METHODS

Characterizing Mutations in Samples WithLow-Level Mosaicism by Collection andAnalysis of DHPLC Fractionated Heteroduplexes

Paul Emmerson,1 Julie Maynard,1 Sian Jones,1 Rachel Butler,2 Julian R. Sampson,1

and Jeremy P. Cheadle1*1Institute of Medical Genetics, University of Wales College of Medicine, Heath Park, Cardiff, UK; 2Medical Genetics Service for Wales,University Hospital of Wales, Heath Park, Cardiff, UK

Communicated by Michael Dean

Somatic mosaicism is a frequent phenomenon in mendelian disorders that exhibit a high proportion ofnew mutations; however, mutant alleles present at low frequency are difficult to detect and characterize.We have previously shown that denaturing high-performance liquid chromatography (DHPLC) candetect TSC1 and TSC2 mutations in tuberous sclerosis patients with low-level somatic mosaicism, evenwhen direct sequencing cannot identify the causative lesion. Characterization of these mutationstraditionally involves extensive sequencing of cloned products. To overcome this limitation, we haveutilized DHPLC with an in-line fraction collector to isolate low-level heteroduplex peaks that can bedirectly sequenced to reveal the mutation. We have successfully applied this technique to resolve themutations 2724-1G4C in TSC1 and 1462-28del42bp, 1774del4bp, and N1643K (4947C4G) in TSC2,which were present in only 6.5–17% of the patients’ alleles. We have also applied this technique tosuccessfully resolve seven somatic APC mutations in colorectal tumor samples that were previouslyundetectable by direct PCR product sequencing. This method may simplify many of the currentlychallenging goals in mutation detection. Hum Mutat 21:112–115, 2003. r 2003 Wiley-Liss, Inc.

KEYWORDS: mosaicism; DHPLC; fraction collection; mutation detection; APC; TSC1; TSC2; MYH; MUTYH

DATABASES:

TSC1 – MIM 605284, 191100 (TSC); GenBank: AF013168TSC2 – MIM 191092; GenBank: X75621APC – MIM# 175100; GenBank: M74088, M73548MUTYH – OMIM# 604933; GenBank: U63329

INTRODUCTION

Tuberous sclerosis complex (TSC; MIM# 191100)is an autosomal dominant disorder characterized bythe development of hamartomatous growths in avariety of tissues and organs [Gomez et al., 1999].Involvement of the brain is associated with intellec-tual handicap, epilepsy, and abnormal behavioralphenotypes. Other organs commonly involved includethe skin, kidneys, and heart, where the associatedfeatures include facial angiofibromas, subungualfibromas, forehead plaques, shagreen patches, renalangiomyolipomas and cysts, and cardiac rhabdomyo-mas. About 60% of TSC cases are sporadic,representing new mutations, and TSC occurs in atleast 1 in 10,000 live births, without apparent ethnicclustering [Osborne et al., 1991]. Patients with TSCharbor a wide variety of germline mutations in eitherthe TSC1 gene (MIM# 605284) on chromosome

9q34 or the TSC2 gene (MIM# 191092) on chromo-some 16p13.3 [reviewed in Cheadle et al., 2000].Somatic mosaicism is a frequent phenomenon in

disorders exhibiting a high mutation rate and at least10% of sporadic TSC cases are thought to be somaticmosaics for TSC1 or TSC2 mutations [Verhoef et al.,

Received 8 July 2002; accepted revised manuscript 1 October2002.

nCorrespondence to: Dr. Jeremy P. Cheadle, Institute of MedicalGenetics, University of Wales College of Medicine, Heath Park,Cardi¡, CF14 4XN,UK. E-mail: cheadlejp@cardi¡.ac.uk

Grant sponsors:Tuberous Sclerosis Alliance;Tuberous SclerosisAssociation, UK; ELWA, Knowledge Exploitation Fund; W.D.A.,Centres of Expertise in Technology and Industrial Collaboration(CETIC) award.

DOI10.1002/humu.10159Published online in Wiley InterScience (www.interscience.wiley.com).

rr2003 WILEY-LISS, INC.

HUMANMUTATION 21:112^115 (2003)

1999]. This may result in the failure of moleculargenetic diagnosis, due to an inability to detect mutantalleles present at low frequency [Kwiatkowska et al.,1999]. We have recently formatted denaturing highperformance liquid chromatography (DHPLC) [re-viewed in Xiao and Oefner, 2001] for mutationscreening of the TSC1 and TSC2 genes [Jones et al.,2000]. In four TSC cases with mosaicism, we observedaberrant DHPLC elution profiles that were notassociated with any detectable sequence variation ondirect manual or automated sequence analysis of thePCR products [Jones et al., 2001; Antonarakis et al.,2002]. After extensive sequencing of cloned products,the mutations were characterized in 6.5–17% of thepatients’ alleles from peripheral blood leukocyte DNA.Cloning and sequencing of numerous cloned productsis both time-consuming and expensive. Here, wedescribe a method for characterizing mutations inpatients with mosaicism by isolating and directlysequencing low-level mutant heteroduplex peaks thatare separated by DHPLC from the high-level,predominantly wild-type, homoduplex peaks. Thismethod has been successfully applied to resolve themutations in the four previously reported patientswith mosaicism. To further demonstrate the utility ofthis technique, we have also successfully resolved aseries of somatic APC mutations in colorectal tumorsamples that could not be characterized by sequencingunfractionated PCR products.

MATERIALSANDMETHODSPatient Samples

DNA was extracted from peripheral blood leukocytes (PBL)by standard methods. Patient 169f carried the TSC1 exon 20splice mutation 2724-1G4C in 6.5% of alleles from PBL DNA,Patient 395 carried the TSC2 exon 14 mutation 1462-28del42bp in 17% of alleles from PBL DNA, Patient 22715harbored the TSC2 exon 16 mutation 1774del4bp in 7.5% ofalleles from PBL DNA, and Patient 367 carried the TSC2 exon37 missense mutation N1643K (4947C4G) in 15% of allelesfrom PBL DNA [Jones et al., 2001; Antonarakis et al., 2002].

Tumor Samples

DNA was extracted from colorectal adenoma and carcinomatissue that had been micro-dissected from paraffin blocks. Thenature of all tissues was verified histologically.

PCR Ampli¢cation

Relevant exons were amplified using AmpliTaq Gold DNApolymerase (Applied Biosystems (ABI), Warrington, UK), aspreviously described [Jones et al., 2000; Jones et al., 2002].Amplification cycles consisted of an initial denaturation at941C for 10 min, followed by 32 cycles of 941C for 1 min, 53–571C for 1min, and 721C for 1min, and a final extension step of721C for 7 min. To enhance the formation of heteroduplexesfor DHPLC analysis, samples were denatured at 941C for 30 secthen slowly cooled to 401C at a rate of 11C/min. Details ofprimer sequences and annealing temperatures are availableat www.uwcm.ac.uk/study/medicine/medical_genetics/research/tmg/projects/On-line_resources.htm.

DHPLC Analysis and Fraction Collection

DHPLC analysis was carried out using the WAVE 3500HTDNA fragment analysis system (Transgenomic; Crewe, UK),run in rapid mode with a gradient time of 2.5 min and a totalrun time of 3 min. The start ratio of buffers A (0.1Mtriethylammonium acetate [TEAA], 0.025% acetonitrile[ACN] [v/v], pH7.0), and B (0.1M TEAA, 25% ACN [v/v],pH7.0) was calculated using Wavemaker 4.1 software (Trans-genomic; Crewe, UK), and the stop ratio was set as the buffer Bstart value +12%. The optimum temperature for separation oflow-level heteroduplex and high-level homoduplex peaks wasdetermined empirically by performing a temperature titration atthe Tm721C at 0.51C intervals. Fraction collection was carriedout using a Transgenomic FCW-200 in-line fragment collectoron the whole window setting [Kuklin et al., 1999] andboundaries were initially determined using 5ml injections ofsamples. Fraction collection was performed using 100mlinjections of samples and products were eluted in 200ml ofbuffer D (8% ACN).

DNA Sequencing

Collected products were purified using the QIAquick PCRpurification kit (Qiagen; Crawley, UK), according to themanufacturer’s specifications, and eluted in 30ml of milli-Qwater. A total of 10ml of purified products were sequenced usingboth forward and reverse primers with the ABI Prism BigDyeterminator cycle sequencing kit, and analyzed on an ABI Prism3100 DNA sequencer.

RESULTS

DHPLC analysis of exon 20 of TSC1 for Patient169f, exon 14 of TSC2 for Patient 395, exon 16 ofTSC2 for Patient 22715, and exon 37 of TSC2 forPatient 367 was carried out at the empiricallydetermined optimum temperature for separation ofheteroduplex and homoduplex peaks (59.31C, 62.41C,63.41C, and 62.41C, respectively; data not shown).DHPLC analysis of exon 20 of TSC1 from Patient169f showed three distinct peaks (Fig. 1A). Fragmentcollection and sequencing clearly resolved the splicemutation 2724-1G4C only in the reverse direction ofcollection 1 and the forward direction of collection 2.Normal sequence was observed in collection 3(Fig. 1A). Sequencing of the three fragments collectedfrom DHPLC analysis of exon 14 of TSC2 fromPatient 395 revealed the 1462-28del42bp deletion inboth the forward and reverse directions of collections1 and 2, but revealed only wild-type sequence incollection 3 (Fig. 1B). Sequencing of the twofragments collected from DHPLC analysis of exon16 of TSC2 from Patient 22715 revealed the mutation1774del4bp only in the forward direction of collection1 (Fig. 1C). DHPLC analysis of exon 37 of TSC2from Patient 367 showed a complex trace with sixpeaks; this is due to the presence of both the mosaicmutation and a constitutional polymorphism(C4977T) in the same amplimere [Antonarakiset al., 2002]. The mosaic mutation N1643K wasclearly resolved in both forward and reverse directionsof collections 1–3, but was not present in collections

CHARACTERIZINGMOSAIC MUTATIONS 113

4 or 5 (Fig. 1D); the polymorphism was resolved in allfive collections (data not shown).

DHPLC analysis of a 2.8 kb region of the APC genein 108 colorectal tumors from seven patients withbiallelic germline MYH mutations (MUTYH; MIM#604933) was carried out as previously described [Joneset al, 2002]. Seven out of 50 (14%) somatic APCmutations could not be characterized by directsequencing of unfractionated PCR products and wereonly resolved after the collection and analysis of low-level mutant heteroduplex peaks (Fig. 2A–C).

DISCUSSION

By collecting and sequencing DHPLC separatedheteroduplex peaks, we have successfully resolved

mutations that were undetectable by standard PCRproduct sequencing. This technique works by isolatingand analyzing fractions with significantly increasedratios of mutant to wild-type alleles: whereas mutantalleles may be present at very low levels inunfractionated PCR products, they will be present atsubstantially higher levels (B50%) in the hetero-duplex fractions. Interestingly, sequencing of theheteroduplex fractions from Patients 169f and 22715clearly resolved the causative mutations in only theforward or reverse sequencing reactions. This is likelyto be due to different heteroduplex species in differentfractions (e.g., mutant-sense/wild-type antisensestrands in one fraction, and mutant antisense/wild-type sense strands in the other fraction). In allpatients, only wild-type sequence was obtained fromthe major homoduplex peak, as expected.The main advantages of our methodology over

cloning are the considerable savings in time, money,and effort. Cloning can be technically challenging andif mutant alleles are present in only 5% of the sample,59 clones would have to be sequenced to have

FIGURE 1. Fraction collection and sequencing of DHPLC se-parated low-levelmutant heteroduplexpeaks to resolvegerm-line mutations in patients with somatic mosaicism. DHPLCelution pro¢les are shown on the left with the positions ofthe collection boundaries indicated by vertical hashed lines.Collections are numbered with the corresponding sequencesshown on the right. A: 2724-1G4C in exon 20 of TSC1 inPatient 169f was clearly resolved in the reverse direction ofcollection 1 and the forward direction of collection 2. B:1462-28del42bp in exon 14 of TSC2 in Patient 395 was re-solved in both the forward and reverse (illustrated) directionsof collections 1 and 2. C:1774del4bp in exon 16 of TSC2 inPatient 22715 was resolved in the forward direction of collec-tion1. D: N1643K in exon 37 ofTSC2 in Patient 367 was re-solved in the forward (illustrated) and reverse directions ofcollections1^3.

FIGURE 2. Fraction collection and sequencing of DHPLC se-parated low-level mutant heteroduplex peaks to resolvesomatic mutations in colorectal tumors. DHPLC elution pro-¢les are shown on the left, sequences of unfractionated PCRproducts are shown in the center, and sequences of DHPLCfractionated heteroduplexes (boundaries indicated by verti-cal hashed lines on DHPLC traces) are shown on the right.Seven out of 50 somatic APC mutations were only resolvedafter thecollectionandanalysis of low-levelmutant heterodu-plex peaks, e.g., (A) S836X (2507C4A) in tumor sampleMA22_10B, (B) E1374X (4120G4T) in sample MA20_11A,and (C) S1356X (4067C4A) in sample MA22_13B. All se-quences are shown in the forward direction except in A.

114 EMMERSON ETAL.

a Z95% chance of identifying a single pathogeniclesion. In practice, two mutant clones should beidentified to rule out sporadic PCR-induced errors, inwhich case 93 clones would need to be sequenced.The major technical issue encountered in ourmethodology was ensuring that appropriate amountsof PCR product were loaded onto the column forefficient fragment collection, as insufficient quantitiesof DNA can lead to a failure of subsequent sequencingreactions. In general, we have found that 480ng of‘‘heteroduplex peak DNA’’ generates quality se-quence, although as little as 20ng has been analyzedsuccessfully.

In conclusion, we describe a straightforwardmethod for characterizing germline mutations inpatients with low-level mosaicism and somatic muta-tions present at low levels in tumor samples due to thepresence of contaminating normal tissue. This meth-odology will also have applications for the character-ization of single nucleotide polymorphisms in pooledDNA samples, and could therefore substantiallyfacilitate many of the technically challenging goalsin human genetic research.

ACKNOWLEDGMENTS

We thank Mr. P. Davies for technical help andProf. F. Dunstan for statistical assistance.

REFERENCES

Antonarakis ES, Sampson JR, Cheadle JP. 2002. Temperaturemodulation of DHPLC analysis for detection of coexistingconstitutional and mosaic sequence variants in TSC2.J Biochem Biophys Methods 51:161–164.

Cheadle JP, Reeve MP, Sampson JR, Kwiatkowski DJ. 2000.Molecular genetic advances in tuberous sclerosis. HumGenet 107:97–114.

Gomez MR, Sampson JR, Holets-Whittemore V. 1999.Tuberous sclerosis complex. 3rd ed. New York: OxfordUniversity Press.

Jones AC, Sampson JR, Hoogendoorn B, Cohen D, Cheadle JP.2000. Application and evaluation of denaturing HPLC formolecular genetic analysis in tuberous sclerosis. Hum Genet106:663–668.

Jones AC, Sampson JR, Cheadle JP. 2001. Low level mosaicismdetectable by DHPLC but not by direct sequencing. HumMutat 17:233–234.

Jones S, Emmerson P, Maynard J, Best JM, Jordan S, WilliamsGT, Sampson JR, Cheadle JP. 2002. Biallelic germlinemutations in MYH predispose to multiple colorectaladenoma and somatic G:C-T:A mutations. Hum MolGenet 11:2961–2967.

Kuklin A, Munson K, Taylor P, Gjerde D. 1999. Isolation andanalysis of amplified cDNA fragments during detection ofunknown polymorphisms with temperature modulated het-eroduplex chromatography. Mol Biotechnology 11:257–261.

Kwiatkowska J, Wigowska-Sowinska J, Napierala D, Slomski R,Kwiatkowski DJ. 1999. Mosaicism in tuberous sclerosis as apotential cause of the failure of molecular diagnosis. N Engl JMed 340:703–707.

Osborne JP, Fryer A, Webb D. 1991. Epidemiology of tuberoussclerosis. Ann NY Acad Sci 615:125–127.

Verhoef S, Bakker L, Tempelaars AM, Hesseling-Janssen AL,Mazurczak T, Jozwiak S, Fois A, Bartalini G, Zonnenberg BA,van Essen AJ, Lindhout D, Halley DJ, van den OuwelandAM. 1999. High rate of mosaicism in tuberous sclerosiscomplex. Am J Hum Genet 64:1632–1637.

Xiao W, Oefner PJ. 2001. Denaturing high-performance liquidchromatography: a review. Hum Mutat 17:439–474.

CHARACTERIZINGMOSAIC MUTATIONS 115