Suppression of autophagy permits successful enzyme replacement therapy in a lysosomal storage disorder—murine Pompe disease

Characteristics of the MLCcre:Atg7^F/F:GAA^−/− strain.

We have generated GAA^−/− mice in which a large segment of the Atg7 gene is excised in fast muscle fibers by Cre recombinase; these mice are referred to as MLCcre:Atg7^F/F:GAA^−/−. We have also made a control MLCcre:Atg7^F/F:WT strain in which Atg7 is excised in fast muscle fibers of WT mice.

Consistent with our previous data,¹⁹ autophagy is upregulated in muscle of the GAA^−/− mice as shown by an increase in both cytosolic LC3-I and autophagosomal membrane-bound LC3-II (Fig. 1A). As expected, in the MLCcre:Atg7^F/F:GAA^−/− mice autophagy was considerably suppressed in fast (gastrocnemius, Fig. 1A) but less so in slow (soleus, Fig. 1B) muscles as indicated by the levels of LC3-II. The suppression of autophagy in fast muscle is further supported by the increase in the level of p62 protein (Fig. 1C). The degree of the suppression of autophagy in fast muscles differs in young and old mice, as shown by a progressive decline with age in the levels of LC3-II and Atg7 (Fig. 1A and Suppl. Fig. 1). Small clusters of double-membrane vesicles containing undigested cytoplasmic material and possibly glycogen were detected by electron microscopy (Fig. 2), but autophagic buildup, which is so prominent in GAA^−/− muscle, was not observed in fibers from MLCcre:Atg7^F/F:GAA^−/− mice of any age (Fig. 1D and Suppl. Video 1).

Expansion of lysosomes, a hallmark of Pompe disease, persists in muscle fibers from MLCcre:Atg7^F/F:GAA^−/− mice, but the lysosomes in the majority of the fibers are smaller (5–6 µm) than those in the GAA^−/− mice (12–15 µm) (Fig. 1C). Furthermore, the fibers from MLCcre:Atg7^F/F:GAA^−/− contain a large number of normal-looking, dot-like LAMP-1-positive lysosomes which appear to maintain their acidic pH (Suppl. Fig. 2A). Other prominent features of these autophagy-deficient fibers are clustering of lysosomes in the core of fibers (Suppl. Fig. 2B) and an age-dependent accumulation of ubiquitinated (Ub) proteins (Suppl. Fig. 2C–E).

The most striking finding in MLCcre:Atg7^F/F:GAA^−/− mice, however, is a ∼60% reduction in the level of glycogen accumulation in fast muscle compared to that in the GAA^−/− mice, suggesting that the suppression of autophagy serves as substrate reduction therapy (Table 1 and Suppl. Table 1). The reduced glycogen level is detected in fast but not in cardiac (Table 1) or slow muscles (Suppl. Table 1), indicating that suppression of autophagy, rather than differences in glycogen metabolism in the two strains, is responsible. Clinically, the MLCcre:Atg7^F/F:GAA^−/− mice are no different from the GAA^−/−,²¹ in terms of muscle wasting, kyphosis and life span.

The absence of autophagic buildup, a significantly reduced glycogen load and the lack of additional clinical manifestations made the MLCcre:Atg7^F/F:GAA^−/− model a promising candidate for enzyme replacement therapy.

Enzyme replacement therapy.

Suppression of autophagy combined with ERT resulted in a ∼90% decrease of glycogen in fast muscles from MLCcre:Atg7^F/F:GAA^−/− mice when compared to the levels in untreated GAA^−/− mice (Fig. 3A and Table 1). This glycogen clearance was also shown by PAS staining of muscle biopsies (Fig. 3B) and by immunostaining of single fibers for a lysosomal (LAMP-1) marker (Fig. 3C). In contrast, ERT-treated GAA^−/− mice with genetically intact autophagy had only a ∼30% reduction in fast muscles (Fig. 3A and Table 1). Furthermore, the autophagic buildup persisted in ERT-treated GAA^−/− muscle, as shown by immunostaining of single fibers for an autophagosomal marker, LC3 (Fig. 3C and Suppl. Video 2). As expected, in both strains ERT cleared glycogen from cardiac muscle very efficiently (Table 1).

We also tested ERT in another, previously described muscle-specific, autophagy-deficient Pompe model with a higher level of glycogen accumulation.¹⁹ In this model, HSAcre:Atg5^F/F:GAA^−/−, the Atg5 gene was excised by Cre recombinase driven by the human skeletal actin promoter. Unlike in the MLCcre:Atg7^F/F:GAA^−/− strain, glycogen levels in HSAcre:Atg5^F/F:GAA^−/− mice were insignificantly reduced by only ∼20% compared to the GAA^−/− (4.0 ± 2.6% wet weight and 5.0 ± 2.8% for HSAcre:Atg5^F/F:GAA^−/− and GAA^−/− respectively).¹⁹ With the additional samples the difference (∼29%) reached statistical significance, but the glycogen levels still remained much higher than in MLCcre:Atg7^F/F:GAA^−/− mice (Fig. 3A, Table 1 and Suppl. Table 1).

Despite the higher level of glycogen, the effect of ERT in HSAcre:Atg5^F/F:GAA^−/− was equally impressive—a complete removal of glycogen was observed (Fig. 3A and Table 1). These data were supported by PAS staining of muscle biopsies (Suppl. Fig. 3) and by immunostaining of fibers for LC3 and LAMP-1 (shown for MLCcre:Atg7^F/F:GAA^−/− in Fig. 3C). Thus, it is the suppression of autophagy rather than the initial level of glycogen accumulation that has an effect on therapy.

We next examined the functionality of the lysosomes in ERT-treated mice by analyzing the level of Ub-proteins. A number of transmembrane proteins and receptors are ubiquitinated and delivered via the endocytic pathway to the lysosomes for degradation.^22,23 Another pool is delivered to the lysosomes via constitutive autophagy.^24–26 Both pools contribute to the accumulation of Ub-proteins in fast GAA^−/− fibers,¹⁹ as well as in cells from patients with different lysosomal storage diseases.²⁷ We observed no appreciable reduction of Ub-proteins in ERT-treated GAA^−/− mice (Fig. 4A). In contrast, ERT treatment in autophagy-deficient GAA^−/− strains resulted in a decrease in the amount of Ub-proteins in both soluble and nonsoluble fractions as shown by western blotting (Fig. 4B) and by immunostaining of myofibers (Fig. 4C). These data suggest that the lysosomal function in muscles from treated autophagy-deficient GAA^−/− mice is largely restored. Thus, the suppression of autophagy in skeletal muscle greatly facilitates the effect of ERT, resulting in an outcome which has not been observed in Pompe mice with genetically intact autophagy.

The functional effect of ERT, however, as assessed by the force of the EDL muscle from treated and untreated autophagy-deficient GAA^−/− mice, was not measurable because at the age of these animals (4 months) even untreated GAA^−/− mice show no significant difference in the specific isometric tetanic force when compared to the WT (Suppl. Table 2).

Although successful reversal of lysosomal pathology is achieved in MLCcre:Atg7^F/F:GAA^−/− mice by ERT, it should be remembered that these animals remain autophagy-deficient in skeletal muscle. Consistent with the previously published data,²⁸ suppression of autophagy in the control MLCcre:Atg7^F/F:WT or HSAcre:Atg5^F/F:WT mice resulted in accumulation of Ub-proteins; this accumulation, however, was modest when compared to that in the GAA^−/− (Fig. 4B). No weight loss was recorded in 4-month-old MLCcre:Atg7^F/F:WT animals [21.1 ± 1.0 g (n = 11) and 21.2 ± 2.3 g (n = 7) in WT and MLCcre:Atg7^F/F:WT respectively] and no major clinical abnormalities were noted in these mice during an observation period of up to two years. These autophagy-deficient mice show muscle atrophy, as evidenced by a tendency toward a reduction in cross-sectional area of EDL muscle (Suppl. Table 2) and by a reduced diameter of single psoas (fast) fibers (Suppl. Table 3). The maximum Ca⁺²-activated isometric force generated by single psoas fibers were lower than in the WT but higher than in the GAA^−/− (Suppl. Table 3). In addition, the overall muscle strength, as measured by a grip strength test, of the autophagy-deficient mice was lower than the WT (Suppl. Table 4).

Possible triggers of autophagy in fast muscle of GAA^−/− mice.

We have previously shown that autophagy is upregulated in Pompe muscle.¹⁹ We now add two additional pieces of information supporting the upregulation of autophagy: Beclin 1, a protein known to be activated during autophagy induction,²⁹ and the Atg7 protein are increased in GAA^−/− fast muscle (Figs. 1A and 5A).

Unraveling the mechanisms of autophagic disturbance in skeletal muscle under pathological conditions is a daunting task, particularly because skeletal muscle is a peculiar tissue in terms of how it responds to the classic inducers of autophagy—starvation and an mTOR inhibitor, rapamycin. Unlike cardiac muscle, WT fast muscle showed no appreciable increase in the amount of LC3-II following 24 or 48 hours of starvation (Suppl. Fig. 4A and B). This outcome was unexpected because it had been previously shown that starvation did result in conversion of LC3-I to LC3-II in muscle.³⁰ The discrepancy may be attributed to the differences in the kind of muscles used (we consistently use only the white part of the gastrocnemius muscle), the genetic background and the age of the animals. For starvation experiments, we routinely use 4–5 month-old animals; in young, ∼1 month old mice, we, too, see a mild response to starvation but the results are inconsistent (not shown).

Rapamycin also does not induce autophagy in WT skeletal muscle as shown by us (Suppl. Fig. 4C) and others.³¹ Similar to the effect of rapamycin in the WT, no increase in LC3-II was detected in muscle from GAA^−/− mice although mTOR activity was suppressed by the drug, as evidenced by a decrease in the amount of the hyper-phosphorylated (γ) 4E-BP1 (Suppl. Fig. 4C). Furthermore, the mTOR regulation of protein synthesis in GAA^−/− muscle does not fit the typical mold. mTOR regulates protein synthesis through the phosphorylation and inactivation of 4E-BP1, a repressor of mRNA translation.³² In the GAA^−/− muscle there is a significant increase in both hypo- (α and β) and hyper-phosphorylated (γ) forms of 4E-BP1 as if conflicting messages are sent for protein synthesis (Suppl. Fig. 4D).

The role of FOXO transcription factors, known to be involved in the induction of autophagy,³¹ is also not clear in GAA^−/− muscle. We have previously shown that FOXO1 was not upregulated in fast muscles of GAA^−/− mice.¹⁹ We now demonstrate by realtime PCR that FOXO3 is, in fact, downregulated (∼1.5 fold) in fast muscles of GAA^−/− mice (n = 5).

Although starvation does not appear to induce autophagy in WT muscle, in GAA^−/− muscle we did observe an increase in LC3-II upon starvation (Suppl. Fig. 4B). This sensitivity to starvation of fast muscle in the GAA^−/− mice may be one of the factors contributing to the increase in autophagy in this disease.

Considering the lack of clarity concerning the classical triggers of autophagy in GAA^−/− muscle, we looked at a recently appreciated regulator of autophagy, glycogen synthase kinase 3β (GSK-3β), a protein long known to be involved in glycogen metabolism. A significant decrease in phosphorylation (activation) of GSK-3β is seen in skeletal muscle in both young and old GAA^−/− mice (Fig. 5B and C) leading to an increase in phosphorylation (inactivation) of glycogen synthase (GS) (Fig. 5D). The inactivation of glycogen synthase may reflect a homeostatic adjustment in muscle cells to reduce the cytoplasmic glycogen burden, but it appears to come at a price—the induction of autophagy by activation of GSK-3β, with unfortunate consequences for Pompe skeletal muscle.

To investigate whether or not GSK-3β can induce autophagy in muscle cells, we expressed constitutively active GSK-3β^S9A in C2C12 cells. Indeed, immunostaining and western analysis for LC3 showed an induction of autophagy in both C2C12 myoblasts and myotubes (Fig. 6A and B). Furthermore, a similar induction of autophagy was observed in GAA^−/− derived primary myoblasts,³³ as demonstrated by a two-fold increase of LC3-II/α-tubulin ratio in the cells expressing GSK-3β^S9A as compared to those expressing the vector alone (not shown). Consistent with the data in whole muscle, mTOR signaling does not appear to regulate autophagy in C2C12 myotubes, as shown by the absence of changes in the levels of phosphorylated 4E-BP1 in the cells expressing GSK-3β^S9A (Fig. 6D).

Thus, a homeostatic attempt to downregulate the synthesis of glycogen may contribute to the upregulation of autophagy in this glycogen storage disorder.

Generation and genotyping of muscle-specific autophagy-deficient Pompe mice.

Atg7-conditional knockout mice²⁶ (Atg7^flox/flox, referred to as Atg7^F/F), provided by Dr. Komatsu, were crossed with GAA^−/− mice⁵⁸ (C57BL/6/129/Sv inbred strain) to produce Atg7^F/F:GAA^−/− mice. We then crossed Atg7^F/F:GAA^−/− mice to a skeletal muscle specific Cre line, which has been generated by homologous recombination of Cre into the myosin light chain 1f (MLC) locus (knock-in) (C57BL/6).⁵⁹ The Cre mice were generously provided by Dr. Burden (Skirball Institute for Molecular Medicine, New York University Medical center, NY, New York). Cre expression in this line is restricted to fast fibers and is activated early during muscle development. After several rounds of crossing we produced GAA^−/− mice deficient for Atg7 specifically in skeletal muscle (MLCcre:Atg7^F/F:GAA^−/−).

Genomic DNA was isolated from tail clips using the QuickGene DNA tissue kit (AutoGene, FK-DTS) according to the manufacturer's instructions. The presence of Cre recombinase is indicated by a ∼400 bp PCR product made with either primer pair Cre sense/antisense: 5′-CCG GTG AAC GTG CAA AAC AGC CTC TA-3′/5′-CTT CCA GGG CGC GAG TGG ATA GC-3′ or 5′-AAG CCC TGA CCC TTT AGA TTC CAT TT-3′ (MLC1f promoter)/5′-CCT CAT CAC TCG TTG CAT CGA-3′ (anti-sense in Cre). The presence of a wild-type MLC1f gene (Cre-negative) is indicated by a ∼300 bp PCR product made with MLC1f promoter primer (see above) and an antisense primer in exon 1 of the MLC1F gene (5′-GCA GCG GGC TTC TTC ACG TCT TTC T-3′). The Atg7 wild type, Atg7^F/+ and Atg7^F/F alleles were detected with the primer pair 5′-TGG CTG CTA CTT CTG CAA TGA TGT-3′/5′-GAA TAT TCT AAT TCA ACC AGA TCT AGG T-3′ (Atg7ex13s/Atg7intr13as) that amplifies a ∼1.5 kb fragment from the wt allele and a ∼0.35 kb fragment from the floxed allele. The wild-type Atg7 can also be detected as a PCR fragment of ∼0.6 kb with primer pair 5′-GCT GGT TAA AGA CTG TCT AAT AAA GAG CA-3′/5′-CTG CCG CTG AGC CCT GAG AGA GGC CT-3′ (intron 13s/mid-intron 13as). The GAA wild type, GAA^+/− and GAA^−/− alleles were detected as described.⁵⁸ Generation and genotyping of HSAcre:Atg5^F/F:GAA^−/− mice was described previously.¹⁹ In these muscle-specific autophagy-deficient mice, the Atg5 gene is inactivated in muscle by Cre recombinase driven by the human skeletal actin (HSA) promoter.

The removal of the LoxP-flanked fragment from the Atg7 gene was confirmed by PCR of genomic DNA with Atg7ex13s primer (5′-TGG CTG CTA CTT CTG CAA TGA TGT-3′) and an antisense primer 5′-CAC ACA GAT GTG ACA CTG GAG TGC AGC AT-3− in intron 14. The excision of the loxP-flanked fragment [which includes all the exons starting at exon 14 of the Atg7 gene²⁶] is indicated by the presence of a ∼1.2 kb PCR fragment. A combination of three primers—Atg7ex13s/Atg7intr13as/Intron 14as—allows for the detection of both excised (∼1.2 kb) and nonexcised floxed (∼0.4 kb) alleles. Analysis was performed using tail or tissue (muscle or brain) DNA, which was isolated according to the standard procedure. The MLCcre:Atg7^F/F:GAA^−/− colony contained ∼150 mice. The clinical signs of muscle disease were monitored in ∼50 mice.

Enzyme replacement therapy.

Two and a half month-old GAA^−/−, MLCcre:Atg7^F/F:GAA^−/− and HSAcre:Atg5^F/F:GAA^−/− mice received 3 i.v. injections of alglucosidase alfa (provided under a CRADA between the NIH and the Genzyme Corporation) at a dose of 100 mg/kg every other week. Diphenhydramine hydrochloride was injected i.p at a dose of 5 mg/kg 15 minutes before the second and third injections of rhGAA as described.⁶⁰ The mice were sacrificed 5 days after the last injection. Twelve GAA^−/−, 11 HSAcre:Atg5^F/F:GAA^−/− and 9 MLCcre:Atg7^F/F:GAA^−/− were treated with rhGAA.

Isolation of fixed single muscle fibers and immunofluorescence microscopy.

Muscle fixation, isolation of single fibers and immunostaining are described in detail in Raben et al.¹² At least three animals from each genotype were used to obtain single muscle fibers for immunostaining. For each immunostaining and for confocal analysis, at least 20 fibers were isolated from psoas muscle.

Isolation of live single muscle fibers and staining with acridine orange.

Single live myofibers were isolated from gastrocnemius muscle. The dissected muscle was placed in a Sylgaard plate containing PBS. Muscles were pinned at resting length and cleaned of any extraneous material under a dissection microscope. Muscles were then digested in 0.2% (w/v) collagenase type 1 (Sigma-Aldrich, C0130-1G) in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen, 11054-020) for 1 h at 37°C in an atmosphere of 5% CO₂. Individual fibers were released under a dissecting microscope by gentle pipeting of partially digested muscle bundles. Individual translucent myofibers (∼20 fibers) were plated in each well of a BD cell TAK™-coated 2-chambered cover glass (BD Pharmingen, 354241) and incubated in plating medium [10% horse serum in DMEM (HyClone Laboratories, SH30074.03)] overnight to let the fibers attach to the surface; the fibers were then incubated with acridine orange (AO) at 2.5 µg/ml for 10 minutes (Molecular Probes, A3568). Fibers were washed three times with medium and then analyzed on a Zeiss 510 confocal microscope (AO emission was collected with a 500–560 nm band pass filter (green) and a 590–660 nm band pass filter (red)).

Western blot.

Whole muscle tissues were homogenized in RIPA buffer [PBS containing 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS and a protease inhibitor cocktail tablet (Roche Diagnostics, 11836170001)]. Samples were centrifuged for 30 min at 13,000 rpm at 4°C. Alternatively, detergent-soluble and -insoluble fractions from muscle tissues were obtained as described.²⁴ Protein concentrations of the supernatants of the total lysates or soluble fractions were measured using the Bio-Rad Protein Assay (Bio-Rad Laboratories, 500-0006). Equal amounts of protein were run on SDS-PAGE gels (Invitrogen, EC6498BOX and EC6078BOX) followed by electro-transfer onto nitrocellulose membranes (Invitrogen, LC2001). Membranes were blocked in 1:1 PBS and Odyssey Blocking Buffer (LI-COR Biosciences, 927-40003), incubated with primary antibodies overnight at 4°C, washed, incubated with secondary antibodies and washed again. Blots were scanned on an infrared imager (LI-COR Biosciences).

The following primary antibodies were used for western blots and immunostaining of fixed fibers: rabbit anti-LC3B (microtubule-associated protein 1 light chain 3) (Sigma, L7543); rat anti-mouse LAMP-1 (Lysosomal-Associated Membrane Protein 1) (BD Pharmingen, 553792); rabbit polyclonal anti-Atg7 (Cell Signaling Technology, 2631); mouse anti-poly-ubiquitinated conjugates (FK2) (BIOMOL International, PW8810); goat polyclonal anti-BECN1 (Beclin 1) (Cell Signaling Technology, 3738); rabbit monoclonal anti-GSK-3β (Cell Signaling Technology, 9315); rabbit monoclonal anti-phospho-GSK-3β (Ser9) (Cell Signaling Technology, 9323); rabbit monoclonal anti-glycogen synthase (Cell Signaling Technology, 3886) and rabbit polyclonal anti-phospho-glycogen synthase (Ser641) (Cell Signaling Technology, 3891); rabbit polyclonal anti-eIF4EBP1 (Abcam, ab2606) and anti-phospho-4E-BP1 (Cell Signaling Technology, 9459); guinea pig polyclonal anti-p62 (ProGen, GP62-N); rabbit polyclonal anti-HA-probe (Santa Cruz Biotechnology, sc-805); mouse monoclonal anti-vinculin (Sigma, V 9131) and mouse monoclonal anti-GAPDH antibody (Abcam, ab9484) served as loading controls. Alexa Fluor-conjugated antibodies (Molecular Probes, A21057, A21076, A21096) were used as secondary antibodies.

Quantitative real-time PCR.

Total RNA was extracted from gastrocnemius muscle using TRIzol reagent (Gibco-BRL, 15596-026) according to the standard procedure. The RNA cleanup protocol from the Qiagen RNeasy Mini Kit (Qiagen Sciences, 74104) was used to eliminate short transcripts. 5 µg total RNA was used for cDNA synthesis using the High Capacity cDNA Archive Kit from Applied Biosystems (4368814). The cDNA was then diluted 1:5 and 1 µL of the diluted cDNA was used to perform real-time PCR in 20 µL reactions in 96-well optical plates according to the manufacturer's instructions. SYBR® Green Mouse Foxo3 primer set [SABioscienes, (Mm. 338613)] was use for the analysis. Mouse β-actin was used as an endogenous control.

Electron microscopy.

Muscles were fixed and treated as described¹² except that in some cases osmication was done in the presence of 1.5% potassium ferrocyanide (“reduced osmium”) for darker glycogen staining.

The source of fast and slow muscles.

The white part of gastrocnemius, quadriceps, psoas and extensor digitorum longus (EDL) muscles in mice are a good source of glycolytic fast-twitch type II fibers, whereas soleus muscle is a good source of oxidative slow-twitch type I fibers.^41,61

Glycogen measurement and light microscopy.

Glycogen concentration in skeletal muscle, periodic acid-Schiff (PAS) staining of muscle biopsies and fiber type analysis were performed as described.^19,53,62

Plasmids, C2C12 cell cultures and transfection.

The EcoRI/XhoI fragment containing HA (hemagglutinin)-tagged GSK-3β S⁹A was released from the pcDNA3 plasmid (Addgene plasmid #14754). This fragment was then cloned into the EcoRI/SalI sites of the pHan-Puro retroviral expression vector (kindly provided by Dr. M. Fulco, NIAMS, NIH). GSK-3β S⁹A is a glycogen synthase kinase in which serine 9 of human GSK-3β is mutated to alanine. When expressed, the mutant kinase is constitutively active (ca). The pHan-Puro-GFP vector was used as a control. Culturing and transfection of C2C12 myoblasts and myotubes were done according to the standard procedures with some modifications. Viral supernatants were obtained by transfection of Phoenix cells (Orbigen, RVC-10002) with the retroviral expression vectors (pHan-Puro or pHan-Puro containing constitutively active GSK-3βS9A) using the calcium phosphate method. C2C12 mouse myoblasts were grown in DMEM supplemented with 20% fetal bovine serum (Atlanta Biologicals, S11550), 0.5% chick embryo extract (Sera Laboratories International Ltd., CE-650-T) and 1% penicillin-streptomycin-glutamine (Invitrogen, 10378-016). The cells were seeded into 6-well plates (3 × 10⁴ cells/well) and infected with the filtered viral supernatant containing 8 µg/ml Polybrene (Sigma, H9268-5G). The plates were centrifuged at 3,700 rpm at room temperature for 1.5 h, placed at 37°C overnight, and then selected in the presence of 2 µg/ml of puromycin (Invitrogen, A1113803). For differentiation of C2C12 cells into myotubes, the cells were grown to near confluence and the culture medium was changed to DMEM containing 1% penicillin-streptomycin-glutamine and 5% horse serum (HyClone Laboratories, SH30074.03).

For starvation, the cells (myoblasts or myotubes) were incubated in Krebs-Ringer solution at 37°C for 4 h. Alternatively, cells were treated with 400 nM bafilomycin [(Sigma, 131793) in DMEM containing serum and supplements to inhibit autophagosome-lysosome fusion⁶³]. For western analysis the cells were lysed in RIPA buffer with inhibitors. For immunostaining, the cells were fixed with 2% paraformaldehyde (Electron Microscopy Sciences, 15710) in 0.1 M phosphate buffer for 1 h. After several washes with PBS, myoblasts were incubated with blocking reagent (MOM kit; Vector Laboratories, BMK-2202) for 1 h at room temperature; the cells were then incubated with primary antibodies overnight at 4°C, washed with PBS, incubated with secondary antibody for 2 h and washed again with PBS before examination by confocal microscopy (Zeiss LSM 510).

Force measurements.

Force measurements on the EDL (extensor digitorum longus) muscles were conducted as described.⁶⁴ Force measurements on single fibers isolated from psoas muscle were done as described.⁶⁵ The overall muscle strength was evaluated by grip strength measurements using a grip-strength meter (Columbus Instruments, 1027MPB); the data were normalized by body weight and expressed as KGF/kg as described.⁶⁶

Animal care and experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.