A human skeletal muscle interactome centered on proteins involved in muscular dystrophies: LGMD interactome
- Gaëlle Blandin1, 2,
- Sylvie Marchand1,
- Karine Charton1,
- Nathalie Danièle1,
- Evelyne Gicquel1,
- Jean-Baptiste Boucheteil1,
- Azéddine Bentaib1,
- Laetitia Barrault1,
- Daniel Stockholm1,
- Marc Bartoli1, 2 and
- Isabelle Richard1Email author
© Blandin et al.; licensee BioMed Central Ltd. 2013
Received: 3 August 2012
Accepted: 7 February 2013
Published: 15 February 2013
The complexity of the skeletal muscle and the identification of numerous human disease-causing mutations in its constitutive proteins make it an interesting tissue for proteomic studies aimed at understanding functional relationships of interacting proteins in both health and diseases.
We undertook a large-scale study using two-hybrid screens and a human skeletal-muscle cDNA library to establish a proteome-scale map of protein-protein interactions centered on proteins involved in limb-girdle muscular dystrophies (LGMD). LGMD is a group of more than 20 different neuromuscular disorders that principally affect the proximal pelvic and shoulder girdle muscles.
Results and conclusion
The interaction network we unraveled incorporates 1018 proteins connected by 1492 direct binary interactions and includes 1420 novel protein-protein interactions. Computational, experimental and literature-based analyses were performed to assess the overall quality of this network. Interestingly, LGMD proteins were shown to be highly interconnected, in particular indirectly through sarcomeric proteins. In-depth mining of the LGMD-centered interactome identified new candidate genes for orphan LGMDs and other neuromuscular disorders. The data also suggest the existence of functional links between LGMD2B/dysferlin and gene regulation, between LGMD2C/γ-sarcoglycan and energy control and between LGMD2G/telethonin and maintenance of genome integrity. This dataset represents a valuable resource for future functional investigations.
KeywordsInteractome Muscular dystrophies Yeast-two hybrid
Selected Interacting Domains
Predicted Biological Score
Proximity ligation assay
- M F:
Limb-girdle muscular dystrophy
Limb-girdle muscular dystrophy, type 2
Congenital muscular dystrophy 1B
Autosomal dominant myopathy with proximal muscle weakness and early respiratory muscle involvement
Hyaline body myopathy
Adult onset distal myopathy
- MFM/ ARVC:
Myofibrillar myopathy with arrythmogenic right ventricular cardiomyopathy
Distal myopathy with pes cavus and areflexia
Myopathy with excessive autophagia.
The skeletal muscle tissue with its movement generation capacity is the organ of voluntary action but it plays also a major role in metabolic homeostasis. It is composed of long multinucleated cells, the myofibers, which possess a highly structured organization to ensure the dynamics and coordination of muscle contraction and resistance to resulting physical stresses. Notwithstanding its very well organized structure, the muscular tissue presents an important plasticity, which is necessary for adaptation to physical and metabolic constraints. These combined characteristics of structure and plasticity depend on the concerted action of protein complexes and of metabolic and signaling pathways. These features together with the complexity of the skeletal muscle organization and its connection with human disorders makes it an interesting tissue for proteomic studies aimed at understanding functional relationships of interacting proteins in both health and diseases.
During the last decade, an increasing number of studies have been carried out to produce and analyze large-scale protein-protein interaction networks in various bacterial and eukaryotic model systems. Several studies have investigated the human interacting-proteins network, either at a genome-wide scale [1–3] or with the aim of exploring a targeted interaction network [4–6]. Remarkably, several disease-targeted projects have proven to offer a very powerful strategy to address the role and function of the proteins involved, as exemplified by Huntington’s disease  and in spinocerebellar ataxia . These works have capitalized on several high-throughput technologies (yeast-two hybrid (Y2H) system and Tandem Affinity Purification) to detect binary protein-protein interactions (PPI) or protein complexes, and have combined them with computational methods to propose “interactome” networks.
As an approach towards the identification of skeletal muscle functional networks, we undertook a large-scale study to establish a proteome-scale map of protein-protein interactions centered on proteins involved in limb-girdle muscular dystrophies (LGMD). LGMD is a group of neuromuscular disorders with autosomal dominant (LGMD1) or recessive (LGMD2) inheritance modes that principally affect the proximal pelvic and shoulder girdle muscles (for a recent review see ). More than twenty different genetic entities were identified so far but it is estimated that 30-40% of patients clinically diagnosed with LGMD still do not have a genetic signature of their disease  with at least 25% of families who are not linked to any known locus and 40% of isolated cases with no detected mutation in any known LGMD gene. The known LGMD-causing genes encode proteins expressed in a variety of cellular compartments and involved in diverse biological functions that are not yet fully understood. Yet, the present knowledge has highlighted the importance of the components of the dystrophin-glycoprotein complex and α-dystrophycan glycosyltransferases for membrane stability, and the implication of other LGMD proteins in processes such as regulation of sarcomere structure or nuclear stability, for the survival of the muscle fibers . No curative treatment is currently available for LGMD, which pleads for the elucidation of the pathological mechanisms implicated in the diseases in order to propose innovative therapeutic approaches.
In this context, we selected a large-scale strategy to identify novel protein-protein interactions that could shed light on the biological processes at the origin of LGMD pathogenesis. We selected 13 LGMD-causing proteins and related proteins and ran high-throughput Y2H assays to build a first interactome network that we further expanded by performing additional secondary and tertiary Y2H screenings with new bait proteins of interest identified as preys in the primary screenings. The resulting LGMD-centered interaction network was established by combining results of 87 screenings based on 76 different bait proteins and incorporates 1018 proteins connected by 1492 direct binary interactions. Using both experimental and bioinformatics tools, we assessed the overall quality of this Y2H network and isolated a high-confidence (HC) sub-network of 705 PPIs associating 497 proteins. The Y2H LGMD-centred network and its HC sub-network were compared to a literature-based network. Gene Ontology (GO) term analysis showed that the LGMD-centered interactome and especially its HC sub-network, is much more specifically enriched in proteins associated with the muscular tissue and the cytoskeleton than the literature-based dataset. Among the interesting outcomes of our study are the strong inter-connectivity of the LGMD proteins which, in addition to several direct links, present a number of indirect associations thanks to sarcomeric proteins, the identification of candidate genes for orphan neuromuscular disorders (NMDs) and the discovery of new possible functions for LGMD proteins. In particular, we suggest the existence of functional links between LGMD2B/dysferlin and gene regulation, between LGMD2C/γ-sarcoglycan and energy control and between LGMD2G/telethonin and maintenance of genome integrity.
The protein interactions from this publication have been submitted to the IMEx ( http://www.imexconsortium.org/) consortium through IntAct (pmid 22121220) and assigned the identifier IM-16425. The entire network can be browsed using the PIMRider software at http://pimr.hybrigenics.com.
Y2H bait design
Bait design was organized in three successive rounds in which primary, secondary and tertiary baits were selected. Structural and functional domain predictions from TM-HMM , SignalP3.0  or PFAM (PFAM 23.0, release 19/08/2008 ) were used to exclude hydrophobic trans-membrane domains, signal peptides and transcriptional trans-activation domains from bait constructs. In addition, to favor the identification of novel protein-protein interactions, we usually selected regions on the proteins that had not been previously documented for their functional role. For selection of secondary and tertiary bait proteins and for design of their bait sequences, we used additional bibliographic searches and other criteria computed from our Y2H results such as the Predicted Biological Score (PBS) categories and information from Selected Interacting Domains (SID; see below section “Identification of interacting fragments and scoring of the interactions”). Some examples of bait candidates that came to our attention included targets of choice for therapeutic strategies such as proteins that participate in signaling pathways, proteins involved in various forms of myopathies or proteins expressed in typical muscle cellular compartments such as the sarcomere or sarcoplasmic reticulum.
Bait cloning and library construction
Bait sequences were PCR-amplified from MRC Gene Service or Invitrogen plasmids or from a random primed cDNA library obtained by reverse transcription of a poly(A) RNA library isolated from adult (Ambion AM7983) or 18-19 week-old fetal (Stratagene #778020) human skeletal muscles. Bait PCR products were cloned in the pB27 plasmid, a plasmid derived from the original pBTM116 , as a LexA C-terminal fusion . Plasmid DNA was purified with the QIAprep Spin Miniprep (QIAGEN), verified by full insert sequencing and introduced into the L40ΔGal4 (MATa) bait yeast strain . Alternatively, prey fragments were directly extracted from the prey plasmid and subsequently cloned in pB27 to use them as secondary or tertiary baits.
The prey library in yeast was constructed from adult (Ambion AM7983) and fetal (Stratagene #778020) human skeletal muscle poly(A) RNA. Random-primed cDNA fragments were prepared from these two RNA pools and cloned in the pP6 plasmid derived from the original pACT2 (Clontech) as a C-terminal fusion of the Gal4 transcription activating domain. Altogether, 90% of the plasmids contained a cDNA insert with an average size of 600 bp. After amplification in Escherichia coli (50-100 million independent clones), the Y187 (MATalpha) yeast strain was transformed with an equimolar pool of the adult and fetal cDNA libraries. Ten million independent yeast colonies were then collected, pooled and stored at -80°C as equivalent aliquot fractions of the same prey library. Validation of the prey library was performed by recapitulating several published interactions as described in . Bait proteins belonging to different functional classes were used: a GTPase (Rac1), a transcription factor (TP53), a splicing factor (SF1) and a component of a E3-ligase complex (BTRC).
Screening procedure and identification of prey fragments
Y2H screens were performed using a mating method as described in  at the Hybrigenics facility (Hybrigenics, Paris). As first step, small-scale screenings were performed to assess toxicity and auto-activation capacity of the baits and to adjust selective pressure of the screens accordingly. In particular, the optimal concentration of 3-aminotriazol (3-AT) was determined prior to performing each large-scale screen. Auto-activating baits able to activate transcription of the reporter gene by themselves were identified and were not considered for large-scale screenings. Subsequently, each bait clone was tested in a full-size screen against an average of 103 million yeast prey clones, equivalent to ten-fold coverage of the library. All positive clones were picked and the corresponding prey fragments were PCR-amplified and sequenced at their 5′ and 3′ junctions. Sequence contigs were built and identified by comparison to the NCBI Human RefSeq database as described in .
Identification of interacting fragments and scoring of the interactions
Following contig assembly of positive clones, the common sequence shared by the assembled prey fragments was used to define the SID along each prey protein. Furthermore, for each interaction, a PBS was computed with E-values ranging from 0 to 1 to establish six distinct categories: PBS-A to -E (see  for details on calculation). The technically most reliable interactions were associated with the PBS-A, -B or -C categories (with P values < 1 e-10 for PBS-A;< 1 e-5 for PBS-B and < 1 e-2.5 for PBS-C) and are found in two reciprocal and independent screens (X->Y and Y->X) and/or in interaction cycles (X-Y, Y-Z and X-Z) and/or in a single screen but with many overlapping prey fragments. Interactions were assigned to the PBS-D category when they were supported either by a single experimental clone from a screen or by several clones bearing the same start and stop positions, the SID being identified by a singleton fragment instead of a family of several overlapping fragments. This PBS-D category corresponds to a heterogeneous group of interactions that theoretically could consist of technical false-positive interactions as well as true-positive interactions hardly detectable by Y2H systems (due to constraints in tri-dimensional conformation of bait or prey domains, toxicity in yeast, poor mRNA representation of the prey in the library, …). All the PBS-D should therefore be considered as putative unless validated by a second technique. The PBS-E category characterizes SID that have been found as prey in more than ten independent screens with unrelated bait proteins in all screenings performed with human libraries at the Hybrigenics facility. These interactions potentially represent possible false-positives of the Y2H system as well as interactions with proteins known to be highly connected due to their biological function or with proteins containing a biochemically promiscuous motif. Finally, interactions with proteins or domains corresponding to known false positives of the Y2H system as it is described above were removed from the data and from our analyses. Examples of yeast growth assays describing interactions with the different PBS categories using the same experimental procedures can be found in [16–18].
The antibodies used for immunoprecipitation of the baits are BD Biosciences anti-TCAP (T26820-050), Novocastra Laboratories Ltd anti-DYSF (NCL-Hamlet) and Santa Cruz Biotechnology anti-ABI1 (sc-30038), anti-ACTN2 (sc-15335), anti-DES (sc-14026), anti-MYOM1 (sc-30390) and anti-TCAP (sc-8725).
The antibodies used for prey detection by western blot are Abcam anti-SNAPIN (ab37496), Abnova anti-ADPGK (H00083440-M01), anti-APPL1 (H00026060-A01) and anti-ENO1 (H00002023-M04), Aviva anti-KBTBD10 (ARP38732_T100) and Santa Cruz Biotechnology anti-KIF1B (sc28540) and anti-KTN1 (sc33562).
The antibodies used for immunochemistry and Duolink assays and their corresponding dilutions are: Abcam anti-CMYA5 (ab75351, 1:50) and anti-OPTN (ab23666, 1:100), Abgent anti-DGKD (AP8126b, 1:50), Abnova anti-DNAJB6 (H00010049-M01, 1:100) and anti-EEF1G (H00001937-M01, 1:50), Novocastra Laboratories Ltd anti-DYSF (NCL-Hamlet, 1:20), Proteintech Group anti-SNAPIN (10055-1 AP, 1:50), Santa Cruz Biotechnology anti-ACTN2 (sc-15335, 1:100), anti-ALMS1 (sc-54507, 1:50), anti-APPL1 (sc-67402, 1:50), anti-DES (sc-14026, 1:100), anti-FLNC (sc-48495, 1:100), anti-KIF1B (sc-28540, 1:50), anti-MYOM1 (sc-30390, 1:100), anti-MYOM2 (sc-50435, 1:200) and anti-NEB (sc-28286, 1:100) and Sigma anti-NPHP3 (HPA009150, 1:75).
The bait proteins were isolated from R9 cell extracts (a gift from Dr. Anne Galy, Inserm U790, Evry, France) at myoblast or myotube stage (7 to 10 days of differentiation) or from gastrocnemius muscle excised from four week-old mice and homogenized in 6 ml lysis buffer (Tris 20 mM, pH 7.5, NaCl 50 mM, EGTA 2 mM, Triton 1%, Protease Inhibitor Cocktail (Complete mini, Roche), E64 2 μM) using a FastPrep-24 apparatus (MP Biomedicals). The mouse samples correspond to a protocol approved by Genethon’s ethics committee under the number CE11_014 and performed in accordance with the directive of 24 November 1986 (86/609/EEC) of the Council of the European Communities. After centrifugation of the lysates, 500 μg to 1 mg of proteins in 1 ml were incubated with 30 μl of protein G–Sepharose beads (Amersham) for 1 h to clear from nonspecific binders. The protein extract was then subjected to immunoprecipitation by 1 h incubation at +4°C with 2 to 4 μg of primary antibodies corresponding to the baits, then 30 μl of protein G Sepharose beads (Amersham) were added and incubation was carried out for 2 h or overnight at +4°C.
After centrifugation at 1000 g for 5 min, the immunocomplex was washed three times with 1 ml of buffer and resuspended in 15 μl 4x NuPAGE LDS sample buffer (Invitrogen) and dithiothreitol reducing agent. Samples were then heated at +70°C for 10 min and centrifuged briefly. Protein complexes were separated by electrophoresis on SDS-PAGE NUPAGE 4-12% Bis-Tris gel (Invitrogen). Transfer of the proteins was performed on PVDF membrane and verified by staining with Ponceau red. Immunostainings were performed with primary antibodies corresponding to prey and IRD-680 or 800 donkey anti-mouse, -rabbit or -goat as secondary antibodies according to LI-COR’s protocol. Bands were then visualized with the Odyssey infra-red imaging system (LI-COR-Biosciences) at 700 nm (red) and 800 nm (green).
Indirect immunofluorescence microscopy assays were carried out on transversal cryosections prepared from normal human paravertebral striated muscles of a 13-year old female biopsy obtained from the biobank Myobank under the validation number AC-2008-87 from the French ministry of research (Institute of Myology, Paris). The sample was treated anonymously. Frozen slides were air-dried for 30 min at room temperature, fixed with 4% PAF for 5 min, washed 3 × 5 min in PBS, incubated in a blocking buffer (4% BSA, 0.02% Triton) for 30 min, washed in PBS, then incubated with a biotin blocking solution (Vector Laboratories, SP-2001) for 15 min and washed in PBS for 5 min. Slides were stained at room temperature for 1 h or at +4°C overnight with primary antibodies diluted in the labeling solution (1% BSA / PBS). Slides were then incubated with a donkey anti-mouse-Alexa 488 for dysferlin and a donkey (anti-rabbit or anti-goat) biotinylated secondary antibody for its partner (dilution 1:1000) for 45 min, washed 3x 5 min in PBS and stained with streptavidin coupled to Alexa-594 (Molecular Probes, dilution 1:500 in PBS) for 30 min. For nucleic acid staining, slides were then incubated with TOPRO-3 (Molecular Probes, dilution 1:2000) for 5 min, washed 2 x 5 min in PBS and 1x in water for 2 min. Slides were subsequently mounted in Fluoromount-G™ (SouthernBiotech, 0100-01). Images were acquired using the 40x or 63x objective of a Zeiss Axiovert 100 M. LSM.510 Meta laser scanning confocal microscope and the constructer software. Colocalization analyses were performed by statistical analysis of the correlation between the intensity values of red and green pixels in a dual- channel image. The JACop plug-in  for ImageJ (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, http://imagej.nih.gov/ij/, 1997-2011 ) was used to calculate Pearson’s Correlation coefficient. Co-localization was defined as strong for 0.5<R≤1, medium for 0.25<R≤0.5 and low for R≤0.25.
Proximity ligation assays
The Duolink® kit (Olink Bioscience) is based on the use of two unique and bi-functional probes called PLA™, each probe consisting of a secondary antibody attached to a unique synthetic oligonucleotide that acts as a reporter. After a 10 min fixation with paraformaldehyde 4% and blocking (BSA 5% in PBS) steps, muscle sections were stained with one or two primary antibodies depending on the experiment (single protein detection or detection of interacting proteins) over-night at +4°C. After washing, the sections were incubated with the secondary oligonucleotide-linked antibodies (PLA probes) provided in the kit. The oligonucleotides bound to the antibodies were hybridized, ligated, amplified, and detected using a fluorescent probe (Detection Kit 563). Dots were detected with the Zeiss laser scanning confocal microscope and intensity signal counted using ImageJ software ( http://imagej.nih.gov/ij/). A series of controls were performed for each analysis (bait antibody only, prey antibody only and negative control for which the primary antibody is omitted).
For quantification analysis: three images were acquired under the same conditions (laser power, PMT gain and pinhole) for each experiment. For each image, five fibers were randomly selected and used to count all positive spots within each compartment (total of 15 cells). The regions of interest (ROI) for membrane and cytoplasm compartments were separately delimitated manually and signal quantification was performed on all identified spots using the ImageJ software. For each compartment, we considered that the PPI was validated by the assay when the mean signal ratio between the PLA images of the PPI, “PPI signal”, and the control images of the prey, “PREY signal”, was superior to 0.2, indicating that the interaction with the prey potentially recruited more than 20% of the interacting partner in the delimited compartment.
Bioinformatics and statistical analyses
IpScan  with Interpro 17.0  was used to annotate the protein sequences. The SID coordinates were compared with the position of the different Interpro domains. Cytoscape tools ( http://www.cytoscape.org) were used to infer connectivity, a parameter that indicates the number of proteins that directly interacts with a given protein. Comparison of PPIs identified by our Y2H screenings with previously published PPIs was performed using the iRefWeb interface ( ; http://wodaklab.org/iRefWeb/) by considering direct interaction found in mammals.
GO mapping and clustering were performed with the DAVID 6.7 web interface [23, 24] using the Functional Annotation Clustering tool and the GOTERM_FAT annotation categories in order to filter the broadest GO terms. For the LGMD-centered dataset, a list of official gene symbols was used to identify the proteins and within each identified GO cluster, GO terms were analyzed in terms of hierarchy to identify the most specialized children terms common to all proteins within the cluster and these terms were reported as “shared GO” annotations (Additional file 1: Table S1). To analyze whether our datasets were statistically different from a random dataset, GO clustering was also performed with a list of Uniprot accessions for the 19220 human protein-coding genes (HGNC, http://www.genenames.org/). The number (Shared-i) of human proteins with which a given bait protein (Bait-I) shared a GO cluster was calculated for all three GO classes (BP, MF and CC) and all baits and the number of protein pairs not sharing a GO cluster was deduced (NonShared-i = 19220 - Pi). The overall frequency of expected shared and non shared protein pairs was calculated as the ratio between the sum of Shared-i and the sum of all pairs (76 × 19220), and the sum of NonShared-i and the sum of all pairs, respectively. A Chi-2 test (P<0.05) was used to compare expected values with observed values from the LGMD-centred dataset or the subset consisting of all PBS-A to -C categories.
GO enrichment analyses were performed using the DAVID Functional Annotation tool with Uniprot accession numbers as identifiers, the Homo sapiens background and the GOTERM_FAT annotation categories. Enrichment at 1% significance level was defined with a modified Fisher exact P value (the “EASE” score) as recommended by the DAVID interface.
Statistical analysis of obtained proportions for the other analyses was done using the Fisher test function in R.
Bait design and screening procedure
A/Description of primary baits
Protein symbol and LGMD form
Bait domain coordinates (aa)*
Bait domain description
C2-like domain + exons15-16
full length protein with the C129S mutation in the autocatalytic site
N-terminal DYSF regions containing the first three C2 domains
central DYSF domain
C-terminal DYSF domain containing the last four C2 domains and excluding the transmembrane span
full length protein
tripartite motif-containing 32
full length protein minus the RING domain
exons 4-8 from the Z-Disc region
exons 14-17 from the Z-Disc region
exons 28-33 from the Z-Disc region
exons 108-114 from the N2A-PEVK region
exons 358-363 from the Mline region
ankyrin repeat domain 1
full length protein
full length protein minus the actin-binding C-terminal domain
F-box protein 32 (MAFbx)
full length protein
tripartite motif-containing 63 (MuRF1)
full length protein
For each bait domain, we first assessed its toxicity and auto-activation capacity by a small-scale Y2H screen and then performed a large-scale Y2H assay by screening a high-complexity cDNA prey library obtained by random priming of poly(A)+ RNA from adult and fetal human skeletal muscles that we constructed for this purpose. The bait interaction was tested against an average of 103 million prey clones to insure a ten-fold coverage of the prey library. Positive prey clones were sequenced and compared to the NCBI human RefSeq database for prey identification. Contig assembly of positive clones was performed to isolate the minimum interacting domain(s) on each prey sequence [Selected Interacting Domains (SID)]. We used clone coverage and local topology information to compute a confidence score [Predicted Biological Score (PBS)] and classify each PPI into five categories: PBS-A, -B or -C for the most reliable interactions, PBS-D for putative interactions involving a single bait clone and PBS-E for interactions involving highly connected proteins.
We then examined the interaction networks resulting from the first screenings according to the PBS categories and literature data and conducted selection of secondary and tertiary baits. First, we isolated 54 prey proteins of interest to design 57 new bait domains for a second round of screenings and then, we used the resulting Y2H network to select 10 additional proteins corresponding to 11 baits for a third and last round of screening. Two of the chosen baits showed autoactivation capacities (CMYA5 Q3501-K4069 and RCOR3 M1-L296 fragments) and were therefore discarded. Overall, we successfully carried out 87 large-scale Y2H screenings using 76 different bait proteins. The comprehensive set of baits is listed in Additional file 2: Table S2.
General properties of the Y2H interaction map
General features of the LGMD-centered interaction map
Total number of bait domains/proteins
87 / 76
Average number of tested diploids
103 106 per screen
20 106 - 203 106
Average number of processed positive clones
155 per bait domain
Total number of proteins/PPIs
Average number of SIDs per bait domain
Average number of partners per prey protein
Total number/average size of identified SIDs
Examination of interaction domains
Most frequent domains in the LGMD-centered dataset and representativeness in the human proteome
% in the total of SID
% in the human proteome (according to Müller et al., Genome Res 2002)
Immunoglobulin (IPR003599; IPR003598; IPR007110)
4.2% (n=1214; r=2)
Zinc finger, C2H2-type (IPR007087)
17.6% (n=5092; r=1)
Fibronectin type III (IPR003961)
2.9% (n=842; r=5)
Ankyrin repeat (IPR002110)
0.9% (n=278; r=14)
Nebulin 35 residue motif (IPR000900)
Quality assessment of the Y2H network
For all large-scale studies aimed at experimentally identifying molecular interactions, the technical false positive rate is of special concern. We addressed this problem with the assignment of a PBS category. Furthermore, literature mining, cross-validation assays, and functional correlations were used to further estimate the overall solidity of the network.
Comparison with known interactions data
We first compared our LGMD-centered Y2H dataset with literature-curated interactions between the corresponding mammalian proteins that are referenced in the molecular interaction database iRefWeb ( http://wodaklab.org/iRefWeb/). This web interface reports data on PPIs consolidated from major public databases such as IntAct, BIND or HPRD. Among our PPIs, 72 were already reported, representing 4.8% of our map (Additional file 1: Table S1), a figure slightly above others found in large-scale Y2H studies aimed at exploring the human interactome (3.4%  and 3.8% ). We observed a strong enrichment for literature-based interactions within the PBS-A category for which 18% of PPI (43) correspond to previously known interactions (p-value = 1.54e-18), thus confirming the correlation between the PBS and the biological significance of the interaction.
In total and among the 54 PPIs investigated using these three different techniques, 40 interactions were considered as relevant by at least one technique (74%) with 20 of them from the PBS-A, -B, or -C categories and the others from the PBS-D category. The 26% remaining interactions identified by our Y2H screens that could not be experimentally cross-validated in our hands consist in 3, 3, 2 and 5 interactions classified in the PBS-A to –D categories, respectively.
Examination of GO annotations
The NMD proteins in the LGMD-centered network
We examined our LGMD-centered interactome map to pinpoint interactions involving proteins identified as the genetic cause of one or more hereditary NMDs by searching the OMIM database. We identified 199 proteins of the network that are associated with human monogenic diseases including 77 proteins whose defects have been described as the genetic cause of one or more hereditary NMD (Additional file 4: Table S4) . Among this last group, 43 proteins correspond to myopathies including the recently described DNAJB6 , 20 to cardiomyopathies, 17 to neuropathies, 8 to metabolic muscle diseases, 3 to excitation abnormalities and 8 to unclassified NMD. From these 199 proteins, it was possible to construct a protein interaction network including 88 proteins in 113 interactions (Additional file 5: Figure S1). No specific repartition of NMD and non-NMD proteins was noticed.
Considering that the LGMD-centered interactome is strongly enriched in known NMD-causing proteins, we expect that some yet-uncharacterized NMD-causing proteins are part of our interaction network. We examined in more detail the candidate genes for the orphan LGMDs. In addition to the causative genes for 4 dominant and 17 recessive LGMD forms that are known so far, genetic linkage analyses have been used to map loci for three new LGMDs. These orphan LGMD loci contain between 13 and 45 genes, with a total of 2 genes encoding proteins of the LGMD-centered interactome. Consequently, FLNC and protein transport protein Sec31A (SEC31A) were isolated as unique candidates for LGMD1F and LGMD1G, respectively. No gene from the LGMD-centered interactome was found within the LGMD1E locus.
Properties of the LGMD-centered network and biological functions of the proteins
To investigate which biological functions are associated with the LGMD-centered network, we performed a GO term enrichment analysis. In this analysis, we compared the LGMD dataset to a high-confidence (HC) network encompassing the most meaningful PPIs as defined with combined results from the experimental and computational analyses described above and a literature-based network.
The HC dataset was built by extracting from the LGMD-centered map the 491 PPIs classified in the PBS-A, -B, or -C categories plus PPIs from the PBS-D and -E categories for which we obtained additional evidence of the interaction. First, we added the 22 of our Y2H interactions from the PBS-D or -E categories that have been described in other studies (Additional file 1: Table S1). Second, PPIs from the PBS-D or -E categories were added to the HC dataset if experimental results from co-immunoprecipitation, immunofluorescence or PLA as described above confirmed the Y2H interaction. In total, 20 experimentally cross-validated interactions from the PBS-D category were included in the HC dataset. Third, were also included in the HC network, 107 pairs of interacting proteins from the PBS-D or -E categories that either share a BP GO annotation or share a CC and a MF GO annotation. Finally, since our dataset unraveled a high average level of relationship between NMD proteins, 65 PPIs in the PBS-D and -E categories and involving a NMD prey protein were added to the HC network. The resulting HC dataset consists of 497 proteins and 705 PPIs (Additional file 1: Table S1) and includes 174 and 40 PPIs from the PBS-D and -E categories, respectively.
The literature-based dataset was constructed by combining all direct protein-protein interactions reported in the literature for proteins of the HC dataset or LGMD-causing proteins that were not included in our initial set of baits. It is interesting to note that only two interactions for the five LGMD glycosyltransferases have been identified in previous reports [between Fukutin and protein O-linked-mannose beta-1,2-N-acetylglucosaminyltransferase 1 (POMGnT1) and between protein O-mannosyl-transferase 1 (POMT1) and 2 (POMT2)]. Since the glycosyltransferases are all described or predicted to be single or multipass membrane proteins, the rarity of PPI and their absence in our LGMD-centered network are possibly explained, since identification of interactions in the vicinity of membranes has proven to be difficult with the Y2H method. We searched for previously published experimental binary and direct interactions using the iRefWeb interface and identified 2675 direct binary interactions (Additional file 6: Table S5). This literature-based network consists of 2239 proteins and 3304 PPIs.
Comparison of the three networks
Number of proteins
Number of PPIs
Average number of partners per bait protein
Additional file contents
Additional file 1: Table S1
gives a comprehensive list of all the identified PPI
Additional file 2: Table S2
gives comprehensive information about the baits and Y2H screenings
Additional file 3: Table S3
gives the list of all the SIDs
Additional file 4: Table S4
presents the Human diseases associated with the proteins of the network
Additional file 5: Figure S1
presents a sub-network composed of the disease-related proteins only
Additional file 6: Table S5
gives the list of the PPIs for the literature-based network
Our Y2H experimental strategy was a large-scale protein domain-based approach using a methodology that has been previously successfully implemented . As compared to approaches where baits are screened against full-length prey proteins, a domain-based approach offers several advantages including the possibility of narrowing down interaction domains and of reducing the false negative discovery rate by allowing a more efficient folding of domains . A similar approach used in the field of ataxia identified interactions that were missed in a previous full-length protein-based approach . Interestingly, this domain-based approach enabled us to analyze possible bias in interacting domains. We found that the three most frequent domains present in the SIDs (Immunoglobulin-like, Zinc Finger and Fibronectin domains) are also amongst the most frequent ones in the human proteome, emphasizing how important biomolecular interactions are for cellular processes. Two other domains (Ankyrin repeat and Nebulin motif) were frequently reported in the SIDs whereas they are rare in the human proteome. These domains were established as useful for meeting the demands of skeletal muscle physiology and constraints [34, 35]. Auto-binding capacities of ankyrin repeats lead to mechanical-resistant dimers and nebulin-like domains play a role in the regulation of muscle contraction, especially through their interaction with actin and the thin filament. On the other hand, two domains frequently found in the human proteome (EGF-like and P-loop domains) are underrepresented in our set of interacting domains. This observation could be explained either by the function of the domain; the P-loop motif is known to be involved in hydrolysis of ATP and GTP but not in protein-protein interaction; or by a technical bias as in the case of the EGF-like domain which principally serves as an interacting domain within extracellular protein modules. The underrepresentation of this last domain is probably related to the fact that the LGMD proteins are intracellular proteins with the exception of sarcoglycans but for which the intracellular domain was selected for the Y2H screening.
The utility of any network obviously depends on the quality of the data. Our approach resides in the exploitation of random-primed cDNA libraries constructed from human skeletal muscle poly(A) RNA. It minimizes the risk of detecting interactions between proteins that would not be co-expressed in the muscle tissue and therefore should reduce the number of biological false positives. In addition, assignment of a PBS score using a statistical method allowed us to classify each PPI identified in our Y2H assays into five predictive categories, ranging from the most reliable interactions to possible technical false positives. The PBS score was previously demonstrated to successfully predict the reliability of a PPI on a subset of Y2H interactions where the authors experimentally confirmed 79% of all PPIs from the PBS-A, -B or -C categories by pull-down or co-immunoprecipitation assays . The PBS takes into account both local parameters, such as the number of identical or independent fragments found for each partner and global information derived from the entire network such as reciprocal interactions, highly connected domains etc… Our screens identified a high percentage of the PBS-D interactions (53.6%) which reflects the high complexity of the constructed muscle library and the fact that this library is screened to saturation. As a consequence, the rate of false negatives should be extremely low, as even rare transcripts in the library, or weak or transient interactions, could be detected. In line with the notion that PBS-D interactions might be more difficult to detect, a higher proportion of interactions from the ABC categories than from the D category can be confirmed by other methods, such as co-immunoprecipitation or pull-down . However, they could also represent false positive interactions and should be considered with caution. Finally, over 210 publications (full list available http://www.hybrigenics-services.com/publications/index/list). Of note, 27% of the publications reporting the functional validation of protein interactions identified using the same Y2H methodology correspond to the validation of a D interaction. This indicates that a significant proportion of such low confidence Y2H interactions are functionally relevant.
In the present study, we further confirmed the relevance of the PBS score with computational and experimental evidence. First, nearly a fifth of the PBS-A PPIs were found to have been previously reported whereas this figure falls to less than 5% when considering the whole interactome. Second, co-immunoprecipitation assays confirmed as positive 11 out of 16 technically conclusive tests (69%) and PLA validated 34 of the 44 tested interactions (77%). Although there is often limited overlap between studies [2, 36], probably in part because of the spatial and temporal aspects of the proteome interaction, we compared our results obtained for DYSF with a recent study that analyzed the composition of DYSF complexes in cultured myoblasts, myotubes and skeletal muscle tissues by mass spectrometry and bioinformatics methods . Interestingly, 19 (14%) of the 136 DYSF interactions identified by our large-scale Y2H screening were also found in the de Moree’s study, a ratio that is a little higher than what is usually observed . In contrast, only 3 interactions (TTN, ACTN2 and DES) were found in a study based on Fisher’ method  and none in another in silico study .
Remarkably, results of our Y2H screens led to a single connected network where the different LGMD proteins are highly connected. The strong inter-connectivity between LGMD proteins is illustrated by a high number of direct interactions. This was quite surprising since, even if the different LGMD forms share seemingly close clinical phenotypes, the LGMD-causing proteins have been described to have quite diverse locations and biological functions. In addition to identifying a remarkable number of direct interactions between LGMD proteins, mining the data revealed that LGMD proteins belong to a highly connected network of interacting proteins with, in particular, the sarcomeric proteins ACTN2, MYBPC1, MYOM1 and MYOM2 identified as hub proteins sharing the highest number of links with the LGMD proteins. Interestingly, these proteins are structural proteins located at key places on the sarcomere, the Z-disc, the N2A-line and the M-band. GO analyses further support the crucial place of the cytoskeleton in the connections between LGMD proteins. Taken together, these data suggest that common molecular mechanisms underlie the pathogenesis of these diseases and highlight the sarcomere as an important platform for skeletal muscle homeostasis and myofiber survival.
We expect that our interaction map can serve as a new tool to accelerate discovery of the causative mutated genes for orphan LGMDs or for other orphan NMDs already or not yet described as well as to identify modifier genes. Examination of the chromosomal location of all the genes coding for the proteins that are part of our original interactome map revealed putative candidates for orphan NMDs of which FLN and, SEC31A appeared to be of particular interest for LGMD1F and LGMD1G, respectively. Defects in the FLNC gene coding for an actin-binding protein, are already known to cause myofibrillar myopathies  but the gene was supposedly excluded as being involved in the pathogenesis of LGMD1F . Nevertheless, non-coding sequences were not fully investigated for pathogenic mutations and, as mentioned by the authors, the gene remains a possible candidate. The SEC31A gene coding for a component of a protein complex responsible for vesicle budding from the endoplasmic reticulum , has not been associated with any disease. It is an interesting candidate since LGMD1G is associated with progressive limitation of fingers and toes flexion  and SEC31A has been linked to collagen secretion . In addition to the LGMD proteins, there is a high proportion of proteins involved in congenital and metabolic myopathies in our network, therefore it is very likely that causative genes for genetically uncharacterized forms of these two groups of diseases lie within our interactome.
An interesting outcome of our study is to provide new PPIs that further support and extend previous findings and pinpoint new pathways of interest that could be affected in LGMD. For example, two novel interactions of calpain 3 (CAPN3) with, ring finger protein 167 (RNF167), an E3 ubiquitin-protein ligase, and the proteasome maturation factor (POMP) are of particular interest considering a previous report indicating that CAPN3 acts upstream of the ubiquitin-proteasome system . For DYSF, a number of new interacting partners can be categorized in three different cellular processes: endocytosis, microtubule-related transport and regulation of gene expression. The first two pathways fit well with previous knowledge about DYSF functions but, interestingly, the third pathway indicates a new possible role for this protein. Another interesting finding in view of the fact that the pathogenesis of SGCG deficiency does not seem strictly related to membrane stability [46, 47], is the possible relationship of this protein with energy controlling pathways since interaction with proteins involved in glycolysis or glycogenolysis (enolases 1 and 3 and PYGM) or in the TCA cycle (SUCLG2, ACO1) was identified. Finally, several partners detected for TCAP suggest that it may play a role in the maintenance of genome integrity, in accordance to the recent report showing a relationship between TCAP and p53 turn-over . These elements provide new avenues to explore for a better understanding of the pathophysiology of the various forms of LGMD.
In conclusion, this study presents new interacting partners for LGMD proteins and other proteins known to be involved in NMD. In this sense, it has the potential to reveal new candidate genes for NMD but also modifiers of the phenotype. This broad dataset should also help to take a step further towards the understanding of skeletal muscle tissue. In particular, it will help to improve our knowledge about the cellular functions and roles of NMD proteins in the muscle cell and about their participation in the diseases they trigger thereby speeding up the identification of putative drug targets.
This work was supported by the Association Française contre les Myopathies and the Jain Foundation. We thank Dr. Anne Friedrich for technical assistance on gene coordinates and cytogenetic bands studies and Roseline Yao and Julie Dumonceaux from the Inserm U790 for the R9 cells. We are grateful to the sequencing and histological departments of Genethon and to Jean Pascal Lepetit-Stoffaes, Drs Alain Meil, Vincent Collura and Etienne Formstecher from Hybrigenics for excellent assistance.
- Stelzl U, Worm U, Lalowski M, Haenig C, Brembeck FH, Goehler H, Stroedicke M, Zenkner M, Schoenherr A, Koeppen S: A human protein-protein interaction network: a resource for annotating the proteome. Cell 2005, 122:957–968.PubMedView Article
- Rual JF, Venkatesan K, Hao T, Hirozane-Kishikawa T, Dricot A, Li N, Berriz GF, Gibbons FD, Dreze M, Ayivi-Guedehoussou N: Towards a proteome-scale map of the human protein-protein interaction network. Nature 2005, 437:1173–1178.PubMedView Article
- Ewing RM, Chu P, Elisma F, Li H, Taylor P, Climie S, McBroom-Cerajewski L, Robinson MD, O'Connor L, Li M: Large-scale mapping of human protein-protein interactions by mass spectrometry. Mol Syst Biol 2007, 3:89.PubMedView Article
- Bouwmeester T, Bauch A, Ruffner H, Angrand PO, Bergamini G, Croughton K, Cruciat C, Eberhard D, Gagneur J, Ghidelli S: A physical and functional map of the human TNF-alpha/NF-kappa B signal transduction pathway. Nat Cell Biol 2004, 6:97–105.PubMedView Article
- Goehler H, Lalowski M, Stelzl U, Waelter S, Stroedicke M, Worm U, Droege A, Lindenberg KS, Knoblich M, Haenig C: A protein interaction network links GIT1, an enhancer of huntingtin aggregation, to Huntington's disease. Mol Cell 2004, 15:853–865.PubMedView Article
- Lim J, Hao T, Shaw C, Patel AJ, Szabo G, Rual JF, Fisk CJ, Li N, Smolyar A, Hill DE: A protein-protein interaction network for human inherited ataxias and disorders of Purkinje cell degeneration. Cell 2006, 125:801–814.PubMedView Article
- Nigro V, Aurino S, Piluso G: Limb girdle muscular dystrophies: update on genetic diagnosis and therapeutic approaches. Curr Opin Neurol 2011, 24:429–436.PubMedView Article
- Aurino S, Piluso G, Saccone V, Cacciottolo M, D'Amico F, Dionisi M, Totaro A, Belsito A, Di Vicino U, Nigro V: Candidate-gene testing for orphan limb-girdle muscular dystrophies. Acta Myol 2008, 27:90–97.PubMed
- Guglieri M, Straub V, Bushby K, Lochmuller H: Limb-girdle muscular dystrophies. Curr Opin Neurol 2008, 21:576–584.PubMedView Article
- Krogh A, Larsson B, von Heijne G, Sonnhammer EL: Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes. J Mol Biol 2001, 305:567–580.PubMedView Article
- Nielsen H, Engelbrecht J, Brunak S, von Heijne G: A neural network method for identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites. Int J Neural Syst 1997, 8:581–599.PubMedView Article
- Finn RD, Tate J, Mistry J, Coggill PC, Sammut SJ, Hotz HR, Ceric G, Forslund K, Eddy SR, Sonnhammer EL, Bateman A: The Pfam protein families database. Nucleic Acids Res 2008, 36:D281-D288.PubMedView Article
- Vojtek AB, Hollenberg SM: Ras-Raf interaction: two-hybrid analysis. Methods Enzymol 1995, 255:331–342.PubMedView Article
- Formstecher E, Aresta S, Collura V, Hamburger A, Meil A, Trehin A, Reverdy C, Betin V, Maire S, Brun C: Protein interaction mapping: a Drosophila case study. Genome Res 2005, 15:376–384.PubMedView Article
- Fromont-Racine M, Rain JC, Legrain P: Toward a functional analysis of the yeast genome through exhaustive two-hybrid screens. Nat Genet 1997, 16:277–282.PubMedView Article
- Maiweilidan Y, Klauza I, Kordeli E: Novel interactions of ankyrins-G at the costameres: the muscle-specific Obscurin/Titin-Binding-related Domain (OTBD) binds plectin and filamin C. Exp Cell Res 2011, 317:724–736.PubMedView Article
- Alhamidi M, Kjeldsen Buvang E, Fagerheim T, Brox V, Lindal S, Van Ghelue M, Nilssen O: Fukutin-related protein resides in the Golgi cisternae of skeletal muscle fibres and forms disulfide-linked homodimers via an N-terminal interaction. PLoS One 2011, 6:e22968.PubMedView Article
- Burgo A, Proux-Gillardeaux V, Sotirakis E, Bun P, Casano A, Verraes A, Liem RK, Formstecher E, Coppey-Moisan M, Galli T: A molecular network for the transport of the TI-VAMP/VAMP7 vesicles from cell center to periphery. Dev Cell 2012, 23:166–180.PubMedView Article
- Bolte S, Cordelieres FP: A guided tour into subcellular colocalization analysis in light microscopy. J Microsc 2006, 224:213–232.PubMedView Article
- Quevillon E, Silventoinen V, Pillai S, Harte N, Mulder N, Apweiler R, Lopez R: InterProScan: protein domains identifier. Nucleic Acids Res 2005, 33:W116-W120.PubMedView Article
- Mulder NJ, Apweiler R, Attwood TK, Bairoch A, Bateman A, Binns D, Bork P, Buillard V, Cerutti L, Copley R: New developments in the InterPro database. Nucleic Acids Res 2007, 35:D224-D228.PubMedView Article
- Turner B, Razick S, Turinsky AL, Vlasblom J, Crowdy EK, Cho E, Morrison K, Donaldson IM, Wodak SJ: iRefWeb: interactive analysis of consolidated protein interaction data and their supporting evidence. Database (Oxford) 2010, 2010:baq023.View Article
- Sherman BT, Huang da W, Tan Q, Guo Y, Bour S, Liu D, Stephens R, Baseler MW, Lane HC, Lempicki RA: DAVID Knowledgebase: a gene-centered database integrating heterogeneous gene annotation resources to facilitate high-throughput gene functional analysis. BMC Bioinformatics 2007, 8:426.PubMedView Article
- da Huang W, Sherman BT, Lempicki RA: Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 2009, 4:44–57.PubMedView Article
- Muller A, MacCallum RM, Sternberg MJ: Structural characterization of the human proteome. Genome Res 2002, 12:1625–1641.PubMedView Article
- Hilfiker S, Greengard P, Augustine GJ: Coupling calcium to SNARE-mediated synaptic vesicle fusion. Nat Neurosci 1999, 2:104–106.PubMedView Article
- Bansal D, Miyake K, Vogel SS, Groh S, Chen CC, Williamson R, McNeil PL, Campbell KP: Defective membrane repair in dysferlin-deficient muscular dystrophy. Nature 2003, 423:168–172.PubMedView Article
- Vignal E, Blangy A, Martin M, Gauthier-Rouviere C, Fort P: Kinectin is a key effector of RhoG microtubule-dependent cellular activity. Mol Cell Biol 2001, 21:8022–8034.PubMedView Article
- Scita G, Nordstrom J, Carbone R, Tenca P, Giardina G, Gutkind S, Bjarnegard M, Betsholtz C, Di Fiore PP: EPS8 and E3B1 transduce signals from Ras to Rac. Nature 1999, 401:290–293.PubMedView Article
- Sarparanta J, Jonson PH, Golzio C, Sandell S, Luque H, Screen M, McDonald K, Stajich JM, Mahjneh I, Vihola A: Mutations affecting the cytoplasmic functions of the co-chaperone DNAJB6 cause limb-girdle muscular dystrophy. Nat Genet 2012, 44:450–455-S451–452.PubMedView Article
- Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T: Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 2003, 13:2498–2504.PubMedView Article
- Legrain P, Selig L: Genome-wide protein interaction maps using two-hybrid systems. FEBS Lett 2000, 480:32–36.PubMedView Article
- Kahle JJ, Gulbahce N, Shaw CA, Lim J, Hill DE, Barabasi AL, Zoghbi HY: Comparison of an expanded ataxia interactome with patient medical records reveals a relationship between macular degeneration and ataxia. Hum Mol Genet 2011, 20:510–527.PubMedView Article
- Tee JM, Peppelenbosch MP: Anchoring skeletal muscle development and disease: the role of ankyrin repeat domain containing proteins in muscle physiology. Crit Rev Biochem Mol Biol 2010, 45:318–330.PubMedView Article
- Ogut O, Hossain MM, Jin JP: Interactions between nebulin-like motifs and thin filament regulatory proteins. J Biol Chem 2003, 278:3089–3097.PubMedView Article
- Bader GD, Hogue CW: Analyzing yeast protein-protein interaction data obtained from different sources. Nat Biotechnol 2002, 20:991–997.PubMedView Article
- de Morree A, Hensbergen PJ, van Haagen HH, Dragan I, Deelder AM, Hoen PA t, Frants RR, van der Maarel SM: Proteomic analysis of the dysferlin protein complex unveils its importance for sarcolemmal maintenance and integrity. PLoS One 2010, 5:e13854.PubMedView Article
- van Haagen HH, Hoen PA t, de Morree A, van Roon-Mom WM, Peters DJ, Roos M, Mons B, van Ommen GJ, Schuemie MJ: In silico discovery and experimental validation of new protein-protein interactions. Proteomics 2011, 11:843–853.PubMedView Article
- Cacciottolo M, Belcastro V, Laval S, Bushby K, di Bernardo D, Nigro V: Reverse engineering gene network identifies new dysferlin-interacting proteins. J Biol Chem 2011, 286:5404–5413.PubMedView Article
- Comi GP, Fortunato F, Lucchiari S, Bordoni A, Prelle A, Jann S, Keller A, Ciscato P, Galbiati S, Chiveri L: Beta-enolase deficiency, a new metabolic myopathy of distal glycolysis. Ann Neurol 2001, 50:202–207.PubMedView Article
- Palenzuela L, Andreu AL, Gamez J, Vila MR, Kunimatsu T, Meseguer A, Cervera C, Fernandez Cadenas I, van der Ven PF, Nygaard TG: A novel autosomal dominant limb-girdle muscular dystrophy (LGMD 1F) maps to 7q32.1–32.2. Neurology 2003, 61:404–406.PubMedView Article
- Tang BL, Zhang T, Low DY, Wong ET, Horstmann H, Hong W: Mammalian homologues of yeast sec31p. An ubiquitously expressed form is localized to endoplasmic reticulum (ER) exit sites and is essential for ER-Golgi transport. J Biol Chem 2000, 275:13597–13604.PubMedView Article
- Starling A, Kok F, Passos-Bueno MR, Vainzof M, Zatz M: A new form of autosomal dominant limb-girdle muscular dystrophy (LGMD1G) with progressive fingers and toes flexion limitation maps to chromosome 4p21. Eur J Hum Genet 2004, 12:1033–1040.PubMedView Article
- Townley AK, Feng Y, Schmidt K, Carter DA, Porter R, Verkade P, Stephens DJ: Efficient coupling of Sec23-Sec24 to Sec13-Sec31 drives COPII-dependent collagen secretion and is essential for normal craniofacial development. J Cell Sci 2008, 121:3025–3034.PubMedView Article
- Kramerova I, Kudryashova E, Venkatraman G, Spencer MJ: Calpain 3 participates in sarcomere remodeling by acting upstream of the ubiquitin-proteasome pathway. Hum Mol Genet 2007, 16:1006.PubMedView Article
- Hack AA, Cordier L, Shoturma DI, Lam MY, Sweeney HL, McNally EM: Muscle degeneration without mechanical injury in sarcoglycan deficiency. Proc Natl Acad Sci U S A 1999, 96:10723–10728.PubMedView Article
- Barton ER: Restoration of gamma-sarcoglycan localization and mechanical signal transduction are independent in murine skeletal muscle. J Biol Chem 2010, 285:17263–17270.PubMedView Article
- Knoll R, Linke WA, Zou P, Miocic S, Kostin S, Buyandelger B, Ku CH, Neef S, Bug M, Schafer K: Telethonin deficiency is associated with maladaptation to biomechanical stress in the mammalian heart. Circ Res 2011, 109:758–769.PubMedView Article
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.