Splice Site Mutations in the ATP7A Gene

Menkes disease (MD) is caused by mutations in the ATP7A gene. We describe 33 novel splice site mutations detected in patients with MD or the milder phenotypic form, Occipital Horn Syndrome. We review these 33 mutations together with 28 previously published splice site mutations. We investigate 12 mutations for their effect on the mRNA transcript in vivo. Transcriptional data from another 16 mutations were collected from the literature. The theoretical consequences of splice site mutations, predicted with the bioinformatics tool Human Splice Finder, were investigated and evaluated in relation to in vivo results. Ninety-six percent of the mutations identified in 45 patients with classical MD were predicted to have a significant effect on splicing, which concurs with the absence of any detectable wild-type transcript in all 19 patients investigated in vivo. Sixty-seven percent of the mutations identified in 12 patients with milder phenotypes were predicted to have no significant effect on splicing, which concurs with the presence of wild-type transcript in 7 out of 9 patients investigated in vivo. Both the in silico predictions and the in vivo results support the hypothesis previously suggested by us and others, that the presence of some wild-type transcript is correlated to a milder phenotype.


Introduction
Menkes disease (MD) is an X-linked, multisystemic lethal disorder of copper metabolism caused by mutations in the ATP7A gene. The severe classical form of MD is characterised by progressive neurodegeneration, connective tissue abnormalities, distinctive ''kinky'' hair and ultimately death in early childhood. Though most patients (90-95%) exhibit the severe clinical course, there are various milder atypical forms with longer survival of the affected patients (atypical MD). The Occipital Horn Syndrome (OHS), mainly characterized by connective tissue manifestations is the mildest form [1].
A broad spectrum of ATP7A mutations (from single amino acid substitutions to large deletions and chromosome aberrations) has been identified in patients with classical or one of the milder MD forms [2]. Approximately 22% of these ATP7A mutations are splice site mutations [2].
Splice site mutations are DNA sequence changes that alter or abolish correct mRNA splicing during the process of precursor mRNA maturation. Splice site mutations can result in the complete skipping of one or more exons, retention of introns, creation of a pseudo-exon, or activation of a cryptic splice site within an exon or an intron [3]. The resulting transcript may either be in-frame or out-of-frame. An in-frame transcript may lead to the production of a functional or partially functional protein, whereas out-of-frame transcripts are likely to cause the formation of a premature termination codon and subsequent rapid nonsense-mediated mRNA decay [4,5]. Some splice site mutations do not abolish the wild-type transcript expression completely, which may lead to less severe phenotypes [6,7,8].
This study presents a comprehensive overview of 61 splice site mutations in the ATP7A gene including 33 novel mutations, identified in patients with classical MD, atypical MD or OHS. Studies of splice site mutations are highly relevant in order to identify a correlation between the effects of splice site mutations and patient phenotypes, to enable a prediction of the severity of the disease, and to design therapies. It has been proposed that MD patients might especially benefit from early copper therapy when some functional ATP7A protein is present [9]. We investigated the effect of 12 mutations in vivo, and 61 mutations with the bioinformatics tool, Human Splicing Finder (HSF). We compared the in silico predictions with in vivo results and with the observed phenotypes of the patients.

Samples and mutation detection
The 33 novel splice site mutations were found in a cohort of MD patients who were referred to the Kennedy Center for molecular and/or biochemical diagnosis. Mutations were identified by PCR amplification and direct sequencing of the 23 exons using an ABI3130XL sequencer (Applied Biosystems, Foster, CA).

Bioinformatics
We have compiled 61 ATP7A splice site mutations. All were analysed with HSF version 2.4 [10] (http://www.umd.be/HSF/). Only those mutations that are located in the defined splice site consensus sequences are included here. The consensus sequences are (C/A)AG|gt(a/g)agt and cag|G for the donor splice site and the acceptor splice site, respectively [11]. The mutation designations are based on ATP7A transcript ENST00000341514 from Ensembl. A in the ATG start codon is defined as c.1. The relative strength of the splice sites obtained from the bioinformatics tool is   given as the consensus value (CV), which ranges from 0 to 100. Splice sites with CVs higher than 80 are strong splice sites; less strong sites have CVs that range from 70 to 80. Splice sites with CVs of 65 to 70 are weak, as only a few of these sites are active [10]. Hence, splice sites with CVs below 70 are herein considered non-functional. The CV-output from HSF only indicates the strength of the actual splice site and does not take the possible effect of (and on) other cis-acting elements into account. The effect of a splice site mutation does not solely depend on the CV value, but also on the relative change in DCV. Desmet et al. (2009) [10] emphasised that DCV reductions of at least 10% for a mutation in any position, or of 7% for a mutation in position +4, are likely to affect splicing.

Results
When screening the coding region of the ATP7A gene in patients with Menkes disease or Occipital Horn Syndrome, we identified a total of 33 novel splice site mutations. They are listed in Table 1 together with 28 previously identified splice site mutations. Forty-five of these mutations lead to classical MD, 6 lead to the less severe atypical MD, and 6 mutations were identified in patients with the mildest phenotype, OHS. Only 4 of the mutations were identified in MD patients with unknown clinical phenotypes.
To explore whether a bioinformatics tool can be used to predict the effect of the genomic mutations on mRNA processing, and possibly the severity of the disease, we compared the experimental results and the observed phenotypes with the predictions obtained with the Human Splicing Finder.
The Human Splicing Finder results are listed in Table 1. Seventy-six percent (34/45) of the classical MD cases have a splice site CV below 70. Ninety-three percent (42/45) of the cases have a DCV reduction over 10% (or 7% for +4 mutations) ( Table 2). Ninety-six percent of the same mutations (43/45) lead to either a CV below 70 or a DCV reduction over 10% (or 7% for +4 mutations). Only two of the mutations among the classical MD cases lead to both a CV above 70 and a DCV reduction below 10% (cases: 14, 49).
In contrast, among the mutations identified in the 6 atypical MD patients and the 6 OHS patients, only four mutations (33%) lead to a DCV reduction over 10% (or 7% for +4 mutations) (cases: 16, 39, 46, 57). One of these mutations (16) also leads to a splice site CV below 70 and, as expected, no detectable wild-type transcript in the fibroblast material. In 67% of the mutations that lead to the mild phenotypes, the splice site CVs were above 70 in combination with DCV reductions below 10%, indicating that these sites are active ( Table 2). In agreement with this, detectable amounts of wild-type transcript were found in 7 of the 9 analysed patients with mild phenotypes. Only two patients (cases: 16,25) lacked detectable wild-type transcript.
In 15 cases (2,12,13,17,24,30,35,36,38,39,46, 50, 53, 57, 60) the CV-and DCV-parameters contradicted in the prediction of activity of the splice site. Mutations at the +5 position in particular, are responsible for a large fraction of these cases, as they constitute half of the total inconsistent incidences. In fact, the predictions for 8 out of 9 +5 point mutations are inconsistent. This indicates that the determination of both the CV and the DCV is essential.
As splice site mutations may lead to the use of cryptic splice sites, potential splice sites in the vicinity of wild-type splice sites were identified with HSF. Most cryptic splice sites are likely to be located +/2 100 bp on each side of the exon-intron boundary [12]. Therefore only this area was investigated for the presence of potential splice sites. All experimentally identified cryptic splice sites in the patient material were among the potential splice sites suggested by HSF.

Discussion
This study reviews disease-causing splice site mutations in ATP7A in relation to their effect on transcript processing and patient phenotype. Of the 57 mutations identified in patients with known clinical phenotypes, 21% of the mutations lead to the milder phenotypes, atypical MD or OHS (Table 2). We have previously reported that only 9% of all MD patients display the milder phenotypes [2]. Thus, splice site mutations relatively often lead to the milder phenotypes when compared to ATP7A mutations in general. This fact renders these mutations highly interesting for further research, and makes predictions of effects on pre-mRNA splicing, using bioinformatics tools, very relevant.
Wild-type transcript was detected in vivo in most OHS and atypical MD cases, but not in any of the tested, classical MD cases. We and others have previously shown that even a small amount of wild-type protein may have a major impact on the outcome of the disease in individuals with splice site mutations in ATP7A [13,14]. For instance, 2-5% correctly spliced ATP7A mRNA, as compared to the level in unaffected individuals, is enough to develop the mild phenotype, OHS [14].
Assuming that a splice site CV above 70 combined with a DCV reduction below 10% (or 7% for +4 mutations) allow the production of some amount of wild-type transcript, whereas a CV below 70 combined with a DCV reduction over 10% completely block the production of wild type transcript, 96% of the mutations identified in classical MD patients are predicted to have a significant effect on splicing with no production of wild type transcript. However, this is only the case for 33% of the mutations identified in patients with the milder phenotype. Overall, when using the CV parameter in combination with the DCV parameter, 51 of 57 mutations (89%) fit the hypothesis that mild phenotypes develop when wild-type transcript is produced, whereas classical MD is the phenotypic outcome when no wild-type transcript is produced (Table 2). Only, in vivo result from one case with classical MD does not fit this hypothesis. In case 14, with a CV above 70 and a DCV reduction below 10%, we were, in contrast to the prediction, unable to detect any wild-type transcript in vivo. There are also two exceptions, based on in vivo results, among cases with the mild phenotypes. In case 16, no wild-type transcript was detected in vivo in fibroblasts with the identified mutation, which concurs with the prediction given by HSF. The mild phenotype is thought to be caused by the production of an in-frame transcript, in which exon 8 is skipped and exon 9 is extended with 5 bp [15]; exon 8 encodes the region of the ATP7A-protein between the last metal-binding domain and the first transmembrane domain. In Case 25 with OHS phenotype: we were unable to detect any wildtype transcript; this contradicts the prediction given by HSF. The mild phenotype is also in this case thought to be related to the production of a partially functional protein-variant encoded by an in-frame transcript, which in this case lacks exon 10. The resulting ATP7A-protein lacks transmembrane domains 3 and 4, and is located in the endoplasmatic reticulum instead of its correct location in the trans-Golgi network [16]. It has been proposed that this ''mis-localised'' protein may retain some of its copper transporting function, resulting in the less severe phenotype.
In approximately one third of the represented exon-intron boundaries (8 of 26) the CV for one or more of the potential splice sites identified with HSF is higher than the CV for the un-mutated wild-type splice site. A similar observation has previously been reported in e.g. the b-globin intron 1 [17]. In b-globin intron 1 the authentic donor splice site is always used in the wild-type sequence, although there is a slightly stronger cryptic splice site close by which is only used when the wild-type splice site is mutationally weakened.
In case 2, HSF found several possible alternative donor splice sites (217, +16, +50, +55, +86), in the vicinity of the wild-type splice site. We have previously reported that a cryptic splice site at position +50 is used in the mutated sequence in this patient [14]. This means that the active cryptic splice site is not the potential splice site closest to the wild-type splice site. This concurs with observations made by Roca and colleagues [17] who found that 26 of 61 transcripts (43%), in which cryptic splice sites were analysed, skip the nearest potential splice site and use a site further away. The choice of site involves a very complex interplay between various factors, such as sequence context, available transacting factors and secondary structure of the pre-mRNA [17,18,19].
In conclusion, both the in silico predictions and in vivo results support the hypothesis previously suggested by us and others, that the presence of some wild-type transcript correlates with a milder phenotype. Notably, disease-causing mutations identified in patients with mild phenotypes might be difficult to distinguish from benign mutations, which do not cause disease. However, if the family history confirms beyond any doubt, that a splice site mutation is the cause of disease, HSF is very successful in predicting the phenotype. A l l 5 7 -----Using HSF, each mutation is analysed for the effect on the given splice site based on the two parameters CV and DCV. Mutated splice sites with CVs above 70 are likely to retain some activity. Conversely, splice sites with CVs below 70 are considered inactive. DCV-reductions of less than 10% (or * less than 7% at position +4) are likely to retain some wild-type splice site activity, whereas DCV-reductions of more than 10% (7% at position +4) are considered broken and inactive [10]. The mutations are categorized based on the CV-and DCV-values obtained in Table 1