Rapid detection of pathological mutations and deletions of the haemoglobin beta gene (HBB) by High Resolution Melting (HRM) analysis and Gene Ratio Analysis Copy Enumeration PCR (GRACE-PCR)

Quantification of Hb A₂ is routinely used for the detection of ?-thalassaemia carriers. The normal range for Hb A₂ is 2.0–3.3 % with levels above 3.5 % being considered indicative of ?-thalassaemia trait [11]. The current study included 22 samples with Hb A₂ levels greater than 3.5 %, which did not have detectable mutations or deletions of the HBB gene. This is in line with previous observations that false positive results can be a significant problem when screening for ?-thalassaemia carriers based on Hb A₂ levels [40]. Indeed, elevated Hb A₂ in the absence of ?-thalassaemia has recently been linked to mutations in the KLF1 gene [14], suggesting that molecular confirmation may be required. In our laboratory, HBB gene mutations are routinely detected with a commercial reverse hybridization assay (ViennaLab Diagnostics). This commercial kit was designed to detect the 22 most prevalent mutations found in Indian and Middle Eastern populations. The current study identified 63 samples (15.4 %) with mutations not targeted by this kit (Table 3). This further illustrates the limitations of conventional PCR techniques when screening for ?-thalassaemia trait in highly heterogeneous populations, and highlights the advantages of using a more universal assay, such as the one described here.

The first application of HRM-PCR to the HBB gene was described by Wittwer et al. [41]. This was a small-scale study designed to detect the mutations responsible for the abnormal haemoglobins Hb S and Hb C [41]. Despite the fact that they amplified approximately the same section of the gene covered by the H2 primer set in the current study, the authors did not report any problems with the HBB:c.9C??T SNP. This is presumably due to the relatively small number of samples tested (n?=?12). In contrast, in the current study the HBB:c.9C??T SNP was encountered frequently and it was not possible to resolve all genotypes without the addition of the unlabelled probe.

Three previous studies have described HRM-PCR assays to scan the HBB gene in different populations. Shih et al. [30], Saetung et al. [31] and Lin et al. [29] used the technique to detect the most prevalent mutations in Taiwanese, Thai and Chinese populations respectively [29–31]. These assays scanned selected areas of the HBB gene and the assay described by Saetung et al. also included a tube for the detection of the 3.4 kb deletion. Both Shih et al. [30] and Lin et al. [29] reported problems with the HBB:c.9C??T SNP, which were resolved by placing the primer over the SNP. A possible concern with this approach is that it might result in allele dropout, particularly if a SNP is located close to the 3’ end of a primer. The primers described by Shih et al. [30] put the SNP close to the 3’ end, while the primer used by Lin et al. [29] place the SNP in the middle of the primer where it is less likely to have a significant effect. However, the HBB:c.19G??A and HBB:c.20A??T mutations are located at the 3’ end of the primer used by Lin et al. [29], which would probably affect the amplification of these common pathologically significant alleles. Since these mutations are rare in China [42] and were not included in the specific mutations being targeted this would not have been a significant concern when screening that population. However, these mutations are very common throughout Africa and the Middle East [43], thus the assay described by Lin et al. [29] is not well suited for screening in these regions. In contrast to overlaying the HBB:c.9C??T SNP with a primer, the approach described here of using the unlabelled probe, allowed detection of all the mutations we encountered in this area of the HBB gene.

Saetung et al. [31] avoided complications from the HBB:c9.C??T SNP by simply not scanning this region of the HBB gene. This approach limits the usefulness of their assay in populations where mutations in this area of the gene are common.

Other high frequency SNPs also complicated HRM analysis of the HBB gene. These were HBB:c.315?+?16G??C (rs10768683) and HBB:c.316-185C??T (rs1609812). We addressed this issue by substituting the respective SNP with inosine in each primer designed to cover these bases. This approach, as previously demonstrated for the BRCA genes [37], eliminated the detection of this high frequency SNP and provided clean HRM profiles (Fig. 2; 1a and 1b).

Until recently, deletions of the HBB gene were considered to be rare with only the 619 base deletion being common, accounting for around 20 % of ?-thalassaemia alleles in some parts of India and Pakistan [2]. However, it has now become apparent that HBB gene deletions are more frequent than was previously believed. The 3.4 kb deletion has a prevalence of up to 4 % in some parts of Thailand [44] and the Filipino ?-thalassaemia deletion has been described as being a common cause of ?-thalassaemia in Filipinos [36, 45]. Indeed, it has been suggested that deletions involving the ?-globin gene cluster account for as many as 10 % of all ?-thalassaemia mutations [4].

Conventional HRM-PCR works well for the detection of point mutations and small indels, but is not a useful technique for the detection of larger gene rearrangements. Melting curve assays have been described to detect the 3.4 kb deletion of the HBB gene [31, 44]. However, for other deletions Gap-PCR is still commonly used. Although Gap-PCR is a simple and robust technique, it is not ideal for high volume screening, since the use of agarose gel electrophoresis for detection makes it a time consuming open tube technique. An important limitation of both Gap-PCR and the existing melt curve assays is that they can only be used to detect specific deletions with well-defined breakpoints and consequently are not suitable for the detection of novel or rare deletions. In contrast, the GRACE-PCRs included in the current study determine the copy number ratio between the HBB gene and the reference gene. Thus GRACE-PCR can be used to detect large rearrangements of the HBB gene without prior knowledge of the breakpoints. In order to maximise the number of deletions that could be detected our assay used two GRACE-PCR reactions, one targeting the promoter and the other targeting the third exon.

Allele dropout due to a SNP within a priming site is a potential risk with any PCR based assay. The use of the probe and primers containing inosine in the current assay addressed this concern for the most common high frequency SNPs associated with the HBB gene. However, with 868 known pathological mutations of the HBB gene, as well as a number of non-pathological ones, it is inevitable that there are SNPs associated with most priming sites. The risk of misinterpretation due to allele dropout is significantly mitigated by the overlapping nature of the HRM amplicons and by the use of two GRACE-PCR reactions. Furthermore, the fact that no problems due to allele dropout were encountered in the 410 samples evaluated in the study indicated that the risk is indeed small. Nevertheless, it is recommended that the possibility of allele dropout be considered when results of the assay do not correlate with other laboratory findings or the clinical picture.

The general purpose of gene scanning is to identify amplicons carrying variant sequences. The HRM-PCR profiles of the amplicons not only indicated the presence of a variation in the sequence, but also often allowed the identification of the exact nucleotide change. In the 68 blindly analysed samples all variant sequences were correctly detected. Furthermore, in 84 % of these cases the melting curves were sufficiently distinctive for the correct genotype to be assigned prior to sequencing (Table 3; columns Sample set 2 versus Sample set 1). We believe, as more experience is gained with this assay, provisional direct genotyping of almost all samples should be possible. The observed genotyping ability of HRM-PCR for the HBB gene contrasts favourably with denaturing HPLC (DHPLC) assays, as DHPLC may require either a two-step assay [46], or may require to be repeated after spiking with control DNA from a known variant [47].

When a mutation occurred in overlapping amplicon regions (Fig. 1 and Table 1) it was detected in both amplicons (Table 3). This provided additional information assisting in the assignment of the correct genotype. For example, a sample found to be positive with both primer sets H4 and H5 was likely to be either HBB:c.126_129delCTTT, HBB:c.135delC or HBB:c.140G??A, since they were the only mutations identified in the current study that occurred in this region of overlap.

Interestingly, SNPs in linkage disequilibrium with pathological mutations also provide additional information which assists in assigning the correct genotype. For example, an association between Hb Monroe (HBB:c.92G??C) and the SNP (HBB:c.-92C??G) in the HBB gene promoter has been reported [48]. In the current study three cases of Hb Monroe in association with HBB:c.-92C??G were found (Table 3; Genotype 22). The Hb Monroe mutation and the SNP resulted in distinctive melting curves in amplicons obtained with primer sets H3 and H1, respectively.

No mutations were found in the region of the HBB gene scanned by primer sets H7, H10 and H11, indicating that mutations in this part of the gene are comparatively rare in the population studied (Table 3). It is interesting to note that even in the highly heterogeneous population of the UAE, 91 % of positive cases could be detected using just 5 of the 13 primer sets. This suggests that a stepwise approach could be used starting with high mutation frequency amplicons, thus reducing the number of PCR reactions required to obtain results.

The combination of the HRM-PCR and GRACE-PCR assays theoretically cover 97 % of the mutations listed in the HbVar database [3]. This compares favourably with previously described assays that range from 35 to 90 % theoretical coverage [29, 30]. An example of a mutation that would not be detected by the current assays is HBB:c.203_204delTG [35] due to lack of overlap between primer sets H5 and H6. Similarly, HBB:c.316-146 T??G and HBB:c.316-125A??G [3] would not be detected due to lack of overlap between primer sets H7 and H8.

Only negative (wild type) controls were included in the current study, since all samples screened were also sequenced. However, when using the assays for clinical testing, the inclusion of appropriate positive and negative controls with each run is highly recommended. Such controls could be genomic DNA from confirmed cases or plasmids [29].

In our laboratory, the costs for scanning the HBB gene with the combined HRM-PCR/GRACE-PCR assay were approximately one eighth the cost of sequencing. Given that in at least 84 % of cases HRM-PCR allowed the assignment of a provisional genotype, this represents an opportunity for significant cost savings. Should sequencing be required to evaluate an unexpected melt curve, this can be performed by directly sequencing the positive HRM amplicon [20], which is still cheaper and faster than sequencing the entire HBB gene. In addition, the HRM-PCR / GRACE-PCR approach allows the detection of large deletions which are not readily detected by sequencing. In the current study, the samples were manually pipetted and processed in batches of 72 PCR reactions. Since both HRM-PCR and GRACE-PCR lend themselves to automation, robotic reaction setup is possible, and if a micro-titre plate based PCR system was used, up to 384 reactions could be run simultaneously on a single plate. Although we used different annealing temperatures, all the primers used for HRM-PCR will perform adequately at 55 °C, which would simplify batch processing if micro-titre plates were adopted.

In conclusion, the current study has resulted in the development of a robust assay for the detection of pathological HBB gene mutations. The approaches used for handling the high frequency SNPs mean that the assay is universally applicable, rather than being population specific. The inclusion of GRACE-PCR targeting both the gene promoter and third exon means that the assay has the ability to detect a wide range of large deletions of the HBB gene. The assay has sensitivity and specificity comparable to sequencing and was able to detect all mutations in the samples tested. Furthermore, this assay can be run at a fraction of the cost of a full sequencing approach as 84 % of positive samples could be provisionally genotyped. The assay has provided useful data on the spectrum of HBB gene mutations in the highly heterogeneous population of the United Arab Emirates. In future it should be able to serve as a screening test for suspected ?-thalassaemia carriers and variant haemoglobins.