Progressive myoclonus epilepsy of Lafora type (LD, MIM 254780) is a fatal autosomal recessive disorder characterized by the presence of progressive neurological deterioration, myoclonus, epilepsy and polyglucosan intracellular inclusion bodies, called Lafora bodies. Lafora bodies resemble glycogen with reduced branching, suggesting an alteration in glycogen metabolism. Linkage analysis and homozygosity mapping localized EPM2A, a major gene for LD, to chromosome 6q24. EPM2A encodes a protein of 331 amino acids (named laforin) with two domains, a dual-specificity phosphatase domain and a carbohydrate binding domain. Here we show that, in addition, laforin interacts with itself and with the glycogen targeting regulatory subunit R5 of protein phosphatase 1 (PP1). R5 is the human homolog of the murine Protein Targeting to Glycogen, a protein that also acts as a molecular scaffold assembling PP1 with its substrate, glycogen synthase, at the intracellular glycogen particles. The laforin–R5 interaction was confirmed by pull-down and co-localization experiments. Full-length laforin is required for the interaction. However, a minimal central region of R5 (amino acids 116–238), including the binding sites for glycogen and for glycogen synthase, is sufficient to interact with laforin. Point-mutagenesis of the glycogen synthase-binding site completely blocked the interaction with laforin. The majority of the EPM2A missense mutations found in LD patients result in lack of phosphatase activity, absence of binding to glycogen and lack of interaction with R5. Interestingly, we have found that the LD-associated EPM2A missense mutation G240S has no effect on the phosphatase or glycogen binding activities of laforin but disrupts the interaction with R5, suggesting that binding to R5 is critical for the laforin function. These results place laforin in the context of a multiprotein complex associated with intracellular glycogen particles, reinforcing the concept that laforin is involved in the regulation of glycogen metabolism.
Progressive myoclonus epilepsy (PME) of Lafora type (LD, MIM 254780) is a fatal autosomal recessive disorder characterized by the presence of progressive neurological deterioration, myoclonus and epilepsy (1). LD initially manifests during adolescence (usually between 10 and 17 years) with generalized tonic–clonic seizures, absences, drop attacks or partial visual seizures. As the disease progresses, the myoclonus increases in frequency and becomes constant. A rapidly progressive dementia with apraxia, aphasia and visual loss ensues, leading patients to a vegetative stage and death, usually within the first decade from onset of the first symptoms. LD was first described in 1911 by Gonzalo R. Lafora (2,3). He originally described the presence of polyglucosan intracellular inclusion bodies, so-called Lafora bodies, in the brain and spinal cord of an adolescent patient (2). It was then shown that these inclusion bodies are pathognomonic of LD and are not limited to the central nervous system. This led to the suggestion that the disease could be a generalized storage disorder and that it might be related to the glycogenoses (4). Later, it was shown that Lafora bodies have a composition which resembles that of glycogen or amylopeptin (5,6).
LD occurs worldwide, but is particularly common in the Mediterranean countries of Southern Europe and Northern Africa, in Southern India and in the Middle East. Linkage analysis and homozygosity mapping localized a major gene for LD to chromosome 6q24 (7,8). However, genetic heterogeneity is expected since in ∼10–20% of the LD families, linkage to chromosome 6q24 can be excluded (9–11). In 1998, we and others independently identified the gene responsible for LD on 6q24, EPM2A, and showed that 80% of the affected individuals were homozygous or compound heterozygous for mutations in this gene (9,12). A total of 30 different disease-related mutations have been described (13,14) of which 12 are missense mutations. Targeting disruption of Epm2a in mice results in a phenotype similar to that observed in human patients, including presence of Lafora bodies, neurodegeneration and profound neurological disturbances (15). Interestingly, in the Epm2a-knockout mice, the neurodegeneration and Lafora bodies predates the behavioral abnormalities, the ataxia and the epileptic myoclonias by many months.
EPM2A encodes a protein of 331 amino acids (so-called laforin) containing a dual-specificity protein phosphatase catalytic domain in the C-terminus and a carbohydrate binding domain in the N-terminus (16). Recently, it has been shown that recombinant laforin is able to hydrolyze phosphotyrosine as well as phosphoserine/threonine substrates (17) and binds directly to glycogen (16). Using yeast two-hybrid assays we report here that laforin interacts with itself and with R5, one of the glycogen targeting regulatory subunits of protein phosphatase 1 (PP1) (18–20). Critical regions for the interaction with laforin in R5 were delineated using deletion and site-directed mutagenesis analysis. In addition, functional characterization of LD-associated EPM2A missense mutations illustrated the relevance of the laforin–R5 interaction in LD, suggesting that the interaction with R5 is critical for the laforin function. In conclusion, we report here the identification of an interaction partner of laforin that is related to glycogen metabolism. Although the functional consequences of the laforin–R5 interaction are currently unknown, these data place laforin in the context of a multiprotein complex associated with intracellular glycogen particles, suggesting that laforin could be involved in the regulation of proteins important in the glycogen metabolism.
Expression and biochemical analyses of LD-associated EPM2A mutations
Homology searches with the laforin sequence have revealed the presence of a carbohydrate binding domain (CBD, residues 1–116) and a dual-specificity phosphatase catalytic domain (DSPD, residues 156–320) that have been shown to be functionally active domains in recombinant laforin proteins (16,17) (Fig. 1A). Several missense mutations altering single amino acid residues within the CBD and DSPD domains have been described in LD patients. However, functional characterization of these single amino acid substitutions is only available for the W32G mutation that showed both decreased phosphatase activity and loss of glycogen binding capacity (16).
To establish whether the missense mutations described in LD patients are loss-of-function mutations, and to determine the relative contributions of the CBD and DSPD domains to the LD phenotype, we have constructed and performed analyses in 10 different laforin recombinant proteins carrying LD-associated mutations. Nine of these mutations (W32G, F84L, R108C, T194I, G240S, G279S, Q293L, Y294N and P301L) are predicted to result in single amino acid substitutions and one (R241X) results in a truncation of laforin that eliminates the DSPD domain (Fig. 1A). This truncation has been found in 40% of the patients studied (13). In addition to these mutations, we have expressed two mutations (C266S and D235A), not reported in LD patients, but which are expected to disrupt the phosphatase activity of laforin without an effect on the correct folding of the protein (16,21). Recombinant proteins were expressed both in E. coli as GST::laforin fusion proteins and in a cell-free in vitro system using the TNT kit (Promega). All GST::laforin fusion proteins were expressed in E. coli in soluble form, purified with similar yields from the bacterial lysate supernatants by affinity chromatography (Fig. 1B), and assayed for phosphatase activity (Fig. 1C). Laforin recombinant proteins expressed in the TNT system were used to determine the glycogen binding capacity (Fig. 1D).
Figure 1C and D illustrates that the wild-type laforin recombinant protein shows both phosphatase activity and binding to glycogen and that the laforin proteins carrying the C266S or the D235A mutations show, as expected, no phosphatase activity but normal binding to glycogen. Nine of the laforin proteins carrying LD mutations (W32G, F84L, R108C, T194I, R241X, G279S, Q293L, Y294N and P301L) show no or very low levels of phosphatase enzyme activity and also lack the capacity to bind to glycogen.
It is noticeable that the W32G, F84L and R108C mutations, located in the CBD, have a profound effect on the phosphatase activity of laforin and that, similarly, the T194I, R241X, G279S, Q293L, Y294N and P301L mutations in the DSPD domain lack capacity to bind glycogen. These data formally establish that these EPM2A mutations are LD causative mutations.
An interesting mutation is the G240S in the DSPD domain. This mutant protein shows a phosphatase specific activity approximating that of the wild-type laforin (80%) and presents normal binding to glycogen (Fig. 1C and D). Formally, the G240S mutation could be a EPM2A polymorphism, with no functional consequences, and in linkage disequilibrium with an undetected causative mutation in EPM2A. Alternatively, the G240S mutation could alter another domain of laforin that is necessary for the correct functioning of the protein. In fact, laforin is expected to have additional functional domains involved in the interaction with putative regulatory subunits or with the in vivo substrates. We have thus used a yeast two-hybrid system to identify potential proteins that interact with laforin in brain and skeletal muscle and have analyzed how the LD-associated laforin mutations, including G240S, affect these interactions.
Yeast two-hybrid screens with laforin as bait
Two screenings of both a human skeletal muscle cDNA library and a human brain cDNA library (total of four screens) with laforin as a bait generated a total of 91 positive colonies. The plasmids containing the cDNA sequence of the putative laforin interaction partners were extracted from the conjugated AH109/Y187 yeast cells, transformed in E. coli and sequenced. cDNA sequences were characterized using a local alignment search tool (BLAST, NCBI). Sixty-five of the positive clones (∼70%) were excluded from further analysis because they lacked biological sense (i.e. the cDNA sequence was out of frame or corresponded to untranslated or genomic sequences). The remaining 26 positive clones (∼30%) encoded full-length, or almost full-length, in-frame cDNA inserts. These 26 plasmids were retransformed into yeast cells to exclude DNA-binding activity. In addition, the 26 positive plasmids were co-transformed into yeast cells with the pGTB9-laforin, or empty pGBT9, to confirm a specific interaction.
Fourteen of the 26 positive cDNAs (51%) corresponded to laforin, suggesting that laforin interacts with itself in brain and skeletal muscle. These laforin-encoding cDNAs could be grouped into six different classes of clones regarding whether they were full-length or started at nucleotide positions 163, 193, 196, 286 or 388 from the translation start site, respectively. It is interesting that many of the laforin clones isolated from cDNA libraries were truncated at their N-terminal region, suggesting that this N-terminal region is not required for the interaction of laforin with itself.
Three additional positive cDNA clones identified R5. R5, the human homolog of the murine Protein Targeting to Glycogen (PTG), is one of the four targeting proteins that bind to PP1 and glycogen, targeting the phosphatase to the intracellular glycogen particles. R5 is a 36 kDa protein that is expressed in a variety of tissues, with the highest levels being in skeletal muscle, liver and heart. Laforin and R5 present overlapping expression patterns with similar relative levels of expression in different human tissues, except in liver where laforin is low (9) and R5 is very high (Fig. 2). In addition to targeting PP1 to the glycogen particles, R5 acts as a molecular scaffold, assembling PP1 with its substrates at the glycogen particle and priming them for activation by insulin (20,22).
The three human R5 cDNA clones identified in the yeast two-hybrid screenings are identical and include an almost full-length cDNA insert, lacking 47 amino acids at the N-terminal region of the R5 protein. We reconstructed a full-length R5 cDNA in the pACT2 plasmid vector and co-transformed it in yeast with pGBT9–laforin or with empty pGBT9 to confirm the specificity of the interaction with the complete R5 protein.
Interestingly, when we co-transformed in yeast one of the pGBT9–laforin clones lacking the first 193 nucleotides of laforin with the pACT2-R5 plasmid the interaction was lost, indicating that the N-terminal region of laforin is required for the interaction with R5.
As a whole, the results of the yeast two hybrid screenings of the human skeletal muscle and brain cDNA libraries provide evidence that laforin interacts with itself and with R5. Furthermore, these data also suggest that different laforin domains are implicated in these interactions. The N-terminal region of laforin, including the glycogen binding domain, is necessary for the interaction between R5 and laforin, but it is not necessary for the interaction of laforin with itself.
In addition to the identification of interacting partners laforin and R5, we have isolated nine additional plasmid clones corresponding to a total of four different transcriptional units encoding a putative E3 ubiquiting ligase with a RING finger domain, a Ran binding protein, a member of the PIAS family and a protein related to Pals1, respectively. These clones are currently under investigation.
R5 domains involved in the interaction with laforin
PTG, the murine homolog of R5 (18), binds independently to glycogen, PP1 and glycogen synthase via three different binding sites. These PTG binding sites have been delineated through a combination of deletion and site-directed mutagenesis analyses (20) and are conserved at the amino acid level in R5. To characterize the regions of R5 involved in the interaction with laforin we constructed a series of R5 deletions in the pACT2 vector and used them in co-transformation experiments with pGBT9–laforin. The R5 constructs, named R51-183, R5116-183, R5116-238 and R5183-292, are depicted schematically in Figure 3A. We found that the deletion in R5 of the first 116 amino acids including the PP1-binding site, and the last 44 amino acids (R5116-238), has no effect in the interaction with laforin (data not shown). On the other hand, fragments of R5 including only the glycogen-binding site (R5116-183) or only the glycogen synthase-binding site (R5183-292) did not interact with laforin (Fig. 3A). Our conclusion from these experiments was that the interaction between laforin and R5 requires a minimal R5 fragment (amino acids 116–238) that includes both the glycogen synthase-binding domain and the glycogen-binding domain. In addition, an important conclusion from these data is that the interaction between R5 and laforin in the yeast cell is not mediated by glycogen through the glycogen binding site present in both proteins. This was shown by both R51-183 and R5116-183 fragments, as well as the R5D222A,D225A mutant (see below), which maintain glycogen binding domains and show no interaction with laforin.
To confirm the interaction between R5 and laforin by a non-yeast based assay we expressed the R51-183, R5116-183 and R5116-238 fragments in E. coli as a GST fusion proteins and purified them from the bacterial lysate supernatants by affinity chromatography. Recombinant human laforin expressed in COS7 cells were mixed with the GST–R5 fragments, the complexes precipitated with glutathione–agarose beads and the proteins in the precipitate analyzed by western blot with a rabbit anti-human laforin antibody. In agreement with the yeast two hybrids experiments, Figure 3B shows that laforin is precipitated by the R5116-238 fragment but not by the R51-183 nor the R5116-183 fragments, confirming a direct interaction between laforin and R5 through the R5 amino acids 116–238 region.
To further characterize the interaction between laforin and R5 we have constructed two R5 double mutants (R5V60A,F62A and R5D222A,D225A) in the pACT2 yeast expression vector and tested them for binding to laforin in co-transformation experiments with the pGBT9–laforin plasmid. The R5V60A,F62A and R5D222A,D225A mutants reproduce those used earlier to characterize the binding sites for PP1 and glycogen synthase in PTG (20). R5V60A,F62A includes two amino acid substitutions in the PP1-binding site and R5D222A,D225A two amino acid substitutions in the glycogen synthase-binding site.
We validated these mutants in our yeast two hybrid assay by co-transforming the R5 double mutants in the pACT2 vector with the skeletal muscle form of glycogen synthase (GYS1), in the pGBT9 vector. These preliminary experiments confirmed interaction between R5 and glycogen synthase with the pACT2–R5V60A,F62A construct, but not with the pACT2–R5D222A,D225A construct. In addition, co-transformation experiments using pACT2-laforin and pGBT9-GYS1 failed to demonstrate an interaction between laforin and glycogen synthase 1 (data not shown).
Figure 3C summarizes the results of the co-transformation experiments with laforin and the R5 double mutants, showing clearly that the interaction between laforin and R5 does not occur with the R5D222A,D225A mutant. To exclude that these results could be influenced by a lack of expression of the mutant R5 protein we performed a western blot with rabbit anti-human GST–R5 antibody and confirmed that the levels of expression in the yeast cells of the mutant and wild-type R5 proteins were equivalent (Fig. 3C). From these data we concluded that the laforin–R5 interaction requires a functional glycogen synthase-binding domain. In this respect, laforin shows identical structural requirements that present the PP1 substrates in their interaction with R5.
Immunolocalization and cell fractionation experiments
Co-transfection experiments in C2C12 murine myoblasts with R5 and laforin demonstrate co-localization of both proteins (Fig. 4A). Interestingly, the punctuated expression pattern of laforin in these experiments contrasts with the cytoplasmic distribution of laforin observed in single transfections, or in cloned C2C12 laforin stable transfectants (C2C12laforin; six independent clones; Fig. 4B). Exhaustive examination of the C2C12laforin preparations revealed that, occasionally, clustering of laforin was present in a few cells in all the C2C12laforin clones (Fig. 4B), perhaps suggesting increased glycogen synthesis in these particular cells. To further analyze whether laforin translocates to glycogen particles under glycogenic conditions we transfected the skeletal muscle form of glycogen synthase into the C2C12laforin clones. Figure 4C shows that in glycogen synthase-transfected C2C12laforin cells, the phosphatase laforin translocates with glycogen synthase to structures that resemble glycogen particles associated with the sarcoplasmic reticulum (Fig. 4C and D).
In a different set of experiments, we performed subcellular fractionation of C2C12laforin clones cultured in the presence or absence of glucose and insulin. In the absence of glucose, laforin shows a cytoplasmic distribution and appears in the 100 000g supernatant. In the presence of glucose, laforin translocates to the 100 000g precipitate containing the microsomal/glycogen particle fractions (Fig. 4E). Similar translocation was observed for other glycogen metabolizing enzymes, such as glycogen synthase. These data suggest that, under conditions of increased glycogen synthesis, like those resulting from the overexpression of R5 (PTG) (19) or glycogen synthase (23), or from culturing in the presence of glucose and insulin, after depletion of glucose, laforin translocates to the intracellular glycogen particles together with other enzymes of the glycogen metabolism. Laforin, like many other glycogen metabolizing enzymes that interact with R5, presents a glycogen binding domain that has been shown in vitro to be sufficient to mediate the binding of laforin to glycogen (16). Whether translocation of laforin to the glycogen particles requires R5 cannot be determined from the experiments presented here. Our results, however, clearly indicate that, under these experimental conditions, most, if not all laforin, is found associated with the glycogen particles and that therefore the activity of laforin in C2C12 muscle cells is probably restricted to this subcellular location.
The LD-associated EPM2A mutations disrupt the interaction between laforin and R5
To determine whether the LD-associated EPM2A mutations affect the interactions of laforin with itself and with R5, full-length laforin cDNA inserts carrying the W32G, F84L, R108C, T194I, G240S, G279S, Q293L, Y294N and P301L mutations were cloned in the pGBT9 vector and co-transformed in yeast with the pACT2–laforin, or the pACT2–R5 plasmids, respectively. To provide an additional positive control in these experiments we also included the analysis of the C266S mutation. As described earlier, the C266S–laforin lacks phosphatase activity but it is expected to maintain the folding and the structural properties of the wild-type laforin. Figure 5 summarizes the data obtained for all the mutated laforin proteins regarding the interaction with itself and with R5, respectively.
Only two of the LD-associated mutations reduce the interaction of laforin with itself below 40% of that observed for the wild-type laforin. These are R108C and T194I. This is in contrast with the observation that most of the LD-associated mutations lack phosphatase activity, glycogen binding capacity and do not interact with R5 (see below), suggesting that the interaction of laforin with itself depends on a linear, non-conformation-dependent, sequence located between the CBD and DSPD domains.
On the other hand, all the LD-associated EPM2A mutations disrupt the interaction between laforin and R5, suggesting that this interaction requires a conformation-dependent binding site. This would be in agreement with the observation that most of the tested LD-associated mutations also lack phosphatase activity and glycogen binding capacity, indicating that these mutations most likely, result in the profound disorganization of the structure of the protein.
The exception is the G240S mutation. Interestingly, while the G240S mutation hardly affects the phosphatase and binding to glycogen activities (Fig. 1), it causes a profound effect in the laforin–R5 interaction (Fig. 5B), suggesting that this mutation alters a residue directly involved in the interaction between the two proteins. Moreover, these data provide an explanation for why the G240S mutation results in LD, and supports the concept that the interaction with R5 is critical for the correct functioning of laforin.
It should be mentioned that the C266S mutation, that abolishes phosphatase activity but not glycogen binding (Fig. 1), increases significantly the interaction between laforin and R5, and does not affect the interaction of laforin with itself, suggesting that R5 could be one of the in vivo substrates of the phosphatase.
PPP1R5 as a second EPM2 locus
Genetic heterogeneity is expected in LD because linkage to chromosome 6q24 can be excluded in approximately 10–20% of the LD families (9–11). Based on the data presented in this report, PPP1R5, the gene encoding R5, should be considered a candidate gene in those LD pedigrees in which linkage to the EPM2A locus is excluded. We selected five of such LD pedigrees and sequenced the two exons of PPP1R5 in one affected member from each LD pedigree. We failed to identify any mutation in PPP1R5. To rule out linkage of LD to the PPP1R5 locus, we have selected and carried out segregation analysis in the five families for four genetic markers (GATA115E01, D10S583, D10S571 and D10S677) within a 10 cM region of chromosome 10q23–q24, including PPP1R5. Absence of linkage to PPP1R5 was evident. Thus, these analyses formally excluded PPP1R5 as responsible for LD in these five families. The PPP1R5 exclusion is based on the following findings: (i) four of the LD pedigrees non-linked to the EPM2A locus are consanguineous but the affected individual is heterozygous at the PPP1R5 locus; and (ii) the fifth pedigree is non-consanguineous but the three affected children in this kindred do not share the paternal and maternal PPP1R5 haplotypes.
It is likely that genetic heterogeneity in LD involves more than two loci in addition to EPM2A (Serratosa et al., unpublished observations). Therefore, our conclusion that PPP1R5 is not responsible of LD in the five LD pedigrees reported here does not necessarily mean that PPP1R5 should not be responsible of LD in other LD pedigrees non-linked to EPM2A.
To decipher the functional in vivo role of the dual-specificity phosphatase laforin it is essential to identify its potential regulatory or targeting subunits and its putative in vivo substrates. One way to approach this objective is to characterize the protein–protein interactions in which laforin may be involved. We have extensively used the yeast two-hybrid assay to identify potential proteins interacting with laforin in brain and skeletal muscle. Here we report two of these interactions. First, we provide evidence that laforin interacts with itself and second, that laforin interacts with R5 (PTG), one of the four glycogen targeting subunits, GL, GM, R5 and R6, which form complexes with the catalytic subunit of PP1. Identifying an interaction of laforin with proteins of the glycogen metabolism is very relevant to LD because Lafora bodies, the most characteristic feature of LD, have a composition that resemble glycogen (5,6). Moreover, our findings are particularly interesting because, in addition to PP1, R5 also binds to glycogen synthase and glycogen phosphorylase and, when overexpressed, R5 functions as an enhancer of glycogen accumulation (19,20,24,25).
Laforin presents multiple functional domains that are likely to be necessary for the correct functioning of the protein. In addition to the DSPD and the CBD reported earlier, the data presented here provide evidence of additional laforin domains that are responsible for the laforin–laforin and the laforin–R5 interactions. Based on the analysis of LD-associated EPM2A mutations, the laforin–laforin interaction site was mapped to the region between the CBD and the DSPD. This self-interaction site is probably a linear, non-conformation-dependent, domain because it is not affected by most LD-associated EPM2A mutations. On the other hand, the R5 interaction site in laforin is most likely a conformation-dependent site that is disrupted by the same mutations that cause the loss of phosphatase activity and glycogen binding capacity. This is perhaps not surprising if one considers that these mutations are causative of LD. LD mutations probably result in a profound disorganization of the structure of laforin. One exception among these mutations is the G240S mutation that interferes with the laforin–R5 interaction but does not affect the phosphatase and glycogen-binding, suggesting that this mutation may be directly involved in the interaction between the two proteins. These data provide an explanation for why the G240S mutation results in LD, illustrate the relevance of the laforin–R5 interaction in LD and suggest that the interaction with R5 is critical for the correct functioning of laforin.
Another important finding of these studies is the observation that laforin, like other glycogen metabolizing enzymes, translocates to the glycogen particles under conditions of increased glycogen synthesis. Moreover, the data from the translocation experiments indicate that, at least in skeletal muscle, the role of laforin is probably restricted to the glycogen particles because under glycogenic conditions laforin is found almost exclusively in that subcellular fraction. The specific function of laforin at the multiprotein complexes associated with the glycogen particles is, however, a matter of speculation. Overexpression of R5 results in an increased basal and insulin-dependent glycogen synthesis in many tissues including liver and skeletal muscle. This glycogen accumulation correlates with an elevated activity of PP1 on both glycogen synthase and glycogen phosphorylase, although overexpression of R5 predominantly increases the glycogen phosphorylase phosphatase activity of PP1 (22). The precise mechanism by which R5 overexpression enhances glycogen synthesis is not completely understood. One possibility that has been proposed is that R5 simply targets more PP1 to the glycogen particles (26). Additionally, it has been suggested that R5 may act as a molecular scaffold assembling glycogen synthase, phosphory-lase, phosphorylase kinase and the catalytic subunit of PP1 onto the glycogen particle (19). In this regard, it has been shown that R5 presents independent binding sites for PP1, glycogen and the PP1 substrates. Interestingly, glycogen synthase, phosphorylase and phosphorylase kinase interact with R5 at the same site between the glycogen binding domain and the C-terminus, so that they are not able to bind to R5 at the same time (20). Using R5 mutagenesis and R5 deletion constructs we have shown that the laforin-binding site in R5 coincides with the site that mediates the interaction of R5 with the PP1 substrates. These findings may suggest that laforin is also a substrate of PP1 and that the interaction between laforin and R5 facilitates the action of PP1 on laforin. This implies that laforin may be regulated by PP1. Alternatively, R5 may be a substrate for laforin and perhaps in this way laforin modulates R5 ligand specificity. Recently we have obtained evidence that both R5 and laforin can be phosphorylated in vitro (SRdeC and PS, manuscript in preparation) which may provide support to this hypothesis.
The functional analyses of laforin presented here represent a significant advance in our understanding of LD. This work demonstrates that laforin requires a complex pattern of interactions for activity that can be inactivated at multiple levels by single residue substitutions. Most important, our data provide evidence that laforin interacts with R5 and it is targeted to the intracellular glycogen complexes where the phosphatase forms part of a multiprotein complex associated with proteins important in the glycogen metabolism. Taken together these data reinforce the concept that LD is related to the glycogenoses and suggest that laforin could be involved in the regulation of the glycogen metabolism. Whether this regulation is achieved at the level of R5, altering its binding capacity for PP1 or the PP1 substrates, at the level of the glycogen enzymes, PP1, glycogen synthase, phosphorylase or phosphorylase kinase, modifying their activity, or at a cellular level, controlling the intracellular translocation of the multiprotein complexes including all these glycogen enzymes, is presently under investigation.
MATERIALS AND METHODS
Laforin was amplified from human liver cDNA by PCR and cloned into the prokaryote vector pGEX-A (Invitrogen). The final construct, pGST::laforin, encoded a recombinant laforin protein fused to GST at its NH2-terminal. The cDNA of laforin was also cloned into the mammalian expression vectors pCINeo (Promega), pcDNA3 (Invitrogen), pGBT9 and pACT2 (Clontech). The cDNA of R5 was amplified from the IMAGE clone 5088205 and cloned into the vectors pACT2 (Clontech) and pcDNA3/Flag (Invitrogen). The fusion protein expressed from this vector, pFlag::R5, contains an NH2-terminal FLAG epitope of eight amino acids. To produce fusion proteins with an NH2-terminal enhanced green fluorescence protein (EGFP), the cDNAs of laforin and R5 were cloned into the pEGFP-C1 vector (Clontech). These plasmids were used as templates for the introduction of LD missense mutations by PCR, using the QuickChange Site-Directed Mutagenesis Kit (Stratagene) and appropriate mutagenic oligonucleotides. The cDNA of human glycogen synthase was amplified from the IMAGE clone 3953692 and also cloned into the pEGFP-C1 and pGBT9 vectors (Clontech). All expression constructs were fully sequenced to rule out the presence of undesired mutations resulting from PCR amplification.
Expression and purification of recombinant proteins
Expression constructs were transformed into E. coli for expression. Bacterial clones were inoculated into LB cultures in the presence of 100 µg/ml ampicillin and grown at 37°C until the OD600 reached 0.9–1.0. GST–fusion protein synthesis was induced upon addition of 100 µM IPTG and the cultures were incubated for a further 1 h at 25°C. Bacteria were collected, washed, resuspended in lysis buffer (PBS, 10 µM benzamidine, 1.5 µM aprotinin, 5.25 mM leupeptin, and 0.5 mM PMSF) and lysed with a French Press. Proteins were solubilized with 0.1% Triton X-100 for 1 h at 4°C and lysates were then cleared by centrifugation for 10 min at 10 000g at 4°C and finally passed through a 0.45 µm filter. The soluble, GST-fused proteins were purified by affinity chromatography using Glutathione–agarose beads (Sigma) and eluted with 50 mM reduced Glutathione (Sigma), 250 mM Tris–HCl pH 8.0. The proteins were stabilized with 2 mM DTT and 15% glycerol, and later dialysed against 50 mM Tris–HCl pH 8.0, 150 mM NaCl, 2 mM DTT and 15% glycerol. The purity of enzyme preparations was monitored by 10% SDS–PAGE and Coomassie staining. Additionally, we used TNT Coupled Reticulocyte Lysate System (Promega) to express proteins in vitro. Synthesis was performed as recommended by the manufacturer. Briefly, 0.5 µg of each plasmid was added to 12.5 µl of reticulocyte lysate and incubated for 2 h at 30°C. Products were analysed by SDS–PAGE and western blot.
Polyclonal mouse antibodies against recombinant human laforin was raised as previously described (27). In brief, 40 µg of GST::laforin, purified as described above, was mixed with Ribi adjuvant (Ribi ImmunoChem Research, MT) and injected intraperitonealy in S/W mice. Five immunizations spaced 2 weeks apart were conducted before injection of 106 T-180 sarcoma cells. Ascites fluid was drained from the peritoneal cavity after the mice developed an obvious tumor. Polyclonal rabbit antibodies against recombinant human laforin and R5 were obtained by subcutaneous immunization of GST recombinant proteins in rabbits. The IgG fraction of the antisera were obtained by chromatography on DEAE-Sephacel. A mouse anti-skeletal muscle glycogen synthase antibody was also used (Chemicon).
Phosphatase activities of the recombinant GST::laforin proteins were assayed using 20 mMp-nitrophenyl phosphate (pNPP, Sigma Diagnostics, St Louis, MO, USA) as the substrate in a final volume of 150 µl buffer (50 mM Hepes pH 6.0, 50 mM NaCl, 5 mM EDTA and 50 mM β-mercaptoethanol) at 37°C for 15–60 min. Absorbances were read at 405 nm and a molar extinction coefficient of 1.78×104 M−1 cm−1 was used to calculate the concentration of the p-nitrophenolate ion produced in the reaction.
Glycogen binding assays
For in vitro experiments, 15 µl of each TNT reaction, containing wild-type or mutant laforin proteins, was diluted in 85 µl of binding buffer [50 mM Tris–HCl pH 7.5, 150 mM NaCl, 0.1% (v/v) 2-mercaptoethanol, 0.1 mg/ml bovine serum albumin] and then was ultracentrifugated in a TL-100 (Beckman) at 100 000g for 90 min at 4°C, to pellet insoluble proteins. Soluble fractions were incubated in the absence or presence of 2 mg of glycogen (Roche) previously pelleted by ultracentrifugation. After 1 h at 4°C, samples were ultracentrifugated at 100 000g for 90 min. Supernatants and pellets fractions were run in a 10% SDS–PAGE, transferred to nitrocellulose membranes and developed using mouse polyclonal anti-laforin antibodies.
A northern blot (Clontech, Palo Alto, CA, USA) containing poly (A)+ RNA from different human tissues was hybridized with a 2220 bp R5 probe encoding amino acids 73–317 and the 3′-UTR. The probe was radiolabeled with [32P]-α-dCTP using the random primer labeling method (28).
Yeast two-hybrid screens
We used the Matchmaker Two-Hybrid System 3 (Clontech) with pGBT9-[GAL4BD]–laforin as a bait and a human skeletal muscle or brain cDNA library (pACT2–[GAL4AD]) pre-transformed in the yeast strain Y187 (Mat a) of Saccharomyces cerevisiae. The assay was performed as recommended by the manufacturer. In brief, yeast cells (AH109, Mat α) containing the bait plasmid were conjugated with the cDNA library for 20–24 h at 30°C with mild agitation (30–50 rpm). We screened approximately 3×106 transformants for growth, in synthetic dropout (SD) interacting media lacking leucine, tryptophan, adenine and histidine amino acids. The plates were left for 5–12 days. The plasmids containing the cDNA sequence of the putative laforin interaction partners were extracted from the conjugated AH109/Y187 yeast cells, transformed in E. coli DH5α′F by electroporation and sequenced. The plasmids were then retransformed into the conjugated AH109/Y187 yeast cells and assayed for growth on interaction selective SD plates to exclude DNA-binding activity. Appropriate controls were included. In a second round of testing, the library plasmids and the pGTB9–laforin were co-transformed into empty AH109 yeast cells using the LiAc/ssDNA/PEG-heat shock protocol. Transformants were assayed again for growth on interaction selective SD plates with appropriate controls. The interaction in the co-transformations with the different recombinant laforin and R5 proteins were determined as follows: co-transformations were plated both on SD interacting selective media plates and SD plasmid selective plates (-leucine, -tryptophan) and the number of colonies in the SD plates expressed as a percentage of the colonies in the plasmid plates and normalized to the results obtained for the co-transformations with the wild-type laforin and R5 proteins. β-Galactosidase assays on duplicates of these co-transformations showed a linear correlation with the data obtained from the selective SD plates (data not shown).
Transfection experiments and expression of recombinant proteins in eukaryotic cell lines
Cos7 cells were transfected with 1 µg cDNA of laforin (pCINeo::laforin) and GFP::R5 fusion proteins using FuGENE6 (Roche). After 24–36 h, the cells were washed twice with PBS. Cells were then trypsinized and the trypsin then blocked firstly with complete Dulbecco's modified Eagle medium (DMEM; GIBCO) and 4 µg/ml soybean inhibitor and secondly with PBS and 0.2 µg/ml soybean inhibitor. Cells were then harvested with ice-cold protein lysis buffer consisting of 25 mM Hepes, pH 8.5, 150 mM KCl, 2 mM EDTA, 0.1% IEGPAL (Sigma), 1 mM PMSF, 10 µM benzamidine, 5.25 mM leupeptin and 0.5 mM DTT. Cells were lysed by three rounds of freezing and heat-shock. The cell lysates were then fractionated by firstly centrifuging at 10 000g for 10 min. The soluble fraction was collected for immunoprecipitation.
To generate laforin stable transfected cell lines, 5×105 fibroblast cells or C2C12 (murine myoblats) cells were transfected with 4 µg of pCIneo::laforin using Nucleofector (Amaxa biosystems) 48 h post-transfection; transfected cells were selected by using 1.5 mg/ml of G418 (GIBCO) diluted in DMEM. Medium was replaced every 3–4 days. After 2 weeks, cells were cloned by limit dilution and individual clones were isolated and grown. The presence of transfected laforin in the clones was confirmed by western blot and immunofluorescence.
GST and GST-fusions of R5 fragments were pre-incubated for 30 min at 4°C with Glutathione–agarose beads (Sigma) that had been preblocked with TBS plus bovine serum albumin (1 mg/ml), 0.5% Triton X-100, and 1 mM DTT. After washing with TBS plus 0.1% IEGPAL (Sigma), the beads were incubated with 10 000g supernatants of laforin-transfected COS7 cell lysates for 1 h at 4°C. Subsequently the beads were sedimented (30 s at 1000g), washed three times with 500 µl of TBS plus 0.1% NP40 and probed in western blot for the presence of R5 laforin. The COS7 cells were lysed in a buffer containing 25 mM Hepes, pH 8.5, 150 mM KCl, 2 mM EDTA, 0.1% IEGPAL (Sigma), 1 mM PMSF, 10 µM benzamidine, 5.25 mM leupeptin and 0.5 mM DTT.
Cells were grown on chamber slides (NUNC) and transfected with 1 µg DNA using FuGENE6 (Roche). Eighteen hours post-transfection, cultures were rinsed twice in PBS and fixed for 20 min at RT in 4% paraformaldehyde. After washing with TBST buffer (50 mM Tris–HCl pH 7.4, 150 mM NaCl, 3% BSA and 0.2% Tween-20), cells were incubated with rabbit anti-laforin antibody diluted in the same buffer for 1 h at 37°C. Biotinylated secondary antibodies and Texas Red-conjugated streptoavidin (Molecular probes) were used to localize the primary antibodies. Cells were mounted with moviol and immunostaining was visualized under fluorescence microscopy.
We thank A. Zorzano, J. Guinovart, J. Domiguez and M. Garcia-Rocha for many helpful discussions regarding glycogen metabolism. We also thank the DNA Sequencing Laboratory at the Centro de Investigaciones Biológicas for their contribution to this work. This work was supported by grants from the Asociación Lafora España, the Spanish CICYT (SAF2001-0613) and the Instituto Carlos III (PI02/0536, G03/054 and G03/011). K.E.H. was supported by a Marie Curie postdoctoral fellowship (HPMD-CT-2001-01259) from the European Commission and P.G.-G. by a fellowship from the Asociación Lafora España and by the Centro de Investigación en Enfermedades Neurológicas.
NOTE ADDED IN PROOF
Following submission of this manuscript, Chang et al. reported the identification of a second gene responsible for Lafora Disease (EPM2B) encoding a putative E3 ubiquitin ligase with a RING finger domain, called malin. [Chang, E.M., Young, E.J., Ianzano, L., Munteanu, I., Zhao, X., Christopoulos, C.C., Avanzini, G., Elia, M., Ackerley, C.A., Jovic, N.J. et al. (2003) Mutations in NHLRC1 cause progressive myoclonus epilepsy. Nat. Genet., 35, 125–127]. Based on their findings, Chang et al. propose a defect in the ubiquitin-degradation pathway in Lafora disease. However, the biochemical composition of Lafora bodies, very similar to glycogen, shows a very low protein content, which is in contrast to that of the aggregates and inclusion bodies observed in other disorders involving the ubiquitin-degradation pathway. In addition to the traditional role of ubiquitination in protein degradation, it is increasingly recognized that incorporation of ubiquitin represents an important post-translational modification that regulates target protein localization and activity, much like phosphorylation. Charac-terization of potential interactions between laforin and malin, as well as identification of the pathways through which both laforin and malin operate, are needed to understand the abnormalities in the cellular physiology that results in lafora disease.
To whom correspondence should be addressed at: Departamento de Inmunología, Centro de Investigaciones Biológicas, Consejo Superior de Investigaciones Científicas, Ramiro de Maeztu 9, 28040 Madrid, Spain. Tel: +34 918373112; Fax: +34 915340432; Email: email@example.com
The authors wish it to be known that, in their opinion, the first three authors should be regarded as joint First Authors.
Departamento de Pediatria, Facultad de Medicina, Universidad Autónoma de Madrid, Madrid, Spain.