crossmark

Research

Author’s Choice © 2016 by The American Society for Biochemistry and Molecular Biology, Inc. This paper is available on line at http://www.mcponline.org Molecular Signatures of Membrane Complexes Underlying *□S

Rolf Turk‡§¶ʈ**, Jordy J. Hsiao¶, Melinda M. Smits¶, Brandon H. Ng¶, Tyler C. Pospisil‡§¶ʈ**, Kayla S. Jones‡§¶ʈ**, Kevin P. Campbell‡§¶ʈ**, and Michael E. Wright¶‡‡

Mutations in encoding components of the sar- The muscular dystrophies are hereditary diseases charac- colemmal -glycoprotein complex (DGC) are re- terized primarily by the progressive degeneration and weak- sponsible for a large number of muscular dystrophies. As ness of . Most are caused by deficiencies in such, molecular dissection of the DGC is expected to both associated with the cell membrane (i.e. the sarco- reveal pathological mechanisms, and provides a biologi- lemma in skeletal muscle), and typical features include insta- cal framework for validating new DGC components. Es- bility of the and consequent death of the myofi- tablishment of the molecular composition of plasma- ber (1). complexes has been hampered by a One class of muscular dystrophies is caused by mutations lack of suitable biochemical approaches. Here we present in genes that encode components of the sarcolemmal dys- an analytical workflow based upon the principles of pro- tein correlation profiling that has enabled us to model the trophin-glycoprotein complex (DGC). In differentiated skeletal molecular composition of the DGC in mouse skeletal mus- muscle, this structure links the to the cle. We also report our analysis of protein complexes in intracellular . The DGC consists of mice harboring mutations in DGC components. Bioinfor- (DG)1, the - complex, dystrophin (DMD), matic analyses suggested that cell-adhesion pathways and dystrophin-associated proteins. In its mature form, the were under the transcriptional control of NF␬BinDGC DG component is comprised of a cell-surface-associated al- mutant mice, which is a finding that is supported by pre- pha subunit (␣-DG) and a transmembrane beta subunit (␤-DG) vious studies that showed NF␬B-regulated pathways un- (2). Whereas ␣-DG functions as receptor for extracellular pro- derlie the pathophysiology of DGC-related muscular dys- teins (3), ␤-DG binds dystrophin via its cytoplasmic domain trophies. Moreover, the bioinformatic analyses suggested and thereby links the cytoskeleton to the plasma mem- that inflammatory and compensatory mechanisms were brane (4). The dystrophin-associated proteins are alpha-dys- activated in skeletal muscle of DGC mutant mice. Addi- trobrevin (DTNA), alpha--1 (SNTA1), beta-syntro- tionally, this proteomic study provides a molecular frame- work to refine our understanding of the DGC, identifica- tion of protein biomarkers of neuromuscular disease, 1 The abbreviations used are: DG, Dystroglycan; DGC, Dystrophin- and pharmacological interrogation of the DGC in adult glycoprotein complex; DMD, Dystrophin; ␣-DG, Dystroglycan alpha skeletal muscle https://www.mda.org/disease/congenital- subunit; ␤-DG, Dystroglycan beta subunit; DTNA, alpha-; muscular-dystrophy/research. Molecular & Cellular SNTA1, alpha-syntrophin-1; SNTB1, beta-syntrophin-1; NNOS, neu- Proteomics 15: 10.1074/mcp.M116.059188, 2169–2185, ronal nitric oxide synthase; SGCA, alpha-sarcoglycan; SGCB, beta- 2016. sarcoglycan; SGCD, delta-sarcoglycan; SGCG, gamma-sarcoglycan; SSPN, Sarcospan; 2DGE, 2-dimensional gel electrophoresis; dMS, Directed mass spectrometry; FASP, Filter-aided sample preparation; SCX, Strong-cation exchange; LC, Liquid chromatography; BSA, Bo- From the ‡Howard Hughes Medical Institute, §Senator Paul D. vine serum albumin; AMU, Atomic mass unit; MW, Molecular weight; Wellstone Muscular Dystrophy Cooperative Research Center, ¶De- FDR, False discovery rate; ontology (GO), Wild type (WT), partment of Molecular Physiology and Biophysics, ʈDepartment of Protein-protein interaction (PPI), -1 (PLEC1), DES, ; Neurology, **Department of Internal Medicine, Roy J. and Lucille A. UTRN, Utrophin; GRB2, receptor-bound 2; PLCB2, Carver College of Medicine, The University of Iowa, Iowa City, Iowa Phospholipase C-beta-2; NMJ, ; NF␬B, Nu- Author’s Choice—Final version free via Creative Commons clear factor kappa Beta; ITGA7, -alpha-7; ITGB1, Integrin- CC-BY license. beta-1; ITGA5, Integrin-alpha-5; ITGA6, Integrin-alpha-6; ITGAM, In- Received February 19, 2016, and in revised form, March 23, 2016 tegrin-alpha-M; ITGB2, Integrin-beta-2; ITGB3, Integrin-beta-3; sCK, Published, MCP Papers in Press, April 20, 2016, DOI 10.1074/ Serum creatine ; B2M, Beta-alpha-2-microglobulin; CLU, Clus- mcp.M116.059188 terin; EGFR, Epidermal growth factor receptor; FN1, Fibronectin; TTN, Author contributions: RT designed and performed experiments, ; SRM, Selective-reaction monitoring; SID, Stable-isotope dilu- processed and analyzed data and wrote the manuscript. JJH, MMS, tion; GR, Glucocorticoid receptor; WGA, Wheat-germ agglutinin; and BHN performed experiments and processed MS data. KPC wrote SDS-PAGE, Sodium dodecyl sulfate-polyacrylamide gel electropho- the manuscript. MEW designed experiments, performed experiments, resis; DTT, Dithiothreitol; ACN, Acetonitrile; Q-TOF, Quadrupole and wrote the manuscript. Time-of-Flight.

Molecular & Cellular Proteomics 15.6 2169 Molecular Signatures of Membrane Protein Complexes phin-1 (SNTB1), and neuronal nitric oxide synthase (NNOS). ple; this shortcoming is borne out by the underrepresentation Further stabilization of the DG-axis occurs through the of this class of proteins in data sets from such analyses (20). sarcoglycan-complex, which consists of five subunits: alpha- Second, as stated above, integral membrane proteins are sarcoglycan (SGCA), beta-sarcoglycan (SGCB), delta-sar- poorly resolved by 2DGE because they are insoluble (21). coglycan (SGCD), gamma-sarcoglycan (SGCG), and Sar- Although recent improvements in sample preparation have cospan (SSPN) (5). made the 2DGE method more amenable to the proteomic A common feature of the DGC-related muscular dystro- analysis of skeletal muscle (e.g. biochemical fractionation re- phies is loss or severe reduction of the entire DGC, because moving high-abundance contractile proteins from crude mus- of genetic abnormalities of a single component (6–8). This cle extracts facilitated proteomic analysis of low-abundance general loss of proteins leads to a pathogenic mechanism that membrane proteins by the 2DGE strategy (22)), this approach is hypothesized to account for the significant overlap in path- did not improve the detection of integral membrane proteins. ological features of numerous muscular dystrophies. Although Nongel based proteomic approaches have gained in use gene-expression profiling has been used to identify the com- over the past years and also been used to study the DGC in mon pathogenic mechanisms underlying disease in the con- muscular dystrophy. For example, DGC components were text of DGC mutations (9), this approach failed to provide a directly immunoprecipitated and the samples were then bio- deeper understanding of the structural interface between the chemically analyzed using shotgun proteomic methods (23– extracellular matrix and the cytoskeleton, because no infer- 25). Although this approach identified many components of ences could be made about the cellular localization of the the DGC, and even putative novel interacting proteins, the proteins affected by differential . scale of this proteomic analysis was relatively small. More- A complete molecular understanding of communication be- over, the antibody used to immunoprecipitate the DGC might tween the extracellular matrix and intracellular environment have influenced the stability of the complex, and thus the via integral membrane proteins on the cell surface requires proteomic results likely depended on both antigen and anti- comprehensive biochemical characterization of the plasma- body. A second approach was based on analysis of soluble membrane proteome. However, despite recent advances in protein fractions from skeletal muscle by label-free, reverse- proteomic approaches, the analysis of intact membrane-pro- phase liquid chromatography followed by MS/MS (26). In a tein complexes continues to pose technical difficulties (10). third (and similar) approach, a label-free shotgun proteomic First, the experimental conditions under which particular analysis was used to analyze from the dys- membrane protein complexes can be extracted and pre- trophin-deficient (mdx) and WT mice, revealing differential served must often be determined empirically. Second, integral expression of a number of proteins involved in stabilizing the membrane proteins are inherently insoluble and notoriously basal lamina and organizing the cytoskeleton (27). difficult to manipulate and characterize biochemically. Third, Recent advances in mass spectrometer instrumentation tandem mass spectrometry (MS)-derived polypeptide se- (i.e. linear quadrupole-Orbitrap mass analyzer) in conjunction quences typically exclude the mostly hydrophobic amino ac- with optimized sample preparation workflows (i.e. isoelectric ids that comprise the transmembrane domains of integral peptide focusing OffGel fractionator) and software platforms membrane proteins (11); typically, highly concentrated acids (i.e. MaxQuant) has facilitated the in-depth proteomic analysis (e.g. formic acid) or volatile aqueous/organic solvent mixtures of a skeletal line, skeletal , and single (e.g. acetonitrile) are required for the analysis of membrane muscle fibers using label-free shotgun proteomic acquisition proteins by such approaches (12, 13). schemes (28, 29). Similar label-free shotgun proteomic work- Several proteomic strategies have been used to detect flows have recently identified global proteomic changes in differentially expressed integral membrane proteins involved contraction, energy metabolism, extracellular matrix, and cy- in muscular dystrophy. The predominant method has been toskeleton of skeletal muscle of mdx mice (16). Moreover, 2-dimensional gel electrophoresis (2DGE) followed by in-gel interfacing differential subcellular protein fractionation of skel- trypsin digestion and tandem MS (MS/MS) (14–16). More etal muscle tissue with the label-free shotgun approach un- recently, 1-dimensional gel electrophoresis (1DGE) has been covered a reduction in full-length dystrophin isoform Dp427 utilized in the characterization of normal, aging, and dystro- expression and perturbations in the expression of protein phic skeletal muscle (17–19). Although differentially ex- networks involved in metabolism, signaling, contraction, ion- pressed proteins in muscular dystrophy samples were iden- regulation, protein folding, the extracellular matrix, and the tified using the 2DGE proteomic strategy, a number of cytoskeleton in mdx4cv mice (15). Despite these findings, the technical shortcomings make the use of this approach im- proteomic identification of the DGC expressed in skeletal practical for the study of integral membrane proteins by MS/ muscle remains incomplete to date. MS. First, the low resolving power and limited dynamic range In this report, we present a novel biochemical workflow for of the 2DGE method make it difficult to separate complex the isolation, identification, and mass spectrometry-based protein samples and acquire MS/MS data on low-abundance quantification of protein complexes on the plasma membrane integral membrane proteins that may be present in the sam- of mouse skeletal muscle, and report on several important

2170 Molecular & Cellular Proteomics 15.6 Molecular Signatures of Membrane Protein Complexes

discoveries related to muscular dystrophy using this work- were eluted by 1-h incubation with 0.3 M N-acetyl-D-glucosamine in flow. Using lectin (wheat-germ)-affinity chromatography in washing buffer, at 4 °C. Buffer exchange (0.1% digitonin, 50 mM combination with sucrose-gradient fractionation to isolate Tris-HCl pH 7.4, 150 mM NaCl, and proteinase inhibitors) was per- formed to remove N-acetyl-D-glucosamine. Antibody beads were such protein complexes, we carried out MS/MS-based pro- generated by coupling ϳ3.5 mg of anti-dystroglycan antibody (IIH6), teomic analysis of the DGC in both WT mice and mouse anti-ryanodine receptor antibody (XA7), or bovine serum albumin models of DGC-related muscular dystrophy. Specifically, la- (BSA) to Sepharose beads using activated CNBr according to man- bel-free, directed mass spectrometry analysis (dMS) (30, 31) ufacturer’s protocol (GE Healthcare Life Sciences, Cambridge, United of proteins fractionated by sucrose-gradient centrifugation Kingdom). WGA-enriched proteins were incubated with antibody beads, followed by washing 3ϫ (0.1% digitonin, 50 mM Tris-HCl pH validated the comigration of components of the DGC in skel- 7.4, 150 mM NaCl, proteinase inhibitors), and eluted in 0.1 M triethyl- etal muscle from WT mice, but not the dystrophin-deficient amine, pH 11.5. Western blots were incubated with antibodies against mdx and delta-sarcoglycan-null (Sgcd-null) models of mus- ␣-DG and ␤-DG (AF6868), DMD (Abcam), SGCA (20A6), SGCB (G26), cular dystrophy (32, 33). Moreover, comparative analysis of SGCG (21B5), SGCD (R214), ADBN (Transduction Laboratories, San protein expression between the WT and DGC-mutant mice Jose, CA), and NNOS (R200) as described elsewhere (36). Experimental Design and Statistical Rationale— revealed transcriptional up-regulation of a protein network Processing of MS Samples—Using large-scale filter assisted sam- involved in cell adhesion in the DGC-mutant mice, strongly ple preparation, samples were buffer exchanged four times with 8 M suggesting that compensatory cell adhesion network(s) are Urea, 50 mM Tris-HCl, pH 8.5, and 100 mM beta-mercapto-ethanol. activated in these models of muscular dystrophy. In summary, The first three rounds of buffer exchange were performed for 2 h, and our biochemical workflow effectively interrogated the compo- the fourth overnight. Next, samples were concentrated (Amicon Ultra, sition of the DGC and DGC-related muscular dystrophies in 10 kDa cut-off MW, Millipore, Billerica, MA) and separated by SDS- PAGE on 4–12% gels (Invitrogen) for analysis by silver staining. skeletal muscle, and also provides an experimental strategy Samples were reduced with 10 mM dithiothreitol (DTT) at 37 °C in 8 M to validate novel DGC components in skeletal muscle using Urea, 50 mM Tris-HCl, pH 8.5 for 30 min, and subsequently alkylated targeted proteomic methods. with 55 mM iodoacetamide for 30 min in the dark. Samples were diluted to a final concentration of 0.5 M urea with 50 mM Tris-HCl pH EXPERIMENTAL PROCEDURES 8.5. Each sample was treated with 1 ␮g of sequence-grade trypsin (Promega, Sunnyvale, CA) and incubated overnight at 37 °C. Samples Animals—Animal care, ethical usage, and procedures were ap- were then spiked with a tryptic digest of bovine serum albumin (BSA, proved and performed in accordance with the standards set forth Fremont, CA) containing iodoacetic-acid alkylated cysteine residues by the National Institutes of Health and the Animal Care Use and (Michrom Bioresources, Auburn, CA), at a 1:75 molar ratio. Samples Review Committee at the University of Iowa. C57BL6/J and mdx mice were obtained from The Jackson Laboratory. The Sgcd-null were acidified and desalted on Vydac C18 spin-columns (The Nest mice were developed in the Campbell laboratory as described Group), and then subjected to SCX fractionation on polysulfoethyl-A elsewhere (33). packed spin-columns (The Nest Group) according to the manufactu- Immunohistochemistry—Histopathological studies were per- rer’s protocol. Briefly, desalted samples were dissolved in SCX buffer formed as before (33, 34). Polyclonal antibodies against DMD (Ab- B(5mM KHPO4, 25% acetonitrile (ACN, Southborough, MA)) and cam, Cambridge, UK) and SGCD (R214), and monoclonal antibod- loaded onto SCX spin-columns, and the tryptic digests were then ies against ␣-DG (IIH6), ␤-DG (AP83), SGCA (20A6), and SGCG released from the SCX spin-columns using a three-step KCl elution (21B5) were used for immunohistochemistry. gradient developed from a mixture of buffer B and buffer C (5 mM Sucrose Density Gradient Biochemistry—Skeletal muscle was iso- KHPO4, 25% ACN, 350 mM KCl). Salt-eluted fractions were desalted, lated from adult mice, and microsomes were prepared as described dried down, and dissolved in MS loading buffer (1% acetic acid, 1% ϭ elsewhere (35). Nonwashed microsomes were solubilized overnight at ACN). A biological replicate n 1 for the proteomic comparison 4 °C while rotating in a solution of 1% digitonin, 50 mM Tris-HCl pH between wt, mdx, and Sgcd mouse models. The application of or- 7.4, 150 mM NaCl, and proteinase inhibitors. Nonsolubilized proteins thogonal protein fractionation methods (i.e. sucrose gradient density were pelleted at 100,000 ϫ g. The supernatant was incubated over- centrifugation, and strong-cation exchange peptide chromatography) night with wheat-germ agglutinin (WGA)-agarose beads (Vector Lab- coupled with the directed mass spectrometry workflow facilitated an oratories, Burlingame, CA) at 4 °C. The beads were washed 3x in in-depth proteomic comparison of skeletal muscle tissue between washing buffer (0.1% digitonin, 50 mM Tris-HCl pH 7.4, 150 mM NaCl, wildtype and muscular dystrophy mouse models. WGA-enriched proteinase inhibitors), and proteins were eluted by 1-hour incubation trypsin-digested samples from each mouse strain were subjected to with 0.3 M N-acetyl-D-glucosamine in washing buffer, at 4 °C. The spin-column strong cation exchange chromatography. Each salt- eluted proteins were then fractionated by centrifugation (2 h at bumped SCX fraction was subjected to in-depth proteomic coverage 238,000 ϫ g) through a 5–30% sucrose gradient. Fractions were using the directed MS workflow. The iodoacetic-acid, trypsin-di- collected from the top of the gradient and analyzed by SDS-PAGE. gested BSA protein standard was spiked into each protein sample to Coimmunoprecipitation Biochemistry and Immunoblotting—Skele- provide independent markers to detect systematic errors within and tal muscle was isolated from rabbit skeletal muscle (Pelfreeze, 41225– between independent samples. The spiked standard also provided a 2), and microsomes were prepared as described elsewhere (4). Total method to detect systematic errors in mass spectrometry instrumen- microsomes were solubilized overnight at 4 °C while rotating in a tation that could skew the proteomic findings. The proteomic findings solution of 1% digitonin, 50 mM Tris-HCl pH 7.4, 150 mM NaCl, and for each mouse strain were derived from pooled skeletal muscle proteinase inhibitors. Nonsolubilized proteins were pelleted at samples representing 10–15 individual mice from the wildtype, mdx, 100,000 ϫ g. The supernatant was incubated overnight with wheat- or Sgcd mouse colonies. germ agglutinin (WGA)-agarose beads (Vector Laboratories) at 4 °C. Mass Spectrometry—The samples were subjected to LC-MS/MS The beads were washed 3x in washing buffer (0.1% digitonin, 50 mM on the 6520 Agilent Accurate-Mass Quadrupole Time-of-Flight (Q- Tris-HCl pH 7.4, 150 mM NaCl, proteinase inhibitors), and proteins TOF) mass spectrometer, interfaced with the HPLC Chip Cube. The

Molecular & Cellular Proteomics 15.6 2171 Molecular Signatures of Membrane Protein Complexes

samples were loaded onto the large-capacity C18 Chip II (160 nL formed on a Bio-Rad iCycler. Statistical analyses were performed as enrichment column, 9 mm analytical column) and subjected to LC- described elsewhere (40). MS/MS analysis for 90 min in a gradient of 1.5% to 35% buffer B (100% ACN, 0.8% AA). Full MS (MS1) data were acquired with a mass RESULTS range of 400–1250 m/z and an acquisition rate of 1 spectra/second. From these data, an ion preference list was generated using Agilent Isolation and Separation of N-linked Glycosylated Proteins MassHunter Qualitative Software. dMS was performed using the fol- from Skeletal-muscle Microsomes—We developed a compar- lowing settings: a maximum of 10 ions per cycle, a narrow isolation ative proteomic workflow and applied it to skeletal muscle ϳ width ( 1.3 atomic mass units), precursor masses dynamically ex- tissue from WT, mdx, and Sgcd-null mice. This consisted of the cluded for 30 s after 8 MS/MS in a 30-s time window, and use of the preferred ion list. The capillary voltage and capillary temperature following steps (also summarized in Fig. 1). First, nonwashed settings for MS were set to 1800 V and 330 °C, respectively. The microsomes were prepared from muscle isolated from WT, infused reference mass of 1221.9906 was used to correct precursor mdx, and Sgcd-null mice, and solubilized in 1% digitonin. These m/z masses in each LC-MS/MS experiment. samples were subjected to lectin-affinity chromatography using Protein Identification—The raw.d files were searched against the wheat-germ agglutinin, and then eluted with N-acetyl glucosa- UniProt mouse database (downloaded April 18th, 2012; with 55190 mine. Next, the samples were layered on a 5–30% sucrose entries) using SpectrumMill Software version B.04.00.127 and the following settings: precursor mass tolerance of 25 parts per million gradient and subjected to centrifugation. Select sucrose-den- (ppm), product mass tolerance of 200 ppm, and a maximum of two sity gradient fractions (4–8) were subjected to large-scale, filter- trypsin miss cleavages. Searches for post-translational modifications aided sample preparation (FASP) and digested with trypsin (41). included a static carbamidomethylation on cysteine residues (C ϭ The C18-desalted peptide samples were further fractionated by 57.02146 AMU), differential modifications for oxidized methionine strong-cation exchange (SCX) spin-column chromatography, (M ϭ 15.9949 AMU), phosphorylated serine, threonine, tyrosine (STY ϭ 79.9663 AMU), and ubiquitinated (K ϭ 114.0429 AMU). and then subjected to high-resolution LC-MS to generate a For normalization between the samples, the raw.d files were searched preferred list of 2ϩ and 3ϩ peptide ions. Select samples were against the UniProt bovine database with same search settings as then subjected to directed mass spectrometry, and the resulting above except the files were searched with a static carboxymethyla- data files were searched with the Spectrum Mill program for ϭ tion on cysteine residues (C 58.005 AMU). A total of 14 cysteine- nonredundant protein identifications (IDs). containing BSA peptides were quantified and the total intensities of these peptides were collated for each sample and the differences Relative quantification of proteins across sucrose fractions between samples were normalized (supplemental Table S1). The pep- was made possible by spiking each FASP sample with ap- tide hits were autovalidated with the “Fixed thresholds” setting using proximately ϳ250 picomoles of alkylated (using iodoacetic- the following values: Score Threshold of 3, % Spectral Intensity (SPI) acid) trypsin-digested bovine serum albumin (BSA). This step Threshold of 30, and Rank 1-Rank 2 Threshold of 2, with delta was performed prior to the fractionation of tryptic peptides by Fwd-Rev not checked such that the initial False-Discovery Rate (FDR) SCX spin-column chromatography, so that BSA peptides (in was Յ 7%. To reduce the FDR well below ϳ7% the proteomic results were further filtered by applying a delta Fwd-Rev score of 1.2 to particular alkylated-cysteine BSA peptides) would cofraction- provide the most stringent filter in the removal of false positive protein ate with sample tryptic peptides based on their charge state. identifications. When there were distinct peptides that uniquely rep- The ϩ1 atomic mass unit (AMU) differential between iodoa- resent multiple protein isoforms, the individual member is reported, cetic acid (i.e. MW-ϩ58, carboxymethyl) and iodoacetamide and the quantification was based on the mean of the peptide spec- (i.e. MW-ϩ57) in the spiked samples makes it possible to trum matches (PSM) for that specific isoform. Data Deposition—The mass spectrometry proteomics data have distinguish between peptides containing cysteines with iodo- been deposited to the ProteomeXchange Consortium (http://proteome acetic-acid or iodoacetamide. Trypsin-digested iodoacetic- central.proteomexchange.org) via the PRIDE partner repository with acid labeled BSA was selected as the external standard be- the data set identifier ϽPXD004020Ͼ. The annotated spectra are ac- cause it incorporates 26 tryptic cysteine-containing peptide cessible through the MS-Viewer module at http://prospector2.ucsf.edu/ ions (ϩ2, ϩ3 peptide ions, m/z window of 400–1250) that are prospector/cgi-bin/msform.cgi?formϭmsviewer. Key: 61ze1ebl1o. Network Visualization and Public Protein Interaction Databases fractionated based on their charge state during SCX spin- Used—Protein interaction networks were visualized using Cytoscape column chromatography. Subjecting the desalted peptide 3.1.0 (http://www.cytoscape.org)(37). Interactions were identified us- samples to high-resolution LC-MS on the Agilent 6520 Qua- ing the GeneGo bioinformatic software (http://www.genego.com). drupole Time-of-Flight high-resolution mass spectrometer Functional Analyses— analyses were performed us- yielded a list of preferred peptide ions, and these were tar- ing the “WebGestalt” software package (38). Cluster analysis was geted for MS/MS sequencing using a dMS protocol and the performed using the “Cluster” software package (39). K-means clus- ters were calculated based on absolute correlation (centered). MassHunter data acquisition software. Pilot proteomic exper- Quantitative RT-PCR—Total RNA was isolated from skeletal mus- iments demonstrated that the dMS acquisition protocol was cle from WT, Sgcd-null, and mdx mice using TRIzol (Ambion, South superior to the traditional data-dependent acquisition (DDA) San Francisco, CA) according to the manufacturer’s protocol. Five protocol in the identification and quantification of trypsin- micrograms of total RNA was treated with DNase I (Promega), and digested bovine serum albumin on the 6520 QTOF (Supple- then reverse transcribed using AMV reverse transcriptase (Roche, Basel, Switzerland) according to the manufacturer’s protocol. qRT- mental Table 2). This label-free analytical workflow was ex- PCR was performed using RT2 qPCR primers (Qiagen, Hilden, Ger- pected to facilitate the relative quantification of proteins in many) against Itga5, Itga6, Cd151, and Vav1. Reactions were per- muscle samples, and to provide a platform for monitoring and

2172 Molecular & Cellular Proteomics 15.6 Molecular Signatures of Membrane Protein Complexes

1

Skeletal Muscle Isolation 6

Intensity LC/MS Full MS (MS1) Time 2

Non-Washed Microsome Preparation

7

Directed List Generation 3

WGA Enrichment

8

3% Intensity LC-MS/MS directed Time MS (dMS)

15% 4

Sucrose Gradient Fractionation 9

Database Search LCVLHEK Spectrum Mill SHCIAEVEK NECFLSHK Spiked Cysteine Carboxymethylated Tryptic BSA peptides into Samples

5 10

Ion Exchange Data Curation Chromatography Fractionation

FIG.1.Schematic illustration of workflow. Skeletal muscle is dissected (1) and subjected to subcellular fractionation to enrich for membrane proteins (2). Glycoproteins and associated proteins are isolated using wheat-germ agglutinin (WGA) chromatography (3), and are separated by sucrose-gradient fractionation (4). Fractions of interest are spiked with cysteine carboxymethylated tryptic BSA peptides and subjected to ion-exchange chromatography to reduce sample volume; they are eluted using increasing salt concentrations (5). A first LC/MS run is performed on each elution fraction (6), and the peptides identified are used to generate a directed list (7). A directed LC-MS/MS run is then performed to provide quantitative data (8). Proteins are annotated based on queries of the Spectrum Mill database (9). The annotated quantitative data is then curated (10). detecting systematic errors that would skew the relative pro- petting error), and changes (e.g. decreases) in sensitivity of tein quantification between samples. Sources of systematic the mass spectrometer. error might include reagent failure (i.e. corrupted SCX and/or We next determined whether skeletal muscle from WT, C18 spin columns), differences in sample handling (e.g. pi- mdx, and Sgcd-null mice subjected to lectin-affinity chroma-

Molecular & Cellular Proteomics 15.6 2173 Molecular Signatures of Membrane Protein Complexes

A B C57Bl6 Sgcd-null mdx MS 1 2 3 45678910111213

α-D G α-DG

SGCA

β-D G β-DG WT

MS 1 2 3 45678910111213 DMD α-D G

SGCA

SGCA β-D G Sgcd-null

MS12345678910111213 SGCG α-D G

SGCA SGCD β-D G mdx

FIG.2.Disruption of the dystrophin-glycoprotein complex in DGC mutants. A, Sucrose-gradient sedimentation was used to analyze protein complexes in wheat-germ agglutinin-enriched, nonwashed microsomes from WT, Sgcd-null and mdx mice. A 5–30% sucrose gradient was run. Light-to-heavy fractions (1–13) were separated by SDS-PAGE, and the expression of ␣-DG, SGCA, and ␤-DG was detected by immunoblotting. Fractions depicted in red were subjected to proteomic analysis. B, Immunohistochemistry was used to detect DGC subunits in skeletal muscle cryosections from WT, Sgcd-null, and mdx mice. tography and sucrose-gradient fractionation expressed the Sgcd-null mice suggests that the complex was perturbed in DGC at detectable levels, as this was critical to quantitative these contexts (Fig. 2A). Indeed, visualization of DGC sub- proteomic analysis (Fig. 1). Western blot analysis of DGC units using immunohistochemistry showed that the DGC is subunits ␣-DG, SGCA, and ␤-DG in the WT samples revealed localized at the plasma membrane in WT mice (Fig. 2B). that all three proteins were detectable in sucrose fractions Conversely, in mdx and Sgcd-null mice the membrane ex- 6–8 (Fig. 2A, top panel). In the mdx sample, the migration pression of subunits of the DGC is lost, indicating loss of the pattern was broader, with each subunit detectable across DGC as a stable higher MW complex (Fig. 2B). sucrose fractions 4–10 (Fig. 2A, bottom panel). In the Sgcd- Quantitative Proteomic Analysis of Microsomes from WT, null sample, where SGCA was not detectable in any fraction, Sgcd-null, and mdx Skeletal Muscle—Sucrose-gradient frac- ␣-DG was detectable across fractions 5–10, and ␤-DG was tions 4–8 from WT, mdx, and Sgcd-null skeletal muscle were restricted to fractions 3–6 (Fig. 2A, middle panel). Measure- subjected to quantitative mass spectrometry (summarized in ment of sucrose densities per fraction confirmed equal gradi- Fig. 1). dMS analysis of these samples resulted in 460, 563, and ents for each mouse model (supplemental Fig. S1). The dem- 744, respectively, nonredundant protein IDs. As shown in the onstration of co-fractionation of the DGC subunits in the WT Venn diagram in Fig. 3A, 215 proteins were shared among all sample supports the notion that the DGC is normally a stable three samples, 40 between the WT and mdx samples, 75 be- higher MW complex in skeletal muscle. The discordant migra- tween the WT and Sgcd-null samples, and 155 between the tion of DGC subunits through the sucrose gradient in mdx and mdx and Sgcd-null samples. A further 130, 153, and 299 pro-

2174 Molecular & Cellular Proteomics 15.6 Molecular Signatures of Membrane Protein Complexes

A Our results prompted us to implement an unbiased strategy WT Sgcd-null for detecting groups of proteins that co-migrate across the 75 130 299 sucrose-density gradient, with the aim of identifying com-

215 plexes that form in the membranes of skeletal muscle. To this 40 155 end, we applied empirical K-means clustering to the pro- teomic data sets, partitioning them based on their pattern of 153 migration through the sucrose density gradient. This ap- proach identifies putative protein complexes based on the mdx principle that proteins belonging to a complex will behave B 70 similarly, and it resulted in the grouping of 1067 nonredundant IDs from the three genotypes into 14 protein clusters (Fig. 4A). 60 To determine whether each of the 14 protein clusters is 50 linked to specific biological processes that occur in the skel- etal muscle, we uploaded them individually into the GeneGo 40 WT Sgcd-null bioinformatics software program. The top-ranked biological 30 mdx process networks associated with each cluster were identified Percentage (%) 20 (supplemental Table S3). As expected based on the regener-

10 ation processes that are active in muscular dystrophy, the top-ranked process network was development/neuromus- 0 e r n n d cular junction, and was associated with cluster K5 (p ϭ n s o o ic e la t ri t ifi u d a s cleu n m s Ϫ11 u o a N olec las um l 2.88E ). Other highly ranked biological process networks m lex toskele ch l c Membra y o u n ro C U ac omp Mit ndop M C E Retic included cell-adhesion/platelet-endothelium-leukocyte inter- ϭ Ϫ10 FIG.3.Classification of data set. A, A Venn diagram shows the actions for cluster K10 (p 1.46E ), and cell adhesion/cell Ϫ distribution and number of proteins detected in microsomes from WT matrix-interactions for cluster K11 (p ϭ 6.24E 10). Notably, mice and the Sgcd-null and mdx disease models. B, Bar graphs the proteins in clusters K10 and K11 were identified mainly in showing the percentages of the top six classifications for the com- the Sgcd-null and mdx mice, consistent with their up-regula- bined protein expression data per mouse model listed below, along with the group “Unclassified.” For each gene ontology term, the tion in muscular dystrophy. Overall, this bioinformatic analysis percentage refers to the fraction of classified proteins per total num- identified biological process networks (in specific protein ber of proteins (all classes). clusters) that are related to the physiology of skeletal muscle tissue in WT, Sgcd-null, and mdx mice. Our results allowed for the development of a molecular teins were restricted to the WT, mdx, and Sgcd-null skeletal framework to examine known components of the DGC, and muscle samples, respectively. In total, 1067 proteins were iden- possibly identify novel DGC components in skeletal muscle. tified across the WT, Sgcd-null and mdx samples (supplemental To this end, we generated physical protein-protein interaction Table S5). (PPI) maps using the GeneGo software, applying the “shortest We next evaluated the cellular distributions of the identified path” algorithm and taking only direct binding interactions proteins to determine if lectin-affinity chromatography had into account for each of the 14 clusters (supplemental Fig. sufficient power to enrich for proteins in, and associated with, S3). Close inspection of these PPI maps revealed that the the skeletal-muscle membrane. To this end, we uploaded the DGC was part of cluster K5 (supplemental Fig. S5). This proteins identified in the WT, Sgcd-null, and mdx samples into cluster is characterized by the co-migration of proteins in the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt) and sucrose-gradient fraction 7 of the WT sample (Fig. 4B). The searched the Gene Ontology (GO) annotations (38). This analy- DGC components in this cluster contained dystroglycan sis indicated that, for each genotype, more than 50% of the (listed as DAG1), dystrophin (DMD), SGCA, SGCB, SGCD, identified proteins are membrane related (WT 51%, Sgcd-null SGCG, SSPN, SNTA1, and DTNA. Notably, several of these 58%, and mdx 58%). The remaining proteins identified in each DGC components were visibly reduced in the mdx sample, sample were either unclassified or annotated as components of and most were undetectable in the Sgcd-null sample. These the nucleus, macromolecular complexes, cytoskeleton, mito- results were concordant with the Western blot and immuno- chondrion, or endoplasmic reticulum. Interestingly, although the histochemistry results in skeletal muscle between wt, mdx, number of proteins identified in the samples from Sgcd-null and and Sgcd-null mice (Fig. 2). mdx mice was higher than those in WT mice, the overall dis- The comigration of proteins through a sucrose density gradi- tributions (percentages in various membranes within the ent does not provide definitive proof that the proteins interact as cell) were similar (Fig. 3B). These findings confirm that lec- a protein complex, thus we performed a co-immunoprecipita- tin-affinity chromatography was effective in enriching for tion experiment to test if the detergent-solubilized DGC membrane and membrane-associated glycoproteins. formed a macromolecular protein complex in solution. ␣-DG

Molecular & Cellular Proteomics 15.6 2175 Molecular Signatures of Membrane Protein Complexes

A B K5 K1 WT4 WT5 WT6 WT7 WT8 SGCD4 SGCD5 SGCD6 SGCD7 SGCD8 MDX4 MDX5 MDX6 MDX7 MDX8 DMD SGCD SGCA DAG1 K10 DTNA SGCG SGCB K2 SSPN SNTA1 MCM9 ASH1L ITGA4 FHL1 RPL7 ASAP2 BIK CDCA5 SLC39A7 NXN FAM16A SH3BP4 1500032L24Rik KLHL13 K3 K11 INTS1 1110002E22Rik SERPINE1 DNAH3 ARID5A HERC2 PRCC MYZAP ATXN2L GVIN1 RGNEF LIMD2 SNRPA K4 CLIP1 NAV1 RPTN F8 BAZ1A DMPK PLCB2 TPRN K12 RNF2 ABCA15 FAM199X SUPT5H PRKCB RASSF6 MPP1 K5 ZFAT RNF17 CUL7 ZKSCAN16 SERPINB1A MTX1 DYRK2 TEX9 KCTD18 MILL2 CASC5 ZFP462 K6 C3ORF62 HES7 OLFR955 PRDX6 NFKIBL1 K13 PHYHIPL DLG5 K7 S1PR1 TDRD12 PABPC1L TBCB SPRR1A EIF3G APC2 K8 PIK3C2B MTRF1 RELT ADAMTS5 SPG11 RASGRF1 STK3 K14 NXPH1 PALLD K9 CRTC3 ANAPC5 LASS6 HTR2B

FIG.4.K-means clustering. A, Partitioning of the 1067 proteins detected in our screen into 14 clusters using K-means clustering. In all cases, the columns represent sucrose-gradient fractions from muscle taken from WT, Sgcd-null, and mdx mice (n ϭ 5 per genotype). Heatmap intensities represent relative protein expression. B, Expanded view of K-means cluster 5 (K5), which contains known DGC components (red labels).

(IIH6) antibody-coupled Sepharose beads were generated Bovine Serum Albumin (BSA) for the immunoprecipitation ex- alongside negative control beads composed of Ryanodine periments. WGA-enriched rabbit skeletal muscle microsomes Receptor antibody (XA7) (i.e. negative control IgM antibody) or were subjected to immunoprecipitation to test if DGC sub-

2176 Molecular & Cellular Proteomics 15.6 Molecular Signatures of Membrane Protein Complexes

Immunoprecipita on to regulate calcium influx into skeletal muscle, and this regu- lation was abrogated in muscular dystrophy (47). Alpha-neur- on IIH6 XA7 BSA exin, which is present at the neuromuscular junction (NMJ) on on on A Elu and is a known DG ligand (48), also formed an edge to DG. G ash W Void Void W Void Wash Wash Elu Elu Elu Overall, this PPI map demonstrates that the K5 protein cluster includes known components of the DGC, and can serve as a α-DG platform for examining putative DGC-interacting proteins as β-DG candidates for novel muscular dystrophy-causing genes. We manually constructed a force-directed graph of the DMD DGC to facilitate more in-depth quantification of the DGC SGCA proteins in the K5 protein cluster (Fig. 6). For the purposes of this analysis, DG expression is represented as separate ex- SGCB pression of the ␣-DG and ␤-DG subunits. When the WT, SGCG Sgcd-null, and mdx samples were compared, the intensity of DGC components ranged across four orders of magnitude. SGCD For example, whereas in the WT samples all DGC compo- ADBN nents were present in sucrose fractions 6 and 7 (with the intensity for all components being between 1 ϫ 106 and 1 ϫ NNOS 108), in the mdx samples the DMD, SNTA, and SGCG subunits 12345678910 were completely undetectable, and the levels of others were 3 6 FIG.5.Co-immunoprecipitation of DGC subunits. Lane 1, input; reduced (e.g. SSPN in fraction 7 at 1 ϫ 10 versus 1 ϫ 10 ). lanes 2–4, anti-alpha-dystroglycan beads (IIH6); lanes 5–7, anti-ryan- Moreover, in the Sgcd-null sample, only the ␣-DG and ␤-DG odine receptor beads (XA7); lanes 8–10, bovine serum albumin beads subunits of the DGC were detectable, and at levels signifi- ␣ ␤ (BSA). Western blot analysis of -DG, -DG, DMD, SGCA, SGCB, ␤ SGCG, SGCD, ADBN, NNOS. cantly lower than in the WT sample (e.g. -DG in fraction 7 at 1 ϫ 103 versus 1 ϫ 106), corroborating earlier findings indi- units were specifically bound to IIH6 (Fig. 5). As predicted, the cating that the skeletal-muscle DGC requires SGCD expres- DGC subunits specifically bound to IIH6, whereas no DGC sion. Overall, these quantitative proteomic results provide subunits were detectable in the negative control immunopre- strong evidence that the DGC is compromised in mouse cipitations (Fig. 5, compare lane 4 to lanes 7 and 10). models of muscular dystrophy, validating the results from These results allowed for the development of a molecular earlier biochemical and immunohistochemical analyses. framework to examine known components of the DGC, and to Putative Compensatory Cell-adhesion Networks Identified identify novel DGC components in skeletal muscle. To this in Mouse Models of Muscular Dystrophy—Our lack of insight end, we generated a more comprehensive PPI map of the into the regulatory processes that lead to shared pathological DGC-containing K5 cluster using the GeneGo bioinformatics features of DGC-related muscular dystrophies (49) is a signif- software, again applying the shortest path algorithm but in- icant gap, and our comparative proteomic analysis of mem- cluding only those direct interactions that involve maximally brane protein complexes in WT, Sgcd-null, and mdx skeletal two nodes (supplemental Fig. S4). The proteins in cluster K5 muscle provided a unique opportunity to address this. For that were identified in both this and the original analysis are example, K-means clustering revealed that the expression of depicted in blue, whereas the proteins depicted in gray rep- protein clusters K10 and K11 was increased in both the resent “missing links.” Known DGC components are shown mutant versus the WT samples. This preliminary finding with red borders, and proteins that bind directly to them with prompted us to examine in greater depth which protein(s) and green borders. Inspection of this PPI map revealed proteins protein networks might represent biomarkers of disease that are mutated in muscular dystrophies and had not previ- processes that underlie the skeletal muscle pathology in ously been linked to the classical DGC (42, 43). For example, dystrophies related to DGC dysfunction. First we performed plectin1 (PLEC1) is linked to several DGC components, and GeneGo transcriptional network analysis on proteins whose also to MYZAP, mutant forms of which are associated with expression was elevated at least twofold in the Sgcd-null or severe skeletal-muscle dysfunction (44). Mutant forms of mdx versus WT samples. This analysis had the potential to desmin (DES) also cause muscular dystrophy, and our anal- identify factors that globally regulate protein ex- ysis associated this protein with the DGC for the first time. pression in skeletal muscle. Other known DGC-interacting proteins identified include utro- A total of 302 proteins were subjected to this analysis, and the phin (UTRN), growth factor receptor-bound 2 (GRB2), and top-ranked subnetwork centered on NF␬B(Supplemental Table neuronal nitric oxide synthase (NNOS) (45). The PPI map also 4). The NF␬B (RelA-p65)-centric network consisted of 129 included phospholipase C beta2 (PLCB2), which can bind to nodes and 478 edges, including 91 of the 129 nodes identified DTNA or SNTA1 (46, 47). Notably, PLCB2 was recently shown in the proteomic screen (Fig. 7A). The identification of this net-

Molecular & Cellular Proteomics 15.6 2177 Molecular Signatures of Membrane Protein Complexes

α-DG β-DG α-DG β-DG α-DG β-DG α-DG β-DG α-DG β-DG

SGCA DMD SGCA DMD SGCA DMD SGCA DMD SGCA DMD

SGCB DTNA SGCB DTNA SGCB DTNA SGCB DTNA SGCB DTNA WT

SGCG SNTA1 SGCG SNTA1 SGCG SNTA1 SGCG SNTA1 SGCG SNTA1

SGCD SSPN SGCD SSPN SGCD SSPN SGCD SSPN SGCD SSPN

α-DG β-DG α-DG β-DG α-DG β-DG α-DG β-DG α-DG β-DG

SGCA DMD SGCA DMD SGCA DMD SGCA DMD SGCA DMD

SGCB DTNA SGCB DTNA SGCB DTNA SGCB DTNA SGCB DTNA

SGCG SNTA1 SGCG SNTA1 SGCG SNTA1 SGCG SNTA1 SGCG SNTA1 Sgcd-null

SGCD SSPN SGCD SSPN SGCD SSPN SGCD SSPN SGCD SSPN

α-DG β-DG α-DG β-DG α-DG β-DG α-DG β-DG α-DG β-DG

SGCA DMD SGCA DMD SGCA DMD SGCA DMD SGCA DMD

SGCB DTNA SGCB DTNA SGCB DTNA SGCB DTNA SGCB DTNA mdx

SGCG SNTA1 SGCG SNTA1 SGCG SNTA1 SGCG SNTA1 SGCG SNTA1

SGCD SSPN SGCD SSPN SGCD SSPN SGCD SSPN SGCD SSPN

Fr. 4 Fr. 5 Fr. 6 Fr. 7 Fr. 8

Intensity

3 4 5 6 7 8

1x10 1x10 1x10 1x10 1x10 1x10

FIG.6.Quantitative analysis of expression of DGC components. Intensities of expression of DGC components present in cluster K5, for sucrose-gradient fractions 4–8 from the WT, Sgcd-null and mdx mice. work was notable in of two other observations: patients teins in greater depth. Instability of the sarcolemma is a major afflicted with Duchenne muscular dystrophy commonly exhibit pathology of skeletal muscle in DGC-mediated muscular dys- NF␬B-mediated inflammatory morbidity, for which they are trophies, and have the potential to restore cell ad- treated with glucocorticoids; and the recently developed anti- hesion in skeletal muscle with this type of damage. We thus inflammatory steroid VBP15 reduces muscular dystrophy in generated a force-directed graph of the proteins present in mdx mice, and does so by inhibiting the NF␬B pathway, and cluster K11, which contains the majority of the integrins that does not cause the deleterious side effects of glucocorticoid- were identified in fraction 7 of the sucrose-density gradient in based therapeutics that are related to immune function (50). the mdx and Sgcd-null samples, and identified the proteins We next set out to determine if the 91 proteins of the NF␬B regulated by NF␬B (blue border) (Fig. 8A). These PPI maps hub fall into specific clusters, which would establish whether consist of 9, 34, and 14 proteins in the WT, Sgcd-null, and NF␬B-regulated proteins comigrate, i.e. are likely to form mdx samples, respectively. Whereas the WT sample ex- protein complexes, in both the Sgcd-null and mdx samples. pressed only two integrins (itga7 and itgb1), the mdx sample Coordinated regulation of these protein complexes by NF␬B expressed five (itga5, itga6, itga7, itgb1, and itgb2), and the might reflect disease processes that are active in both mouse Sgcd-null sample expressed six (all those expressed in mdx models of muscular dystrophy. A PPI map of the 91 proteins mice and ITGB3). Notably, the expression of ITGA5 and identified 184 edges (Fig. 7B). Forty-eight of the proteins were ITGA6 has been linked to active regeneration of skeletal mus- restricted to clusters K2 (18 proteins), K10 (18 proteins), and cle in muscular dystrophy (51), and both of these integrins K11 (11 proteins), networks whose functions are associated were expressed only in the mutant samples (with the intensity with cell-adhesion/cell-matrix and platelet/endothelium inter- for both components being between 1 ϫ 106 and 1 ϫ 108). actions (Fig. 7B, supplemental Table S3). Closer inspection of Moreover, transgenic overexpression of ITGA7 reduces the these protein clusters showed that integrins, which mediate skeletal-muscle pathology associated with dystrophin defi- adhesion between cells and the extracellular matrix, were ciency (52), and this integrin was highly up-regulated in both up-regulated in both the Sgcd-null and mdx samples (Fig. 7B). mutant samples (both at 1 ϫ 107 versus 1 ϫ 106 in WT). This The finding of integrin up-regulation in the DGC mutant finding supports the notion (52, 53) that ITGA7 can ameliorate mice prompted us to examine the expression of these pro- cell adhesion defects in skeletal muscle in the context of

2178 Molecular & Cellular Proteomics 15.6 Molecular Signatures of Membrane Protein Complexes

A Gnai2 M6pr Tmem2 Tgm2 Pltp Rab11b Clu Tmem87a Pvr Ppap2a Abcg2 Ctsc Stim1 Slc16a1 Cpd Itgb6 Thbd Gdf15 Cd151 Slc12a2 Lamb2 NF−kB Sypl1 Hadha Pdgfrb Cd47 Erap1 p52/p65 Icam1 Ctsb Clec10a Cd68 Tpp1 Cd44 Acadvl Jund Mpo Acta2 Atp8a1 Orai1 Cybb Itga6 Itga5 Brca2 Stx4 Pld3 Kit Slc4a7 Alox5 Vav1 Ncam1 Vamp8 Dst Psap Fn1 Mme NF−kB RelA (p65 Abcb6 Bcam Cyba Gpc1 B2m Ttn Cd37 Slc3a2 Rps27a NF−kB Atp1b3 Lgals3bp subunit) Stab1 Hrg Cd55 Ifitm3 Cd34 Man2b1 Tfrc Lama3 H2−Aa IKK−gamma Pecam1 LUBAC E3 Galectin−3 Myadm TAB1 Asph Il6st Slc2a4 ErbB2 NIK(MAP3K14) Cd74 c−IAP2 E2N(UBC13) TAB2 IKK−beta Cspg4 Hspa5 Slc7a5 IKK (cat)RIPK1 TRADD TNF−alpha Bin3 Vti1b NF−kB Fyn Adam10 PKC−deltap50/p65 PKC−epsilonRBCK1 CD3 zeta TRAF2ErbB3 ADAM17 MAP3K3 Ptprc NFKBIA TNF−R1 ErbB4 CD3 MALT1 NF−kB1 I−kB CARD11 Cd82 (p50) c−IAP1 TAK1(MAP3K7) GRB7 Bcl−10 Egfr

Itgb1 6 8 B Vti1b 9 Asph Gdf15 3 Cd44 Myadm Jund Slc4a7 Ppap2a Abcb6 10 Acadvl Vamp8 Pltp Gnai2 Rab11b Thbd Sypl1 B2m Pecam1

Clec10a Cd82 Mpo Cd68

Icam1 Stab1

2 Cd37 Cspg4 Ctsb Clu Pdgfrb Pvr Psap Egfr Ctsc Ncam1 Tgm2 Man2b1 Cd34 Kit Tpp1 Ptprc Erap1 Pld3 Bcam Acta2 Cd55 Il6st Hrg Stim1 Lgals3bp 11 Fn1 Brca2 Gpc1 Itgb1 Cd151 H2−Aa Ifitm3

Atp1b3 Cybb

Slc3a2 Rps27a Itga5 Slc12a2 Hadha Itga6 Abcg2 Vav1 Atp8a1 Ttn Dst 14 Tmem87a M6pr Slc7a5 Bin3 Cd74 Cd47 Lama3 Mme

Cyba Adam10 Slc2a4 Tmem2 Lamb2 Tfrc Orai1 Alox5 12 Hspa5 Cpd Stx4 Slc16a1 Itgb6 13 Adam17

FIG.7.NF␬B-regulated protein expression. A,NF␬B-responsive proteins that are up-regulated twofold and more in mouse models of muscular dystrophy are shaded in red. Proteins that are associated with the network according to GeneGo are shaded in gray. B, NF␬B-regulated proteins sorted according to K-means cluster.

Molecular & Cellular Proteomics 15.6 2179 Molecular Signatures of Membrane Protein Complexes

A WT SGCD-null mdx Fr.7 Fr.7 Fr.7

Kif15 Kras Tenm4 Atp5i

Atp5j2 Atp5f1 Cd151 Gpc1 Anpep Enpep Anpep Enpep Ptgfrn Ptgfrn Itga5 Pdia3 Itgb1 Dpp4 Igf2r Itga6 Itgb2 Itga7 Nipbl Anpep Cd151 Itgb1 Iigp1 Ifitm3 Enpep Itga7 Hadha Vdac2 Atp5h Susd2 Gpc1 Gmppb Ifitm 3 Itga7 Hadha Igf2r She Itga6 Dpp4 Prrc2a Hadha Itgb2 Prrc2a Lamtor1 Itga5 Itgb1 Dpp4 Ntan1

Arpc1b Tmem4 3 Igf2r Itgb3 Atp5f1

Log(Intensity) 010

B Itga5 Itga6 Cd151 8 8 8

6 *** 6 6

*** 4 *** 4 4 *** *** Fold Change Fold Change Fold Change *** 2 2 2

0 0 0 C57 Sgcd-null C57 mdx C57 Sgcd-null C57 mdx C57 Sgcd-null C57 mdx

FIG.8. Quantitative analysis of expression of components of the integrin-related network. A, The integrin-related PPI map was generated based on cluster K10, using GeneGo. Intensities of expression are depicted for each protein detected in sucrose-gradient fraction 7 for the WT, Sgcd-null and mdx mouse models. B, Transcriptional up-regulation of NF␬B-regulated genes in Sgcd-null and mdx mouse models detected by quantitative RT-PCR. muscular dystrophy. Expression levels of three NF␬B-regu- our study by examining the levels of expression of proteins lated genes (Itga5, Itga6, and Cd151) validated the increase in associated with other neuromuscular disorders, we found that protein abundance (Fig. 8B). A total of 13 distinct integrins 42 proteins of the proteins in our proteomic data set are listed were detected across the three strains of mice (supplemental in the “Gene Table of Neuromuscular Disorders”, a curated Table S6), where the majority of integrins (12 out of 13) database for genes associated with neuromuscular diseases showed either an increase in abundance or were only ex- (54). Notably, hierarchical clustering of the 42 proteins across pressed in the mdx and Sgcd-null models (i.e. ITGAV, ITGA5, the WT, mdx, and Sgcd-null samples revealed that certain ITGA6, ITGAM, ITGB1, ITGB2, ITGB3, ITGA7, ITGAL, ITGB4, neuromuscular disorders group together (Fig. 9). Not surpris- ITGAX, AND ITGB6). Overall, these data support a compen- ingly, given that the mdx and Sgcd-null models of muscular satory cell-adhesion model, whereby NF␬B coordinates the dystrophy were used for the comparative analysis, the DGC- up-regulation of integrin networks in response to perturbed related disorders clustered together. However, other types of DGC function in the context of muscular dystrophy. neuromuscular disorders were also grouped; for example, Proteomic Profiling of Neuromuscular Disorders—The ex- glycogen-storage and related diseases (McArdle, Pompe, and perimental approach presented here initially focused on pro- Danon disease), as well as channel/transporter-related dis- teomic analysis of biochemical fractions from distinct DGC- eases (hypokalaemic periodic paralysis and Charlevoix dis- related mouse models and WT mice. Broadening the scope of ease). These findings were interesting on several levels. First,

2180 Molecular & Cellular Proteomics 15.6 Molecular Signatures of Membrane Protein Complexes WT8 SGCD4 MDX6 MDX8 WT4 WT6 SGCD5 WT7 SGCD7 SGCD8 MDX4 MDX5 MDX7 WT5 SGCD6 Gene Name Symboly Associated Neuromuscular disorder FFourour andand a halfhalf LIMLIM domainsdomains 1 Fhl1Fhl1 Emery-DreifussEmery-Dreifuss muscularmuscular dystrophy,dystrophy, typetype 2 ; RigidRigid spinespine syndromsyndrom DystroglycanDystroglycan Dag1Dag1 RecessiveRecessive limb-girdlelimb-girdle muscularmuscular dystrophydystrophy withwith primaryprimary alpha-dystroglycanalpha-dystroglycan defectdefect Alpha-sarcoglycanAlpha-sarcoglycan SgcaSgca Limb-girdleLimb-girdle muscularmuscular dystrophy,dystrophy, typetype 2D2D Gamma-sarcoglycanGamma-sarcoglycan SgcgSgcg Limb-girdleLimb-girdle muscularmuscular dystrophy,dystrophy, typetype 2C2C Delta-sarcoglycanDelta-sarcoglycan SgcdSgcd Limb-girdleLimb-girdle muscularmuscular dystrophy,dystrophy, typetype 2F2F DystrophinDystrophin DmdDmd DuchenneDuchenne musculamuscular dystrophydystrophy / BeckerBecker muscularmuscular dystrophydystrophy Beta-sarcoglycanBeta-sarcoglycan SgcbSgcb Limb-girdleLimb-girdle muscularmuscular dystrophy,dystrophy, typetype 2E2E SarcospanSarcospan SspnSspn notnot associatedassociated withwith diseasedisease DystrobrevinDystrobrevin alphaalpha DtnaDtna NoncompactionNoncompaction ofof leftleft ventricularventricular myocardiummyocardium withwith congenitalcongenital heartheart defectsdefects Alpha-1-syntrophinAlpha-1-syntrophin Snta1Snta1 LongLong QTQT syndromesyndrome 1212 IsoformIsoform 8 ofof Myotonin-proteinMyotonin-protein kinasekinase DmpkDmpk MyotonicMyotonic dystrophydystrophy (Steinert)(Steinert) Lysosome-associatedLysosome-associated membranemembrane glycoproteinglycoprotein 2 Lamp2Lamp2 DanonDanon diseaseisease Sarcoplasmic/endoplasmicSarcoplasmic/endoplasmic reticulumreticulum calciumcalcium ATPaseATPase 1 Atp2a1Atp2a1 BrodyBrody diseasedisease VeryVery long-chainlong-chain specificspecific acyl-CoAacyl-CoA dehydrogenasedehydrogenase AcadvlAcadvl VLCADVLCAD deficiencydeficiency GlycogenGlycogen phosphorylase,phosphorylase, musclemuscle formform PygmPygm GlycogenGlycogen storagestorage disease,disease, typetype V (McArdle)(McArdle) HeparanHeparan sulfatesulfate proteoglycanproteoglycan 2 Hspg2Hspg2 Swartz-JampelSwartz-Jampel syndromesyndrome LysosomalLysosomal alpha-glucosidasealpha-glucosidase GaaGaa GlycogenGlycogen storagestorage disease,disease, typetype IIII (Pompe)(Pompe) ProteinProtein Col6a3Col6a3 Col6a3Col6a3 UllrichUllrich syndromesyndrome ; BethlemBethlem myopathymyopathy Ras-relatedRas-related proteinprotein -7aRab-7a Rab7aRab7a Carcot-Marie-ToothCarcot-Marie-Tooth neuropathy,neuropathy, typetype 2B2B BiglycanBiglycan BgnBgn notnot associatedassociated withwith diseasedisease TitinTitin TtnTtn Limb-girdleLimb-girdle muscularmuscular dystrophy,dystrophy, typetype 2J2J ; TibialTibial muscularmuscular dystrophydystrophy (Udd)(Udd) CollagenCollagen alpha-1(VI)alpha-1(VI) chainchain Col6a1Col6a1 UllrichUllrich syndromesyndrome ; BethlemBethlem myopathymyopathy 6-phosphofructokinase,6-phosphofructokinase, musclemuscle typetype PfkmPfkm GlycogenGlycogen storagestorage disease,disease, typetype VIIVII (Tarui)(Tarui) L-lactateL-lactate dehydrogenasedehydrogenase LdhaLdha LactateLactate dehydrogenase-Adehydrogenase-A deficdeficiencyiency SodiumSodium channelchannel proteinprotein typetype 4 subunitsubunit alphaalpha Scn4aScn4a HypokalaemicHypokalaemic periodicperiodic paralysis,paralysis, typetype 2 L-typeL-type calciumcalcium cchannelhannel ssubunitubunit aalpha-1Slpha-1S CCacna1sacna1s H Hypokalaemicypokalaemic pperiodiceriodic pparalysis,aralysis, ttypeype 1 Caveolin-3Caveolin-3 Cav3Cav3 Limb-girdleLimb-girdle muscularmuscular dystrophy,dystrophy, typetype 1C1C SoluteSolute carriercarrier familyfamily 1212 membermember 6 Slc12a6Slc12a6 PeripheralPeripheral neuropathyneuropathy andand agenesisagenesis ofof thethe corpuscorpus callosumcallosum (Charlevoix(Charlevoix disease)disease) LamininLaminin subunitsubunit alpha-2alpha-2 Lama2Lama2 CongenitalCongenital muscularmuscular dystdystrophyrophy withwith merosinmerosin deficiencydeficiency TransmembraneTransmembrane proteinprotein 4343 Tmem43Tmem43 LUMALUMA relatedrelated muscularmuscular dystrophydystrophy LamininLaminin subunitsubunit beta-1beta-1 Lamb1Lamb1 notnot associatedassociated withwith diseasedisease LamininLaminin ssubunitubunit ggamma-1amma-1 LLamc1amc1 n notot aassociatedssociated wwithith ddiseaseisease LamininLaminin subunitsubunit beta-2beta-2 Lamb2Lamb2 CongenitalCongenital MyasthenicMyasthenic syndromesyndrome withwith b2-lamininb2- deficiencydeficiency IsoformIsoform Alpha-7X1AAlpha-7X1A ofof IntegrinIntegrin alpha-7alpha-7 Itga7Itga7 CongenitalCongenital muscularmuscular dystrophydystrophy withwith integrinintegrin defectdefect MyelinMyelin pproteinrotein PP00 MMpzpz C Carcarcot-Marie-Toothot-Marie-Tooth neuropathy,neuropathy, typetype 1B1B & 4E4E ; CMTDI,CMTDI, typetype D ; CMT3CMT3 Nesprin-1Nesprin-1 Syne1Syne1 N Nesprin-1esprin-1 rrelatedelated mmuscularuscular ddystrophyystrophy DysferlinDysferlin DysfDysf L Limb-girdleimb-girdle mmuscularuscular ddystrophy,ystrophy, ttypeype 22BB ; DDistalistal rrecessiveecessive mmyopathyyopathy ((Miyoshi)Miyoshi) ATP-bindingATP-binding cassettecassette sub-familysub-family C membermember 9 Abcc9Abcc9 DilatedDilated cardiomyopathy,cardiomyopathy, typetype 1O1O DesminDesmin DesDes MyofibrillarMyofibrillar mmyopathy,yopathy, ddesmin-relatedesmin-related mmyopathyyopathy ; DDilatedilated ccardiomyopathy,ardiomyopathy, ttypeype 11II Nesprin-2Nesprin-2 Syne2Syne2 Nesprin-2Nesprin-2 relatedrelated muscularmuscular dystrophydystrophy Junctophilin-2Junctophilin-2 Jph2Jph2 H Hypertrophicypertrophic ccardiomyopathyardiomyopathy LamininLaminin subunitsubunit alpha-4alpha-4 Lama4Lama4 DilatedDilated cardiomyopathycardiomyopathy duedue toto laminin-alpha4laminin-alpha4 defectdefect

FIG.9.Relationships among proteins associated with neuromuscular disorders. Hierarchical clustering of the relative levels of protein expression per sucrose gradient fraction for WT, Sgcd-null and mdx mice. Proteins are grouped based on their similar migration patterns across mouse models, and similar distributions through the sucrose gradient. they suggest that certain biochemical pathways linked to flow facilitated molecular characterization of the DGC in WT muscular dystrophies that are unrelated to DGC defects are skeletal muscle and the mdx and Sgcd-null mouse models of similarly affected in the mdx and Sgcd-null mouse models. muscular dystrophy, making possible the first comprehen- Second, they indicate that pathways leading to different types sive, comparative proteomic study of the DGC in skeletal of neuromuscular disorders are affected in, and shared by, muscle in mice. various neuromuscular disorders. Therefore, interrogation of This MS analysis of the DGC in skeletal muscle was suc- these proteins in the context of different neuromuscular dis- cessful with respect to several bioanalytical achievements. orders might lead to an understanding of the molecular rela- First, the nonionic detergent digitonin effectively promoted tionships, as well as of some of the clinical similarities and the water solubilization of membrane-bound DGC, facilitating phenotypes associated with certain neuromuscular disorders, downstream biochemical analyses. Second, lectin-affinity although further research into the neuromuscular disease- chromatography made it possible to enrich for detergent- associated pathways will be necessary to validate these solubilized glycosylated proteins. Third, sucrose-density gra- hypotheses. Our findings are also interesting in that they dient centrifugation proved an effective method for sizing suggest that the analysis of other mouse models of neuro- glycoprotein complexes. Fourth, the FASP method facilitated muscular disease in a proteomic comparison like that pre- removal of MS-incompatible analytes (i.e. sucrose, detergent, sented here will expand our knowledge of, and clarify the salts) from sucrose-gradient protein fractions prior to dMS relationships among, proteins that are involved in other neu- (56). Fifth, the dMS acquisition scheme enabled us to acquire romuscular diseases. higher quality MS/MS spectra of targeted peptide ions pres- ent in the tryptic-peptide samples from the complex. As such, DISCUSSION our proteomic workflow facilitated the in-depth MS-based Proteomic Workflow—In this study we demonstrate a label- analysis of the DGC in mouse skeletal muscle tissue. free proteomic workflow for studying the molecular composi- A technical strength of our label-free proteomic workflow tion of membrane protein complexes in mouse skeletal mus- was the use of trypsin-digested, iodoacetic-acid labeled bo- cle, based upon the principles of protein correlation profiling vine serum albumin (BSA) as an internal standard to normalize and related proteomic studies (15, 16, 34, 40, 55). This work- measurements across samples from WT, mdx, and Sgcd-null

Molecular & Cellular Proteomics 15.6 2181 Molecular Signatures of Membrane Protein Complexes skeletal muscle. Previous label-free proteomic studies used brane-related protein complexes that are linked to disease in trypsin-digested iodoacetamide-labeled BSA for this purpose tissues other than skeletal muscle. (57) (58). However, a recent proteomic study in Arabidopsis Validation of DGC Composition and Potential Identification thaliana showed that the intensities of BSA protein ions were of Novel DGC Components—Our discovery by proteomic sufficiently reproducible for protein normalization (58). Our analysis of sucrose-gradient fractions that DGC subunits re- proteomic study builds upon this theme, demonstrating that solved into a single cluster (K5) was validated using a coim- iodoacetic-acid labeled BSA is a superior to iodoacetamide munoprecipitation study and corroborates earlier findings labeled BSA as an internal standard, based upon the fact that suggesting that the DGC is a stable plasma-membrane pro- the majority of proteomic samples are alkylated with iodoac- tein complex (reviewed in (63)). Our analysis revealed comi- etamide. This ready distinction is especially important be- gration between the known DGC components and a subset of cause of the high degree of conservation of serum albumins proteins that had not previously been implicated in the DGC in across species (i.e. Ͼ70% protein identity between WT mice, and reductions in these associations in mouse , monkey, dog, cow, mouse, and rat) (59) The iodo- models of muscular dystrophy. Therefore, these proteins are acetic-acid labeled cysteine-containing BSA peptides harbor candidates as novel DGC-interacting proteins. As protein- a ϩ1 AMU mass shift difference (ϩ58) relative to sample- protein interaction maps are built from published data based derived cysteine containing BSA peptides labeled with iodo- either on actual experiments or literature-based methodolo- acetamide (ϩ57). On a high-resolution mass spectrometer, gies, novel protein-protein interactions cannot be determined. this mass differential is sufficient to distinguish sample-de- However, our analysis provides a powerful enrichment strat- rived BSA peptides from standard BSA peptides. egy to provide a set of candidates which need to be tested for physical binding to the DGC in future experiments. This in- Another advantage of using iodoacetic-acid labeled BSA as cludes the application of stable-isotope dilution mass spec- an internal standard is the fact that trypsin-digested iodo- trometry (SID-MS) methods to validate and monitor the DGC acetic-acid BSA contains 26 tryptic cysteine-containing pep- across different muscular dystrophy models. tides that can be used to normalize quantified proteins across Putative Compensatory Cell Adhesion Pathways in Muscu- samples. Theoretically, these 26 iodoacetic-acid labeled cys- lar Dystrophy Models—A major finding from our study was the teine peptides (restricted to 2ϩ and 3ϩ charged ions) consti- detection of cluster-specific protein complexes in the skeletal tute a library that provides ϳ31% proteomic coverage of the muscle of the mdx and Sgcd-null models of muscular dystro- BSA standard, based upon their ionization in the 400–1250 phy. Our direct comparison of such complexes between WT m/z MS1 window during a dMS experiment. In essence they and mutant mice allowed us to establish if protein complexes function as external standards for the detection of systematic were up- or downregulated in the latter by combining the error that may arise during sample handling (i.e. sample pep- information present in Fig. 4 and supplemental Fig. S3. Inter- tide fraction-SCX spin column, sample processing-C18 spin estingly, there were similarities and differences in clustered column), or changes in sensitivity of the mass spectrometer protein signals between the mdx and Sgcd mice (Fig. 4A). that may arise during the acquisition of peptide ions in MS1 Inspection of clusters K6 (mdx), K9 (mdx), K12 (mdx), and K13 and dMS experiments (e.g. because of ion suppression, prob- (Sgcd) demonstrate an overlap in inflammatory and cell ad- lems with the nano-LC system). Obviously the introduction of hesion protein network processes. Specific proteins repre- protein/peptide standards into an experimental sample is not sentative of these broad protein network processes might without limitations; an excess of standard can induce ion reflect differences in the pathological state of each between suppression and reduce the amount of proteins detected in the mdx and Sgcd muscular dystrophies. For example, asyn- the experimental sample. Empirical studies in our laboratory chronously regenerating microenvironments have been iden- have shown that a 1:75 molar ratio of standard peptide to tified as an underlying driver of fibrosis and failed regeneration sample peptide had no deleterious impact on the identifica- in muscular dystrophies (28). Differences in clustered protein tion and quantification of proteins in complex peptide sam- networks between the mdx and Sgcd mice might reflect dis- ples (unpublished data). An optimal molar ratio of BSA to similarities in asynchronously regeneration between the mus- target protein (1–4 pmol to 5–25 ␮g of protein) was also cular dystrophy models. Future experimentation should be obtained in Arabidopsis thaliana (58). able to confirm this hypothesis between mdx and Sgcd mice. A final advantage of our targeted proteomic workflow is that Furthermore, our bioinformatic analysis showed that the selective-reaction monitoring (SRM) methods combined with NF␬B pathway was significantly up-regulated in both mouse stable-isotope dilution (SID) can be easily incorporated into models for muscular dystrophy, and that the majority of the our label-free proteomic workflow to quantify the DGC, and NF␬B regulated proteins were present in three clusters: K2, also to measure dynamic changes in the DGC under a variety K10, and K11. The fact that cluster K11 exhibited the most of experimental conditions (60–62). We anticipate that it will PPIs suggests that these proteins form an NF␬B-regulated be possible to adapt our label-free proteomic workflow for the protein complex. Our further bioinformatic analysis of the study of changes in the composition of other plasma mem- NF␬B-regulated complex showed that it was enriched for

2182 Molecular & Cellular Proteomics 15.6 Molecular Signatures of Membrane Protein Complexes

several integrins, among which some contribute to the devel- It remains to be determined whether our biochemical work- opment of muscle tissue (ITGA5 and ITGA6) and others link flow is suitable for to the study of other well-characterized the extracellular matrix to the plasma membrane (ITGA7). membrane protein complexes in mammalian systems. More Indeed, recent studies showed that several integrins (i.e. importantly, our analytical framework validated components Itga2, Itga5, Itgb1, and Itgb4) are NF␬B regulated (64–69). of the DGC in skeletal muscle tissue and showed how muta- These findings suggest that integrin-based cellular function tions in the DGC perturbed the composition of the DGC in (i.e. cell adhesion) is important for normal muscle develop- established mouse models of muscular dystrophy. The mo- ment and activity, and that NF␬B-regulated pathways may lecular interrogation of the DGC in skeletal muscle tissue will become activated when muscle homeostasis is disrupted. facilitate the identification of novel components of the DGC, Interestingly, the only pharmacological intervention for provide deeper molecular insights into how drugs influence Duchenne Muscular Dystrophy (DMD) is treatment with glu- the biochemical composition of the DGC, and uncover active cocorticoids, agents that are thought to act by suppressing compensatory pathways and/or mechanisms at the molecular inflammation via inhibition of NF␬B, stabilizing the sarco- and cellular level to influence the skeletal muscle phenotype in lemma, reducing necrosis, and increasing muscle mass (70). muscle dystrophies. Finally, the proteomic findings will be of Thus, glucocorticoid therapy slows the progression of weak- general interest to skeletal muscle biologists, as many of the ness, reduces the development of scoliosis, and delays res- reported proteins may encode novel biomarkers of skeletal piratory insufficiency (71). However, long-term glucocorticoid muscle diseases. treatment in DMD patients can result in detrimental side ef- Acknowledgments—We thank David Venzke and Keith Garringer fects such as the development of Cushingoid syndrome, for technical assistance, and members of the Wright and Campbell weight gain, compromised bone health, and dose-dependent laboratories, as well as Christine Blaumueller, for fruitful discussions. growth retardation (71). Moreover, in mdx mice long-term We also like to thank Christine Miller, Joe Roark, and Steve Fischer glucocorticoid application promotes muscle wasting (72). (Agilent Technologies) for their help in software development and Therefore, despite the benefits of glucocorticoids in treating analysis of MS data. DMD patients, potentially deleterious effects in the long term * This work was supported by a University of Iowa Cardiovascular might limit their therapeutic value. Center Institutional Research Fellowship (5T32HL007121-37) to Recently, a promising novel drug for the treatment of DMD, RT, and American Reinvestment and Recovery Act Grant VBP15, was discovered. Like glucocorticoids, it has beneficial (1RC2NS069521-01) to KPC, Muscular Dystrophy Association Research Grants (157538 and 238219) to KPC, and Paul D. Well- NF␬B-inhibiting effects, but it does not transactivate the glu- stone Muscular Dystrophy Cooperative Research Center Grant cocorticoid receptor (GR) and thus does not reproduce the (1U54NS053672) to KPC. KPC is an investigator of the Howard deleterious side effects of steroids with respect to muscle Hughes Medical Institute. This work was also supported by University repair and strength (50). The beneficial effects of VBP15 may of Iowa start-up funds to MEW. □ result from altered regulation of specific genes, because the S This article contains supplemental material. ‡‡ To whom correspondence should be addressed: Department of transcriptional activities of the glucocorticoid receptor and Molecular Physiology & Biophysics, University of Iowa Carver College ␬ NF B have been described to physically interact (73). We of Medicine, 5-630 Bowen Science Building, 51 Newton Road, Iowa speculate that although glucocorticoids reduce muscle ne- City, IA 52242. Tel.: 319-384-1764; Fax: 319-335-7330; E-mail: crosis, their immunosuppressive effects interfere with muscle [email protected]. regeneration, consistent with the fact that analyses of gluco- The authors declare that they have no competing interests. corticoid-treated mdx mice provided no evidence for in- REFERENCES creased regeneration (70). In contrast, drugs (like VBP15) that lack immunosuppressive effects could inhibit specifically 1. Mokri, B., and Engel, A. G. (1975) Duchenne dystrophy: electron micro- ␬ scopic findings pointing to a basic or early abnormality in the plasma NF B-mediated transcription. Moreover, despite the fact that membrane of the muscle fiber. Neurology 25, 1111–1120 NF␬B signaling normally promotes cell adhesion, the appar- 2. Ibraghimov-Beskrovnaya, O., Ervasti, J. M., Leveille, C. J., Slaughter, C. A., ent compensatory up-regulation of various integrin signaling Sernett, S. W., and Campbell, K. P. (1992) Primary structure of dystro- phin-associated glycoproteins linking dystrophin to the extracellular ma- pathways in the skeletal muscle of mdx and Sgcd-null skeletal trix. Nature 355, 696–702 mice may be sufficient to rescue muscle regeneration in these 3. Hohenester, E., Tisi, D., Talts, J. F., and Timpl, R. (1999) The crystal contexts. Quantitative protein profiling of the integrin net- structure of a laminin G-like module reveals the molecular basis of alpha-dystroglycan binding to , perlecan, and agrin. Mol. Cell 4, works expressed in mdx and Sgcd-null mice in the context of 783–792 glucocorticoid and VBP15 treatment is expected to address 4. Campbell, K. P., and Kahl, S. D. (1989) Association of dystrophin and an these possibilities in the future. integral membrane glycoprotein. Nature 338, 259–262 5. Ozawa, E., Yoshida, M., Suzuki, A., Mizuno, Y., Hagiwara, Y., and Noguchi, This study describes an analytical workflow to study the S. (1995) Dystrophin-associated proteins in muscular dystrophy. Human composition of plasma membrane protein complexes in skel- Mol. Genetics 4, 1711–1716 etal muscle tissue. Under the experimental conditions em- 6. Ohlendieck, K., and Campbell, K. P. (1991) Dystrophin-associated proteins are greatly reduced in skeletal muscle from mdx mice. J. Cell Biol. 115, ployed in this study our proteomic workflow is only amenable 1685–1694 to the study of the DGC expressed in mouse skeletal muscle. 7. Hack, A. A., Lam, M. Y., Cordier, L., Shoturma, D. I., Ly, C. T., Hadhazy,

Molecular & Cellular Proteomics 15.6 2183 Molecular Signatures of Membrane Protein Complexes

M. A., Hadhazy, M. R., Sweeney, H. L., and McNally, E. M. (2000) profiling of skeletal muscle reveals heart-type fatty acid binding protein Differential requirement for individual and dystrophin in the as a candidate biomarker of aerobic capacity. Proteomes 1, 290–308 assembly and function of the dystrophin-glycoprotein complex. J. Cell 27. Holland, A., Dowling, P., Zweyer, M., Swandulla, D., Henry, M., Clynes, M., Sci. 113, 2535–2544 and Ohlendieck, K. (2013) Proteomic profiling of cardiomyopathic tissue 8. Iwata, Y., Nakamura, H., Mizuno, Y., Yoshida, M., Ozawa, E., and from the aged mdx model of Duchenne muscular dystrophy reveals a Shigekawa, M. (1993) Defective association of dystrophin with sarcolem- drastic decrease in laminin, nidogen and . Proteomics 13, mal glycoproteins in the cardiomyopathic hamster heart. FEBS Lett. 329, 2312–2323 227–231 28. Dadgar, S., Wang, Z., Johnston, H., Kesari, A., Nagaraju, K., Chen, Y.-W., 9. Turk, R., Sterrenburg, E., van der Wees, C. G., de Meijer, E. J., de Menezes, Hill, D. A., Partridge, T. A., Giri, M., Freishtat, R. J., Nazarian, J., Xuan, J., R. X., Groh, S., Campbell, K. P., Noguchi, S., van Ommen, G. J., den Wang, Y., and Hoffman, E. P. (2014) Asynchronous remodeling is a driver Dunnen, J. T., and t Hoen, P. A. (2006) Common pathological mecha- of failed regeneration in Duchenne muscular dystrophy. J. Cell Biol. 207, nisms in mouse models for muscular dystrophies. FASEB J. 20, 127–129 139–158 10. Barrera, N. P., Betts, J., You, H., Henderson, R. M., Martin, I. L., Dunn, 29. Murgia, M., Nagaraj, N., Deshmukh, A. S., Zeiler, M., Cancellara, P., Moretti, S. M., and Edwardson, J. M. (2008) Atomic force microscopy reveals the I., Reggiani, C., Schiaffino, S., and Mann, M. (2015) Single muscle fiber stoichiometry and subunit arrangement of the alpha4beta3delta GABA(A) proteomics reveals unexpected mitochondrial specialization. EMBO Re- receptor. Mol. Pharmacol. 73, 960–967 ports 16, 387–395 11. Griffin, N. M., and Schnitzer, J. E. (2011) Overcoming key technological 30. Rudomin, E. L., Carr, S. A., and Jaffe, J. D. (2009) Directed sample inter- challenges in using mass spectrometry for mapping cell surfaces in rogation utilizing an accurate mass exclusion-based data-dependent tissues. Mol. Cell. Proteomics 10, R110.000935 acquisition strategy (AMEx). J. Proteome Res. 8, 3154–3160 12. Whitelegge, J. P., Gundersen, C. B., and Faull, K. F. (1998) Electrospray- 31. Schmidt, A., Claassen, M., and Aebersold, R. (2009) Directed mass spec- ionization mass spectrometry of intact intrinsic membrane proteins. Pro- trometry: towards hypothesis-driven proteomics. Current Opinion Chem. tein Sci. 7, 1423–1430 Biol. 13, 510–517 13. Souda, P., Ryan, C. M., Cramer, W. A., and Whitelegge, J. (2011) Profiling 32. Bulfield, G., Siller, W. G., Wight, P. A., and Moore, K. J. (1984) X chromo- of integral membrane proteins and their post translational modifications some-linked muscular dystrophy (mdx) in the mouse. Proc. Natl. Acad. using high-resolution mass spectrometry. Methods 55, 330–336 Sci. U.S.A. 81, 1189–1192 14. Ohlendieck, K. (2011) Skeletal muscle proteomics: current approaches, 33. Coral-Vazquez, R., Cohn, R. D., Moore, S. A., Hill, J. A., Weiss, R. M., technical challenges and emerging techniques. Skeletal Muscle 1, 6 Davisson, R. L., Straub, V., Barresi, R., Bansal, D., Hrstka, R. F., William- 15. Andersen, J. S., Wilkinson, C. J., Mayor, T., Mortensen, P., Nigg, E. A., and son, R., and Campbell, K. P. (1999) Disruption of the sarcoglycan- Mann, M. (2003) Proteomic characterization of the human by sarcospan complex in vascular : a novel mechanism for protein correlation profiling. Nature 426, 570–574 cardiomyopathy and muscular dystrophy. Cell 98, 465–474 16. Dunkley, T. P., Dupree, P., Watson, R. B., and Lilley, K. S. (2004) The use 34. Dunkley, T. P. J., Watson, R., Griffin, J. L., Dupree, P., and Lilley, K. S. of isotope-coded affinity tags (ICAT) to study organelle proteomes in (2004) Localization of Organelle Proteins by Isotope Tagging (LOPIT). Arabidopsis thaliana. Biochem. Soc. Trans. 32, 520–523 Mol. Cell Proteomics 3, 1128–1134 17. Hojlund, K., Yi, Z., Hwang, H., Bowen, B., Lefort, N., Flynn, C. R., Langlais, 35. Ohlendieck, K., Ervasti, J. M., Snook, J. B., and Campbell, K. P. (1991) P., Weintraub, S. T., and Mandarino, L. J. (2008) Characterization of the Dystrophin-glycoprotein complex is highly enriched in isolated skeletal human skeletal muscle proteome by one-dimensional gel electrophore- muscle sarcolemma. J. Cell Biol. 112, 135–148 sis and HPLC-ESI-MS/MS. Mol. Cell. Proteomics 7, 257–267 36. Duclos, F., Straub, V., Moore, S. A., Venzke, D. P., Hrstka, R. F., Crosbie, 18. Theron, L., Gueugneau, M., Coudy, C., Viala, D., Bijlsma, A., Butler-Browne, R. H., Durbeej, M., Lebakken, C. S., Ettinger, A. J., van der Meulen, J., G., Maier, A., Bechet, D., and Chambon, C. (2014) Label-free quantitative Holt, K. H., Lim, L. E., Sanes, J. R., Davidson, B. L., Faulkner, J. A., protein profiling of vastus lateralis muscle during human aging. Mol. Cell. Williamson, R., and Campbell, K. P. (1998) Progressive muscular dys- Proteomics 13, 283–294 trophy in alpha-sarcoglycan-deficient mice. J. Cell Biol. 142, 1461–1471 19. Rayavarapu, S., Coley, W., Cakir, E., Jahnke, V., Takeda, S., Aoki, Y., 37. Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang, J. T., Ramage, D., Grodish-Dressman, H., Jaiswal, J. K., Hoffman, E. P., Brown, K. J., Amin, N., Schwikowski, B., and Ideker, T. (2003) Cytoscape: a software Hathout, Y., and Nagaraju, K. (2013) Identification of disease specific environment for integrated models of biomolecular interaction networks. pathways using in vivo SILAC proteomics in dystrophin deficient mdx Genome Res. 13, 2498–2504 mouse. Mol. Cell. Proteomics 12, 1061–1073 38. Wang, J., Duncan, D., Shi, Z., and Zhang, B. (2013) WEB-based GEne SeT 20. Gygi, S. P., Rochon, Y., Franza, B. R., and Aebersold, R. (1999) Correlation AnaLysis Toolkit (WebGestalt): update 2013. Nucleic Acids Res. 41, between protein and mRNA abundance in yeast. Mol. Cell. Biol. 19, W77–83 1720–1730 39. Eisen, M. B., Spellman, P. T., Brown, P. O., and Botstein, D. (1998) Cluster 21. Santoni, V., Kieffer, S., Desclaux, D., Masson, F., and Rabilloud, T. (2000) analysis and display of genome-wide expression patterns. Proc. Natl. Membrane proteomics: use of additive main effects with multiplicative Acad. Sci. U.S.A. 95, 14863–14868 interaction model to classify plasma membrane proteins according to 40. Hartman, N. T., Sicilia, F., Lilley, K. S., and Dupree, P. (2007) Proteomic their solubility and electrophoretic properties. Electrophoresis 21, Complex Detection Using Sedimentation. Anal. Chem. 79, 2078–2083 3329–3344 41. Andersen, J. S. (2003) Proteomic characterization of the human centro- 22. Carberry, S., Zweyer, M., Swandulla, D., and Ohlendieck, K. (2014) Com- some by protein correlation profiling. Nature 426, 570–574 parative proteomic analysis of the contractile-protein-depleted fraction 42. Smith, F. J., Eady, R. A., Leigh, I. M., McMillan, J. R., Rugg, E. L., Kelsell, from normal versus dystrophic skeletal muscle. Anal. Biochem. 446, D. P., Bryant, S. P., Spurr, N. K., Geddes, J. F., Kirtschig, G., Milana, G., 108–115 de Bono, A. G., Owaribe, K., Wiche, G., Pulkkinen, L., Uitto, J., McLean, 23. Yoon, J. H., Johnson, E., Xu, R., Martin, L. T., Martin, P. T., and Montanaro, W. H., and Lane, E. B. (1996) Plectin deficiency results in muscular F. (2012) Comparative proteomic profiling of dystroglycan-associated dystrophy with epidermolysis bullosa. Nat. Gen. 13, 450–457 proteins in wild type, mdx, and Galgt2 transgenic mouse skeletal muscle. 43. Cetin, N., Balci-Hayta, B., Gundesli, H., Korkusuz, P., Purali, N., Talim, J. Proteome Res. 11, 4413–4424 B., Tan, E., Selcen, D., Erdem-Ozdamar, S., and Dincer, P. (2013) A 24. Johnson, E. K., Zhang, L., Adams, M. E., Phillips, A., Freitas, M. A., novel desmin mutation leading to autosomal recessive limb-girdle Froehner, S. C., Green-Church, K. B., and Montanaro, F. (2012) Pro- muscular dystrophy: distinct histopathological outcomes compared teomic analysis reveals new cardiac-specific dystrophin-associated pro- with desminopathies. J. Med. Gen. 50, 437–443 teins. PloS one 7, e43515 44. Seeger, T. S., Frank, D., Rohr, C., Will, R., Just, S., Grund, C., Lyon, R., 25. Johnson, E. K., Li, B., Yoon, J. H., Flanigan, K. M., Martin, P. T., Ervasti, J., Luedde, M., Koegl, M., Sheikh, F., Rottbauer, W., Franke, W. W., Katus, and Montanaro, F. (2013) Identification of new dystroglycan complexes H. A., Olson, E. N., and Frey, N. (2010) Myozap, a novel in skeletal muscle. PloS One 8, e73224 protein, activates serum response factor-dependent signaling and is 26. Malik, Z. A., Cobley, J. N., Morton, J. P., Close, G. L., Edwards, B. J., Koch, required to maintain cardiac function in vivo. Circulation Res. 106, L. G., Britton, S. L., and Burniston, J. G. (2013) Label-free LC-MS 880–890

2184 Molecular & Cellular Proteomics 15.6 Molecular Signatures of Membrane Protein Complexes

45. Rando, T. A. (2001) The dystrophin-glycoprotein complex, cellular signal- 60. Keshishian, H., Addona, T., Burgess, M., Mani, D. R., Shi, X., Kuhn, E., ing, and the regulation of cell survival in the muscular dystrophies. Sabatine, M. S., Gerszten, R. E., and Carr, S. A. (2009) Quantification of Muscle Nerve 24, 1575–1594 Cardiovascular Biomarkers in Patient Plasma by Targeted Mass Spec- 46. Lyssand, J. S., Whiting, J. L., Lee, K. S., Kastl, R., Wacker, J. L., Bruchas, trometry and Stable Isotope Dilution. Mol. Cell Proteomics 8, 2339–2349 M. R., Miyatake, M., Langeberg, L. K., Chavkin, C., Scott, J. D., Gardner, 61. Picotti, P., and Aebersold, R. (2012) Selected reaction monitoring-based R. G., Adams, M. E., and Hague, C. (2010) Alpha-dystrobrevin-1 recruits proteomics: workflows, potential, pitfalls and future directions. Nat. alpha-catulin to the alpha1D-adrenergic receptor/dystrophin-associated Methods 9, 555–566 protein complex signalosome. Proc. Natl. Acad. Sci. U.S.A. 107, 62. Ludwig, C., Claassen, M., Schmidt, A., and Aebersold, R. (2012) Estimation 21854–21859 of absolute protein quantities of unlabeled samples by selected reaction 47. Sabourin, J., Harisseh, R., Harnois, T., Magaud, C., Bourmeyster, N., Deliot, monitoring mass spectrometry. Mol. Cell. Proteomics 11, M111 013987 N., and Constantin, B. (2012) Dystrophin/alpha1-syntrophin scaffold reg- 63. Ozawa, E. (2004) The Muscle Fiber Cytoskeleton: The Dystrophin System. ulated PLC/PKC-dependent store-operated calcium entry in myotubes. in Myology (Engel, A. G., and Franzini-Armstrong, C. eds.), 3rd Ed., Cell Calcium 52, 445–456 McGraw-Hill. pp 455–470 48. Sugita, S., Saito, F., Tang, J., Satz, J., Campbell, K., and Sudhof, T. C. 64. Chen, Y. W., Nagaraju, K., Bakay, M., McIntyre, O., Rawat, R., Shi, R., and (2001) A stoichiometric complex of neurexins and dystroglycan in brain. Hoffman, E. P. (2005) Early onset of inflammation and later involvement J. Cell Biol. 154, 435–445 of TGFbeta in Duchenne muscular dystrophy. Neurology 65, 826–834 49. Mercuri, E., and Muntoni, F. (2013) Muscular dystrophies. Lancet 381, 65. Becker-Weimann, S., Xiong, G., Furuta, S., Han, J., Kuhn, I., Akavia, U. D., 845–860 Pe’er, D., Bissell, M. J., and Xu, R. (2013) NFkB disrupts tissue polarity in 50. Heier, C. R., Damsker, J. M., Yu, Q., Dillingham, B. C., Huynh, T., Van der 3D by preventing integration of microenvironmental signals. Oncotarget Meulen, J. H., Sali, A., Miller, B. K., Phadke, A., Scheffer, L., Quinn, J., 4, 2010–2020 Tatem, K., Jordan, S., Dadgar, S., Rodriguez, O. C., Albanese, C., 66. Wang, J. H., Manning, B. J., Wu, Q. D., Blankson, S., Bouchier-Hayes, D., Calhoun, M., Gordish-Dressman, H., Jaiswal, J. K., Connor, E. M., Mc- and Redmond, H. P. (2003) Endotoxin/lipopolysaccharide activates NF- Call, J. M., Hoffman, E. P., Reeves, E. K., and Nagaraju, K. (2013) VBP15, kappa B and enhances tumor cell adhesion and invasion through a beta a novel anti-inflammatory and membrane-stabilizer, improves muscular 1 integrin-dependent mechanism. J. Immunol. 170, 795–804 dystrophy without side effects. EMBO Mol. Med. 5, 1569–1585 67. Yan, B., Chen, G., Saigal, K., Yang, X., Jensen, S. T., Van Waes, C., 51. Gullberg, D., Sjoberg, G., Velling, T., and Sejersen, T. (1995) Analysis of Stoeckert, C. J., and Chen, Z. (2008) Systems biology-defined NF- fibronectin and vitronectin receptors on human fetal skeletal muscle cells kappaB regulons, interacting signal pathways and networks are impli- upon differentiation. Experimental Cell Res. 220, 112–123 cated in the malignant phenotype of head and neck cancer cell lines 52. Burkin, D. J., Wallace, G. Q., Nicol, K. J., Kaufman, D. J., and Kaufman, S. J. differing in p53 status. Genome Biol. 9, R53 (2001) Enhanced expression of the alpha 7 beta 1 integrin reduces 68. Nam, J. M., Ahmed, K. M., Costes, S., Zhang, H., Onodera, Y., Olshen, muscular dystrophy and restores viability in dystrophic mice. J. Cell Biol. A. B., Hatanaka, K. C., Kinoshita, R., Ishikawa, M., Sabe, H., Shirato, H., 152, 1207–1218 and Park, C. C. (2013) beta1-Integrin via NF-kappaB signaling is essen- 53. Burkin, D. J., Wallace, G. Q., Milner, D. J., Chaney, E. J., Mulligan, J. A., and Kaufman, S. J. (2005) Transgenic expression of {alpha}7{beta}1 integrin tial for acquisition of invasiveness in a model of radiation treated in situ maintains muscle integrity, increases regenerative capacity, promotes breast cancer. Breast Cancer Res. 15, R60 hypertrophy, and reduces cardiomyopathy in dystrophic mice. Am. J. 69. Ahmed, K. M., Zhang, H., and Park, C. C. (2013) NF-kappaB regulates Pathol. 166, 253–263 radioresistance mediated by beta1-integrin in three-dimensional culture 54. Kaplan, J. C., and Hamroun, D. (2013) The 2014 version of the gene table of breast cancer cells. Cancer Res. 73, 3737–3748 of monogenic neuromuscular disorders (nuclear genome). Neuromuscu- 70. Fisher, I., Abraham, D., Bouri, K., Hoffman, E. P., Muntoni, F., and Morgan, lar Disorders 23, 1081–1111 J. (2005) Prednisolone-induced changes in dystrophic skeletal muscle. 55. Foster, L. J., de Hoog, C. L., Zhang, Y., Zhang, Y., Xie, X., Mootha, V. K., FASEB J. 19, 834–836 and Mann, M. (2006) A mammalian organelle map by protein correlation 71. Ricotti, V., Ridout, D. A., Scott, E., Quinlivan, R., Robb, S. A., Manzur, A. Y., profiling. Cell 125, 187–199 Muntoni, F., and NorthStar Clinical, N. (2013) Long-term benefits and 56. Wisniewski, J. R., Zougman, A., Nagaraj, N., and Mann, M. (2009) Universal adverse effects of intermittent versus daily glucocorticoids in boys with sample preparation method for proteome analysis. Nat. Methods 6, Duchenne muscular dystrophy. J. Neurol. Neurosurg. Psychiatry 84, 359–362 698–705 57. Brambilla, F., Resta, D., Isak, I., Zanotti, M., and Arnoldi, A. (2009) A 72. Sali, A., Guerron, A. D., Gordish-Dressman, H., Spurney, C. F., Iantorno, M., label-free internal standard method for the differential analysis of bioac- Hoffman, E. P., and Nagaraju, K. (2012) Glucocorticoid-treated mice are tive lupin proteins using nano HPLC-Chip coupled with Ion Trap mass an inappropriate positive control for long-term preclinical studies in the spectrometry. Proteomics 9, 272–286 mdx mouse. PloS One 7, e34204 58. Zauber, H., Schuler, V., and Schulze, W. (2013) Systematic evaluation of 73. Rao, N. A., McCalman, M. T., Moulos, P., Francoijs, K. J., Chatziioannou, reference protein normalization in proteomic experiments. Frontiers Plant A., Kolisis, F. N., Alexis, M. N., Mitsiou, D. J., and Stunnenberg, H. G. Sci. 4, 25 (2011) Coactivation of GR and NFKB alters the repertoire of their binding 59. Little, S. (2010) Orthologs of Human Serum Albumin. Davidson College sites and target genes. Genome Res. 21, 1404–1416

Molecular & Cellular Proteomics 15.6 2185 Supplemental Figure/Table Legends

Supplemental Figure 1: Sucrose-gradient fractionation. Graph depiction of the sucrose

percentage in each fraction that demonstrated comparable gradients between mouse

models.

Supplemental Figure 2: Silver staining of sucrose-gradient fractions. Protein distribution and abundance prior to further fractionation by ion-exchange chromatography are shown.

Supplemental Figure 3: PPI maps for each K-means cluster. Protein-protein interactions based on direct binding were generated for each cluster, using the GeneGo software.

Supplemental Figure 4: Protein-protein interaction (PPI) map of DGC-associated proteins. Proteins present in cluster K5 are shown in blue, and among these the DGC- related proteins have a red border; others are shown in gray. Proteins directly associated with the DGC (first neighbors) have a green border.

Supplemental Figure 5: Protein-protein interactions of DGC-associated proteins.

Protein-protein interactions present in cluster K5 are shown.

Supplemental Table 1: Total intensities of 14 iodoacetic-labeled BSA peptides used for protein sample normalization.

1

Supplemental Table 2: dMS protocol vs. DDA protocol.

Supplemental Table 3: Process network analysis of K-means clusters. Statistically

significant process networks were determined for each cluster, using the GeneGo software. In each case the ratio represents the number of proteins in the specified cluster over the total number of proteins in the network

Supplemental Table 4: Transcriptional network analysis on proteins whose expression was elevated at least 2-fold in the Sgcd-null or mdx vs WT samples.

Supplemental Table 5: List of identified proteins across the WT, Sgcd-null and mdx samples.

Supplemental Table 6: List of integrins detected in WT, Sgcd-null and mdx samples.

Supplemental Table 7: List of all peptide sequence identifications and post-translational modifications.

2

35

30

25

20

15

10 1 2 3 4 5 6 7 8 9 10 11 12 13 WT 11.8 12.7 14.1 15.5 16.9 18.5 19.9 21.5 22.9 24.5 26.7 28.7 31.3 Sgcd 11.4 12.9 14 15.5 16.9 18.5 19.9 21.7 23.1 25.3 26.8 28.9 31.1 mdx 11.4 12.5 14 15.5 16.9 18.5 20.1 21.7 23.5 25.4 27.5 29.7 31.5

Supplemental Figure 1

Cluster K1

Decorin COL6A1

Keratin 6A 17

PLU−1 RPAP1

Cluster K2

HBB Hemopexin gp130

HP Saposin A TSC22D4 HBD Adult hemoglobin TGOLN2 Saposin D Alpha1−globin ACTB Acid sphingomyelinase Clusterin CD166 HPRG ASAH Actin cytoplasmic 2 HYOU1 Saposin C MTCH1 EMAPII ATP5B P4HB

Calpastatin CLN2 (Tripeptidyl−peptidase I) BCAM 14−3−3 beta/alpha ATP5A Actin muscle 14−3−3 sigma 14−3−3 zeta/delta SYNJ1 HRC Triadin Calsequestrin

HSP12A NCAM1 SRRM1 14−3−3 gamma FRY Proepithelin MCAM Fibrillin 1 RHG7 14−3−3 theta PP1G regulatory subunit GYS1 GRK6 EDNRB SAE1

Rictor 14−3−3 epsilon ELKS C1 inhibitor RhoGAP4 CXCR4 Actin 14−3−3 eta Apotransferrin Transferrin 14−3−3 COL4A2 Collagen XIX MEX3D Fibronectin STIM1 C10orf79 Lca5 ACTA2 COL4A1 ERC2/CAST UBE1DC1

COL4A6 90K Collagen IV AMPL COL4A3 Galectin−1 COL4A4 COL4A5 Cathepsin B SOD3 (EC−SOD) IL8RB

Supplemental Figure 3A Cluster K3 Thioredoxin HLAB HLA−C

LAMP2

OSP94 TFEB HLA−B

LAMP1 HLA−Cw3

MITF HLA−A GLAST1/EAAT1

CD82 PEDF (serpinF1)

TAF1 DHX16 U5−200KD CD9 RAB43 Fc epsilon RI gamma CD44 NAR1

BCDIN3

ATAD3A TAF15 SLIT2 TFIID PPCKC

Ca−ATPase1

Cluster K4

MyHC MELC Kallikrein 5 Corneodesmosin

MRLC

MYH9 Kallikrein 7 MYH10 CHD7 BAF180

p120GAP Beta−2 adrenergic receptor ALDOA

Kallikrein 6 (Neurosin) II Neurobromin Beta−1 adrenergic receptor PI3K cat class IB (p110−gamma) ZNF238 IQSEC1 SRP68

Kallikrein 13 TEX15

CGI−99 TIP120A

Supplemental Figure 3B Cluster K5

Delta−sarcoglycan HTR1A Palladin PKC−alpha HES7 Sarcospan Epsilon−sarcoglycan RASGRF1

Gamma−sarcoglycan Beta−sarcoglycan S1P1 receptor RING2 Nucleoredoxin

Metaxin 1

Ash1 Alpha−sarcoglycan Dystrophin HTR1D HTR1B PRCC

Dystrobrevin alpha cPKC (conventional) Dystroglycan APCL ANAPC5 ANAPC2 APC/hCDH1 complex Sororin

PLC−beta2 HTR2B PLA2 Syntrophin A PKC−mu NAV1 MPP1 DLG5(P−dlg) PRK1 CLIP170 CLIP115 PKC Coagulation factor VIII STK3 PLC−beta1 eIF3S4 PKC−zeta NSGPeroxidase FHL1 (SLIM1) PAI1

RASSF6 HTR2A PKC−beta DYRK2 RPL7

Cluster K6

VAMP8 ITGAV Reticulon 4 BVES

VTI1B VAMP2 ENTPD2−alpha CD36

Cyr61 AP−1 B3GN1 JunD

ppl2/A4 MLL4 SR−BI Calumenin

Supplemental Figure 3C Cluster K7

ATP1B2 CD147 SLC16A3

Cluster K8

A2M receptor Impas 1 CRCM OST48 gamma−Secretase complex

MAP3K2 (MEKK2) COX II COX Vb Endophilin B2

PDF ABCA1 Tropomyosin−1

Cluster K9

SCN4B RHAG PP2C alpha Na(v) beta 1 SLC4A1 Aquaporin 1 CACNA1S

gamma−Secretase complex EPB41 SHP−2 Rab−33B C730036D15Rik AP−1 SMEK1 FOP ALK−2 Caveolin−3

ASCT2 (SLC1A5) SCN4A(SkM1) LPP1 CREB3 iASPP U5−102 kDa CREB4 CREB1

TD−60 BATF Csk Tensin 3 C1orf77 Nucleolin Synaptopodin SLC12A6 MutYH L−type Ca(II) channel, alpha 1C 1 subunit Tho2 BAF250A

CAP−E NFAT−90 Ku80 ATF−2

ACM2 FALZ CLN5 ATF−4 ATF−2/c−Jun PRIM1 ATF−1 Rad50

Supplemental Figure 3D Cluster K10

Syntaxin 12 VAMP5

PLTP SNAP−23

Ephrin−A receptors Perlecan

PTK7 Ephrin−A receptor 4 VAMP7

Ephexin CSPG4 (NG2) Csp

Collagen VI AKT1

C1qa Fc gamma RI Ephrin−A receptor 3 VEGFR−1 Ephrin−A receptor 2 PDGF−R−alpha RDS C1qRp

Peripherin AKT(PKB) AL7A1 C1q ACSL1

CACNA2D2 Thrombospondin 4 −1 AHNAK VLDLR CYB5R3 PRNP PDGF−R−beta Clathrin Stabilin−1 PKD1−like G−protein alpha−i2 TGM2 GAA

Liprin−beta 2 HGF receptor (Met) NACA MYOG PERM GLCM CD45 Collagen XII CARM1 HFE2 Rab−7 DEP−1 NPM3 Clathrin heavy chain TSLC1 Beta−2−microglobulin FCGRT C20orf3 Cathepsin D EGFR TXNIP (VDUP1) c−Kit

GNPTA FAT10 PDGF receptor DDX9 NELL1

CLOCK ICAM1 SLTM G−protein alpha−i3 SHPS−1 CAPZ beta

Fam47e

Thrombomodulin sICAM1

Protein C CPB2

Supplemental Figure 3E Cluster K11

ITGAM alpha−V/beta−3 integrin

Laminin 2 ITGB2 VAV−1 Laminin 4

Laminin 8 ITGA7

alpha−7/beta−1 integrin Laminin 6 PDIA3 alpha−6/beta−1 integrin

ITGB1 LAMA2 Laminin 1 Laminin 10 ITGB3 PTPR−zeta ITGA6 LAMB1 CD151 alpha−4/beta−1 integrin alpha−5/beta−1 integrin IGF−2 receptor alpha−3/beta−1 integrin ITGA5 DPP4 PTGFRN K−RAS

gp91−phox CD13 HADHA alpha−M/beta−2 integrin FAT1 alpha−2/beta−1 integrin

ENPEP Dopamine transporter FLJ20625 AMOTL2 GCC1 RPS27A VDAC 2

Ubiquitin RP40

KIAA1211 Ferroportin 1 PPAL p130 ARPC1B ATP5F1 ATPK SF3B1

ATP5H ATP5I SF3B2

Supplemental Figure 3F Cluster K12

Rhbdf2 ADAM17 PKC

SUR1

SPAG1 Adenylate cyclase type IX Kir6.2 PTPR−beta SUR2 ROS1

PKC−alpha

ITGB4 SLC7A6 hnRNP D−like PKC−epsilon SLC7A5 Phospholemman VAPA

gamma−Secretase complex Fascin Syntaxin 4 PKC−gamma NHE7 MTMR15 Adenylate cyclase type VIII Syntaxin 8

Calpain small subunit HBS1L Sortilin CRACM1 PKC−delta REEP2

Calmodulin GLUT4

PKC−beta Nephrocystin−5 PKC−zeta ITFG3 PA2G6 Utrophin ACAA1 Albumin RalA

Neprilysin PKC−eta Pitx2 ITGB2

TBCD4 DSCAML1 MCT1 (SLC16A1) SEC31L1 IRS−4 CDK5RAP2 MLL1 (HRX) ABCA5 26S proteasome (19S regulator)

MICAL3 MKLP1

APEH IST1 SUR K(+) channel, subfamily J Peg10

AIRE TLR3 TOLLIP

alpha−L/beta−2 integrin SKAP55 A−Raf−1 PKC−theta

Supplemental Figure 3G Cluster K13 SSTR4 FZD4

Plexin B2 TM9SF2

SSTR1 Chapsyn−110 Alpha− 2 Alpha−actinin 1

Plexin A1 FZD7

AATF (Che−1) PRK1 PLA2G10

NNP−1 RRP1B (KIAA0179) ITGAD ITGAX Kendrin

RAR−beta/RXR−alpha CPD PLA2R1 PLA2G1B RARbeta

C2orf18 SYH CRM1 RARgamma RAR−alpha/RXR−beta APOA1 Podocalyxin−like 1 ADAM10 B2

BPAG1 LAMG1 p70 S6 kinase1 TER GRP75 PAR2 MNS1 CD63 Nidogen RAR Laminin 6 TCFbeta1 RARalpha C5AR MPP9 PAR1 RAR−alpha/RXR−alpha SYNE2 LAMB2 SEC61 complex PTPR−alpha TMS−1 PYPAF5 EGLN3 AKAP6 ARMC4 SEC62 PCGF1 IGF−2 SLC22A23 Homer 2 NPY4R SSTR2 Homer Desmin MA2A2 IP3R1 ABCC2 WDR37 IP3 receptor G−protein beta/gamma HSP70 PLA2 RBM15B ACOX2 BST2 ENT1 Homer 3 LDLR APOE AUF1 RSG5 G−protein beta−2

GLUT1 V2 receptor Delta−type opioid receptor TfR1 SHP−2 alpha−X/beta−2 integrin LDL HSC70 ITGB2 Alpha−actinin 4 Alpha−actinin alpha−M/beta−2 integrin GRP78

LPA1 receptor EDD Neogenin CD74 SLC25A3 alpha−L/beta−2 integrin SEC61 alpha LPA3 receptor IL8RA SSTR5 SLC25A4 BRD1 alpha−D/beta−2 integrin BGLR

HDL proteins

PA2G6 SLC25A11

UGT1A4 Cytochrome B5 CD36L2 CXCR4

SR−A1 HGK(MAP4K4) Mu−type opioid receptorG−protein beta SLC25A6 SLC25A5 FPRL1 PPAPDC2

APOD

DLL4 NOTCH2 IP3R2 MPP2

Cluster K14

TSGA10 Titin

SLC7A8 SLC3A2

ATP1B1 ATP1A2 Supplemental Figure 3H Supplemental Figure 4 Delta−sarcoglycan

Sarcospan Epsilon−sarcoglycan

Gamma−sarcoglycan Beta−sarcoglycan

Alpha−sarcoglycan Dystrophin

Dystrobrevin alpha cPKC (conventional) Dystroglycan

PLC−beta2

Syntrophin A PKC−mu NAV1

PKC Coagulation factor VIII

FHL1 (SLIM1) PAI1

PKC−zeta NSGPeroxidase

DYRK2 RPL7

HTR1A PKC−alpha

PLC−beta1 eIF3S4

S1P1 receptor HTR2A PKC−beta

HTR1D HTR1B

Palladin HES7 APC/hCDH1 complex Sororin RASGRF1

RING2 Nucleoredoxin HTR2B PLA2

Metaxin 1 PRK1 CLIP170 CLIP115 Ash1 PRCC

MPP1 DLG5(P−dlg) APCL ANAPC5 ANAPC2

STK3

RASSF6