High Resolution Melting Analysis

Albert Lam

Introduction

High resolution melting analysis (HRMA)is a method for analysis of DNA fragments. Using the melting temperature and melting curve for a fragment, it is possible to derive information about the sequence. It is quick and cheap and does not require special training, making it a widely accessible technique in many laboratories1.

Figure 1: Overview of High Resolution Melting Analysis2. For a heterozygous locus, there are two species of homoduplex DNA, one for each allele. Polymerase chain reaction is used to amplify the four strands of DNA, which are then all melted and reanneal randomly to produce two homoduplexes and two heteroduplexes. A saturating, intercalating fluorescent dye binds to the double stranded DNA and fluoresces. Gradual heating causes the DNA to melt and release the fluorescent dye.

The first step in HRMA is the amplification of DNA by qPCR in a reaction containing a saturating, intercalating DNA binding dye. The saturating dye only fluoresces when bound to double-stranded DNA, which allows for detection of duplex DNA. By measuring fluorescence, we can determine the amount of double stranded DNA. As the temperature increases, the DNA melts and releases the dye, which is detected as a decrease in fluorescence. The melting curve produced by the release of the dye is used for further analysis1,2.

Three commonly used interpretations of the melting curve are the difference plot, the negative first derivative plot, and the HRM score. A difference plot shows the differences between samples by subtracting their melting curves from a reference curve.A negative first derivative plot finds the slope at each point in the melting plot to determine the start of amplicon melting and the end of melting. Using the negative first derivative plot, we can determine a quantitative measure, the HRM score, by subtracting the temperature at which amplicon melting is complete minus the temperature at which melting begins3.

Several factors affect the melting curve of a particular DNA fragment, most notably the GC content. GC base pairs in the DNA form 3 hydrogen bonds as opposed to 2 hydrogen bonds between AT base pairs1. Tong and Giffard have made an approximation for the melting temperature of a fragment using Tm = 81.5 + 16.69(log[Na+]) + 0.41(%G+C) - (500/sequence length)1. The sequence length and ion concentrations in solution also contribute to the melting temperature and melting curve of the fragment. Longer sequences will have higher melting temperatures and may contain multiple melting domains. Multiple melting domains may either help discern sequence variation of mask them. Cousins et al. has shown that for sequences longer than 400 bp, single base pair changes, single base pair deletions, and 3 base pairs and 9 base pair insertions do not greatly affect HRM score4.

There are many reported uses for HRMA over the past decade. Several of the most common ones are for SNP determination, bisulfite sequencing, and assessment of sequence diversity. SNP determination is by far the most reported use for HRMA. In the analysis of microbiological species, different species can be differentiated from one another by HRMA analysis of conserved genes with polymorphisms, such as ribosomal RNA. Genotyping by examining SNPs is also commonly done. SNPs can be associated with antibiotic resistance in pathogens, but they are also associated with certain alleles in humans. The genotype at a particular locus can be determined using HRMA.SNP determination makes use of the formation of heteroduplexes, duplex DNA composed of strands that are not perfectly complementary. Reactions containing heterduplexes have different melting curve shapes than reactions containing homoduplexes. In general, heteroduplexes cause the melting curve to expand towards the lower temperatures5,6.

Heteroduplexes are created after the last step in the qPCR reaction. The samples are heated to a temperature to melt all duplexes and then cooled to allow for annealing of DNA to form heteroduplexes. The formation and detection of heteroduplexes is affected by several factors. Gundry et al. has shown that by increasing the rate of the cooling step after initial DNA melting, more heteroduplexes are formed in relation to the amount of homoduplexes in heterozygotes. During the generation of the melting curve by incrementally increasing the temperature to melt the DNA, a faster rate of heating allows for greater detection of heteroduplexes. Gundry et al. suggests that for maximal detection of heteroduplexes, a cooling rate of >2°C/s, a heating rate of >0.2°C/s, and low magnesium concentrations are preferred7.

HRMA is also commonly used to determine the extent of the methylation near the promoter of genes. Methylation of the promoter of a gene may indicate gene silencing, which is particularly problematic if a tumor suppressor gene is silenced.The first step in using HRMA for DNA methylation analysis is bisulfite conversion of unmethylated cytosine to thymine. During the amplification step, the thymine produces A-T base pairs, contributing to a decreased melting temperature compared to the sequence before bisulfite conversion. When the melting curve of DNA amplicons are analyzed, segments that are more methylated will produce melting curves with a lower melting temperature8.

Towler et al. proposed to use HRMA as a technique to assess HIV diversity. In their study, they amplified a region of the HIV gag gene (HXB2 2068-2278) and measured the melting curve using HRMA. They also sequenced the HIV genome and calculated sequence diversity %. The two groups they used were mothers positive for HIV and their infants who are also positive for HIV. The infants were considered to be incident infections, while the mothers were not classified as some could have been infected recently while others were infected long before sample collection. By comparing the HMR score to sequence diversity, they were able to determine that higher HRM scores were associated with greater HIV sequence diversity. However, plotting their data in a scatter plot, Figure 2, shows low correlation between HRM score and sequence diversity. What is consistent, though, is that the HRM scores of the infants are lower than most of the HRM scores of the mothers9.

Cousins et al.has also published several papers on using HRM score as a measure for HIV incidence. Using 6 different sequences of the HIV genome within the gag, pol, and env genes,they found significant differences between the HRM scores of patients with acute, recent, and non-recent infections in 5 of the 6 sequences, as can be seen in Figure 4. Only ENV2 did not show significant differences between acute, recent, and non–recent samples. By comparing HRM scores of patient samples to plasmid controls, they proposed that their HRM diversity assay is capable of detecting even low levels of diversity early in infection, the acute phase10.

Additionally, Cousins et al. has developed a multi-assay algorithm (MAA)using HRMA to determine HIV incidence. Using a combination of 4 different assays, they were able to design an assay with a mean window period of 154 days and a shadow of 179 days (Figure 5).Mean window period is the number of days that a person who is a recent infection would be determined as MAA positive. The shadow is the number of days prior to the date of sample collection that is being assessed. For example, the MAA that Cousins has developed has a mean window period if 154 days and a shadow of 179 days. If an average patient were to have his sample collected every day after infection, he would test positive for 154 days11.You can then expect that the person was infected 179 days ago when he tests positive. In general, a large mean window period allows for greater detection of incident infections, allowing for smaller cross-sectional surveys12. A shadow of less than a year is needed to accurately determine incidence of HIV infection within the past year12.

There are some limitations to HRMA. Most significantly, it is difficult to detect small changes if the GC content of a fragment does not change. For example, A->T and T->A transitions do not cause a large shift in the melting curve (ref). Detection of SNPs may also be limited when nearby variation affects the melting curve. The overcome these limitations, 3’ blocked probes can be used to target a particular region or SNP of interest within a longer amplicon sing asymmetric PCR. These probes can mask nearby variation so that they do not affect detection of the SNP, and they can also detect changes even when the amplicon is long5,6.

There are many uses for HRMA, and as more techniques are used to modify the method, more applications may be possible. Its advantage over sequencing lies in its cost effectiveness, speed, and simplicity, though he information gathered from HRMA may not be as abundant as from sequencing.

Rationale

It should be possible to distinguish between incident and chronic samples using high resolution melting. Although the Cousins et al. group failed to do so, their analysis was solely dependent on the HRM score, leaving out a wealth of other information. In addition, the regions they targeted may not be ideal for HRM analysis. By targeting a different region, changing the methods of how to do high resolution melting, or examining more than just the HRM score, a deeperanalysis can be done on the high resolution melting curves to discover whether or not a difference can be found between incident and chronic patient samples.

Incident samples can be divided into single founder and multiple founder samples. Chronic samples contain a high diversity of sequences without sequences that have low hamming distances (Q10) to other ones within the sample13. Incident samples are expected to have distinctly different melting curves than either multiple founder samples or chronic samples. Between the multiple founder and chronic samples, the differences may be less distinct and obvious, and thus, a very careful analysis must be done to determine a criterion that can be used to distinguish between the two.

Methods

High Resolution Melting

I used the Roche Light Cycler 480 instrument and software for high resolution melting detection and analysis. Reagents, plates, and optical film were all provided by Roche for use with the Light Cycler 480 in the 96-well block. Amplification was done using 45 cycles of 95° for 10 seconds, 65° for 10 seconds with a starting temperature of 55° and increasing by 1° for each cycle until 65°, and 72° for 25 seconds. The number of cycles was later reduced to 40. The MgCl2 concentration used was 3 mM. I used primers HIAFor4 (5’-TACAAGACCCAACAACAATACA-3’) and HIARev4 (5’-GCTTTTCCTACTTCCTGCCACAT-3’) to amplify a 411 base pair sequence in the HIV subtype B envelope.

High resolution melting was performed by first melting the amplified PCR product to 95°at the maximum rate of 4.4°C/s for 1 minute and then cooling to 40° at a maximum rate of 2.2°C/s for 1 minute. The temperature was raised to 65° before beginning fluorescent capture for HRMA. 25 acquisitions were made per °C at a heating rate of 0.02°C/s. These conditions are part of the recommended protocol supplied with the reagents Roche provided and are similar to the conditions Cousins et al. used10.

To obtain multiple melt data, amplification of samples was done as described above, but directly after amplification, the melting curve was obtained by heating to 95° at 0.02°C/s. Afterwards, the samples were maintained at 95° for 1 minute before cooling to 40° for 1 minute. A second capture of the melting curve was performed as described above. After the second capture, successive cycles of heating to 95° for 1 minute and cooling to 40° for 1 minute were performed until the PCR amplified product had been melted and annealed ten times. A final melting curve capture was performed at the end of the heating and cooling cycles.

Hamming Distances

I calculated the hamming distances for 500 envelope sequences for each of the HIV B, HIV C, and HIV D subtypes. From there, I created dendrograms using R to create 5 different groupsfor each subtypeand found the representative sequences for each group. From there, I generated multiple melting curves using uMelt (

In addition, 1000 sequences were analyzed using BioEdit to examine conserved sequences within the HIV B subtype gag and pol genes. Primers were designed at the conserved regions to target variable regions within the genes and an average hamming distance for each variable region was calculated. Average hamming distances for the regions in gag, pol, and env analyzed by Cousins et al. was also calculated10.

Results

High resolution melting curves was obtained for several samples of single genome amplification, which should resemble incident infection with a single founder strain, and HIV cDNA from patients with a chronic HIV infection (Figure 6). I did not test incident infection with multiple founder strains. However, performing HRMA on samples from an incident infection with multiple founder strains is necessary to determine if a difference can be found. Plotting the negative first derivative of the melting curves for the different samples suggests that there are differences between the shapes of the melting curves. Notably, however, is the difference seen between HIA8 C9 and HIA8 C12; both sequences come from the same patient, but the difference is that they are from different strains. There are a total of 7 nucleotide differences between the two sequences, and all differences are an A/T base from HIA8C12 compared to a G/C base from HIA8 C9. These differences lead to the observed melting curves.

Figure 6 High resolution melting curve for single genome amplification (SGA) samples and samples from chronic patients. HIA8 C9, HIA9 G2, and HIA8 C12 are SGA samples; HIA2, HIA19-1, and HIA35-1 are samples from chronic patients.

In order to test if the samples were fully mixed, I tested whether or not melting the amplified products and annealing them multiple times would have a significant effect on the melting curves (Figure 7). Three separate melting curves were obtained using the multiple melt method as described above. Most of the curves have the same shape within each set, with the exception of HIA35-1; after multiple melt cycles, there is a shift in intensity towards the peak with a higher melting temperature. Contrary to what we expect, not melting the DNA homoduplexes after amplification but before HRMA did not contribute to a significant difference to after melting the DNA duplexes.

However, a caveat of the HRMA experiments already conducted is that the melting times may not have been sufficient to allow for complete melting and annealing. Analysis using the Light Cycler software showed that the fluorescence for each sample did not all start at the same intensity. Furthermore, the ending fluorescence after amplification was not the same as the starting fluorescence for the first melting curve in the multiple melt curves (data not shown).Further experiments should examine whether or not a longer cooling time or longer melting time may affect the melting curves. A recent presentation (unknown source, from Dr. Park) has suggested that relaxation time for annealing of duplex DNA may require longer than 200 seconds.

Figure 7 Multiple melting curves. HIA2 and HIA4 are cDNA samples from chronic samples, HIA8 C9 is an SGA, and HIA35-1 is a sample from a chronic patient. The Y-axis is in –d(Fluorescence)/d(Temperature).

The representative sequences for 5 different groups of HIV B, HIV C, and HIV D envelope genes were determined from a sample of 500 sequences from each subtype (supplemental file). Some sequences have more than one peak, but the majority of the representative sequences throughout the three subtypes have only one peak. This may suggest that many of the sequences we come across in incident samples should have only one peak.

In order to better evaluate the Cousins et al. paper examining the different regions of HIV using HRM, I wanted to see whether or not the variability of the regions they examined were high in diversity. I calculated the average hamming distances for each region examined (Table 1). All of the regions have fairly high average hamming distances, and all are above 0.02. However, their variable region spans only around 100 base pairs each. I have examined other variable regions that can be targeted, and although their average hamming distances may be lower, the regions span over 300 bp each (Table 2).

Table 1: Average hamming distance for variable regions examined in Cousins et al. 2011
Cousins Variable Region / Average Hamming Distance
Pol / 0.0310876395521
Gag 1 / 0.0280855773807
Gag 2 / 0.0337753118973
Env 1 / 0.0257867050724
Env 2 / 0.0365863158756
Env 3 / 0.0469734697384
Table 2: Average hamming distance for potential variable regions to be targeted
Variable Region / Average Hamming Distance
Gag Region 1 / 0.0270205290036
Gag Region 2 / 0.0407733897419
Pol Region 1 / 0.0266858706281

Discussion

Our preliminary data suggests that there are differences that can be found between incident HIV infections and chronic HIV infections using HRMA. However, we must firstunderstand why the starting fluorescence for each melting curve within the same sample is not consistent. Afterwards, we must also perform HRMA on more sample types, especially on the incident infection with multiple founder strains. We expect that that will be the most challenging type of incident samples to differentiate from chronic samples. Once we have completed those two tasks, we may begin to look for a criterion to distinguish and develop and assay to determine incident and chronic infections.

References

1.Tong, S. Y. C. & Giffard, P. M. Microbiological applications of high-resolution melting analysis. J. Clin. Microbiol.50, 3418–21 (2012).

2.Roche Diagnostics. High Resolution Melting : Optimization Strategies High Resolution Melting : Optimization Strategies. (2008).