Discovering rare variants by use of melting temperature shifts seen in melting curve analysis.

Two reports in this issue of Clinical Chemistry describe novel mutations in thrombophilia genes (1, 2). The authors of the reports describe variants detected by the identification of a melting temperature (Tm) different from the expected Tm of the target mutation and from the Tm of the wild-type sequence. The variants reported in this issue are A20218G in the prothrombin (factor II) gene and C1690T in the factor V gene. These cases add to the reported variants identified by fluorescent hybridization probes and melting analysis on the LightCycler. In addition to factor II and V variants, variants have been discovered similarly in genes coding for MTHFR and HFE (3, 4). Melting analyses, as performed on the LightCycler or similar instruments, can distinguish variants that lie within the region of the target probe. Differences in base composition, nucleotide position, and nearest-neighbor environments all affect Tm. These differences are detected by monitoring fluorescence during an increase in temperature. Melting is visualized by a loss of fluorescence as the probes dissociate from the template. Differences in Tm from the expected Tm for wild-type and target mutations are indicative of a nontarget (a variant other than the mutation the assay was designed to detect) mutation or variant. In contrast, assays that use restriction enzymes, allelespecific amplification, or hybridization probes at a single detection temperature may not detect a variant or may show an indeterminate genotype (3, 4). With these 3 types of assays, when one allele carries the variant and the opposite allele is a wild type, the variant allele may be undetected, and the genotype would be reported as wild type. If the variant is in conjunction with a target mutation on the opposite allele, the genotype could be reported as a homozygous mutant, rather than a mutant/variant compound heterozygote. The melting analysis avoids these pitfalls. Variants with Tm shifts of 2–5 °C are easily visualized, as was the case of the prothrombin A20218G case reported by Tag et al. (1 ). However, if the base change produces a sequence with a temperature stability similar to that of the target mutation, the Tm shift may be subtle. Other variants, as discussed by Mahadevan and Benson (2 ), could be mistaken for a Leiden mutation if careful qualitycontrol measures are not followed. By tracking Tms over time, a Tm range can be established for each allele. However, Tms can vary between samples as a result of several factors, including salt or DNA concentrations. Calculating the difference in Tms between the wild-type and mutant peaks ( Tm) in heterozygous samples is more precise than relying on Tm alone. Tms are less affected by sample-to-sample differences because both alleles are affected equally. The Tm variation of any specific allele typically ranges from 0.2 to 0.5 °C, whereas Tm variation is often 0.1 °C or less (5 ). Ninety-five percent confidence intervals (2 SDs) of Tms are 0.2 °C. For factor V and MTHFR assays, we have identified variants with Tms of 0.4 °C. In the report of Mahadevan and Benson (2 ), the variant they found was readily identified by both the Tm and the Tm. The excellent precision exhibited in these assays led us to hope that different variants may be identified by Tm alone, eliminating the need for sequencing. After confirming an HFE variant T189C (6 ) by sequencing more than a dozen samples, we now identify this variant by Tm alone. By contrast, for the prothrombin gene, we have found 3 different variants with Tms that were within 0.5 °C of each other and had overlapping 95% confidence intervals. We therefore routinely sequence all prothrombin samples with Tm shifts to confirm the variant. The cost of sequencing is not an important factor because these variants constitute a small proportion of tested samples. For samples that appear homozygous (wild type or mutant), identifying variants is more difficult because Tms cannot apply. In these cases, Tm alone, or in conjunction with the shape of the curve (a shoulder or broad peak), can often suggest a variant. The frequencies of these variants differ among target genes. Factor V variants are rare, with 1 detected in 17 000 samples (1 in 34 000 chromosomes). In addition to the reported factor V C1690T variant (which prematurely terminates the protein chain), we have also seen A1696G (7 ), which produces an isoleucine-to-valine change in the activated protein C (APC)-resistance pocket (8 ); 1690delC (7 ), which causes a frameshift; and G1689A (no amino acid change) (9 ). Prothrombin variants C20209T (10 ), A20207C (5 ), A20218G (1 ), and C20221T (11 ) are detected in 1 in 1660 samples (1 in 3300 chromosomes). Of these variants, C20209T accounts for nearly 85%. Of the HFE variants (1 in 1000 samples) T189C (6 ), a silent mutation, is present in 88% of the variants. HFE variants C842A (Thr281Lys) (4 ) and A854C (Glu285Ala) at the C282Y locus and G197A (Arg66His; H63D locus) (6 ) were each found once. MTHFR has several common variants (3 ); C685G (Iso225Val) has an allele frequency of 1 in 4700, and G679A (Asp223Asn) has an allele frequency of 1 in 3300 in our sample set. Overall, MTHFR variants are detected in 1 of 750 samples. A common misconception is that rare variants are on different chromosomes. We have found several cases in which a rare variant is on the same chromosome (i.e., in cis) as a target mutation (4 ). The phase can be determined by testing parents to confirm co-inheritance of the mutation and the variant from one parent. Alternatively, a variant seen with a homozygous mutation confirms that one chromosome contains both a mutation and the variant. After identifying a variant, determining its consequence can be challenging. A variant that causes a frameshift or a stop codon is reported as causative or suspected causative, as is the case for the factor V C1690T variant. However, as Mahadevan and Benson (2) point out, a nonsense mutation at this position could have clinical consequences different from those of the Leiden mutation. Silent mutations such as Editorials