Background
p53 is a tumour suppressor that plays a major role in regulating the cellular response to environmental and genotoxic stress, through cell cycle inhibition and promotion of programmed cell death or senescence [
1‐
4]. There are a variety of stresses that have been shown to activate p53 including DNA damage, cell-cycle aberrations, hypoxia, and aberrant growth signals resulting from expression of oncogenes [
5‐
10]. Moreover, a wide body of work has established the role of p53 in response to DNA damage [
5‐
8].
It has been estimated that the gene encoding p53 (
TP53) is mutated in more than 50% of human cancers [
9]. It appears that inactivation of p53, by mutation or by other means, is highly advantageous and perhaps an absolute requirement for cancer progression (reviewed in Soussi, 2007). The prevalence of p53 mutations is highest in ovarian cancer (48.3%), followed by colorectal cancer (43.6%), oesophageal cancer (42.6%), head and neck cancer (41.5%), and lung cancer (38.4%) [
10].
In ovarian cancer, mutation of the
TP53 gene appears to play an important role in serous carcinogenesis, where
TP53 mutations are present in approximately 50% of high-grade serous carcinomas but rare in serous borderline tumours and low-grade serous carcinoma [
11,
12]. Mutations in
TP53 are also common in ovarian and breast tumours that are
BRCA1-related [
13,
14].
Mutations can occur throughout
TP53. However there are also "hot spots" that account for many mutations [
10,
15]. In the majority of cancers, p53 dysfunction is caused through a direct mutation within the DNA-binding domain of the gene [
9]. Moreover mutations in exons 5 to 8 of
TP53 comprise 94.2% of all somatic mutations in the IARC database, version R11 [
10].
Different
TP53 mutations appear to have different consequences [
16‐
18]. Work on the common p53 mutants, R273H and R175H, using knock-in
in vivo mice models showed that each mutation was responsible for distinct tumour patterns and characteristics, compared to mice with complete loss of one p53 allele [
19,
20]. These studies demonstrate that p53 mutant alleles may have an oncogenic potential beyond the simple loss of p53 function. This is also borne out by clinical studies in human cancer where certain types of
TP53 mutations are associated with a poorer prognosis compared to other types of mutations. For example, studies have shown that
TP53 mutations affecting the zinc binding domains or missense mutations in the DNA binding domain correlate with poorer prognosis in breast cancer [
16,
21].
Overabundance of p53 by immunohistochemistry has been commonly used as a surrogate marker for
TP53 mutation in a wide range of cancers and although the method is cost effective, it does not have the required sensitivity to predict
TP53 mutations [
12,
22]. Dideoxy sequencing remains the gold standard method to detect these mutations. However, it has the disadvantages of high cost in terms of labour and reagents. Sequencing is used most effectively as a confirmatory method after pre-screening with a mutation scanning technique for samples that ideally have had enrichment of the tumour component before DNA extraction.
High resolution melting (HRM) is a mutation scanning technique that monitors the progressive change in fluorescence caused by the release of an intercalating DNA dye from a DNA duplex as it is denatured by marginal increases in temperature [
23]. It is an in-tube method requiring the inclusion of an saturating intercalating dye in the PCR mix and the addition of a high resolution melt step after PCR. The technique has already been employed to scan for somatic mutations in the
KIT,
BRAF,
EGFR, ERBB2, and
KRAS genes [
24‐
27]. In this study, we show that high resolution melting is a highly effective scanning technique for mutations in
TP53.
Discussion
There have been numerous methods employed to detect
TP53 mutations, each with its particular advantages and disadvantages. Scanning methodologies such as SSCP, gradient gel electrophoresis and denaturing high performance liquid chromatography (DHPLC) are advantageous because they significantly reduce the amount of sequencing that ultimately needs to be performed [
31‐
36], streamlining the mutation detection process and making it more cost effective. The high resolution melting methodology presented here has the additional advantages over other scanning methodologies of increased simplicity and rapid turn-around-time, because it is performed directly after PCR amplification in the same tube. We chose exons 5 to 8 because over 94% of reported mutations occur in these regions [
10].
Although the sensitivity is likely to be dependent on the particular mutation, the cell line testing results indicate a high sensitivity for HRM down to 5% mutant sequence. The sensitivity testing results also indicate that the overall variation of biological replicates of wild-type melt profiles will affect the ability to discriminate mutant samples from wild-type, which will ultimately affect the sensitivity of the technique.
Samples with pure mutant DNA could pose a problem for detection with HRM. For example, in cell lines containing only a mutant allele due to LOH of the remaining WT allele, in some cases (e.g. T47D) the resultant PCR product has minimal temperature difference compared to that amplified from wild-type DNA. However, this is a theoretical rather than a practical consideration as a proportion of wild-type DNA will be present in a tumour biopsy even with microdissection of the tissue specimen. Indeed, the addition of a small amount of wild-type DNA (5%) to T47D caused a sufficient heteroduplex effect to facilitate detection by HRM [see Additional file
1]. Addition of 5–10% wild type DNA should be considered for samples are known to be extremely pure, such as those that have been purified by laser-capture microdissection.
HRM analysis requires careful attention to the design of the amplicons as mutation detection is easier when there is a single melting domain. This is exemplified by exon 5 in which the 394A>C mutation in the 5' end of the exon was not initially observed under the conditions used. This was even the case when the cell line was diluted with normal DNA to create a heterozygous mutation (data not shown). We found that separating the amplicon into 2 domains allowed the detection of this mutation and all other mutations tested. We furthermore manipulated the primers, which encompassed a GC region between the 2 domains by introducing sequence changes into the primers, which promoted greater homogeneity in the melting behavior of the amplicon.
The ovarian tumour DNA panel was used to assess the sensitivity and positive predictive value of HRM for
TP53 exons 5 to 8. The samples chosen were preselected for mutations in these exons but otherwise we performed a blinded analysis. There were a number of samples that had large shifts from the melt profiles of wild-type samples whereas others had changes that were more subtle. The aberrant curve produced for sample 22 in exon 6 appeared closer to a wild-type sample than a mutant because the profile existed within the spread for samples resembling wild type. However, the sample was classified as having an aberrant melt profile because of its conspicuously different shape, the earlier melting that is characteristic of heteroduplex formation, and because replicates gave the same pattern (Figure
6, panel B). The sequencing data showed that sample 22 had an A to G change at base 578. In this sample set, the HRM methodology had 100% sensitivity and 91% positive predictive value.
The screening of 20 breast tumour DNA samples with an unknown
TP53 mutation status followed by sequencing of all the samples allowed us to further assess the sensitivity and positive predictive value of the methodology. Seven samples had aberrant melt profiles indicating the presence of a mutation and the sequencing results confirmed the presence of the mutations (see Table
5). In addition, there were no mutations detected by sequencing in samples scored as wild-type by HRM. This gave 100% sensitivity and 100% positive predictive value for HRM for the breast tumour sample set. Larger studies are needed to further assess the sensitivity and positive predictive value of the current methodology.
In the breast tumour sample set, sample B14 was interesting as it was scored as a query by HRM and subsequent sequencing showed the presence of a 531_533delC mutation that was present at a low percentage (Figure
8). Data from other HRM studies (our unpublished results) indicate that deletions/insertion of a single base do not have the same magnitude of change in melt profile compared to that of a heterozygous SNP. Another sample, B4 was interesting as the HRM data indicated a clear mutation but the sequencing data showed the mutation at the lower limit of detection. These cases reflected the lower proportion of tumour in some of the samples as there was no enrichment of the tumour component before DNA extraction.
One of the major advantages of HRM is for the detection of somatic mutations in genes that can have alterations at different positions in the coding sequence. Because in addition to pathogenic mutations, HRM will detect SNPs, silent mutations and intronic variation of unknown significance, and may have an intrinsic false positive rate, any aberrant melt curves should be validated by sequencing.
Acknowledgements
We would like to thank Katie Huang for assisting with blinded analysis of the HRM data and Angela Tan for careful reading of the manuscript. We would also like thank Stephen Lade and Maurice Loughrey for pathological review of frozen ovarian tissue sections and Adrian Harris, University of Oxford, for the breast cancer samples. AOCS Biospecimen Core was supported by the U.S. Army Medical Research and Materiel Command under DAMD17-01-1-0729, The Cancer Council Tasmania and The Cancer Foundation of Western Australia and the National Health and Medical Research Council of Australia (NHMRC). WGOTB was also supported by NHMRC. This work was supported by grants to AD from Novartis, the National Health Medical Research Council of Australia (NHMRC), the CLL Global Research Foundation and to SF from the Victorian Breast Cancer Research Consortium. AAA holds a Cancer Research UK Clinician Scientist Fellowship. We would also like to thank John Zalcberg for his continued support.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MK participated in primer design, carried out the molecular genetic studies and data analysis, and wrote the manuscript. AAA, SJH and JDB were responsible for conducting sequencing analysis of the ovarian cancer samples for TP53 mutations. DE was responsible for coordinating the samples, micro-dissection of the tumour material and preparation of the DNA. AdF provided ovarian tumour samples from the Westmead Gynaecological Oncology Tissue Bank. AOCS and the Westmead Gynaecological Oncology Tissue Bank were responsible for collection, processing and storage of their respective samples. SJB provided the breast cancer DNA and contributed to the manuscript. DDB suggested the use of the sequenced ovarian cancer resource, was involved in the study design and contributed to the manuscript. AD originated the study, designed the primers, reviewed the data analysis and co-wrote the manuscript. All authors read and approved the final manuscript.