Thalassaemias are inherited haemoglobin disorders characterised by a quantitative reduction of the α- or β-globin chains.^1–3 Genomic deletions involving the α-globin gene cluster on chromosome 16p13.3 are the most common molecular cause of α-thalassaemia (∼80–90% of cases). Rearrangements in the β-globin gene cluster on 11p15.4 account for ∼10% of all β-thalassaemia mutations and hereditary persistence of fetal haemoglobin (HPFH) syndromes. Besides the most common deletions, a large variety of less frequently occurring thalassaemia deletions have been found in different populations. At least 60 different deletions involving the β- and more than 50 involving the α-globin gene cluster have been described to date^4,5 (http://globin.cse.psu.edu/hbvar/menu.html).

The molecular tests commonly used to identify these deletions are gap-PCR, Southern blot analysis, and fluorescent in situ hybridisation (FISH) analysis.^6,7,8,9,10 Gap-PCR can only be applied to known deletions, Southern blot analysis is time consuming and technically demanding and success is very much dependent upon the hybridisation probes available, and FISH analysis involves laborious cell culturing to generate metaphase chromosome spreads and has a low resolution (>20 kb).

Recently, a simple technique suitable for rapid quantitative analysis, multiplex ligation-dependent probe amplification (MLPA), has been described.¹¹ This technique is based on the ligation and PCR amplification of two adjacently hybridising oligonucleotides. Each oligonucleotide pair is designed to give a product of a unique length, and by using common ends all probes can be amplified with one primer pair. Using a fluorescent label allows probe separation on a capillary sequencing system. This method has been applied successfully in a number of genes in which deletions and duplications are common.^12–14 In the original description, the probes were generated by cloning into specially developed M13 vectors. Recently, we have simplified this method by using chemically synthesised oligonucleotides. Discrimination of probes based on chemically synthesised oligonucleotides (∼40–60 nt) was doubled using two universal primer sets each labelled with a different fluorophore, allowing up to 40 probes to be used in a single reaction.¹⁵

To simplify the detection of α- and β-thalassaemia deletions and increase the resolutions, we designed two probe sets for each cluster. For the α-cluster, two probe sets of 35 probe pairs in total were designed with an average distance of ∼20 kb, covering a genomic region of ∼700 kb. For the β-cluster, a total of three probe sets consisting of 50 probe pairs were designed covering a region of ∼500 kb and an average distance of ∼10 kb. Control DNA of known α- and β-thalassaemia deletion carriers was used and the deletion characterised by an independent method. Two groups of patient samples suspected of having a (large) deletion in either the α- or β-globin gene cluster, based on haematological findings, were analysed in this assay.

METHODS

Patients

Patients suspected of having haemoglobinopathies were referred to our laboratory for haematological, biochemical, and DNA analysis.¹⁶ Based on this analysis, they were diagnosed as α- and/or β-thalassaemia carriers. The patients suspected of having α-thalassaemia in whom no abnormalities were found by gap-PCR for the seven most common α-thalassaemia deletions and non-deletion types of α-thalassaemia were excluded by direct sequencing of the α-genes, were selected for MLPA. Some showed either an unbalanced α/β chain synthesis ratio (<0.8) and/or inclusion bodies¹⁷ indicative for a deletion of both α-genes on the same allele.¹⁸ In addition, a few patients presented with haemoglobin H (HbH) disease, but analysis thus far only revealed one mutation, suggesting a deleted allele in trans. Some showed the presence of possible junction fragments by Southern blot, in which the deletion could not be characterised due to lack of probes in the region flanking the potential deletion. In total, 38 possible α-thalassaemia carriers were selected to be screened for rearrangements in 16p13.3. These samples were collected during a period of approximately 5 years.

Patients were selected who presented with a microcytic hypochromic anaemia in the presence of elevated HbA₂ levels, for which standard DNA analysis revealed no abnormalities in the β-globin gene sequence or the 5′ and 3′UTR. These samples include patients showing a high HbF expression, indicative for HPFH, (δβ)⁰- or ^Gγ(^Aγδβ)⁰-thalassaemia, and patients showing normal HbA₂ and HbF levels with α/β chain synthesis ratios higher than 1.5, indicative for deletions involving the complete cluster and/or the regulatory elements. A total of 51 samples were analysed by MLPA.

As positive controls for MLPA of the α-globin gene cluster, we used seven deletions confirmed previously by gap-PCR (- - ^SEA, -α^3.7, -α^4.2, - - ^{Med I}, - - ^FIL, - - ^THAI, and -(α)^20.5, indicated as black bars in fig 1B). Two other deletions (33 kb - -^{Dutch I} and -α^7.9) were previously characterised by Southern blot analysis and direct sequencing of the amplified break point fragments.^19,20 For MLPA of the β-cluster, the Dutch III (ε^Gγ^Aγ)°δβ-thalassaemia of 112 kb,²¹ the 50 kb Belgian (γδβ)°-thalassaemia,²² the 25–30 kb Chinese β⁰-thalassaemia,²³ the 12.6 kb Dutch I β⁰-thalassaemia,²⁴ and the Indian (−619 bp) β⁰-thalassaemia deletions,²⁵ all previously characterised by Southern blot analysis, were used as positive controls (fig 2).

Probe design

In total 35 probe pairs were designed to detect rearrangements on 16p13.3, covering approximately 700 kb from the telomere to the MSLN gene (table 3, fig 1). For each probe pair, the common ends correspond to either the MLPA amplification primers (forward tag 5′-GGGTTCCCTAAGGGTTGGA-3′; reverse tag 5′-TCTAGATTGGATCTTGCTGGC-3′)¹¹ or the multiplex amplifiable probe hybridisation (MAPH) primers (forward tag 5′-GGCCGCGGGAATTCGATT-3′; reverse tag 5′-CACTAGTGAATTCGCGGC-3′),²⁶ which allows simultaneous amplification and detection of the separated fragments in different colours.

Similarly, 34 probe pairs to be analysed in two colours were designed to detect rearrangements on 11p15.4 (table 1, fig 2). A third probe set, consisting of an additional 16 probe pairs, was designed for fine mapping some of the deletions found by MLPA (table 2). In order to detect all 50 probe sets in the same fragment analysis sample run, a third common extension was used for the additional probe set, which allowed the use of a third colour (M13 forward tag 5′-GGCGATTAAGTTGGGTAAC-3′; M13 reverse tag 5′-GTTCACACAGGAAACAGC-3′).

Unique sequence was identified using the BLAT program (http://genome.ucsc.edu),²⁷ and care was taken that no known sequence variants were present in the primer annealing site. Probes within each set were designed to produce PCR products differing by 2 bp in length to allow separation in the size range of 80–125 bp using capillary electrophoresis on an ABI 3700 sequencer (Applied Biosystems, Foster City, CA). Primers have been designed using the RAW program (MRC-Holland, Amsterdam, The Netherlands) such that the melting temperature of the hybridising regions of each probe was at least 65°C, with a GC percentage between 35% and 60%.

The oligonucleotides were from Illumina (San Diego, CA), synthesised in a salt-free environment (50 nmol scale) and used without further purification. The downstream primer of each probe pair was 5′ phosphorylated to allow ligation. Separate probe mixes were prepared to allow the detection of deletions in either the α- or β-globin gene clusters, combining two sets of probes with MLPA and MAPH common ends at a final concentration of 4 fmol/μl. The α- and β-globin gene MLPA probe mixes are available upon request (http://www.LGTC.nl).

MLPA reaction

Reagents for MLPA and subsequent PCR amplification were purchased from MRC-Holland. All primers used for PCR amplification were purchased from Sigma-Genosys (Cambridge, UK). The MLPA reactions were performed as described by Schouten et al¹¹ and White et al.¹⁵ In brief, approximately 200 ng of genomic DNA in a final volume of 5 μl was heated for 5 min at 98°C. After cooling to room temperature, 1.5 μl of the probe mix and 1.5 μl SALSA hybridisation buffer (MRC-Holland) were added to each sample, followed by heat denaturation (2 min at 95°C) and hybridisation (16 h at 60°C). Ligation was performed by adding 32 μl of ligation mix at 54°C for 10 min and the reaction was stopped by incubating for 5 min at 95°C. PCR amplification was carried out for 33 cycles in a final volume of 25 μl, adding both the MAPH-F and -R and the MLPA-F and -R primer sets to a final concentration of 100 and 200 nM, respectively, with MAPH-F being labelled with HEX and MLPA-F being labelled with FAM. The third common primer set used for the β-globin gene cluster were M13-F and M13-R; the primers were labelled with ROX and added to a final concentration of 100 nM. A size standard (0.05 μl ROX 500; Applied Biosystems, http://www.appliedbiosystems.com) was added to each well and products were separated by capillary electrophoresis on the ABI 3700 sequencer (fig 3).

Data analysis

For quantitative analysis, trace data from GeneScan (Applied Biosystems) were exported to Excel (Microsoft, www.microsoft.com) to calculate allelic loss in the patient samples tested.¹⁵ In brief, two probes for unlinked loci were included per probe set as a reference in each sample. The height of each α- (or β-) globin cluster specific probe peak was divided by the sum of the heights of the two reference probe peaks to give a ratio. The median ratio for each probe across all samples was calculated and this value was used to normalise each probe to 1.0, which corresponds to a copy number of two. The upper threshold for deletions was set at 0.75 and the lower threshold for duplications at 1.25. The normalising factor for each sample was calculated as the mean value of the unaffected probes within a sample (defined as falling between 0.8 and 1.2) and dividing all values within that sample by this value.

All samples were tested at least twice. Detection of deletions is simplified by the fact that a series of flanking probes all generate a decreased signal. In cases of unlinked or single probe deletions, the region covering the MLPA probes is amplified and sequenced to rule out the presence of rare sequence variants under the ligation site.

RESULTS

Design of the MLPA assay for α-thalassaemia rearrangements

Fragment analysis in the size range of 80–125 bp allows the simultaneous amplification of approximately 20 probes differing by 2 bp in length. To maximise the number of loci that can be analysed in a single MLPA assay, we used a second primer set with common ends, to allow co-amplification of the two primer sets under the same PCR conditions. Probes were designed for each gene and pseudo-gene in the α-globin gene cluster, in the unique sequences L₀ and L₁, at the HS-40, the MPG gene, and more proximal at conserved sequences, respectively, 20 and 9 kb from the MPG gene (fig 1). More distally, two probes were designed flanking the 3′HVR, known to be involved in many rearrangements of the α-cluster, and 15 probes at approximate intervals of 13–50 kb with the most proximal probe localised in the MSLN gene, known to be deleted in the α-thalassaemia mental retardation syndrome (ATR-16).^28,29 The 35 probe pairs shown in table 3 can detect all the deletions described to date.

Of the 35 probes tested in triplicate on 14 healthy individuals, two gave a standard deviation of greater than 12% (table 3, probes 17a and 21a). These probes were considered to be unreliable and were excluded from further calculations. To investigate the efficacy of the assay, DNA samples of nine carriers with known deletions were used as positive controls. All could be detected unequivocally and their extent could be confirmed (black bars in fig 1B).

To demonstrate that duplications are also reliably detected, we tested a homozygote and heterozygote carrier for the common -α^3.7 deletion, which results in the loss of the α₂-specific 3′UTR and a heterozygote for the so called α-triplication, which is characterised by a duplication of the α₂-specific 3′UTR. The results are summarised in fig 3.

MLPA for β-thalassaemia rearrangements and HPFH

Similar to the α-cluster, 34 probes were designed for loci in the β-globin gene cluster and flanking regions. The region spans from the olfactory receptor gene OR52D1 to OR52A4 and covers an area of approximately 370 kb (table 1, fig 2). Most large deletions reported so far are located in this region and all should be detectable. In order to detect small deletions removing part or all of the β-gene,³⁰ a subset of closely spaced probes (fig 2B) surrounding the β-globin gene were selected. A third probe set was designed with different common ends (M13-F and -R) to allow amplification and detection with a third colour. Loci were selected in between some widely spaced probes and towards the centromere. Standard deviations for these probe sets were calculated in 19 healthy individuals; none showed standard deviations greater than 12%.

Positive controls (marked as black bars in fig 2) were used to test the capacity of the MLPA assay to detect the deletions found by other methods in these patients. Probes covering deleted loci showed half the intensity of the surrounding probes, matching the positions and extensions of all the six known deletions.

Patient samples for α-thalassaemia

Our MLPA analysis revealed a large deletion involving the α-globin genes in 19 out of 38 patients. In the remaining 19 patients, 11 different deletions were detected, affecting either the α-globin genes or the regulatory elements known to be involved in globin gene expression. Six showed no resemblance to previously described deletions and were considered to be new (- - ^GZ, - -^OH, (αα)^L, (αα)^ZW, - - ^AB, - - ^MK). One has been described (Dutch II α⁰-thalassaemia) but the breakpoint position and deletion length could not be determined at the time³¹; FISH analysis performed in John Radcliff Hospital in Oxford revealed an approximate deletion length of 300 kb (Higgs, personal communication). Four deletions show similarity with previously described deletions (fig 1B, last four deletions). One 14 year old Dutch girl showed haematological parameters typical for an α⁰-thalassaemia carrier (MCV 65 fl, MCH 19.5 pg, RBC 5.79×10¹², and positive HbH inclusion bodies test). The α-genes were structurally intact and we only detected the deletion of a single probe 5a (fig 1A (αα)^ZW). The location of this probe coincides with one of the cis acting elements that regulate α-gene expression, known as the HS-40.

Patient sample for β-thalassaemia

Analysis of the 51 samples suspected for β-thalassemic rearrangements or HPFH using MLPA revealed 10 different deletions in 31 out of 51 patient samples. In the remaining 20 samples, a deletion of the probe sets tested could be excluded. In three cases, deletions were detected which do not match those described to date and are considered to be new. All three deletions, found in Dutch carriers, silence the expression of the complete globin gene locus and were named Dutch IV (ε^Gγ^Aγ)δβ⁰-, Dutch V ε^Gγ^Aγδβ⁰-, and Dutch VI (ε^Gγ^Aγδβ)⁰-thalassaemia. One matched the HPFH-2 deletion and was confirmed by breakpoint PCR.⁸ One sample belonged to a patient described in 1996 by Abels et al³² as a carrier of Dutch II (ε^Gγ^Aγδβ)⁰-thalassaemia, however the deletion length was not determined at that time. Now the deletion length is estimated to be at least larger than 400 kb and the 5′ breakpoint located between position 5408246 and 5387552 (UCSC Genome Browser, May 2004) (fig 2A). Five deletions match the length and breakpoint locations of previously described deletions, two of which, the Croatian (εγδβ)⁰- (at least >108 kb) and the Filipino β⁰-thalassaemia (at least >45 kb), were incompletely mapped. More accurate length estimations were obtained by MLPA, being between 128–143 kb and 109–122 kb, respectively. The other three showed similarity to the Dutch I 12.6 kb β⁰-thalassaemia deletion (in seven independent patients of Dutch origin), the 13.4 kb Sicilian (δβ)⁰-thalassaemia deletion, which are also frequently found in the Mediterranean basin,^33,34 and the 32.6 kb Indian ^Gγ^Aγ(δβ)⁰-thalassaemia,³⁵ found in four independent chromosomes from Surinam-Hindustani subjects (fig 2B).

DISCUSSION

We describe the application of MLPA for high resolution mapping of deletions causing α- and β-thalassaemia. Using synthetic oligonucleotides, 35 loci along a genomic region of 700 kb from the tip of the short arm of chromosome 16, containing the α-globin gene cluster, could be analysed in two colours in a single reaction. More loci could be analysed simultaneously by using a third pair of amplification primers, labelled with a third fluorophore. This increased the number of probes to 50 loci spanning a genomic region of 500 kb on 11p15.4 and used to detect rearrangements causing β-thalassaemia or HPFH. Although slightly better results can be obtained when performing the PCR with the three sets of labelled universal primers separately, the ligation of all 50 probes was done in a single tube reaction. The fragment analysis was performed on a single sample of the three pooled PCR products per patient, which allowed the simultaneous analysis of 86 patient samples along with 10 normal controls in a 96 well format fragment analysis run on the ABI 3700.

The use of chemically synthesised oligonucleotides instead of cloning the half probes into M13 vectors, as originally described for MLPA,¹¹ allows cheap and rapid probe development, which increases the flexibility of MLPA for characterising genomic rearrangements. Only two out of 85 probes (2%) were excluded from further calculations due to standard deviations higher than 12% when tested on a validation set of 12 wildtype controls. The majority showed standard deviations of between 0.05 and 0.08. Although these deviations seem significant, please note that due to the probe density rearrangements are mostly detected using a series of flanking probes (>2).

The ability to detect rearrangements in both regions was tested using positive controls, heterozygous for the seven most common α-thalassaemia deletions confirmed by gap-PCR, and for two less frequent mutations, Dutch I and -α^7.9, confirmed by Southern blot analysis. By selecting 12 probes closely distributed along the 40 kb α-globin gene cluster, all of the common deletion types (except for the - -^FIL and - -^THAI) could be distinguished from each other by MLPA. In our eyes, the simplicity, work-load, and cost make MLPA a superior alternative to Southern blot analysis when a single technique is preferred for the detection of deletions causing α-thalassaemia in a research setting. When desired, gap-PCR can be used for independent confirmation. Similarly, six positive controls were selected, based on confirmation by different methods (Southern blotting and/or direct sequencing of break point fragments) and tested for the β-cluster probe set. All the probes expected to be deleted were confirmed in the heterozygotes tested.

In 19 and 20 samples large rearrangements involving the α- and β-globin genes, respectively, could be excluded. Point mutations or micro-deletions affecting expression and located in between the probes, would not be picked up by MLPA. However, since iron levels were not known for some patients and anaemia due to iron deficiency could easily be mistaken for α- or normal HbA₂ β-thalassaemia, we believe that negative samples may fall into this category.

Polymorphisms in the genome, interfering with probe annealing and ligation of the two probe pairs, may cause the loss of probe signal leading to a false positive MLPA result.^11,14,36 During the screening of patient samples suspected for α-thalassaemia, one case repeatedly showed the deletion of a single probe 5a (in fig 2A), named the (αα)^ZW deletion found in an adopted child. This probe was selected in a highly conserved region of the HS-40, not containing the polymorphic sites known to be present in human populations.³⁷ Deletion of this regulatory element is expected to cause severe down regulation of α-gene expression of the affected chromosome. Even though nothing can be said about the extent of deletion, the fact that the HbH inclusion bodies test was positive and that no other rearrangements involving the α-genes were found, is strongly in favour of a deletion involving the HS-40. Whether or not this deletion, which is at maximum 30 kb in length, also involves HS-33 as found by Higgs et al³⁸ needs further analysis. These types of deletions in human carriers may contribute to understanding the mechanisms involved in regulation of downstream α-gene expression³⁹ and will be studied further.

In conclusion, MLPA is an attractive alternative for FISH analysis for screening large deletions, for example in ATR-16 syndrome.^9,10 The tiling paths of cloned probes currently available for cytogenetic analysis of 16p13.3 and 11p15.4 are shown in figs 1 and 2. The distribution of synthetic probes coincides with the available cosmids, and allows a higher resolution of mapping than the available BAC or PAC probes. In contrast to in situ hybridisation, no laborious cell culture to generate metaphase spreads is necessary. MLPA can be performed directly on (stored) DNA samples.

MLPA uses standard technology only, that is, hybridisation, ligation, PCR, and capillary electrophoresis. Since most diagnostic laboratories have these technologies operational, implementation of MLPA should be straightforward. The robustness, simplicity, and intrinsic redundancy (probe density) of this approach, and the additional specificity offered by the ligation step, make MLPA an attractive technique for the detection and characterisation of copy number variation (deletions/duplications) in any region of the genome, particularly for high resolution analysis, and those regions not amenable to analysis by array comparative genomic hybridisation (array CGH).⁴⁰