM forward and reverse primer, 0.4μl AmpliTaq Gold in a total reaction volume of 50μl. We then purified the PCR products using the QIAquick PCR Purification Kit Spin Protocol (Qiagen) prior to sequencing according to manufacturer’s instructions using BigDye Terminator v3.1 and an ABI 3100 Genetic Analyser (PE Applied Biosystems, Foster City, CA, USA).

We used a total of six short tandem repeat microsatellite markers from a 2Mb region around PPARGC1A to haplotype around the PPARGC1A gene in members of Family B. We selected four markers (D4S3017, D4S2953, D4S425, D4S23013) (deCODE Genetics; http://​www.​decode.​com) and designed two (19xTA and 20xTG) more around short tandem repeats within the DNA sequence of PPARGC1A and labelled either the forward or reverse primer with a 6-FAM label. We amplified 500ng DNA by polymerase chain reaction (PCR) with 1X PCR Buffer, 2.5mM MgCl₂, 0.2mM dNTP, 0.25μM forward and reverse primer, 0.4μl AmpliTaq Gold in a 50μl final volume. Amplified PCR products were diluted 1:800 with Hi-Di formamide/500 LIZ size standard mix (formamide to size standard ratio of 67:1) (Applied Biosystems) and the mixture was then denatured for 5 min at 95°C. The samples were separated in an ABI PRISM3700 DNA Sequencer and subjected to the fragment analysis protocol. Allele scoring was performed using the GeneMapper 4.0 software (Applied Biosystems).

We used HT29 cells to test the robustness of the GINI technique with caffeine treatment. Results from semi-quantitative real-time RT-PCR validation of ten replicate experiments identified a consistent 2.5- to 3.7-fold upregulation of SMAD4 mRNA after 10mM caffeine treatment (Figure 1). Microarray results of three replicates identified a total of 553 probes corresponding to 495 genes that were significantly differentially expressed ≥2-fold between untreated and caffeine-treated HT29 cells. When sorted by adjusted P-value, SMAD4 ranked as the 48^th most differentially expressed gene and the 28^th most significantly upregulated gene (fold change = 2.78; adj P-value = 7.83-E07; B-statistic = 12.94).

In parallel with positive control HT29 cells, we treated LCLs with known BRCA1 or BRCA2 mutations with a caffeine concentration gradient and used semi-quantitative real-time RT-PCR to find the concentration that best stabilises their respective target genes (Additional File 1). The highest level of target gene stabilisation in the LCLs was 2.4-fold at 7.5mM caffeine. We repeated the GINI experiments a total of three times for the BRCA1 and BRCA2 cell lines, and each replicate of every cell line showed at least a 1.5-fold stabilisation of their respective target genes after treatment with 7.5mM caffeine (Figure 2).

We identified candidate protein truncated transcripts within each family as those with statistically significant increased expression after caffeine treatment in affected individuals of that family, but not in unaffected individuals of the other two families, and which were not part of the “global caffeine response” (Figure 3). We identified a total of two, two and five candidate genes for Family A, B and C, respectively (Table 2). Interestingly, only one gene, peroxisome proliferator-activated receptor-γ coactivator-1 α (PPARGC1A), was identified as a candidate in more than one family.

We tried to validate all nine candidate genes in with semi-quantitative real-time RT-PCR but only PPARGC1A showed consistent stabilisation of mRNA transcript across replicates of Family B in individuals affected by breast cancer and not in their healthy family members (Figure 4 and Additional File 3 9, Additional File 4, Additional File 5, Additional File 6, Additional File 7, Additional File 8). Primer sequences are provided in Supplementary Table 1. The youngest individual in Family B (individual 4: aged 26) showed statistically significant stabilisation of the PPARGC1A transcript in two out of three experimental replicates, suggesting that this individual may be a carrier for a family-specific breast cancer mutation but has not yet developed disease. However, sequencing of all coding regions in all affected and unaffected family members and flanking splice sites did not reveal any protein truncating mutations in PPARGC1A. We identified three coding (rs2970847, rs3755863 and rs8192678) and one non-coding (rs2946385) single nucleotide polymorphisms (SNPs). The only variant identified that was not listed in dbSNP as a common polymorphism was IVS7delT (Figure 5). However, this variant was found in all members sequenced including affected and unaffected individuals. Furthermore, Human Splicing Finder [35] suggested that this intronic variant has no effect on splicing. Haplotype analysis using short tandem repeat microsatellite markers spanning 2Mb around PPARGC1A showed no evidence of a haplotype that might carry a protein truncating mutations segregating with disease within the family (Figure 5).