Altered PLP1 splicing causes hypomyelination of early myelinating structures.

OBJECTIVE: The objective of this study was to investigate the genetic etiology of the X-linked disorder "Hypomyelination of Early Myelinating Structures" (HEMS).
METHODS: We included 16 patients from 10 families diagnosed with HEMS by brain MRI criteria. Exome sequencing was used to search for causal mutations. In silico analysis of effects of the mutations on splicing and RNA folding was performed. In vitro gene splicing was examined in RNA from patients' fibroblasts and an immortalized immature oligodendrocyte cell line after transfection with mutant minigene splicing constructs.
RESULTS: All patients had unusual hemizygous mutations of PLP1 located in exon 3B (one deletion, one missense and two silent), which is spliced out in isoform DM20, or in intron 3 (five mutations). The deletion led to truncation of PLP1, but not DM20. Four mutations were predicted to affect PLP1/DM20 alternative splicing by creating exonic splicing silencer motifs or new splice donor sites or by affecting the local RNA structure of the PLP1 splice donor site. Four deep intronic mutations were predicted to destabilize a long-distance interaction structure in the secondary PLP1 RNA fragment involved in regulating PLP1/DM20 alternative splicing. Splicing studies in fibroblasts and transfected cells confirmed a decreased PLP1/DM20 ratio.
INTERPRETATION: Brain structures that normally myelinate early are poorly myelinated in HEMS, while they are the best myelinated structures in Pelizaeus-Merzbacher disease, also caused by PLP1 alterations. Our data extend the phenotypic spectrum of PLP1-related disorders indicating that normal PLP1/DM20 alternative splicing is essential for early myelination and support the need to include intron 3 in diagnostic sequencing.

Introduction

Among the childhood leukodystrophies, hypomyelinating disorders constitute a large, highly heterogeneous group of patients, many of whom remain without a genetically confirmed diagnosis.1–4 Using magnetic resonance imaging (MRI) pattern recognition analysis, we previously identified a novel hypomyelinating disorder in four male patients in which hypomyelination is specifically pronounced in early myelinating structures.1 As this distribution of hypomyelination is different from other hypomyelinating disorders, in which these early myelinating structures as a rule contain more myelin than the later myelinating structures,4,5 we proposed a new disease called “Hypomyelination of Early Myelinating Structures” (HEMS).1 Family history suggested X-linked inheritance, supported by the report of two brothers with the same clinical picture and MRI pattern.1,6 The clinical phenotype of HEMS patients resembled that of other well-known hypomyelinating disorders, with onset of symptoms in late infancy, including ataxia and increasing spasticity, and relatively preserved cognition.1,4,7–9 We hypothesized that HEMS could be caused by mutations of an X-chromosomal gene involved in the regulation of early myelination.1,6 Diagnostic Sanger sequencing for known causes of hypomyelination, including Pelizaeus–Merzbacher disease (PMD), caused by X-linked PLP1 mutations, and Pelizaeus–Merzbacher-like disease (PMLD), caused by GJC2 mutations, had been unrevealing.1

The combination of MRI pattern recognition analysis, which is used for the categorization of homogeneous groups of patients with an unclassified white matter disorder and exome sequencing has been shown to be successful in identifying novel disease genes and new phenotypes associated with known disease genes.10–12 In this study, we used the same approach and ascertained mutations in a specific region of PLP1 that had initially not been identified. PLP1 is located on the X-chromosome, contains seven exons and encodes both proteolipid protein 1 (PLP1) and its smaller isoform DM20 that is derived by the use of an alternative splice donor site within exon 3.13 In all HEMS patients, the PLP1 mutations are located either in the PLP1-specific region encoded by exon 3B that is spliced out in DM20 or in intron 3. Using in silico splicing prediction programs, in silico analysis of predicted secondary RNA structures, and in vitro analysis of gene splicing in RNA prepared from patients’ fibroblasts and transfection studies, we show that these mutations play a role in alternative splicing of PLP1.

Patients and Methods

MRI studies and clinical examination

We included 16 male patients from 10 unrelated families (Table1). Thirteen of the 16 patients were included based on their specific MRI pattern compatible with HEMS.1 Seven of these 13 patients were identified from our MRI-database of over 3000 cases with an unclassified leukoencephalopathy using MRI pattern recognition analysis,3 and six patients were previously identified and described (patients 8, 9, 11, and 12,1 and patients 4 and 106). S. H. K. and N. I. W. evaluated the MRIs according to a published protocol.3 Three affected male siblings (patients 5, 15, and 16) were included without MR images. Clinical and laboratory investigations were retrospectively reviewed. Diagnostic Sanger sequencing of PLP1 in patients 1, 2, 3, 8, 9, 11, 12, and 14 had previously been performed in diverse laboratories and reported unrevealing.

Table 1

PLP1 mutations, in silico predictions and PLP1/DM20 ratio.

Patients	c.DNA	Protein	Mother carrier	Nuc. conservation1	In silico predictions						In vitro studies
					Missense prediction	Splicing effects			Mfold analysis		PLP1/DM20 ratio of normal
					SIFT/poly-phen2/mutation Taster	Predicted new splice sites4	Predicted change in strength natural PLP1 splice donor site4	Predicted change number ESE5/ESS6 motifs	LDIS-5′ & LDIS-3′ normal [ΔG]-mutant [ΔG]7	Predicted conformational changes PLP1 splice donor site exon 3B	Transfection study	Patients’ fibroblasts
1	c.380_392del	p.(Arg127Lysfs*16)	Yes	1	n.a.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.
48,9,108,9	c.404T>G	p.(Leu135Trp)	Yes	1	D/B/B	No	No	+2 ESS	n.d.	Yes	n.d.	n.d.
139, 149	c.436C>T	p.=	Yes	1	n.a.	Yes10	No	+2 ESS	n.d.	No	0.18	n.d.
811	c.441A>T	p.=	No	1	n.a.	Inconclusive12	No	+1 ESS	n.d.	Yes	0.14	0.01
6	c.453+7A>G	p.?	n.d.	1	n.a.	Inconclusive13	Inconclusive14	+1 ESE (sc35)	n.d.	Yes	0.05	0.06
215,315,59, 1111,91515,1615	c.453+159G>A	p.?	2,3,15,16 n.d.;5&11 yes	1	n.a.	No	No	No	+5.1	n.d.	n.d.	n.d.
7	c.453+164G>A	p.?	No	1	n.a.	Inconclusive16	No	Change strength ESE (sc35)	+5.9	n.d.	n.d.	n.d.
911	c.454−312C>G	p.?	Yes	1	n.a.	No	No	+3 ESS, +1 ESE (sc35)	+4.1	n.d.	n.d.	n.d.
1211	c.454−314T>G	p.?	No	1	n.a.	No	No	No	+1.6	n.d.	0.1217	n.d.

For a detailed description of methods and programs used see Data S1. c.DNA, complementary DNA; nuc., nucleotide; n.a., not applicable; n.d., not done; B, benign; D, damaging; LDIS-5′, long-distance interaction site 5′; LDIS-3′, long-distance interaction site 3′; SC35, serine/arginine-rich splicing factor 2.

Assessed using Phastcons scores (0 = no conservation, 1 = high conservation).

Predicted new splice donor or acceptor sites. Yes = significant change, inconclusive = inconclusive change.

Predicted change in number exonic splicing enhancers (ESE) motifs identified by ESE Finder 3.0.

Predicted change in number exonic splicing silencer (ESS) motifs identified by FAS-ESS web server using the FAS-hex2 set.

Difference of [ΔG] = minimal Gibbs energy, free energy, kcal/mole of the intra-intronic RNA structure fragment between the normal and mutant. Normal [ΔG] = −17.6 kcal/mole.

Patients previously published by Tonduti et al.6

Sibling pairs are as follows: 4 and 10, 5 and 11, and 13 and 14.

Predicted strength of new splice donor site at c.434 of 87 by Human splice site finder (HSF) (normal range 0–100), 0.8 by NNsplice (normal range 0–1), 76.3 by Splice Site Finder (normal range 0–100), 0.8 confidence by Netgene2 (normal confidence range 0–1) and 3.4 by MaxEnt (normal range 0–12).

Patients previously published by Steenweg et al.1

Predicted strength of new splice donor site at c.439 of 79, by HSF (normal range 0–100).

Predicted increase in potential acceptor splice site at c.453+15 strength of 45.7% (2.6–3.8) by MaxEnt (normal rang 0–16).

Predicted decrease of 21% (0.7–0.5) by NNsplice (normal range 0–1).

Patients belong to the same family.

Predicted strength new splice donor site at c.453+161 of 2.3 by MaxEnt (normal range 0–12).

Previously investigated and reported by us.23

Informed consent

We received approval of the ethical standards committee for our gene identification research on patients with unclassified leukoencephalopathies at the VU University Medical Center in Amsterdam. All guardians of the patients participating in this study gave written informed consent. Approval was also obtained from the Institutional Review Board at Nemours/Alfred I. duPont Hospital for Children and the BC Children’s Hospital, University of British Columbia, Canada, and informed consent was obtained as appropriate on the patients studied at these institutions.

Whole-exome sequencing

Whole-exome sequencing (WES) on DNA from patients 5 and 11 (brothers), their mother, and patient 9, was performed using SeqCap EZ Human Library v3.0 kit (Nimblegen, Madison, WI, US) on a HiSeq2000 (Illumina, San Diego, CA). Coverage of at least 20× was reached for >96% of the targeted regions. Average sequencing depth ranged between 58 and 71. Data analysis was performed as described previously.14 For three families, WES had been initiated in three other institutes, and available WES data were used for identification of mutations after recognition of the gene.

Targeted sequencing of the X-chromosome exome

X-chromosome exome sequencing on DNA from patients 4 and 10 (brothers), 5 and 11 (brothers), 8 (including parents), 9 and 12 (including parents), was performed using SureSelectXT X-chromosome kit (Agilent, Santa Clara, CA, US) on a HiSeq2000 (Illumina, San Diego, CA). For all samples, coverage of at least 30× was reached for >80% of the targeted regions. Average sequencing depth ranged between 335 and 645. Data analysis was performed as described previously.15

Validation and detection of PLP1 variants

Validation and segregation of the identified PLP1 variants in patients 1, 4, 5, and 7–12 by exome sequencing analysis was performed using standard Sanger sequencing. Primers were designed (Primer 3 V.0.4.0)16,17 using reference sequence: PLP1, NM_000533.3. In patient 6, all 7 exons and intron–exon boundaries of PLP1 were sequenced; in the remaining patients, only exon 3 and intron 3 of PLP1 were sequenced using suitable primers (available upon request).

In silico analysis of effects of the identified variants

Pathogenicity of missense variants was predicted using SIFT,18 PolyPhen-219 and Mutation Taster.20 Conservation of nucleotides was analyzed using Phastcons scores, obtained via Alamut version 2.4.

In silico splicing analysis

The predicted effects of identified variants in PLP1 on splicing were analyzed using different programs (details see in Data S1). A deviation of ≥10% from the normal score was considered as a change likely to affect splicing if present in at least three splice site prediction tools21 (indicated by “yes” in Table1). A deviation of ≥10% from the normal score in less than 3 splice site predictions tools was considered an inconclusive result. Changes of <10% were regarded as not significant and not reported. Predicted changes of number of exonic splice enhancer and silencer motifs were determined using default threshold values.

In silico analysis of secondary PLP1 RNA folding

The mfold program accessed via their web server with standard parameters was used to analyze PLP1 normal and mutant RNA sequences for changes in the secondary pre-mRNA structure and stability (http://mfold.rna.albany.edu/?q=mfold/RNA-Folding-Form).22 Using mfold, the four intronic mutations c.453+159G>A, c.453+164G>A, c.454−312C>G, and c.454−314T>G were analyzed using mutant RNA sequences including two regions within intron 3 of PLP1: 20 bases from c.453+150 to +169 and 20 bases from c.454−326 to −307, separated by 15 random bases (N15), as previously described by us.23 The four mutations c.404T>G, c.436C>T, c.441A>T, and c.453+7A>G were analyzed using mutant PLP1 RNA sequences comprising exon 3 (c.192−c.453) and 40 flanking intronic nucleotides. As a measurement for the splice donor site accessibility by the spliceosome, conformational changes of the PLP1 splice donor site in exon 3B and the DM20 splice donor site in exon 3A, were visually examined and quantified by counting the nucleotides of the consensus PLP1 splice donor site (c.451 to c.453+6: AAGGUGAUC) that were predicted to form base pairs.

PLP1/DM20 alternative splicing studies

Minigene splicing construct transfection assay

The effects of three PLP1 mutations c.436C>T, c.441A>T, and c.453+7G>A on PLP1/DM20 alternative splicing were individually investigated in a minigene splicing reporter construct transfection assay using Oli-neu cells, an immortalized cell line representing immature oligodendrocytes, as previously described, with modified reverse transcription (RT) (details in Data S1).23,24

Skin fibroblast cultures

PLP1/DM20 alternative splicing was investigated in skin fibroblasts from patient 6 harboring the c.453+7A>G intronic mutation. In addition, skin fibroblasts were available from a patient described in 1991 with the c.441A>T silent mutation,25 the same as identified in patient 8 from our HEMS cohort. RT-PCR analysis of RNA prepared from cultured skin fibroblasts was performed as previously described with modified RT reaction (details in Data S1).23

Results

MRI findings

Detailed MRI findings of all patients in our cohort except three affected siblings are provided in Table S1 and Figures1, 2. These patients had the characteristic MRI features corresponding to the previously described disorder “HEMS”.1 On initial MRIs, patients had mild T2 hyperintensity of the medulla oblongata, the pons, especially at the border with the medulla oblongata, and the hilus of the dentate nucleus, with variable T2 hyperintensity of the peridentate white matter, indicating hypomyelination of these regions (Figs.1A and B, 2I and J). Mild T2 hyperintensity of the optic radiation, periventricular white matter and parietal white matter was observed in all patients (Figs.1C, 2K), with extensions into the subcortical white matter under the pericentral cortex in eight patients (Figs.1D, 2L). Patients 1 and 6 had more extensive subcortical white matter hypomyelination. All patients had alternating T2 hyperintense–hypointense–hyperintense stripes in the posterior limb of the internal capsule (Figs.1C, 2K). The thalamus had a mildly elevated T2 signal, except for its ventrolateral part, which was darker, a configuration that is normally seen in neonates (Figs.1C, 2K). On follow-up, the T2 signal abnormalities of the medulla oblongata and pons improved in four of the eight patients in whom repeat images were available, indicating progressing myelination (Fig.1I and J). The periventricular and deep white matter and/or corpus callosum showed more extensive T2 hyperintensity than before in seven of the eight patients, indicating myelin loss (Fig.1K and L). After identification of mutations in the gene PLP1 (see below), two clinically asymptomatic female carriers (c.454−312C>G, age 40 years, and c.453+159G>A, age 53 years) underwent brain MRI, which demonstrated mild global atrophy of the cerebral hemispheres and diffuse mild T2 hyperintensity of the supratentorial white matter (data not shown).

Figure 1

MRI of HEMS patients. Initial MRIs of patient 9 (A–D, age 3 years), and of an age-matched control (E–H, 2.5 years) and follow-up MRIs of patient 9 (I–L, age 12.4 years), and of an age-matched control (M–P, 12 years). All images are axial T2-weighted. MRIs of patient 9 show mild T2 hyperintensity of the medulla oblongata (A, arrow), the pons (B, white arrow) and the hilus of the dentate nucleus (B, black arrow). The thalamus is mildly T2 hyperintense, except for its ventrolateral part, which is dark (C, white arrow). The posterior limb of the internal capsule shows alternating hyperintense–hypointense–hyperintense stripes (C, black arrow). There is mild T2 hyperintensity of the optic radiation (C, white arrowhead) and the periventricular white matter that extends into the central subcortical white matter (D, black arrow). At later stages, the T2 signal in the medulla and pons improves (I and J), but the T2 hyperintensity of the optic radiation (K, white arrow), the corpus callosum (K, black arrow) and the periventricular white matter and subcortical white matter increases. Control images (E–H and M–P) show the normal myelination pattern with respect to age.

Figure 2

MRI of PLP1-related disorders. MRI of a patient with: classic Pelizaeus–Merzbacher disease (PMD) harboring a PLP1 duplication (A–D, age 23 months), Spastic Paraplegia type 2 (SPG2) with a frameshift mutation (c.263delC, p.(Ala88Valfs*26)) (E–H, age 11.4 years) and patient 11 with hypomyelination of early myelinating structures (HEMS) harboring an intronic mutation (c.453+159G>A) (I–L, age 20 months). All MR images are axial T2-weighted. Note that in the patients with PMD (A and B) and SPG2 (E and F), the medulla oblongata and the pons show a normal dark T2 signal, in contrast to the patient with HEMS (I and J), in whom these structures are mildly T2 hyperintense. The alternating hyperintense–hypointense–hyperintense stripes in the posterior limb of the internal capsule seen commonly in HEMS patients (K) are not present in the patient with PMD (C) or SPG2 (G). All the three illustrated patients have T2 hyperintensity of the periventricular white matter, deep white matter and subcortical white matter, which is diffuse and extensive in the patient with PMD (C and D), restricted in the patient with SPG2 (G and H), and intermediate in the patient with HEMS (K and L).

Clinical profiles and laboratory results

Detailed clinical characteristics are provided in Table S2. Twelve of the 16 patients presented with nystagmus in the first or second year of life. The other four patients presented first with cerebellar dysfunction or developmental delay. All patients continued to gain motor skills, but mostly delayed. All developed progressive spasticity of the legs and signs of cerebellar dysfunction. At last clinical follow-up, five patients (age range 3–12 years) could still walk without support, although they had an evident spastic-ataxic gait. Five patients used braces or a walker for ambulation, and five needed a wheelchair. One was still too young to judge his ability to walk. Cognitive capabilities were normal or mildly impaired. All female carriers were clinically normal. Nerve conduction studies performed in eight patients revealed normal motor and sensory conduction velocities in all except patient 1 who had delayed conduction velocities in his legs.

Genetic analysis

We first performed WES in two brothers and their mother and an unrelated patient. Under the hypothesis of an X-linked recessive inheritance model, we selected all rare hemizygous variants (variants with a minor allele frequency of ≤1% in known public control databases; 1000 Genomes, dbSNP137 and National Heart, Lung, and Blood Institute Exome Sequencing Project (http://evs.gs.washington.edu/EVS/), and absence from our in-house control samples), located on the X-chromosome. With this approach we were unable to identify a candidate gene. We subsequently performed targeted sequencing of the X-chromosome exome including four additional HEMS patients and parents. The same filtering strategy was used, and genes were selected if variants were present in at least two unrelated patients. This led us to the identification of one candidate gene: PLP1 (MIM 300401), encoding PLP1 (shown in Fig.3A). Two exonic variants were identified: a c.404T>G missense variant, predicting p.(Leu135Trp) and a c.441A>T silent variant. Manual analysis of the intron data revealed three variants located deep in intron 3. Coverage of two of these variants, c.454−312C>G and c.454−314T>G, was very poor for both the WES and the X-chromosome exome approach (<5 reads), while the c.453+159G>A variant was sufficiently covered with the X-chromosome exome (63–97 reads) (data not shown). In the nine remaining HEMS patients PLP1 variants were identified either by Sanger sequencing or by WES data retrieved from other institutes. In total, nine different PLP1 variants were identified in 16 patients (10 families) (Table1, Fig.3B). All variants were located in exon 3B or intron 3 (Fig.3A and B). The mutations were maternally inherited in five families and de novo in three families. In two families, the mother was not available for testing. However, multiple affected family members indicated obligate carriership.

Figure 3

Overview of PLP1 mutations and predicted secondary PLP1 RNA structure. This overview illustrates PLP1 gene structure with its seven exons (black boxes) and the two alternative transcripts, PLP1 or DM20, resulting from a splice donor site within exon 3 (A). Close-up of exon 3B and intron 3 is depicted in (B). All nine mutations found in our hypomyelination of early myelinating structures (HEMS) cohort are shown. The frameshift mutation is indicated with a black solid line, the two silent mutations with a dashed line and the single missense mutation with two striped lines. Intronic mutations are indicated with an interrupted line. The secondary PLP1 RNA structure is displayed in (C), illustrating the formation of the predicted long-distance interaction (LDI) structure between LDIS-5′ and LDIS-3′ within intron 3. Mutated positions in our HEMS cohort are indicated with stars. Adapted from Taube et al.23 with permission of Oxford University press.

In silico analyses of effects of the identified PLP1 variants

The c.380_392del variant was predicted to result in a frameshift p.(Arg127Lysfs*16) truncating PLP1, but not DM20. For the other eight identified PLP1 variants, we performed several in silico bioinformatics prediction analyses to investigate pathogenicity. An overview of the results is depicted in Table1.

Of the eight variants, only the silent c.436C>T change was predicted to have an effect on splice sites by creating a new splice donor site in exon 3B at position c.434. Analysis of splice regulatory elements showed that for all three exonic variants and the c.454−312C>G variant 1–3 new exonic splicing silencer motifs were predicted.

Using mfold, we previously found that two regions within intron 3 of PLP1 form an intramolecular base-pairing interaction in the predicted secondary PLP1 RNA structure and play an important role in the control of PLP1/DM20 alternative splicing.23 In our HEMS cohort, we found four variants in either of these two regions (Fig.3B and C). The 5′ region (c.453+151 to c.453+166) is referred to as the long-distance interaction (LDI) site 5′ (LDIS-5′) and the 3′ region (c.454−323 to c.454−308) as the LDIS-3′ (Fig.3C). Mutations created in LDIS-5′ and LDIS-3′ sequences that were predicted to destabilize this secondary PLP1 RNA structure (initial Gibbs free energy [ΔG] value less than −16.4 kcal/mole) decreased the ratio of alternatively spliced products PLP1 to DM20 in a minigene reporter system, while mutations of the central nonpairing bases, c.453+158 and c.454−315, and mutations that were predicted to maintain the secondary structure did not.23 We investigated the effect of the four variants identified in our families, two in LDIS-5′ (c.453+159G>A and c.453+164G>A) and two in LDIS-3′ (c.454−312C>G and c.454−314T>G), on the LDI structure and found that in all cases the stability of the mutant LDI structures was predicted to be reduced ([ΔG] values of less than −16.4 kcal/mole) compared with the normal fragment ([ΔG of −17.6 kcal/mole]) (Fig. S1).

Analysis of the predicted secondary PLP1 RNA structures for the remaining four variants c.404T>G, c.436C>T, c.441A>T, and c.453+7A>G showed that all except the c.436C>T variant increased the stability of the local RNA structure of the PLP1 splice donor site, while leaving the structure of the DM20 splice donor site unchanged (Table S3). These predictions suggest that the PLP1 splice donor site is more engaged in intramolecular base pairing in the mutated RNA fragments than in the normal RNA fragments.

PLP1/DM20 alternative splicing studies in a minigene splicing construct assay

In a reporter construct transfection assay,23,24 we investigated the effects of the c.436C>T, the c.441A>T, and the c.453+7A>G mutation on PLP1/DM20 alternative splicing. These mutations resulted in a significant reduction in the PLP1/DM20 ratio to 0.18 (c.436C>T), 0.14 (c.441A>T), and 0.05 (c.453+7A>G) of normal (Fig.4A). We could not detect the potentially aberrantly spliced product from the new splice donor site predicted for the c.436C>T mutation. The effect of the c.454−314T>G mutation was previously reported by us and resulted in a PLP1/DM20 ratio of 0.15 of normal.23

Figure 4

Dysregulation of PLP1 splicing by patients’ mutations. Quantitation of PLP1/DM20 mRNA produced in an immature immortalized oligodendrocyte cell line from the transfected splicing minigene reporter construct indicated a significant decrease in the PLP1/DM20 ratio with patients’ mutations compared with normal construct (A). The graph shows the ratio of PLP1 product to the DM20 product from each patient construct normalized to the normal construct (mean of replicates ±SD). All constructs were tested in triplicate. All three patient constructs had a ratio significantly different from the normal construct, as measured by Student’s t-test: ***, P < 0.001. Analysis of skin fibroblast RNA by RT-PCR indicated a decrease in PLP1/DM20 ratio with patients’ mutations (B). The primers that amplify from exon 2 to exon 5 PLP1 readily detect PLP1 and DM20 signal in the same reaction as different sized bands. A heteroduplex of PLP1 and DM20 is also visible (PLP1 and DM20 heteroduplex formation has been described previously).24 Quantitation performed by dividing PLP1 + ½ heteroduplex band by DM20 and normalizing, indicated PLP1/DM20 ratio in patient fibroblasts was 0.01 of normal for c.441A>T and 0.06 of normal for c.453+7A>G. The PLP1 band was 105 bases larger than the DM20 band.

PLP1/DM20 alternative splicing studies in fibroblasts

In patient fibroblasts, we investigated the effect of the c.441A>T and the c.453+7A>G mutation on the PLP1/DM20 ratio and found it reduced to 0.01 and 0.06 of normal (Fig.4B). We consider this to be a rough estimate of the reduction in PLP1/DM20 ratio because there is very little PLP1 compared with DM20 in skin fibroblasts and we noted that the amount of total PLP1 (PLP1 + DM20) decreases with passage number.

Discussion

Using exome sequencing, we identified the genetic defect causing HEMS.1 All patients had hemizygous mutations in the PLP1 gene, which encodes both PLP1 and its smaller isoform DM20 that is derived by the use of an alternative splice donor site within exon 3.13 Together the proteins constitute more than half of the total protein mass of myelin in the central nervous system (CNS).26 The DM20 transcript is preferentially expressed in the developing CNS before initiation of myelination, whereas the PLP1 transcript dominates during myelination and adulthood, suggesting tightly regulated PLP1/DM20 alternative splicing.

PLP1 mutations are known to be associated with a broad continuum of neurological phenotypes ranging from connatal PMD with severe hypomyelination to pure X-linked spastic paraplegia type 2 (SPG2) with even normal brain imaging in some cases.5,7–9 Different genetic alterations have been identified to cause these different phenotypes, with duplication of the entire PLP1 gene as the most common change usually leading to the classic form of PMD.27 More severe forms are associated with missense mutations in highly conserved regions or, in rare cases, with triplications and higher copy number of PLP1, while patients with null mutations or missense mutations in less conserved regions present with milder signs.28,29 Previously, we had defined HEMS based on the distinctive abnormalities seen on brain MRI. With the identification of PLP1 mutations in this group, HEMS should be added as a recognizable new MRI phenotype within the broad spectrum of PLP1-related disorders.

Clinically, most patients with HEMS have a relatively mild functional disability and can be classified as complicated SPG2. All patients, except for patient 1 harboring the truncating PLP1 mutation (p.(Arg127Lysfs*16)) have normal nerve conduction studies or sensory function on physical examination. This is in agreement with the hypothesis that only patients lacking PLP1 due to a functional null mutation or a truncating mutation in the PLP1-specific region have a peripheral neuropathy.28,30–32 In five families, the mutation was inherited from the mother. All female carriers were clinically asymptomatic, but the two females who underwent MR imaging showed abnormalities consisting of mild atrophy of the cerebral hemispheres and a diffuse T2 hyperintensity of the white matter. This phenomenon has also been observed in carriers of mutations associated with a mild PMD phenotype,32,33 and can be explained by random X-inactivation due to the mild effects of these mutations.8,9

Remarkably, in our HEMS cohort, the identified PLP1 mutations were restricted to the PLP1 specific region encoded by exon 3B that is spliced out in DM20, and to intron 3 (Fig.3A and B). Only one family carried an obvious pathogenic PLP1 mutation (p.(Arg127Lysfs*16)). The other nine families had presumed “subtle” PLP1 mutations: these mutations were located in the noncoding region outside the splice donor and acceptor sites of exon 3B in six families; a silent mutation was identified in two families, and a missense mutation for which pathogenicity predictions were contradictory was identified in one family. This is in contrast to the commonly reported mutations in exon 3B that are mainly missense, truncating or located at the splice donor site sequence of exon 3B.28,30–32,34–47

We provide evidence that the mutations ascertained in our HEMS cohort alter PLP1/DM20 alternative splicing and must therefore be interpreted as pathogenic. Splicing is a complex mechanism not only controlled by the canonical splice donor and acceptor sites and branch sites but also by exonic or intronic enhancers and silencers and by regulation through the secondary RNA structure.23,24,42,45,48–51 We first performed in silico prediction analysis on the effects of the variants concerning these regulatory factors. Splice site predictions for the c.436C>T mutation showed a new PLP1 splice donor site at c.434. For all three exonic substitutions (including c.436C>T) and for the intronic c.454−312C>G mutation, the creation of putative new exonic silencer motifs was predicted. This mechanism had previously been found associated with loss of the PLP1 isoform for a PLP1 c.436C>G missense mutation.52 Analysis of predicted secondary PLP1 RNA structures containing the c.404T>G, c.441A>T and 453+7A>G mutations revealed potential conformational changes of the PLP1 splice donor site that could result in a less accessible splice site for the spliceosome. This change may shift the balance between the usage of DM20 and the PLP1 exon 3 splice donor sites in disadvantage of the latter. Interestingly, the four noncoding mutations restricted to two specific regions deep within intron 3 of PLP1 (LDIS-5′ and LDIS-3′) were predicted to reduce the stability of the secondary LDI PLP1 RNA structure, which has been found associated with a decreased PLP1 to DM20 ratio.23 Overall, these results indicate that the identified variants have an impact on PLP1/DM20 alternative splicing. For the three tested variants (c.453+7A>G, c.436C>T, c.441A>T), we could confirm these predictions in vitro by detecting a decreased PLP1/DM20 ratio in fibroblasts or in our minigene reporter transfection assay. Several additional algorithms applied for prediction of the impact of the single missense variant (c.404T>G) on protein structure and function showed contradictory results.

Although mutations in PLP1 are also associated with PMD or SPG2, the MRI pattern of HEMS contrasts with the MRI findings in these disorders (see Fig.2).1,4,5 Normally, tracts become myelinated at the time they become functional, resulting in a fixed spatiotemporal sequence of myelination.53 In patients with HEMS, the brain structures that normally myelinate early (e.g., brainstem, hilus of the dentate nucleus, posterior limb of the internal capsule, optic tracts and tracts to the pericentral cortex) are poorly myelinated in contrast to structures that normally myelinate at a later developmental stage, which show better myelination.1,53 In patients with PMD and SPG2, however, the early myelinated structures are relatively better myelinated than other brain structures (Fig.2A–H). This indicates that PLP1/DM20 alternative splicing and maintenance of a certain PLP1 to DM20 ratio are important for early myelination.

Noteworthy, three mutations identified in our HEMS cohort have been reported before (c.436C>T,34 c.441A>T,25 and c.453+7A>G46). The patient with the c.436C>T had mild functional disability.34 For the patients with the c.441A>T and c.453+7A>G mutation only the clinical diagnosis “PMD” was provided.25,46 We recently also identified three PMD families with mutations in the LDIS-3′ region (including c.454−314T>G), with a mild clinical phenotype, comparable to HEMS patients.23 No MRI data were available for these patients, so we were not able to evaluate whether these patients have an MRI phenotype compatible with HEMS.

Our HEMS cohort presents a challenge for PLP1 Sanger sequencing in a diagnostic setting. For most mutations, pathogenicity was difficult to prove. For example, the two silent mutations, c.436C>T and c.441A>T, had already been identified in the past in our patients, but were thought to be benign. Moreover, the noncoding regions LDIS-5′ and LDIS-3′ in intron 3 are currently not included in PLP1 diagnostic sequencing. Our data support the proposal that sequencing of intron 3 of PLP1 should be included in standard diagnostic procedures because of the importance of this region for controlling PLP1/DM20 alternative splicing.23 Furthermore, this study illustrates that caution is warranted in case of negative exome sequencing results, as coverage is often not optimal for intronic regions. Thanks to the number of patients with identical clinical and MRI presentation, we were able to identify the genetic basis of HEMS. The identification of more patients with mutations in this or other regions of PLP1 altering PLP1/DM20 alternative splicing will further elucidate the specificity of the HEMS MRI phenotype within the PMD spectrum. Future research will show whether these mutations that affect PLP1/DM20 alternative splicing could be potential targets for treatment aimed at correction of the PLP1 to DM20 ratio.