|Home | About | Journals | Submit | Contact Us | Français|
The intracellular Ca2+ ([Ca2+]i) level of skeletal muscles must be rapidly regulated during the excitation-contraction-relaxation process 1. However, the signaling components involved in such rapid Ca2+ movement are not fully understood. Here, we report that mice deficient in the novel phosphatidylinositol phosphate (PIP) phosphatase MIP displayed muscle weakness and fatigue. Muscles isolated from MIP−/− mice produced less contractile force, markedly prolonged relaxation, and exhibited exacerbated fatigue. Further analyses revealed that MIP deficiency resulted in spontaneous Ca2+ leak from the internal store — the sarcoplasmic reticulum (SR). This was attributed to the decreased metabolism/dephosphorylation and the subsequent accumulation of MIP substrates, especially PI(3,5)P2 and PI(3,4)P2. Furthermore, we found that PI(3,5)P2 and PI(3,4)P2 bound to and directly activated the Ca2+ release channel/ryanodine receptor (RyR1) of the SR. These studies provide the first evidence that finely controlled PIP levels in muscle cells are essential for maintaining Ca2+ homeostasis and muscle performance.
During our systematic genome-wide survey for tyrosine/dual specificity phosphatases (unpublished work), we discovered a novel phosphatase by hidden Markov database mining using the conserved catalytic motif ([V/I][V/I]HCXXGXXR[T/S]) as the bait sequence. Both human (BC035690) and mouse (BC018294) homologies were identified. They share 90% identity in amino acid sequences (Supplementary Information, Fig. S1). Northern blotting analyses illustrated that this phosphatase was predominantly expressed in skeletal muscle and heart (Fig. 1a). Immunostaining indicates that it is primarily localized in the cytoplasm (data not shown). To verify its phosphatase property, we generated a GST fusion protein and tested its catalytic activity using pNPP (p-Nitrophenyl Phosphate), a widely used non-specific substrate for tyrosine/dual specificity phosphatases. This new phosphatase did not significantly hydrolyze pNPP (Fig. 1b). Instead, it dephosphorylated a variety of PIPs, especially PI(3,5)P2 (Fig. 1c), similar to PTEN and myotubularin and myopathy related (MTMR) phosphatases that also favor PIPs as substrates despite containing tyrosine phosphatase domains 2. As this new phosphatase is mainly expressed in skeletal muscle and heart, we named it MIP (muscle-specific inositol phosphatase). While our gene knockout work on MIP was ongoing, the Mustalin group also identified this phosphatase (FLJ20133) in their comprehensive collection of tyrosine phosphatases from the human genome and listed it as the 14th member of the MTMR family (MTMR14) based on the homology of its catalytic motif to myotubularin 3. More recently, inactivating mutations in this phosphatase were identified in two cases of human autosomal centronuclear myopathy 4. However, whether these mutations play a causal role in this disease has not yet been determined.
To characterize the physiological function of MIP, we generated MIP deficient mice through gene targeting (Supplementary Information, Fig. S2). MIP−/− mice were born at the Mendelian ratio and showed no obvious deficiencies during the first 12 months of life. It appears that heart development and basal functions of MIP−/− mice under resting conditions were not significantly affected (Supplementary Information, Table S1 and S2). Interestingly, animal behavior tests revealed a decreased motor function in MIP−/− mice (Supplementary Information, Fig. S3). We then reasoned that a muscle specific phenotype might manifest itself more evidently during or after a stress response. Thus, untrained 16 to 24-week old female wildtype (WT) and MIP−/− mice were run in a rodent treadmill until exhausted following the protocol previously reported 5. The running time of the knockout mice was dramatically decreased (Fig. 2a). These strenuous exercise results indicate that MIP−/− mice are prone to greater skeletal muscle fatigue. However, treadmill performance does not necessarily reflect intrinsic changes within the essential contractile features of intact skeletal muscles and may be confounded with other systemic changes. To directly measure muscle function, extensor digitorum longus (EDL), a fast-twitch muscle, was isolated from the animals used in the treadmill experiments 1–2 weeks after the fatigue tests and assessed for contractility and fatigability characteristics. As illustrated in Fig. 2b, the force-generating capacity (magnitude of the maximal fused isometric tetanic contraction, Tmax) of EDL muscles from the MIP−/− mice was reduced by ~ 60% compared to the control muscles. In addition, the relaxation profile of MIP−/− muscles was substantially prolonged. To further determine the performance of isolated muscles, in vitro intermittent fatigue of the muscles 6,7 was studied. Compared to WT controls, upon exposure of MIP mutant muscles to repeated electrical stimulations, resting baseline force increased, while maximum tetanic force quickly decreased and muscles became non-responsive to subsequent stimulations (Fig. 2c).
To systematically compare fatigability and recovery from fatigue in WT and MIP−/− muscles, we tested the effects of in vitro fatiguing stimulation and recovery from fatigue in the slow-twitch, fatigue-resistant soleus muscle. We found that soleus muscles from MIP−/− mice fatigued to a greater extent and recovered significantly less than soleus from WT animals (Fig. 2d and 2e). In addition, the kinetics of single tetanic contractions were altered in soleus muscles from MIP−/− mice as seen in Fig. 2f. The plateau phase of the contraction was not well maintained in MIP−/− muscles. We next determined the force-generating capacity (Tmax) of soleus muscles from MIP−/− mice and found that Tmax of MIP−/− soleus was reduced by ~15% (Fig. 2g). Finally, the normalized force vs. frequency relationship was shifted to the right in MIP−/− muscles (Fig. 2g), i.e. at any given frequency of stimulation less force was generated. These results obtained from isolated intact muscles reaffirm that the contractility and relaxation defects of MIP−/− muscles are muscle cell autonomous. Remarkably, the muscle phenotypes appear to be exacerbated in aged mutant mice. About 30% of MIP−/− mice older than 18 months displayed accelerated muscle wasting/atrophy compared to WT littermates (Fig. 2h).
The poor performance of MIP−/− muscles may indicate a defect in the regulation of [Ca2+]i since rapid cycling of [Ca2+]i is critical for force production, relaxation, and recovery from fatigue 1. As Ca2+ handling in muscle cells primarily takes place in a highly specialized junctional structure known as the triad junction that is formed by the transverse tubule invagination of the plasma membrane and the terminal cisternae of SR 8, we surveyed morphology of this anatomic structure in mutant muscle cells. Electron microscopic examination revealed that the terminal cisternae of SR were slightly swollen (Fig. 3a). Ultrastructures of mutant cells otherwise appeared normal.
To functionally test whether MIP−/− muscle cells might have an impaired Ca2+ signaling, we prepared primary myotubes and assessed [Ca2+]i by the ratiometric fluorescence technique. As shown in Fig. 3b and 3c, the basal F340/F380 ratio in MIP−/− myotubes was higher than that in WT cells, indicating elevated resting [Ca2+]i levels in MIP−/− myotubes. Changing the extracellular buffer (ECB) from Ca2+-containing (1.5 mM) to Ca2+-free ECB in tandem with the addition of caffeine and ryanodine (which are known to induce sustained opening of the RyR channel 9,10) (Fig. 3b) or thapsigargin (TG), a potent inhibitor of the SR Ca2+ pump (Sarco/Endoplasmic Reticulum Ca2+-ATPase, SERCA) that inhibits the reuptake of Ca2+ into the SR 11 (Fig. 3c), resulted in a rapid and transient elevation of [Ca2+]i followed by loss of Ca2+ from the myotubes. Under these conditions, the magnitude of the initial Ca2+ transient was reduced in MIP−/− cells as compared to WT myotubes (Fig. 3b and 3c), suggesting decreased Ca2+ content of the mutant SR. Moreover, the duration of the decaying phase was markedly prolonged in the mutant cells, indicating a prolonged release or a defective clearance of [Ca2+]i in the mutant cells.
Store operated Ca2+ entry (SOCE), the capacitative Ca2+ entry through store-operated Ca2+ channels on the plasma membrane following the depletion or the decrease of Ca2+ from internal stores, is a universal mechanism that provides a direct way of refilling intracellular Ca2+ stores 12. Since MIP−/− muscles had lower levels of SR Ca2+, we reasoned that a compromised SOCE could contribute to the reduced Ca2+ storage capacity in MIP−/− muscles. In fact, our data showed that SOCE was decreased in the mutant myotubes, as evidenced by the reduced [Ca2+]i response to the restoration of 2 mM extracellular Ca2+ in the ECB medium (Fig. 3b and 3c). Subsequent treatment with SKF-96365, a known blocker of the store-operated Ca2+ channel 13 in the plasma membrane, quickly decreases [Ca2+]i concentration. By comparing [Ca2+]i traces of WT and MIP−/− cells (Fig. 3b and 3c), it is clear that while SOCE is still functional in MIP−/− myotubes, this response is significantly blunted.
To further determine if SR Ca2+ storage is reduced in MIP−/− adult muscles, we prepared single intact flexor digitorum brevis (FDB) fibers and exposed them to ionomycin (Ca2+ ionophore), an experimental condition designed to assess total Ca2+ storage within the SR 14,15. As shown in Fig. 3d and 3e, ionomycin-mobilized Ca2+ was greatly decreased in MIP−/− muscle fibers. In addition, we found that resting levels of [Ca2+]i were increased in MIP−/− fibers (Fig. 3d and 3f), consistent with earlier observations in myotubes (Fig. 3b and 3c). Thus, ablation of MIP causes diminished Ca2+ storage within the SR and an elevated [Ca2+]i, findings that are fully consistent with and supported by the phenotypic changes of muscle weakness and impairment of muscle relaxation in MIP−/− mice (Fig. 2b, 2c, 2d, and 2f).
MIP dephosphorylates multiple PIPs (Fig. 1d), especially PI(3,5)P2, the most recently identified phosphatidylinositol-bisphosphate (PIP2) isomer whose subcellular localization and function in mammalian cells have not been characterized 16,17. To test whether the dysfunction of store-operated Ca2+ signaling in MIP deficient cells is associated with decreased dephosphorylation/metabolism of PI(3,5)P2, we first determined subcellular localization of PI(3,5)P2 in myotubes. Confocal microscopic examination following immunostaining showed that it resided in the SR, as evidenced by co-localization with the endoplasmic reticulum specific protein disulfide isomerase (PDI) (Fig. 4a). We then assessed PI(3,5)P2 levels in MIP−/− SR by immunostaining of myotubes with anti-PI(3,5)P2 and anti-PDI antibodies followed by laser scanning cytometric analyses 18,19. As shown in Fig. 4b, immunofluorescence intensity of PI(3,5)P2 in MIP−/− SR defined by PDI positive areas is doubled compared to that in WT counterparts, confirming excessive accumulation of PI(3,5)P2 in the mutant SR. We next dialyzed WT and MIP−/− myotubes with 200 nM PI(3,5)P2 or PI(5)P, the hydrolyzing product of PI(3,5)P2, and monitored Ca2+ signaling. As illustrated in Fig. 4c, perfusion of PI(3,5)P2 resulted in dysfunction of the SR in WT myotubes, recapitulating the defects of Ca2+ signaling in MIP−/− cells (Fig. 3b and 3c). The decaying phase of the Ca2+ transient triggered by TG was substantially prolonged by PI(3,5)P2 overloading. By contrast, addition of PI(5)P did not rescue Ca2+ signaling in MIP−/− cells, nor were the Ca2+ profiles in WT cells changed (Fig. 4d). Furthermore, we tested other PIPs that can be dephosphorylated by MIP. As shown in Fig. 4e, overloading of PI(3,4)P2 and PI(3)P (to a less extent), but not PI(4,5)P2 and PI(3,4,5)P3, also prolonged the decaying phase of TG-triggered Ca2+ transients in WT myotubes. Overloading of these PIPs into MIP−/− cells either did not or only slightly increased Ca2+ signaling defects (Fig. 4f). Together, these PIP perfusion results support the concept that dysfunction of the Ca2+ store (SR) in MIP deficient muscle cells is attributed to the increased levels of PIP substrates, especially PI(3,5)P2 and PI(3,4)P2.
To define the molecular mechanisms underlying the disrupted Ca2+ homeostasis in MIP−/− cells, we assessed SERCA ATPase activity that is responsible for reuptake of Ca2+ into the SR, as reported 20,21. We also tested for the contribution of a potential functional change in Na+/Ca2+ exchangers in the plasma membrane to the phenotypes by treating myotubes with KB-R9743, a selective inhibitor for the reverse mode of the Na+/Ca2+ exchanger 22. Neither SERCA ATPase nor the Na+/Ca2+ exchanger appeared to be affected in MIP−/− cells (Supplementary Information, Fig. S4). These data thus led us to the hypothesis that Ca2+ stores in MIP−/− cells were spontaneously leaky. To test this hypothesis, we focused on potential functional changes in RyR1, the skeletal muscle Ca2+ release channel on the SR 23,24. First, we tested potential physical interactions of MIP PIP substrates with the RyR1 protein by PIP array analyses. As shown in Fig. 5a, PI(3)P, PI(5)P, and PI(3,5)P2 clearly bound to purified RyR1. In addition, PI(4)P and PI(3,4,5)P2 showed weak binding to RyR1. The direct physical interaction between PI(3,5)P2 and RyR1 was further verified by the PI(3,5)P2-agarose beads pull-down assay and anti- PI(3,5)P2 co-immunoprecipitation analyses (Fig. 5b). We then determined whether the interactions of PIPs with RyR1 on the purified SR microsomes produced functional effects. We used the [3H]ryanodine binding assay to assess RyR1 channel activity because [3H]ryanodine binds with high affinity to the open conformational state of the channel 25,26. The dose-response curve conducted at near-optimal (pCa 5) free Ca2+ (Fig. 5c) showed that among all PIPs tested, PI(3,5)P2 had the most effective activation, even though its physical interaction with the RyR1 protein was weaker than PI(3)P and PI(5)P (Fig. 5a). In addition, PI(3,4)P2 also activated RyR1 despite very weak binding to RyR1 protein. The observation that the RyR1 binding affinity of PIPs does not directly correlate with their effects on activation of the RyR1 channel suggests that some RyR1-PIP interactions are functionally more productive than others and that each PIP elicits a distinct re-arrangement of the channel protein that has different repercussions on channel gating. The complexity of PIP function in regulating ion channels and transporters is exemplified by PI(4,5)P2, the best characterized PIP. The effects of PI(4,5)P2 on various channels and transporters include activation, inhibition, and insensitivity 27. Clearly, further systemic studies are needed to determine the structure-activity relationship of the PIPs on the RyR1 channel function.
We next conducted Ca2+-dependence of [3H]ryanodine binding curves to determine whether PI(3,5)P2 activates RyR1 at free Ca2+ concentrations compatible to those found in myofibers at rest (pCa 7) or during contraction (pCa 6-5). The Ca2+-dependence of [3H]ryanodine binding to RyR1 has been well established. In the range of pCa 9 (10−9 M) to pCa 5 (10−5 M), Ca2+ has an activating effect and increases [3H]ryanodine binding to RyR1, whereas in the range of pCa 4 to pCa 2, Ca2+ has an inactivating effect. This dual effect of Ca2+ gives rise to a characteristic bell-shaped curve (Fig. 5d), which is similar to the Ca2+-Po (open probability) relationship of single RyR1 channels 25,26. In the presence of PI(3,5)P2, the binding curve maintained the same bell shape; however, the absolute binding values were dramatically increased, especially between pCa 7 and pCa 4. These results strongly suggest that the partial Ca2+ depletion of SR in MIP−/− muscle cells may be caused by direct activation of the RyR1 Ca2+ release channel by accumulated MIP PIP substrates.
Finally, we directly tested the functional significance of the interaction between PI(3,5)P2 and RyR1. PI(3,5)P2 was added to RyR1 reconstituted lipid bilayers. A nominally-free Ca2+ solution (~5 μM contaminant Ca2+) in the cis (cytosolic) side served to activate RyR1 (Fig. 5e). Before addition of PI(3,5)P2, NPo, the probability of the RyR1 channel being open (P0) times the number of channels in the recording, was low (1.15±0.3, n=3). Addition of 30 and 60 μM PI(3,5)P2 induced a significant increase in open events, increasing NP0 to 1.92±0.4 and 2.6±0.6 (n=3), respectively (Fig. 5e, and Supplementary Information, Fig. S5a). The effect of PI(3,5)P2 on the RyR1 channel was even more dramatic at a low concentration of free Ca2+ (pCa7) (Fig. 5f), suggesting an enhancement of Ca2+ leak and Ca2+-induced Ca2+ release (CICR) activities by PI(3,5)P2. The basal activity of the RyR1 channel was minimal under this condition (Po = 0.08±0.03, n=3) (Fig. 5f and Supplementary Information, Fig. S5b), but increased to Po = 0.28±0.04 and 0.51±0.11 (n=3) immediately after addition of 30 and 60 μM PI(3,5)P2, respectively (n=3). Unitary conductance (~700 pS) remained unchanged after addition of PI(3,5)P2. These electrophysiological analyses demonstrate that PI(3,5)P2 binds to RyR1 and directly modifies channel activity at contracting (pCa 5) and resting (pCa 7) levels of Ca2+, suggesting that the enhanced CICR and increased Ca2+ leak observed in MIP−/− cells may be due, at least in part, to the increased PI(3,5)P2 levels resulting from MIP absence.
In summary, our studies provide the first evidence that finely controlled PIP levels in the SR membrane and possibly the sarcolemmal membrane are critical for the proper function of the SR as the Ca2+ storage compartment and in rapidly regulating intracellular Ca2+. However, the mechanisms that regulate MIP activity and PIP levels in normal skeletal muscle physiology remain to be addressed. Since PIPs, especially PI(3,5)P2, are also the major substrates of other phosphatases of the MTMR family 2, dysregulated metabolism/dephosphorylation of these PIPs may be responsible for MTMR phosphatase-associated muscle disorders in general. Further exploration of the PIP pathway involved in intracellular Ca2+ homeostasis may provide new therapeutic targets for these muscle diseases.
The MIP allele was targeted by homologous recombination. Exon 3 containing the start codon ATG to Exon 5 was replaced with the Neo cassette. The targeting vector was constructed using the ‘recombineering’ technique. It was then electroporated into D1 mouse embryonic stem (ES) cells with the 129S6 × C57BL/6J hybrid background. G418 resistant ES clones were isolated and screened by PCR genotyping using the primers as shown in Supplementary Information, Fig. S2a. Two ES cell clones containing a correctly targeted MIP allele were used to generate chimeric mice. Three germline transmitted chimeric mice were obtained and used to cross C57BL6J mice to produce heterozygous mice. MIP+/− mice were backcrossed with C57BL6J mice for 3 generations and F4 generation of mice were used for the studies. No differences in the two lines of mutant mice derived from the original two ES cell clones were detected.
We tested the in vitro contractility properties of intact, anatomical muscles from the animals as previously described 5,28. WT and mutant EDL and soleus muscles were dissected intact and placed into 2.5 mM Ca2+ modified Ringer solution (142 mM NaCl, 5 mM KCl, 2.5 mM CaCl2, 2 mM MgCl2, 10 mM HEPES, 10 mM glucose, pH 7.4) bubbled with 100% O2. We also compared our results using a bicarbonate buffer system and did not observe significant differences. After careful dissection of the intact muscles, a pair of control and mutant intact muscles were mounted vertically onto 20-ml Radnotti glass chambers with built-in platinum stimulating electrodes (Radnoti Products, Monrovia, CA). One tendon of the muscle was attached to a force transducer and the other one to a stationary arm in preparation for in vitro fatiguing experiments. A Powerlab computer-interface program (AD Instruments, Colorado Springs, CO) was used to control the electrical stimulation protocols and to record, digitize, and store force output data. Field stimulation (squared-waves electrical currents of 500-msec duration, 300 mA using frequencies of stimulation in the range of 1–180 Hz) was accomplished with platinum electrodes running on both sides of intact muscles. After mounting, WT and MIP−/− muscles were mounted in parallel, and their resting lengths were adjusted to produce maximal isometric twitch force. The muscles were then subjected to the force vs. frequency relationship, and the stimulating frequencies that produced maximal isometric tetanic force (Tmax) were determined. Frequencies that produced Tmax were then employed for the remaining of the protocols. After 20 min of equilibration (1-min interval, 0.83% duty cycle) at a high frequency, muscles were exposed to a fatiguing stimulation protocol (1-sec interval, 50% duty cycle) using the same frequency of stimulation for 5 min. The fatiguing protocol used in this study was modified from our previous protocol 6,7. Force output data during fatigue were normalized to Tmax before the onset of fatigue and analyzed by Origin software (OriginLab Corp. Northampton, MA). Tmax was normalized to the cross sectional area of each muscle and is reported in kPa. All experiments were conducted at room temperature (23 ± 2°C).
Primary myoblasts were isolated from hind-limb muscles of WT and MIP−/− neonates, grown and differentiated into myotubes as described previously 29. Three independent isolates of each genotype were analyzed. Spatial and temporal distribution of intracellular Ca2+ concentration in individual myotubes was determined as previously reported 10,30. Myotubes derived from primary myoblasts after 4–5 days of differentiation were loaded with 5 μM Fura-2-AM for 45 min at 25°C in ECB (130 mM NaCl, 5 mM KCl, 1.5 mM CaCl2, 1 mM MgCl2, 25 mM HEPES, pH7.5, 1 mg/ml BSA, and 5 mM glucose). The buffer was replaced with fresh ECB and the incubation was continued for 45 min at 25°C to permit deesterification. A dual-wavelength spectrofluorometer (excitation at 340 nm and 380 nm) was used to determine the kinetic changes of [Ca2+]i concentration, which is reflected by the average ratio of F340/F380 of all cells (4–5 cells) in the field. Changes in intracellular Ca2+ were measured for 8 min following exposure to 10 μM TG or 10 mM caffeine plus 1 μM ryanodine in the Ca2+-free balanced salt solution (ECB was deficient in Ca2+ and supplemented with 0.5 mM EGTA). The extracellular bath solution was then changed with ECB with 2 mM Ca2+ and the average ratio of F340/F380 was continuously monitored. In some cases, after 3 min, the extracellular buffer was changed to ECB containing 50 μM or 100 μM SKF-96365. The [Ca2+]i concentration was monitored for additional 8 min.
For determination of resting cytosolic Ca2+ and total SR Ca2+ content in intact muscle fibers directly dissected from adult mice, FDB fibers were used for the experiments. In brief, FDB fibers were isolated by enzymatic disassociation in 0.2% type I collagenase (Sigma-Aldrich, St. Louis, MO). Mean FDB fiber size was 1 mm × 20 μM. FDB fibers were loaded with 10 μM Fura-2 AM for 45 min at room temperature in Tyrode solution. N-benzyl-p-toluene sulphonamide (20 μM), a myosin II inhibitor, was applied for 15 min to prevent motion artifact from muscle contraction 31. Fibers were also embedded into silicone grease to maintain their position in the culture dish 32. The ratio of Fura-2 fluorescence at excitation wavelength of 350 and 380 nm was measured on a PTI spectrofluorometer (Photon Technology International) to assess the [Ca2+]i level. Total SR Ca2+ storage was measured by addition of 5 μM ionomycin in the presence of 2 mM CaCl2 in the bath solution.
[3H]ryanodine binding to purified rabbit skeletal muscle SR was carried out as previously described 25,26. The incubation medium contained 0.2 M KCl, 40 mM Na-HEPES (pH 7.2), 7 nM [3H]ryanodine, 1 mM EGTA and various CaCl2 concentrations necessary to set free Ca2+ [Ca2+] from 10 nM to 10 mM. Sixty μg of mouse skeletal muscle SR-enriched microsomes and PI(3,5)P2 were added into the incubation medium to the indicated concentrations. Samples (0.1 ml) were incubated at 37 °C for 90 min. Mathematical fitting of data was accomplished with the computer program Origin (version 7.0, Micro-cal Inc., Northampton, MA).
Purified rabbit skeletal muscle RyR1 protein was reconstituted into Muller-Rudin planar lipid bilayers as described previously 25,26. Single channel data were collected at steady voltages (+35 mV, cis chamber grounded) for 2 min in symmetrical 300 mM cesium methane-sulfonate and 10 mM Na-HEPES (pH 7.2). The recording solution contained ~5 μM free Ca2+ as assessed by a calibration curve, which was sufficient to activate the RyR1 channel. Different concentrations of PI(3,5)P2 were added to the cis chamber, which corresponded to the cytosolic side of the channel. Signals were digitized at 4 kHz and analyzed after filtering with a low-pass 8-pole Bessel filter at a sampling frequency of 1.5 kHz. Data acquisition and analyses were performed with Axon Instruments software and hardware (pClamp version 8.03, Digidata 200 AD/DA interface). The full and non-conducting current values were obtained from Gaussian fits to the amplitude histograms, as previously described 24,25.
We are very grateful to Mr. David Chess, Dr. Margaret Chandler, Ms. Tammy Stefan, Dr. James Jacobberger, Dr. Leticia Brotto, and Dr. Jianjie Ma for technical assistance and helpful discussions. Motor function tests were performed by the Case Western Reserve University Rodent Behavior Core. This work was supported by The National Institutes of Health grants HL068212 and HL082670 (to C.K.Q.), HL55438 (to H.H.V.), The American Heart Association grant 0535555N (to M.B.), and a pilot grant from Case Center for Transdisciplinary Research on Energetics and Cancer (to T.M.N., C.K.Q., and M.B.).
AUTHOR CONTRIBUTIONSJ.S., W.M. Y., M.B., J.A.S., and C.S. conducted the research and summarized the data. C.K.Q., M.B., H.H.V., T.M.N., and C.G. designed the experiments and wrote the manuscript.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.