Deletion of Mbtps1 (Pcsk8, S1p, Ski-1) Gene in Osteocytes Stimulates Soleus Muscle Regeneration and Increased Size and Contractile Force with Age*

Conditional deletion of Mbtps1 (cKO) protease in bone osteocytes leads to an age-related increase in mass (12%) and in contractile force (30%) in adult slow twitch soleus muscles (SOL) with no effect on fast twitch extensor digitorum longus muscles. Surprisingly, bone from 10–12-month-old cKO animals was indistinguishable from controls in size, density, and morphology except for a 25% increase in stiffness. cKO SOL exhibited increased expression of Pax7, Myog, Myod1, Notch, and Myh3 and 6-fold more centralized nuclei, characteristics of postnatal regenerating muscle, but only in type I myosin heavy chain-expressing cells. Increased expression of gene pathways mediating EGF receptor signaling, circadian exercise, striated muscle contraction, and lipid and carbohydrate oxidative metabolism were also observed in cKO SOL. This muscle phenotype was not observed in 3-month-old mice. Although Mbtps1 mRNA and protein expression was reduced in cKO bone osteocytes, no differences in Mbtps1 or cre recombinase expression were observed in cKO SOL, explaining this age-related phenotype. Understanding bone-muscle cross-talk may provide a fresh and novel approach to prevention and treatment of age-related muscle loss.

Bone osteocytes are multifunctional endocrine cells. Osteoid osteocytes control bone mineralization by secretion of DMP1, phosphate-regulating endopeptidase homolog, X-linked (PHEX), and matrix extracellular phosphoglycoprotein (MEPE) (1), whereas mature osteocytes secrete sclerostin, an inhibitor of bone formation, and regulate phosphate homeostasis and, indi-rectly, bone mineralization via production of FGF23 (2). Osteocytes are the mechanosensory cells in bone and appear to regulate muscle function and myogenesis (3)(4)(5). Specifically, deletion of connexin43 in osteoblasts/osteocytes shows it is involved in postnatal bone-muscle cross-talk via an osteocalcin-mediated stimulation of muscle formation and function (5). Differentiated muscle cells, in return, secrete as yet unidentified factors that protect osteocytes against apoptosis and produce a factor that synergizes with fluid flow shear stress to increase prostaglandin E 2 production by osteocytes (6). Myogenesis in adult muscle occurs via activation of satellite (stem) cells during muscle regeneration after injury or exercise-induced damage (7,8). Distinct populations of satellite cells appear to be responsible for regeneration of fast twitch and slow twitch myofibers (9,10), respectively. MBTPS1 (SKI-1, PCSK8, S1P) 3 is the eighth member of the proprotein convertase family of proteases (11). Localized to the cis/medial Golgi, MBTPS1 catalyzes the mandatory first cleavage of transmembrane-bound transcription factors like SREBP1 and cAMP-response element-binding protein; once released, their DNA binding domains are imported into the nucleus where they regulate transcription (12). Interestingly, MBTPS1 is required for the transcription of a number of bone matrix and mineralization-related genes including type XI collagen, Phex, Dmp1, fibronectin, and fibrillin in bone osteoblasts and osteocytes. Correspondingly, inactivation of MBTPS1 by AEBSF or specific inhibitor decanoyl-Arg-Arg-Leu-Leu-chloromethyl ketone blocks transcription of the aforementioned genes and inhibits mineralization (13,14). These results demonstrated that the differentiated phenotype of osteoblastic cells and possibly osteocytes depends upon MBTPS1.
Having identified a functional role for MBTPS1 in osteoblasts and osteocytes in vitro, we asked whether skeletal functions require MBTPS1 in vivo. Because deletion of Mbtps1 in TGG-3Ј (male ϭ 331 and 302 bp, female ϭ 331 bp). PCR products were evaluated by electrophoresis on 2.1% agarose gels followed by staining with ethidium bromide and digitally imaged with a Fuji LAS4000 charge-coupled device detector.
Analysis of MBTPS1 Protein from Muscle and Bone-Forelimbs and tibiae from control and Mbtps1 cKO mice were dissected free of soft tissues, then frozen in liquid nitrogen, and pulverized in a bone mill. Bone powder was then washed extensively with PBS to remove marrow cells and blood prior to extraction at 4°C by mixing for 72 h with 0.1 M Tris acetate buffer, pH 7.4, containing 8 M urea, 0.2% CHAPS, 0.5 M EDTA, and 0.02% sodium azide. A ratio of 20 ml of extraction solution/g of bone powder was used. After clarification of the extracts by centrifugation, supernatant fractions were dialyzed separately first against distilled water and then against two changes of 5% acetic acid at 4°C. Extracts were then lyophilized to dryness in several aliquots. Following protein quantitation using the Geno Technology Non-Interfering Protein Analysis Assay kit, samples were hydrated with SDS-PAGE sample buffer and an excess of dithiothreitol and heated at 95°C for 10 min. 52 g of bone extracts were electrophoresed per lane of 4 -20% linear gradient SDS gels (Precise Tris-HEPES gels, Pierce) at 50 V for 2.5 h and then electroblotted onto PVDF membrane at 100 V for 2 h. Coomassie Blue prestained globular protein standards were also electrophoresed on each gel. Blots were processed for immunodetection with rabbit anti-human or mouse MBTPS1 C-terminal primary antibodies (21) using a SuperSignal West Femto Maximum Sensitivity Chemiluminescent Substrate kit (Thermo Scientific) as described previously (13). Blots were digitally imaged using a Fuji LAS4000 system with a charge-coupled device camera.
Soleus muscles of control and Mbtps1 cKO mice were used for contractility studies and then immediately flash frozen in liquid nitrogen and stored at Ϫ80°C. Tissues were then homogenized, and the total protein was isolated using radioimmune precipitation assay buffer (Sigma-Aldrich) according to the manufacturer's instructions. The lysates were centrifuged at 12,000 rpm for 10 min at 4°C, and the supernatants were collected. The protein concentration of the lysates was detected using a Micro BCA Protein Assay kit (Pierce). The lysate and sample buffer were then mixed and boiled for 10 min before loading onto the gel. Proteins (20 and 40 g/lane) were separated on a Mini-PROTEIN precast gel (Bio-Rad) under constant voltage (200 V) and then transferred electrophoretically to a nitrocellulose membrane by using the Trans-Blot Turbo transfer system using the TGX gel program (Bio-Rad). Membranes were blocked in 5% nonfat dry milk at 4°C and then incubated with the primary antibody (anti-Mbtps1 C-terminal antibody (21)) overnight at 4°C. Blots were incubated with secondary antibody (goat anti-rabbit IgG, HRP conjugate (1:20,000); EMD Millipore, Billerica, MA) for 2 h at room temperature and then visualized as noted above for bone extracts.
Immunohistochemical Analysis of Myosin Heavy Chains-Soleus muscles and extensor digitorum longus muscles (EDL) were dissected immediately after sacrifice and flash frozen in optimum cutting temperature medium for storage. Muscles were cut into 14-m-thick sections using a cryostat and then processed for immunofluorescence staining using a protocol adapted from that published by Phillips et al. (22). Briefly, sections were air-dried, acetone-fixed, and then blocked with mouse M.O.M. IgG blocking reagent (Vector Laboratories, Inc.) for 1 h at room temperature. After washing in PBS, sections were incubated overnight in the following mixture of primary antibodies dissolved in PBS containing 2% normal goat serum: 1:100 anti-type I myosin heavy chain (BS.D5 IgG2b, Developmental Studies Hybridoma Bank), 1:200 anti-type IIA myosin heavy chain (SC.71 IgG1, Developmental Studies Hybridoma Bank), and 1:100 anti-type IIB myosin heavy chain (BF.F3 IgM, Developmental Studies Hybridoma Bank). After washing in PBS, sections were treated for 2 h with the following mixture of secondary antibodies dissolved in PBS containing 2% normal goat serum: Alexa Fluor 647-conjugated goat antimouse IgG2b (Invitrogen), Alexa Fluor 488-conjugated goat anti-mouse IgG1 (Invitrogen), and Alexa Fluor 594-conjugated goat anti-mouse IgM (Invitrogen). After washing extensively with PBS, sections were postfixed in methanol for 5 min, rinsed in PBS, and then coverslipped with mounting medium including DAPI (Vector Laboratories, Inc.). Sections were imaged using an inverted Nikon TE2000-E epifluorescence microscope and Metamorph software. ImageJ software was used to overlay different colored images of the same field, and then contrast and brightness were adjusted in Photoshop.
Measurement of Food Intake-Mice were housed individually in Tecniplast metabolic cages for 6 days. The first 3 days represented an acclimatization phase. Daily food intake, urine output, body weight, and solid waste output were then measured for the 2-day period from days 4 to 6, which were considered the study period. All mice were returned to normal caging at the end of the study period.
Isolation of Osteocyte-enriched Bone and Extraction of Proteins for Analysis-Three cKO male mice, two control littermates, and two wild type male mice aged 10 months were processed separately via the procedure described below to obtain osteocyte-enriched bone matrix. Mice were cervically dislocated and sprayed with 70% ethanol prior to dissection of the femora, tibiae, humerus, ulna, and radius while leaving the end of joints intact. Feet were removed from forelimbs and hind limbs distal to the ankle joints and placed in ␣-minimum Eagle's medium containing penicillin/streptomycin in a laminar flow hood. Limbs were sequentially passed through four dishes of sterile PBS to dilute possible contaminants and then placed in ␣-minimum Eagle's medium. Each limb was then dissected carefully in a small amount of medium, removing muscle, tendons, and ligaments from the still intact joints. Femora were separated from tibiae, and "clean" bones were placed in ␣-minimum Eagle's medium. Both ends of each bone were cut off using a scalpel, and the marrow was flushed out using PBS. Dissected bones were then placed in ␣-minimum Eagle's medium and split longitudinally with a scalpel, and after cutting into 1.5-2.0-mm sections, the pieces were divided into 2 wells of a 6-well culture dish containing PBS. Bone pieces were monitored microscopically throughout the digestion protocol. For Digest 1, the PBS was removed, and the bone pieces were treated with 4 ml of collagenase (2 mg/ml) for 30 min at 37°C. Cells released were removed, the bone pieces were rinsed with 2 ml of Hanks' balanced salt solution, and the collagenase diges-tion step was repeated twice more before the bone chips were incubated with 5 mM EDTA and 0.1% BSA in PBS for 30 min at 37°C. Cells released were removed. The bone chips were then rinsed with 2 ml of Hanks' balanced salt solution and retreated with collagenase for 30 min at 37°C to obtain an "osteocyteenriched bone" fraction.
The osteocyte-enriched bone fraction was immediately incubated for 1 h at 4°C with 0.4 mM AEBSF in PBS to block autolysis of MBTPS1 prior to freezing at Ϫ80°C. Bone chips (0.2-0.3 g) derived from each mouse were then pulverized individually in liquid nitrogen into small particles and resuspended in 10 ml 0.1 M Tris acetate buffer, pH 7.5, containing 8 M urea, 0.5 M EDTA, and 0.2% CHAPS. Each sample was heated initially at 95°C for 10 min to inactivate degradative enzymes and then extracted for 3 days at 4°C with mixing. The extracts were then desalted by dialysis in 3500-molecular weight-cutoff tubing against two changes of deionized water and then four changes of 5% acetic acid at 4°C. Protein content of extracts was determined with the Non-Interfering Protein Analysis Assay kit. Proteins were solubilized in SDS-and 8 M urea-containing sample buffer and electrophoresed under reducing conditions on 4 -20% gradient gels (Precise, Thermo Fisher) as described previously (13). Gels were then electroblotted onto PVDF membranes in a CAPS/MeOH buffer, pH 11.0 and subjected to immunodetection with a 1:20,000 dilution of primary antibody recognizing the C-terminal domain of MBTPS1 (21). Chemiluminescence signals were generated using HRP-conjugated goat anti-mouse IgG (Bio-Rad) and SuperSignal West Femto Chemiluminescent Substrate (Pierce). Digital images were collected using a Fuji LAS4000 imager.
Three-point Bending to Failure-To characterize the biomechanical properties of the bones, displacement-controlled three-point bending to failure tests were conducted on excised femora ex vivo. Femora were harvested from freshly euthanized mice. For each femur, all associated soft tissue was removed, and the bone was individually wrapped in salinesoaked gauze and then stored at Ϫ20°C until testing. On the day of testing, samples were removed from Ϫ20°C storage and allowed to thaw and reach room temperature. Prior to and throughout the tests, bone samples were kept hydrated with saline-soaked gauze and not allowed to dry out. The femora were placed into the test fixtures so they would be impacted in what would be the anterior to posterior direction in vivo (ElectroForce 3200, Bose Corp., Minnetonka, MN). Cross-head displacement and axial load were recorded at a rate of 70 Hz. The stiffness and ultimate force were calculated from the resulting load versus displacement curves for each sample. The Young's modulus (E) for each bone was calculated using the following equation.
where S is the stiffness, L is the span length, and I is the area moment of inertia. The area moment of inertia was calculated across the midspan at the fracture location using 10 slices midspan of the realigned microCT scans using the BoneJ plug-in for the image-processing program ImageJ (National Institutes of Health, Bethesda, MD) (23,24).
MicroCT Analyses on Bone-MicroCT scans were made on isolated femora using a Scanco Vivact 40 instrument. The instrument was calibrated weekly using a series of hydroxyapatite phantoms. Cortical and trabecular bone content was determined using three-dimensional reconstructed bone volume, proprietary software, and standardized protocols with a threshold of 260 g/cm 2 . For example, cortical content was assessed using a 50-m-thick cross-section extending distally from the base of the third trochanter, which was used as an anatomical landmark. Trabecular content was estimated using a 100-m-thick cross-section of the proximal growth plate containing exclusively secondary spongiosa. Values for total volume, bone volume, hydroxyapatite content per total volume, and hydroxyapatite content per bone volume were obtained from these analyses.
Ex Vivo Muscle Function Studies-Intact EDL and soleus from mice of each genotype were used for collection of isometric force data using an eight-chambered system (25,26). Experiments were collected and analyzed with ADInstruments Power-Lab software customized for these experiments. Muscles were first equilibrated for 20 min to mimic conditions of normal activity (low duty cycle, ϳ1%). They were then subjected to the length-force relationship test to determine optimal length at which maximal force is achieved. Muscles were stimulated with frequencies ranging from 1 to 130 Hz to generate force versus frequency relationships. From the force versus frequency, the frequency producing maximal tetanic force (T max ) was then used for the rest of the experiments. Muscles were stimulated every minute with T max for 5-10 min to assure that the preparation was stable at which point the tissue bathing buffer was replaced with one containing 0 mM Ca 2ϩ ϩ 0.1 mM EGTA. The muscles were then stimulated every minute at T max to determine the impact of removal of external calcium on muscle force. After 20 min, the 0 mM Ca 2ϩ buffer was washed out and replaced by a buffer containing 2.5 mM Ca 2ϩ , and the muscles were stimulated every minute for 20 min at T max to allow for force recovery. The muscles were then were subjected to fatiguing stimulation protocols that mimic physiological fatigue (50% duty cycle; intermittent fatigue) (27,28). At the end of fatigue, muscles were allowed to recover by switching the stimulation parameter to a 1-min interval. At the end of a 30-min recovery period, muscles were treated with 5 mM caffeine, which effectively releases calcium from the sarcoplasmic reticulum to test in the intact whole muscle whether excitationcontraction coupling was defective at the end of the recovery process in any of the genotypes.
After force protocols, muscle dimensions and masses were measured for determination of cross-sectional normalized forces and for comparative morphometric analyses. After coating lightly in optimum cooling temperature medium, isolated muscles were flash frozen in liquid N 2 -cooled isopentane, sectioned frozen, and then analyzed for general histology by hematoxylin staining or for myosin heavy chain expression by immunofluorescence staining.
Isolation of Muscle and Osteocyte-enriched RNA, Labeling, and Processing for Whole Genome Array-Osteocyte-enriched mouse bone chips were prepared as described above and then treated sequentially four times with alternating EDTA treat-ments followed by collagenase digestions. Total RNA was isolated from frozen bone chips and from flash frozen muscle samples using an RNeasy Micro kit according to standard protocols along with the inclusion of a proteinase K digestion step. The RNA was analyzed for quality using a Nanodrop and an Agilent Bioanalyzer after which 100 ng of total RNA were used for the Agilent Mouse Transcriptome Array 1.0 protocol. Briefly, the total RNA was primed with primers containing a T7 promoter sequence. Single-stranded cDNA was synthesized with a T7 promoter sequence at the 5Ј-end, converted to doublestranded cDNA, and in vitro transcribed with T7 RNA polymerase to generate antisense RNA. The cRNA was cleaned up using a bead protocol and then used to synthesize sense strand cDNA in a reaction that included dUTP at a fixed ratio relative to dTTP. RNase H was used to hydrolyze the cRNA strand after which the sense strand cDNA was purified using beads and then fragmented at the unnatural dUTP residues using uracil-DNA glycosylase and apurinic/apyrimidinic endonuclease 1. The fragmented cDNA was labeled by terminal deoxynucleotidyltransferase using the Affymetrix proprietary DNA Labeling Reagent that is covalently linked to biotin. The biotinylated and fragmented cDNA was hybridized overnight to the Mouse Transcriptome 1.0 arrays after which it was washed and stained on a FS450 GeneChip system and scanned using a GCS30007G scanner.
The CEL files from muscle or bone osteocytes then were analyzed using Affymetrix Expression Console and Transcriptome Analysis Console 2.0. For detailed statistical analysis, the CEL files were imported into GeneSpring v13, quantile-normalized using the PLIER16 algorithm, and baseline-transformed to the median of all samples. The log 2 -normalized signal values were then filtered to remove entities that showed signal in the bottom 20th percentile across all samples. The above lists were subjected to a t test (p Ͻ 0.05) with a Benjamini-Hochberg false discovery rate correction. A 1.5-fold filter was applied to identify genes that were differentially expressed between control and cKO mice.
Quantitative PCR-Flash frozen muscle was disrupted in TRIzol reagent using a miniature motorized homogenizer. Total RNA was then isolated using a Direct-zol RNA MiniPrep Plus column (Zymo Research) and amplified using a High Capacity RNA to cDNA kit (Applied Biosystems). Mbtps1 and cre recombinase were quantitated using TaqMan Universal Master Mix II with uracil N-glycosylase (Applied Biosystems) and 6-carboxyfluorescein-labeled/minor grove binding probes and normalized relative to ␤-actin or Gadph.

Results
Mbtps1 mRNA and Protein Are Deleted in Osteocyte-enriched Bone but Not Skeletal Muscle-cKO bone osteocytes exhibited a ϳ70% reduction in Mbtps1 mRNA (array data deposited at NCBI under accession number GSE69975). In terms of protein, osteocyte-enriched control bone extracts contained an 18-kDa MBTPS1 fragment band also enriched in laser microscope-dissected biomineralization foci (Fig. 1a), sites of initial deposition of hydroxyapatite in osteoblastic cultures (29).
Other work 4 has shown that the relative content and half-life of this 18-kDa fragment in bone cells is increased after treatment with AEBSF, which blocks autolytic cleavage of MBTPS1. Consistent with cre recombinase-mediated deletion of Mbtps1, the 18-kDa MBTPS1 band is barely detectable in two of three Mbtps1 cKO bone extracts and noticeably reduced in a third specimen (Fig. 1a).
To determine whether Dmp1-cre was "leaky" in skeletal muscle, adult cKO and control soleus muscles (SOL) and EDL were processed for quantitative PCR. However, the relative content of Mbtps1 mRNA was not statistically different from that in littermate controls (Table 1). Whole genome arrays on SOL from four Mbtps1 cKO and control mice independently confirmed these findings (deposited at NCBI under accession number GSE69985). Furthermore, the expression of cre recombinase was also determined in adult cKO SOL and cKO EDL. Interestingly, no (n ϭ 4) or low (n ϭ 3) levels of cre were detected in cKO mice (Table 2). In contrast, ␤-actin and Gadph (not shown) expression was robust in all muscles tested. Where expressed, cre was present in both EDL and SOL from the same animal despite the fact that the muscle phenotype (see below) was restricted to cKO SOL. cre/actin Ct ratios for cKO EDL and cKO SOL were statistically indistinguishable ( Table 2).
We next asked whether the content of MBTPS1 protein was reduced in cKO SOL. Similar to the mRNA data, the content of an additional immunoreactive 65-kDa-molecular mass MBTPS1 fragment was unchanged when extracts from three separate cKO and littermate control soleus muscles were compared (Fig. 1b). A similar 65-kDa fragment is consistently observed in the medium fraction of osteoblastic cell cultures (Fig. 1b). Therefore, it is apparent that the 18-kDa fragment is enriched in bone osteocytes, whereas the 65-kDa form is present in muscle. Although MBTPS1 is a membrane-bound protease localized to the cis/medial Golgi, a soluble form can also be shed into the medium of kidney cells (30) and bone cells (13).
Long Bones from Adult Mbtps1 cKO Mice Are Stiffer but Not Different in Bone Mineral Content-In view of our use of bonerestricted Dmp1-cre, we initially carried out microCT scans on skeletons of Mbtps1 cKO and control mice. Surprisingly, when microCT data from adult tibiae were compared, no significant differences were found in mineral density, cortical thickness, trabecular number, trabecular density, trabecular thickness, bone volume/total volume, trabecular spacing, bone mineral density, or mineral content (Tables 3 and 4). To analyze biomechanical properties, three-point bending studies were also carried out on isolated femora. Importantly, Young's modulus for cKO femora averaged 7.38 Ϯ 1.65 (mean Ϯ S.D.) compared with 5.90 Ϯ 0.92 for controls, a 25% increase in stiffness (p ϭ 0.013). Because mineral densities were similar (Table 4), we hypothesize that this increase in stiffness reflects a change in organic bone matrix composition. Although neonatal Dmp1cre Mbtps1 cKO mice did not display a bone phenotype (not shown), we did note a characteristic increase in body weight starting abruptly at 15-20 weeks of age, the time at which the mouse approaches skeletal maturity (Fig. 2, a-d). However, when weight data from all litters were pooled at 10 -12 months, Mbtps1 cKO males weighed on average only 1.5 Ϯ 5.9 g more than controls (p ϭ 0.320). Because initial analyses of whole body fat could not account for this difference between Mbtps1 cKO and control mice (data not shown), we analyzed skeletal muscle and found the source of this weight gain.
Adult Mbtps1 cKO SOLs Are Larger and Contract More Forcefully Even When Normalized for Size-SOL and EDL, prototypical slow twitch and fast twitch muscles, respectively, were isolated from adult mice and subjected to contractile studies ex vivo (Fig. 3a). cKO SOLs were 12% larger in mass (p ϭ 0.02) (Fig.  3b) and quantitatively longer in length (data not shown). Interestingly, cKO SOL were also 30% stronger than control littermates both in terms of relative and specific contractile force (p Ͻ 0.05) (Fig. 3, c, d, and e). In contrast, when comparable studies were carried out on adult EDL from cKO and control littermates, no significant differences were observed in size, length, and contractile force (Fig. 3, b, f, and g, and data not shown).
The force-frequency curve for Mbtps1 cKO SOL was also shifted to the right of the control by a small but significant amount; e.g. at the lower frequencies, a higher frequency stim- Control, bones from three separate control littermate mice; cKO, bones from three knock-out mice; BMF, biomineralization foci isolated by laser capture microdissection from AEBSF-treated osteoblastic cultures (29). b, muscle extracts from cKO and control SOL were indistinguishable when immunoblotted for MBTPS1. C1-C3, soleus muscles from three littermate controls; cKO1-cKO3, soleus muscles from three knock-out mice; OB media, osteoblastderived culture media. Both immunoblots were treated with a monospecific polyclonal antibody reacting with the C-terminal region of mouse MBTPS1 (a gift from Dr. N. G. Seidah). Similar blotting results were obtained with a second anti-human MBTPS1 antibody (not shown). Lanes for biomineralization foci and culture media are provided for comparative purposes and were not imaged at the same time as bone or muscle extracts.
ulation is required to elicit the same maximal tetanic force as control SOL (Fig. 4a). In contrast, the responses of control and Mbtps1 cKO SOL were similar when fatigued and then challenged with caffeine (Fig. 4b) and upon depletion and replen-ishment of extracellular calcium (Fig. 4c). Force-frequency curves for control and Mbtps1 cKO EDL were indistinguishable (Fig. 4d) as were the results of fatigue and extracellular calcium depletion protocols (Fig. 4, e and f). In view of these changes in adult Mbtps1 cKO SOL, we asked whether young Mbtps1 cKO SOLs were also larger and more powerful than littermate controls.
Young Mbtps1 cKO SOLs Are Not More Powerful than Controls-Because Dmp1-cre Mbtps1 cKO mice exhibited an increase in weight gain at 15-20 weeks (Fig. 2), we asked whether this finding reflected the temporal onset of the muscle phenotype. Contraction studies were carried out on isolated SOL and EDL from young Mbtps1 cKO and control mice (Fig.  5) at 12 weeks of age, just prior to this observed increase in body weight. Based on this functional testing, cKO and control SOLs were not different in terms of size, force frequency relationship, contractile force, or fatigue properties (Fig. 5, a-e). Large error bars associated with contractile measurements in young SOL are likely due to a developmentally immature excitation-contraction coupling apparatus (31). However, young cKO EDL was about 25% more powerful than control EDL (Fig. 5h), a difference that did not persist in adults (Fig. 3, f and g).
Adult cKO SOLs Exhibit Centralized Nuclei, a Morphological Characteristic of Regenerating Muscle- Fig. 6 shows representative views of hematoxylin-and eosin-stained adult SOL from control and Mbtps1 cKO animals. Black arrows mark the posi-     tions of centralized nuclei (Fig. 6). It is apparent that bigger and more forceful cKO SOLs are enriched in myofibers with central nuclei (21.5 Ϯ 5.7%, mean Ϯ S.D.), whereas control SOL muscles contain background levels (3.5 Ϯ 2.9%, p Ͼ 0.001). A long standing correlation of centralized nuclei with muscle regeneration (32-36) was recently challenged by inconsistent findings in some transgenic muscle models (37,38). However, our results strongly support an association with muscle regeneration because Mbtps1 cKO EDLs, which are phenotypically unaffected, contain only background levels of central nuclei (4.0 Ϯ 2.9%) (Fig. 6, c and d). Young Mbtps1 cKO SOLs Are Not Enriched in Centralized Nuclei-Myofibers from young Dmp1-cre Mbtps1 cKO SOL and control muscles contain few centralized nuclei (Fig. 6, e-h). Quantitatively, only 3.4% Ϯ 3.2 of myofibers in cKO SOL contained centralized nuclei, which was not different from that for control SOL (3.4 Ϯ 2.2%, p ϭ 0.989). Likewise, the content of centralized nuclei in young Mbtps1 cKO and control EDL were indistinguishable and similar to background levels (5.2 Ϯ 2.1 versus 5.2 Ϯ 3.5%).

Myofibers Exhibiting Centralized Nuclei Preferentially Express Type I Myosin Heavy
Chain-To evaluate their distribution within SOL and EDL, type I, type IIA, and type IIB myosin heavy chains were immunofluorescently stained in frozen sections of adult Mbtps1 cKO and control muscles. Consistent with a published report (39), the composition of slow twitch control SOL was composed of 41.8% type Iϩ-expressing myofibers (magenta in color) and 54.5% type IIAϩ-expressing myofibers (green in color) ( Fig. 7a and Table 5). A small percentage of myofibers also stained for type IIB (3.8%), but these were absent from cKO SOL (Fig. 7b). DAPI-stained nuclei appear yellow in this image and were flattened around the inside periphery of most control myofibers (Fig. 7a). Importantly, 15.8% of type I myosin heavy chain-positive cells express central nuclei in cKO SOL muscles (Fig. 6b, arrows, magenta-colored myofibers), whereas, by comparison, only 3% of type IIA-expressing cells contain central nuclei ( Table 5). The sum of these two numbers (18%) agrees well with our prior estimate of 21.5% for cKO SOL after H&E staining (see Fig.  6b). cKO SOL contained similar amounts of type Iϩ myofibers and fewer type IIAϩ myofibers (Table 5), despite the fact that a prior report showed that regeneration of injured soleus muscle led to a selective enrichment in type Iϩ myosin heavy chain myofibers at the expense of type IIAϩ and type Xϩ myofibers (40). Although deficient in type Iϩ myosin heavy chain myofibers, adult fast twitch EDL cKO and control muscles were indis- tinguishable in terms of their contents of type IIA, type IIB, and type X myosin heavy chain expressing myofibers (Table 5).
Gene Expression in Bone and Mechanistic Insights Based on Bigger, More Forceful cKO SOL-To begin to determine the molecular mechanisms responsible for the observed bonemuscle cross-talk, whole genome arrays were performed on both muscle and osteocyte-enriched bone. In addition to a ϳ70% reduction in Mbtps1 mRNA, knock-out bone osteocytes exhibited a significantly reduced expression of 270 genes with only 10 genes increased relative to littermate con-trols (data deposited with NCBI under accession number GSE69975). Specifically, osteocyte markers Mepe, Sost, Phex, and Dmp1 were all reduced 3.0 -1.5-fold along with osteoblast-specific periostin and myogenic repressors Tgf-␤1, Tgf-␤2, and Tgf-␤3. Gene profiling identified alterations in several pathways in knock-out bone including TGF-␤ receptor signaling, TGF-␤ signaling, adipogenesis, pluripotency, and endochondral ossification (data not shown). Any individual factor or a combination of these factors may be responsible for the enhanced cKO soleus muscle. FIGURE 5. EDL and SOL contractile properties from young control and Mbtps1 cKO mice. a, EDL and soleus muscle wet weights from control and Mbtps1 cKO mice. b, force versus frequency curves of intact soleus muscles from control and Mbtps1 cKO mice stimulated at individual frequencies ranging from 1 to 130 Hz. Force is normalized to maximum muscle force (T max ) and expressed as a percentage. c, EDL muscle force versus frequency (1-130 Hz) curves from control and Mbtps1 cKO mice. Contraction force is normalized to T max of the muscle and expressed as a percentage. d, soleus muscle contractile force (millinewtons (mN)) from control mice and Mbtps1 cKO mice at increasing frequencies of stimulation in the range of 1-130 Hz. e, soleus muscle contractile forces from d normalized to cross-sectional area of the muscle (newtons (N)/cm 2 ). f, relative contractile force of soleus muscles from control and Mbtps1 cKO mice after a fatigue-inducing stimulation protocol and after a recovery period of non-fatiguing tetanic stimulations in both the absence and presence of 5 mM caffeine. Data are expressed as a percentage of force produced by the muscle prior to fatigue. g, contractile force (millinewtons) of intact EDL muscles from control and Mbtps1 cKO mice stimulated at frequencies of 1-130 Hz. h, EDL muscle forces from g normalized to muscle cross-sectional area (newtons/cm 2 ). i, EDL muscle relative contractile force from control and Mbtps1 cKO mice after a fatigue-inducing stimulation protocol and after a recovery period of non-fatiguing tetanic stimulations in both the absence and presence of 5 mM caffeine. Data are expressed as a percentage of force produced by the muscle prior to fatigue. Data are represented as mean Ϯ S.E. (error bars). * denotes significant difference (p Ͻ 0.05) when compared with control.
To identify metabolic changes that may explain the Dmp1cre Mbtps1 cKO muscle phenotype, we carried out whole genome arrays on adult SOL (deposited in NCBI under accession number GSE69985). Gene profiling revealed increased expression in pathways mediating EGF receptor signaling, circadian exercise, striated muscle contraction, glycolysis, fatty acid oxidation, and adipogenesis. Briefly, actinin-␣3 was elevated in cKO SOL; Actn3 is a biomarker for increased muscle performance in elite athletes (41). Other striated muscle contraction genes increased included mitsugumin 29 (Mg29), vimentin, desmin, and slow twitch muscle-restricted troponin. Embryonic and perinatal myosin heavy chains (Myh3 and Myh8), which have been used as biomarkers to identify muscle regeneration (42)(43)(44), were also increased. Myf6, which is associated with satellite cell activation during muscle regeneration (45), and Pax7, Myod1, and myogenin (Myog), which are required for postnatal growth and regeneration of adult skeletal muscle (46 -48), were also elevated in Mbtps1 cKO SOL. Metabolically, enhanced expression of triglyceride lipase in cKO SOL should facilitate release and oxidation of fatty acids. Finally, core E-box-dependent circadian clock genes Per1/2 and Cry1/2 were elevated in cKO SOL compared with controls, whereas core clock genes Bmal (Arntl), Dec1, and Dec2 were reduced in cKO SOL muscles, consistent with the known 12-h cycle offset for Bmal expression compared with that for Per1/2 and Cry1/2 (49). Taken together, the gene expression profile for cKO SOL is consistent with observed changes in size and function and remarkably resembles that for regenerating skeletal muscle.

Discussion
Male Dmp1-cre Mbtps1 cKO mice display an age-related slow twitch muscle phenotype with increased specific contractile force and size, which implies novel bone-muscle cross-talk signaling. Although microCT analyses revealed no differences in trabecular or cortical bone mineral density or morphological parameters between cKO and control tibiae, three-point bending studies on femora demonstrated a significant 25% gain in stiffness in Mbtps1-deficient bone (p ϭ 0.01). We rationalize this outcome in terms of a requirement for MBTPS1 for transcription of bone matrix genes including fibronectin, fibrillin2, and collagen XI (13). Because expression was not down-regulated in osteocyte-enriched bone (array data deposited at NCBI under accession number GSE69975), MBTPS1 regulation of these mineralization-related genes may be restricted to osteoblast-like cells (13).
Analysis of muscles from adult Dmp1-cre Mbtps1 cKO mice revealed several unexpected findings. cKO SOL displayed a 30% greater contractile force, 12% bigger mass, and 6-fold increase in centralized nuclei. Central nuclei are a well known morphological characteristic of regenerating muscle (32)(33)(34)(35)(36); however, the effect of conditional deletion of Mbtps1 in bone was restricted to type I myosin heavy chain-expressing myofibers. We interpret this finding as supporting a regenerative response in cKO SOL but not EDL from the same mice. Furthermore, EDL muscles from the same hosts do not show immunochemical evidence of a fast twitch (type II myosin) to slow twitch (type I myosin) transition ( Table 5). The latter finding distinguishes the response of Dmp1-cre Mbtps1 cKO muscle from that expected after endurance training (50).
Consistent with the morphological data, cKO SOL also expressed elevated levels of Myod1, Myf6, Myog, and Pax7 along with embryonic and perinatal myosin heavy chain Myh3 and Myh8 genes, which postnatally represent independent biomarkers of muscle regeneration (45)(46)(47)(48)(49)(51)(52)(53). Regeneration of type Iϩ and type IIϩ myosin heavy chain-expressing myofibers is regulated differently (54) but originates by activation of distinct populations of satellite cells within each muscle (9, 10). In contrast to adult mice, the contractile force and size of young SOL cKO and control muscles were indistinguishable, indicating that the response of SOL to deletion of Mbtps1 in bone occurs after 12 weeks of age.
Although the muscle phenotype is restricted to cKO SOL, analyses of Mbtps1 mRNA and protein content in cKO SOL were indistinguishable from controls. Also, cre recombinase was only detectable in less than half of Mbtps1 cKO mice. Furthermore, when detected, cre expression was equivalent in both cKO EDL and cKO SOL even though changes in muscle function and gene expression were localized to the latter. Taken together, these findings are not consistent with leakiness of the Dmp1-cre in cKO SOL but rather support an indirect bonemuscle cross-talk mechanism.
Transgenic models with a muscle phenotype can be classified into two groups: 1) specific contractile force is either unchanged or 2) specific contractile function is decreased. For example, myostatin-null mice gain mass in both their EDL and SOL; however, functional testing shows that specific contractile force is less than controls for myostatin cKO EDL and the same as controls for cKO SOL (55)(56)(57). Conditional deletion of IGF-1 in muscle was found to be selective for fast twitch EDL where mass and maximum power increased about 30% (58). However, specific contractile force was the same in transgenic and control EDL muscles. Furthermore, a double knock-out of Myod and dystrophin and a knock-out of calpain 3 (59, 60) both display a reduction in muscle contractile force. In contrast, the adult Dmp1-cre Mbtps1 cKO muscle phenotype represents a special case because SOLs exhibit an increase in both specific contractile force and size.
To identify candidate signaling pathways mediating bonemuscle cross-talk, we compared gene array data for adult bigger and more powerful cKO SOL with that for control SOL. Interestingly, expression of basic helix-loop-helix domain-contain-ing transcriptional regulators Dec1 (Bhlhe40, Stra13, Bhlhb2) and Dec2 (Bhlhe41, Sharp1, Bhlhb3) was decreased significantly in cKO SOL compared with controls at the time the mice were euthanized. Bhlhe40 and Bhlhe41 are core circadian clock genes that like Per and Cry are positively regulated by Bmal and Clock (61,62); in return, expression of Dec1 and Dec2 is subject to negative feedback (62,63) (Fig. 8a). Dec1 and Dec2 expression in muscle displays a rhythmic 24 -28-h cycle (64), which because of negative feedback regulation would be predicted to be offset by ϳ12-24 h compared with that for Bmal and Clock (Fig. 8, a and b). Bmal expression is required for satellite cell activation in muscle regeneration (65). Dec1/Dec2, in contrast, were shown previously to repress myogenesis in vitro and in vivo by blocking transcription of Myod1 as well by altering expression of a number of muscle contractile and mitochondrial proteins (66 -68). By analogy with the age-related onset of the Dmp1-cre Mbtps1 cKO phenotype here, Dec2-null mice show no deficit with respect to repair of embryonic muscle; however, as expected for a repressor of myogenesis, regeneration after injury in post-natal Dec2(Ϫ/Ϫ) mice is increased (68). Postnatally, Dec1 represses Notch signaling, which in null mice leads to a complex muscle phenotype (69) (Fig. 8b).
Because SOLs in Dmp1-cre Mbtps1 cKO mice display characteristics of regenerating muscle, we hypothesized that a cycling circadian expression pattern for myogenic repressors Dec1/Dec2 could stimulate or reduce expression of pro-or antimyogenic genes, respectively, during their "down cycle" (Fig. 8,  a and b). Because Lecomte et al. (66) have already identified positive and negative transcriptional muscle-specific targets of Dec1/Dec2, we asked whether expression of these target genes is temporally correlated with that of transcriptional regulators Dec1/Dec2? Importantly, the majority of muscle-specific target genes (79 of 110) displayed the predicted response (66) when Dec1/Dec2 were down-regulated (Table 6). Because expression FIGURE 7. Only type I myosin heavy chain-expressing cKO soleus muscle myofibers exhibit centralized nuclei. Muscles from adult Dmp1-cre Mbtps1 cKO and control mice were dissected immediately following euthanasia and flash frozen, and frozen sections were immunostained as described under "Experimental Procedures." Images of whole muscle cross-sections were pseudocolored using ImageJ. cKO soleus muscle is enriched in type I myosin heavy chain-expressing cells (magenta-colored myofibers) with centralized nuclei (white arrows), whereas type I myosin heavy chain-positive and type IIA heavy chain-positive (green-colored myofibers) control cells contain only background levels of centralized nuclei. Nuclei were identified by DAPI staining (yellow color). a, control soleus muscle; b, Dmp1-cre Mbtps1 cKO muscle. Images shown were photographed at 10ϫ magnification. (Please note that average quantitative results of fiber typing and centralized nucleus counts for SOL and EDL muscles for cKO and control muscles are presented in Table 5.) of both Dec1 and Dec2 is low in Dmp1-cre Mbtps1 cKO SOL (array data deposited at NCBI under accession number GSE69985), this is the expected outcome. Although our transcriptional data represent a single time point and not a systematic timed study of circadian regulation, we believe the high degree of correlation between predicted and observed target gene expression observed in Table 6 supports an indirect role for Dec1/Dec2 in the regeneration and growth of Dmp1-cre Mbtps1 cKO SOL (Table 6 and Fig. 8). Specifically, we propose that bone-muscle cross-talk facilitates circadian cycling of expression of promyogenic genes (including Pax7, Myod1, and Myog) sufficient to form new myofibers from muscle progenitor cells (Table 6 and Fig. 8, a and b). PAX7 is required for specification of satellite cells from muscle-derived stem cells (47). MYOD1, in turn, generates myogenic precursor cells, which are induced to terminally differentiate to myotubes by MYOG (48). We envision that cyclical waves of expression of these promyogenic genes drives progression of satellite cells through the different stages of muscle specification and differentiation while accommodating for the known cross-inhibitory interactions of individual factors (47,70).
Gene profiling identified several other signaling pathways that are more highly expressed in Mbtps1 cKO SOL: EGF receptor signaling, striated muscle contraction, circadian exercise, fatty acid ␤-oxidation, glycolysis, and adipogenesis. We hypothesize that these pathways reflect major metabolic adaptations, e.g. an EGF receptor pathway regulating myoblast differentiation (71), a pathway improving contractile force, and a pathway enhancing oxidative catabolism of fat and sugar nutrients. Increased expression by Mbtps1 cKO SOL of a number of genes can be rationalized in terms of improving contractile performance. For example, expression of actinin3, titin, and desmin is increased compared with controls. Z-disc component actinin3 is strongly associated with improvements in sprint or endurance (72,73), whereas titin is also localized to the Z-disc, which mechanically couples sarcomeric contraction and stretching (74,75). Desmin is required for the tensile strength and integrity of myofibrils (76). Remarkably, Mg29 is also overexpressed in cKO SOL. Mg29 (also known as synaptophysin-like 2) is essential for proper excitation-contraction coupling in skeletal muscle because null mice have reduced muscle performance due to failures in excitation-contraction  Table 6.) b, diagram illustrating the complex inter-relationships and feedback loops that interconnect transcriptional regulators DEC1/DEC2 with circadian core clock genes (Clock/Bmal), MBTPS1-activated transcription factor SREBP1, and NOTCH activation of muscle satellite cells (see "Discussion" for references). coupling, a finding that is also mirrored in aged skeletal muscles, which have an ϳ50% reduction in the protein content of MG29 (77,78). In contrast, improved muscle performance in an animal model where the MG29 level was up-regulated has been reported (79). Therefore, the up-regulation of Mg29 along with the up-regulation of other muscle performance genes helps explain the increased contractile performance of SOL muscles from cKO mice. Consistent with their renewal of type I-expressing myofibers, cKO SOLs also have increased expression of key enzymes that mediate oxidative metabolism of fats and sugars, e.g. adipose triglyceride lipase, adiponectin, phosphoenolpyruvate carboxykinase, and phosphofructokinase.
In summary, male Dmp1-cre Mbtps1 cKO mice display an age-dependent slow twitch muscle phenotype with increased specific contractile force and size. All available evidence supports bone-muscle cross-talk as the cause. Furthermore, gene expression changes in cKO soleus muscles appear entirely consistent with the observed structural and functional changes. Better understanding of bone-muscle cross-talk may provide a fresh and novel approach for the prevention and treatment of age-related muscle loss.

that are consistent with a release from transcriptional regulation
Soleus muscles from Dmp1-cre Mbtps1 cKO and control mice were isolated and processed for whole genome arrays as described under "Experimental Procedures" (deposited at NCBI under accession number GSE69985). Genes exhibiting significant differences in expression were then compared with the list of muscle-specific DEC1/DEC2 target genes identified by Lecomte et al. (66).
a Expression of genes in this column, which are repressed by BHLHB2/BHLHB3, were increased in soleus muscles from Dmp1-cre Mbtps1 cKO mice compared with littermate control muscles. b Expression of genes in this column, which are activated by BHLHB2/BHLHB3, were decreased in soleus muscles from Dmp1-cre Mbtps1 cKO mice compared with littermate control muscles.