In an attempt to identify germline mutations in additional high risk breast cancer susceptibility genes, we have optimised and applied the GINI method on lymphoblastoid cell lines derived from the blood of women from multiple-case non- BRCA1/2 breast cancer families. By using positive control cell lines with known truncating mutations, we have determined the optimal concentration of caffeine that results in significant stabilisation of target genes, which is suggestive of successful inhibition of the nonsense-mediated mRNA decay pathway. Microarray analysis of the transcripts stabilised after NMD inhibition by caffeine treatment in women affected with breast cancer compared to their unaffected relatives identified a total of eight different genes across three families. One gene, peroxisome proliferator- activated receptor-γ coactivator-1 α (PPARGC1A), was a candidate gene in two families and was the only breast cancer susceptibility candidate gene that we could demonstrate by semi-quantitative real-time RT-PCR as being consistently upregulated after GINI in affected members of the family, but not in most unaffected relatives.
PPARGC1A is a master transcriptional regulator of mitochondrial oxidative phosphorylation and cellular energy metabolism. The gene is expressed in a broad range of tissues with higher levels of expression detected in tissues with high oxidative capacity, such as heart, skeletal muscle, brown adipocyte, kidney and brain [36–38]. Upregulation of PPARGC1A in response to oxidative stress can suppress the production of reactive oxygen species . PPARGC1A also plays an important role as an estrogen receptor coactivator in the estrogen receptor (ER) pathway by binding and enhancing transactivation of estrogen receptor alpha (ERα) in a ligand-dependent manner [40, 41]. Persistent estrogen mediated mitogen signalling of ERα has been known to stimulate the growth of a large proportion of breast cancers [42–45]. In fact, over half of all breast cancers overexpress ERα . An association study of ~800 BRCA1/2 mutation-negative familial breast cancer cases and over 1,000 controls from Germany found some evidence that the PPARGC1A Thr612Met polymorphism might be a risk factor for familial breast cancer (OR = 1.35, 95% CI 1.00-1.81, P = 0.049), high-risk familial breast cancer (OR = 1.51, 95% CI 1.08-2.12, P = 0.017) and bilateral familial breast cancer (OR = 2.30, 95% CI 1.24-4.28, P = 0.009) . However, haplotype analysis did not identify any additional association with familial breast cancer .
Although, we did not identify any truncating mutations in the coding or splice site regions of PPARGC1A, we did find an IVS-7delT variant in both affected and unaffected individuals of two families. However, haplotyping analysis around PPARGC1A did not identify a haplotype that segregated with disease in either family, which may contain a cryptic, deeply intronic, mutation that causes protein truncation.
Caffeine can impact on the alternative splicing of a subset of cancer-associated genes . For example, caffeine can result in alternatively spliced isoforms of chaperonin- containing TCP1 subunit 3 (CCT3), asparagine synthetise (ASNS), COMM-domain containing 5 (COMMD5), ATP binding cassette subfamily F member 2 (ABCF2), SLC39A1/ZIRTL, and yippee-like 5 gene (YPEL5) being expressed. Exposure of HeLa cells to caffeine can also result in differential expression of 40 cancer-associated gene (for example, KLF6
YPEL5) isoforms , although it is worth noting that different patterns of gene expression result from differing concentrations of caffeine . Nevertheless, even though increased stability of mutant RNA is suggested to be more likely , GINI does not distinguish between the increased stability of the mutant transcript and the selective depletion of the normal transcript . Therefore, if NMD inhibition by caffeine treatment did result in the stabilisation of one transcript and the reduction of another isoform, then unless the probes present on the microarray can distinguish between these isoforms, the net change in gene expression may not have been detected on the array platform. Furthermore, nonsense codons can reduce the abundance of nuclear mRNA without affecting the abundance of pre-mRNA or the half-life of cytoplasmic mRNA  and this might further reduce the sensitivity of GINI.
In order to acquire a more selective list of nonsense transcripts for a particular cell line, it may be necessary to combine the results of multiple different methods of NMD inhibition: 1) siRNA against UPF1[52, 53], 2) caffeine treatment, and 3) emetine treatment, which inhibits the progression of the ribosome along the mRNA [29, 54]. A major problem with using the GINI approach for identifying pathogenic mutations in yet-unidentified high-risk breast cancer genes in the germline DNA of individuals affected with breast cancer is that the mutation is expected to be present in a heterozygous state (at least, in an autosomal dominant disorder). The stabilisation of only one allele reaches the sensitivity limits of gene expression microarrays. Therefore, genes that may have been mutated but are expressed at a moderate to low level may have been excluded from detection. Tumour suppressor genes are usually inactivated during the process of tumorigenesis by a two-step process involving an inactivating mutation in the target gene accompanied by loss of the wildtype allele. However, in LCLs established from peripheral blood mononuclear leukocytes, the normal wildtype allele could mask the effects of GINI on the mutated allele thus reducing the efficiency of GINI . Furthermore, LCLs may not provide an accurate representation of genes that are active in breast tissue, and if the putative breast cancer susceptibility gene is not expressed in LCLs then GINI will not work to identify susceptibility genes.
There is also evidence to suggest that NMD efficiency varies between different people with the same mutation [56, 57], between different tissue types within an organism [58, 59], and even between different strains of the same cell type . It is possible that variable efficiencies of NMD can influence the clinical outcome of hereditary and acquired genetic disease and thus act as a genetic modifier of human genetic diseases.
Inhibition of the nonsense-mediated mRNA decay pathway followed by microarray analysis has been successfully applied to cancer cell lines to identify protein truncating mutations that may underlie sporadic forms of cancer [19, 23, 25–29, 54]. The GINI method has recently been applied to the LCLs from six prostate cancer patients and their healthy brothers in order to identify susceptibility genes in hereditary prostate cancer . However, despite sequencing 17 candidate genes, no truncating mutations were found. The GINI method has also failed to identify putative tumour suppressor genes in gastric cancer cell lines with siRNA against UPF1 . It is commonly reported that the GINI strategy leads to a high number of false positives [19, 21, 23, 26–28, 55, 61, 62]. The novelty of our approach is the ability to identify transcripts stabilised by NMD inhibition in multiple breast cancer patients within a family and compare this gene list to the transcripts that are stabilised in multiple unaffected members of the same family in an attempt to reduce the number of false positive hits. Despite this analysis identifying few candidate genes per family, we did not identify any detectable nonsense mutations. Therefore, our GINI analysis also results in a high number of false positives. However, it is also possible that the mechanism underlying susceptibility to breast cancer in non-BRCA1/2 families may not be due to truncating mutations in susceptibility genes.