Genotyping of Single-Nucleotide Polymorphisms [619092]

Genotyping of Single-Nucleotide Polymorphisms
by High-Resolution Melting of Small Amplicons
Michael Liew,1Robert Pryor,2Robert Palais,3Cindy Meadows,1Maria Erali,1
Elaine Lyon,1,2and Carl Wittwer1,2*
Background: High-resolution melting of PCR ampli-
cons with the DNA dye LCGreenTMI was recently
introduced as a homogeneous, closed-tube method ofgenotyping that does not require probes or real-timePCR. We adapted this system to genotype single-nucle-otide polymorphisms (SNPs) after rapid-cycle PCR (12min) of small amplicons ( <50 bp).
Methods: Engineered plasmids were used to study all
possible SNP base changes. In addition, clinical proto-cols for factor V (Leiden) 1691G >A, prothrombin
20210G >A, methylenetetrahydrofolate reductase (MTHFR)
1298A >C, hemochromatosis (HFE) 187C >G, and
/H9252-glo-
bin (hemoglobin S) 17A >T were developed. LCGreen I
was included in the reaction mixture before PCR, andhigh-resolution melting was obtained within 2 min afteramplification.Results: In all cases, heterozygotes were easily identi-
fied because heteroduplexes altered the shape of themelting curves. Approximately 84% of human SNPsinvolve a base exchange between A::T and G::C basepairs, and the homozygotes are easily genotyped bymelting temperatures ( T
ms) that differ by 0.8–1.4 °C.
However, in /H1101116% of SNPs, the bases only switch
strands and preserve the base pair, producing very smallT
mdifferences between homozygotes ( <0.4 °C). Al-
though most of these cases can be genotyped by Tm,
one-fourth (4% of total SNPs) show nearest-neighborsymmetry, and, as predicted, the homozygotes cannot beresolved from each other. In these cases, adding 15% ofa known homozygous genotype to unknown samplesallows melting curve separation of all three genotypes.This approach was used for the HFE 187C >G protocol,
but, as predicted from the sequence changes, was notneeded for the other four clinical protocols.Conclusions: SNP genotyping by high-resolution melt-
ing analysis is simple, rapid, and inexpensive, requiringonly PCR, a DNA dye, and melting instrumentation.The method is closed-tube, performed without probesor real-time PCR, and can be completed in less than 2min after completion of PCR.
© 2004 American Association for Clinical Chemistry
LCGreenTMI is a new fluorescent DNA dye designed to
detect heteroduplexes during homogeneous meltingcurve analysis (1). Genotyping of single-nucleotide poly-
morphisms (SNPs)
4by high-resolution melting analysis in
products as large as 544 bp has been reported. UnlikeSYBR
®Green I, LCGreen I saturates the products of PCR
without inhibiting amplification and does not redistributeas the amplicon melts. This allows closed-tube, homoge-neous genotyping without fluorescently labeled probes(2–4) , allele-specific PCR (5,6), or real-time PCR instru-
ments. Heterozygotes are identified by a change in melt-ing curve shape, and different homozygotes are dis-tinguished by a change in melting temperature ( T
m).
However, it was not clear whether all SNPs can begenotyped by this method.
SNP genotyping by amplicon melting analysis requires
high-resolution methods. The differences among geno-types are easier to see when the amplicons are short (7).
Use of small amplicons of /H1102150 bp also allows for very
rapid thermal cycling (8); amplification is complete in
/H1102112 min and is followed by high-resolution melting,
which requires /H110212 min.
We studied all possible homozygous and heterozygous
genotypes with differences at one base position, usingengineered plasmids. In addition, we developed assays to
1Institute for Clinical and Experimental Pathology, ARUP, Salt Lake City,
UT.
2Department of Pathology, University of Utah School of Medicine, Salt
Lake City, UT.
3Department of Mathematics, University of Utah, Salt Lake City, UT.
*Address correspondence to this author at: Department of Pathology,
University of Utah School of Medicine, Salt Lake City, UT 84132. Fax801-581-4517; e-mail carl.wittwer@path.utah.edu.
Received January 30, 2004; accepted March 30, 2004.Previously published online at DOI: 10.1373/clinchem.2004.0321364Nonstandard abbreviations: SNP, single-nucleotide polymorphism; Tm,
melting temperature; MTHFR; methylenetetrahydrofolate reductase; Hb, he-moglobin; and UNG, uracil N-glycosylase.Clinical Chemistry 50:7
1156–1164 (2004) Molecular Diagnostics
and Genetics
1156

genotype the common clinical markers prothrombin
20210G /H11022A(9), factor V (Leiden) 1691G /H11022A(2), methyl-
enetetrahydrofolate reductase (MTHFR) 1298A /H11022C(10),
hemochromatosis (HFE) 187C /H11022G(11), and /H9252-globin [he-
moglobin S (HbS)] 17A /H11022T(12) as examples for each class
of SNP.
Materials and Methods
dna samples
Most of the samples used in this study were blood
specimens submitted to ARUP (Salt Lake City, UT) forroutine clinical genotyping of prothrombin, factor V,MTHFR, or HFE mutations. DNA was usually extractedwith the MagNa Pure instrument (Roche) according to themanufacturer ’s instructions. Additional samples geno-
typed at the
/H9252-globin locus for HbS were provided as
dried blood spots by Pediatrix Screening Inc. (Pittsburgh,PA) and were extracted as described previously (13). All
samples were genotyped at ARUP or Pediatrix Screeningby melting curve analysis on the LightCycler
®(Roche)
using adjacent hybridization probe (HybProbeTM) tech-
nology and either commercial assays (Roche) or in-housemethods (2,11,12). At least three different individuals of
each genotype for prothrombin 20210G /H11022A, MTHFR
1298A /H11022C, HFE 187C /H11022G, and
/H9252-globin 17A /H11022T SNPs were
selected. We obtained 104 samples (35 wild type, 35heterozygous, 34 homozygous mutant) previously geno-typed for factor V (Leiden) 1691G /H11022A. We also studied
three rare DNA samples with mutations near factor VLeiden (1691G /H11022A); the first sample was heterozygous
1690delC, the second was heterozygous 1690C /H11022T, and the
third was compound heterozygous 1696A /H11022G and
1691G/H11022A. These samples were originally identified by
aberrant melting profiles during adjacent hybridizationprobe genotyping and confirmed by sequencing. All sam-ples were deidentified according to a global ARUP pro-tocol under Institutional Review Board no. 7275.
DNA samples obtained with the MagNa Pure or from
dried blood spots were not routinely quantified, butcontained /H1101110–50 ng/
/H9262L. However, for HFE 187C /H11022G
genotyping, DNA was extracted with a QIAamp DNABlood Kit (QIAGEN, Inc.), concentrated by ethanol pre-cipitation, and quantified based on the absorbance at 260nm ( A
260).
Engineered plasmids with an A, C, G, or T at a defined
position amid 50% GC content (14) were kindly provided
by Cambrex BioScience, Inc. (Rockland, ME). Plasmidcopy number was quantified by A
260.
primer selection and synthesis
To maximize the Tmdifference between wild-type and
homozygous mutant genotypes, the amplicons weremade as short as possible. The following process wassystematized as a computer program using LabView(National Instruments) and is available for remote use as“SNPWizard ”, at http:\\DNAWizards.path.utah.edu. Af-
ter input of sequence information surrounding the SNP,the 3/H11032end of each primer was placed immediately adja-
cent to the SNP. The length of each primer was increasedin its 5 /H11032direction until its predicted T
mwas as close to a
user-selectable temperature (usually 55 –60°C) as possi-
ble. The primer pair was then checked for the potential toform primer-dimers or alternative amplicons. If the reac-tion specificity was acceptable, the primers were selected.If alternative products were likely, the 3 /H11032end of one of the
primers was shifted one base away from the SNP and theprocess was repeated until an acceptable pair was found.
Duplex T
ms were calculated with use of nearest-neigh-
bor thermodynamic models described previously (15–22).
Best-fit values of 0.2 /H9262M for the amplicon concentration at
the end of PCR and the Mg2/H11001equivalence (74-fold that of
Na/H11001) were obtained by use of a data set of 475 duplexes
(23). The effective concentration of Mg2/H11001was decreased
by the total deoxynucleotide triphosphate concentration,assuming stoichiometric chelation. The effect of Tris
/H11001was
assumed equal to Na/H11001, and the [Tris/H11001] (20 mM) was
calculated from the buffer concentration and pH. Oligo-nucleotides were obtained from Integrated DNA Technol-ogies, IT Biochem, Qiagen Operon, and the University ofUtah core facility.
For the in silico calculation of the expected /H9004T
mdistri-
bution of SNP homozygotes, the six amplicons studied(factor V, prothrombin, MTHFR, HFE,
/H9252-globin, and
lambda) were considered. For each target, all combina-tions of the three base pairs centered on the SNP wereevaluated. A/A, C/C, G/G, and T/T homozygotes werepaired with each other, giving six pairs of homozygotesfor/H9004T
mcalculations. For each pair, 16 possible combina-
tions of neighboring bases were considered. Therefore,6/H110036/H1100316, or 576 /H9004T
mvalues were calculated and plotted
as a frequency distribution. Because the bases aroundSNPs are not completely random (24), the frequency of all
three base combinations centered on human SNPs wascalculated from the human SNP database (ftp://ftp.ncbi.nih.gov/snp/human, build 120). The analysis included7 291 660 biallelic SNPs on chromosomes 1 –22, X, and Y
having no immediately adjacent polymorphisms. The/H9004T
mdistribution was adjusted for the frequency of each
three-base combination for SNPs in the human genome.
pcr
Reaction conditions for the engineered plasmids and the
/H9252-globin samples consisted of 50 mM Tris (pH 8.3), 500
/H9262g/mL bovine serum albumin, 3 mM MgCl2,200/H9262M each
deoxynucleotide triphosphate, 0.4 U of Taq polymerase
(Roche), 1 /H11003LCGreen I (Idaho Technology), and 0.5 /H9262M
each primer in 10 /H9262L. The DNA templates were used at
106copies (plasmids) or 20 ng (genomic), and a two-
temperature PCR was performed with 35 cycles of 85 °C
with no hold and 55 °Cf o r1so n either the LightCycler
(Roche) or the RapidCycler II (Idaho Technology). PCRwas completed within 12 min.
PCR for the prothrombin, factor V, MTHFR, and HFE
targets was performed in a LightCycler with reagentsClinical Chemistry 50, No. 7, 2004 1157

commonly used in clinical laboratories. The 10- /H9262L reac-
tion mixtures consisted of 10 –50 ng of genomic DNA, 3
mM MgCl2,1/H11003LightCycler FastStart DNA Master Hy-
bridization Probes master mixture, 1 /H11003LCGreen I, 0.5 /H9262M
forward and reverse primers, and 0.01 U/ /H9262LEscherichia
coli uracil N-glycosylase (UNG; Roche). The PCR was
initiated with a 10-min hold at 50 °C for contamination
control by UNG and a 10-min hold at 95 °C for activation
of the polymerase. Rapid thermal cycling was performedbetween 85 °C and the annealing temperature at a pro-
grammed transition rate of 20 °C/s. The primer se-
quences, amplicon sizes, the number of thermal cycles,and the annealing temperatures for each target are listedin Table 1 of the Data Supplement that accompanies theonline version of this article at http://www.clinchem.org/content/vol50/issue7/.
Differentiating HFE wild-type and mutant homozy-
gotes required adding a known genotype to the samples.The known DNA could be added either before or afterPCR. To add the DNA after PCR, equal volumes of knownwild-type and unknown PCR homozygous products weremixed. To add before PCR, precisely 50 ng of unknowngenomic DNA was used as template, along with anadditional 7.5 ng of known wild-type DNA.
melting curve acquisition and analysis
Melting analysis was performed either on the LightCyclerimmediately after cycling or on a high-resolution meltinginstrument (HR-1; Idaho Technology). When the LightCy-cler was used, 20 samples were analyzed at once by firstheating to 94 °C, cooling to 40 °C, heating again to 65 °C
(all at 20 °C/s), and then melting at 0.05 °C/s with con-
tinuous acquisition of fluorescence until 85 °C. LightCy-
cler software was used to calculate the derivative meltingcurves.
When high-resolution melting was used, amplified
samples were heated to 94 °C in the LightCycler and
rapidly cooled to 40 °C. The LightCycler capillaries were
then transferred to the HR-1 high-resolution instrumentand heated at 0.3 °C/s. Samples were analyzed between
65 and 85 °C with a turnaround time of 1 –2 min. High-
resolution melting data were analyzed with HR-1 soft-ware. In most cases, plots of fluorescence vs temperaturewere normalized as described previously (1,7). For direct
comparison with LightCycler data, derivative plots wereused without normalization. All curves were plottedusing Microsoft Excel after export of the data.
Results
Melting analysis of short PCR products in the presence ofthe heteroduplex-detecting dye, LCGreen I, was used togenotype SNPs. Rapid-cycle PCR of short products al-lowed amplification and genotyping in a closed-tubesystem in /H1102115 min without probes or allele-specific am-
plification. The primer locations surrounding the six poly-morphic sites analyzed are shown in Fig. 1. The PCRproducts were 38 –50 bp in length, and the distance from
Fig. 1. Details of the SNPs studied, including the primer positions and
the SNP class (see Table 1).
Both strands of DNA are shown. The large arrows above andbelow the
sequences indicate the 3 /H11032position and directions of the primers. The small
vertical arrows indicate the SNP base change. For the lambda constructs, N
indicates that all possible changes were studied.1158 Liew et al.: SNP Genotyping Using LCGreen I

the 3/H11032end of the primers to the polymorphic site varied
from 1 to 6 bases.
The difference between standard and high-resolution
melting techniques is shown in Fig. 2, which showsderivative melting curves for different factor V (Leiden)genotypes. Although the heterozygotes can be identifiedby the presence of a second, low-temperature meltingtransition even with standard techniques, genotype dif-ferentiation is much easier with high-resolution methods.All subsequent studies were done at high resolution.
Engineered lambda constructs (14) were used to study
all possible SNP base combinations at one position. Fourplasmids (identical except for an A, C, G, or T at oneposition) were either used alone to simulate homozygousgenotypes or in binary combinations to construct “het-
erozygotes ”. The normalized melting curves of the four
homozygotes and six heterozygotes are shown in Fig. 3.All homozygotes melted in a single transition (Fig. 3A),and the order of melting was correctly predicted bynearest-neighbor calculations as A/A /H11021T/T/H11021C/C/H11021
G/G (22). Heterozygotes produced more complex melt-
ing curves (Fig. 3B), arising from contributions of twohomoduplexes and two heteroduplexes (7). Each hetero-
zygote traced a unique melting curve path according tothe four duplex T
ms. The order of melting was again
according to nearest-neighbor calculations (A/T /H11021
A/C/H11021C/T/H11021A/G/H11021G/T/H11021C/G) based on the mean of
the two homoduplex Tms. The six heterozygote curves
merged at high temperatures into three traces, predictedby the highest melting homoduplex present (T/T for the
A/T heterozygote, C/C for the A/C and C/T heterozy-gotes, and G/G for the A/G, G/T, and C/G heterozy-gotes). All genotypes could be distinguished from eachother by high-resolution melting analysis.
The genomic SNPs shown in Fig. 1 include all four
classes of SNPs. Table 1 lists the four classes of SNPs thatresulted from grouping the six different binary combina-tions of bases by the homoduplex and heteroduplexproducts that were produced when a heterozygote wasamplified. Class 1 SNPs are C/T and G/A transitions thatproduce C::G and A::T homoduplexes and C::A and T::Gheteroduplexes. In contrast, class 2 SNPs (C/A and G/T)are transversions that produce C::T and A::G heterodu-plexes. Class 3 SNPs (C/G) produce C::G homoduplexeswith C::C and G::G heteroduplexes. Class 4 SNPs (A/T)produce A::T homoduplexes with A::A and T::T hetero-duplexes. The clinical SNPs studied were chosen to in-clude two examples (factor V and prothrombin) in themost common SNP class and one example in each of theother three classes.
The melting curves for the five clinical SNP targets are
shown in Fig. 4. For all SNP classes, heterozygotes wereeasily identified by a low and/or broad melting transi-tion. For SNPs in class 1 or 2 (factor V, prothrombin,MTHFR), homozygous wild-type and homozygous mu-tant samples were easily distinguished from each other by
a shift in T
m. However, the Tmdifference between ho-
mozygous genotypes for SNPs in class 3 or 4 was smallerthan in class 1 or 2. Homozygous HbS (17A /H11022T; class 4)
could be distinguished from the wild type with a T
m
difference of /H110110.2°C, but the HFE homozygous mutantFig. 2. Derivative melting curves for factor V Leiden genotyping ob-
tained on the LightCycler ( A) and the HR-1 high-resolution instrument
(B).
Three individuals of each genotype were analyzed: wild type ( solid black line ),
homozygous mutant ( dashed black line ), and heterozygous ( solid gray line ).Fig. 3. Normalized, high-resolution melting curves of all possible SNP
genotypes at one position using engineered plasmids.
Three samples of each genotype were analyzed and included four homozygotes
(A) and six heterozygotes ( B).Clinical Chemistry 50, No. 7, 2004 1159

(187C/H11022G; class 3) could not be distinguished from the
wild type.
Complete genotyping of HFE C187G by high-resolu-
tion melting analysis was possible by adding a knowngenotype to the unknown sample. Fig. 5A shows theresult of mixing wild-type amplicons with unknownhomozygous amplicons after PCR. If the unknown sam-ple is wild type, the melting curve does not change.However, if the unknown sample is homozygous mutant,heteroduplexes are produced, and an additional low-temperature transition appears. An alternative option isto add a known genotype to the unknown sample beforePCR. If a small amount of wild-type DNA is added,wild-type samples generate no heteroduplexes, homozy-gous mutant samples show some heteroduplexes, andheterozygous samples show the greatest amount of het-eroduplex formation (Fig. 5B). Table 2 shows that therewas complete concordance between the fluorescent hy-bridization probe (HybProbe
TM) and high-resolution am-
plicon melting methods for 167 samples.
Unexpected sequence alterations under primers or
very near the mutation of interest can adversely affectgenotyping assays. The melting curves for three rareheterozygous sequence changes near factor V Leiden areshown in Fig. 1 of the online Data Supplement. Althoughonly single samples of each rare genotype were available,each curve appeared to have a unique shape, suggestingthat genotype differentiation may be possible.
Shown in Fig. 6 is the in silico frequency distribution
(25) of the calculated /H9004T
ms among the homozygous
genotypes of SNPs, adjusted for the frequency of eachthree base combination centered on SNPs in the humangenome. All possible /H9004T
mcombinations (576) were calcu-
lated by varying the three bases centered on the naturalSNPs of the six amplicons studied here. Class 1 and 2SNPs form the broad cluster around 0.8 –1.4°C and are
easily distinguishable by melting analysis. Class 3 and 4SNPs include the minor peaks around 0.00 and 0.25 °C.
Although most class 3 and 4 SNPs can be fully genotypedby high-resolution melting analysis, one-fourth haveidentical predicted T
ms, and the homozygotes cannot be
differentiated without use of the addition methods de-scribed above.
Discussion
There are many ways to genotype SNPs (26). Available
techniques that require a separation step include restric-tion fragment length polymorphism analysis, single-nu-cleotide extension, oligonucleotide ligation, and sequenc-ing. Additional methods, including pyrosequencing (27)
and mass spectroscopy (28), are technically complex but
can be automated for high-throughput analysis. Somemethods require two analyses to obtain a genotype.
Homogeneous, closed-tube methods for SNP genotyp-
ing that do not require a separation step are attractive fortheir simplicity and containment of amplified products.Most of these methods are based on PCR and use fluo-rescent oligonucleotide probes. Genotyping occurs eitherby allele-specific fluorescence (29,30)or by melting anal-
ysis (31). Melting analysis has the advantage that multiple
alleles can be genotyped with one probe (32). Most of
these techniques can be performed after amplification iscomplete, although they are often associated with real-time PCR (33–37).
Some closed-tube fluorescent methods for SNP geno-
typing do not require probes. Allele-specific PCR can bemonitored in real time with SYBR Green I (5). The method
requires three primers, two PCR reactions for each SNP,and a real-time PCR instrument that can monitor eachcycle of PCR. An alternative method uses allele-specificamplification, SYBR Green I, and melting curve analysisat the end of PCR (6). Monitoring of each cycle is not
necessary, and a SNP genotype can be obtained in onereaction. However, a melting instrument and three prim-Table 1. SNP classification according to the homoduplexes and heteroduplexes produced after amplification of a
heterozygote and the predicted number of distinct nearest-neighbor thermodynamic duplexes ( Tms).a
Class SNP (frequency)bHomoduplex matches (no. of Tms) Heteroduplex mismatches (no. of Tms) Examples in Figs.
1 C/T or G/A C::G and A::T C::A and T::G 3B, 4A, 4B
(0.662) (2) (2 or 1)c
2 C/A or G/T C::G and A::T C::T and A::G 3B, 4C
(0.176) (2) (2 or 1)c
3 C/G C::G C::C and G::G 3B, 4D, 5
(0.088) (2 or 1)c(2)
4 T/A A::T T::T and A::A 3B, 4E
(0.074) (2 or 1)c(2)
aSNPs are specified with the alternative bases separated by a slash, e.g., C/T indicates that one DNA duplex ha s a C and the othe raTa tt h e same position on
the equivalent strand. There is no bias for one allele over the other, i.e., C/T is equivalent to T/C. Base pairing (whether matched or mismatched) is in dicated by a
double colon and is not directional, i.e., C::G indicates a C::G base pair without specifying which base is on which strand.
bThe human SNP frequencies were calculated from ftp://ftp.ncbi.nih.gov/snp/human dbSNP, build 120, March 18, 2004, based on 7 291 660 biallelic SNPs on
chromosomes 1–22, X, and Y having no immediate neighboring polymorphisms.
cThe number of predicted thermodynamic duplexes depends on the nearest-neighbor symmetry around the base change. One-fourth of the time, nearest-ne ighbor
symmetry is expected, i.e., the position of the base change will be flanked on each side by complementary bases. For example, if a C/G SNP is flanked by an A and
a T on the same strand (Fig. 1D), nearest-neighbor symmetry occurs and only one homoduplex Tmis expected (as observed in Fig. 4D).1160 Liew et al.: SNP Genotyping Using LCGreen I

ers are necessary, with one of the primers modified with
a GC clamp. Both techniques are based on allele-specificPCR, and each allele-specific primer is designed to recog-nize only one allele.
SNP genotyping by high-resolution melting with the
dye LCGreen I does not require probes, allele-specificPCR, or real-time PCR. Only two primers, one PCRreaction, and a melting instrument are required. Reagentcosts for genotyping by amplicon melting are low becauseonly PCR primers and a generic dye are needed. Noprobes or specialized reagents are required.
Although SNPs have been genotyped within ampli-
cons up to 544 bp long (1), use of a small amplicon for
genotyping has numerous advantages. Assay design issimplified because primers are selected as close to theSNP as possible. The T
mdifferences among genotypes
increases as the amplicon size decreases, allowing betterdifferentiation. Cycling times can be minimized becausethe melting temperatures of the amplicons (74 –81°Ci n
Figs. 3 and 4) allow low denaturation temperatures dur-ing cycling that in addition increase specificity. Further-
more, the amplicon length is so small that no temperatureholds are necessary for complete polymerase extension.Potential disadvantages of small amplicons include lessflexibility in the choice of primers, less effective contam-ination control with UNG (38), and difficulty distinguish-
ing between primer-dimers and desired amplificationproducts on gels or during real-time analysis.
Small amplicons allow rapid-cycle protocols that com-
plete PCR in 12 min with popular real-time (LightCycler)or inexpensive (RapidCycler II) instruments. Heterodu-plex detection in small amplicons is favored by rapidcooling before melting, rapid heating during melting, andlow Mg
2/H11001concentrations (7). Although conventional real-
time instruments can be used for melting (Fig. 2), theirresolution is limited. Small T
mdifferences between ho-
mozygotes (see Fig. 4E for an example) are not distin-guished on conventional instruments (data not shown).
The effects of ionic strength and product concentration
on amplicon T
mhave been discussed previously (7).Fig. 4. Normalized, high-resolution melting curves from factor V Leiden 1691G /H11022A (class 1; A); prothrombin 20210G /H11022A (class 1; B); MTHFR
1298A /H11022C (class 2; C); HFE 187C /H11022G (class 3; D); and /H9252-globin 17A /H11022T (class 4; E) SNPs.
Three individuals of each genotype were analyzed and are displayed for each SNP.Clinical Chemistry 50, No. 7, 2004 1161

Because homozygous SNP genotypes are distinguished
only by Tm, these issues are a concern for accurate
genotyping. However, the effect of amplicon concentra-tion on T
mis small and usually dwarfed by the tempera-
ture (in)accuracy of most melting instruments. Further-more, amplifying well into the plateau phase usuallyequalizes by PCR any differences in initial product con-centration. Because T
mis strongly dependent on the ionic
strength, it is important that all DNA samples are ex-tracted in the same way and end up in the same buffers.We did not find it necessary to quantify the DNA samplesbefore PCR unless addition studies were performed andnever attempted to quantify amplicon after PCR butbefore melting.
Dedicated melting instruments have recently become
available (LightTyper and HR-1). The HR-1 provides thehighest resolution and is the least expensive. Althoughonly one sample is analyzed at a time, the turnaroundtime is so fast (1 –2 min) that the throughput is reasonable.
The LightTyper is an interesting platform for high-throughput melting applications. However, the tempera-ture homogeneity across the plate needs to be improvedbefore homozygotes can be reliably distinguished (datanot shown).
Can all SNPs be genotyped by simple high-resolution
melting of small amplicons? Studies with engineeredplasmids of all possible base combinations at one locationinitially suggested that the answer was “yes”(Fig. 3).Heterozygotes were always easily identified. Whether the
different homozygotes were easy to distinguish dependedon the class of SNP (Table 1). The six possible binarycombinations of bases (C/T, G/A, C/A, G/T, C/G, andT/A) group naturally into four classes based on thehomoduplex and heteroduplex base pairings producedwhen a heterozygote is amplified. SNP homozygotes areeasy to distinguish by T
min the first two classes because
one homozygote contains an A::T pair and the other aG::C pair. These short amplicons show homozygote T
mTable 2. Genotype concordance based on adjacent
hybridization probes (HybProbe) and small-amplicon, high-
resolution melting analysis (Amplicon melting).
Marker Genotypes HybProbeaAmplicon
meltingb
Factor V 1691G /H11022A Wild type 35 35
Heterozygous 35 35Homozygous mutant 34 34
Prothrombin
20210G /H11022AWild type 8 8
Heterozygous 3 3Homozygous mutant 11 11
MTHFR 1298A /H11022C Wild type 6 6
Heterozygous 7 7Homozygous mutant 7 7
HFE 187C /H11022G Wild type 4 4
c
Heterozygous 4 4
Homozygous mutant 4 4c
/H9252-Globin 17A /H11022T Wild type 3 3
Heterozygous 3 3Homozygous mutant 3 3
aAll samples were originally genotyped by ARUP (factor V, prothrombin,
MTHFR, and HFE) or Pediatrix Screening ( /H9252-globin) as clinical samples with
adjacent hybridization probes and melting curve analysis.
bGenotyping results for the same samples using LCGreen I, the HR-1
high-resolution melting instrument, and amplicon melting.
cGenotyping required addition of homozygous DNA (see text).
Fig. 5. Genotyping at the HFE 187C /H11022G locus by adding wild-type DNA
to each sample.
(A), wild-type amplicons were mixed with amplicons from three individuals of each
homozygous genotype after PCR. ( B), 15% wild-type genomic DNA was added to
the DNA of three individuals of each genotype before PCR.
Fig. 6. In silico estimation of the Tmdifference among homozygous
genotypes of small amplicon SNPs.
The frequency distribution is adjusted for the relative occurrence of each SNP
and immediately adjacent bases in the human genome. The mean ampliconlength was 43.5 bp. The larger the /H9004T
m, the easier it is to differentiate the
homozygous genotypes. Approximately 4% of human SNPs have a predicted /H9004Tm
of 0.00 °C and are expected to require addition of known homozygous DNA for
genotyping of the homozygotes.1162 Liew et al.: SNP Genotyping Using LCGreen I

differences mostly between 0.8 –1.4°C, and these two
classes make up /H1102284% of human SNPs (39).
It is more difficult to distinguish the homozygotes of
SNPs in classes 3 and 4 (Table 1) because the base pair(A::T or C::G) is simply inverted, that is, the bases switchstrands but the base pair remains the same. Differences inamplicon T
mstill result from different nearest-neighbor
interactions with the bases next to the SNP site, but theyare usually /H110210.4°C (Fig. 3A and Fig. 4, D and E). Class 3
and 4 SNPs make up /H1101116% of human SNPs. Genotyping
of homozygotes is still possible in most cases with high-resolution analysis.
Clinical SNPs of each class were selected for concor-
dance studies with standard genotyping methods. FactorV (Leiden) 1691G /H11022A and prothrombin 20210G /H11022A were
class 1 SNPs, MTHFR 1298A /H11022C was class 2, HFE 187C /H11022G
was class 3, and
/H9252-globin (HbS) 17A /H11022T was class 4.
The class 3 SNP studied (Fig. 4D) was unique in that
we could not differentiate the different homozygotes byT
musing simple melting analysis. Inspection of the bases
neighboring the SNP site revealed why (Fig. 1D). In thiscase, the neighboring bases were complementary, givingnearest-neighbor stability calculations that were identicalfor the two homozygotes. To the extent that nearest-neighbor theory is correct, the duplex stabilities are pre-dicted to be the same. By chance alone, this nearest-neighbor “symmetry ”is expected to occur 25% of the
time. When this occurs in class 1 or 2 SNPs, nearest-neighbor calculations indicate that the stabilities of thetwo heteroduplexes formed are identical. This is not ofconsequence to SNP typing because all three SNP geno-types still have unique melting curves. However, nearest-neighbor symmetry in class 3 or 4 SNPs predicts that thetwo homoduplex T
ms (homozygous genotypes) are iden-
tical. This will occur in /H110114% of human SNPs.
When nearest-neighbor symmetry of class 3 or 4 SNPs
predicts that the homozygotes will not be distinguished,complete genotyping is still possible by adding to thereactions a known genotype either before or after PCR. Ifthe amplicon is added after PCR, only the homozygotesneed to be tested, but potential amplicon contamination isa disadvantage. Adding before PCR requires either thatthe DNA concentration of the sample is carefully con-trolled or that samples are run both with and withoutadded DNA.
Unexpected sequence variants can adversely affect
many different genotyping assays, including the ampli-con melting assay described here. Variants under a primercan cause allele-specific amplification of heterozygotes,giving apparent homozygous genotypes. Variants veryclose to the targeted mutation can confound assays basedon restriction enzyme digestion or hybridization probes.However, different sequence variants under probes (40)
or within amplicons (this study) can often be differenti-ated by melting analysis.
High-resolution amplicon melting with LCGreen I can
also be used to scan for sequence differences between twocopies of DNA (1). In mutation scanning, the method is
similar to other heteroduplex techniques, such as dena-turing HPLC (41) or temperature gradient capillary elec-
trophoresis (42). However, high-resolution melting is
unique in that homozygous sequence changes can oftenbe identified without the use of added DNA. In the case ofSNP genotyping with small amplicons, addition is rarelyrequired.
We thank Jamie Williams for technical assistance, Melissa
Seipp for help with the figures, and Noriko Kusukawa forreviewing the manuscript. This work was supported bygrants from the University of Utah Research Foundation,the State of Utah Center of Excellence program, ARUP,and Idaho Technology (Salt Lake City, UT).
References
1.Wittwer CT, Reed GH, Gundry CN, Vandersteen JG, Pryor RJ.
High-resolution genotyping by amplicon melting analysis usingLCGreen. Clin Chem 2003;49:853 –60.
2.Lay MJ, Wittwer CT. Real-time fluorescence genotyping of factor V
Leiden during rapid-cycle PCR. Clin Chem 1997;43:2262 –7.
3.Livak KJ, Flood SJ, Marmaro J, Giusti W, Deetz K. Oligonucleotides
with fluorescent dyes at opposite ends provide a quenched probesystem useful for detecting PCR product and nucleic acid hybrid-ization. PCR Methods Appl 1995;4:357 –62.
4.Crockett AO, Wittwer CT. Fluorescein-labeled oligonucleotides for
real-time pcr: using the inherent quenching of deoxyguanosinenucleotides. Anal Biochem 2001;290:89 –97.
5.Germer S, Holland MJ, Higuchi R. High-throughput SNP allele-
frequency determination in pooled DNA samples by kinetic PCR.Genome Res 2000;10:258 –66.
6.Germer S, Higuchi R. Single-tube genotyping without oligonucleo-
tide probes. Genome Res 1999;9:72 –8.
7.Gundry CN, Vandersteen JG, Reed GH, Pryor RJ, Chen J, Wittwer
CT. Amplicon melting analysis with labeled primers: a closed-tubemethod for differentiating homozygotes and heterozygotes. ClinChem 2003;49:396 –406.
8.Wittwer CT, Garling DJ. Rapid cycle DNA amplification: time and
temperature optimization. Biotechniques 1991;10:76 –83.
9.Poort SR, Rosendaal FR, Reitsma PH, Bertina RM. A common
genetic variation in the 3 /H11032-untranslated region of the prothrombin
gene is associated with elevated plasma prothrombin levels andan increase in venous thrombosis. Blood 1996;88:3698 –703.
10.Weisberg I, Tran P, Christensen B, Sibani S, Rozen R. A second
genetic polymorphism in methylenetetrahydrofolate reductase(MTHFR) associated with decreased enzyme activity. Mol GenetMetab 1998;64:169 –72.
11.Bernard PS, Ajioka RS, Kushner JP, Wittwer CT. Homogeneous
multiplex genotyping of hemochromatosis mutations with fluores-cent hybridization probes. Am J Pathol 1998;153:1055 –61.
12.Herrmann MG, Dobrowolski SF, Wittwer CT. Rapid
/H9252-globin geno-
typing by multiplexing probe melting temperature and color. ClinChem 2000;46:425 –8.
13.Heath EM, O ’Brien DP, Banas R, Naylor EW, Dobrowolski S.
Optimization of an automated DNA purification protocol for neo-natal screening. Arch Pathol Lab Med 1999;123:1154 –60.
14.Highsmith WE Jr, Jin Q, Nataraj AJ, O ’Connor JM, Burland VD,
Baubonis WR, et al. Use of a DNA toolbox for the characterizationof mutation scanning methods. I: construction of the toolbox andClinical Chemistry 50, No. 7, 2004 1163

evaluation of heteroduplex analysis. Electrophoresis 1999;20:
1186 –94.
15.Allawi HT, SantaLucia J Jr. Nearest neighbor thermodynamic
parameters for internal G.A mismatches in DNA. Biochemistry1998;37:2170 –9.
16.Allawi HT, SantaLucia J Jr. Thermodynamics of internal C.T
mismatches in DNA. Nucleic Acids Res 1998;26:2694 –701.
17.Allawi HT, SantaLucia J Jr. Nearest-neighbor thermodynamics of
internal A.C mismatches in DNA: sequence dependence and pHeffects. Biochemistry 1998;37:9435 –44.
18.Allawi HT, SantaLucia J Jr. Thermodynamics and NMR of internal
G.T mismatches in DNA. Biochemistry 1997;36:10581 –94.
19.Bommarito S, Peyret N, SantaLucia J Jr. Thermodynamic parame-
ters for DNA sequences with dangling ends. Nucleic Acids Res2000;28:1929 –34.
20.Peyret N, Seneviratne PA, Allawi HT, SantaLucia J Jr. Nearest-
neighbor thermodynamics and NMR of DNA sequences withinternal A.A, C.C, G.G, and T.T mismatches. Biochemistry 1999;38:3468 –77.
21.SantaLucia J Jr, Allawi HT, Seneviratne PA. Improved nearest-
neighbor parameters for predicting DNA duplex stability. Biochem-istry 1996;35:3555 –62.
22.SantaLucia J Jr. A unified view of polymer, dumbbell, and oligonu-
cleotide DNA nearest-neighbor thermodynamics. Proc Natl AcadSci U S A 1998;95:1460 –5.
23.von Ahsen N, Wittwer CT, Schutz E. Oligonucleotide melting
temperatures under PCR conditions: nearest-neighbor correctionsfor Mg
2/H11001, deoxynucleotide triphosphate, and dimethyl sulfoxide
concentrations with comparison to alternative empirical formulas.Clin Chem 2001;47:1956 –61.
24.Zhao Z, Boerwinkle E. Neighboring-nucleotide effects on single
nucleotide polymorphisms: a study of 2.6 million polymorphismsacross the human genome. Genome Res 2002;12:1679 –86.
25.von Ahsen N, Oellerich M, Armstrong VW, Schutz E. Application of
a thermodynamic nearest-neighbor model to estimate nucleic acidstability and optimize probe design: prediction of melting points ofmultiple mutations of apolipoprotein B-3500 and factor V with ahybridization probe genotyping assay on the LightCycler. ClinChem 1999;45:2094 –101.
26.Kwok PY, Chen X. Detection of single nucleotide polymorphisms.
Curr Issues Mol Biol 2003;5:43 –60.
27.Ronaghi M. Pyrosequencing for SNP genotyping. Methods Mol Biol
2003;212:189 –95.
28.Sauer S, Gut IG. Genotyping single-nucleotide polymorphisms by
matrix-assisted laser-desorption/ionization time-of-flight massspectrometry. J Chromatogr B Analyt Technol Biomed Life Sci
2002;782:73 –87.
29.Lee LG, Connell CR, Bloch W. Allelic discrimination by nick-
translation PCR with fluorogenic probes. Nucleic Acids Res 1993;21:3761 –6.
30.Mhlanga MM, Malmberg L. Using molecular beacons to detect
single-nucleotide polymorphisms with real-time PCR. Methods2001;25:463 –71.
31.Ronai Z, Sasvari-Szekely M, Guttman A. Miniaturized SNP detec-
tion: quasi-solid-phase RFLP analysis. Biotechniques 2003;34:1172 –3.
32.Wittwer CT, Herrmann MG, Gundry CN, Elenitoba-Johnson KS.
Real-time multiplex PCR assays. Methods 2001;25:430 –42.
33.Wittwer C, Kusukawa N. Real-time PCR. In: Persing D, Tenover F,
Relman D, White T, Tang Y, Versalovic J, Unger B, eds. Diagnosticmolecular microbiology; principles and applications. Washington:ASM Press, 2003:71 –84.
34.Meuer S, Wittwer C, Nakaguwara K, eds. Rapid cycle real-time
PCR: methods and applications. Berlin: Springer-Verlag, 2001:408pp.
35.Reischl U, Wittwer C, Cockerill F, eds. Rapid cycle real-time PCR:
methods and applications —microbiology and food analysis. Ber-
lin: Springer-Verlag, 2002:258pp.
36.Dietmaier W, Wittwer C, Sivasubramanian N, eds. Rapid cycle
real-time PCR: methods and applications —genetics and oncol-
ogy. Berlin: Springer-Verlag, 2002:180pp.
37.Wittwer C, Hahn M, Kaul K, eds. Rapid cycle real-time PCR:
methods and applications —quantification. Berlin: Springer-Ver-
lag, 2004:223pp.
38.Espy MJ, Smith TF, Persing DH. Dependence of polymerase chain
reaction product inactivation protocols on amplicon length andsequence composition. J Clin Microbiol 1993;31:2361 –5.
39.Venter JC, Adams MD, Myers EW, Li PW, Mural RJ, Sutton GG, et
al. The sequence of the human genome. Science 2001;291:1304 –51.
40.Lyon E, Millson A, Phan T, Wittwer C. Detection and identification
of base alterations within the region of factor V Leiden byfluorescent melting curves. Mol Diagn 1998;3:203 –9.
41.Wolford JK, Blunt D, Ballecer C, Prochazka M. High-throughput
SNP detection by using DNA pooling and denaturing high perfor-mance liquid chromatography (DHPLC). Hum Genet 2000;107:483–7.
42.Li Q, Liu Z, Monroe H, Culiat CT. Integrated platform for detection
of DNA sequence variants using capillary array electrophoresis.Electrophoresis 2002;23:1499 –511.1164 Liew et al.: SNP Genotyping Using LCGreen I