High-resolution melting (HRM) re-analysis of a polyposis patients cohort reveals previously undetected heterozygous and mosaic APC gene mutations

Familial adenomatous polyposis is most frequently caused by pathogenic variants in either the APC gene or the MUTYH gene. The detection rate of pathogenic variants depends on the severity of the phenotype and sensitivity of the screening method, including sensitivity for mosaic variants. For 171 patients with multiple colorectal polyps without previously detectable pathogenic variant, APC was reanalyzed in leukocyte DNA by one uniform technique: high-resolution melting (HRM) analysis. Serial dilution of heterozygous DNA resulted in a lowest detectable allelic fraction of 6 % for the majority of variants. HRM analysis and subsequent sequencing detected pathogenic fully heterozygous APC variants in 10 (6 %) of the patients and pathogenic mosaic variants in 2 (1 %). All these variants were previously missed by various conventional scanning methods. In parallel, HRM APC scanning was applied to DNA isolated from polyp tissue of two additional patients with apparently sporadic polyposis and without detectable pathogenic APC variant in leukocyte DNA. In both patients a pathogenic mosaic APC variant was present in multiple polyps. The detection of pathogenic APC variants in 7 % of the patients, including mosaics, illustrates the usefulness of a complete APC gene reanalysis of previously tested patients, by a supplementary scanning method. HRM is a sensitive and fast pre-screening method for reliable detection of heterozygous and mosaic variants, which can be applied to leukocyte and polyp derived DNA. Electronic supplementary material The online version of this article (doi:10.1007/s10689-015-9780-5) contains supplementary material, which is available to authorized users.


Introduction
Familial adenomatous polyposis (FAP) is a hereditary tumor syndrome predisposing to early-onset colorectal cancer (CRC), accounting for 1 % of CRC cases. Classic FAP patients develop well over a hundred of colorectal Electronic supplementary material The online version of this article (doi:10.1007/s10689-015-9780-5) contains supplementary material, which is available to authorized users. polyps starting from early adolescence, leading to CRC approximately a decade later. Attenuated FAP (AFAP) patients develop 10-100 polyps at an older age at onset in their forties [1,2]. Most FAP and AFAP patients carry dominantly inherited pathogenic germline variants in the APC gene (MIM: 175100) or recessive germline variants in the MUTYH gene (MIM: 604933). Pathogenic APC or MUTYH variants can also cause extracolonic features, like duodenal and gastric adenomas and cancer, desmoid tumors, osteomas and benign skin tumors (Gardner syndrome).
The detection rate of pathogenic APC variants, including large structural variation, is approximately 70-80 % in classic FAP patients and 10-30 % in AFAP patients [3][4][5][6]. The MUTYH associated polyposis (MAP) phenotype generally resembles AFAP. Bi-allelic pathogenic MUTYH variants are found in a quarter of polyposis patients negative for pathogenic APC variants [7]. In 20-30 % of adenomatous polyposis patients, no pathogenic variant in APC or MUTYH is identifiable, especially in patients with low polyp counts [7,8]. A few polyposis patients have been linked to other genes, such as SMAD4, BMPR1A, POLE and POLD1 [8,9]. Pathogenic variants in unexplored genes may underlie the remaining genetically unexplained polyposis patients. However, two considerations warrant further examination of the APC gene, with its paramount role in the mechanism of polyposis and somatic defects in almost all adenomatous polyps.
First, in the past 20 years, a variety of methods have been applied to identify pathogenic germline APC variants, including direct sequencing and indirect methods like the protein truncation test (PTT), denaturing high performance liquid chromatography (dHPLC) and denaturing gradient gel electrophoresis (DGGE), southern blot and multiplex ligation-dependent probe amplification (MLPA). The detection rate of pathogenic variants is partly determined by the sensitivity of these methods and to which extend the length of the APC gene was tested [10,11]. Second, mosaic pathogenic APC variants are particularly difficult to detect and probably an underestimated cause of polyposis coli. Of all detected constitutive pathogenic APC variants, 15-25 % occur de novo and 1-4 % are mosaics [12,13].
In this study a reanalysis of the entire coding region of APC was performed in a group of 171 patients with C10 colorectal polyps, without previously detectable pathogenic APC (or MUTYH) variants. APC scanning had been performed before by various methods with different levels of sensitivity and completeness [12]. To identify possible previously missed heterozygous and mosaic variants, we used one uniform technique: high-resolution melting (HRM) analysis, which is known to be able to reliably detect heterozygous variants and lower allelic fractions [14,15]. Also, scanning in polyp DNA of two patients was performed, as a proof of principle, to identify possible mosaic variants.

Patient samples
The study group consisted of 171 index patients, referred for APC and/or MUTYH gene testing at the Laboratory for Diagnostic Genome Analysis (LDGA) in Leiden, The Netherlands, between 1995 and 2007, without detectable pathogenic variants by analysis as earlier described [12]. Of the 244 consecutively referred patients 171 were included, diagnosed with C10 colorectal polyps of adenomatous or predominantly adenomatous histology. The excluded 73 patients had insufficient DNA quality (3), insufficient clinical data (13) or \10 polyps (57). Informed consent was obtained for DNA testing according to protocols approved by local ethics review board. Clinical data was obtained from patient records at the LDGA (Table 1). DNA was extracted from blood leukocytes according to standard protocols. Tissue DNA was analyzed if applicable and available for part of the patients. For HRM validation in total 117 DNA samples with unique heterozygous variants and for each amplicon 10 wild-type samples available at the LDGA were used. Additionally homozygous, heterozygous and wild-type controls of eight common SNPs were tested.
The 10 ll PCR mix contained 20 ng of template DNA, 19 Lightscanner Mastermix (including LC Green Plus dye, Idaho Technology), 2.5 pmol of each primer and distilled water. In our hands results with Lightscanner Mastermix were better compared to separately mixed ingredients (data not shown). Template DNA concentrations were measured by a Nanodrop 1000 Spectrophotometer (Thermo Fisher Scientific, Wilmington, DE). Asymmetric PCRs with unlabeled probes were performed with a 1:5 ratio, with 1 pmol of forward primer, 5 pmol of reverse primer and 5 pmol of the unlabeled probe. Optimal primer annealing temperatures were established by temperature gradient PCR and HRM. Amplicons were redesigned if they showed more than two melting domains [17] (online resource Supp. Fig. S1).

HRM data analysis
Melting of PCR products was performed on a 96 wells Lightscanner instrument, with Lightscanner software version 2.0 (Idaho Technology). Temperature ranges of 70-98 and 55-98°C were used for amplicon scanning and unlabeled probe genotyping, respectively, with increments of 0.1°C/s. Melting curve data analysis was performed as described [16,17]. The software's sensitivity level was set to C3.0. Raw melting curves were normalized at 100 and 0 % fluorescence intensity. They were also temperature shifted, at 5 or 95 % fluorescence intensity, to correct for slight temperature differences across the plate. Melting analysis was performed twice per PCR plate to rule out melting artifacts.
Variant scanning and sequencing PCR and HRM were performed in 96 well plates, including positive controls per amplicon in each plate, with allelic fractions of 50, 13 and 6 %. PCR and HRM were repeated if the first experiment showed an aberrant melting curve. Sequencing was performed if the repeated experiment showed an aberrant melting curve, except for those that were induced by a SNP under a specific probe. The overall portion of succeeded tests, including HRM and sequencing per amplicon, for the 171 patients 9 61 amplicons was 99 %. Direct sequencing was performed as described, directly after HRM, on the same reaction mix and analyzed by Seqscape software version 2.5 [12,17]. Sequences were carefully observed for detection of low peaks. Pyrosequencing was performed as described, to confirm low-level mosaic variants [12]. Detected variants were searched in literature and the UMD and LOVD public databases (www. lovd.nl/apc, www.umd.be/apc, [19][20][21][22]). New variants were analyzed in silico for possible pathogenicity by Alamut version 2.3 (Interactive Biosoftware, Rouen, France).

Pathogenic heterozygous variants
In the 171 scanned patients, HRM analysis and subsequent sequencing detected eight different heterozygous pathogenic variants occurring in ten patients (6 %). Four of the HRM re-analysis of a polyposis patients cohort 249 pathogenic variants were novel and four were previously reported in the literature. Three novel pathogenic variants were frameshifting variants and one was a duplication at the splice donor site of intron 14, (c.1958 ? 1_1958 ? 2dup, patient 7). In silico analysis by Alamut predicted an in frame skip of exon 14, which was confirmed by reverse transcriptase PCR (data not shown). In nine out of the ten patients with a pathogenic variant, it had been missed by previous diagnostic testing methods, respectively PTT at the 5 0 part of exon 15, DGGE in exons 9, 14 and 15 or single-strand conformational polymorphism (SSCP) in exons 9 and 11. One pathogenic variant was located in a region of the gene not tested in that particular patient before ( Table 2, online resource Supp. Table S2).

VUS and polymorphisms
Sixteen different rare variants of unknown significance (VUS) were found in 17 of the 171 scanned patients, of which five were novel, including one silent variant, two missense variants, one in frame deletion and one in frame duplication. No effects on splicing or other in silico clues for pathogenicity were predicted by Alamut. All of these VUS were found in patients without a pathogenic variant, except one (online resource Supp. Table S2). The in-frame deletion c.5501_5506del, p.Val1834_Arg1835del was found in a patient with 100 colorectal polyps (patient 15). Segregation of the variant with the disease in this family suggests that causal relevance of this VUS as a high-  Table 2). The in frame duplication c.6363_6365dup, p.Ala2122dup was found in a patient with [13 polyps (patient 16), for whom no further family members were available for segregation analysis (  Table S2) [23].

Heterozygous variants
Amplicons were tested for sensitivity of the detection of heterozygous variants using the available positive control samples. Of the 117 tested samples with unique heterozygous variants in 50 of the 61 amplicons, 116 were detectable (online resource Supp. Fig. S2). The variant c.423-17dup, a duplication of a T in a T 7 A 13 -stretch in intron 3, was not detectable by HRM. This region has been described as problematic before [24]. Ten wild-type samples were tested per amplicon in duplicate, among which no false positives were seen.

Low allelic fractions
The 116 samples with detectable heterozygous variants were serially diluted to allelic fractions of 25, 13 and 6 % in a wild-type background. All 116 variants were detectable down to at least 13 % and 105 (90 %) down to at least 6 %. For 72 variants the dilutions were continued down to 3 and 2 % and 28 and 6 variants were detectable, respectively. The eleven variants not detectable below 13 % were eight 1 Table S3). Of six variants tested in multiple overlapping amplicons, three showed different detection limits (6 vs. 25 % for one, and 3 vs. 6 % for two variants). Four variants were better detectable with temperature shift set at level 95 % instead of the default 5 % (online resource Supp. Fig. S2 and S3).  Table S2 and Fig. S4).

Multiple variants per amplicon
The presence of one heterozygous variant in an amplicon causes heteroduplexes during HRM, altering the melting curve. The effect of an additional second variant in the same amplicon on the melting curve may be less well visible [25].  Fig. S4 and S5).

Discussion
Re-analysis of APC in leukocyte DNA by HRM in eligible patients previously tested negative for pathogenic variants by other methods, yielded pathogenic variants in 12 out of 171 (7 %) patients, of whom two appeared to be present as mosaicism. Additionally, in each of two patients with apparently sporadic polyposis, without a variant in leukocyte DNA, a mosaic pathogenic variant was detected in polyp tissue DNA only. The detection of these pathogenic variants facilitates genetic counseling and family testing. The previously used pre-screening techniques SSCP, DGGE and PTT have detected most APC variants in the past, including mosaics. However, these methods have been shown to be incapable of detecting all of the variants present, explaining our results [12,13,[25][26][27][28][29].
Our minimal detectable allelic fraction by HRM was 2-6 % for 91 % of variants, and 13 % for those remaining. HRM has been shown to detect allelic fractions of between 1 and 13 %, significantly better than the 10-25 % reported for Sanger sequencing [14,26,[30][31][32][33][34]. Sensitivity has been reported to be optimal for small amplicons of around 100 bp. Our amplicon size was 120-400 bp, in order to span the large APC gene with a limited amplicon number [14,35,36]. Currently, many diagnostic laboratories are using Sanger sequencing without pre-screening for diagnostic gene testing, since it has become cheaper and easier to perform, also for APC [1,20,37]. Sanger sequencing would have detected the 10 heterozygous variants. However, the mosaics in patients 11 and 12 (5 and 15 % allelic fraction) were missed by standard sequencing analysis. They were only visible for us after carefully scrutinizing the sequence trace, with the current knowledge of an aberrant melting curve suggestive for a mosaic variant. The two mosaics with the lowest allelic fraction (5-6 %) from Hes et al. [12] were also not detectable by Sanger sequencing, but only by DGGE. Because of its superior sensitivity to detect mosaic variants, HRM is a method to consider using for scanning genes with high occurrence of mosaic variants, like APC. However, next generation sequencing (NGS) methods may ultimately be the method of choice for mosaic detection in laboratories for which this is feasible. The advantage of NGS is a limit of detection of B1 % and immediate identification of the variant [38][39][40]. Possible disadvantages of NGS in comparison to HRM are the complexity of the method, the expertise needed and purchase costs of the apparatus. HRM is relatively fast, inexpensive and technically less complicated compared to NGS. HRM is also more flexible, with separate analyses per amplicon, while with NGS large amounts of amplicons are pooled together. With HRM it is easier to repeat failed experiments. For a low limit of detection for mosaicism, NGS needs sufficient sequencing depth, making it more expensive. However, the costs for NGS are reducing, and it will probably be more cost-effective (soon and for many laboratories/countries) to optimize an NGS approach instead of HRM.
HRM is a sensitive, inexpensive and convenient diagnostic method, when its limitations are taken into account [25,41]. It can be readily applied in almost every standard diagnostic laboratory. Known limitations of HRM, like limited sensitivity for particular variant types, such as homozygous and base-pair neutral variants, variants located in nucleotide stretches and multiple variants per amplicon, were also shown in our study [17,25]. Variants in known nucleotide stretches, like the T 7 A 13 -stretch in intron 3, can be detected by using an unlabeled probe covering the stretch (data not shown). No adaptations were made for homozygous variant detection, as they are considered as embryonically lethal in APC [42].
Sensitivity of HRM, and also sequencing, for mosaic variants can be improved by Cold-PCR, down to allelic fractions of 0.1-1 % [43]. Cold-PCR might be challenging to optimize for large scale gene scanning [44]. Cold-PCR applications have mostly been described for pathogenic variant hotspot regions and one for multiplex PCR of the coding region of TP53 [43,45]. APC pathogenic hotspot variants and common C[T transitions might be good candidate locations to start with [12]. Other methods suitable for mosaic detection are ultra-sensitive allele specific methods and deep next generation sequencing [12,39,[46][47][48][49][50].
Clinical implications of the detection of APC mosaicism have been described before [12,13,51]. Depending on the timing of occurrence and distribution during embryonic development, a mosaic variant can be present in blood, affected tissue and/or germ cells. The distribution and level of the mosaic variant can indicate its heritability, phenotype and detectability. If a mosaic variant is present in the germ cells, offspring has up to 50 % risk for inheriting the disease. The severity of the phenotype of a heterozygous germline pathogenic APC variant is dependent on its position in the gene. Part of described mosaic patients have milder phenotypes compared to heterozygous carriers of similar variants [12,13]. Our four detected mosaic variants were located at positions related to a severe phenotype, if heterozygous. Patients 11 and 12, with a mosaic variant in both blood and polyps, had a relatively late age of onset, but both had extracolonic features. Patients 13 and 14 had a severe phenotype. However the polyps of patient 13 were limited to the rectosigmoid. None of our four mosaic variants were CGA to TGA transitions, which were previously described to be a significant portion of mosaic variants [12,13].
Blood leukocyte DNA is most commonly used for testing for pathogenic germ line variants [12,51,52]. To demonstrate the value of testing affected tumor tissue for the detection mosaics, and the possibilities of HRM, two cases of whom we had polyp tissue available were analyzed as 'proof of principle' (patients 13 and 14). Both patients showed a pathogenic APC variant recurrent in multiple polyps, without detectable pathogenic variant in blood. Both patients had a clear polyposis phenotype and early age of onset, without detectable pathogenic germ line variants in the APC gene or MUTYH gene. Patient 13 had polyps limited to the rectosigmoid, suggestive of mosaicism. Patient 14 had a typical FAP phenotype, for which the chance of detection of pathogenic APC or MUTYH germ line variants is expected to be very high. Finding the same pathogenic variant recurrent in multiple polyps in each of the two patients (100 % detection ratio of somatic mosaicism), should in this case be seen as a coincidental finding. One comparable mosaic patient was earlier described, with a pathogenic variant present in five analyzed adenomas and not detectable in blood [12]. It is mandatory to build larger series of patients without pathogenic germ line variants in the polyposis genes and collect their polyp material for somatic APC variant analysis [12,13,51]. For distinguishing between mosaicism in a substantial part of the colon and an isolated somatic variant limited to one tumor, analysis of multiple polyps and/or surrounding normal tissue is necessary.
A significant group of patients with multiple colorectal polyps remains genetically unexplained after extensive testing for pathogenic APC and MUTYH variants [8,20]. Yet undetected APC (mosaic) variants will probably explain a small minority of cases. Other screening approaches, like next generation sequencing (NGS), for searching genetic, epigenetic or multifactorial aberrations inside or outside APC need to be explored [53]. Recently, pathogenic variants in the POLE and POLD1 gene were found to explain a small portion of polyposis cases [9].
In conclusion, rescreening of APC by a uniform sensitive detection method like the described HRM method detects heterozygous and mosaic variants previously missed by different conventional methods in group of genetically unexplained patients with multiple colorectal polyps. In addition, scanning APC in DNA isolated from multiple independent polyps of two patients successfully detected mosaic pathogenic APC variants present in affected tissue, but undetectable in blood. By using HRM and by screening of tumor DNA, we have genetically explained a further small portion of unexplained polyposis patients from our cohort.