Abstract
Background: Lafora’s disease is a progressive myoclonus epilepsy with pathognomonic inclusions (polyglucosan bodies) caused by mutations in the EPM2A gene. EPM2A codes for laforin, a protein with unknown function. Mutations have been reported in the last three of the gene’s exons. To date, the first exon has not been determined conclusively. It has been predicted based on genomic DNA sequence analysis including comparison with the mouse homologue.
Objectives: 1) To detect new mutations in exon 1 and establish the role of this exon in Lafora’s disease. 2) To generate hypotheses about the biological function of laforin based on bioinformatic analyses.
Methods: 1) PCR conditions and components were refined to allow amplification and sequencing of the first exon of EPM2A. 2) Extensive bioinformatic analyses of the primary structure of laforin were completed.
Results: 1) Seven new mutations were identified in the putative exon 1. 2) Laforin is predicted not to localize to the cell membrane or any of the organelles. It contains all components of the catalytic active site of the family of dual-specificity phosphatases. It contains a sequence predicted to encode a carbohydrate binding domain (coded by exon 1) and two putative glucohydrolase catalytic sites.
Conclusions: The identification of mutations in exon 1 of EPM2A establishes its role in the pathogenesis of Lafora’s disease. The presence of potential carbohydrate binding and cleaving domains suggest a role for laforin in the prevention of accumulation of polyglucosans in healthy neurons.
Lafora’s disease (LD) is an autosomal recessive progressive myoclonus epilepsy with late childhood or teenage onset of myoclonic and photoconvulsive seizures. Progressive neurodegeneration results in increasingly intractable seizures, dementia, and death within 10 years of onset.1
Pathognomonic Lafora bodies (LB) found in brain2 and in almost every other organ1 stain strongly with periodic acid–Schiff (PAS), indicating a major content of polysaccharides.3 Acid hydrolysis of the polysaccharide yields almost exclusively glucose, indicating it is a polyglucosan (PG).3,4 Unlike glycogen, PG are densely packed, insoluble, and phosphorylated.3 Insolubility and phosphorylation renders their in vitro enzymatic digestion difficult.3,5 Nonetheless, with high concentrations and prolonged digestions, amylolytic enzymes (enzymes that cleave α-linkages between glucose molecules) digest the LB into mono and di-saccharides of glucose).3,5 Equivalent experiments with β-cleaving enzymes have not been performed. Finally, LB also contain up to 30% uncharacterized protein.3
LB are histochemically and chemically almost indistinguishable from corpora amylacea (CA), an inclusion body seen in normal aging cells in most tissues.3,6 In brain, CA are located almost exclusively in glial cells, whereas LB are found primarily in neurons.1,6
LD is caused by mutations in either the EPM2A gene7 or in a second yet unidentified gene.8 EPM2A contains four exons, but when it was first identified, the apparently complete gene was not determined. At the time, the sequence of the first exon was predicted from genomic DNA based on the presence of a eukaryotic translation initiation site and an open reading frame in continuity with the partial sequence available in the cDNA.7 Ganesh et al. provided further support for the predicted first exon by showing a high level of conservation of the human genomic sequence from this region with a putative complete cDNA of the mouse homologue of EPM2A.9
In this report, we show that the largest number of mutations of the EPM2A gene in our data set occur in its first exon, thus confirming the role of this exon in the function of its protein product, laforin. We also present results of extensive analyses of the sequence of laforin. These analyses identify a putative function of laforin in the degradation of PG.
Methods.
Patients.
The specific diagnosis of LD was made in each patient based on the identification of LB in biopsies of skin, liver, or muscle. All patients had normal development during childhood years before onset. The range of ages at onset of the progressively worsening seizures was 8 to 15 years (table). However, in many cases, isolated febrile or nonfebrile seizures had occurred years before the onset of the progressive syndrome. The initial obvious symptoms were generalized convulsions, myoclonic seizures, drop attacks, visual hallucinations, or staring spells. With subsequent affected children, families detected cognitive decline or subtle myoclonias, absences, or visual hallucinations earlier than had been noticed with the first child. The clinical course after onset appears to have been similar in all cases with progressive dementia and increasing myoclonic, photoconvulsive, and other seizures.
EPM2A mutations, predicted effects on the laforin protein, and age and seizures at onset
Mutation analysis in exon 1 of EPM2A.
Mutations were detected by radioactive cycle sequencing using the Thermosequenase Kit (Amersham Life Science; Arlington Heights, IL) with column (Qiagen, Hilden, Germany) purified PCR products. PCR primers were BPF (5′-CTCCAGAGACCAGGTCCAAG-3′) and BPR (5′-GTCTGCTGGCAATACCTTCC-3′); PCR conditions were 35 cycles with annealing temperature 60 °C, annealing and extension times 30 seconds, final extension of 10 minutes, [MgCl2] = 1.5 mM, betaine = 1 M, product size = 771 bp. Sequencing primers were BPR, LD15PF2 (5′-AAGAGGCGCAGAACAGCAC-3′), and H65F2 (5′-AGCTGCTGGTGGTGGGGT-3′). Amplification worked best with the Techne Touchgene Model FTG02TP PCR machine (Techne, Cambridge, U.K.). Oligonucleotide specific hybridization was used to test for the presence of point mutations in the unaffected population.7
Bioinformatic analyses of the sequence of laforin.
The complete sequence of laforin can be accessed at http://www.ncbi.nlm.nih.gov/Entrez/proteins.html (accession AAC83347). To identify possible transmembrane regions, the TMFinder prediction tool (http://www.bioinfo.sickkids.on.ca/TM/login.html) was used.10 To identify secretory signal peptides (endoplasmic reticulum or ER target signals), SPScan was used (Seq Web version 1.1 with the Wisconsin Package version 10). The sequence was further analyzed for the presence of targeting signal sequences for peroxisomes, the nucleus, and mitochondria. The peroxisome signal peptides are PTS1, a tripeptide at the extreme C-terminus composed of Ser or Cys or Ala followed by Lys or Arg or His followed by Leu (S/C/A, K/R/H, L); and PTS2, located near the N-terminus with sequence R/K, L/V/I, x(5), H/Q, L/A, where x represents any amino acid (aa).11 The nuclear localization signal is either a single cluster of approximately five basic residues or two smaller clusters of basic residues separated by any 10 aa.12 The mitochondrial signal peptide is 20–60 aa composed mainly of basic and hydroxylated residues and the near to complete absence of acidic residues.13
BLASTP (http://blast.bioinfo.sickkids.on.ca/BLASTlogin.html) was used to search for sequence homologies between laforin and other proteins. The sequence was also submitted to Pfam at the Sanger Centre (http://www.sanger.ac.uk/Software/Pfam). Pfam is a large collection of protein families grouped together based on sequence homologies in functional domains.14 The alignment program returns positive comparisons between query sequences and Pfam domains as “potential” or “trusted” matches using statistical analyses detailed at the web address. Finally, the PROSCAN tool (http://www.pbil.ibcp.Fr/NPSA/) was used to scan the Prosite database (http://www.expasy.ch/prosite/) at the Expasy Molecular Biology server. Prosite is a large database of biologically significant protein sequence patterns (motifs). PROSCAN’s algorithm allows a search with errors and returns results as percent exact matches between the aa of a submitted sequence and those of matching Prosite family motifs.
Results.
Mutation analysis.
The table summarizes the seven new mutations (point mutations in exon 1) identified in this study, as well as those from our data set from other studies.7,16 Each of the new mutations is unique to the family in which it was found, which is not unexpected as all individuals are from apparently different ethnic backgrounds. In all, there are now 17 mutations found in 22 families including eight missense, five nonsense, two frameshift, and two deletion mutations. The missense changes reported are the ones we considered true mutations (other single nucleotide polymorphisms have been identified), because they result in replacement with nonconserved aa, are not found in normal chromosomes, and/or disrupt a potential functional domain (see Discussion). Finally, four patients are compound heterozygotes (JT, LB, LD103, L6).
Bioinformatic analyses.
TMFinder did not detect transmembrane regions in laforin. Similarly, laforin was not found to contain known targeting signal peptide sequences for entry into the nucleus or any of the organelles. Absence of an ER signal peptide precludes translocation to lysosomes.
Comparison to protein database sequences using BLASTP revealed that laforin’s C-terminus contains the HCxAGxxRS/T motif, which is the catalytic site found in the C-termini of protein tyrosine phosphatases (PTP) (figures 1 and 2⇓). Laforin also contains the conserved aspartate residue 30 aa N-terminus to the catalytic domain and the arginine 27 aa C-terminus to it (see figure 2), both of which are involved in the dephosphorylation mechanism.17,18 PTP are counterparts of protein tyrosine kinases. They regulate a large number of biological processes by dephosphorylating tyrosine residues of specific target proteins. Their N-termini are dissimilar and confer substrate specificity.17
Figure 1. Laforin. Locations of mutations are indicated by vertical lines, except for the two deletion mutations, which are indicated by horizontal lines: CBD-4, amino acid (aa) sequence similarity with carbohydrate binding domain of a number of amylases; w (arrowheads), positions of trp residues possibly involved in hydrophobic interaction with polyglucosan (PG); G10, similarity with the glucohydrolase family 10 active site; PTP, homology with the protein tyrosine phosphatase catalytic region; G1, similarity with the glucohydrolase family 1 N-terminal signature.
Figure 2. Alignment of laforin with six dual-specificity phosphatases (DSP). Stars indicate conserved catalytically important residues in the protein tyrosine phosphatase (PTP) active site. Numbers indicate position in protein of last amino acid (aa) in each line. Laf, laforin; MKP5, most recently described human mitogen-activated protein kinase phosphatase (accession 6005812); HVH3, a human DSP (Q16690); MKP3, a rat MKP (AAB94858); yVH1, a yeast DSP (CAB51765); DSP, a neuronal-specific mouse DSP (CAA64772); MKP1, the first human MKP (P28563).
Submission of laforin to Pfam returned a “trusted match” with dual-specificity phosphatases (DSP). DSP form a branch of the PTP family and utilize the same catalytic site, but they are capable of dephosphorylating serine and threonine residues in addition to tyrosine. Mitogen activated protein kinase phosphatases (MKP) make up an important segment of the DSP family, but laforin lacks key regions that characterize MKP (the CH2 and CB motifs).18,19
Further search for functional domains using Pfam revealed a “potential match” in laforin’s N-terminus (encoded by exon 1) with the carbohydrate binding domain (CBD-4) of a number of bacterial and lower eukaryotic amylases (see figure 1 and figure 3A). CBD-4 containing amylases bind to polysaccharides by hydrophobic interaction using several tryptophans positioned in a semi-regular fashion over an extended region.20,21 This arrangement appears to be present in laforin’s N-terminus (see figure 1).
Figure 3. (A) Alignment of laforin with seven CBD-4 containing amylases from the indicated organisms. Star indicates the conserved tryptophan. This residue is mutated in laforin in family LD52 (table, figure 1). Numbers indicate position in protein of last aa in each line. NCBI accession numbers of the amylases, in order, are D10698, X67291, D49448, Z34466, P29750, P22998, and CAA73926. (B) Alignment of laforin with the consensus sequence of the glucohydrolase family 10 active site. Small case letters show mismatched amino acids (aa). Star indicates the catalytic glutamate nucleophile. Alignment of one member of the GH10 family (a xylanase) with the consensus is also shown. (C) Alignment of laforin and one member of the GH1 family (a phosphoglucosidase) with the consensus sequence of the glucohydrolase family 1 active site.
Finally, PROSCAN detected 75% similarity between sequence AVMNFQTEWDI from laforin and the active site of members of the glucohydrolase family 10 (G10) (Prosite accession number: PS00591) as well as 70% similarity between sequence FLMAKRPAVYIDEEA and the active catalytic N-terminal signature motif of members of the glucohydrolase family 1 (G1) (PS00653) (see figures 1 and 3⇑).22 Glucohydrolases from these families use these respective sites to cleave β linkages of polysaccharides, and family 1 includes several glucohydrolases that act on phosphorylated polysaccharides.23
Mammalian members of the CBD-4 containing amylases and G10 containing glucohydrolases have not been previously described. G1 containing glucohydrolases, however, have been described in mammals (e.g., lactase-phlorizin hydrolase, which splits lactose in the small intestine of neonates).22
Discussion.
At the onset of the progressive syndrome, patients presented with one or more of the following seizures: myoclonic, atonic, generalized tonic–clonic, absence, or occipital (visual hallucinations). There was no correlation between mutation type and age or seizure type at onset (see the table). For example, the mutation in Patient AI would be expected to truncate the protein product to a greater degree than the mutation in Patient LD5, yet AI presented at age 15 with visual hallucinations, whereas LD5 presented at age 8 with absences and myoclonic seizures, and then went on to have recurrent episodes of myoclonic and generalized tonic–clonic status epilepticus by age 10. We were unable to correlate mutation types with the long-term course of the disease following onset, because the patients are in different countries with differing medical care, and because their current years of disease burden are highly variable.
In our patient collection, we have now identified 17 different mutations in the four exons of EPM2A in 22 families (i.e., in ∼70% of our families). The seven new exon 1 mutations establish the role of this exon in the pathogenesis of LD. The remaining families may have mutations in noncoding regions of EPM2A or mutations in the other LD gene(s).8 Until the other LD gene(s) is identified, it will not be possible to estimate reliably the distribution of patients between the two (or more) loci. It is also not possible to estimate the number of patients with mutations in EPM2A that cannot be detected by analyzing its four coding exons.
The EPM2A exon 4 C→T Spanish mutation is the only mutation found in more than two families (see the table). Whereas the extent of mutation heterogeneity might have complicated the genetic diagnosis of LD, the relative simple genomic structure of the gene and availability of established mutation detection assays7,16 should now circumvent this.
Frameshift, stop (nonsense), and deletion mutations result in truncated or no protein products consistent with the autosomal recessive inheritance of LD. However, we have also identified eight missense mutations that will provide information regarding functionally significant regions and aa in laforin (table and figure 1). For example, the exon 1 Trp32Gly mutation occurs in laforin’s putative CBD in a conserved tryptophan residue24 (see figures 1 and 3A⇑), potentially affecting binding to the PG.20,21 Similarly, the Thr194Ile mutation would affect the conserved threonine in the putative G10 glucohydrolase active site (see figures 1 and 3B⇑). This threonine immediately precedes the conserved glutamic acid residue that is directly involved in β linkage cleavage by acting as a nucleophile.25
The occurrence of a putative α-polysaccharide binding domain and β-cleaving motifs in the same protein has not been reported previously, but would be consistent with the catabolism of LB PG, which contain the unusual mixture of α and β anomers. Whereas CBD-4 containing amylases have the CBD in the C-terminus, it is located in the N-terminus of laforin, and whereas G1 and G10 containing glucohydrolases have their active sites in the N-terminus, laforin would have these in the C-terminus (see figure 1).
Polyglucosans normally accumulate over time in the form of CA in glia but not in neurons. Neurons seem particularly vulnerable to PG buildup, as evidenced by the apparent exclusive neurologic symptomatology in LD patients despite massive accumulation of LB in liver, muscle, and other organs.1,3,6 We hypothesize that normal laforin prevents the accumulation of LB in neurons by first interacting with PG using its CBD. This is followed by cleavage of PG β-linkages using the glucohydrolase domains. The PTP domain may be involved in directly dephosphorylating the PG to allow more efficient enzymatic hydrolysis (dephosphorylation of non-aa moieties using the PTP catalytic mechanism has been reported previously26,27). Alternatively, the PTP site may act indirectly by activating downstream elements to recruit additional enzymes including α-cleaving enzymes to complete the catabolism of the PG. Such elements may include the gene product of the second LD gene. Hypotheses generated through this study are now being tested in the laboratory. Immunocytochemical experiments are underway to establish laforin’s subcellular localization. Several in vitro assays are being set up to test whether laforin can cleave phosphates from phosphotyrosine or phosphoserine substrates, digest α or β polysaccharides, and dissolve LB.
Acknowledgments
Supported by a grant from the Medical Research Council of Canada (MRC) to S.W.S. and the Quebec and Sweden Lafora’s disease associations. S.W.S. is a Scholar of the MRC.
Acknowledgment
The authors thank Vicky Sventzouris, Elayne Chan, Dr. Danielle Andrade, and Tamar Minassian for technical assistance; Dr. Stirling Carpenter, Dr. John Callahan, and Dr. Don Mahuran for scientific advice; and Brenda Muskat and Dr. Jamie Cuticchia for bioinformatic support. The authors thank Dr. Arvid Heiberg, Dr. P. Satishchandra, and Dr. Brenda Banwell for referring patients.
Footnotes
-
See also page 331
- Received January 14, 2000.
- Accepted May 24, 2000.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.