|Home | About | Journals | Submit | Contact Us | Français|
Addition of a small peptide called ISG15 is known as ISGylation, which is an ubiquitin (ub)-like posttranslational modification. We currently show that focal ischemia induced by transient middle cerebral artery occlusion (MCAO) in adult mice significantly induces cortical protein ISGylation between 6 and 24hours reperfusion. With two-dimensional western blotting, 45 proteins were observed to be significantly increased in ISGylation (by 1.8- to 9.7-fold) after focal ischemia compared with sham control. Immunochemistry showed that ISGylated proteins are localized in neurons within the ipsilateral striatum and in astroglia within the peri-infarct cortex of ischemic mice. When subjected to transient MCAO, ISG15−/− mice showed increased mortality, exacerbated infarction, and worsened neurologic recovery than did wild-type controls. In addition, mice lacking UBE1L (ub-activating enzyme E1-like protein, the first enzyme of the ISGylation cycle) also showed bigger infarcts when subjected to transient MCAO. Regional cerebral blood flow or other physiologic parameters were not significantly different in both knockouts compared with wild-type controls. These studies indicate that increased protein ISGylation might be an endogenous neuroprotective adaptation to minimize poststroke brain damage.
Proteins undergo several posttranslational modifications (PTMs), which exponentially increase the functional diversity of the proteome (Baumann and Meri, 2004; Sunyer et al, 2008). Many PTMs are reversible and thus activate/deactive proteins like enzymes and second messenger signaling mediators. The ubiquitin (ub) class of PTMs that includes ubiquitination, SUMOylation, and ISGylation add a small protein or peptide to other proteins mediated by a set of specific ligases (Kerscher et al, 2006).
In particular, ISGylation is the conjugation of the protein ISG15 to other proteins. ISG15 is expressed by many mammalian and rodent tissues, but proteins will be ISGylated only based on the need. The C-terminal residue of ISG15 is a glycine, and the carboxyl group of this glycine is the site of attachment to substrates. Lysine side chains are the most common target sites within substrate proteins, resulting in an amide (or isopeptide) bond between ISG15 and substrate (Kerscher et al, 2006). When infected with bacteria or virus, protein ISGylation increases quickly by several fold, presumably stimulated by type I interferon released by virus and by lipopolysaccharide released by bacteria (Kim and Zhang, 2003). Despite its discovery 18 years ago (Loeb and Haas, 1992), the precise functional role of ISGylation in various cell types of the body is still not completely understood. As ISG15 is absent in lower organisms like nematodes and insects, ISGylation is considered to control higher-order cellular functions in eukaryotes (Dao and Zhang, 2005). As ISGylation is induced by bacterial/viral infections, it is considered a regulator of immune functions (Jeon et al, 2010).
ISGylation and de-ISGylation occur in a cyclical manner mediated by a set of ligases (Supplementary Figure 1). The E1 enzyme ub-activating enzyme E1-like protein (UBE1L) a.k.a. ub-like modifier-activating enzyme 7 adenylates ISG15 and forms a thioester bond with it. Activated ISG15 will be passed to the E2 ligase ub-conjugating enzyme H8 by transesterification. There are two E3 ligases for ISG15. The first is an estrogen-responsive finger protein (Efp) that brings the target protein to be ISGylated to the E2–ISG15 complex, and the second ligase is an HECT domain and RCC1-like domain-containing protein 5 that mediates transfer of ISG15 to the substrate protein (Lenschow et al, 2007). The de-ISGylating enzyme ub-processing protease-43 a.k.a. ub-specific peptidase-18 releases ISG15 from the ISGylated proteins (Malakhov et al, 2002).
Recent studies have shown that increased protein ubiquitination and SUMOylation seen after global ischemia, focal ischemia, and hypoxia are neuroprotective (Cimarosti et al, 2008; Lee et al, 2007; Vannucci et al, 1998; Yamashita et al, 1991; Yang et al, 2008a, 2008b). As the significance of ISGylation after cerebral ischemia is not yet studied, we presently evaluated the temporal and cellular patterns of protein ISGylation after transient middle cerebral artery occlusion (MCAO) in adult mice. We further studied the effect of transient MCAO in knockout mice that lack either ISG15 or UBE1L.
Both ISG15−/− and UBE1L−/− mice were developed on a C57BL/6 background. ISG15−/− mice were generated in Dr Klaus-Peter Knobeloch's laboratory in Charité Universitätsmedizin (Berlin, Germany) (Osiak et al, 2005) and subsequently bred in Dr Deborah Lenschow's laboratory at the Washington University (St Louis, MO, USA) (Lenschow et al, 2007). ISG15−/− mice used in this study were bred from Dr Lenschow's colony. UBE1L−/− mice were developed in Dr Dong-Er Zhang's laboratory (The Scripps Research Institute, La Jolla, CA, USA). UBE1L−/− mice used in this study were bred using breeders from Dr Zhang's laboratory. UBE1L−/− mice show normal levels of ISG15, but lack ISG15 conjugation capability (Kim et al, 2006). All surgical procedures were approved by the Research Animal Resources and Care Committee of the University of Wisconsin-Madison and animals were cared for in accordance with the Guide for the Care and Use of Laboratory Animals, US Department of Health and Human Services Publication Number 86–23 (revised).
Focal ischemia was induced in adult, male mice (weighing 23 to 26g) by transient MCAO by an intraluminal suture method as described earlier (Kapadia et al, 2006; Tureyen et al, 2008). In brief, under isoflurane anesthesia (induction: 2% maintenance: 1.2% in an oxygen, and nitrous oxide 50:50 mixture), the left femoral artery was cannulated for continuous monitoring of arterial blood pressure and to obtain measurements of pH, PaO2, PaCO2, hemoglobin, and blood glucose (i-STAT; Senor Devices, Waukesha, WI, USA). The rectal temperature was maintained at 37°C±0.5°C during surgery with a feedback-regulated heating pad. After a midline skin incision, the left external carotid artery was exposed, a surgical 6 to 0 monofilament nylon suture blunted at the end was introduced into its lumen, and gently advanced into the internal carotid artery until regional cerebral blood flow was reduced to ~15% of the baseline (recorded by laser Doppler flowmeter; Vasamedics, LLC, St Paul, MN, USA) as described earlier (Vemuganti et al, 2004). After either 40minutes or 1hour occlusion, the suture was withdrawn (reperfusion confirmed by laser Doppler), the wound sutured, and the mouse allowed to recover from anesthesia and returned to the cage with ad libitum access to food and water. Middle cerebral artery was occluded for 1hour for analyzing postischemic ISGylation and for 40minutes for analyzing infarct volume and neurologic function and immunohistochemistry.
Ischemic infarct volume was measured as described earlier (Kapadia et al, 2006; Tureyen et al, 2008). In brief, mice subjected to transient MCAO were perfused transcardially with buffered paraformaldehyde at 3 days of reperfusion. Each brain was postfixed, cryoprotected, and sectioned (coronal; 40μm thick at an interval of 320μm). The serial sections were stained with Cresyl violet and scanned using the NIH Image program (Rasband, WS, ImageJ, US National Institutes of Health, Bethesda, MD, USA; http://imagej.nih.gov/ij/, 1997-2011). The volume of the ischemic lesion was computed by numeric integration of data from four serial sections with regard to the sectional interval. To account for edema and differential shrinkage resulting from tissue processing, injury volumes were corrected using the Swanson formula: corrected injury volume=contralateral hemisphere volume−(ipsilateral hemisphere volume−measured injury volume) (Swanson et al, 1990).
To evaluate postischemic sensory and motor functions, cohorts of mice were subjected to modified neurologic severity scoring test, adhesive-removal test, rotarod test, and beam balance test. All mice were tested by an evaluator blinded to the study groups at 2 days before MCAO and then on day 3 of reperfusion after MCAO. Postural reflex was measured by modified neurologic severity scoring test on a 5-point scale as described previously (Kapadia et al, 2006; Tureyen et al, 2008). A score of 0 suggests no neurologic deficit (normal), 1 suggests mild neurologic deficit (failure to extend the right forepaw fully), 2 suggests moderate neurologic deficit (circling to the right), 3 suggests severe neurologic deficit (falling to the right), and 4 suggests very severe neurologic deficit (the rat did not walk spontaneously and had a depressed level of consciousness). Sensory motor deficits were evaluated with the adhesive-removal test by placing small adhesive tapes (1 × 1cm2) on each forepaw. The time taken by a mouse to remove tape from each paw was recorded in three trials (Bouet et al, 2009). Motor coordination and motor learning were measured by rotarod test (Jin et al, 2010). After habituation on a stationary rod, mice subjected to transient MCAO were tested with a cylinder that rotated at 8r.p.m. for 3minutes. The latency time to fall from the rod was measured automatically with a stop watch connected to a detector with a photo beam circuit. To examine the vestibulomotor reflex activity, ischemic mice were subjected to beam balance test (Windle et al, 2006). Mice were trained before MCAO to balance on a narrow beam (30 × 1.3cm2) for 60seconds. At 3 days of reperfusion after MCAO, they were scored on a scale of 0 to 6 (steady balance, 0; grasps side of the beam, 1; one limb falls down but hugs the beam, 2; two limbs fall down or spins on beam but will not fall down, 3; falls down between 40 and 60seconds, 4; falls down between 20 and 40seconds, 5; and falls down before 20seconds, 6).
Western blotting was conducted as described earlier (Nakka et al, 2010). In brief, the ipsilateral cortical tissue was homogenized in ice-cold 200mmol/L HEPES buffer (pH 7.4) containing 250mmol/L sucrose, 1mmol/L dithiothreitol, 1.5mmol/L MgCl2, 10mmol/L KCl, 1mmol/L EDTA, 1mmol/L EGTA, 0.1mmol/L phenylmethylsulfonyl fluoride, and 1% protease inhibitor cocktail (Sigma Chemical Co., St Louis, MO, USA). The homogenates were centrifuged at 10,000g for 10minutes at 4°C and proteins were solubilized by adding Lamelli electrophoresis sample buffer (5% SDS, 20% glycerol, 10% 2-mercaptoethanol, 125mmol/L Tri-HCl, pH 6.8, and 0.004% bromophenol blue; Sigma Chemical Co.) and denatured by heating at 94°C for 3minutes. Samples (25μg protein equivalent) were electrophoresed on Bio-Rad Criterion precast gels (Bio-Rad Laboratories, Hercules, CA, USA) (4% to 20% polyacrylamide gradient), transferred onto nitrocellulose membranes, and probed with polyclonal anti-ISG15 (1:500; sc-50368 from Santa Cruz Biotechnology, CA, USA), polyclonal anti-UBE1L (1:500; sc-54437 from Santa Cruz Biotechnology, CA, USA), and monoclonal anti-β-actin (1:1,000; no. 4970 from Cell Signaling Technology, Danvers, MA, USA) antibodies, followed by horseradish peroxidase-coupled anti-rabbit or anti-goat IgG (1:2,000). The protein bands recognized by the antibodies were detected by enhanced chemiluminescence according to the manufacturer's instructions (SuperSignal West Pico Chemiluminescent Substrate Kit; Thermo Scientific, Rockford, IL, USA). Western blots were quantitated using NIH ImageJ software. For each sample, the whole lane was scanned and quantitated. As the ISG15 antibody reacted with a nonspecific ~50kDa band (seen even in ISG15 and UBE1L knockout mice), we subtracted the density of this band of sham lane from MCAO lanes. Densitometric values were normalized against the corresponding β-actin values.
Two-dimensional (2D) electrophoresis and densitometric analysis were conducted as described earlier (Dhodda et al, 2004). In brief, ipsilateral cortical tissue (~50mg) from each mouse was homogenized in 1mL of 10mmol/L Tris-HCl buffer (pH 7.4) containing 0.3% SDS, 50mmol/L magnesium chloride, 500μg/mL RNase, and protease inhibitors (4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, leupeptin, E-64, EDTA, and benzamidine). Samples were incubated on ice for 10minutes and 400μL of SDS boiling buffer (5% SDS, 10% glycerol, and 60mmol/L Tris, pH 6.8) was added. The protein content was estimated (BCA Assay Kit, Pierce, Rockford, IL, USA), adjusted to 1mg/mL with SDS boiling buffer containing β-mercaptoethanol, and samples were heated at 95°C for 5minutes. Two-dimensional PAGE was performed using 200μg of protein equivalent of each sample. In all, 1μg of tropomyosin (served as an external reference standard) was added and samples were subjected to isoelectric focusing with 2% ampholines (pH 3.5 to 10; Amersham Pharmacia Biotech, Piscataway, NJ, USA) for 9,600Volt hour. After equilibration for 10minutes in Tris-HCl buffer (62.5mmol/L, pH 6.8) containing 10% glycerol, 50mmol/L dithiothreitol, and 2.3% SDS, the tube gel was sealed to the top of a 4% stacking gel that overlaid a 10% acrylamide slab gel (0.75mm thick). Electrophoresis in the second dimension was carried out for 4hours at 15mA per gel using myosin (220kDa), phosphorylase A (94kDa), catalase (60kDa), actin (43kDa), carbonic anhydrase (29kDa), and lysozyme (14kDa) as molecular weight (MW) standards. On the 2D gel, tropomyosin added as an external control migrated as a doublet with the lower polypeptide spot of 33kDa MW and 5.2 pI. For each sample, two parallel 2D gels were used. The gels were blotted onto polyvinylidene fluoride membranes, blots were stained with Coomasie Brilliant Blue R-250, and desktop scanned. Blots were then blocked for 2hours in 5% nonfat dry milk in TTBS (Tris-buffered saline) containing Tween-20) and incubated overnight with polyclonal anti-ISG15 antibody (1:500) in TTBS containing 2% nonfat dry milk. Blots were rinsed in TTBS (3 × 10minutes) and probed with horseradish peroxidase-coupled anti-rabbit IgG (1:2,000; GE Healthcare, Piscataway, NJ, USA, Catalog no. NA934V) for 2hours, rinsed in TTBS, and developed with enhanced chemiluminescence method. The duplicate western blot films of each sample were scanned with a laser densitometer and 2D image analysis for spot detection, gel alignment, and spot quantification were performed using the Progenesis PG-240 Software with TT900 (version 2006; Nonlinear Technology, Durham, NC, USA). The scanner was calibrated using neutral density filter set (Melles Griot, Irvine, CA, USA) for linearity. The method for computerized analysis for the pairs included image warping using the TT900 software, followed by automatic spot finding, background subtraction (average on boundary), matching, and quantification in conjunction with detailed manual checking. The identifiable spots were outlined, quantified, and matched between the sham and MCAO blots. In cases in which protein spots were missing in one group of blots and present in the other group, a small area of background was outlined appropriately to facilitate matching. Quantification of proteins was expressed as density of the spots. To correct for variability resulting from staining, results were expressed as relative volumes of all spots in each blot. The fold change was calculated as the percentage difference of a matched spot between the two groups of blots (sham versus ischemia). The software used an inbuilt Student's t-test to find statistically significant differences in spot densities between the sham and ischemia groups.
Mice subjected to 40minutes MCAO were transcardially perfused with 4% buffered paraformaldehyde and the brains were sectioned (coronal; 35μm thick). The sections were rinsed in 0.1mol/L Tris-buffered saline (3 × 5minutes), blocked in 4% bovine serum albumin for 30minutes before primary antibody incubation. For double immunofluorescence staining, sections were incubated with rabbit polyclonal anti-ISG15 (1:100, Santa Cruz) together with mouse monoclonal anti-NeuN antibody (1:500; Chemicon, CA, USA) or mouse monoclonal anti-glial fibrillary acidic protein antibody (1:500; Cell Signaling Technology). Sections were washed in Tris-buffered saline and incubated for 1hour with secondary antibodies (goat anti-rabbit 488 and goat anti-mouse 594; 1:200 each; Molecular Probes, CA, USA) and analyzed with an epifluorescence or confocal microscope. Negative controls included sections in which primary antibodies were omitted, incubated with secondary antibodies only, or sections in which both primary and secondary antibodies were omitted. Three sections were used for each mouse for each antibody.
The number and content of ISG15-conjugated proteins (reactive with ISG15 antibodies) increased progressively from 6 to 24hours reperfusion in the ipsilateral cortex of mice subjected to transient MCAO compared with sham control (Figure 1A). Most of the ISGylated proteins in the postischemic brain were observed to be in the MW range of 25 to 250kDa. There was a significant increase in the total ISGylated protein content at 6hours (by 242%±33%, P<0.05), 12hours (by 222%±41%, P<0.05), and 24hours (by 515%±91%, P<0.05) reperfusion over sham control (n=6 per group) (Figure 1A). Levels of free ISG15 were not altered significantly after transient MCAO compared with sham (Figure 1A). The contralateral cortex of mice subjected to transient MCAO did not show increased ISGylation compared with sham control (data not shown). The protein expression of the E1 ligase of ISGylation cycle UBE1L was also induced between 6 and 24hours of reperfusion after transient MCAO compared with sham control by 172% to 225% P<0.05 at all three reperfusion time points (Figures 1B and 1C). Levels of β-actin used as the loading control were unaltered at any reperfusion time after transient MCAO compared with sham control (Figures 1A and 1B).
Increased protein ISGylation in the ipsilateral cortex of mice subjected to transient MCAO or sham surgery was also confirmed with 2D western blotting. With laser scanning densitometry combined with analysis by Progenesis PG240 software/TT900, 186 protein spots reacted with ISG15 antibodies (those that can be reliably identified in all 2D western blots of either MCAO or sham group) were analyzed. Of the 186 spots, 45 showed significantly increased and 6 showed significantly decreased density in the MCAO group than in the sham group (Supplementary Table 1). The sum of densities of all the ISG15+ spots was 5.34±1.21 for the sham and 13.38±1.72 for the MCAO groups (n=3 per group), and the average spot density for the 45 upregulated spots was 3.94-fold higher in the MCAO group than in the sham group. Figure 1D shows representative 2D blots of sham and MCAO groups and the magnified images of the six protein spots. Of the 45 proteins that showed increased ISGylation, 8 showed >5-fold increase and 16 showed 3- to 5-fold increase (Supplementary Table 1). Of the 45 proteins that showed increased ISGylation in the ischemic brain, 27 are in the MW range of 30 to 65kDa (Supplementary Table 1). The smallest and the biggest proteins that showed increased ISGylation after focal ischemia are 16.5 and 137.5kDa, respectively (Supplementary Table 1). At 24hours reperfusion after transient MCAO, many ISGylated proteins were observed in the peri-infarct area of the ipsilateral cortex in both neurons (positive for NeuN) and astrocytes (positive for glial fibrillary acidic protein) (Figure 2).
As 1hour transient MCAO induced ~45% mortality in ISG15−/−, the occlusion period was decreased to 40minutes to measure infarction and neurologic function. Both ISG15−/− and UBE1L−/− mice developed bigger infarcts than did wild-type mice (Figure 3, top panel). The regional cerebral blood flow measured during MCAO and reperfusion was observed to be similar between the three genotypes (Figure 3, middle panel). There were no significant differences in the physiologic parameters (brain and body temperature, hemoglobin levels, blood glucose, pH, PaCO2, and PaO2) measured before and during occlusion and during reperfusion between the three groups (data not shown). In wild-type mice, transient MCAO/reperfusion (3 days) resulted in an infarct with a total volume of 14.1±2.3mm3 (Figure 3, bottom panel). The infarct volume was observed to be significantly higher in ISG15−/− mice (by 136% P<0.05) (n=9) and UBE1L−/− mice (by 157% P<0.05) (n=6) compared with wild-type mice (n=9) (Figure 3, bottom panel). Immunostaining of brain sections from ISG15−/− and UBE1L−/− mice killed at 1 day of reperfusion after transient MCAO showed absence of cells positive for ISGylated proteins in both knockouts (Figure 4A). Western blot analysis also showed the absence of ISGylated proteins in the ipsilateral cortex of ISG15−/− and UBE1L−/− mice at 1 day of reperfusion with the exception of a ~50kDa nonspecific band (Figure 4B). Previous studies also showed the presence of this ~50kDa protein band in rodents (Malakhova et al, 2003).
The average postischemic neuroscore at 3 days of reperfusion after transient MCAO was observed to be significantly higher (P<0.05) in ISG15−/− mice (3.12.±0.53; very severe neurologic deficits; n=9) compared with wild-type mice (1.67±0.36; mild neurologic deficits; n=9) (Table 1). ISG15−/− mice also fared worse than did wild-type mice in adhesive-removal test (wild-type: 51±6seconds and knockout: 74±11seconds; P<0.05), beam balance test (wild-type: 1.23±0.34 and knockout: 2.55±0.52; P<0.05), and rotarod test (wild-type: 121±14seconds and knockout: 76±13seconds; P<0.05) (n=9 per group) (Table 1). In UBE1L−/− mice, the modified neurologic severity scoring test was observed to be 3.63±0.50 (P<0.05 compared with wild-type; n=9 per group). As these mice were in a poor neurologic status, we were unable to perform the beam balance test, adhesive-removal test, and Rotarod test.
In brief, results of this study showed that transient focal ischemia induces significant protein ISGylation in the brain; and ISGylated proteins are localized in both neurons and astrocytes. Furthermore, transgenic mice that lack the capability to ISGylate proteins show exacerbated ischemic brain damage indicating that postischemic ISGylation might be an endogenous neuroprotective response to an ischemic insult.
We along with others showed that focal ischemia leads to extensive changes in the cerebral mRNA and microRNA expression in rodents (Dharap et al, 2009; Kapadia et al, 2006; Lu et al, 2003; Soriano et al, 2000; Vemuganti et al, 2002). This altered expression of various families of genes has an important role in promoting postischemic pathologic mechanisms like inflammation, ionic imbalance, edema, and receptor dysfunction that precipitate neuronal death after focal ischemia. Many studies showed that pharmacological manipulations can induce neuroprotection after focal ischemia by affecting the gene expression (Chen et al, 2006; Tureyen et al, 2007; Wang et al, 2004; Xu et al, 2005). However, most if not all proteins undergo PTMs that either activate or inactive them and increase the functional diversity of the proteome (Baumann and Meri, 2004; Sunyer et al, 2008). There are several types of PTMs including (1) addition of a small functional group (e.g., phosphorylation, oxidation, hydroxylation, etc.), (2) chemical modification of an amino acid (e.g., deamidation of glut to gln), (3) change in the structure (e.g., disulfide bonding of two cysteines), and (4) addition of a small protein or peptide (e.g., ubiquitination, SUMOylation, and ISGylation) (Kerscher et al, 2006). Although some of these like phosphorylation are well studied, the functional significance of ub class of PTMs in normal and pathologic brains is not evaluated in detail. Furthermore, very few studies evaluated whether cerebral ischemia induces PTMs and if yes, they have a physiologic significance in deciding the functional outcome after stroke. Previous studies showed that protein ubiquitination and SUMOylation increased after global ischemia, focal ischemia, and hypoxia might induce ischemic tolerance (Cimarosti et al, 2008; Lee et al, 2007; Vannucci et al, 1998; Yamashita et al, 1991; Yang et al, 2008a). Conversely, this is the first report that showed increased protein ISGylation after cerebral ischemia. Our studies showed exacerbated ischemic brain damage in knockout mice that lack either ISG15 or UBE1L (the E1 ligase of ISGylation cycle). When subjected to transient MCAO, the two knockout strains that lack the capability to ISGylate proteins showed bigger infarcts and worsened neurologic recovery as measured by sensorimotor deficits, postural reflex, vestibulomotor function, and motor coordination/motor learning compared with wild-type mice. This indicates that similar to other ub class of PTMs, ISGylation is also a neuroprotective adaptation to minimize postischemic brain damage.
In the postischemic brain, ISGylated proteins were observed to be localized in both astrocytes and neurons. Previous studies showed that SUMOylation (an ub-like PTM similar to ISGylation) also increases in the brain by 30minutes to 6hours of reperfusion after transient MCAO in adult rats. We observed that protein ISGylation starts to increase as early as 6hours of reperfusion and progressively more proteins were ISGylated at 1 and 24hours reperfusion. Although both SUMOylation and ubiquitination preferentially modify mostly the higher MW proteins, ISGylation seem to be specific towards mid-MW range proteins (Yang et al, 2008a; Vannucci et al, 1998). Thus, the targets of SUMOylation and ISGylation seem to be different and hence the mechanisms of neuroprotection induced by these two PTMs might also be different.
To conclude, results of this study showed that focal ischemia significantly increases protein ISGylation in the brain and prevention of postischemic ISGylation exacerbates brain damage. At present, the significance of proteins being ISGylated in the ischemic brain is not known. Future studies are required to show whether modulation of specific ISGylated proteins can be a therapeutic option to prevent postischemic pathophysiological events and/or to promote plasticity and regeneration.
The authors declare no conflict of interest.
Supplementary Information accompanies the paper on the Journal of Cerebral Blood Flow & Metabolism website (http://www.nature.com/jcbfm)
These studies were funded by the American Heart Association Grant-in-aid 4580030 and by the United States National Institute of Health Grant GM66955.