Introduction

Pseudoxanthoma elasticum (PXE; OMIM#264800) is an autosomal recessive disorder, affecting the skin (yellowish papules, increased laxity in flexural areas), the eyes (peau d’orange, angioid streaks, retinal hemorrhage and vision loss) and the cardiovascular system (occlusive artery disease, gastrointestinal bleeding).1, 2 The disease results from abnormal calcification and fragmentation of elastic fibers in the middermis, in the Bruch's membrane of the retina and in the elastic laminae of blood vessels.3

PXE is caused by mutations in ABCC6 (OMIM#603234), a gene spanning 74 kb on chromosome 16p13.1. ABCC6 encodes an adenosine triphosphate-binding cassette transporter (subfamily C), previously referred to as multidrug resistance protein 6. This transmembrane protein is expressed mainly in the liver and kidney and to a much lower extent in the tissues affected in PXE. The substrate(s) transported by ABCC6 is unknown, and PXE is now considered a metabolic disease.4, 5

At present, >200 different ABCC6 mutations have been identified. These are primarily located at the 3′ end of ABCC6 between exons 24 and 30. Two of these mutations are particularly prevalent, a deletion of exon 23–29 (del23–29) and p.R1141X (c.3421C>T).6, 7, 8 To date, only 16 different large ABCC6 deletions (entire exons, whole gene deletions) have been identified in PXE patients.7, 9, 10, 11, 12, 13, 14 Nevertheless, ABCC6 is extremely prone to genomic rearrangements because of the high content of repetitive elements in all introns and in the genomic sequences surrounding the gene.14

We hypothesized that because of the documented instability of the ABCC6 genomic region, the unidentified mutant alleles remaining after direct sequencing and screening for the recurrent exon 23–29 deletion, may consist of deletions and/or insertions. In this study, we aimed to screen for the presence of such deletions and/or insertions using the multiplex ligation-dependent probe amplification (MLPA) technique. Our cohort consisted out of 35 patients with a clear-cut diagnosis of PXE but with only one or no ABCC6 mutations identified by current PCR-based techniques (Schouten et al., 2002). Recently, a PXE-like disease (OMIM#610842) that shares a significant phenotypical overlap with PXE was described. Therefore, patients whose genotypes remained incompletely ascertained after MLPA analysis were screened for the possible presence of mutations in GGCX (OMIM#137167), the gene responsible for the PXE-like syndrome.15

Materials and methods

Patients and samples

In a cohort of 331 clinical and biopsy-proven PXE patients of Belgian, French or Italian ancestry, an extended analysis of ABCC6 coding regions and exon/intron boundaries was performed by the traditional approach described earlier.16 After this analysis, we failed to identify ABCC6 mutations in one or both alleles in 35 patients (9 males, 26 females). In 29 of these patients, only a single ABCC6 mutation was found while we could not detect any ABCC6 mutation in 6 affected subjects, representing a total of 41 unidentified alleles.

The PXE diagnosis was made by experienced clinicians (ADP, OMV and LM). The clinical diagnosis was based on the presence of ophthalmological manifestations (including retinal peau d’orange and/or angioid streaks) and skin involvement (macroscopic skin lesions including yellowish papules and/or plaques in the neck and other flexural areas and microscopic skin lesions on full-thickness skin biopsy).17 All patients had a positive skin biopsy with calcification and fragmentation of elastic fibers.

This study was reviewed and approved by the Ethics Committee of the Ghent University Hospital and informed consent was obtained from all participating patients.

MLPA analysis

Genomic DNA was isolated from peripheral blood or fibroblasts using a Puregene kit or QIA-amp DNA extraction, respectively. For MLPA analysis, the commercially available SALSA reagent set p092B kit was used (MRC-Holland, Amsterdam, The Netherlands, http://www.mrc-holland.com). This kit contained 23 probes corresponding to ABCC6 exons 2, 4, 5, 7–15, 17, 18, 21–28 and 30 and 12 control probes for quality control. The p092B kit lacked probes for ABCC6 exons 1, 3, 6, 16, 19, 20, 29 and 31. As ABCC1 is in close proximity to ABCC6 (6.5 kb telomeric), an ABCC1 probe was also included. The construction of the kit precludes generation of signals from the ABCC6 pseudogenes.15

MLPA analysis was performed according to the manufacturer's recommendations (http://www.mlpa.com), using 100 ng of DNA in a 5-μl reaction. MLPA generated fragments were detected using an ABI 3730 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA, http://www3.appliedbiosystems.com) with ROX500 (Applied Biosystems, http://www3.appliedbiosystems.com) as an internal size standard. The genemapper software (Applied Biosystems, http://www3.appliedbiosystems.com) was used to calculate fragment size and concentration, whereas the quantification analysis was performed using Coffalyzer (MRC-Holland, http://www.mrc-holland.com). Coffalyzer is a robust analysis method, which performs quality checks and provides extensive information on the statistical significance of the obtained results. All samples were tested in duplicate.

Confirmation by probe complementarity experiments

During an MLPA experiment, exon-specific probes hybridize to the DNA sample. The presence of deletions and/or duplications can be detected by a decrease/increase in the annealing of probes. However, a decrease in probe/DNA binding can also be the result of differences in nucleotide sequence between MLPA probes and their target sequences because of genomic variations such as single nucleotide polymorphisms. This may give rise to false positive results. To ascertain the presence of single nucleotide polymorphisms, we sequenced the region covered by the probes using an ABI 3730 Genetic Analyzer (Applied Biosystems, http://www3.appliedbiosystems.com). Primers and methods were used as described in Hu et al. (2004). The nucleotide sequence of the probes was obtained from MRC-Holland.

Confirmation by long-range PCR experiments

Long-range PCR was performed using an Iproof high fidelity polymerase (Biorad, Nazareth, Belgium, http://www3.bio-rad.com). In a first step, primer sets flanking the deleted region were designed and optimized. PCR conditions for long-range amplification were as follows [98 °C 30’’ (98 °C 10’’, annealing temperature 30’’, 72 °C 3’)x32, 72 °C 10’]. After amplification, DNA fragments were visualized on a 0.8% agarose gel. Deletion break points were characterized by excision of shorter DNA fragments and direct sequencing. The sequences of the PCR products were compared with the ABCC6 genomic reference sequence (RefSeq NM_001171.5).

Confirmation by array comparative genomic hybridization

To confirm whole ABCC6 gene deletions, a Whole Human Genome Oligo Microarray 4 × 44 K (Agilent technologies, Santa Clara, CA, USA, http://www.agilent.com) was used with particular focus on chromosome 16. For smaller multi-exon deletions, an array specific to chromosome 16 with a high-probe density for the chr16: 15 687 873–17 130 000 region was designed. Probes were selected from the high-density comparative genomic hybridization database (HD CGH, Agilent Technology eArray; https://earray.chem.agilent.com/earray/). After submitting the selected probes to the UCSC genome browser (http://genome.ucsc.edu/), the probe density around all ABCC6 exons was checked. The presence of the ABCC6 pseudogenes rendered probe design more difficult for exons 1 through 9, with little or no probes being available in the HD CGH Database. In each hybridization experiment, patient and control DNA were labeled with CY3 (green) and CY5 (red), respectively.

GGCX sequencing

In those patients with an incomplete genotype following MLPA (n=28), the coding sequence of GGCX (RefSeq NM_000821.4) was verified using primers and PCR conditions as described earlier.18 Patients harboring homozygous or compound heterozygous mutations in specific regions of GGCX develop the PXE-like syndrome, which associates skin and eye features of PXE with a generalized cutis laxa and vitamin K-dependent clotting deficiency.19 The disease potential of identified variants was estimated by using the Polyphen and SIFT tools (http://genetics.bwh.harvard.edu/pph/, http://blocks.fhcrc.org/sift/SIFT.html).

Results

MLPA results

To identify deletions and insertions in ABCC6, we performed MLPA in PXE patients with unidentified mutant alleles. Of the 41 unknown disease alleles from 35 patients, 11 alleles could be further identified by MLPA. These included one intragenic multi-exon deletion, four whole gene deletions and six single-exon deletions (Figure 1). The multi-exon deletion in patient 3 encompassed exons 24 to 27. The four whole gene deletions in patients 13, 20, 26 and 34 expanded beyond ABCC6 as indicated by the 0.5 ratio for the ABCC1 control probe. The single-exon deletions in patients 4, 6, 7, 16, 28 and 32 involved exons 30, 14, 2, 9, 24 and 30, respectively. All these exons encode an intracellular domain of ABCC6, except for exon 2, which corresponds to a transmembrane segment as predicted by the human protein reference database.

Figure 1
figure 1

Overview of MLPA results. Every bar is the ratio result of 1 probe pair PCR product. From left to right: ratio for the ABCC1 probe, ABCC6 exon 30, exon 28–21, exon 18–17, exon 15–7, exon 5, exon 4, exon 2, followed by 12 bars representing the control probes (contr.). The ABCC1 control probe is indicated with an asterisk. All ratios for the negative control sample are 1, indicating that this individual does not show deletions/duplications. As expected, a 0.5 ratio is observed for exon 23, 24, 25, 26, 27 and 28 for the positive control sample. Exon 29 is not incorporated in the kit. A ratio of 1.0 for the ABCC1 control probe (*) is observed for the positive control sample. Each patient with a positive MLPA result is represented in the figure; patient 3: deletion exon 24–27, patient 4: deletion exon 30, patient 6: deletion exon 14, patient 7: deletion exon 2, patient 13: whole ABCC6 gene deletion + deletion ABCC1 gene, patient 16: deletion exon 9, patient 20: whole ABCC6 gene deletion + deletion ABCC1 gene, patient 26: whole ABCC6 gene deletion + deletion ABCC1 gene, patient 28: deletion exon 24, patient 32: deletion exon 30, patient 34: whole ABCC6 gene deletion + deletion ABCC1 gene.

Confirmation of the intragenic multi-exon deletion and the whole gene deletions

The multi-exon and whole gene deletions were confirmed by array comparative genomic hybridization (aCGH) in all patients except patient 20 for whom no appropriate DNA for array analysis was available. The aCGH analysis enabled simultaneously breakpoint determination for patients 3, 13, 26 and 34 (Figure 2a). The minimum/maximum length of the deletions were 5.5–10.5 kb, 784–1642 kb, 1257–1955 kb and 2756–2958 kb, respectively (Table 1). The exact breakpoints of patient 3 could be determined by long-range PCR and measured 8118 bp (c.3307−1006_3882+1582del) (Table 2).

Figure 2
figure 2

Array CGH based results. Red and green lines represent cut-off values and the chromosomal region is indicated underneath each panel. (a) Results for patients 3, 13, 26 and 34 with a multi-exon deletion. Patient 3: deletion exon 24–27. Region chr.16: 16 156 390–16 166 980, where ABCC6 exons 24–27 are located, exhibit decreased DNA hybridization for patients’ 3 DNA (ratio -1). Patients 13, 26, 34: whole ABCC6 gene deletion. A -1 ratio is observed for region chr16:15 399 818–16 183 616, chr16: 14 956 252–16 213 237 and chr16: 15 164 187–17 919 962, respectively, suggesting these three patients have a deletion of several genes surrounding the ABCC6 gene, as further detailed in Table 1. (b) Result for patient 32: deletion exon 30–31. A -1 ratio is observed for region chr.16: 16 150 211–16 155 985.

Table 1 Overview of aCGH results
Table 2 Summary of MLPA results and subsequent confirmation strategy for deletions detected by MLPA

Confirmation of the single-exon deletions

To exclude false positive MLPA results, sequence similarities between the MLPA probes and the patient's DNA were compared in the case of single-exon deletions. The sequences were identical for patients 4, 16 and 32, validating our results (Table 2). Patient 7 exhibited a homozygous nucleotide variant in the probe-binding site. However, we verified that this nucleotide change had no influence on probe-binding efficacy, with an MLPA experiment using a control sample containing the same variant (data not shown). In contrast, a false positive MLPA result was obtained as a consequence of the presence of a known heterozygous point mutation in the probe-annealing region in patient 6 (c.1798C>T; p.R600C). The presence of a heterozygous variant in the same region demonstrates that the patient is not hemizygous for this region (Table 2).

In patient 32, an aCGH experiment could be performed, demonstrating that the deletion comprises exon 30 and 31 and is 5–6 kb in length (Figure 2b). For the other patients, no fresh DNA was available for aCGH analysis.

To characterize the deletion in patient 28, long-range PCR was carried out to amplify across the deleted region and the resulting PCR product was sequenced. The deletion measured 1754 bp (c.3307−904_3506+660del) (Table 2). Unfortunately, our various attempts to determine the breakpoints for patients 4, 7 and 16 using many different primer pairs were unsuccessful.

Finally, we performed MLPA analysis in eight patients with a previously identified apparently homozygous p.R1141X mutation (exon 24) to verify whether the subjects were indeed homozygotes or whether this variant was paired with a deletion of the exon 24 region. Surprisingly, we identified deletions removing the region of exon 24 in two of these patients, indicating that a non-negligible proportion of patients with R1141X were in fact compound heterozygous.

Overall, out of 35 patients with incomplete genotypes, MLPA analysis allowed us to obtain the complete genotype for 7 patients (20%) and a partial characterization for 25 affected individuals, whereas the disease-causing alleles remained undetected in only 3 subjects with biopsy-proven diagnosis.

Subsequently, the DNA samples of the 28 patients remaining with an incomplete genotype after MLPA were screened for possible nucleotide changes in GGCX that could account for the missing causative alleles. As no MLPA probes were available to search for GGCX deletions, we used direct sequencing of exonic regions only. No mutation was detected, though a heterozygous variant was found in exon 10 (p.S452 T) for one patient. This nucleotide variant was not present in 100 control samples (200 alleles) but was predicted to be benign by in silico analysis.

Discussion

PXE, a heritable disorder affecting the skin, eyes and the cardiovascular system, is caused by mutations in the ABCC6 gene.3, 10, 20, 21 Of the various techniques applied to mutation analysis of ABCC6, the currently prevailing and most efficient technique is direct sequencing, with a mutation detection rate of about 90%. Various deletions involving parts or the whole ABCC6 gene have been reported earlier.7, 10, 11, 12, 13, 14 Small or large deletions are expected to make up for the bulk of the unidentified alleles because heterozygous middle-sized deletions are notoriously difficult to detect with traditional PCR-based assays. Furthermore, the presence of numerous repetitive elements makes the ABCC6 region subject to genomic rearrangements.

Deletions often originate from homologous recombination between identical sequences of repeated DNA. Several types of repeats (long and short interspersed nuclear elements, Alu repeats and so on) are abundantly present in the intra- and extragenic region of ABCC6 (UCSC genome browser, http://genome.ucsc.edu/).11, 22 Some of them are strongly suspected to have caused the recurring 23–29 deletion and the unique exon 15 deletion.12, 14 Similarly, such rearrangements may have been responsible for the existence of the two ABCC6 pseudogenes, which further illustrates the recombinational potential of this region of chromosome 16. Indeed, Cai et al. have shown that the ABCC6 pseudogenes contribute to mutations in the parent gene.22, 23 Until now, a total of 16 different large deletions (exons, whole gene deletions) have been described.7, 9, 10, 11, 12, 13, 14 Because of the wide variety of repeat elements in the ABCC6 region, we anticipated that several of the unidentified mutant alleles in our PXE cohort would consist of deletions and insertions.

In our patients with incomplete genotypes, we identified a deletion in 25% of the uncharacterized alleles. Nine of the deletions were novel, hereby increasing the spectrum of known large ABCC6 deletions from 16 to 25.

Of the 10 patients in whom a deletion was detected, four carried a deletion removing the whole ABCC6 gene. In these four patients, the absence of ABCC1 was also observed, indicating that these deletions were not restricted to a single gene. ABCC1 is located 6.5 kb telomeric to ABCC6 and encodes a plasma membrane drug-efflux pump (ABCC1) closely related to ABCC6. Array CGH revealed that the whole gene deletions were of variable length (784–1642 kb, 1257–1955 kb and 2756–2958 kb). Moreover, aCGH showed the absence of several other genes as summarized in Table 1, indicating that the deletions extend far beyond ABCC1–ABCC6. The deletion of these genes was not associated with additional phenotypic characteristics in these PXE patients, which can be explained by the fact that an intact copy of the deleted genes still is present on the other chromosome.

Besides expanding the ABCC6 mutation spectrum, our results have several practical implications. First, by increasing the mutation detection rate, familial screening and genetic counseling can be improved. Second, the identification of deletions is of relevance for the interpretation of direct sequencing results. Every patient diagnosed with a homozygous ABCC6 mutation could in fact be compound heterozygous for that particular mutation and a deletion of the corresponding region. In the molecular analysis of ABCC6, verification of the multi-exon deletion 23–29 and analysis of other recurrent mutations followed by sequencing of the whole coding region remain the first essential steps.2 If after this approach, one or more mutant alleles remain undetected or in the case of a homozygous mutation (and unavailable parents), MLPA can be applied to search for deletions or duplications.

Because of the phenotypic overlap between PXE and the PXE-like syndrome, we analyzed the GGCX gene in the 28 patients remaining with an incomplete genotype after MLPA but did not detect mutations nor functional polymorphisms in this gene.

One may note that the MLPA kit we used in this study covered only 23 of the 31 exons of ABCC6. Nevertheless, the mutational screening we applied was quite effective, as we identified 10 deletions, 9 of them being novel. By adding the missing eight ABCC6 exon probes (exons 1, 3, 6, 16, 19, 20, 29, 31), the deletion detection will be further improved. Especially, the addition of an exon 29 probe would be of interest, as many mutations have been described in this functionally important protein region.2 Although other techniques, which are comparable in costs and time of handling, can detect genomic deletions (for example QMPSF), MLPA is a validated technique, which is accompanied by a user-friendly software program (Coffalyzer) to determine the statistical significance of results.

In conclusion, our approach has increased the ABCC6 mutation detection rate, the mutation spectrum and has also confirmed that a second gene locus for PXE is unlikely. In addition, our results underline the importance of genomic instability in the ABCC6 region. Furthermore, we have demonstrated the heterogeneous nature of the ABCC6 whole gene deletions and have shown that most of these expand far beyond the ABCC6 genomic region. Our results clearly showed that deletions of various sizes constitute a sizable proportion of ABCC6 mutations accounting for some of the previously undetectable alleles. Furthermore, we showed that in certain cases, the mutation status of patients may erroneously be interpreted as homozygous because of the technical limitation of direct sequencing. As it is now evident that small deletions are more common than previously thought, we propose MLPA as an efficient complementary technique to ABCC6 molecular diagnosis.

Conflict of interest

The authors declare no conflict of interest.