The advent of molecularly targeted therapy is shifting the paradigm of management of cancer patients from generalised chemotherapy and/or radiotherapy to personalised treatments with better efficacy and lower toxicity. An example of this is the development of EGFR tyrosine kinase inhibitors in the management of NSCLC patients where the presence of predictive markers such as EGFR and KRAS mutations in their tumours stratifies patients to receive the appropriate treatment.
Currently, direct sequencing is the standard method for EGFR mutation detection, but its limited sensitivity, high cost and long turnaround time have prompted the development of alternative methods for routine clinical testing which have greater diagnostic practicality for somatic mutation detection.
HRM has recently been introduced as a screening method for mutation detection. It is an in-tube method which can be performed in a fast, cheap, and robust manner. It has been applied to the detection of both germline and somatic mutations. For heterozygous germline mutations, HRM has sensitivities approaching 100% [32
]. For somatic mutations in tumours, detection can be compromised by a low proportion of tumour cells in the biopsy. In practice, it has been shown previously that mutant alleles at levels as low as 5 to 10% can be detected by HRM for KRAS
exon 2 mutations [29
HRM has previously been used for detection of EGFR
mutations. In lung cancer, the two most common types of EGFR
mutation, exon 19 deletions and exon 21 p.L858R, were screened by HRM, with a reported 92% sensitivity compared with direct sequencing [29
]. In head and neck cancer, EGFR
exons 18 to 21 were screened by HRM, resulting in the detection of two EGFR
mutations in 24 squamous cell carcinomas [28
We designed this study to determine whether HRM can be a diagnostically useful screening method for EGFR and KRAS mutations in clinical FFPE specimens by validating HRM against sequencing in a large sample cohort containing various types of EGFR and KRAS mutations.
Each of the previously described HRM mutation screening assays for EGFR
have limitations preventing them from being practical for one or more of the following reasons: they do not cover all of exons 18 to 21 [29
], they do not discriminate against common SNPs leading to unnecessary sequencing [28
], they do not detect all mutations that are detectable by sequencing [29
] and they have not been validated against a large panel of clinical samples with previously determined mutations [28
In one assay, a primer is located over the c.2184+19 common SNP potentially resulting in non-amplification of the mutant allele [28
]. Allele dropout due to sequence variation at the primer binding site has been previously demonstrated [41
]. If the allele that has not amplified contains the mutation, the mutation will not be detected.
In this study, HRM assays were designed to maximise the benefits of HRM screening and to minimise SNP interference. EGFR exon 18 to 21 assays covering the entire coding regions were developed. The incorporation of mismatched bases at SNP loci within the primer sequences made it possible to exclude two SNPs, c.2361G>A and c.2184+19G>A. Under the conditions used in our assays, this did not preclude efficient amplification of both alleles, allowing us to detect all 15 mutations from exons 18 and 20.
Due to its high frequency, the exonic SNP, c.2361G>A, would normally necessitate nearly 50% of samples being sequenced for exon 20 because the heterozygous melting pattern given by the SNP can not readily be distinguished from mutation by HRM. Fortunately, its position, in the middle of a large exon, allowed us to divide exon 20 into two fragments using two overlapping amplicons with PCR product sizes of 121 bp and 146 bp. Primers which overlaid the SNP and contained a mismatched residue at the SNP location were used for both amplicons to exclude the SNP from being detected by HRM. A 'G' and an 'A' respectively (mismatched to both alleles) were introduced at c.2361 into the exon 20a reverse and exon 20b forward primers. This strategy led to a short region of 15 bp (c.2353_2367) flanking the SNP for which mutations could not be detected. This is not a serious limitation as no mutations have so far been reported in that region [42
The other common exonic SNP, c.2508C>T in exon 21, is present at a much lower frequency and thus necessitated only a comparatively small increase in the amount of sequencing that would be required. The SNP was detected in nine of the 200 samples.
It is important to consider the amplification information when interpreting HRM results. Instruments allowing the real time monitoring of amplification are advantageous in this regard. We observed that melting curves from samples with insufficient amplification tended to be shifted to the right relative to the wild-type curves in the normalised plots. This implies that the amount of the amplifiable (functional) templates varies in each sample depending on the degree of DNA degradation even though they are all adjusted to the same concentration (2.5 ng/μl). For those samples, we diluted the wild-type control DNA to get similar amplification to the samples with insufficient amplification. We also increased the amplification cycle number to 60 to allow sufficient amplification for melting analysis. As shown in Figure , with sufficient amplification, the right shifting of melting curves was corrected and thus the patient DNA could be reliably compared to the wild-type.
Figure 5 Influence of amplification status on melting analysis. Panel A: The amplification plot of EGFR exon 19 shows insufficient amplification of several samples relative to the wild-type (blue). Panel B: The subsequent melting analysis of the same samples on (more ...)
Accurate identification of mutations is a crucial aspect of all mutation screening methods. This current study demonstrates the accuracy of HRM in the detection of EGFR
mutations in a panel of 200 NSCLC samples. Seventy-five EGFR
mutations and 25 KRAS
mutations were identified by HRM analysis in concordance with sequencing results. The mutation types for each exon were similar to those reported in previous studies, in-frame deletions in exon 19, insertions in exon 20 and missense mutations in both exon 18 and 21 [7
]. However, the overall mutation rate (37.5%) in this Australian study was much higher than in North America and Western Europe (10%) but similar to East Asian populations (30–50%) [43
]. Although most Australians are of European descent, the higher EGFR
mutation rate is likely to be due to selection bias of patients by referring oncologists for features associated with EGFR mutation. The pre-selection criteria included tumours that were histologically adenocarcinomas (70% of samples tested) and patients that were female (60% of patients tested) and/or of Asian ethnicity as many of the patients had a name that was consistent with Asian ancestry. The KRAS
mutation rate (12.5%) which is lower than that reported in previous studies [16
] also supports the notion of pre-selection bias.
Although all positive sequencing results were detected by HRM, some samples were considered positive by HRM but were negative by sequencing (Table ). There are several possible explanations. One possible explanation is that the adverse effects of formalin fixation on DNA can cause PCR artefacts during amplification. At least four chemical reactions occur between formaldehyde and DNA; methylol formation, methylene bridge formation, apurinic and apyrimidinic site formation, and hydrolysis of the phosphodiester bonds [48
]. Compared to the DNA extracted from frozen tissues, a higher frequency of non-reproducible sequence alterations have been reported with DNA isolated from the formalin fixed tissues [49
]. Therefore, the cumulative effects of the PCR artefacts either from Taq polymerase error and errors attributed from chemical reactions of formalin on DNA influence the melting profile of the amplicon depending on the degree of DNA damages.
Seven samples gave positive results in more than three EGFR HRM assays, supporting the hypothesis that the quality of DNA is one of the factors causing aberrant variation of melting in HRM analysis. It has been observed that the wild-type variation in melting analysis is much greater with FFPE DNA than DNA from frozen tissues or peripheral blood (data not shown). The amount of false positives decreased with decreasing amplicon length, with EGFR exon 19 (amplicon size 250 bp) giving the most false positives and KRAS exon 2 (92 bp) giving the least.
Another possibility is that some samples contained levels of mutation below the sensitivity of sequencing detection as a result of a low percentage of tumour in the sample or genetic heterogeneity within the tumour. Where sequencing requires the mutation to be present at a level of 20% of the sample, HRM can detect heterozygous genetic changes down to 10% or below [26
]. We are now adopting digital techniques to confirm that some specimens have true mutations present at low levels. Other HRM false positives can arise from PCR errors due to amplification from very low levels of template. True mutations can be distinguished from artefacts by confirming the identical sequence variations from independent amplification (Do and Dobrovic, manuscript submitted).
HRM is a suitable methodology to test FFPE samples as well as samples with a very low quantity of DNA. It has been reported that ten percent buffered formalin, an aqueous dilution of formaldehyde, can interact with DNA and initiate irreversible DNA degradation resulting in an adverse effect on DNA quality [35
]. In our HRM assays, all the samples were successfully amplified and analysed. Our results show that HRM can be performed with as little as 1 ng template in the EGFR
exon 19 HRM assay. This level of sensitivity of HRM, together with the possibility of sequencing of the HRM product, will extend our ability to screen even clinical samples with extremely low DNA quantity such as samples taken from patients with inoperable tumours. HRM analysis can now provide a genetic testing option for these patients, in which the results might prove useful in directing treatment and may ultimately improve outcomes.
mutations are predominantly mutually exclusive with very rare tumours containing both genes mutated [44
]. The coexistence of mutations in both EGFR
has only been reported in two patients [52
]. In the current study, all 25 KRAS
positive samples were wild-type for EGFR
, supporting the general mutual exclusiveness of the two mutations.