| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Molecular Diagnostics and Genetics |
1 ARUP Institute for Clinical and Experimental Pathology, Salt Lake City, Utah, USA;2 Department of Pathology, University of Utah, Salt Lake City, Utah, USA;3 Institute of Veterinary Medicine, Georg-August University, Göttingen, Germany.
aAddress correspondence to this author at: Dr. Pont-Kingdon, ARUP Institute for Clinical and Experimental Pathology, ARUP Laboratories. 500 Chipeta Way, Salt Lake City, UT 84108. E-mail pontkig{at}aruplab.com.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: We used software to predict hybridization structures and Tms from 10 noncontinuous probes and 54 different templates. Predicted Tms were compared to existing experimental data. The bulging template’s sequences (omitted in the probe) ranged in size from 1 to 73 nucleotides. In 36 cases, we compared observed and predicted 
Tms between alleles complementary to the probe and mismatched alleles. In addition, using software that predicts effects of bulges, we designed a probe and then tested it experimentally.
Results: The mean differences between predicted and observed Tms were 0.65 (2.51) °C with the Visual OMP software and 0.28 (1.67) °C with the MeltCalc software. Tms were within a mean (SD) of 0.36 (1.23) °C (Visual OMP) and –0.01 (1.02) °C (MeltCalc) of observed values. An increase in the size of the template bulge resulted in a decrease in Tms. In 2 templates, the presence of a variant in the bulge influenced the experimental Tm of 2 noncontinuous probes, a result that was not predicted by the software programs.
Conclusions: The use of software prediction should prove useful for the design of noncontinuous probes that can be used as tools for molecular haplotyping, multiplex genotyping, or masking sequence variants.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Probes that do not bind continuously on the template DNA have been described recently. Probes that span several loci [loci-spanning probes (LSProbes)1 ] allow molecular haplotyping(1) and multiplex genotyping in single step melting experiments(2). In LSProbes, the sequence of the template between the analyzed loci is partially omitted. On binding of the probe, the nonhybridized template sequence between the analyzed loci creates a bulge. By hybridizing simultaneously on several loci, LSProbes can analyze several variants separated by up to 80 nucleotides (nt). As with continuous probes, mismatches in the hybridized region of the LSProbe lowers the melting temperature (Tm) (temperature at which 50% of duplex strands are separated). This characteristic enables the identification of different variants and haplotypes.
To simplify melting interpretation of other sequence changes detected by the probe, masking probes incorporate a neutralizing mismatch at the position of nontargeted sequence variation(3). The masking probes in this report were designed with a deletion of the nontargeted sequence-variation position from the probe sequence. This design creates a template bulge between all possible alleles when hybridized to the masking deletion probe. Wild-type and masked variant alleles will have only the template bulge and a similar Tm. In contrast, any other sequence variation elsewhere under the probe will have an additional mismatch with the probe and a lower Tm than the wild-type and masked variant alleles.
The thermodynamic properties of nucleic acid hybridization are well studied(4)(5)(6)(7)(8)(9)(10)(11)(12)(13), and software prediction of probe stability is useful in the design of molecular assays(14)(15)(16). Several software programs are available to predict the Tms of continuous probes and primers, the most accurate using the thermodynamic nearest-neighbor model(5)(7)(12). The effect of single and multiple mismatches on Tm are well documented and predicted by several software programs(4)(5)(6)(7)(11). Only a few, however, allow for the presence of bulges in either the template or the probe. We have evaluated the Visual OMP (oligonucleotide modeling platform) and the MeltCalc(14) software programs for their prediction of binding of noncontinuous probes on templates with and without mismatches. In addition, prediction of structure between templates and probes, effect of mismatches, size of the template bulge, and presence of variants in the bulge were examined.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
) used in this study were designed without the aid of software that can incorporate into either template or probe the thermodynamic instability due to bulges. The noncontinuous probes with omission of 1–3 nt of the template sequence were designed to mask nontargeted sequence variants such as benign polymorphic sites in RET protooncogene and the fibroblast growth factor receptor 3 gene (FGFR3)2
(3). The probes with longer template omissions of 13–73 nt were designed for multiplex genotyping of the beta hemoglobin gene (HBB)(2) or haplotyping of the beta 2 adrenergic receptor (ADRB2) or of the WIAF 1537–1538 markers on chromosome 21(1)(17)]. We used the MeltCalc software that incorporates the thermodynamic effect of bulges on Tm predictions to design 1 probe with a 40-nt omission in the bovine MC1R melanocortin 1 receptor gene (MC1R), which we then tested experimentally.
|
In addition to probes published previously, a masking deletion probe was designed for the detection of the c.1948A>G mutation in the FGFR3 gene, which is responsible for more than 99% of cases of thanathophoric dysplasia type II. The probe was a fluorescence resonance energy transfer (FRET) hybridization probe in which the site c.1953 has been omitted (Table 1
) to avoid detection of an A>G benign polymorphism present in 4% of the general population. With this probe, the c.1953 A>G polymorphism is undetectable, and the masked polymorphic allele (G) has an identical Tm to the wild-type allele (A) of 47.5 °C, wereas the c.1948G disease-causing allele has a Tm of 56.6 °C (data not shown).
experimental tMS and tMS
Experimental Tms were all obtained by analysis of changes of a fluorescence signal during probe melting. Both adjacent hybridization probes detected by FRET and unlabeled probes detected by double-strand DNA–specific dyes(1)(2)(3)(18) have been used. Briefly, all PCR experiments were performed using a LightCycler instrument, and melting curves of hybridization probes were analyzed with the LightCycler data analysis software version 3.5 (Roche Diagnostics). Tms of unlabeled probes were analyzed by high-resolution melting in an HR1 instrument (Idaho Technology) in the presence of the double-strand–specific dye LCGreen®Plus+ (Idaho Technology).
For each set of probes and templates, Tms were calculated by subtracting the lowest allele Tm from the highest allele Tm.
software analysis of noncontinuous probes
Visual OMP (DNA Software; http://www.DNAsofware.com) simulates hybridization of nucleic acids in solution and provides structural and thermodynamic parameters in intra- or intermolecular situations. We used version 6.0 of the software. Data entry included the sequences of the probe and the templates (complete PCR fragment) or the area of the templates that hybridized with the probe (see Fig. 2
of(2)). Structure settings in the program were adjusted to allow formation of bulges up to 80 nt. After simulation, the optimal structure of the probe hybridized to the template was selected as the predicted structure and Tm value.
|
MeltCalc is an Excel-based worksheet (http://www.meltcalc.com) for thermodynamic calculation and automatic design of hybridization probes. Version 2.95 was used for this study, upgraded from the commercially available 2.0 for introduction of bulges in the template. We entered the sequence of the probe, the "dangling ends" (nts corresponding to the nts found on the template, both in 5' and 3' of the probe), the length of the potential template bulge, and the location of the bulge in the probe.
In both programs, the salt conditions (214 mmol/L Na+) and oligonucleotide concentrations (0.2 µmol/L for adjacent hybridization probes and 0.5 µmol/L for unlabeled probes) corresponded to the conditions used in our experiments(19). In MeltCalc, a correction for double-strand dye (SYBR-Green I) was used with unlabeled probes.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Fig. 1A
shows the correlation between observed and calculated Tms and Bland-Altman plots(20). Whereas the correlations do not significantly differ, the 95% CI ranges of the predicted values were somewhat different (Visual OMP, ±4.8 °C vs MeltCalc, ±3.3 °C). With both software programs, there is a slight trend toward underestimation of higher Tms; this trend is more pronounced with Visual OMP, mainly owing to the noncontinuous probes that form small template bulges (
3 nt).
|
In 36 experiments, 9 of the probes were hybridized with templates of different genotypes or haplotypes, and Tms were analyzed. As shown in Fig. 1B
, there was no significant deviation from the line of identity within the given 95% CI of the correlation. The Tms are predicted with better precision (95% CI of predicted value: Visual OMP, ±2.6 °C vs MeltCalc, ±2.1 °C) than the Tm values.
To evaluate if the size of the DNA bulge affected the Tm and Tms predictions, the calculated minus observed values were plotted against the bulge size (Fig. 1C
and 1D
). Table 2
provides an overview of the calculations and shows that the average differences from both programs are similar for both Tm and Tms predictions. A more detailed evaluation of the Tm data (Fig. 1C
) shows a significant trend toward underestimation of the destabilizing effect of bulges with Visual OMP, with a dependency of prediction on the bulge size that reaches statistical significance (significant deviation from slope = 1). If Tms are considered (Fig. 1D
), the underestimation of the destabilization by larger bulges becomes substantially smaller, and both programs show almost identical predictions over the investigated range.
|
We tested the use of the MeltCalc software by designing a noncontinuous probe to analyze the linkage between 2 single-nucleotide polymorphisms separated by 53 nt in the promoter region of the bovine MC1R gene(21). The probe must reliably allow differentiation of the 2 linked haplotypes (WT and UU) from the 2 other possible haplotypes (nonlinkage types, U1 and U2). We selected the antisense wild-type probe that predicted a Tm of 12.2 °C (see Supplemental Data Table 1
in the Data Supplement that accompanies the online version of this report at http://www.clinchem.org/content/vol54/issue6) between the 2 common haplotypes WT and UU. After experimentation, the observed Tm between the haplotypes was 9.6 °C (SD 0.7 °C), which is within the error range calculated above.
detection of sequence variants in the bulge
As noted above, both Visual OMP and MeltCalc can calculate the predicted Tms and Tms for noncontinuous probes. We used Visual OMP to predict the effect of differences in nucleotide sequences in the bulge. Bulged template sequences are not used in MeltCalc, and therefore bulge analysis could not be performed with this software. With the use of 3 noncontinuous probes, 46/79-GG in the ADRB2 receptor and HbCmatch and HbEmatch in the HBB gene (Table 1
), on templates with a variant within the template bulge, we observed Tm shifts compared to WT templates (Fig. 2
). The 46/79-GG LSProbe was designed to haplotype the 2 variants c.46A>G and c.79C>G of the ADRB2 gene by binding to both loci and creating a 22-nt bulge of template sequence(1). Fig. 2A
shows 5 samples heterozygous for the c.46G/79G (GG) and c.46A/C79C (AC) haplotypes. In 1 sample (black trace), a Tm shift of 2 °C (50.2 vs 48.2 °C) from wild-type samples can be observed. Sequencing showed that this sample had a c.66C>T variant in the 22-nt template bulge (AtC). The second example used 2 LSProbes that simultaneously genotyped 3 loci of the HBB gene and clearly identified the 4 possible HBB alleles (HbC, HbE, HbS, and WT) (Fig. 2B
and (2). One probe matched to the HbC allele and the other to the HbE allele. Forty nine nucleotides of template sequence are omitted between the HbS/C and the HbE loci. During validation of the assay, we tested 6 samples heterozygous for diverse variants in the 49-nt template bulge. The variants were not detected in 5 of the samples that appeared as wild type, but in one sample a Tm shift from WT was observed. The presence of a Coushatta variant (c.68A>C) on 1 chromosome, confirmed by sequencing, allowed a slight increase of Tm with both the probe matched to the HbC allele and the probe matched to the HbE allele (respectively 0.85 and 0.66 °C). Analysis of the Tm of these probes with the different templates using Visual OMP did not reveal these Tm variations. Our observation of differences in stability of the probes due to template sequence variants in the bulge indicate that in some cases, nontargeted polymorphisms can also be detected, increasing the analysis power of noncontinuous probes.
structure prediction between probe and template in the design of noncontinuous probes
We explored the usefulness of using structure prediction provided by software in the design of noncontinuous probes used for multiplex genotyping and/or haplotyping. Both Visual OMP and MeltCalc allow prediction of the structure created by a noncontinuous probe annealed on its template. Visual OMP has the added feature of providing visual representation of the hybridization of the probe with its template.
As shown above, the HBB-HbE matched probe allows genotyping of the 4 HBB alleles HbE, HbS, HbC, and WT. The optimal structure predicted by Visual OMP between the probe and the template shows the probe hybridized on each side of the 49-nt template bulge (Fig. 3A
). Fig. 3B
shows data obtained with a probe that was shorten and had only 8 nt homologous to the template region containing the HbE allele. This probe does not distinguish the HbE genotype from WT, whereas the HbC and HbS alleles are clearly distinguished from WT and from each other. Therefore, this "short" probe allowed genotyping of the HbS and HbC alleles, localized 1 nt from each other, but did not allow simultaneous genotyping if the HbE locus that is 59 nt from the HbS locus. The most stable structure predicted for the short probe had a dangling end that does not bind on the HbE locus. A visual representation using Visual OMP is shown in Fig. 3
. Hybridization of the template to only the 3' region of the short probe is more stable than a complete probe hybridization, owing to the relatively higher destabilizing effect of the bulge in this short probe compared to longer probes. Both Visual OMP and MeltCalc predicted this occurrence. This prediction was confirmed by the similar Tms observed for the HbE allele and the WT allele. Both software programs predicted that the short probe would fail to genotype the 3 loci simultaneously, and that additional nucleotides were needed to hybridize with the HbE locus to allow multiplex genotyping.
|
In the second example, we assessed the structure prediction of both programs in the design of more complex haplotyping probes. We have previously described a noncontinuous probe that spans 3 loci of the ADRB2 gene and determines the 3 main haplotypes of the gene(22). These haplotypes are a combination of variants at the 3 positions c.1–20 C>T, c.46 G>A, and c.79 G>C. The haplotypes are c.1–20C/46G/79G (CGG), c.1–20T/46G/79C (TGC), or c.1–20T/46A/79C (TAC). The probe was designed to anneal with the 3 loci and have 2 intervening sequences of 22 and 55 nt(1). The structure predicted by Visual OMP for the probe annealed to a perfectly matched template (CGG) is shown on the left of Fig. 4
. The probe had 2 mismatches with TGC (not shown) and 3 mismatches with TAC (Fig. 4
right). Tms observed with the 3 templates allow clear distinction of the 3 haplotypes (Fig. 4
, bottom). Both Visual OMP and MeltCalc predicted the probes to dangle and not anneal on the first mismatched nucleotide (C:A), as illustrated by the Visual OMP predicted structure with a TAC template that has a C:A mismatch, a G:T mismatch, and a G:G mismatch with the probe. The TGC template that does not have the G:T mismatch has an intermediate Tm. This example shows that even if the probe design does not allow complete hybridization of the noncontinuous probe with all the haplotypes, the differences in probe hybridization stability can still allow unambiguous haplotyping, as predicted by both Visual OMP and MeltCalc.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In the design of any probe system (e.g., FRET hybridization probes, molecular beacons, Eclipse probes, unlabeled probes) maximizing the effect of mismatch on probe stability is important and is accomplished by varying the length of the probe, by choosing the strand of DNA exhibiting the most destabilizing mismatch, by designing short probes with a minor groove binder moiety, or by incorporating nonconventional nucleotides such as locked nucleic acids(30). Tm prediction software is available for continuous probe design, but these programs cannot be used for the design of noncontinuous probes.
By comparing experimental Tm data from several noncontinuous probes to Tms predicted by software programs capable of incorporating the effect of bulges as well as mismatches, we tested the usefulness of software analysis in the design of noncontinuous probes. Also, because noncontinuous oligonucleotides could be used as PCR primers to avoid polymorphic loci, Tm prediction will help optimization of PCR using noncontinuous primers.
The noncontinuous probes ranged from 20–35 nt, which is in the range observed with continuous probes. The omitted template sequences were positioned in a central area of the probe. We found that in general, 11 nt of probe sequence on each side of the template bulge was sufficient for the probes to span several hybridization regions of the template. This number of nucleotides strictly depends on the nearest-neighbor composition of the duplex-forming area of the probe. In comparing the observed and the predicted Tms with software that uses the thermodynamic NN-model, we showed that a noncontinuous probe Tm prediction is possible with an average error of <1 °C. This is possible because the 2-state melting model can describe the underlying biophysical mechanism. Additionally, the software can predict the Tm difference between loci matched and mismatched with the probes with an average error of <0.5 °C. As expected, the error range is narrower with Tms than Tms because Tms values are relative and depend less on experimental conditions, concentration of probes, and PCR product and bulge effects. For haplotyping and genotyping applications of noncontinuous probes, the Tms rather than the absolute Tms determine how many different haplotypes and genotypes are distinguishable by the probe.
Structure prediction is also very useful in the design of noncontinuous probes. As demonstrated in Fig. 3
, data obtained from a probe that will not allow hybridization across the different regions indicates the need for redesign of the probe (i.e., increase the nucleotide length on the dangling end of the probe). On the other hand, a combination of different mismatches and different structures might actually contribute to the differentiation of several haplotypes. The binding of the probe that spans 3 loci, and creates 2 bulges in the template sequence, is different on a matched template (CGG) than on templates with 2 or 3 mismatches (TGC and TAC) (Fig. 4
). The presence of both mismatches and structure changes must contribute to the Tm difference between the CGG and the TGC template, whereas the Tm difference between TGC and TAC is due only to additional mismatches between TAC and the probe, because the structures are the same.
Limitations of LSProbe design for haplotyping include the number of haplotypes that can be distinguished and the maximum number of loci tested. It seems reasonable to incorporate 2 bulges for haplotyping of a 3-loci block, as shown with the ADRB2 20/46/79 probe, because only 4 haplotypes have been described at this locus. In theory, a variant at 3 loci can lead to up to 8 haplotypes, and the probe might not have the ability to distinguish all 8 haplotypes, although use of high-resolution instruments could increase the number of distinguishable haplotypes. For diagnostic purposes, it is usually not necessary to distinguish all possible haplotypes, because some will not naturally occur.
In the majority of the samples that had a known polymorphism in the bulge, the polymorphism did not alter the Tm of the probe, as expected. On the other hand, in the 2 cases in which we observed Tm variations, slight changes in secondary structures inside the bulge could have affected the probe Tm. No secondary structures were predicted in the bulge, and Visual OMP did not predict the effect of the sequence variant. As with continuous hybridization probes or unlabeled probes, samples with Tms that differ significantly from the expected Tm (>2SD of the within run precision) should be sequenced for verification.
From our observations, the following guidelines can be proposed for the design of successful noncontinuous probes used for haplotyping and multiplex genotyping: (a) like with continuous probes, the probe should be as short as possible to maximize Tm and the number of alleles they can distinguish; (b) the occurrence of dangling ends should be avoided for the probe to interrogate several loci simultaneously; and (c) the position of the mismatch should be close to the middle of the hybridized stems.
Software prediction that includes the effect of bulges in addition of the effect of mismatches should be very useful in the design of noncontinuous probes, specifically when large (>3 nt) bulges are expected and when multiple loci, genotypes or haplotypes are interrogated.
|
Acknowledgments |
|---|
Financial Disclosures: Dr. E. Schütz holds copyrights for the MeltCalc software.
Acknowledgments: We wish to thank Astrid Thuid for help with the Visual OMP software and C. Wittwer for critical reading of the manuscript.
|
Footnotes |
|---|
2 Genes: FGFR3, fibroblast growth factor receptor 3; HBB, hemoglobin, beta; ADRB2, adrenergic, beta-2-, receptor; MC1R, bovine MC1R melanocortin 1 receptor.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  E. Lyon and C. T. Wittwer LightCycler Technology in Molecular Diagnostics J. Mol. Diagn., March 1, 2009; 11(2): 93 - 101. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
