2404-2412 Nucleic Acids Research, 1995, Vol. 23, No. 13

Electrophoresis for genotyping: temporal thermalgradient gel electrophoresis for profiling ofoligonucleotide dissociationIan N. M. Dayl,*, Sandra D. O'Dell1, Ian D. Cash1'2, Steve E. Humphrlies1 andGlenn P. Weavind2

1 Division of Cardiovascular Genetics, Department of Medicine, University College London Medical School, TheRayne Institute, 5 University Street, London WC1 E 6JJ, UK and 2NHS Chemical Pathology and University ClinicalBiochemistry, Level D, South Laboratory Block, Southampton General Hospital, Tremona Road, SouthamptonS09 4XY, UK

Received April 13, 1995; Revised and Accepted May 24, 1995

ABSTRACT

Traditional use of an o ligonueotde probe to detenninegenotype depends on p base pairing to a single-stranded target which is stable to a higher t e rethan when im ft binding occurs due to a mismatchin the taget sequence. Bound oligonuceotde isdtted at a prmined single tempeatre 'snap-shot' ofth meling profile, allowing the distnction ofperfect from m t base pairing. In hetothe sence of the alenative sequence must beverfied with a second oligonuleotide ry tothe variant. Here we describe a system of reatimevariable temperature Ieect rss during which theoligonucleotide dissociakts from its target. In 20%polyacrylamide the target strand has minimal mobilityand eleased oligonlode migrates ex y quicidyso that the 'eed' rather than the 'bound' is displayed.The full profile of oligonuceotide dissociation during geleectrphores IsIr along the gel track,and asingle oligonucleotde is sufficint to confirn hetero-zygosity, since the profile displays two separate peaks.Resolution is great, with use of short track lengthsenabling analysis of dense arrays of samples. Each geltrack can contain a different target or oligonuclotideand the temperature gradlent can accommodate oligo-nucleotides of d ifernt melting temperatures. Thisprovides a convenient systm to examine the interactionof many diffleret oflgonucldes and target sequencessimultaneously and requires no prior knowldg of themutant sequence(s) nor of oligonuceotideb meltingtemperatures. The application of the technique isdescribed for screening of a hotspot for mutations in theLDL or gene in pnts with familial hyper-cholserolaemia.

INTRODUCTIONA common approach to the 'genotyping' of sequence substitu-tions in DNA is the use of oligonucleotide binding assays (1,2).

An oligonucleotide with perfect base pairing remains annealed toits cognate single-stranded target at a higher temperature than willoccur if there is imperfect base pairing, such as a one basemismatch. The use of a stringent wash at a predeterminedtemperature in a defined ionic milieu enables demonstration ofperfect base pairing. Traditional techniques observe the boundfraction of oligonucleotide and are commonly configured in aformat which is convenient only for a single temperature'snapshot' of the melting profile. In addition, it becomes necessaryin samples of DNA from individuals heterozygous for thesequence change to demonstrate the alternate sequence with asecond oligonucleotide representing the variant sequence (3).These techniques are well-known as allele-specific oligonucleotide(ASO) assays. Typical configurations have included oligonucleo-tide labelled with 32p at its 5' end, used with target DNA fromcloned libraries or PCR reactions, denatured and immobilised onfilters by library blotting, Southern blotting of gel bands or spotblotting (4). Newer approaches have employed affinity capturegroups such as biotin incorporated into PCR products andnon-isotopic labels and detection systems for the ASO (e.g. 5). Thereverse procedure has also been described in which the oligo-nucleotide(s) is immobilised and the labelled target is hybridisedto the oligonucleotide or oligonucleotide array (6). However, in allof these procedures, the existent formats undertake a wash step atsome defined salt and temperature condition, followed by adetection step to observe the fraction still bound. The fraction stillbound depends on the conditions used and the affinity of binding.If the results from different wash temperatures are to be observed,sequential wash/detect cycles must be undertaken, which isusually either a laborious process or an impossible process fordetection procedures which are destructive to the annealingreaction or its constituents. Two approaches would make itpossible easily to gather all of the information of the ASOcomplete melting profile. First, real-time detection of the fractionstill bound during the wash would enable continuous observationof the melting profile: in the future such systems will becomeavailable as microsensors on microchips (7). Alternatively, thecollection of the 'freed' component for detection as a series of

* To whom correspondence should be addressed

Q-E.) 1995 Oxford University Press

Nucleic Acids Research, 1995, Vol. 23, No. 13 2405

fractions, or as a continuum, sequentially representing higherstringencies of washing, would achieve the same goal.

In this paper we describe an approach using electrophoresis as

the means of separating free from bound oligonucleotide, andtemperature as the stringency variable for dissociation of oligo-nucleotide from its target. Several advantages over traditional ASOmethods are considered. We have used this technique fordetermining profiles of oligonucleotide dissociation and forscreening for mutations at a hotspot in the LDL receptor gene infamilial hypercholesterolaemia patients.

MATERIALS AND METHODS

Oligonucleotides

Oligonucleotides were from Genosys, Cambridge, UK.Five oligonucleotides were used. The PCR primers used to

amplify a 100 bp sequence at the 3' end of exon 4 of the LDLreceptor gene were:

FH43 (sense primer): 5'-GATGGTGGCCCCGACTGCAAG-3';FH56 (antisense primer): biotin-5'-GGGACCCAGGGACAGGT-GATAGGAC-3'.The allele specific oligonucleotides used were:

SB 101: 5'-AAATCTGACGAGGAAAAC-3';SB102: 5'-AAATCTGACAAGGAAAAC-3';FH153: 5'-AAATCTGAGGAGGAAAAC-3'.The use of biotinylated oligonucleotide for PCR purely reflected

availability of oligonucleotide in the laboratory at the time. Theselection of oligonucleotides was based on availability in thelaboratory for research on a hotspot for mutations in exon 4 of theLDL receptor gene in patients with familial hypercholesterolaemia.

PCR products

APCR product of 100 bp was prepared from normal humanDNA(PCR-N/N) which was a perfect match with SB101, and fromheterozygotes for mutations mismatched to the CpG in SB101reading on the target strand either ...AGT-FlT1CCTCGTCGTCA-GATIT.. (PCR-N/M1, a one base substitution) or ...AGTITl-CCTCCAGATTT... (PCR-N/M2, a two base deletion). Allsymmetric PCR products were prepared using 20 pl reactionswith 20 pl oil in Omniplates on a Hybaid Omnigene, (Hybaid Ltd,Teddington, UK), PCR conditions were as follows: 1 cycle, 96°CS min; 68°C S min; 35 cycles, 96°C 1 min; 68°C 5 min; 1 cycle,68°C 1 min. Each 20 pl reaction mix contained 1.5 mM MgCl2,8 pmol each PCR primer, estimated average 40 ng human DNA,0.2 U Taq DNA polymerase (Gibco-BRL), 200 pM each dNTP,5% (v/v) WI (Gibco-BRL), 50 mM KCI, 10 mM Tris-HCl (pH8.3), 0.01% (w/v) gelatin. Asymmetric PCRs were prepared by anidentical PCR, but containing 80 fmol rather than 8 pmol of theoligonucleotide FH43 priming sense strand synthesis.

Oligonucleotide labelling

Each ASO was 5'-labelled with 32P, by incubating at 37°C for 1 hin a 50 pl mix containing 20 pmol oligonucleotide, 10 U T4polynucleotide kinase (Gibco-BRL), 25 pCi [y-32P]ATP (Amer-sham), 50 mM Tris-HCl (pH 7.6), 10 mM MgCl2, 5 mMdithiothreitol.

Oligonucleotide annealing reaction

32P-labelled oligonucleotide (1 pl of a 1/50 dilution of thelabelling reaction) was heated at 95 °C for 1 min with 4 ,ul ofPCRproduct, 1 gl 80mM MgC12 in 400mM Tris-HCl pH 7.5, and 2 plloading dye mix [0.25% bromophenol blue, 0.25% xylenecyanol, 40% (w/v) sucrose in water], then left at room tempera-ture (23°C) for 30 min.

Temporal thermal gradient electrophoresis

The electrophoresis buffer was pre-cooled to 10°C. Afterannealing of oligonucleotide to target was complete, 7 pl ofannealing mixture was loaded under TBE buffer into the wells ofa 2 mm thick horizontal 20% polyacrylamide gel, and preparedas described (8). The loaded gel was overlaid with a glass plate,transferred to a carrier rack and lowered into the electrophoresistank. A stirrer delay of 10 min was used to exclude the risk ofturbulence dislodging the samples from the wells at the start oftherun, although in practice this was prevented by the glass platecover. During electrophoresis the temperature was raised evenly,for example from 10 to 50°C over 300 min, after a run in of 25min at 10°C. This ensured that 'never-bound' labelled oligo-nucleotide would be resolved well in front of the dissociationprofile of oligonucleotide from mismatched or perfectly basepaired target PCR strands.

Autoradiography

The wet gel, attached to its glass plate as described (8), waswrapped in 'Clingo-rap', with a dry contact ensured between the'Clingo-rap' and photographic film. The gel was autoradio-graphed overnight at -70°C using Kodak Hyperfilm MP with anenhancement screen.

Mark I apparatus

A 2 mm thick horizontal 20% polyacrylamide gel (8), whichallows oligonucleotide migration but retards larger PCR targetstrands near the origin, was used. A very thin plastic electrophore-sis tank, to accommodate the glass plate with gel laid directly inits base, minimally covered by buffer, was made to fit on top ofthe block of a PCR machine, so that the PCR machine couldcontrol the temperature of the gel during electrophoresis. Apolystyrene block was floated on the buffer to achieve thermalinsulation.

Mark H apparatus

The design features of this equipment (Fig. 1) are: (i) a deepcuboid tank to accommodate 2 1 of buffer and a rack of 10-20gels. The tank is of silica glass to withstand thermal stresses; (ii) aplastic impellor to ensure efficient stirring and even temperatureat all points in the tank; (iii) an infrared lamp beneath the tank anda black perspex platform in the base of the tank to absorb infraredirradiation, whence heat is transferred into the circulating buffer.This system of heating evades the need for a heating element(which would have to be ceramic coated on account of theelectrolyte/electrophoresis) in the tank and has minimal intrinsicthermal mass thus making both small and rapid temperaturechanges feasible; (iv) a serpent cooling pipe passing through thebase of the tank to enable cooling by a cooling recirculator;(v) standard electrophoresis electrodes; (vi) thermocouples in the

2406 Nucleic Acids Research, 1995, Vol. 23, No. 13

( ooIat--t -1 e+. ou+ e-

D Housing to-*nfra e anm

Figure 1. Scheme of apparatus constructed for development of PODGE technique.

tank and on the cooling coil inlet for feedback of temperatureinformation to the control electronics/software; (vii) an analogue-digital control box interfacing computer control by purpose-written software with thermocouple feedbacks, power supply tothe infrared elements, solenoid control offlux through the coolingcoil, and electrophoretic power supply to the tank.

Equating ditance of oligo migration with temperatureof dissoiation

The physical resolution achieved on the gel track is dependent onthe programmed temperature gradient with time and on theelectrophoretic conditions employed. TBE conductivity as adependent variable oftemperature was calculated from measuredcurrent supplied for constant voltage. In the range 4-600C thefunction appears to be linear and is reflected also in the relativemobilities of dyes and oligonucleotides (data not shown).Defming total migration distance, D; voltage applied, V;

temperature, T; time, t; and a conductivity parameter, m,

m = kl + k2T

where kl and k2 are constantsand,

AD = Vmt 1

With a linear temperature (7) gradient against time (t),

T= k3 + k4t

Therefore,

AD = V(k1 + k11t)&t,hence,

D = V J (kI + k11t)dt= Vk1t + (Vk11t2)/2 2

Thus for electrophoresis under constant voltage, distancemigrated increases parabolically with time. It may be preferableto have a linear relationship between migration distance with timeand hence a linear relation between migration distance andtemperature of dissociation for oligonucleotides, so that calibra-

tion is simplified and spatial resolution is linearly proportionateto temperature (for linear temperature rise with time). Use ofconstant current (I) rather than constant voltage achieves this,appreciable from the terms above by substituting V with Tim, i.e.the dependent variablem is cancelled out in equation 1. However,in the absence of a power pack capable of supplying constantcurrent (in the mark II system, preferably up to 1 A), calibrationof migration using a model of the form of equation 2 isappropriate. The studies described in this paper were undertakenusing constant voltage, for trivial reasons (delivery delay for amore suitable power pack). The relative mobility of oligonucleo-tide to bromophenol blue was constant at different temperatures.

RESULTSEquipment

The mark I prototype using an electrophoresis apparatus fashionedon the surface of the block of a PCR machine, only enabled aneffective gel area of 60 x 80 mm2. Nevertheless, reasonabletemperature control was possible with the crude design employed,sufficient to demonstrate that the method was a viable way todisplay both perfect match and mismatch binding events for asingle oligonucleotide (Fig. 2).The purpose-built mark II apparatus was capable of accurate,

precise temperature control homogeneously throughout a 2 1electrophoresis tank, to within 0.1-0.2°C of a programmeddesired temperature profile. Maximal rate of heat input was 1200W, maximal rate of heat dissipation was 70 W at 10°C, typicalamperometric heating was 10-15 W at 100C, 40-50W at 400C.The tank has the capacity for 10-20 horizontal polyacrylamide orMADGE gels (8) to be electrophoresed simultaneously, enabling1000-2000 target oligonuceotide dissociation profiles to bedetermined simultaneously.

Establishing conditions for ASO SB1O1 binding toPCR-N/N, PCR-N/M1 and PCR-N/M2 products

Comparison of SB101 binding to PCR targets generated bysymmetric and asymmetric PCR. The oligonucleotide SB101anneals to the antisense strand of the 100 bp product. Prior toannealing, the double-stranded symmetric PCR product was

Nucleic Acids Research, 1995, Vol. 23, No. 13 2407

N/ N/ N/ oligoN! N/ N/N Ml M2only M2 M1 N

tI T

origin --

matched oligo-j-{-

mismatched oligo --

never bound oligo -

Figure 2. An early proof-of-principle of profiling of oligonucleotide dissocia-tion by gel electrophoresis during temporal thermal gradient of the gel. Aninitial proof-of-principle experiment using a simple adapter enabling thecombined use of a PCR block and an electrophoresis power pack for temporalthermal gradient electrophoresis using a horizontal 20% polyacrylamide gel (8)used symmetric PCR product and a high concentration of 32P-labelledoligonucleotide. Although perfect match (N/N) and mismatch (NIM1 andN/M2) binding were demonstrated, signal was weak relative to the excess ofnever-bound SB101. This reflects the competition between SB101 andequivalent strand of the PCR product (which is in considerable molar excess

over SB101). An improved adapter would make this approach feasible for lowthroughput applications and the experiment also illustrates the reason andrationale for the use of asymmetric PCR in the final approach.

melted at 95 'C. In a standard PCR as the mixture cools during theannealing stage the duplex reforms, with the sense strands incompetition with the oligonucleotide for the antisense bindingsite. As a result, limited binding of oligonucleotide to target wasseen using DNA generated by standard PCR (Fig. 2).

In order to minimise competition between oligonucleotide andPCR antisense strands during reannealing, an excess of antisensestrands is required. These were generated in an asymmetric PCRby creating a 100-fold depletion of sense primer FH43 relative toantisense primer FH56. The excess antisense strands obligatorilyremain as single strands available for oligonucleotide binding.

Antisense:sense primer ratios of 100:1, 10:1, 3:1, 2:1 and 1:1(standard PCR) were tested using genomic DNA template.Autoradiography confirmed that the 100:1 ratio originally usedproduced the greatest binding ofSB101 to target, but satisfactorybinding (i.e. sufficient yield of single-stranded target) was alsoachieved with ratios down to 3:1. From separate experiments,(data not shown) quantitating asymmetrical PCR single strandyield, we estimated that compared with the amount of oligo-nucleotide in an annealing reaction, the target was in -10-foldmolar excess over the oligonucleotide.

Effect of salt concentration in the annealing reaction. Finalconcentrations in the annealing mixture of 0.1, 0.5 and 1.0 MNaCl and 1 and 10 mM MgCl2 were compared with a watercontrol. The autoradiograph showed that there was no never-bound oligonucleotide remaining when 0.5 and 1.0 M NaCl or 10mM MgCl2 were used; all the oligonucleotide bound to the targetstrands. However there was preferential binding of oligonucleo-tide to the mutant strand of the heterozygotes, suggesting that themutant had a higher affinity than the normal strand for theoligonucleotide under these annealing conditions, a paradoxicalresult because SB101 matches the normal sequence. Thisproblem was addressed later (next section).

Effect of annealing time and temperature on oligonucleotidebinding (asymmetric PCRs). A variety of conditions wereexplored, including presence or absence of initial denaturationsteps of 95 or 60°C, different temperatures of annealing [wet ice,room temperature (23°C) and higher temperatures] and differenttimes of annealing. The conditions were as otherwise described inmethods, including 10 mM MgCl2 (final concentration).

(i) No denaturation step, annealing on wet ice. Cold conditionsresulted in slow annealing, for example 60 min was required toobtain a good level of binding in the presence of 10 mM MgCl2.There still remained substantial amounts of never-bound oligo-nucleotide.

(ii) No denaturation step, annealing at room temperature (23 °C).The autoradiograph showed that binding to the mutant strand of theheterozygote remained constant at all times tested, and completebinding was effected within 5 min. Binding to the normal strand ofthe heterozygote took 25-30 min to reach the same level asestimated from the intensity of the signal on the autoradiograph.The reason for the paradoxically faster binding of the oligo-

nucleotide to the mutant sequence is assumed to be due toconformational differences in the two strands; normal and mutantsingle-stranded DNAs assume different single strand conforma-tion polymorphisms (SSCPs) as a result of minor sequencechanges, (see Discussion). Accordingly the next step was toattempt to equalise initial oligonucleotide binding to mutant andnormal strands in the heterozygote by preheating the annealingmixture to 95°C for 1 min, followed by annealing at roomtemperature (23°C) for 30 min in the presence of 10mM MgCl2.

(iii) 95°C denaturation step, annealing at 23 °C. The autoradio-graph revealed that this procedure was very successful inimproving binding of oligonucleotides to the normal strand of theheterozygote. There was very little never-bound oligonucleotidein the heterozygote track; most was bound to target andreasonably equally distributed between perfect and mismatchedtarget strands. Thus although pre-heating at 95°C had beenrejected as unnecessary from the viewpoint of liberating addi-tional antisense target strands from the double-stranded compo-nent of the PCR product (see above), it proved to be important todestroy a conformation adopted by normal single strands whichtended to exclude oligonucleotide binding. However, we believethat in our first experiments using symmetric PCR, the limitedamount of oligonucleotide binding (detectable using 50-fold morelabelled oligonucleotide than in subsequent experiments usingasymmetrical PCR), may have mainly represented oligonucleotideannealed in a bubble within subsequently reannealed antisense/sense duplexes. To avoid this additional potential molecularcomplex when using asymmetric PCRs, denaturation at 60°C,which does not denature duplex PCR products was tried.

2408 Nucleic Acids Research, 1995, Vol. 23, No. 13

2 4 (I. 8 9 V) ... .SES] i@

matched ol'go .9:-:...

T'jIdt.C hed 0 I

mite bo.-iri oi.gc

Figure 3. Screening of a set of samples for mutation within the binding regionfor an individual oligo. 32P-labelled oligonucleotide SB1O1 was annealed toasymmetric PCR product from DNA templates representing a set of probandsfrom familial hypercholesterolaemia families. The mixes were then loaded on

a submerged horizontal 20% polyacrylamide gel and subjected to electrophore-sis with a temporal thennal gradient (linear and increasing) of the buffer.Never-bound oligonucleotide migrates the furthest, oligonucleotide bound andperfectly matched to its target migrates the least, and oligonucleotide bound butwith a mismatch (samples 3, 5, 6 and 12) migrates an intermediate distance.

(iv) 60°C denaturation step, annealing at 23°C. The autoradio-graph showed a similar level of oligonucleotide binding to normaland mutant strands in the heterozygote, with minimal oligo-nucleotide remaining never-bound.

Effect ofpH on oligonucleotide binding. The pH of the annealingmixture was adjusted using Tris-HCl to a final concentration of50 mM. Three pHs were tested, 7.5, 8.2 and 8.7 at 23°C, theannealing temperature. From the autoradiograph, there was no

visible binding at pH 8.7 but satisfactory results were obtained atpH 7.5 and 8.2. As the pH of Tris rises with falling temperature,the lower pH, 7.5 was chosen for subsequent trials because the pHof the mixture would rise on loading the gel at 10°C. The higherpH would risk guanine deprotonation (9).

Screening for mutations at a 'hotspot': LDL receptorgene exon 4 3' end as an example

In this phase of development of the profile of oligonucleotidedissociation during gel electrophoresis (PODGE) system, thefeasibility of testing multiple samples simultaneously for variationat the site targeted by the SB 101 oligonucleotide was assessed. Thissite contains a CpG dinucleotide sequence which is a welldocumented mutational 'hotspot' in the LDL receptor gene (10).Initially a set of 12 samples comprising genomic DNA of FHheterozygotes with known mutations in exon 4 was tested. Thesehad been included as positive controls in a large scale analysis of800 FH probands undertaken in this laboratory for LDL receptormutations (11; O'Dell et al., in preparation). Only four of the 12mutations were expected to be identifiable by SB101; the rest layoutside the targeted sequence.One short gel was required forarow of 12 samples and an SB 101

marker. The autoradiograph of the gel after electrophoresis isshown in Figure 3. D3, D5, D6 and D12 all show the dissociationpattern on the autoradiograph characteristic of a heterozygote.They are all frameshift mutations at codon 206, a 2 bp deletion,which was the M2 mutation used during our method development.

Subsequent experiments (data not shown), in which oligo-nucleotide is first purified from unincorporated [y-32P]ATP byspun column, have demonstrated the feasibility of shorteningelectrophoresis time and track length, and using multiplerows/arrays of tracks, so that >100 target oligonucleotidedissociation profiles could be determined per gel.

Dissociation prordes of oligonucleotides from target

Single oligonucleotides SBIO, SB102, FH153, FH43. In thePODGE approach we are using electrophoresis as the means ofseparating bound from free oligonucleotide and temperature isthe wash stringency variable. The continuous melting profile ofa single oligonucleotide complementary to the normal sequenceis determined as temperature rises over a set range. Thus, in theheterozygote, the single ASO displays three spots in order ofdecreasing migration. The first fraction released (at the lowesttemperature) is the never-bound oligonucleotide, migrating as afree species for the duration of the run. The second fraction isreleased at a low temperature because it has a mismatch with themutant PCR strand to which it has bound. The third fraction isreleased at the highest temperature because it was annealed withperfectly cognate strands in the PCR product. The greater thesequence difference between the normal and mutant strands, theearlier the release of the mismatched oligonucleotide with respectto the oligonucleotide perfectly matched to the normal sequenceand, consequently, the greater the migration distance between thedissociated oligonucleotide spots.

Results from the four oligonucleotides used are shown inFigure 4 and were each tested singly against three targetsPCR-N/M 1, PCR-N/M2 and PCR-N/N, using the same annealingconditions established for effective binding of SB101.The autoradiograph (Fig. 4) clearly differentiates the behaviour

of SB 101, the perfect match to the normal sequence, and SB 102,which differs from normal by one base. SB101 is released fromthe M2 strand of the heterozygote before its release from M1because it differs by one base from Ml, and by two bases fromM2, (estimated temperature of release from M2 is 28°C and fromMl is 31°C). SB102 is released from the N strand earlier thanSB 101 (estimated temperature of SB 102 release from N is 31 °Cand of SB101 from N is 45°C). SB101 matches N, SB102 has aone base difference to the normal sequence. FH43 is the sensestrand primer used in the PCR but here is used as an ASO, bindingto its complementary sequence on the antisense target generatedin excess. It has a Td greater than SB 101 and SB 102 [owing to itshigher %(G+C) content] and is released later but simultaneouslyfrom the same sequence on the M and N strands (estimatedtemperature ofFH43 release fromM andN strands is 50°C). Thisoligonucleotide was not included to detect mutations but as ademonstration of oligonucleotide dissociation from a second siteon the target. FH153 matches MI, which represents the D206Emutation in exon 4 of the LDL receptor gene (12). FH153 isreleased from the M2 strand first, which has a 2 bp deletion at therecognition site (10) and so has the greatest mismatch to theoligonucleotide (estimated temperature of release is 25.5°C).FH153 is released from the N strand next, from which it differsby a single base (estimated temperature of release is 31 °C). It isreleased from the MI strand last, with which it is perfectlymatched (estimated temperature of release is 36°C). SB102differs at adjacent centrally placed bases from N, Ml and M2 byone, two and three bases, respectively, and the respective melting

. i}. . .

Nucleic Acids Research, 1995, Vol. 23, No. 13 2409

aC->G = "Ml"

oligoFH43-> deletion of AC="M2"

gatggtggccccgactgcaagGACAAATCTGAcgAGGAAAACTGCGGTAT............+.+++

region bound by SB101,SBl02,FH153

GGGCGGGGCAGGGTGGGGGCGGGG Cgtcctatca cctgtccctg ggtccc

<-oligoFH56

s.lddfffil.i; k, .1imik..

Ooft% uAsff jw,b W

..:.....

c

Normol Aflelesand/orProbes

Mutant Allelesand

Probes

Ff143 perfect rnctch tocil torget stron.:as

SBtOI perfect match toN torget stranas 3 ,

FH 153 perfect rralch toMI torget strands :_

o-cse mismtches:SB 01IM I:SBl02/N:FH153JN.SB101 mlsmatch to 2odeltion af M2 taoget st-arrtw2 & 3 oase misrnatches-SB 102/M 1.:SB 102tM2:FH 153/Ms.

Never boound oligos = p

o0 0 0 0~~~ ~~ ~~~~~~~~~00o~~~~~~~~~~~~~~~~~~1o3u O a Lo o

o~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~uuO t)I- Ugu O3(iK < -K -K < C < <-K <- <u0oQteUouucsu<D o o O uu00 00 -0 -0 < 0

< _ < - U< <j -K < -K u< d:>l O lU a,L<- K- < - 0- 0- 0u

- 'K- 0- <- 0- 0 0 oil__ .< < _ < - - < < < K I

: _ <, < Uo U19UOU U t9 U O U O tS t- -0 -0 -0 0 0 0 0XoX U e eJ o o U o o @ U e o Uo~~~~~8U a uIs (9 11a<

ei <.; <u " < <i < < 4v <a u ab .

° Ll ° 4u

J 000-00 00 0 0 -0 -0 --ua u Xu d U e O- <1 U e Q d uu 0 <B udukOdO0J 0 SUO i_< _< _ _ -<4

<

- _< _ < -< _ u e

CD

ou(9

U "U IJe u u o u u u0o 0 003U 0u _ 0- - -< -K K < < - < _V U @o<_<-< < < - < _~~~~~~~~~~~<- < u_

4:- < - < - ^ < F < - < < < - <; - < - < -00u 00u L 0

m m urnd 1 d o o e d u urU

SB101 SB 112 FH1153 FH43 FH43+SB I 01M I M2r N M I M2 N M I M2 Ns N M I NU N N Mv 1

N 2/N NN/O

., I0. 0.. . !I'*~~~ @- 0- 00-000_m--_-------

u uu Qa cU u

'i

.., I LI< 4-C - -1.K - <

a L, LI L.a (i 0 u-C -

a L) .l uU.2 v ..

'iI

1 2 Vu a C, 0- -k

-K :: " I< K -

0 (3

a

1-

I4

LqdL

I

m

Iuu(Bu-C

L,I

(9a !.A0 (i

.1 0.

u-1

'i Es-

-E ::a

z

ct

I

L)(4 :-C

11:0 (0 (.c0 (

1.1.

K :

.Kz

z

eiu

e

Q

Li 0u U11 0u u.c K0

u

uI6 (D

2f.) u

0 ::I

..J.& ....

L) 0u t,CB 0u-K .K

1 2- 'Ku 0u a u

-C. .u0 U-

U U-C .KC 'i 0- I

:: 4K .:O a

L)

-1

2410 Nucleic Acids Research, 1995, Vol. 23, No. 13

events and order are evident from the results in Figure 4(estimated temperatures of release are 20.5°C from M2, 22.5°Cfrom Ml and 31 °C from N).

Binding ofSB101 and FH43 to the same MI and N targets. Thepairs ofoligonucleotides were pre-mixed so that on addition to theannealing mixture they were both at the same fmal concentrationas the single oligonucleotides. The object of this trial was todetermine whether more than one oligonucleotide was resolvableon a single track, thereby improving the throughput for potentialmutational screening purposes. Ml heterozygote targets withsingle oligonucleotides served as markers for the positions ofoligonucleotides released from M and N strands.

Figure 4 shows the pattern of binding and release of the twooligonucleotides in combination that would be expected fromtheir behaviour as single probes against the targets. FH43 bindsto and is released from both targets in the same way, as bindingis to the sequence on the antisense target strand complementaryto this sense primer, used as an ASO. SB 101 shows earlier releasefrom the mutant strand with the single base mismatch than fromthe normal strand to which it is perfectly matched.

DISCUSSION

The mark I approach was sufficient to establish proof-of-prin-ciple, and although very crude, shows that it would be possible tomanufacture a simple 'adapter' electrophoresis tank which couldutilise a standard PCR block for variable temperature control anda standard power pack for electrophoresis.The mark H system described offers a convenient means to

visualise the full dissociation profiles of oligonucleotides fromtargets to which they have either fully or partially base paired.Electrophoresis buffers are usually of low salt concentration, sincehigh salt buffers would result in much slower migration of thesample or would demand much higher current supply andnecessitate means to counteract amperometric heating. Thisconflicts with commonly used conditions for oligonucleotidehybridization, where very high concenutaions of salt are used. Thesystem described here illustrates that it is possible to combine theuse of moderate length oligonucleotides with standard electro-phoresis buffer, and we have in addition developed an apparatus forachieving a highly precise, programmable temperature variableduring the time course of the electrophoresis. Oligonucleotides inthe range 18-21mer had melting profiles in lx TBE for match andmismatch binding, falling in the range 15-55°C, which wasconvenient for the apparatus constructed. Shorter oligonucleo-tides with mismatches would be expected to display much lowerdissociation temperatures which would necessitate much lowerannealing temperatures (if possible at all) or much higher saltannealing and electrophoresis buffer: our data (not shown)support this expectation.

Given oligonucleotide bound to target, the method enables theautomated collection of the full dissociation profile, with theoriginal variable of temperature represented after analysis as aspatial display on a gel track. This approach has a number ofadvantages and potential advantages enumerated below.

Advantages of PODGE

Visualisation of full melting profile of oligonucleotide. Thecontinuous profile of oligonucleotide melting is determined, incontrast with standard allele-specific oligonucleotide methodswhich only look at the oligonucleotide still bound in a tempera-ture 'snapshot' of the profile. Thus in a heterozygote, an ASOrepresenting the normal sequence (rather than mutant mismatch)displays three bands/smears in order of decreasing migration:first, the never-bound oligonucleotide migrating as oligonucleo-tide for the duration ofthe run; secondly, oligonucleotide releasedat a lower temperature because it has a mismatch with the mutantPCR strands to which it has bound; finally, oligonucleotidereleased at a higher temperature because it was annealed withperfectly cognate target strands in the PCR product.

No predetermination of melting temperatures necessary. Thecontinuous nature of the temperature profile means that fullmelting profiles are always visualised, evading any requirementfor prior determination of melting temperature representingperfect match binding.

No needforprior knowledge ofmutant sequence. Since only oneoligonucleotide, representing the known/normal sequence, isused, it is not necessary to know in advance the variantsequence(s). This may be useful in screening potential hotspotsfor postulated but unknown variations.

High resolution enables short track length, compatible with highdensity arrays. The gel shift of the more mobile species (oligo)by the larger, less mobile PCR target, is far greater than thereciprocal gel shift. In high percentage gels the mobility of PCRtargets approaches zero. The mobility shift of oligonucleotidedependent on base pairing (perfect or imperfect match) to targetand consequent resolution is massive.and therefore very shorttrack lengths such as those on MADGE gels (8) can readily beused. During the development of PODGE, we have used 2 mmthick horizontal 20% polyacrylamide gels autoradiographed wet,and 32P-labelled oligonucleotide for detection. On account of thethick gel and long path length of ,B-emission, the resultantautoradiograph is not truly representative of the actual meltingprofiles. Nevertheless, resolution is clearcut and massive and hasreadily been adapted down to 2-3 cm track lengths. We fullyexpect that a combination of non-isotopic detection and closerjuxtaposition of reporter to detector will enable much higherresolution and dense arraying of tracks.

Figure 4. Dissociation profiles of a set of oligos with a range of zero, one, two or three base mismatches to target strands, and an illustration of combinatoric use. (a)Schematic showing the relationship of the oligonucleotides used, representing a PCR product and site of multiple mutations in the LDL receptor gene in patients withfamilial hypercholesterolaemia. PCR primer oligonucleotides and CpG hotspot shown in lower case letters. (b) Autoradiographs of dissociation profiles, oligos andtargets as explained in schematic (C). Wells are at the top of the photograph and electrophoresis is in a downward direction. (c) Schematic interpretation of resultsobtained. Base pairing of oligos to targets is shown at the top: differences between oligos and between targets (relative to SB101 and N) are shown in small case, withmismatch base pairs marked with an x. It should be noted that FH153 has perfect match to mutant sequence MI and mismatch to normal sequence. Two N/N tracksare included for FH1153, one track had a low yield of PCR product and hence, in contrast with the adjacent track, there is some never-bound oligonucleotide present.The identity of each dissociation event is indicated at the left-hand side of the figure.

Nucleic Acids Research, 1995, Vol. 23, No. 13 2411

Advantage of temporal thermal gradient. A two-dimensionalarray of gel slots for sample loading is made possible by thetemporal thermal gradient, whereas it would not be possible witha spatial thermal gradient.

All-in-one result, no re-synthesis of data. Apart from obtainingsufficient signal, interpretation is not dependent on the concentra-tion of the target strands. This contrasts with ASO analyses on spotblots and gel blots, where a 'grey' signal may be the result eitherof a perfect match target in low concentration or an imperfectmatch target in high concentration. In these instances, review oftheseparate results from separate filters is inevitably necessary. Thesignal representing the full melting profile of the oligonucleotideprovides an internal check of the success of the PCR, a separatecheck is not necessary. We have found that there is usuallycomplete binding of oligonucleotide to PCR product, so thatpresence of never-bound-oligo signal typically marks a PCRdropout. In a typical asymmetric PCR (with one oligonucleotidereduced 100-fold in concentration) drop-out rate in a 96well plateis -5%. It should be noted that a gel autoradiographed wet cansubsequently be stained or undergo a second electrophoresis, forexample with a profile in a higher temperature range.

Each gel track represents an independent hybridisation event. Thecompartmentalization of target and oligonucleotide down anelectrophoresis track enables different targets and different oli-gonucleotides to be analysed on the same run. In effect, each gelslot equates with a separate hybridization bag and each gel track isan axis for display of a complete melting profile.

Multiplex oligo decoration of target is feasible. The spatialresolution and display of both perfect and imperfect oligonucleo-tide binding in a mixture has high combinatoric potential, forexample to ask the question 'does any one of a large number ofoligonucleotides decorating a target dissociate at a lower tempera-ture than normal?'. In effect this would be a short track length assayfor sequence variation, in contrast with SSCP (13) or heteroduplexanalysis (14) which depend on the detection of small mobilityshifts. The method represents a different potential approach tosequencing-by-hybridisation (15).

Additional concepts ofPODGE which may be of interest

(i) Concept of observing the 'freed' rather than the 'bound'. Thisis the first allele-specific oligonucleotide method, to our knowl-edge, to look at the freed as well as, or rather than, the bound.(ii) Target could also be labelled. It does not matter if the targethappens also to be labelled, because of the eventual spatialresolution from oligonucleotide signal. (iii) No bulk flow or wash,therefore no dilution of 'eluted' labelled oligo. Since electrophore-sis, rather than bulk flow, is used to separate freed from target, thereis no dilution of signal. (iv) Entirely liquid phase, no solid phaseanchorage. The oligo/target annealing is entirely in liquid phase,thus avoiding the distortions and lesser access of solid phaseanchorage which may artefactually flatten the melting curve.(v) Free oligo is at infinite dilution (instantly separated), andprofiles are pure dissociation profiles. Little is known about thekinetics of oligonucleotide annealing/dissociation events, butstandard hybridisation approaches usually employ washes with anexcess of buffer with a combination of temperature and timeselected to eliminate all but perfect match binding. Either a

used, or a time short enough not to eliminate all binding must beused at the dissociation temperature. Whichever approach is used,the signal observed does not represent the dissociation event. Thismay be important when the discrimination between match andmismatch events is small, for example with terminal rather thancentral mismatch, or possibly with oligonucleotides directed toshort (e.g. four base) homopolymeric runs with length polymorph-ism of one base where a spectrum of different annealing speciesmay tend to broaden the respective melting profiles.An altemative to the electrophoretic approach described here

would be to use a capture method (e.g. biotin/streptavidin) to trap thetarget strand, for example on a column, then to elute the boundlabelled 'interrogation' oligonucleotide using bulk flow in combina-tion with a temperature gradient. A column capture approach wouldhave the advantage that higher salt concentrations (and hence shorteroligonucleotides) could be used than is possible with electrophoresis.The apparent volume of the bands in our gels is high because we are

currently using 32P detection on 2 mm thick 20% polyacrylamidegels which have to be autoradiographed wet. Howeverwe know thatthe true bands are much more closely confined, the main spreadbeing that of oligo diffusion which would occur in any system. Abulk flow approach would likely result in dilution of the elutedlabelled oligonucleotide, but with sensitive reporter groups andconvenient means of fraction collection it would have potential as

an alternative to an electrophoresis approach. However, this woulddemand immobilisation of the target (which PODGE does not) andwould lack the convenience of continuous spatially resolvedcollection of the 'eluate' and of only requiring slots in a gel (e.g.rather than a thermally controllable set of columns and collectiondevices) to set up.

In summary, we have demonstrated the feasibility ofan approachfor determining the complete profiles of dissociation of oligo-nucleotides (annealed to target strands) by gel electrophoresisusing temporal thermal gradient in the buffer immersing the gelduring the course of the run. A variety of advantages has beendemonstrated over traditional ASO binding assays and some ofthese are now being utilised in the authors' laboratory for studiesof the relationships between human genetic diversity and cardio-vascular disease. In particular, the technique has first been appliedto the construction of a convenient screening assay selective formutations at a mutational hotspot in the LDL receptor gene infamilial hypercholesterolaemia patients, which has the potential forvery high throughput. Oligonucleotides sited at this hotspot havealso been used to demonstrate some of the other anticipatedadvantages of the technique.

Note added in proof

Experience with different PCR targets and oligonucleotidesindicates that the method should have widespread applicability.Our main concerns are with the equalisation of competition inheterozygotes for binding of labelled oligonucleotide (limiting inthe current format) by target alleles (in excess in the currentformat) some of which adopt different conformations of differentaccessibility, and low yields of asymmetric PCR target if longerPCR products are used. However, the use of 100 nucleotide targetstrands has been generally successful, and with a non-

radioisotopic detection system it will be possible to use conditionsof binding oligonucleotide in excess to minimise potential

temperature below the actual dissociation temperature must be competition between target strand alleles in heterozygotes.

2412 Nucleic Acids Research, 1995, Vol. 23, No. 13

ACKNOWLEDGEMENTSThis work was supported by UK MRC ROPA award G9414230to INMD and SEH. INMD is the recipient of a British HeartFoundation Intermediate Fellowship. Dr V. Gudnason is thankedfor oligo SB 101 and for template DNAs used for Figure 3. SEHis supported by a Chair award from the British Heart Foundation.The method is subject to patent application.

REFERENCES1 Wallace, R.B., Johnson,M.J., Hirose,T., Miyake,T., Kawashima,E.H., and

Itakura,K. (1981) Nucleic Acids Res., 9, 879-894.2 Wood, W.I., Gitschier,J., Lasky, L.A. and Lawn, R.M. (1985) Proc. Natl.

Acad Sci. USA, 82, 1585-1588.3 Erlich, H.A. and Bugawan, T.L. (1989) in Erlich, H.A. (ed.), PCR

Technology: Principles and Applications for DNA Amplification. StocktonPress, USA Chapter 16, pp.193-208.

4 Sambrook, J., Fritsh, E.F. and Maniatis, T. (1989) Molecular Cloning: ALaboratory Manual 2nd edn. Cold Spring Harbor Laboratory Press, ColdSpring Harbor, NY.

5 Vlieger, A.M., Medenblik, A.M., van-Gijlswijk, R.P., Tanke, H.J., van derPloeg, M., Gratama, J.W. and Raap, A.K. (1992) Anal. Biochem., 205,1-7.

6 Saiki, R.K., Walsh, P.S., Levenson, C.H. and Erlich, H.A. (1989) Proc.Natl. Acad. Sci. USA, 86, 6230-6234.

7 Beattie,K. Eggers,M., HoganM.,Varma,R., Lamture,J., Hollis,M.,Ehrlich,D. and Rathman,D. (1993) Clin. Chem., 39, 719-722.

8 Day, I.N.M. and Humphries, S.E. (1994) Anal. Biochem., 222, 389-395.9 Dawson, R.M.C., Elliott, D.C., Elliott, W.H. and Jones, K.M. (1986) Data

for Biochemical Research. Clarendon Press, Oxford, UK, Chapter 5.10 Gudnason,V., Day,I.N.M. and Humphries,S.E. (1994) Atherosclerosis and

Th7rombosis, 14, 1717-1722.11 Whittall,R., Gudnason,V., Weavind,G.P., Day,L.B., Humphries,S.E. and

Day,I.N.M. (1995) J. Med. Genet., in press.12 Kotze,M.J., Langenhoven,E., Warnich,L., Du Plessis,L., Marx,M.P.,

Oosterhuizen, C.J.J. and Retief,A.E. (1989) S. Afr: J. Med., 76, 399-401.13 Orita, M., Iwahana,H., Kanazawa,H., Hayashi, K., Sekiya, T., (1989).

Proc. Natl. Acad. Sci. USA, 86, 2766-2770.14 Keen,J., LesterD.,Ingleheam,C., Curtis,A. and Bhattacharya,S. (1991)

Trends Genet., 7, 5.15 Maskos, U. and Southern, E.M. (1992) Nucleic Acids Res., 20, 1675-1678.