Proteomic analysis reveals that wildtype and alanine-expanded nuclear poly(A)-binding protein exhibit differential interactions in skeletal muscle

Oculopharyngeal muscular dystrophy (OPMD) is a late-onset, primarily autosomal dominant disease caused by a short GCN expansion in the PABPN1 (polyadenylate-binding protein nuclear 1) gene that results in an alanine expansion at the N terminus of the PABPN1 protein. Expression of alanine-expanded PABPN1 is linked to the formation of nuclear aggregates in tissues from individuals with OPMD. However, as with other nuclear aggregate-associated diseases, controversy exists over whether these aggregates are the direct cause of pathology. An emerging hypothesis is that a loss of PABPN1 function and/or aberrant protein interactions contribute to pathology in OPMD. Here, we present the first global proteomic analysis of the protein interactions of WT and alanine-expanded PABPN1 in skeletal muscle tissue. These data provide both insight into the function of PABPN1 in muscle and evidence that the alanine expansion alters the protein–protein interactions of PABPN1. We extended this analysis to demonstrate altered complex formation with and loss of function of TDP-43 (TAR DNA-binding protein 43), which we show interacts with alanine-expanded but not WT PABPN1. The results from our study support a model where altered protein interactions with alanine-expanded PABPN1 that lead to loss or gain of function could contribute to pathology in OPMD.

Polyalanine expansion diseases are caused by an abnormal number of GCN trinucleotide repeats that give rise to protein variants with expanded alanine tracts (1,2). Oculopharyngeal muscular dystrophy (OPMD), 4 a late-onset muscle disease, is one such polyalanine expansion disorder that results from expansion of GCN trinucleotide repeats in the gene encoding the nuclear poly(A)-binding protein, PABPN1 (3). The N-terminal domain of PABPN1 contains a 10-alanine tract of unknown function that directly follows the initial methionine. In individuals with OPMD, this 10-alanine tract is expanded to 11-18 alanine residues with the most common disease variant containing a 13-alanine tract (3,4). OPMD is commonly inherited in an autosomal dominant manner, so only a single modestly expanded copy of PABPN1 is sufficient to confer pathology in a subset of skeletal muscles (3). How such a modest change in a single copy of the ubiquitously expressed PABPN1 gene causes muscle-specific pathology is poorly understood.
Like other polyalanine expansion diseases, OPMD is characterized by the formation of insoluble protein aggregates (5). In OPMD, these aggregates are found in the nucleus and contain PABPN1, polyadenylated RNA, and other RNA-binding proteins (6,7). Whether the formation of these PABPN1 aggregates causes toxicity and cell death or is a protective mechanism is unclear (8). However, the presence of nuclear PABPN1 aggregates in unaffected muscles and neurons (9 -12) argues against a model based solely on aggregate-mediated toxicity. In fact, recent studies suggest that the pathogenic mechanism of OPMD is related to sequestration of PABPN1, other proteins, and RNAs in nuclear aggregates, thus decreasing the functional pools of these important molecules (7,13,14). Given that PABPN1 protein levels in muscle are already low (15), sequestration into aggregates or interaction with alanine-expanded PABPN1 may decrease available PABPN1 below some threshold required for normal muscle function (15,16). If alanine expansion also impairs normal PABPN1 function, this could compound any defects associated with decreased PABPN1 availability and exacerbate pathology. However, few studies have probed how alanine expansion affects the function of the PABPN1 protein in muscle tissue in vivo.
As is true for many RNA-binding proteins (17), PABPN1 performs multiple critical functions in post-transcriptional regulation of gene expression (18). The canonical function of PABPN1 is enhancing the processivity of poly(A) polymerase during poly(A) tail addition to nascent RNAs (19). However, PABPN1 also plays important roles in polyadenylation and cleavage sequence selection (20,21), nuclear export (22,23), and the pioneer round of translation in the cytoplasm (24). Considering that alanine domains have been implicated in protein-protein interactions (25,26), alanine expansion could affect interactions between PABPN1 and other proteins important for mediating these processes. For example, alanine-expanded PABPN1 could make abnormal protein contacts that lead to either a loss or gain of function. Alternatively, alanineexpanded PABPN1 could fail to interact with a subset of proteins critical for the proper function of PABPN1. A previous report demonstrated increased binding of alanine-expanded human PABPN1 to heat-shock proteins and arginine methyltransferases in lysates of C2C12 myoblasts and myotubes, supporting a model where alanine-expansion alters protein interactions with PABPN1 (27).
There is limited knowledge about PABPN1 functional interactions in skeletal muscle (28,29). Characterizing and comparing the protein-protein interactions of WT and alanine-expanded PABPN1 in skeletal muscle tissue could provide valuable insight into the muscle-specific functions of PABPN1 while also identifying pathways most susceptible to dysfunction in OPMD. Here, we present the first global comparison of the protein interactions of WT and alanine-expanded PABPN1 expressed in skeletal muscle tissue. We detected an increased number of proteins associated with the alanine-expanded PABPN1 compared with WT PABPN1. Among the proteins that interact preferentially with alanine-expanded PABPN1 is the ALS-associated protein TDP-43, which forms aggregates in OPMD (30), associates with PABPN1 in neurons (31) and has recently been linked to key functions in regenerating muscle (32). The present study reveals that soluble TDP-43 is associated with higher-molecular-weight complexes and that TDP-43 is prone to oxidative stress-mediated aggregation in primary myoblasts that express alanine-expanded PABPN1. We also demonstrate alternative splicing of known TDP-43 targets, which suggests a loss of TDP-43 function. In addition to TDP-43, we identify other putative PABPN1 interactors that could be predisposed to dysfunction via a similar mechanism. This study provides both a comprehensive list of PABPN1-interacting proteins in skeletal muscle and insight into how these interactions are altered for alanine-expanded PABPN1. Our analysis also provides proof of principle that aberrant interactions with alanine-expanded PABPN1 can lead to changes in downstream cellular functions.

Electroporation can be used to express epitope-tagged PABPN1 in skeletal muscle
Levels of PABPN1 protein are very low in skeletal muscle (15), making biochemical studies of PABPN1 protein interac-tions in this disease-relevant tissue challenging (29). Although several mouse models exist that express PABPN1, they either express native levels of alanine-expanded PABPN1 (33) or do not equally overexpress WT and alanine-expanded PABPN1 (34,35). Thus, we used electroporation into skeletal muscle to allow direct comparison of the proteins that interact with WT PABPN1 and alanine-expanded PABPN1 in skeletal muscle.
To confirm the proper localization of exogenously expressed PABPN1, plasmids expressing GFP-tagged WT or alanine-expanded PABPN1 were electroporated into gastrocnemius muscle. As shown in Fig. 1C, GFP signal co-localizes with the DAPI signal at the position of muscle cell nuclei in muscle sections confirming the expected nuclear localization of both Ala-10 and Ala-17 PABPN1 in electroporated muscle.

Differential protein interactions with WT and alanine-expanded PABPN1
To identify PABPN1-interacting proteins, soluble FLAGtagged Ala-10 and Ala-17 PABPN1 were immunoprecipitated from the electroporated skeletal muscle tissue, and bound samples were analyzed by MS as illustrated in the schematic shown in Fig. 1A. We first confirmed comparable immunoprecipitation of Ala-10 and Ala-17 PABPN1 (Table S1 and see Fig. 4B) and then analyzed the co-immunoprecipitating proteins. Mass spectrometry revealed a total of 381 proteins that were enriched in the FLAG-immunoprecipitations from muscle lysates expressing FLAG-tagged Ala-10 or Ala-17 PABPN1 but not from muscle lysates electroporated with the control pcDNA3 plasmid. PABPN1-interacting proteins were defined as those with a 2-fold or greater increase in FLAG immunoprecipitations in samples expressing FLAG-Ala-10 or FLAG-Ala-17 PABPN1 relative to control muscles (Table S1). A threshold of 2-fold or greater difference in the number of peptides co-immunoprecipitating with Ala-10 versus Ala-17 PABPN1 was used to define proteins as preferentially interacting with Ala-10 or Ala-17 PABPN1. Notably, the C-terminal domain of PABPN1 is required for RNA binding (37), and there is evidence that Ala-10 and Ala-17 PABPN1 interact to the same extent with RNA (38). Thus, these differential interactions are not likely to reflect RNA-dependent interactions.

Differential protein interactions with PABPN1 in muscle
were detected only in the Ala-10 PABPN1 immunoprecipitation, whereas 129 proteins were found only in the Ala-17 PABPN1 immunoprecipitation (Table S1). These results reveal that more proteins interact with Ala-17 PABPN1 than Ala-10 PABPN1, which is consistent with the propensity for alanineexpanded PABPN1 to bind to other proteins with higher affinity than WT PABPN1 (27).
Gene Ontology (GO) analysis was performed on the three groups of proteins: those that immunoprecipitated similarly with both Ala-10 and Ala-17 PABPN1 (Fig. 2B); those that immunoprecipitated preferentially with Ala-10 PABPN1 (Fig.  2C); and those that immunoprecipitated preferentially with Ala-17 PABPN1 (Fig. 2D). Interacting proteins were grouped based on biological process, cellular component, and molecular function with respect to Z score. Regulation of RNA stability and mRNA processing are two of the top biological processes identified for the proteins interacting equally with Ala-10 and Ala-17 PABPN1, whereas regulation of transcription and regulation of RNA stability are the top biological processes for proteins interacting more with Ala-10 and Ala-17 PABPN1, respectively. Several proteins previously identified as PABPN1 interactors including PABPC1, MATR3, and SKIP/SNW (28,29) were identified in these immunoprecipitation experiments. The identification of known interactors and of other regulators of RNA metabolism provides further confidence in the validity of this proteomic analysis.
The co-precipitating proteins that belong to each GO term for Ala-10 PABPN1 interactors (Ala-10 Ͼ Ala-17) (Fig. 2C) or Ala-17 PABPN1 interactors (Ala-17 Ͼ Ala-10) (Fig. 2D) are shown in Fig. 3, panel A (Ala-10 PABPN1) and panel B (Ala-17 PABPN1) as circular graphs, with the colors representing the GO categories and the names of the proteins listed around the circumference. Of the 49 proteins that immunoprecipitated preferentially with Ala-10 PABPN1, SKIP/SNW (Fig. 3A), one is a transcriptional regulator important for skeletal muscle differentiation and is a known interactor of PABPN1 (28). The proteins identified that interact preferentially with alanine-expanded PABPN1 are more numerous and include MATR3 (Fig.  3B), which was recently characterized as a PABPN1 interactor in skeletal muscle (29). Numerous RNA-binding proteins including hnRNPC, hnRNPM, ELAVL1, TDP-43, and the RNA helicases DHX9 and DDX39B also preferentially immunoprecipitated with alanine-expanded PABPN1.
To validate the MS results, we confirmed the interaction of PABPN1 with hnRNPUL2, which immunoprecipitated similarly with Ala-10 and Ala-17 PABPN1, as well as the previously reported interaction with SKIP/SNW1, which immunoprecipitated preferentially with Ala-10 PABPN1. An antibody specific A, a schematic outlining the experimental approach to identify proteins that associate with PABPN1. For each plasmid, n ϭ 2 hind limb (tibialis anterior and gastrocnemius) muscles from two mice were analyzed. B, lysates from muscles electroporated with plasmids expressing FLAG-tagged WT (Ala-10) PABPN1 or alanine-expanded (Ala-17) PABPN1 or a control plasmid were resolved by SDS-PAGE and analyzed by immunoblotting with either an anti-FLAG antibody (top two blots) or using an antibody that specifically detects alanine-expanded PABPN1 (36) (bottom two blots). HSP90 serves as a loading control. Shown is a representative image from n ϭ 2 immunoblots using lysate from four electroporated mice. C, limb muscle (gastrocnemius) was electroporated with plasmids expressing WT (Ala-10) GFP-PABPN1 or alanine-expanded (Ala-17) GFP-PABPN1. Muscles were collected and sectioned 6 days after electroporation and analyzed by microscopy for GFP localization to detect GFP-PABPN1. DAPI and laminin staining mark the nuclei and muscle basal lamina, respectively.

Differential protein interactions with PABPN1 in muscle
for PABPN1 (22) or control IgG was used to perform immunoprecipitations from control mouse muscle lysates that express only endogenous PABPN1. Immunoblot analysis of input and bound samples demonstrates that both hnRNPUL2 and SKIP/ SNW immunoprecipitate with endogenous PABPN1 from skeletal muscle lysate (Fig. 4A).
To validate differential interactions with Ala-10 and Ala-17 PABPN1, WT and alanine-expanded FLAG-tagged PABPN1 were immunoprecipitated from electroporated skeletal muscle lysate using FLAG antibody. Immunoblot analysis with FLAG antibody shows that equal amounts of Ala-10 and Ala-17 PABPN1 are immunoprecipitated from the muscle lysate (Fig.  4B). The bound samples were then immunoblotted for PABPC1, which immunoprecipitated equally with Ala-10 and Ala-17 PABPN1, and TDP-43, which immunoprecipitated only with Ala-17 PABPN1 (Fig. 3). Immunoblot analysis shown in Fig. 4B confirms that PABPC1 interacts similarly with both Ala-10 and Ala-17 PABPN1, whereas TDP-43 is enriched only in the immunoprecipitation of Ala-17 PABPN1. These results are consistent with results obtained from mass spectrometric analysis that revealed differential interactions between Ala-10 and Ala-17 PABPN1.
Most previous studies have focused on proteins that interact with alanine-expanded PABPN1 in the context of intranuclear aggregates (13,14,27). The studies described here were performed utilizing only soluble cellular fractions, so identified Ala-17 PABPN1 interactions do not represent aggregated proteins. However, they may be components of higher-molecularweight complexes. To assess whether Ala-17 PABPN1 exists in larger protein complexes than Ala-10 PABPN1, lysates from electroporated muscles were resolved on a density gradient followed by immunoblotting for FLAG-tagged PABPN1. As shown in Fig. 4C, Ala-17 PABPN1 migrates in heavier fractions than Ala-10 PABPN1, which is consistent with the finding that Ala-17 PABPN1 interacts with multiple proteins and thus can form complexes with higher molecular weights than Ala-10 PABPN1.

TDP-43 is present in high-molecular-weight complexes that contain alanine-expanded PABPN1
To better understand how aberrant protein interactions with Ala-17 PABPN1 may be linked to pathology, we sought to further characterize the interaction between Ala-17 PABPN1 and TDP-43, which immunoprecipitated only with Ala-17 PABPN1 (Figs. 3B and 4B). TDP-43 is an RNA-binding protein that is linked to the pathogenesis of ALS and myopathies, which is prone to aggregate formation in the context of neuromuscular disease (39 -41). Furthermore, recent work shows that TDP-43 plays a critical role in muscle formation and forms amyloid-like structures during myogenesis (32). Given that Ala-17 PABPN1 forms high-molecular-weight complexes as shown by altered migration on a glycerol gradient (Fig. 4C), we performed a series of experiments to examine the characteristics of TDP-43 in the presence of Ala-17 PABPN1.
To assess the solubility of TDP-43 in the presence of Ala-17 PABPN1, we exploited a recently published knock-in mouse model of OPMD (33). These mice genocopy OPMD patients with one allele encoding Ala-10 PABPN1 and one encoding Ala-17 PABPN1 under the control of the native Pabpn1 promoter. As with a number of RNA-binding proteins, the levels of  A, to examine interaction with endogenous PABPN1, skeletal muscle (gastrocnemius) lysate was incubated with control IgG or an antibody against PABPN1 (22). Input and bound fractions were resolved by SDS-PAGE and immunoblotted with anti-PABPN1, anti-hnRNPUL2, or anti-SKIP/SNW antibody. Shown is a representative image from n ϭ 2 immunoblots using lysate from four electroporated mice. B, to examine differential interactions with WT and alanine-expanded PABPN1, lysates obtained from muscles electroporated with plasmids expressing WT (Ala-10) or alanine-expanded (Ala-17) PABPN1 were incubated with anti-FLAG antibody conjugated beads. Bound fractions were resolved by SDS-PAGE followed by immunoblotting with anti-FLAG, anti-PABPC1, and anti-TDP-43 antibodies. Shown is a representative image from n ϭ 2 immunoblots using lysate from four electroporated mice. C, lysates obtained from muscles electroporated with plasmids expressing FLAG-tagged WT (Ala-10) or alanine-expanded (Ala-17) PABPN1 were analyzed by sedimentation through 10 -30% glycerol gradients. Alternate gradient fractions from top to bottom were loaded from left to right together with an input lane and resolved by SDS-PAGE. Immunoblotting with an anti-FLAG antibody was used to detect WT (Ala-10) PABPN1 or alanine-expanded (Ala-17) PABPN1 in the fractions. Shown is a representative immunoblot from three separate fractionation experiments. The box highlights Ala-17 PABPN1 present in high-molecular-weight complexes.

Differential protein interactions with PABPN1 in muscle
TDP-43 are low in mature muscle but are increased in isolated primary myoblasts (42), which can be differentiated into myotubes in vitro (43). Thus, we examined TDP-43 in both primary myoblasts and myotubes prepared from control WT Ala-10/ Ala-10 PABPN1 mice and the OPMD model Ala-17/Ala-10 PABPN1 mice.
Primary myoblasts isolated from WT mice (Ala-10/Ala-10) or OPMD mice (Ala-17/Ala-10) were differentiated into myotubes and lysed in a gentle lysis buffer followed by centrifugation. Immunoblot analysis of the soluble and pellet fractions shows that more TDP-43 is present in the pellet fraction obtained from myotubes of the OPMD model (Ala-17/Ala-10) as compared with pellet fraction obtained from WT myotubes (Ala-10/Ala-10) (Fig. 5A). The soluble fractions from WT and OPMD model myotubes were also loaded and resolved on 10 -30% glycerol density gradients. Immunoblot analysis of fractions collected from top to bottom of the density gradients shows that TDP-43 in lysate from OPMD model (Ala-17/Ala-10) myotubes migrates in heavier fractions when compared with lysate prepared from WT (Ala-10/Ala-10) myoblasts (Fig.  5B). These results imply that soluble TDP-43 exists in larger protein complexes in myotubes expressing Ala-17 PABPN1 than in control myotubes.
TDP-43 is regulated by redox-mediated cysteine oxidation that results in high-molecular-weight (HMW) species, which are prone to aggregation (44). Given the mitochondrial dysfunction found in the OPMD mouse model (33), the presence of TDP-43 in pellet fractions may be related to redox-mediated aggregation. To determine whether TDP-43 is more susceptible to redox-mediated aggregation in cells that express Ala-17 PABPN1 compared with control, myoblasts from OPMD Ala-17/Ala-10 and WT Ala-10/Ala-10 mice were treated with hydrogen peroxide (ϩ) or left untreated (Ϫ) prior to lysis in the absence of reducing agents. As shown in Fig. 5C, immunoblot analysis reveals that there is HMW TDP-43 in the pellet fraction from Ala-17 PABPN-expressing cells prior to treatment with hydrogen peroxide (Ϫ). After treatment with hydrogen peroxide, even more HMW TDP-43 was detected in the pellet fraction from OPMD model cells expressing Ala-17 PABPN1. HMW TDP-43 was not detected in the pellet fraction in control cells that express Ala-10/Ala-10 PABPN1 in either the absence (Ϫ) or the presence (ϩ) of hydrogen peroxide treatment. This result demonstrates that TDP-43 is more prone to redox-mediated aggregation in cells obtained from the OPMD Ala-17/ Ala-10 mouse model compared with WT muscle cells.
The presence of TDP-43 in HMW species in primary muscle cells isolated from an OPMD Ala-17/Ala-10 mouse model could suggest that TDP-43 is present in PABPN1-containing intranuclear aggregates associated with OPMD. To determine whether TDP-43 associates with PABPN1 Ala-17 aggregates, we employed a transgenic Ala-17-overexpressing mouse model of OPMD (Ala-17.1), in which PABPN1 forms readily detectable aggregates in skeletal muscle (34). These aggregates are functionally defined as KCl-insoluble and are thus detectable after KCl washout of soluble proteins (7). Muscle sections obtained from WT mice or the Ala-17 PABPN1-overexpressing mice (Ala-17.1) were treated with salt (ϩKCl) to wash away soluble proteins and reveal aggregates or left untreated (ϪKCl) and then immunostained using an antibody to detect TDP-43. Immunofluorescence staining reveals nuclear TDP-43 staining in the KCl-treated muscle sections (Fig. 5D), indicating that TDP-43 can accumulate in nuclear aggregates that occur in muscle overexpressing alanine-expanded PABPN1.

TDP-43 protein dysfunction in OPMD model cells
The presence of TDP-43 in higher-molecular-weight and insoluble complexes may render it unavailable to perform normal functions related to RNA splicing. To assess whether TDP-43 interaction with Ala-17 PABPN1 impairs TDP-43 function, we evaluated the alternative splicing of known TDP-43 RNA targets (45)(46)(47)(48). If TDP-43 function is impaired by association with alanine-expanded PABPN1, these alternative splicing events might show changes comparable with the loss of TDP-43 function. For this analysis, RNA from WT Ala-10/Ala-10 and OPMD Ala-17/Ala-10 myoblasts was reversetranscribed into cDNA, followed by RT-PCR to detect alternative splicing of the previously characterized TDP-43 targets Mef2D, Sort1, and Dnajc5 (45)(46)(47). As shown in Fig. 6 (A and B), we detect increased inclusion of cassette exons in the Mef2D (exon 8) and Sort1 (exon 17b) transcripts but not Dnajc5 (exon 4) in myoblasts obtained from the OPMD Ala-17/Ala-10 mouse model as compared with WT Ala-10/Ala-10 myoblasts. Exon inclusion in Sort1 was also quantified using quantitative RT-PCR with primers specific for the exon 17b-included and exon 17b-excluded transcripts. Results from quantitative RT-PCR analysis confirm the increased inclusion of the Sort1 17b cassette exon in myoblasts isolated from the OPMD Ala-17/Ala-10 mouse model (Fig. 6C). These data suggest that TDP-43 function is impaired in muscle cells expressing Ala-17 PABPN1.
To extend this analysis of the consequences of TDP-43 highmolecular-weight complex formation in Ala-17 PABPN1-expressing cells, we sought to understand the functional implications of altered Sort1 splicing. SORT1 is a sorting receptor expressed on the cell surface, endoplasmic reticulum, and Golgi apparatus, which is involved in the transport of a wide variety of proteins including the glucose transporter GLUT4 and progranulin (PGRN) (49 -52). TDP-43 represses the inclusion of Sort1 exon 17b, suggesting that SORT1 function could be dysregulated under conditions where functional levels of TDP-43 are decreased.

Differential protein interactions with PABPN1 in muscle
evidence that SORT1-mediated import is responsible for enhanced uptake of N-TAP PGRN in OPMD Ala-17/Ala-10 myotubes (Fig. 7D). The increased uptake of both glucose and N-TAP PGRN suggests that the inclusion of the cassette exon in Sort1 affects the function of the SORT1 protein in muscle cells that express Ala-17 PABPN1.

Discussion
Characterizing and comparing the protein-protein interactions of Ala-10 and Ala-17 PABPN1 in muscle is critical to understand the various functions of PABPN1 while identifying pathways most susceptible to dysfunction. Ideally, affinity purification of endogenous WT or alanine-expanded PABPN1 from skeletal muscle followed by MS would identify the co-immunoprecipitating protein interactors of PABPN1. However, existing disease models of OPMD either have variable degrees of overexpression of the WT and alanine-expanded proteins (34) or express low endogenous levels of PABPN1 in skeletal muscle (15,33), which hinders such a proteomics approach. To circumvent these technical challenges, we exploited several different mouse models. To achieve comparable expression of WT and alanine-expanded PABPN1, we employed electroporated mouse skeletal muscles with plasmids expressing similar levels of FLAG-tagged Ala-10 or Ala-17 PABPN1. Immunoprecipitation of FLAG-PABPN1 from these electroporated muscles followed by MS allowed for the first comparison of the protein interactions of the WT and OPMD-associated alanine-expanded PABPN1 in skeletal muscle tissue. We then used several additional mouse models and experimental conditions to validate and explore the functional consequences of the interactions identified. We validated interactions with WT PABPN1 using control mouse skeletal muscle (Fig. 4A). We also examined the possible consequences of TDP-43 interactions with alanine-expanded PABPN1 by employing two additional OPMD mouse models. To assess whether TDP-43 is detected in aggregates in an OPMD mouse model, we used a transgenic mouse model (Ala-17.1 PABPN1 (34)) that overexpresses Ala-17 PABPN1 in skeletal muscle (Fig. 5D). This model was employed for this experiment because insoluble aggregates are readily detectable (34). We extended our analysis of TDP-43 function to a mouse model where alanine-expanded PABPN1 is expressed from the endogenous Pabpn1 locus (33). This model genocopies OPMD patients who have one WT and one mutant allele of PABPN1. This model was selected because it most closely reflects the situation that would occur in muscle tissue from individuals with OPMD. Thus, each model was carefully selected to provide insight into the PABPN1 protein-protein interactions in skeletal muscle as well as to identify potential functional consequences that could occur when alanine-expanded PABPN1 is expressed in disease.
Co-immunoprecipitation followed by MS identified three discrete groups of PABPN1 protein interactors. The first group immunoprecipitated equally with Ala-10 and Ala-17 PABPN1, whereas the second showed preferential interaction with Ala-10 PABPN1 and potentially represents proteins that the OPMD-associated Ala-17 PABPN1 cannot access. The third group of proteins was much larger and showed preferential interaction with Ala-17. For the group of proteins that interacts  Sort1 (exon 17b), and Dnajc5 (exon 4) using RNA isolated from WT Ala-10/Ala-10 (A10/A10) or OPMD Ala-17/Ala-10 (A17/A10) primary mouse myoblasts. Gapdh serves as a loading control. The upper and lower bands are the exon-included (ϩ) and exon-excluded (Ϫ) PCR products, respectively. B, densitometric quantification of the exon-included and exon-excluded RT-PCR products from A is presented as a ratio of included/excluded PCR products. Shown is mean plus or minus standard deviation for n ϭ 3; * indicates p Ͻ 0.05. C, quantitative RT-PCR analysis of Sort1 alternative splicing in WT (A10/A10) or OPMD (A17/A10) primary mouse myoblasts. The steady-state level of Sort1 RNA with exon 17b included or excluded was quantified by quantitative (q)RT-PCR using variant-specific primers and is represented as the Sort1 included/excluded ratio normalized to Gapdh. The data shown represent mean plus or minus standard deviation for n ϭ 16. * indicates p Ͻ 0.05.

Differential protein interactions with PABPN1 in muscle
similarly with WT and alanine-expanded PABPN1, we cannot rule out the possibility that some of these interactions could be RNA-mediated. However, we employed conditions for the immunoprecipitation where we have not detected a difference upon RNase treatment (29).
The increased number of proteins that immunoprecipitate with Ala-17 PABPN1 is consistent with model that alanine expansion increases the affinity of PABPN1 for other proteins (56). Given that immunoprecipitations were performed on soluble lysate from muscle tissue expressing Ala-17 PABPN1, the increased protein interactions with Ala-17 PABPN1 do not represent the intranuclear aggregates detected in the muscles of individuals with OPMD and OPMD mouse models. However, these interactions may represent high-molecular-weight complexes that could seed the formation of PABPN1-containing aggregates. Therefore, the proteins that immunoprecipitated preferentially with Ala-17 PABPN1 could be most susceptible to sequestration in OPMD-associated aggregates. Indeed, a previously published study suggested the formation of preinclusion-like structures in cells expressing alanine-expanded PABPN1, but the study did not report the components of these structures or the mechanistic implications of their formation (57).
The proteins that interact equally with both Ala-10 and Ala-17 PABPN1 include poly(A)-binding proteins (PABPC1, PABPC4, and PAIP1), hnRNPs (HNRNPUL2, Q, and H1/H2), and other critical RNA-binding proteins, including EIF4A, AUF1, and NPM1. In addition to RNA-binding proteins, the methyltransferase PRMT1 was also identified as a PABPN1 interactor. We previously identified PABPC1, PABPC4, and measured with a plate reader as a reporter for glucose uptake. Shown is mean plus or minus standard deviation for n ϭ 3 experiments. * indicates p Ͻ 0.05. B, WT (Ala-10/Ala-10) and OPMD (Ala-17/Ala-10) myotubes were cultured in the presence or absence of recombinant N-TAP human progranulin (PGRN) for 16 h. Cell lysates were resolved by SDS-PAGE followed by immunoblotting for PGRN for three independent samples for each condition analyzed. C, band intensities for PGRN uptake from B were quantified and normalized to total protein (Ponceau S staining). Shown is mean plus or minus standard deviation for n ϭ 4 experiments. * indicates p Ͻ 0.05. D, to control for the specificity of PGRN uptake, WT (Ala-10/Ala-10) and OPMD (Ala-17/Ala-10) myotubes were cultured in the presence or absence of recombinant N-terminally TAP-tagged (N-TAP) or C-terminally TAP-tagged (C-TAP) human PGRN for 16 h. Cell lysates were resolved by SDS-PAGE followed by immunoblotting for PGRN. HSP90 serves as a loading control.

Differential protein interactions with PABPN1 in muscle
HNRNPQ as protein interactors of PABPN1 in skeletal muscle (29). Furthermore, previous reports demonstrated that PABPN1 interacts with arginine methyltransferases (27,56), and arginine methylation regulates PABPN1 function (58). The identification of established interactors in these proteomic data sets provides validation to support the novel protein interactions identified. These newly identified PABPN1 protein interactors could not only provide insight into the currently known functions of PABPN1 but also uncover new functions of PABPN1 in muscle cells.
The largest group of PABPN1 interactors includes proteins that interact preferentially with the OPMD-associated Ala-17 PABPN1. The interactors of Ala-17 PABPN1 fall under a wide range of functional and biological classifications and could represent aberrant protein-protein interactions with the alanineexpanded PABPN1. These proteins and the cellular processes they regulate could be most susceptible to loss of function caused by sequestration in high-molecular-weight complexes or intranuclear aggregates in OPMD. Alternatively, they could mediate some deleterious gained function when in complex with alanine-expanded PABPN1. The Ala-17 PABPN1 interactors include numerous proteins that are involved in post-transcriptional regulation of RNA including DHX9 (59), CSDE1 (60), ELAVL1 (HuR) (61,62), and RENT1 (63,64). DHX9 is a nuclear RNA helicase that processes dsRNA and circular RNA (59,65). DHX9 also associates with proteins such as FUS, which is linked to aggregation in amyotrophic lateral sclerosis (ALS) (66,67). We also detected a number of mitochondrial ribosomal proteins that preferentially interact with Ala-17 PABPN1. Given the high abundance of mitochondrial ribosomal proteins in skeletal muscle, these proteins are often considered contaminants in immunoprecipitation experiments. However, mitochondrial dysfunction has been detected in patient tissues and animal models of OPMD and other repeat expansion disorders (33, 68 -70). Polyalanine stretches directly associate with mitochondrial proteins in mouse brain and muscle lysate (71) and are associated with apoptosis resulting from mitochondrial dysfunction (72). Additionally, dipeptide repeats expressed from C9ORF72 repeat expansions associate with mitochondrial ribosome subunits and compromise mitochondrial function (73). In addition to these expanded proteins, mutations in genes encoding other disease-linked proteins such as TDP-43 are also associated with aberrant mitochondrial localization of mutant proteins and mitochondrial dysfunction (74,75).
Among the proteins identified to interact more with Ala-17 PABPN1 is the ALS-associated protein TDP-43 (76,77), which interacts with PABPN1 in mammalian cells (31,78). TDP-43 was recently identified as a component of amyloid-like myogranules that are localized near the sarcomeres in the cytosol of nascent myotubes in culture and myofibers in vivo (32). These myo-granules contain TDP-43 bound to mRNAs encoding critical myogenic factors including embryonic myosin heavy chain, thus highlighting the importance of TDP-43 in skeletal muscle formation. We detected soluble TDP-43 in high-molecularweight fractions in glycerol gradient fractionation experiments and insoluble TDP-43 in muscle cells expressing Ala-17 PABPN1. These high-molecular-weight complexes may repre-sent increased formation or persistence of myo-granules in Ala-17 PABPN1-expressing cells, early formation of insoluble nuclear TDP-43 aggregates that have been detected in mice overexpressing Ala-17 PABPN1, or an alternative high-molecular-weight complex. However, any of these scenarios have the potential to sequester TDP-43 and impair its normal functions. Indeed, we detected altered splicing of the known TDP-43 targets Mef2d and Sort1. The changes we detected in both Sort1 and Mef2D splicing are comparable with changes that are detected when TDP-43 is depleted by siRNA (45,48), supporting the idea that TDP-43 function is decreased in muscle cells that express alanine-expanded PABPN1.
TDP-43 binds a large complement of coding and noncoding RNAs (45,79) and is a common component of aggregates in various neuromuscular diseases (30). TDP-43 interacts and aggregates with MATR3 (80, 81), which was identified as a PABPN1 interactor here and in our previous study (29). Furthermore, TDP-43 aggregation is found in skeletal muscles in some cases of sporadic and familial ALS (40), as well as in aggregates associated with inclusion body myositis and OPMD (30,41,82). Importantly, PABPN1 interacts with and ameliorates the toxicity of aggregation-prone TDP-43 variants in multiple model systems (31). In that study, expression of PABPN1 had minimal effect on endogenous TDP-43, but alanine-expanded PABPN1 was not tested. Our study suggests that alanine-expanded PABPN1 could have the opposite effect of WT PABPN1 by aberrantly binding to and sequestering functional TDP-43.
The group of proteins that interacts preferentially with Ala-10 PABPN1 when compared with Ala-17 PABPN1 includes proteins involved in transcription such as GTF2F1 (83), SAFB1 (84), and SKIP/SNW (85,86). SAFB1 interacts with FUS (Fused-in-sarcoma) and MATR3, a PABPN1 interactor (29), to regulate splicing and transcription (87). SKIP/SNW interacts with PABPN1 to regulate expression of genes important for myogenesis (28). The identification of proteins that preferentially interact with WT PABPN1 suggests that alanine-expanded PABPN1 may not retain all of the functions of the WT protein. Thus, alanine expansion in PABPN1 can not only lead to a gain-of-function in the form of abnormal protein-protein contacts but also result in a loss-of-function by precluding critical protein interactions.
Although the role of protein aggregation in polyalanine and polyglutamine expansion disorders is widely accepted, it is unclear whether a loss or gain of protein function underlies disease pathogenesis. Pathology could arise from a combination of these effects, and the combined effects could be specific to certain tissues, providing some insight into the tissue specificity of disease. With respect to the different mechanisms, critical cellular proteins can be sequestered through specific or nonspecific interactions with aggregation prone disease-associated proteins such as PABPN1, leading to a loss of function. The aggregation prone proteins might also be unable to make appropriate protein interactions essential for their function. This study both identifies PABPN1-interacting proteins in muscle, a disease-relevant tissue, and sheds light on the complex mechanisms that can contribute to cellular dysfunction as a result of altered interactions with alanine-expanded protein.

Electroporation
All animals were housed and handled according Association for Assessment and Accreditation of Laboratory Animal Care guidelines and were approved by the Emory University Institutional Animal Care and Use Committee. Age-and sex-matched mice were anesthetized with an intraperitoneal injection of 91 mg/kg ketamine and 9.1 mg/kg xylazine. Tibialis anterior or gastrocnemius muscles were injected with 30 l of 0.4 unit/ microliter hyaluronidase solution in 0.9% NaCl with a 30-gauge needle. After 2 h, the mice were reanesthetized and injected with 30 l of plasmid DNA (1 g/microliter). The limbs to be electroporated were coated in ultrasound jelly and electroporated (ten 20-ms pulses of 175 V/cm with 480-ms intervals between pulses) using 7-mm platinum tweezer electrodes connected to an ECM 830 in vivo electroporator (BTX Harvard Apparatus). For each plasmid type, two animals were electroporated, and both tibialis anterior and gastrocnemius muscles were electroporated for each animal. The mice were euthanized 7 days following electroporation using a CO 2 chamber as per institutional animal care and use committee guidelines. Electroporated muscles were collected and flash frozen in cryofreezing medium (Sakura OCT-TissueTek) for sectioning with a cryostat (Leica Biosystems) or homogenized for immunoblot/ immunoprecipitation analysis

Primary myoblast (satellite cell) isolation
Primary myoblasts were isolated from hindlimb muscles using the MACS satellite cell isolation kit (Miltenyi Biotec) according to the manufacturer's instructions. Briefly, dissected hindlimb muscles were rinsed in PBS, cleaned and then minced using razor blades in DMEM. Minced muscles were treated with 0.1% Pronase with slow stirring for 1 h at 37°C followed by trituration in DMEM (ϩ10% FBS, 100 units/ml penicillin G, 100 mg/ml streptomycin). Triturated muscle solution was passed through a Steriflip 100-m vacuum filter and centrifuged at 900 ϫ g for 5 min. The resulting pellet was washed in PBS and resuspended in ACK lysing buffer for 2 min at room temperature to lyse red blood cells. At the end of the incubation, samples were centrifuged at 900 ϫ g for 5 min, and the pellet was resuspended with the addition of a satellite cell isolation kit followed by a 15-min incubation on ice. MACS buffer was added, and samples were applied to LS columns in a MACS magnetic field. Flowthrough containing satellite cells was collected, centrifuged at 900 ϫ g for 5 min, and resuspended in growth medium (Ham's F10, 20% FBS, 5 ng/ml basic fibroblast growth factor, 100 units/ml penicillin G, 100 mg/ml streptomycin) on collagen-coated (bovine collagen I; Gibco) dishes.

Immunoprecipitation
Primary myoblasts or electroporated muscles were homogenized in immunoprecipitation (IP) buffer (50 mM HEPES, 125 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, 1 mM DTT, complete mini protease inhibitor mixture (Roche)). Homogenate was further lysed by sonication on ice five times at 0.5% output for 10 s and then passed through a 27-gauge syringe five times. Insoluble components were removed by centrifugation at 13,000 rpm for 15 min. The resulting supernatant was collected, and concentration of soluble proteins was determined using a bicinchoninic acid assay. 10% of lysate was removed to be used as the input sample, and M2 FLAG magnetic beads (Sigma) for IP from electroporated muscle or PABPN1 antibody (22) and control rabbit IgG for IP of endogenous PABPN1 were added to soluble cell lysates. The samples were incubated on an endover-end rotator at 4°C overnight. For endogenous PABPN1 immunoprecipitation, protein G magnetic beads (Dynabeads; Invitrogen) were added to the lysate for 1 h and incubated endover-end at room temperature. The beads were then magnetized and washed three times with cold IP buffer. The washed beads were either submitted for mass spectrometric analysis or used for immunoblotting. To immunoblot IP samples, the proteins were eluted with reducing sample buffer (250 mM Tris-HCl, 500 mM DTT, 10% SDS, 0.5% bromphenol blue, 50% glycerol).

Immunoblotting
Cell lysates were prepared in radioimmunoprecipitation assay buffer with cOmplete protease inhibitor tablet (Roche Applied Science). Protein concentrations were measured using bicinchoninic acid assay, and equal total protein amounts were boiled in reducing sample buffer and resolved on 4 -20% Criterion TGX polyacrylamide gels (Bio-Rad). Separated proteins were electrophoretically transferred to nitrocellulose membranes and incubated in blocking buffer (5% nonfat dry milk in TBS-Tween (0.1%)) for 1 h. Blocked blots were incubated overnight at 4°C in primary antibody diluted in blocking buffer. Primary antibodies were detected using species-specific secondary antibodies conjugated to horseradish peroxidase plus incubation in enhanced chemiluminescence substrate (ECL, Sigma). Chemiluminescent signals were detected by exposing blots to autoradiography film (Daigger) or a Bio-Rad ChemiDoc imaging system. Primary antibodies used in this study:

Liquid chromatography and tandem MS
We performed on-bead MS to identify PABPN1-interacting proteins. Lysate from muscles electroporated with each plasmid were pooled prior to immunoprecipitation to generate one Differential protein interactions with PABPN1 in muscle sample per plasmid. Following wash steps from the immunoprecipitation, Dynabeads from control and FLAG-PABPN1 immunoprecipitates were resuspended in 8 M urea, 100 mM NaHPO 4 , pH 8.5 (50 l final volume) and treated with 1 mM DTT at 25°C for 30 min, followed by 5 mM iodoacetimide at 25°C for 30 min in the dark. The samples were then diluted to 1 M urea with 50 mM ammonium bicarbonate (final volume, 400 l) and digested with lysyl endopeptidase (Wako; final concentration, 1.25 ng/l) at 25°C for 4 h and further digested overnight with trypsin (Promega; final concentration, 1.25 ng/l) at 25°C. The resulting peptides were desalted with a Sep-Pak C18 column (Waters) and dried under vacuum. Each sample was resuspended in loading buffer (0.1% formic acid, 0.03% TFA, 1% acetonitrile) and analyzed independently by reverse-phase LC-MS/MS as essentially previously described with slight modification (88). Briefly, peptide mixtures were loaded onto a C18 nano-LC column (75-m-inner diameter, 15-cm-long, 1.9-m resin from Dr. Maisch GmbH) and eluted over a 5-30% gradient (buffer A: 0.1% formic acid; buffer B: 0.1% formic acid in 100% AcN). Eluates were monitored in a MS survey scan followed by 10 data-dependent MS/MS scans on a Q-Exactive plus Orbitrap mass spectrometer (Thermo Scientific, San Jose, CA). The acquired MS/MS spectra were searched against a concatenated target decoy mouse reference database (version 62) of the National Center for Biotechnology Information (downloaded November 14, 2013, with 30,267 target entries) using the SEQUEST Sorcerer algorithm (version 4.3.0, SAGE-N). Search parameters included: fully tryptic restriction, parent ion mass tolerance (Ϯ 50 ppm), up to two missed trypsin cleavages, and dynamic modifications for oxidized Met (ϩ15.9949 Da). The peptides were classified by charge state and first filtered by mass accuracy (10 ppm for high-resolution MS) and then dynamically by increasing XCorr and ⌬Cn values to reduce protein false discovery rate to less than 1%. If peptides were shared by multiple members of a protein family, the matched members were clustered into a single group in which each protein identified by a unique peptide represented a subgroup.

GO enrichment and network analysis
Functional enrichment of the proteins detected with Ala-10 PABPN1 or Ala-17 PABPN1 was determined using the GO-Elite (v1.2.5) package (89). The set of total proteins identified was used as the background. The Z score determines overrepresentation of ontologies in a module, and the permutation p value was used to assess the significance of the Z score cutoff of 1.96 and the p value cutoff 0.05, with a minimum of 5 proteins/ category used as filters in pruning the ontologies. Horizontal bar graphs were plotted in R. The networks were constructed using the circlize package in R.

Solubility assay
Myotubes were lysed in cold lysis buffer (50 mM Tris, pH 8, 150 mM NaCl, 1% Igepal CA-630) and incubated on ice with for 5 min. Lysates were sonicated with three 5-s pulses in a benchtop Branson sonicator (20% output) followed by centrifugation at 13,000 rpm at 4°C for 20 min. The supernatant was collected as the soluble fraction, and the pellet was washed once with cold PBS and centrifuged for 10 min at 4°C. The resulting pellet represented the insoluble fraction and was resuspended in 8 M urea, 2% SDS. Immunoblot analysis was performed on the soluble and insoluble fractions.

RNA isolation and reverse transcription
Total RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's protocol and treated with DNase I (Invitrogen). RNA was reverse-transcribed to cDNA using random hexamers and Moloney murine leukemia virus reverse transcriptase (Invitrogen). One microgram of RNA was used for each reverse transcription reaction and was then employed for quantitative real time PCR.

Quantitative real-time PCR
Relative mRNA levels were measured by quantitative PCR analysis of triplicate samples of 10 ng of cDNA with QuantiTect SYBR Green Master Mix on an Applied Biosystems StepOne Plus real time PCR machine (ABI). Fold change was calculated using the ⌬⌬Ct method. The Gapdh transcript was used as the normalizer for the first ⌬Ct and to primers sets for the coding sequence of the specific target for the second ⌬Ct. Statistical significance was determined using Student's t test.

Glucose uptake assay
Glucose uptake was measured using a glucose uptake cellbased assay kit (catalog no. 600470) from Cayman Chemical as

Differential protein interactions with PABPN1 in muscle
per the manufacturer's instructions. Briefly, myoblasts and myotubes were grown in black, clear-bottomed 96-well plates and treated for 6 h with medium containing 200 g/ml of the fluorescent deoxyglucose analog (2-NBDG) provided with the kit. Following this treatment, supernatant was aspirated, and cell-based assay buffer was added without disturbing the monolayer of cells. 2-NBDG taken up by the cells was detected using a Biotek microplate reader with filters designed to detect fluorescein (excitation/emission ϭ 485/535 nm).

PGRN uptake assay
Myoblasts and myotubes were incubated for 16 h with recombinant N-TAP or C-TAP human Progranulin (a kind gift from the Kukar Lab, Emory University) (54). At the end of the incubation, the cells were washed three times with PBS and lysed in radioimmunoprecipitation assay buffer followed by immunoblotting to detect human PGRN.

Statistical analysis
All error bars represent standard error of the mean with statistical significance being determined with a Student's t test using GraphPad Prism7. In all cases samples with p Ͻ 0.05 were considered to be statistically significant.