dystrophy
Morrée, A. de
Citation
Morrée, A. de. (2011, January 12). Functional protein networks unifying limb girdle muscular dystrophy. Retrieved from https://hdl.handle.net/1887/16329
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/16329
Note: To cite this publication please use the final published version (if
applicable).
Functional protein networks unifying Limb Girdle Muscular
Dystrophy
Antoine de Morrée
Reprinted with kind permission of the Rijksmuseum.
Layout by E Savelkoul, CV de Morrée and A de Morrée
Printing by Off Page, www.offpage.nl
ISBN 978-94-90371-60-9
Copyright 2010 by Antoine de Morrée. All rights reserved. Copyright of the individual chapters rests with the authors, with the following exceptions:
Chapers 2 and 4: Public Library of Science Chapter 3: Oxford University Press
No part of this book may be reproduced, stored in a retrieval system, or
transmitted in any form or by any means, without prior permission of the author.
Limb Girdle Muscular Dystrophy
Proefschrift ter verkrijging van
de graad van Doctor aan de Universiteit Leiden,
op gezag van Rector Magnificus prof.mr. P.F. van der Heijden, volgens besluit van het College voor Promoties
te verdedigen op woensdag 12 januari 2011 klokke 16.15 uur
door
Antoine de Morrée
geboren te Utrecht in 1982
Promotores: Prof. Dr. Ir. Silvère M van der Maarel1 Prof. Dr. Rune R Frants1
Overige leden: Prof. Dr. Isabel Illa2 Prof. Dr. Peter ten Dijke3 Prof. Dr. Rene Toes4
1 Afdeling Humane Genetica, Leids Universitair Medisch Centrum, Leiden, Nederland
2 Laboratori de Neurologia Experimental, Universitat Autònoma de Barcelona, Spanje.
3 Afdeling Moleculaire Cel Biologie, Leids Universitair Medisch Centrum, Leiden, Nederland
4 Afdeling Reumatologie, Leids Universitair Medisch Centrum, Leiden, Nederland
The studies described in this thesis have been performed at the Leiden University Medical Center, department of human genetics. This work was financially supported by the Dutch Prinses Beatrix Fonds (MAR05-0112) and the Jain Foundation.
Publication of this thesis was financially supported by the Dutch J.E. Jurriaanse Stichting and the Jain foundation.
James D. Watson
Abbreviations 11
Foreword 15
1. Protein networks unifying Limb Girdle Muscular Dystrophy 19
2. Proteomic analysis of the Dysferlin protein complex unveils its importance for sarcolemmal maintenance and
integrity 33
-Chapter 2 is currently in press, PLoS ONE-
3. Calpain 3 is a modulator of the Dysferlin protein complex in skeletal
muscle 63
-Chapter 3 has been published, HMG 2008-
4. Calpain 3 is a rapid-action, unidirectional proteolytic switch central to muscle remodeling 85
-Chapter 4 has been published, PLoS ONE 2010-
5. Self-regulated alternative splicing at
the AHNAK locus 105
-Chapter 5 has been submitted for publication-
6. AHNAK redistributes upon Integrin inhibition in cultured myoblasts 131
7. Crosstalk between Dysferlin and Integrin ß3 regulates cell contacts in
human monocytes 147
8. Maintenance and remodeling are central to skeletal muscle physiology and disturbed in Limb Girdle Muscular
Dystrophy 171
Supplement 189
References 213
Summary 237
Nederlandse samenvatting zonder
vakjargon 241
Curriculum Vitae 249
Publication list 251
Acknowledgements 253
ACTN Actinin
ADP Adenosine diphosphate ANXA1 Annexin A1
ANXA2 Annexin A2 AP2 Adaptin 2
ATP Adenosine triphosphate ATP5b ATP Synthase ß Subunit AuC Area under the ROC Curve BCA Bicinchoninic acid protein
assay
BG Beta-Galatosidae-GFP fusion protein
CAPN3 CAPN3
CHAPS 3-[(3-cholamidopropyl)- dimethylammonio]- 1-propanesulfonate CLTA Clathrin alpha
CMT Charcot-Marie-Tooth disease, demyelinating
DACM Distal Anterior Compartment Myopathy
DAPI 4’,6-diamidino- 2-phenylindole
DAVID Database for Annotation, Visualization and Integrated Discovery
DFNB9 Neurosensory nonsundromic recessive deafness 9
DGC Dystophin Glycoprotein Complex
DHPR Dihydropyridene Receptor DMD Dystrophin
DYSF Dysferlin ECL Enzymatic
Chemiluminescence ELISA Enzyme-Linked
ImmunoSorbent Assay ER Endopplasmic Reticulum FCS Fetal Calf Serum
Fer1 Ferlin Fer1L Ferlin-like FLNC Filamin C
FSHD Facioscapulohumeral muscular dystrophy GFP Green Fluorescent Protein GO Gene Ontology
GST Glutathione S-transferase HCAb Heavy Chain Antibody IGF2R Mannose-6 phosphate
receptor
IP Immunoprecipitation IPTG Isopropyl ß-D-
1-thiogalactopyranoside IS Insertion Sequence ITGA Integrin Alpha ITGB Integrin Beta
and Genomes
L-AHNAK Large AHNAK isoform LGMD Limb Girdle Muscular
Dystrophy
L-PRX Large Periaxin isoform LVGCC L-Type Voltage-Gated
Calcium Channel MD Muscular Dystrophy
MG53 Matsagumin 53, also known as Trim72
MH Malignant hyperthermia MM Miyoshi Myopathy MS Mass Spectrometry MTJ Myotendinous Junctions NCBI National Center for
Biotechnology Information NES Nuclear Export Signal NLS Nuclear Localization Signal NP40 nonyl
phenoxypolyethoxylethanol OPMD Oculopharyngeal muscular
dystrophy
PABPN1 Poly(A) Binding Protein N1 PAGE Poly-Acrylamide Gel
Electrophoresis
PBS Phosphate Buffered Saline PCR Polymerase Chain Reaction PIAS Protein Inhibitor of Activated
STATs
PKA(B)(C) Protein Kinase A(B)(C)
PRX Periaxin
PVDF Polyvinylidene difluoride ROC Receiver operating
characteristic RT Room Temperature S-AHNAK Small AHNAK isoform SC35 Spliceosome Component 35 SDS Sodium-Dodecyl-Sulphate S-PRX Small Periaxin isoform SUMO Small Unbiquitin-like Modifier TEA TriEthanolAmine
TLN Talin
TPM TropoMyosin TUBA Alpha-Tubulin
UMLS Unified Medical Language System
VHH Variable domain of a Heavy Chain Antibody
VINC Vinculin
One of the most versatile tissues of the human body is the muscle. There are three different types of muscle: cardiac muscle, smooth muscle and skeletal muscle.
Cardiac muscle cells make up the heart, which drives blood circulation. The smooth muscle cells are part of the blood vessel walls and aid blood flow by maintaining blood pressure. Finally, the skeletal muscle connects skeletal structures to allow movement of body parts in relation to one another.
Muscle tissue is highly adaptive, which is apparent in growth, exercise and degeneration and regeneration. During normal contractions it needs effective and efficient mechanisms to repair the continuous damage to the cytoskeletal anchors, the contractile apparatus, and the cellular membrane. When repair is not effective, the damaged fibers die and are rapidly cleared by immune cells. A new fiber will replace the former.
Each of these maintenance and repair processes can be affected by genetic mutations, resulting in a muscular dystrophy phenotype. Much can be learned about normal muscle function from such mutations. Such knowledge can be used in the development of treatments for these diseases. Limb Girdle Muscular Dystrophy (LGMD) is a rare heterogeneous disorder that can be caused by mutations in at least 21 different genes. These genes encode proteins with highly differing functions and yet mutations in all of them give rise to a similar clinical presentation.
In this thesis I will explore a potential molecular mechanism that unifies the different genetic defects that individually can cause a limb girdle muscular dystrophy. I will start with a thorough survey of recent literature, venturing the hypothesis that all forms of LGMD suffer from impaired muscle maintenance capacity (Chapter 1). I then describe several lines of experimental research that investigate this hypothesis; work that started with the membrane repair (and maintenance) protein Dysferlin.
Dysferlin is strongly implicated in muscle membrane repair. However, it is
functionality of Dysferlin. To answer these questions a robust and reproducible method for purification of Dysferlin containing protein complexes was set up, yielding insight into which processes Dysferlin is involved (Chapter 2). The interaction with two binding partners, Calpain 3 (CAPN3) and AHNAK was further investigated, showing that CAPN3 modulates the direct interaction between Dysferlin and AHNAK through enzymatic cleavage of AHNAK (Chapter 3). CAPN3 is in addition to Dysferlin another protein to cause a LGMD phenotype. It is a proteolytic enzyme. Using the identified CAPN3 cleavage sites in AHNAK a general CAPN3 cleavage motif was uncovered, allowing for the identification of novel substrates and a functional model for this protein (Chapter 4). CAPN3 is central to muscle maintenance through cleavage of structural proteins. The other Dysferlin interaction partner, AHNAK, localizes to vesicles and binds structural proteins, and could link such two processes. As little is known about the AHNAK gene and protein family, Chapter 5 describes its evolution and transcriptional regulation in skeletal muscle. A function for AHNAK is explored by chemically inhibiting Integrins (Chapter 6). Muscle tissue strongly communicates with immune cells, which also express many skeletal muscle proteins, including Dysferlin and AHNAK. In Chapter 7 a monocyte-macrophage model system is used to indicate a role for Dysferlin and AHNAK in Integrin signaling and monocyte differentiation and phagocytosis, which are deregulated in LGMD. Finally, in Chapter 8 all the experimental conclusions are summarized, and held against the light of recent scientific literature, resulting in the conclusion that Dysferlin and CAPN3 are involved in muscle maintenance and remodeling. This Thesis ends with the description of possible experimental research lines that will help to continue this research.
Protein networks unifying
Limb Girdle Muscular Dystrophy
Introduction
Limb Girdle Muscular Dystrophy (LGMD) is a rare progressive disorder that mainly affects the skeletal muscles of the pelvic and shoulder girdle. LGMD is characterized by a period of good health until disease onset, which depending on the gene involved can start ranging from early teens, to midlife. After disease manifestation, LGMD progresses and within several years the muscle tissue shows severe wasting and increasing weakness, and patients often end wheelchair bound [32]. While in most cases the proximal muscles are first affected, with time other more distal muscles follow. LGMD is not deadly per se, but in many cases there is an increased risk for cardiac and respiratory failure [32].
Because of these risk factors it is important that a correct diagnosis is reached rapidly [32]. LGMD is genetically heterogeneous and can be transmitted in an autosomal dominant (LGMD1) and autosomal recessive (LGMD2) manner. For many LGMD loci the disease gene has been identified. Strikingly, there are at least 21 different genes that when mutated give rise to an LGMD phenotype (Table 1) [160]. Being this heteregoneous, diagnosis of LGMD is not straightforward. It has been estimated however, that it should be possible to reach a precise diagnosis in roughly 75% of the LGMD patients [32]. Diagnosis is based on clinical presentation of affected muscles, creatine kinase levels, and often western blotting and DNA analysis [32]. An additional complication with LGMD is that the genes involved are often causally linked to other muscle diseases as well. Phenotypes may be partially overlapping, resulting in a risk of misdiagnosis (for examples see below) [32].
With so many genes involved in a single phenotype LGMD is often considered a group of progressive muscle disorders [32,160]. However, while it is certainly possible to regard the distinct subtypes as separate disease entities, the possibility of a common mechanism underlying all forms of LGMD is worthwhile to explore;
certainly in the light of a potential intervention strategy. Most LGMD genes are not solely expressed in muscle, yet their phenotype seems to be restricted predominantly to skeletal muscle. For most of the encoded proteins a function has been ascertained (Table 1) [160].
Based on protein function LGMD has been divided into major disease mechanisms: 1) a structural defect at the muscle membrane (Sarcoglycans [221]), 2) muscle membrane repair deficiency (Dysferlin [13]), 3) defects in sarcomere remodeling, cytoskeleton structure and cytoskeleton-membrane interactions (Calpain 3 (CAPN3) [17]) [118]. With the identification of ANO5 mutations in LGMD2L a possible fourth disease mechanism has been uncovered that regards ion channel dysfunction (see discussion, [25]). Closer inspection of these pathogenic mechanisms may give an idea of commonalities between the LGMD types
VINC
TLN
PARVB ILK
ACTN
ITGAITGB
Caveolin 3
Sarcolemma
F-ACTIN Dysferlin
Calpain 3
Trim 32 FKRP
LMNA
MYOT SGCA
SGCB SGCD SGCG
TCAP TTN
ANO5
Fukutin POMT2
POMT1 Dystrophin
Laminin
beta-Dystroglycan
alpha-Dystroglycan
Extra-cellular matrix
Endoplasmic Reticulum
Nucleus Golgi Network
Vesicles
Figure 1: Schematic overview of the Dystrophin-glycoprotein anchoring complex. All identified LGMD proteins are localized in the cartoon.
LGMD and membrane stability
The skeletal muscle fibers provide contractile strength to the body. Such force is generated by the internal sarcomere. At the Z-disk the sarcomere connects to the internal Actin cytoskeleton. The filamentous Actin network is bound by the Dystrophin-glycoprotein complex (DGC) which links to the extracellular matrix and thereby allows for transduction of the mechanical force (Figure 1) [16]. Mutations in partners of the DGC have been identified in various muscle disorders, the most severe and prevalent disorder being Duchenne Muscular Dystrophy, which is caused by mutations in the DMD gene encoding for Dystrophin [115]. Dystrophin localizes to the intercellular side of the sarcolemma. It is a modular rod-like protein with an Actin binding domain at its N-terminus and a large C-terminus, connected by a large number of repeat domains. Dystrophin directly interacts with filamentous Actin and to a large protein complex at the sarcolemma. It thereby provides a molecular link between the internal Actin cytoskeleton and the extracellular matrix.
The central part of the anchoring complex is built by the transmembrane glycoprotein Dystroglycan. Dystroglycan is post-translationally cleaved. The resulting α-Dystroglycan is fully extracellular and interacts with extracellular matrix proteins such as Laminin. These interactions are mediated via glycosyl groups that are enzymatically added to Dystroglycan as it travels through the ER and Golgi. Interestingly, several mutations in the genes encoding of at least four different glycosylation enzymes have been causally linked to LGMD (Table 1). The extracellular α-Dystroglycan stays in complex with the transmembrane β-Dystroglycan. This protein interacts on the cytosolic side of the sarcolemma with Dystrophin.
The anchoring complex is stabilized in the sarcolemma by a group of proteins called Sarcoglycans, together with Sarcospan [221]. These proteins prevent extreme movement of the anchor when much force is generated by the muscle.
The four different Sarcoglycans form a tight complex, and loss of one often results in secondary loss of the others, indicating that they depend on each other’s presence for correct localization and function [221]. Mutations in all four Sarcoglycan genes have been identified in LGMD [221]. Intriguingly, α-Sarcoglycan contains a putative ATP binding site [221], suggesting additional functionalities for this protein.
LGMD and membrane repair
One of the best studied LGMD genes is Dysferlin (DYSF). The mutational spectrum for Dysferlin is extensive. Mutations are found throughout the gene without apparent hotspots and include nonsense, missense, and splice site mutations
(www.dmd.nl/dysf). Moreover, mutations in the Dysferlin gene can also cause other disease entities, such as Miyoshi Myopathy (MM [165]) and Distal Anterior Compartment Myopathy (DACM [126]. All of them are collectively referred to as “Dysferlinopathies” [160]. Intriguingly, it has been reported that a single Dysferlin mutation resulted in LGMD in one patient and MM in another [255].
The Dysferlinopathies are characterized by a late onset at age 17-25. Proteomic analysis indicated a decrease in glycolytic type I fiber marker proteins and a concomitant increase in oxidative type II marker proteins, suggesting differences in muscle fiber composition [70]. In addition, Dysferlinopathy patients have a higher number of immature fibers than healthy controls [45]. The muscle tissue is characterized by a strong inflammation of mainly monocytes and macrophages [193], and the disease is often misdiagnosed as polymyositis [193], an autoimmune disease of the muscle.
Transmembrane domain O toferlin
Fer1L6 Dysferlin
M yoferlin
C2 C2 Fer1Fer1 C2 Fer1Fer1 Fer1Fer1 DYS F DYS F C2 C2 C2
C2 Fer1Fer1 C2 Fer1Fer1 C2 C2 C2
C2 C2 Fer1Fer1 C2 Fer1Fer1 Fer1Fer1 DYS F DYS F C2 C2
C2 C2 Fer1Fer1 C2 Fer1Fer1 Fer1Fer1 DYS F C2 C2 C2
C2 C2 C2
Fer1 Fer1
C2 Fer1Fer1 C2 Fer1Fer1 C2 C2
C2 C2
C2
C2 C2
Fer1L5
Cytosolic Extra cellular
DysF domain
DYS F
Ferlin domain
Fer1
C2 domain Fer1 C2
C2
Figure 2: Schematic overview of the Dysferlin protein and its family members. Domain prediction is based on CCD tool (NCBI). The sequence N-terminal of the transmembrane domain is on the cytosolic side of the cell membrane. For Dysferlin, the extracellular sequence is ~11 amino acids only.
The Dysferlin gene encodes a 230 kDa protein that contains a single pass transmembrane domain at its C-terminus, and has seven C2 domains, three ferlin domains and two DysF domains (Figure 2). The larger part of Dysferlin is intracellular. The C2 domains are calcium-sensitive phospholipid-binding domains and are believed to be essential for Dysferlin function [243,244]. For the first C2 domain, C2A, calcium-dependent phospholipid-binding activity has been shown [243]. Through the C2 domains Dysferlin can also interact with other proteins, in certain cases also in a calcium-sensitive manner (AHNAK2 [119]). The function of the Ferlin and DysF domains are not clear. Based on structural experiments it is thought that mutations in this domain cause the protein to misfold [198], and such misfolding might affect the entire protein.
Cellular studies showed that Dysferlin is a vesicular protein that travels to and from the sarcolemma, and responds to changes in intracellular free calcium [13].
When the plasma membrane of cultured myotubes is ruptured, for instance by laser-mediated wounding, Dysferlin is rapidly recruited to the plasma membrane in a calcium-dependent manner, and accumulates at the site of the lesion [13].
This recruitment is essential for patch-fusion repair of the damaged membrane [12,13], and it has been hypothesized that a disturbed membrane repair capacity underlies the different Dysferlinopathies [214].
Dysferlin is part of the Ferlin family of proteins named after the Caenorhabditis elegans gene Ferlin (fer1) [15]. Ferlin is involved in vesicular function in C.
elegans spermatids [253]. Mutations in Ferlin result in infertility [253]. Recently it was shown that these worms also have a muscle phenotype [150]. The human genome encodes six Ferlin-like (Fer1L) genes.
Otoferlin (Fer1L2) is expressed in hair cells of the inner ear, and participates in cellular transmission of sound [217]. Mutations in Otoferlin cause an autosomal recessive form of congenital deafness (DFNB9 MIM#601071) [187]. Otoferlin is expressed in a long and a short isoform (Figure 1) [262]. In humans, the long isoform is found in brain, while the short isoform is expressed in cochlea, vestibule and brain [27,262]. In mice, only the long isoform is observed, yet a mutant mouse model recapitulates the hearing impairment [166]. Otoferlin binds to Syntaxin 1a and Snap25 (both through C2F) and is involved in calcium-regulated vesicle trafficking [209]. In addition Otoferlin interacts with the L-type voltage gated calcium channel (via C2D domain) and aids in regulating neurotransmission in hair cells [209]. Interestingly, the C2 domains that have thus far been described in Otoferlin protein-protein interactions, are present in both Otoferlin isoforms.
Myoferlin (Fer1L3) is like Dysferlin expressed in skeletal muscle, but at a lower level [63]. Myoferlin is important for myoblast differentiation, as knockdown impairs the fusion of cultured myoblasts [75,76]. Myoferlin expression increases
with differentiation [75]. While Dysferlin is mostly found at the sarcolemma of mature skeletal muscle fibers, Myoferlin is predominantly found at peri-nuclear membranes [75]. No pathogenic mutations in Myoferilin have yet been described [63]. Knockdown of Myoferlin in endothelial cells attenuated membrane repair [21], suggesting that Myoferlin has a membrane repair function as well and that it might be able to substitute Dysferlin as a potential strategy for therapy.
However, while the introduction of human Myoferlin in Dysferlin deficient mice yielded improved the membrane resealing capacity of isolated myoblasts, the general dystrophic phenotype was not rescued [2]. This suggests that Dysferlin has functions beyond membrane repair which cannot be readily substituted by Myoferlin.
The Fer1L4, Fer1L5 and Fer1L6 proteins have not yet been characterized.
Several mouse models have been described that have altered or absent Dysferlin expression [13,22]. These models recapitulate the human phenotype and show a mild progressive muscular dystrophy, and a disturbed membrane repair capability, both in vivo and in cell culture [13,22]. A high level (10-fold) of Dysferlin overexpression also causes a muscle pathology [93], different from membrane repair defects, with increased Endoplasmic Reticulum (ER) stress levels. Like Dysferlinopathy patient muscle tissue, Dysferlin deficient mouse models show strong immune infiltrate after massive muscle damage, either through repeated eccentric contractions [215] or through repeated injections with the snake venom notoxin [45,168].
It is believed that infiltration of immune cells in skeletal muscle tissue is required for normal physiology [9,231]. Throughout life and development there is ongoing communication between muscle tissue and immune cells. This communication is not only important for the clearing of pathogens and dead cells, but also seems to be important for maintenance and differentiation of skeletal muscle tissue [231]. When muscle fibers are damaged, immune cells (first neutrophils, and then monocytes/macrophages) are recruited to clean up the damaged cells that are beyond repair [9,231]. This is mediated by M1 pro-inflammatory macrophages.
In a second phase of the immune response M2 contra-inflammatory macrophages are needed to stimulate regeneration [9]. It has been shown that soluble factors from such macrophages can activate satellite cells [252].
In the absence of Dysferlin the neutrophils and monocytes appear later in the damaged muscle tissue and stay around longer [45]. Both observations may be of importance to the dystrophic phenotype, and hint at a function of Dysferlin beyond membrane repair. No increased sensitivity of Dysferlin patients to infections has been documented. It could well be that Dysferlin serves a communicative role,
between monocyte and muscle.
Recent studies showed Dysferlin expression in cells of non-muscle lineage [6].
These cells include neuronal cells and monocytes [6]. While it is interesting to speculate that Dysferlin is a membrane repair protein in these cell types as well, this is less likely in the case of monocytes/macrophages. These cells destroy other cells, clear up debris and matrix proteins, and participate in cytokine signaling, and should one be damaged others will be recruited and take its place.
It was shown that monocytes from Dysferlin knockout mice, and Dysferlinopathy patients are more aggressive in phagocytosis assays, indicating altered function [193]. This might contribute to the phenotype, but is unlikely to explain all of the pathology. In recent experiments Dysferlin was introduced in Dysferlin knockout mouse muscle tissue, by transgenesis [185] and by AAV treatment [168]. Both studies reported a complete rescue of the contraction induced phenotype [168,185].
The presence of Dysferlin in monocytes however is indicative of Dysferlin functions other than membrane repair (see Chapter 7 for further reading). And indeed, at later ages the rescued mouse models still developed a mild phenotype [168], indicative of a role of non-muscle Dysferlin in the pathogenicity.
Proteomic studies have shown Dysferlin to interact with a number of proteins, many of which complement the membrane repair function of Dysferlin. Annexins A1 and A2 are ubiquitous calcium-sensitive membrane fusogens that interact with Dysferlin, and presumably aid in patching membrane tears (for further interactions see Chapter 2) [33,162,179]. Trim72 is a redox sensor that acts upstream of Dysferlin and is essential for Dysferlin vesicle nucleation [34,36]. A mutant mouse model for Trim72 has a mild muscular dystrophy [34], but to date no pathogenic mutations have been reported in humans. This is different for the Dysferlin binding partner Caveolin 3 [172]. Mutations in this protein cause LGMD1C [186]. Caveolin 3 is an essential component of caveolae, small invaginations of the membrane [85]. It participates in Dysferlin trafficking and endocytosis. Recent experimental data, showed that Caveolin 3 affects Dysferlin endocytotosis, and that several pathogenic Caveolin 3 mutations impaired this phenomenon [36,110,111]. Both proteins are also found at T-tubuli [142,143], where they may also serve a common role. In LGMD2B loss of Dysferlin often coincides with a secondary loss of Caveolin 3 [251]. Recently, CAPN3 was also identified in the Dysferlin protein complex [120]. Mutations in the cysteine protease CAPN3 cause LGMD2A [211], and like Caveolin 3 [251], CAPN3 is also often lost in LGMD2B patient tissue [7]. Interestingly, mutations in CAPN3 were considered to represent the third pathogenic mechanism in LGMD [118].
LGMD and structural stability
Skeletal muscle is very adaptive and continuously undergoes cycles of degeneration and regeneration. To do this it needs to remodel its cytoskeleton and contractile apparatus. The contractile apparatus is formed around the giant protein Titin (3500 kDa). Titin functions as a molecular ruler along which the contractile proteins are aligned [147]. Upon contraction-induced damage this giant protein complex needs to be disassembled, to remove toxic protein fragments, and to allow for rapid reassembly [17]. It is believed that the cysteine protease CAPN3 is centrally involved in this process.
CAPN3 is a member of the Calpain family of cysteine proteases which participate in limited proteolysis in response to free calcium levels [139]. It is the only disease causing member of the family, and its expression is limited to muscle [139]. Contrary to its ubiquitous family members it is non-responsive to the endogenous inhibitor Calpastatin (which in fact is a substrate [195]), and it is extremely sensitive to fluctuations in intracellular free calcium levels (within nanomolar range [191,195]). CAPN3 is an unstable protease that autolyses upon activation. Its estimated in vitro half-life is less than ten minutes [138] and consequently not much is known of its substrates, or function. Within skeletal muscle most of the CAPN3 is found in its full-length non-autolysed form of 94 kDa [139]. This form is believed to be proteolytically inactive [72]. It localizes for 90% to the sarcomere [17], and directly interacts with Titin [107,195]. It is hypothesized that the Titin interaction is used to store inactive CAPN3, to allow for local proteolysis upon activation [107,138]. Interestingly, mutations in Titin at the CAPN3 binding site also cause LGMD [104]. Elegant proof for this model comes from experiments with transgenic mice. Mice that overexpress human CAPN3 were crossed with mice that had mutations in the CAPN3 binding domain in Titin [122,195]. While the CAPN3 transgenic mouse itself had no phenotype, the Titin mutant mouse showed a mild muscular dystrophy. Crossing both mouse models aggravated the muscle phenotype, strongly indicating the importance of Titin in buffering CAPN3 activity [122,195].
Several CAPN3 transgenic mouse models have been described, including knock- out and overexpression (of wild-type, proteolyticaly inactive and constitutively active CAPN3) mice [14,152,153,237,240]. Unlike Dysferlin, overexpression of CAPN3 is without apparent phenotype [237], suggesting that the muscle can cope with extra copies of CAPN3. It has been estimated that the number of CAPN3 binding sites on Titin greatly exceeds the amount of wild-type CAPN3 protein molecules, possibly explaining the large buffering capacity of muscle for CAPN3 and the absence of an overt phenotype [17]. All the other models however, show a mild muscular dystrophy phenotype. A first direct clue for the function of CAPN3
was found in hind-limb suspension experiments [153]. Where wild-type muscle showed strong muscle atrophy, degeneration and sarcomere remodeling during the experiment, and clear regeneration thereafter, this process was impaired in CAPN3 deficient muscle [153]. Therefore, CAPN3 is considered to be essential for muscle atrophy and possibly degeneration and regeneration. In agreement with this, the very few substrates reported are all structural proteins [98,242], and ectopic overexpression of CAPN3 causes cell rounding and detachment, indicative of cytoskeleton remodeling [242] (see Chapter 3 and 4 for further reading).
Over 400 pathogenic CAPN3 mutations have been described and roughly one third of these are miss sense (www.dmd.nl/capn3). There is no clear mutational hotspot [151]. It has been estimated that roughly one third of all mutations does not impair the proteolytic activity [184]. Modeling experiments on the experimentally determined structure of Calpain 2 indicated that several mutations affect the autolytic activity of CAPN3, showing that a change in the activity time span might already be deleterious [87,135]. This resulted in a model where locally stored inactive CAPN3 (on Titin molecules [107]) allows for local proteolysis upon activation [17]. However, the activation signal remains elusive and biochemical support for this hypothesis is wanting [17] (see Chapter 4 for further reading).
Due to the identification of CAPN3 in the Dysferlin protein complex, the hypothesis was put forward that CAPN3 functions in membrane repair [118]. It has recently been shown that ubiquitous Calpains, Calpain 1 and 2, are essential components of the non-muscle membrane repair system [183]. However, membrane repair assays on CAPN3 deficient primary myoblasts showed no clear difference in membrane repair capacity [182], indicating that the link between LGMD2A and 2B is not that straightforward.
Linking LGMD proteins
In all forms of LGMD muscle tissue is correctly formed and functional.
Concomitantly, in all three above described pathogenic mechanisms deregulation of muscle maintenance seems to be central, rather than muscle development.
In all forms of LGMD an increased level of regeneration has been observed. This suggests that when muscle maintenance processes fail, damaged muscle fibers need to be replaced. Muscle has a high capacity for regeneration, but it is not unlimited [56,57].
Muscle tissue contains a large number of satellite cells: a self renewing population of mononucleated cells that can differentiate into myoblasts, and thus give rise to new myofibers [56,57]. The self renewing capacity of these cells is a limiting factor in muscle regeneration. A possible model for LGMD could be that fibers that fail in proper maintenance are removed and replaced as long as
possible. This stress on the satellite cell population results in a quicker loss of regenerative capacity [31]. When regeneration can no longer match the loss of existing fibers, and the rate of myofiber loss exceeds the rate of replenishment, LGMD disease onset would eventually follow.
In line with this model the earliest onset of disease is described for CAPN3 (LGMD2A) patients [28]. Extrapolating from the mouse models[152,153], these patients likely have an impaired muscle degeneration and regeneration capacity, making them especially prone to defects in muscle maintenance. Dysferlin patients (LGMD2B) can still regenerate their myofibers. Onset of LGMD2B is thus later than for CAPN3 patients.
To maintain correct muscle function specific molecular signaling pathways exist. Pathways that mechanically or chemically sense contraction and resulting damage, and that need to respond appropriately. The molecular interactions between Dysferlin and Caveolin 3 [172] and CAPN3 [120] provide a first indication for a connecting molecular network that underlies muscle maintenance and LGMD pathogenicity. Maintenance processes require rapid and clear-cut communication at a molecular level. An intriguing protein in such network might be AHNAK, which has been implicated in many different processes that are important to muscle, ranging from membrane repair, vesicle trafficking, excitation-contraction coupling, and (neuronal) remodeling [3,19,52,53,90,100,106,118,119,146,175,176,20 6,220,222] (for further reading on AHNAK see Chapters 3, 5 and 6). AHNAK consists of a large body of repeat units that fold into a B-propellor [146], and a N- and C-terminus that are intrinsically disordered [113,146,227]. Given the large number of hitherto described ANAK binding partners (>10) [5,99,177], and its partially disordered folding, we propose that AHNAK functions as a Hub protein, which are important central components of protein-protein networks [109].
We hypothesized that unraveling further molecular communications would provide access to a potential protein network that underlies and unifies all forms of LGMD.
In Chapter 2 therefore a robust and reproducible method for uncovering protein- protein interactions is described for the LGMD protein Dysferlin. The uncovered interactions provide a new basis for extending the molecular networks in muscle maintenance and LGMD pathogenicity.
Table 1
Disease MIM Gene Protein Expression Function
LGMD1A #159000 MYOT Myotillin Skeletal muscle,
heart * Sarcomere
component LGMD1B #159001 LMNA Lamin A/C Ubiquitous Nuclear lamina
component LGMD1C #607801 CAV3 Caveolin 3 Skeletal muscle,
heart, brain, adrenal gland
Essential component of caveolae
LGMD1D %603511 unknown Unknown LGMD1E %602067 Unknown Unknown LGMD1F %608423 Unknown Unknown LGMD1G %609115 Unknown Unknown
LGMD2A #253600 CAPN3 CAPN3 Skeletal muscle, heart *
Proteolytic regulation of the cytoskeleton
LGMD2B #253601 DYSF Dysferlin
Skeletal muscle, heart, monocytes, kidney, trophoblast
Membrane repair
LGMD2C #253700 SGCG γ-sarcoglycan
Skeletal muscle, heart, monocytes, mesenchym
Dystrophin glycoprotein complex stability LGMD2D #608099 SGCA α-sarcoglycan Skeletal muscle, heart, monocytes
Dystrophin glycoprotein complex stability LGMD2E #604286 SGCB β-sarcoglycan Skeletal muscle, heart
Dystrophin glycoprotein complex stability LGMD2F #601287 SGCD δ-sarcoglycan Skeletal muscle, heart, intestine
Dystrophin glycoprotein complex stability LGMD2G #601954 TCAP Theletonin Skeletal muscle
and heart Sarcomere component LGMD2H #254110 TRIM32 Trim32 Skeletal muscle,
brain *
E3 ubiquitin ligase and microRNA regulator
LGMD2I #607155 FKRP Fukutin- Related Protein
Skeletal muscle, heart, thymus, kidney, placenta
*
Glycosylation enzyme
LGMD2J #608807 TTN Titin Skeletal
muscle, heart, trophoblast
Sarcomere component
LGMD2K #609308 POMT1 Protein O- Mannosyl transferase 1
Skeletal muscle, heart, cerebellum, lymph node, placenta *
Glycosylation enzyme
LGMD2L #611307 ANO5 Tmem16 Skeletal muscle,
heart, brain* Calcium-dependent Chloride channel LGMD2M #611588 FKTN Fukutin Heart, intestine,
trophoblast Glycosylation enzyme LGMD2N *607439 POMT2 Protein O-
Mannosyl transferase 2
Adrenal gland, thyroid gland, monocytes, kidney
Glycosylation enzyme
Table 1: Overview of LGMD subtypes. For each disease variant the protein and gene
identifier is given, together with the MIM number, expression summary and protein function. The expression profile is based on proteinatlas, or when marked with * genecards. Protein function was gathered from PubMed references.
Proteomic analysis of the Dysferlin protein complex unveils its importance for sarcolemmal maintenance and integrity
Antoine de Morrée 1, Paul J. Hensbergen2, Herman H.H.B.M. van Haagen1, Irina Dragan2, André M Deelder2, Peter A.C. ’t Hoen1, Rune R. Frants1, Silvère M. van der Maarel1
1 Department of human genetics, Leiden University Medical Center, Leiden, The Netherlands
2 Biomolecular Mass Spectrometry Unit, Department of parasitology, Leiden University Medical Center, Leiden, The Netherlands
-Chapter 2 has been published, PLoS One. 2010 Nov 5;5(11):e13854.-
Abstract
Dysferlin is critical for repair of muscle membranes after damage. Mutations in Dysferlin lead to a progressive muscular dystrophy. Recent studies suggest additional roles for Dysferlin. We set out to study Dysferlin’s protein-protein interactions to obtain comprehensive knowledge of Dysferlin functionalities in a myogenic context. We developed a robust and reproducible method to isolate Dysferlin protein complexes from cells and tissue. We analyzed the composition of these complexes in cultured myoblasts, myotubes and skeletal muscle tissue by mass spectrometry and subsequently inferred potential protein functions through bioinformatics analyses. Our data confirm previously reported interactions and support a function for Dysferlin as a vesicle trafficking protein. In addition novel potential functionalities were uncovered, including phagocytosis and focal adhesion. Our data reveal that the Dysferlin protein complex has a dynamic composition as a function of myogenic differentiation.
We provide additional experimental evidence and show Dysferlin localization to, and interaction with the focal adhesion protein Vinculin at the sarcolemma.
Finally, our studies reveal evidence for cross-talk between Dysferlin and its protein family member Myoferlin. Together our analyses show that Dysferlin is not only a membrane repair protein but also important for muscle membrane maintenance and integrity.
Author summary
Mutations in the gene encoding for Dysferlin result in a variety of progressive muscular dystrophies. The pathogenic mechanism of these diseases is far from clear, although it was shown that Dysferlin is critical for calcium-dependent membrane repair. To date only few interaction partners have been described, however, the identification of such complex partners does help elucidate the function of Dysferlin. Here we describe the large-scale analyses of Dysferlin’s complex partners from myoblasts, myotubes and skeletal muscle tissue. We identify >100 novel complex partners, that confirm existing hypotheses and in addition point to new putative functionalities of Dysferlin. We observed that Dysferlin participates in diverse processes in a spatiotemporal dependent manner, that together ensure proper maintenance of skeletal muscle membrane integrity.
Introduction
Dysferlin (DYSF, MIM*603009) is a 230 kDa large transmembrane protein highly expressed in striated muscle and to a lesser extent in other tissues, including monocytes, syncytiotrophoblast, endothelium, brain, pancreas, and kidney [6].
Dysferlin is found intracellularly on vesicles and at the plasma membrane. Upon laser-inflicted membrane damage Dysferlin rapidly accumulates at the site of the lesion in a calcium-dependent manner, and participates in patch-fusion repair. In the absence of Dysferlin the membrane tear is not adequately repaired and the myofiber will undergo necrosis [13].
Mutations in the Dysferlin gene cause a spectrum of adult-onset progressive muscular dystrophies including Limb Girdle Muscular Dystrophy type 2B (LGMD2B, MIM#253601), Myoshi Myopathy (MM, MIM#254130), and Distal Anterior Compartment Myopathy (DACM, MIM#606768), commonly referred to as Dysferlinopathies [15,126,165]. There is no clear genotype-phenotype correlation and the ~150 described mutations cover the complete open reading frame (www.dmd.nl/dysf). It is therefore unclear how defects in the DYSF gene cause muscular dystrophy. It has been suggested that the skeletal muscle membrane is continuously subject to mechanical wear and tear, and that the Dysferlin deficiency phenotype results from inefficient membrane repair in response to continued membrane damage [214]. Dysferlin knockout mice develop a phenotype similar to the Dysferlinopathies [13].
Dysferlin contains seven C2 domains, two DysF domains and a C-terminal transmembrane domain. C2 domains are calcium-sensitive phospholipid- binding domains, as was also shown for the first C2 domain (C2A) of Dysferlin [243], and are thought to be important for regulating Dysferlin trafficking. These domains have also been shown to interact with proteins [119]. The function of the DysF domain remains unclear [198].
Dysferlin (also known as ferlin1-like 1, FER1L1) belongs to the family of ferlin-like proteins that includes otoferlin (FER1L2, MIM *603681), Myoferlin (FER1L3, MIM *604603), FER1L4, FER1L5 and FER1L6. The family is named after ferlin, a Caenorhabditis elegans gene that when mutated causes infertility [253] and muscle dysfunction [150]. Ferlin is essential for the fusion of vesicles with the cell membrane [253]. Mutations in otoferlin cause an autosomal recessive form of congenital deafness (DFNB9, MIM#601071) [263]. Otoferlin is expressed in hair cells in the inner ear [217], and participates in the trafficking of synaptosomal vesicles [262]. Myoferlin is like Dysferlin strongly expressed in muscle [63]. It has been suggested that Myoferlin might be able to replace Dysferlin as a potential strategy for therapy of Dysferlinopathies [63,75]. The exact function of Myoferlin remains unclear. FER1L4, FER1L5 and FER1L6 proteins
have not yet been characterized.
While the role of Dysferlin in membrane repair is well established, it is less clear how the protein is regulated. Likely it requires binding partners that aid in vesicle nucleation, localization, targeting and recycling. To date only few of such cofactors have been identified, yet they yielded important insight into Dysferlin function.
MG53 (TRIM72, MIM *613288) is a redox sensor that participates in vesicle nucleation [34]. It can interact with Dysferlin [36]. Together with the Dysferlin interacting protein caveolin 3 it is involved in the trafficking of Dysferlin to and from the sarcolemma [36,110,111]. In addition, annexins A1 and A2 can bind Dysferlin in a calcium-dependent manner [162]. These proteins are membrane fusogens that participate in lysosome exocytosis [179]. They are thought to aid in Dysferlin vesicle targeting. The cysteine protease Calpain 3 (CAPN3, MIM *114240) co- immunoprecipitates with Dysferlin in skeletal muscle [118,120]. It is hypothesized to play a role in cytoskeleton remodeling [17,118], and is predicted to remodel cytoskeletal structures to allow for patch fusion repair [118,182,183]. Finally, AHNAK (MIM*103390) is found on enlargosomes which have been implicated in membrane enlargement and repair [26,52,167]. AHNAK interacts with Dysferlin in skeletal muscle, an interaction that is regulated by CAPN3 activity [118]. These protein interactions have thus yielded some information on Dysferlin function in membrane repair.
Recent data however, suggest that Dysferlin is more than a membrane repair protein. It has been shown to be involved in cytokine [45] and chemokine [68]
secretion and associates with developing t-tubules [143]. In fertilized sea urchin embryo’s Dysferlin participates in extracellular ATP signaling [60]. Moreover, a defect in membrane repair cannot fully explain the patient’s phenotype, which has been reported to include renal [130] and cardiac failure [169]. In addition, Dysferlin is considered to be a very dynamic protein, which is found in the cytosol in myoblasts and regenerating myofibers, but shows a more prominent membranous localization in mature skeletal muscle tissue [45]. It was shown that Dysferlin trafficking and endocytosis depend on its direct interaction with caveolin 3 [110,111,172].
We hypothesized that a comprehensive overview of Dysferlin function can be inferred from large-scale proteomics analysis of Dysferlin protein complexes.
Dissecting complex protein structures requires highly specific affinity binders. We have previously described heavy chain antibody fragments (HCAb) that specifically recognize Dysferlin in mice and humans [120]. These HCAb make for exceptional tools to dissect the Dysferlin protein complex due to their strong performance in immunoprecipitation (IP) experiments. By immunoprecipitation of endogenous
Dysferlin from human skeletal muscle followed by mass-spectrometry and western blotting we could previously identify CAPN3 and AHNAK to interact with Dysferlin [119,120].
In this study we compared Dysferlin protein complexes from proliferating myoblasts, differentiated myotubes and mature skeletal muscle tissue. We show that 1) we can efficiently and reproducibly immunoprecipitate Dysferlin protein complexes from these sources, 2) bioinformatics analyses of mass spectrometry identified proteins confirm a role for Dysferlin as a vesicle protein, and 3) such analyses reveal new layers of Dysferlin function which we substantiated by further exploring interactions with Myoferlin, and focal adhesion components.
Results
Robust and reproducible isolation of Dysferlin protein complexes
As Dysferlin expression is reported to increase with myogenic differentiation we hypothesized that the composition of its protein complex might also change during this process. Therefore we aimed to isolate Dysferlin protein complexes from different stages of myogenic differentiation. We used the IM2 cell model [189] as a source for Dysferlin protein complexes to establish the optimal immunoprecipitation conditions. This cell line can be grown indefinitely under
250 kDa 120 kDa
100 kDa 72 kDa 55 kDa
Loading control
CAPN3 DYSF Day 0 Day 1 Day 3 Day 5
Figure 1: Dysferlin is upregulated during myogenic differentiation in IM2 cells. IM2 cells were differentiated and harvested at day 0, 1, 3 and 5. Protein lysates were probed on western blot for Dysferlin (upper panel) and CAPN3 (lower panel).
permissive conditions, and switched to myogenic differentiation by serum deprivation. The IM2 cell model performs more consistently than the C2C12 cell line in terms of myogenic capacity and differentiates within a shorter time-span.
We first established by western blotting that IM2 cells express Dysferlin (Figure 1). Dysferlin expression increases with differentiation, in line with previous data from C2C12 myoblast cells [81]. CAPN3, an established partner in the Dysferlin protein complex is also expressed in these cells (Figure 1).
We previously reported two Dysferlin specific llama VHH heavy chain antibody fragments (HCAb) that can specifically immunoprecipitate Dysferlin from human skeletal muscle [23]. We adapted the immunoprecipitation method for cultured cells by testing different buffer conditions (Figure 2). We used the two reported Dysferlin HCAb fragments, F4 and H7 [23], and compared their performance to a non-specific HCAb (3A), selected against amyloid-β [40] and which does not recognize Dysferlin. We selected three different lysis buffers for our experiments.
The first buffer is of low ionic strength and contains 0.2% Triton. Triton is a comparatively mild non-denaturing non-ionic detergent. The second buffer is of high ionic strength containing SDS and TEA and has a strong solubilizing capacity. At the used concentration (0.1%) SDS acts as a partially denaturing detergent that is efficient at breaking low-affinity protein-protein interactions. The third buffer contains CHAPS (0.15%) and is of intermediate strength. CHAPS is a nondenaturing zwitterionic detergent that is useful for solubilizing membrane proteins. The CHAPS buffer has been used previously to precipitate Dysferlin protein complexes from muscle tissue with HCAb [23] and with conventional antibodies [37].
Both anti-Dysferlin HCAb can precipitate protein complexes in a mild buffer with low Triton concentration (Figure 2A) whereas the non-specific HCAb 3A fails to immunoprecipitate Dysferlin or other proteins. The zwitter-ionic buffer (CHAPS) results in a reduction of IP efficiency, consistent with the increased stringency of the buffer. Again, the non-specific HCAb does not immunoprecipitate detectable protein amounts. Finally we lysed the cells in SDS buffer (0.1%). The SDS is predicted to disfavor transient interactions and should therefore result in decreased IP of Dysferlin complexes. As predicted, only a limited amount of proteins coimmunoprecipitate with H7 in SDS buffer. F4 however immunoprecipitates an increased number of proteins. A concomitant western blot for Dysferlin and its interaction partner CAPN3 confirmed the Coomassie blue stained gels (Figures 2B (myoblasts) and 2C (myotubes)). From this we conclude that the HCAb can be used in different buffer conditions, and that H7 is the HCAb of choice for qualitative IP experiments, while F4 can be used for confirmation experiments.
Moreover, the Triton buffer yields the highest level of protein complexes and is
therefore preferred as lysis buffer. Because it is more difficult to homogenize tissue than cultured cells, we decided to use CHAPS buffer for skeletal muscle tissue as we have done previously [120].
Triton Chaps SDS
F4 3A H7 F4 3A H7 F4 H7 3A
HCAb 250 kDa
170 kDa 100 kDa 72 kDa 55 kDa
34 kDa 28 kDa 17 kDa
250 kDa 170 kDa 100 kDa 72 kDa 55 kDa
DYSF
CAPN3
250 kDa 170 kDa 100 kDa 72 kDa 55 kDa
DYSF CAPN3
Triton Chaps SDS
F4 3A H7 F4 3A H7 F4 3A H7 Myoblasts
Myotubes
A
B
C
Figure 2: Reproducible Dysferlin immunoprecipitation under different conditions.
IM2 myoblasts and myotubes were lysed in three different buffers and subjected to an HCAb Dysferlin immuno precipitation protocol. F4 and H7 are specific for Dysferlin while 3A is a non- specfic control HCAb. A) Coomassie stained gel of immunoprecipitation fractions from myoblasts.
B) Western blot for Dysferlin and CAPN3 corresponding to the gel in A. C) A similar western blot on myotubes IP fractions (corresponding gel not shown).
Myoblast Myotube Muscle F4 3A H7 F4 3A H7 F4 3A H7
Myoblast Myotube Muscle F4 3A H7 F4 3A H7 F4 3A H7
DYSF
250 kDa 170 kDa 100 kDa 72 kDa 55 kDa
34 kDa 28 kDa
250 kDa 170 kDa
M yoblasts: M yotubes: Muscle:
D ysferlin D ysferlin D ysferlin AH N AK AH N AK AH N AK C alpain 3 C alpain 3 C alpain 3 Annexin A2 Annexin A2
Tubulin Tubulin Tubulin
A
B
C
Described interaction partners validated by western blotPARVB PARVB PARVB
DHPR DHPR
Figure 3: Immunoprecipitation of Dysferlin from different myogenic sources is highly reproducible. A) Coomassie blue stained gel of IP fractions from IM2 myoblasts, myotubes and skeletal muscle tissue. IP samples of two Dysferlin specific HCAb (F4 and H7) yield highly similar staining patterns, contrary to a non-specific HCAb (3A). B) A concomitant western blot for Dysferlin confirms the IP. C) All protein bands stained in A), lanes H7, were excised, in-gel trypsin digested and analyzed by mass spectrometry. Many described interaction partners of Dysferlin were identified by western blot (Figure S2) and thus confirmed.
Dysferlin interactions in myoblasts, myotubes, and skeletal muscle tissue
We continued with preparing lysates for IM2 myoblasts, IM2 myotubes at 5 days post-differentiation (when spontaneous contraction was observed), and human skeletal muscle tissue. After cell lysis in Triton buffer or tissue homogenization in CHAPS buffer, Dysferlin protein complexes were immunoprecipitated from these lysates using HCAb H7, separated on SDS-PAGE gels and stained with Coomassie- blue for further mass spectrometry analysis (Figure 3A). We monitored the IP procedure with western blot (Figure S1). The IP was confirmed by western blot with a conventional antibody specific for Dysferlin (Figures S1 and 3B). With western blotting we could confirm six of the described interactions (TUBA [41]
(Figure S1) CAPN3 [23], AHNAK [8] (both in Figure S2), PARVB [42], DHPR [43], ANXA2 [21] (data not shown), summarized in Figure 3C. Mass spectrometry analysis confirmed the presence of one additional reported Dysferlin interaction partner (MG53 [18] Table S1), but also resulted in the identification of many new putative binding partners (complete lists in table S1). Skeletal muscle cells contain a large amount of sarcomeric proteins, and components of the dystrophin glycoprotein complex (DGC). We did not observe the core DGC components in the IP fractions, and the amount of sarcomeric proteins is low, consistent with the fact that the HCAb specifically recognize and immunoprecipitate Dysferlin.
We conclude that the IP method efficiently and reproducibly enriches for Dysferlin protein complexes.
In our IP dataset for myoblasts we identified 521 proteins annotated in online databases, 344 proteins in myotubes and 229 proteins in skeletal muscle tissue (full lists in supplemental tables S1 and S2). 115 proteins are shared by all datasets (Figure 4). These proteins we therefore consider a core-set of most likely
115 126
84 0
103 196
30
Tissue 229
M yotube 344 M yoblast 521
Figure 4: Venn diagram showing the overlap of Dysferlin interaction partners in myoblast, myotubes and skeletal muscle tissue. All Coomassie stained bands from the H7 IP shown in Figure 3A were excised, in-gel digested and analyzed by mass spectrometry.
115 proteins are consistently identified in all three protein sources.
Dysferlin interacting proteins, or representative of functionalities of Dysferlin that do not change during differentiation. As mentioned before, the tissue IP was performed under more stringent conditions. Interestingly, 85% of the proteins identified in the tissue IP were also identified in the cultured cells. This underlines the reproducibility of the IP experiment, and indicates a relatively low number of proteins that nonspecifically co-immunoprecipitate with Dysferlin. An unexpected finding was that there are no proteins shared by IP from tissue and myotube alone. This might be explained by the more stringent conditions used for tissue homogenization.
Potential interaction between Dysferlin and Myoferlin
We were surprised to consistently identify the Dysferlin homolog Myoferlin to co-immunoprecipitate with Dysferlin (Mascot protein score of 300, 12 unique peptides in myoblast IP sample). We do not think this occurs due to cross reactivity of our antibodies. Myoferlin shares 68% amino acid sequence similarity with Dysferlin [63]. However, the Dysferlin HCAb were previously shown not to cross- react with recombinant Myoferlin [120]. Moreover, the Dysferlin HCAb failed to immunoprecipitate proteins from Dysferlin negative myoblasts and myotubes, as judged from Coomassie stained gels and western blots for Dysferlin and Myoferlin (not shown). Finally, when we used western blotting to confirm the presence of full-length Myoferlin in the H7 and F4 IP fractions (Figure 5), we observed that it is absent in the control HCAb 3A fractions. Highest levels of Myoferlin are detected in the myoblast IP fractions. We therefore think that Dysferlin and Myoferlin can interact. We further verified this interaction using U2OS cells which express Myoferlin but not Dysferlin (Figure 5B). We performed IP experiments on non- transfected U2OS cells, and U2OS cells transiently transfected with recombinant Dysferlin and analyzed IP fractions on western blot. As expected, F4 and H7 cannot immunoprecipitate Myoferlin from wild-type U2OS cells (Figure 5B). We could only detect Myoferlin, but not Dysferlin, in the non-bound fractions, showing that U2OS cells only express Myoferlin. This is also additional support that the HCAb are specific for Dysferlin and cannot directly bind to Myoferlin. However, Myoferlin is co-immunoprecipitated from U2OS cells expressing recombinant Dysferlin (Figure 5B, F4). Thus, F4 can immunoprecipitate Myoferlin, albeit weakly, only when Dysferlin is co-expressed. According to previous studies Myoferlin is strongly expressed in myoblasts, prominently in mature muscle myonuclear membranes, and to a lesser extent at the sarcolemma [63]. Dysferlin is mostly observed at the sarcolemma. This indicates that only a small fraction of both proteins co-occur in muscle tissue, consistent with the observation that only a fraction of Myoferlin co-immunoprecipitates with Dysferlin. All together this indicates that Myoferlin
and Dysferlin can be part of the same protein complex, and opens the possibility of their co-existence in a single protein complex in muscle.
250 kDa 130 kDa 100 kDa
250 kDa 130 kDa
DYS F F4 H7 MYOF
U2OS
Non-bound Immunoprecipitation
F4 H7 F4 H7
Wild type Dysferlin Mock Wt Dysf Mock
F4 F4 F4
250 kDa
F4
Myoblast
3A H7 F4Myotube
3A H7 F4Muscle
3A H7 MYOFA
B
*
130 kDa
Figure 5: Myoferlin and Dysferlin may be present in the same protein complex. A) Dysferlin IP fractions were stained in western blot with a Myoferlin specific antibody, to verify that full-length Myoferlin is co-immunoprecipitated with Dysferlin, and not with the control HCAb 3A. B) U2OS cells, which express endogenous Myoferlin but not Dysferlin, were transfected with Dysferlin cDNA or empty vector (mock), and both untransfected (wild-type) and transfected cells were lysed and subjected to an IP experiment with F4 and H7. IP and non-bound fractions were analyzed on western blot for Dysferlin (lower panel) and Myoferlin (upper panel). F4 and H7 can only immunoprecipitate Dysferlin from Dysferlin transfected cells (lower panel), consistent with the absence of endogenous Dysferlin expression. Myoferlin is not immunoprecipitated from wild-type cells, though it can be detected in the non-bound fractions. However, Myoferlin is immunoprecipitated by F4 from Dysferlin expressing U2OS cells (band marked by asterisk).
Bioinformatics analysis of all identified complex partners
We continued with the three datasets of putative Dysferlin interacting proteins and performed bioinformatics analyses to characterize the Dysferlin protein complex. We first analyzed the datasets with the STRING interaction database (http://string-db.org/). STRING contains information of known annotated protein- protein interactions. 14 Dysferlin protein-protein interactions are recorded in STRING for human and mouse, of which 6 are direct interactions described in PubMed references (CAV3, CAPN3, AHNAK, PARVB, ANXA1, ANXA2). The others are interactions inferred from co-occurrence in MedLine abstracts and protein databases. Protein-protein interactions described in recently published studies, such as TUBA [10], are not yet annotated, and therefore not included in STRING.
The remaining 8 predicted Dysferlin interactions in STRING are not supported by experimental evidence and include DGC partners. These interactions are not identified in our datasets, and we consider them unlikely given the existing literature on Dysferlin. Thus, STRING does not contain any information on the Dysferlin protein complex partners, and therefore we consider the majority of interactions identified with our IP experiments as novel Dysferlin interactions.
Literature analysis of the Dysferlin protein complex
We then verified whether there was any prior support in the biomedical literature for the identified interactions of Dysferlin by calculating and comparing the textual overlap in MedLine abstracts for Dysferlin and other proteins (Figure S3). We used the concept profile text-mining technique [133]. In this method all concepts (e.g.
proteins, diseases, chemicals, GO terms) that are associated with a given protein identifier in MedLine abstracts are assembled into a concept profile. The amount of overlapping concepts between two proteins is a measure of relatedness; it records both explicit (co-mentioning of two proteins in the same abstract) and implicit (both proteins co-mentioned with a third concept) relationships. We ranked all putative interaction partners based on their match scores and subsequently analyzed, through Area under the ROC Curve (AuC) analysis, whether the identified interaction partners of Dysferlin were among the top ranked proteins.
An AuC of 0.5 would imply that the identified proteins are ranked randomly over the whole range of proteins, while an AuC of 1 would imply that all top-ranked proteins co-immunoprecipiate with Dysferlin. We obtained an AuC of ~0.8 (Figure S3), indicating that the proteins indentified in the IP experiments are related to Dysferlin in MedLine abstracts, and may share similar functions and/or pathways.
In Table 1A we list the top 9 proteins that share strongest conceptual overlap with Dysferlin. This includes two of the described interaction partners (ANXA2 and TRIM72), the others are novel binding partners. To gain a better idea of which type
