BACKGROUND: DNA copy number variation is associated with genetic disorders and cancer. Available methods to discern variation in copy number are typically costly, slow, require specialized equipment, and/or lack precision.
METHODS: Multiplex PCR with different primer pairs and limiting deoxynucleotide triphosphates (dNTPs) (3–12 μmol/L) were used for relative quantification and copy number assessment. Small PCR products (50–121 bp) were designed with 1 melting domain, well-separated Tms, minimal internal sequence variation, and no common homologs. PCR products were displayed as melting curves on derivative plots and normalized to the reference peak. Different copy numbers of each target clustered together and were grouped by unbiased hierarchical clustering.
RESULTS: Duplex PCR of a reference gene and a target gene was used to detect copy number variation in chromosomes X, Y, 13, 18, 21, epidermal growth factor receptor (EGFR), survival of motor neuron 1, telomeric (SMN1), and survival of motor neuron 2, centromeric (SMN2). Triplex PCR was used for X and Y and CFTR exons 2 and 3. Blinded studies of 50 potential trisomic samples (13, 18, 21, or normal) and 50 samples with potential sex chromosome abnormalities were concordant to karyotyping, except for 2 samples that were originally mosaics that displayed a single karyotype after growth. Large cystic fibrosis transmembrane conductance regulator (ATP-binding cassette sub-family C, member 7) (CFTR) deletions, EGFR amplifications, and SMN1 and SMN2 copy number assessments were also demonstrated. Under ideal conditions, copy number changes of 1.11-fold or lower could be discerned with CVs of about 1%.
CONCLUSIONS: Relative quantification by restricting the dNTP concentration with melting curve display is a simple and precise way to assess targeted copy number variation.
Copy number variations are common, involving about 13% of the human genome (1), and some are associated with human disease. Many genetic diseases can be caused by loss or gain of large DNA segments. For example, 1%–3% of cystic fibrosis mutations are large deletions of the cystic fibrosis transmembrane conductance regulator (ATP-binding cassette sub-family C, member 7) (CFTR)4 (2), as are similar percentages of mutations of breast cancer 1, early onset (BRCA1) and breast cancer 2, early onset (BRCA2) (3). Isolated trisomy in chromosomes 13, 18, and 21 can produce viable human fetuses but frequently result in spontaneous miscarriage, while triploidy results in miscarriage. Susceptibility to HIV infection is associated with an increase in chemokine (C-C motif) ligand 3-like 1 (CCL3L1) copy number (4). Higher copy numbers of epidermal growth factor receptor (EGFR) are commonly found in colon cancer and non–small cell lung cancer (5).
Methods for targeted copy number analysis include fluorescent in situ hybridization, multiplex ligation-dependent probe amplification, digital PCR, and real-time PCR. These techniques require multiple probes and specialized instrumentation and/or have limited resolution. Whole genome methods include array comparative genomic hybridization (6), single-nucleotide polymorphism (SNP)5 arrays (7), and massively parallel sequencing (8). They are most efficient as screening or discovery tools for copy number variation and carry higher costs than targeted methods.
High-resolution melting for genotyping or scanning is simple, fast, accurate, and inexpensive (9, 10). However, large deletions or duplications that encompass the primers are not identified unless a deletion is X linked (11). Competitive PCR typically amplifies both a target and a competitor with the same primer set to retain their quantitative relationship (12). Duplex melting with a single primer set that amplifies segmental duplications can quantify trisomies (13), and a similar method using homologous sequences can identify microdeletions or microinsertions (14). The reference and target PCR products are on different chromosomes with distinct melting temperatures (Tms) so they can be compared. These methods are quantitative because identical primers ensure equal efficiencies of the duplexed products, but do require the fortuitous presence of homologous sequences to identify a common primer pair. In addition, the internal sequence differences between reference and target must result in Tms that can be distinguished.
Relative quantification with different PCR primer pairs is also possible if the number of PCR cycles is limited. For example, preamplification methods maintain relative quantification by restricting the number of cycles to 10–15 (15). Quantitative multiplex PCR of short fluorescent fragments uses low cycle number to maintain relative quantification, followed by separation on sequencing gels (16). Comparative high-resolution melting identifies copy number changes by melting at a low cycle number, followed by mutation detection by secondary melting at a high cycle number (17). When PCR is limited by cycle number for relative quantification, the number of cycles must be high enough for adequate signal, yet low enough to maintain the quantitative relationship. There is no way to know the bounds of this interval without experimentation, because it depends on instrument sensitivity, starting DNA input, reaction efficiency, cycling conditions, and reactant concentrations (18). We have found that restricting the deoxynucleotide triphosphate (dNTP) concentration is a more convenient and robust way to automatically limit PCR for precise relative quantification.
DNA from human cell lines with 1 (NA11472), 2 (NA18800), 3 (NA03623), 4 (NA11226), and 5 (NA06061) copies of chromosome X, a heterozygous deletion in CFTR (exons 2 and 3, NA18668), an affected patient with spinal muscular atrophy (SMA) (NA00232), and an SMA carrier (NA003814) were obtained from the Coriell Institute for Medical Research. One hundred clinical products of conception were sent to ARUP Laboratories for karyotyping. Fetal villi were macrodissected away from maternal tissue and metaphase karyotypes obtained after growth. DNA was extracted from residual fetal tissue by lysis, salt precipitation, and washing with ethanol (5 PRIME™ ArchivePure™ DNA purification cell and tissue kit, Fisher Scientific). Fifty of the DNA samples were selected to enrich for trisomy of chromosomes 13, 18, or 21. Another 50 were enriched for sex chromosome abnormalities (Y = 0, 1, or 2; X = 1, 2, or 3). DNA samples from the 2 groups were then deidentified, blinded, and provided to the University of Utah under IRB #7275. On 2 samples with discrepant results, an SNP copy number array was performed (genome-wide human SNP array 6.0, Affymetrix) (19). ARUP also provided deidentified lung tumor DNA extracted from stained slides after formalin fixation and paraffin embedding as previously described (20), and 1 μL was used in each PCR without quantification. Additional normal DNA samples from laboratory personnel were purified by lysis, salt precipitation, and washing with ethanol (Gentra Puregene Blood Kit, Qiagen) and used as controls. All DNA samples (except from the lung tumors) were quantified by absorbance at 260 and 280 nm.
PCR PRODUCT DESIGN
uMelt software (https://dna.utah.edu/umelt/umelt.html) was used to predict melting curves and Tms of PCR products. The Tm difference between target and reference amplicons was designed to be between 2 °C and 10 °C. Human genome databases (http://www.ncbi.nlm.nih.gov and http://genome.ucsc.edu) were searched to confirm that the target and reference primers were unique in the human genome, and that any internal sequence variation was minimal.
Primers for duplex or triplex PCR targeted single copy genes to produce short (50–121 bps) PCR products that differed in Tm. Duplex reactions for chromosomes X, 13, 18, and 21 were paired with reference genes on different chromosomes for relative quantification. Gene amplification of EGFR (7p12) against the reference gene CFTR (7q31.2) used unlinked targets on different arms of the same chromosome. Either SMN1 or SMN2 was selected by allele-specific amplification and compared against a PCR product from a different chromosome. Triplex reactions from chromosomes 7, X, and Y were used to reveal sex chromosome abnormalities. Another triplex reaction compared exon 2 and exon 3 of CFTR against a different chromosome to detect exonic deletions within a gene. The primer sequences, genes, and chromosomes of all references and targets, as well as the PCR product size, Tm and correlation to Figs. in this article are shown in Table 1 in the Data Supplement that accompanies the online version of this report at http://www.clinchem.org/content/vol61/issue5.
SPINAL MUSCULAR ATROPHY GENOTYPE FREQUENCY CALCULATIONS
PCR AND HIGH-RESOLUTION MELTING
PCR was performed on an LC480 (Roche) in 10-μL volumes consisting of 0.5 μmol/L of each primer, 0.4 U KlenTaq1TM (Ab Peptides), 64 ng antiTaq antibody (eEnzyme), 2 mmol/L MgCl2, 50 mmol/L Tris (pH 8.5), and 500 mg/L BSA (Sigma). Also included were 1X LCGreen® Plus dye (BioFire Defense), 50 ng genomic DNA, and 6.25 μmol/L each dNTP, unless otherwise specified. Following an initial denaturation at 95 °C for 2 min, 40 cycles of 10 s denaturation at 95 °C, and 30 s annealing at 65 °C were performed unless otherwise specified. After amplification, the samples were heated to 95 °C momentarily (0 s), cooled to 55 °C for 5 s, and then melted from 65 °C to 95 °C using 15 acquisitions/°C (rate of 0.065 °C/s) to acquire high-resolution melting data. The LC480 with 450-nm excitation and 500-nm emission filters was used for melting unless otherwise specified. SYBR® Green I, used in some experiments instead of LCGreen Plus dye, was from Molecular Probes.
Real-time data were plotted after arithmetic background removal using LC480 software. High-resolution melting data were displayed on negative derivative plots using custom programs written in LabView (National Instruments) to identify and display copy number differences. Fluorescence background was first removed using exponential background subtraction (24) followed by optional fluorescence normalization. Normalization is typically performed to remove fluorescence differences between samples and display the data in terms of percent helicity. However, the absolute fluorescence difference between samples can be informative when studying the compromise between signal strength and the ability to discern copy numbers by varying dNTP concentrations and cycle numbers. Next, curve overlay was performed to remove Tm differences from plate position or chemistry (24, 25). Then, the negative derivative was calculated by Savitzky-Golay differentiation (26). Next, the amplitudes of all reference peaks were normalized to the mean reference peak amplitude to better display copy number differences. Finally, the center of the reference peak and the furthest copy number peak were shifted and stretched horizontally to match the mean temperatures of each peak. Samples were classified into copy number classes by performing unbiased hierarchical clustering on the final target peaks.
Multiplex PCR was used for relative quantification by restricting PCR so that the primers did not become limiting. The products were selected so that they had Tms that were easily distinguishable by melting analysis. The most effective and reliable way we found to restrict PCR amplification was to use a dNTP concentration of 3–12 μmol/L. Fig. 1 shows real-time duplex amplification of a normal control sample with different amounts of dNTPs. At high concentrations, there was little difference in the fluorescence at plateau because dNTPs were in excess. However, the fluorescence dropped and the quantification cycle (Cq) increased as the dNTP concentration decreased until at 3.13 μmol/L the ending fluorescence was only 10% of the highest concentration. When the duplex products were visualized as melting curves, the relative peak heights of the CFTR exon 6 and chromosome X products remained about the same from 3–12 μmol/L of dNTPs. However, from 25 to 200 μmol/L, the peak height of CFTR was constant while that of chromosome X continued to increase. That is, above a certain dNTP concentration, relative quantification failed and other factors affected the efficiency of the 2 products differently.
The effect of dNTP concentration on derivative peak height and relative quantification between 2 duplexed products is shown in Fig. 2. Samples with 1, 2, 3, and 4 copies of chromosome X were used to assess the ability for relative quantification. At very low dNTP concentrations, the fluorescence was very low, reflecting very little amplification (Fig. 2A). The best relative quantification occurred at dNTP concentrations at which the fluorescence was still limited but each copy number was distinguishable (Fig. 2B). At higher dNTP concentrations, the peaks for 3 and 4 copies became admixed (Fig. 2C), and at typical PCR concentrations (Fig. 2D), no relative quantification was possible.
For comparison, PCR was also limited by restricting the number of PCR cycles. Online Supplemental Fig. 1 shows the effect of cycle number on melting curves of 1, 2, 3, and 4 copies of chromosome X. At 21 cycles, the copy number differences could be distinguished although the signal-to-noise ratio was low. At 24 cycles, fluorescence increased and the copy number resolution improved. At 27 cycles, fluorescence was higher still but there was some variation between duplicate melting curves. Finally, at cycle 30, although fluorescence was at maximum, no copy number information remained.
Limiting of dNTPs for relative quantification is robust to differences in initial template concentration. Online Supplemental Fig. 2 compares a 10-fold difference in starting DNA concentration on samples with 1, 2, 3, or 4 copies of chromosome X. With either 5 or 50 ng of genomic DNA, the Cqs clustered by DNA amount (see online Supplemental Fig. 2A). After fluorescence and normalization against a reference gene, the duplex melting curves segregated by the copies of chromosome X, not the initial amount of starting DNA (see online Supplemental Fig. 2B).
We also tried SYBR Green I as a dye instead of LCGreen Plus (see online Supplemental Fig. 3). The optimal range of dNTP concentrations (6.25–12.5 μmol/L) was narrower with SYBR Green I, and although separation of the 1, 2, 3, and 4 copies of chromosome X could sometimes be obtained, the resolution was less consistent than with LCGreen Plus.
Using 6.25 μmol/L dNTPs and LCGreen Plus, 50 blinded samples, including trisomy 13, 18, and 21 and wild type, were amplified using DNA ranging from 10–200 ng. Nine of the samples were identified as trisomy 13 (Fig. 3A), 8 as trisomy 18 (Fig. 3B), 13 as trisomy 21 (Fig. 3C), and 20 had a normal karyotype. The findings correlated without exception to the cytogenetic karyotype. When the same samples were analyzed using standard real-time quantitative PCR (27) copy numbers could not be obtained because the initial template concentrations were not equalized.
Triplex PCR was used to simultaneously quantify chromosome X and Y copy numbers by normalizing against a reference PCR product on a different chromosome (Fig. 4). Fifty blinded samples were analyzed, including products of conception with established abnormal karyotypes as well as male and female normal controls. A control sample, trisomy in X, was also included in duplicate. Ten wild-type male and 22 female samples were easily identified by copy number ratios of chromosome X and Y. However, the normal female samples could not be distinguished from triploidy (69,XXX) because their relative copy numbers do not change. Six samples were monosomy X (45,X), 11 were male triploid karyotypes (69,XXY), and 1 was 47,XYY. The X and Y copy numbers of all samples correlated with the karyotype after cell growth except for 2 samples that karyotyped as 45,X. SNP copy number arrays of DNA of these samples before growth revealed mosaicism at Xp11.4 (the location of the PCR product) compatible with the relative quantification results. The 45,X cell components of both samples were apparently more viable in cell culture and became the only karyotype observed after growth.
In addition to trisomies and sex chromosomes, relative quantification was also demonstrated for exonic deletions and gene amplification. In a triplex assay, a heterozygous deletion including CFTR exon 2 and exon 3 was easy to detect (Fig. 5) against a reference chromosome. When EGFR was targeted in lung tumor samples, gene amplification in most tumor samples was observed to varying degrees (see online Supplemental Fig. 4).
Another disease dependent on copy number changes is spinal muscular atrophy. By combining allele-specific PCR with duplex amplification using restricted dNTPs, the copy numbers of SMN1 and SMN2 were clearly identified (Fig. 6). Zero, 1, and 2 copies of each gene were clearly distinguished with the observed genotype frequencies closely matching the expected population frequencies. Two samples had >2 copies of either SMN1 or SMN2 that were both confirmed by quantitative PCR. By population frequency alone, the most likely copy number of each is 3.
The precision of duplex PCR with restricted dNTPs for copy number analysis depends on the resolution of the melting instrument used and the purity of the DNA. With pure cell line DNA and high-resolution melting, a 1.11-fold increase was easily detected. Although the peak ratio to copy number is not linear and control DNA samples are required for calibration, the precision is excellent, averaging a CV of 0.85% (see online Supplemental Fig. 5).
We introduce competitive PCR with restricted dNTPs, using different primer pairs for reference and target products. After PCR, high-resolution melting distinguishes the products by Tm for relative quantification. Instead of calculating copy number ratios by derivative peak heights, the intensity of the reference peaks are normalized, allowing direct visualization of relative target copy number. Duplex PCR was used to quantify chromosomes X, 13, 18, 21, and the genes EGFR, SMN1, and SMN2 against reference PCR products. Triplex PCR was demonstrated for chromosomes X and Y and CFTR exons 2 and 3. The observed variation within a genotype is greater for dissected clinical samples (Figs. 3 and 4) than with pure DNA (see online Supplemental Fig. 5), likely because of variation in the success of separating all maternal tissue from the fetal samples.
Limiting cycle number for relative quantification of targets with different primer pairs is well established (15–17). Presumably, PCR retains the relative amounts of multiple targets throughout the exponential phase. However, later in PCR, the initial target amounts are no longer reflected by the final products. Because of uncertainty in the appropriate cycle cutoff, we looked for other ways to limit PCR, such as restricting dNTP or polymerase concentrations to maintain relative quantification into the PCR plateau. Although relative quantification by limiting the amount of polymerase was possible, the dynamic range and resolution obtained with this method were not as good as those obtained by restricting dNTPs (data not shown). With dNTPs limiting, the copy number resolution depends on the melting resolution and the purity of the DNA. Although we used melting analysis here, other detection methods, such as relative quantification with fluorescently labeled primers run on a sequencing gel, could be used.
When different primer pairs are used for each target, it is easier to separate their Tms, to keep the products short, and to minimize internal sequence variation. Variation within a PCR product shifts the Tm (if the change is homozygous) or alters the shape of the melting curve (if the change is heterozygous (28)), both undesirable for normalization and clear comparison of the melting curves.
Although we usually prefer rapid temperature cycling (29), relative quantification with restricted dNTPs was better resolved with slower cycling. Annealing temperatures of 1–2 °C higher than the primer Tms were used, along with a 30-s annealing time. When SYBR Green I was used as a dye, the resolution was not as good as with LCGreen (see online Supplemental Fig. 3). This might be expected because SYBR Green I is not a saturating dye and favors high Tm products in multiplex amplifications (30). Other saturating dyes or labeled probe methods (hydrolysis probes, hybridization probes, or molecular beacons) might also be used to assess copy number variation using limiting dNTPs.
Although prior methods of relative quantification estimate copy number ratios by relative peak heights (16, 17), it is not easy to estimate the baseline on derivative melting curve plots. In our hands, better precision and clarity were obtained by normalizing the reference peaks of all derivative plots both vertically (fluorescence) and horizontally (temperature). Different copy numbers of the target are then clustered for simple visual inspection. An additional calibrator (31) is not necessary because the reference peak serves the function of Tm adjustment between curves. One disadvantage is that controls are necessary to validate the absolute copy number. For example, in Fig. 5, SMN1 and SMN2 each have single unknowns that are >2 copies, most likely 3 copies by the population distribution, but this is not certain until either a different method is used for validation or a 3-copy control is available.
Prior methods to determine SMN1 and SMN2 copy numbers relevant to this study include allele-specific PCR followed by quantitative PCR (32) and 3 melting methods. Two of the melting methods amplify a small amplicon around the informative SNV and can identify many copy number ratios (33, 34). However, a single copy at each locus (carrier) cannot be distinguished from 2 copies at each locus (normal). The same concern occurs in the other melting method that uses an unlabeled probe (35). We used allele-specific duplex PCR with limiting dNTPs to independently assess the copy number of each gene.
A comparison of copy number precision between quantitative PCR, digital PCR, and melting with limited dNTPs is instructive. The detectable changes from quantitative PCR are about 1.25- to 1.5-fold while digital PCR can detect changes <1.2-fold (36) and melting analysis easily separates a 1.11-fold difference with a CV of about 1% (see online Supplemental Fig. 5). In digital PCR the CV depends on the number of partitions analyzed and the fraction of partitions that are positive (37). For digital PCR with 20 000 counts, the CV is around 3%–4% (38, 39). Although digital PCR precision can be improved by increasing the number of partitions, not all digital systems are configured to acquire large numbers of partitions, and the analysis time is proportional to the number of partitions counted.
Relative quantification by limiting dNTPs does not require real-time PCR, but melting capability is required. Melting data may be obtained on either the same instrument or on a dedicated melting instrument. Only 5 min or less are required for melting after PCR, although this varies between instruments (40). Copy number resolution depends on the melting resolution, which is a continuous variable even among instruments that claim high-resolution melting. Relative quantification by limiting dNTPs should be useful for targeted evaluation of relative copy numbers, including confirmation of copy number changes revealed by microarrays or massively parallel sequencing.
We thank Derek David, Jesse Montgomery, and Zachary Dwight for helpful discussions and technical assistance.
- cystic fibrosis transmembrane conductance regulator (ATP-binding cassette sub-family C, member 7);
- breast cancer 1, early onset;
- breast cancer 2, early onset;
- chemokine (C-C motif) ligand 3-like 1;
- epidermal growth factor receptor;
- survival of motor neuron 1, telomeric;
- survival of motor neuron 2, centromeric.
↵5 Nonstandard abbreviations:
- single-nucleotide polymorphism;
- melting temperature;
- deoxynucleotide triphosphate;
- spinal muscular atrophy;
- quantification cycle.
(see editorial on page 684)
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.
Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:
Employment or Leadership: L. Zhou, University of Utah; C.T. Wittwer, BioFire Diagnostics and Clinical Chemistry, AACC.
Consultant or Advisory Role: None declared.
Stock Ownership: None declared.
Honoraria: None declared.
Research Funding: C.T. Wittwer, BioFire Diagnostics.
Expert Testimony: None declared.
Patents: L. Zhou, patent pending; R. Palais, patent pending; C.T. Wittwer, patent pending.
Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.
- Received for publication November 22, 2014.
- Accepted for publication February 2, 2015.
- © 2015 American Association for Clinical Chemistry