Introduction

Adrenoleukodystrophy (ALD) is a demyelinating disease caused by mutations of ABCD1 [1]. This disease affects primarily the white matter of the central nervous system occasionally accompanied by adrenal insufficiency [2–4]. Diagnosis of ALD is usually made by the increased contents of very long chain saturated fatty acids (VLCFAs; >C22:0) in plasma as well as by mutational analysis of ABCD1 [5–7]. Since 15% of obligate female carriers have normal VLCFA levels [7], mutational analysis is essential for the diagnosis of the carriers. Since the first report of allogenic HSCT for childhood ALD, there has been an increasing number of reports showing efficacies of HSCT for the childhood cerebral form of ALD, if HSCT is performed at early stages of the disease [8–10]. Thus, availability of rapid molecular diagnosis for patients with ALD and carriers is mandatory in the clinical practice for ALD.

ALD is characterized by a broad spectrum of clinical presentations including childhood cerebral form, adrenomyeloneuropathy (AMN), AMN complicated by cerebral demyelination, adulthood cerebral form, and Addison disease. From clinical experience, patients with different clinical phenotypes can be observed even in a single pedigree. In support of this, no clear genotype–phenotype correlations have been observed [11–16], raising the possibility that other genetic or environmental factors are involved in the pleiomorphic clinical presentations of ALD.

ABCD1 gene encodes a half-ATP-binding cassette (ABC) transporter, adrenoleukodystrophy protein (ALDP), which is localized to the peroxisomal membrane. ABCD2, ABCD3, and ABCD4 genes are the closest homologues of the ABCD1 gene [17, 18]. It has been shown that the majority of mouse liver ALDP and the 70-kDa peroxisomal membrane protein (PMP70) that is encoded by ABCD3 are homomeric proteins [19]. Furthermore, it has been shown that ALDP can form homodimers or a heterodimer with the adrenoleukodystrophy-related protein (ALDR) that is encoded by ABCD2 or the PMP70 that is encoded by ABCD3 [16, 20–22], raising the possibility that these ABCD1-related genes function as disease-modifying genes for ALD.

To provide a rapid mutational analysis for ALD, we developed a microarray-based high-throughput resequencing system of ABCD1 (TKYPD01) [23]. Furthermore, to explore the possibility that these ABCD1-related genes function as disease-modifying genes, we established a comprehensive resequencing system for ABCD1-related genes, ABCD2, ABCD3, and ABCD4. On the basis of the comprehensive resequencing of ABCD1, ABCD2, ABCD3, and ABCD4 genes, we identified 11 novel single nucleotide polymorphism (SNPs). Using these novel SNPs as well as previously described SNPs of these genes, we conducted detailed association studies of these SNPs with the clinical phenotypes of ALD.

Materials and methods

Participants

Forty Japanese ALD patients, consisting of 14 patients with childhood cerebral ALD (CCALD), 8 patients with adulthood cerebral ALD (AdultCer), 2 patients with AMN with later development of cerebral ALD (AMN-Cer), 13 patients with AMN, 1 asymptomatic patient, and 2 patients with unknown form, were enrolled in this study. Among the patients, mutations were previously identified in 16 ALD patients by direct nucleotide sequence analysis, while no mutational analyses were conducted for 24 patients.

For replication studies of the results of association studies on Japanese ALD patients showing potential associations of SNPs in ABCD1-related genes with ALD phenotypes, an independent French ALD cohort with well-defined extreme phenotypes consisting of 118 patients with CCALD and 71 patients with pure AMN (AMN with age >45 years as well as with normal brain magnetic resonance imaging) was studied. In addition, 51 ALD patients with AMN-Cer were also analyzed in the French ALD cohort.

Procedures

Primers specific for ABCD1, ABCD2, ABCD3, and ABCD4 were designed using BLAST search and Smith–Waterman method to avoid amplification of the related homologous genes (Fig. 1; ESM Tables 1, 2, 3, and 4). In particular, since there were many segments homologous to exons 8, 9, and 10 of ABCD1, a specific primer pair was designed (Fig. 1). Fifty nanograms of genomic DNA were subjected to polymerase chain reaction (PCR) amplification in a total volume of 50 μL. The PCR conditions were as follows: 94°C for 1 min, followed by five cycles consisting of 94°C for 30 s, 62°C for 30 s, and 68°C for 2 min; five cycles consisting of 94°C for 30 s, 60°C for 30 s, and 68°C for 2 min; and 25 cycles consisting of 94°C for 30 s, 58°C for 30 s, and 68°C for 2 min, followed by a final extension at 68°C for 7 min, using the LA Taq with GC Buffer PCR system (Takara Bio, Otsu, Shiga, Japan).

Fig. 1
figure 1

Primer design for ABCD1. All the exons of ABCD1 were amplified using six primer pairs. There were pseudogenes at 2p11, 10p11, 16p11, and 22q11, which were similar in sequence to exons 7–10 of ABCD1 gene (92–96%). The forward primer for exons 8–10 was designed to avoid amplification of the related homologous genes. We could design a specific reverse primer for exons 8–10

Full size image

Resequencing DNA microarrays were used in the analyses of the sequences of ABCD1 (TKYPD01) and ABCD2 (TKYAD01). TKYPD01 and TKYAD01 were designed using the platform of GeneChip CustomSeqTM Resequencing Microarray (Affymetrix, Santa Clara, CA, USA). Since there are substantial homologies between ABCD1 and ABCD2, these genes were placed in independent microarrays (TKYPD01 and TKYAD01). Each PCR product of ABCD1 and ABCD2 was quantified using PicoGreen (Molecular Probes, Eugene, OR, USA) and equimolarly pooled. Pooled PCR products of ABCD1 and ABCD2 were fragmented using DNAse I, labeled with biotin, hybridized to DNA microarrays, and subjected to scan and analyses of nucleotide sequences of ABCD1 (TKYPD01) and ABCD2 (TKYAD01) according to the manufacturer's instructions (Affymetrix, Santa Clara, CA, USA). The base calls that were undetermined using the GDAS analysis software (Affymetrix, Santa Clara, CA, USA) were further analyzed by manual inspection. Identified mutations and SNPs were confirmed by the direct nucleotide sequence analysis of the PCR products. All the PCR products of ABCD3 and ABCD4 were analyzed by the direct nucleotide sequence analysis. Identified SNPs of ABCD2, ABCD3, and ABCD4 were examined as to whether they were novel SNPs or known SNPs using the J SNP (http://snp.ims.u-tokyo.ac.jp/index_ja.html) and DB SNP (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=Snp&cmd=Limits).

Statistical analyses

We compared the allele frequencies of detected SNPs between the subgroups of ALD or between the individual subgroup and the controls by Fisher's exact test using the JMP 7 software (SAS Institute, Cary, NC, USA). Deviation of the SNP genotypes from the Hardy–Weinberg equilibrium was evaluated using the PEDSTATS program [24]. Linkage disequilibria among the neighboring SNPs were evaluated using Haploview version 4.1 [25].

Results

Resequencing DNA microarray-based mutational analysis of ABCD1 gene in Japanese ALD patients

All the mutations of ABCD1 were clearly identified using the resequencing DNA microarray system including 26 missense, 2 nonsense, and 12 insertion/deletion mutations of ABCD1 (Figs. 2 and 3; Tables 1 and 2). Mutations of ABCD1 gene were widely scattered in the entire region of ABCD1 gene (Fig. 3; Tables 1 and 2). All types of ABCD1 mutation were distributed among all the phenotypes of ALD (Fig. 3; Tables 1 and 2). Among the 40 mutations, 11 mutations were novel (Tables 1 and 2). Among the deletion/frameshift mutations that are expected to result in complete loss of ALDP functions, the mutations were distributed among all the phenotypes of ALD (Tables 1 and 2), supporting the previous observations of no genotype–phenotype correlations.

Fig. 2
figure 2

Scan data of the resequencing DNA microarray and sequence data of the direct nucleotide sequence analysis (upper panel: patient, lower panel: control). Each column shows a base position, and each row shows a base call DNA in the scan data of the resequencing DNA microarray. Here, a mutation (G277R) was detected, and the signal intensities around the mutation were reduced because of the mismatch of the mutation site. The sequence data of the direct nucleotide sequence analysis confirmed the scan data of the resequencing DNA microarray

Full size image
Fig. 3
figure 3

Identified mutations of ABCD1. Mutations of ABCD1 gene were widely scattered in the entire region of ABCD1 gene. All types of ABCD1 mutations were distributed among all the phenotypes of adrenoleukodystrophy. TM transmembrane domain, EAA-like EAA-like protein motif, Walker A Walker A motif, C sequence nucleotide binding fold conserved sequence, Walker B Walker B motif, fs frameshift

Full size image
Table 1 Identified ABCD1 mutations: mutations of ABCD1 that result in devastating effects (frame shifts or nonsense mutations) on adrenoleukodystrophy protein (ALDP)
Full size table
Table 2 Identified ABCD1 mutations: mutations of ABCD1 that result in amino acid substitutions or in-frame deletions
Full size table

Identification of SNPs of ABCD2, ABCD3, and ABCD4 genes by comprehensive resequencing

Comprehensive resequencing of ABCD2, ABCD3, and ABCD4 genes of the 40 Japanese patients with ALD revealed two novel SNPs, nine SNPs (six known and three novel SNPs), and 13 SNPs (seven known and six novel SNPs), respectively (Fig. 4; Tables 3, 4, 5, and 6; ESM Table 5). Hardy–Weinberg equilibrium was fulfilled for each SNP. The five known SNPs (rs17782508, rs2301345, rs4148077, rs4148078, and rs3742801) were in complete linkage disequilibrium in the Japanese patients with ALD as well as in the controls, as determined using Haploview version 4.1 (Fig. 4).

Fig. 4
figure 4

Identified single nucleotide polymorphisms (SNPs) of ABCD2, ABCD3, and ABCD4 (upper panel). Comprehensive resequencing of ABCD2, ABCD3, and ABCD4 genes of the 40 patients with adrenoleukodystrophy (ALD) revealed two novel SNPs, nine SNPs (six known and three novel SNPs), and 13 SNPs (seven known and six novel SNPs), respectively. Red characters indicate the novel SNPs, blue characters indicate the SNPs identified in the coding region, and black characters indicate the SNPs identified in the noncoding region. Linkage disequilibrium (LD) map of SNPs of ABCD4 in Japanese patients with ALD and the controls using the Haploview version 4.1 (lower panel). The five known SNPs (rs17782508, rs2301345, rs4148077, rs4148078, and rs3742801) were in complete disequilibrium in Japanese patients with ALD and the controls (LOD = 43.97, r 2 = 1.0, D′ = 1.0). Novel SNP7 and the five known SNPs (rs17782508, rs2301345, rs4148077, rs4148078, and rs3742801) were not in strong disequilibrium in Japanese patients with ALD and the controls (LOD = 1.15, r 2 = 0.037, D′ = 0.706), although novel SNP7 and the five known SNPs (rs17782508, rs2301345, rs4148077, rs4148078, and rs3742801) were strong disequilibrium only in Japanese patients with ALD (LOD = 2.02, r 2 = 0.221, D′ = 1.0). The number in the box indicates the data of D′. The color of the box is determined from the LOD score and D′. The block was determined using a confidence interval algorithm [33]

Full size image
Table 3 Summary of identified single nucleotide polymorphism (SNPs) of ABCD2, ABCD3, and ABCD4 in 40 adrenoleukodystrophy patients: novel SNPs
Full size table
Table 4 Summary of identified single nucleotide polymorphism (SNPs) of ABCD2, ABCD3, and ABCD4 in 40 adrenoleukodystrophy patients: known SNPs
Full size table
Table 5 Association studies of detected single nucleotide polymorphism (SNPs) with the clinical phenotypes of Japanese adrenoleukodystrophy patients: novel SNPs
Full size table
Table 6 Association studies of detected single nucleotide polymorphism (SNPs) with the clinical phenotypes of Japanese adrenoleukodystrophy patients: known SNPs
Full size table

Association studies of SNPs of ABCD2, ABCD3, and ABCD4 with the clinical phenotypes of ALD

Using the 11 novel SNPs and 13 previously described SNPs in ABCD2, ABCD3, and ABCD4, we conducted association studies of these SNPs with the clinical phenotypes of ALD (Tables 5 and 6).

For ABCD2, we analyzed two novel SNPs (novel SNP1 and novel SNP2). There were no significant differences in the allele frequencies between patients with cerebral form and those with AMN, or between the patients with individual phenotypes of ALD and the controls. For ABCD3, we analyzed three novel SNPs (novel SNP3, novel SNP4, and novel SNP5) and six previously described SNPs (rs4148058, rs2147794, rs16946, rs681187, rs662813, and rs337592). However, we did not also detect any significant associations.

For ABCD4, we analyzed six novel SNPs (novel SNP6, novel SNP7, novel SNP8, novel SNP9, novel SNP10, and novel SNP11) and seven previously described SNPs (rs17782508, rs17182959, rs17158118, rs2301345, rs4148077, rs4148078, and rs3742801). Interestingly, the five previously described SNPs (rs17782508, rs2301345, rs4148077, rs4148078, and rs3742801) that are in complete linkage disequilibrium were significantly less frequently represented in the patients with Japanese AMN than in the controls in the Japanese population (p = 0.0468), whereas there were no significant differences in the Japanese patients with the cerebral form compared with the controls (Tables 5 and 6).

Given the significant association of the five SNPs (rs17782508, rs2301345, rs4148077, rs4148078, and rs3742801) with the phenotypes of AMN, we then conducted a replication study on an independent French ALD cohort with extreme phenotypes (117 CCALD cases and 71 pure AMN cases). However, we did not find any significant association of these five SNPs with AMN or CCALD. Interestingly, the combination of two intronic SNPs (A (rs17182959) and G (rs7158118)) was significantly more frequently represented in the 51 patients with AMN-Cer than in those with CCALD in the French ALD cohort (p = 0.0049). The combination of these two intronic SNPs (A (rs17182959) and G (rs7158118)), however, was not present in any of the Japanese patients with ALD, although one combination of two intronic SNPs (A (rs17182959) and G (rs7158118)) was present in the Japanese controls (Tables 5 and 6 ; ESM Table 5).

Discussion

Our microarray-based high-throughput mutational analysis system was accurate to detect all the mutations, which were confirmed by direct nucleotide sequence analysis. This system should be highly useful for the mutational analysis of ABCD1 for the diagnosis of patients with ALD, and the diagnosis of the carriers with ALD.

Although reverse transcription (RT)-PCR has been preferentially used for the analysis of ABCD1 gene [11, 12, 26] to overcome the difficulty of specifically amplifying the ABCD1 gene owing to the existence of the related highly homologous genes [13, 27], primers allowing the specific amplification of ABCD1 enable the PCR analysis of genomic DNA, which is much easier than RT-PCR analysis. Similar approaches have also been used for SSCP-based [13, 28] and DNA-based diagnostic testing methods [29].

In the Japanese ALD patients, mutations of ABCD1 gene were widely scattered in the entire region of ABCD1 gene. All types of ABCD1 mutation were distributed among all the phenotypes of ALD, including childhood cerebral form, AMN, and adulthood cerebral form, suggesting that there is no association of a particular phenotype of ALD with individual mutations as previously observed in other ALD populations [1, 11–13, 15] (http://www.x-ald.nl/). Even among the frameshift mutations that are clearly expected to cause a complete loss of ALDP functions, such ABCD1 mutations were distributed among all the phenotypes of ALD.

On the basis of the comprehensive resequencing of ABCD2, ABCD3, and ABCD4 genes, we searched for SNPs of these related genes to explore the possibility of these genes as candidate disease-modifying genes for ALD, although it was shown that ALD phenotypes are independent of ABCD2 genotype in two independent association studies of ABCD2 polymorphisms and ALD phenotypes [30]. Although our study did not reveal SNPs significantly associated with the clinical phenotypes irrespective of the ethnic background, the SNPs in ABCD4 with suggestive association in the Japanese patients (rs17782508, rs2301345, rs4148077, rs4148078, and rs3742801) and French patients (rs17182959 and rs7158118) may still deserve further investigation including association studies on other independent cohorts and studies on the biological effects related to these SNPs. Among the SNPs with suggestive association in the Japanese patients (rs17782508, rs2301345, rs4148077, rs4148078, and rs3742801), rs4148077 (A304T) substitutes a hydrophilic amino acid for a hydrophobic amino acid, and rs3742801 (E368K) substitutes a basic amino acid for an acidic amino acid. It would be interesting to investigate if these amino acid substitutions may have relevance to the function of ABCD4.

In this study, we identified as many as 11 novel SNPs in ABCD2, ABCD3, and ABCD4 genes in addition to the 13 previously described SNPs. These findings indicate that there are still numerous novel SNPs with the number comparable to that of previously described SNPs; furthermore, this study places great emphasis on the role of comprehensive resequencing in the discovery of novel SNPs in relevant genes. The novel SNPs as well as previously described ones in ABCD2, ABCD3, and ABCD4 genes should be useful for further association studies on ALD and other peroxisome diseases and on the biological implications associated with these SNPs.

The diverse phenotypic variations of ALD still remain enigmatic. Recent studies suggest the role of perosixomes of oligodendrocytes in axonal loss and neuroinflammation [31] and microglial apoptosis as an early pathogenic change in CCALD [32]. With the advancement in our understanding of the pathophysiology of ALD, we hope that we can further probe into the disease-modifying factors on the basis of the molecular pathogenesis of ALD. Genome-wide association studies may well serve as an alternative approach for the identification of disease-modifying genetic factors.