Cardiomyocyte-specific perilipin 5 overexpression leads to myocardial steatosis and modest cardiac dysfunction.

Presence of ectopic lipid droplets (LDs) in cardiac muscle is associated to lipotoxicity and tissue dysfunction. However, presence of LDs in heart is also observed in physiological conditions, such as when cellular energy needs and energy production from mitochondria fatty acid β-oxidation are high (fasting). This suggests that development of tissue lipotoxicity and dysfunction is not simply due to the presence of LDs in cardiac muscle but due at least in part to alterations in LD function. To examine the function of cardiac LDs, we obtained transgenic mice with heart-specific perilipin 5 (Plin5) overexpression (MHC-Plin5), a member of the perilipin protein family. Hearts from MHC-Plin5 mice expressed at least 4-fold higher levels of plin5 and exhibited a 3.5-fold increase in triglyceride content versus nontransgenic littermates. Chronic cardiac excess of LDs was found to result in mild heart dysfunction with decreased expression of peroxisome proliferator-activated receptor (PPAR)α target genes, decreased mitochondria function, and left ventricular concentric hypertrophia. Lack of more severe heart function complications may have been prevented by a strong increased expression of oxidative-induced genes via NF-E2-related factor 2 antioxidative pathway. Perilipin 5 regulates the formation and stabilization of cardiac LDs, and it promotes cardiac steatosis without major heart function impairment.

lization of cardiac LDs in vivo. Chronic cardiac excess of Plin5-coated LDs was found to result in a modest cardiac dysfunction with decreased expression of a subset of PPAR ␣ target genes, decreased mitochondria function, and left ventricular concentric hypertrophy. Lack of more severe heart function complications may have been prevented by a strong increased expression of oxidative-induced genes via the nuclear factor (erythroid-derived 2)-like 2 (Nrf2) pathway. Taken together, these fi ndings indicate that Plin5 is an important regulator of cardiac LDs.

Mice
All procedures that involved animal handling were approved by the Institutional Animal Care and Use Committee at the University of Maryland School of Medicine. Mice were housed under the 12 h light/dark cycles (lights on at 7:00 AM, lights off at 7:00 PM) and temperature-controlled room. Mice were kept on a regular chow diet (2018SX Teklad Global 18% Protein Rodent Diet, Harlan Laboratory, 6.2% fat) and allowed at libitum access to food and water except when stated otherwise. Only males were used in these studies. For fasting studies, fasting was started before the dark cycle.

Creation of transgenic mice and mouse breeding
Transgenic mice expressing mouse myc-tagged-Plin5 cDNA selectively in cardiac muscle were made using the cardiac-specifi c promoter from the gene encoding ␣ -myosin heavy chain ( ␣ -MHC). To generate a ␣ MHC-epitope-tagged myc Plin5 transgenic construct, we attached a myc epitope to the NH 2 terminus of Plin5. The pBS II SK vector containing the ␣ -myosin heavy ␣ -chain (MHC) promoter region (GenBank U71441) was kindly obtained from Dr. Robbins (Department of Pediatrics, Children's Hospital Research Foundation, Cincinnati, OH) ( 25 ). The amplifi ed Myc-Plin5 was fused to the ␣ -MHC promoter at the Sal I/Hind III site of the vector. After agarose gel electrophoresis and isolation of the transgene with the QuickCleanII Gel extraction kit (Qiagen, Valencia, CA), the transgene was injected into the pronucleus of B6C3HF/C57BL/6J zygotes and transferred to the oviducts of Swiss Webster pseudopregnant females (University of Maryland Transgenic Core). PCR analysis with genomic DNA isolated from tail tips was performed to identify offspring carrying the Myc-Plin5 transgene. Two founder lines were established showing elevated expression of the myc-Plin5 transgene exclusively in cardiac muscle and were developed after six backcrosses with the wild-type C57BL/CJ (NCI, Frederick, MD). To maximize our ability to examine the effect of excess cardiac LDs on mitochondria and heart function, we used the line with the highest Plin5 expression in the present study.

Phenotypic characterization of MHC-Plin5 hearts
Hearts from three-to four-month-old MHC-Plin5 and wildtype (WT) mice were collected to measure mRNA and protein expression of heart perilipin proteins, TAG content, LD quantification, and mitochondria size, number, and function, as well as heart function (see description below).

Tissue processing for electron microscopy and micrograph analysis
For electron microscopy , hearts were excised, quickly cut in small pieces, fi xed in situ with Trump's fi xative, and postfi xed in or amount of ectopic fat; rather, it might be due to alterations in LD function, resulting in abnormal management of excess energy in diseased tissues.
It was recently demonstrated that LDs are highly dynamic and depend on protein activities associated with their monolayer surface ( 9 ). Depending on physiological and pathological circumstances, some proteins are associated with LDs in a transient manner, whereas others are associated in a more permanent fashion ( 10 ). The surface of LDs are coated with at least one of fi ve perilipin proteins, with perilipin 5 (Plin5) found mostly in oxidative tissues, such as heart, whereas perilipin 1 (Plin1) and perilipin 4 (Plin4) are mostly expressed in adipose tissue, and Plin2 and Plin3 are ubiquitous. Our previous work in cell culture models, along with the work of others, has shown that Plin5 plays an important role by positively regulating LD accumulation (11)(12)(13). We postulated that Plin5 may play a role in LD hydrolysis ( 11 ). At a molecular level, Plin5 acts as a scaffolding protein for lipolytic proteins (14)(15)(16)(17), such as its interaction with ATGL ( 14,17 ), which is a property unique to Plin5 ( 14 ), and with CGI-58, which is a property shared with Plin1 ( 14,16,18,19 ). The significance of intracellular cardiac LD hydrolysis was revealed by an unexpected phenotype of whole-body ATGL-knockout mice ( 20 ), which exhibited massive accumulation of TAG in the heart, cardiomyopathy, and shortened life expectancy. This resembles patients with defective ATGL function who suffer from cardiomyopathy, illustrating the importance of LD regulation in disease ( 21 ). Furthermore, it was recently demonstrated that ATGL-mediated LD hydrolysis can regulate cardiac mitochondrial function via peroxisome proliferator-activated receptor (PPAR) ␣ and PGC-1 ( 22 ). Endogenous ligands for PPAR ␣ activation have low affi nity, and LD hydrolysis may generate acutely high concentrations, inducing PPAR ␣ /PGC-1 activities. The ability of Plin5 to provide a scaffold for ATGL and CGI-58 at the LD surface and to regulate ATGL activity suggests that Plin5 plays an important role regulating cardiac mitochondrial function. In addition, it was recently recognized that Plin5 has a distinctive property of tethering mitochondria to the LD surface ( 23 ); we identifi ed a Plin5 protein domain necessary to support this function ( 23 ). This past work performed mostly in cultured cells supported the view that lipid storage is characterized by a reorganization of organelles to maximize this process and is cell type specifi c. More importantly, it suggested that Plin5 may mediate important metabolic and physical interactions between cardiac LDs and mitochondria. The overall body Plin5 knockout phenotype was recently published, and its main features include absence of cardiac LDs, increased cardiac ␤ -oxidation, and increased reactive oxygen species (ROS) content. These transgenic mice develop heart failure with age ( 24 ).
In the present investigation, we directly address the role of cardiac LDs by overexpressing Plin5 in heart and examine its consequences on cardiac LDs, mitochondrial function, and cardiac function. We provide evidence that by overexpressing Plin5 in cardiac muscle, we have generated a mouse model with a robust increase in cardiac LDs, indicating that Plin5 plays a critical role in formation and stabi-the age of 12 weeks. The total of eight RNA samples was individually processed with the Ambion WT expression kit and GeneChip WT terminal labeling and controls kit (Affymetrix, CA) and hybridized to the Mouse Gene 1.0 ST array (Affymetrix). The array was then read by GCS3000 laser scanner (Affymetrix), and microarray image analysis was carried out using Partek Genomics Suite software. Subsequent analysis was carried out using BioConductor. Data from the eight samples were subjected to background subtraction, quantile normalization, log2 transformation, and probe-set summarization using the robust multichip average algorithm ( 31 ). Only the probe sets that interrogate genes were retained, and those meant for quality control and normalization purposes were excluded from further analysis. The signifi cance analysis of microarrays procedure ( 32 ) was then used to obtain multiple test-corrected q values for differential expression of genes between the hearts from MCH-Plin5 mice and control. Genes upregulated or downregulated in ␣ MCH-Plin5 hearts were selected with a maximum q value of 0.05. Gene ontology analysis was performed in David Informatics Resources 6.7 with GO BP-FAT ( 33 ).

Mitochondrial DNA quantifi cation
Relative amounts of mitochondrial DNA (mtDNA) and nuclear DNA were determined by real-time PCR as described previously ( 34 ). Genomic DNA was extracted from mouse heart using the DNeasy Blood and Tissue Kit (Qiagen), according to the manufacturer's instructions. Real-time PCR amplifi cation was performed using the LightCycler 480 (Roche, Switzerland). Levels of mtDNA were measured by normalizing the mitochondrial gene (cytochrome b) to the nuclear gene (actin). A total of 250 ng genomic DNA was used for mtDNA and nuclear DNA markers, respectively, in a 20 l reaction containing SYBR Green I Master (Roche, IN). The following forward and reverse gene-specifi c primers were used. Mouse cytochrome b, sense: GGC TAC GTC CTT CCA TGA GGA C; antisense: GAA GCC CCC TCA AAT TCA TTC GAC. Mouse actin, sense: CAT CTC CTG CTC GAA GTC TAG; antisense: ATC ATG TTT GAG ACC TTC AAC ACC C.

Preparation of individual mitochondrial subpopulations
Subsarcolemmal mitochondria (SSM) and interfi brillar mitochondria (IFM) were isolated from adult mouse hearts as previously described in detail ( 35 ). Two hearts from the same genotype of mouse were pooled for each experimental data point, and all three genotypes were studied on the same day. Size and membrane potential of SSM and IFM were determined by fl ow cytometry as previously reported ( 9 ).

Mitochondria size and internal complexity
To index mitochondrial subpopulation size and complexity, fl ow cytometry analyses were performed using a FACS as previously described ( 35 ). Briefl y, Mitotracker deep red 633 (Invitrogen, Carlsbad, CA), which moves into intact mitochondria due to membrane potential ( ⌬ ⌿ m ), was used to selectively stain intact mitochondria (emission wavelength, 633 nm; fluorescence, 633 nm; red diode laser), and debris was excluded by gating for only Mitotracker deep red 633-positive events (intact mitochondria). Forward scatter (FSC) and side scatter (SSC) detectors were used to examine size (FSC) and complexity (SSC) in isolated mitochondrial subpopulations, and data were represented as histograms plotted against the number of gated events as previously described ( 35 ).

Electron transport chain respiration
Oxygen consumption in SSM and IFM was measured using a Clark-type oxygen electrode (Qubit Systems, Ontario, Canada ).
1% osmium tetroxide as previously described ( 26 ). Briefl y, tissues were then dehydrated in graded alcohols, en-bloc stained with uranyl acetate, and embedded in epoxy resin. One micron sections and ultrathin sections were cut on either a Porter-Bloom or LKB ultramicrotome, then stained with lead acetate and examined in a JEOL 1200 EX. Thirty-fi ve (heart) fi elds containing longitudinally arrayed myofi brils were photographed from each section at 3,200× or 5,000× magnifi cation. Scanned micrographs were analyzed using ImageJ software to manually generate masks of mitochondrial and LD contours, which were used for the calculation of mitochondrial and LD area maximum and total mitochondrial and LD numbers.

Western blot analysis
For whole-cell extract, individual hearts were homogenized with a glass homogenizer with RIPA buffer (Teknova Hollister, CA) containing 150 mM sodium chloride, 1% Triton X-100,1% deoxycholic acid-sodium salt, 0.1% sodium dodecyl sulfate, 50 mM Tris-HCL (pH 7.5), 2 mM EDTA and a pill of complete inhibitor as directed by the manufacturer (Roche Diagnostics). Samples were then centrifuged at 16,000 g for 30 min at 4°C. For supernatant and fat-cake fractions, fi ve fasted mice hearts from each genotype were pooled and homogenized in Tris and EDTA sucrose buffer supplemented with a pill of complete inhibitor as directed by the manufacturer (Roche Diagnostics), and then samples were processed as previously described ( 26 ). Immunoblot analyses were performed as previously described ( 23 ). For whole extract, 2 mg equivalents of heart tissue homogenates were subjected to SDS-PAGE, transferred to a nitrocellulose membrane, and probed with specifi c antibodies against Plin2-5 and Nrf2 ( 23,27 ). GAPDH (1:1000) (Cell Signaling, Beverly, MA) and ␤ -actin (1:5000) (Sigma, St. Louis, MO) were used as the housekeeping genes. For supernatant and fat-cake samples, the protein content of each sample was determined as described previously ( 15 ), and the precise amounts of protein loaded on gels were as indicated in the fi gure legends. The immunoblot signals were detected with Supersignal chemiluminescence reagents (Pierce, Rockford, IL). Mouse anti-GFP antibody (1:10,000) was from Convance (Emeryville, CA). The membranes were visualized with chemiluminescence reagent ECL (Amersham, NJ) and exposed to Kodak XAR fi lm. Quantitative analysis was performed using Un-scan (Silk Scientifi c Corporation, UT).

Total lipid, total lipid composition, and glycogen content
The hearts were perfused with 3 ml of PBS from the left ventricle. Tissues were excised, weighed, and frozen. Lipids were extracted from 20-40 mg of the left ventricles. Total lipids were extracted by the Folch method ( 28 ), and the TAG content was measured using a commercially available kit (Sigma). Lipid composition was determined using thin-layer chromatography as previously described ( 29 ). The glycogen content was measured by a commercially available kit (BioVision, CA).

Gene expression analysis
Total RNA was extracted using Qiagen RNease kit (Qiagen), quantitated using a ND1000 nanodrop spectrophotometer, and then between 100 and 400 ng was reverse-transcribed using a using the Transcriptor First Strand cDNA Synthesis Kit (Roche, Switzerland). Steady-state mRNA levels were determined by twostep quantitative real time PCR (qRT-PCR) using the LightCycler 480 (Roche) ( 30 ) and Taqman probe/primer sets (Applied Biosystems, CA). GAPDH was used as an internal control for normalization. Primers sequences are listed in supplementary Table II . For microarray (GEO accession number GSE44192 ), hearts were harvested from four MCH-Plin5 mice and four control mice at scraping with a rubber policeman, and homogenization was performed on ice in lysis buffer A. The substrate for the measurement of triglyceride hydrolase activity, containing triolein and [9,10-3 H]triolein (PerkinElmer Life Sciences), was emulsifi ed with phosphatidylcholine/phosphatidylinositol (3:1) using a probe sonicator (Virsonic 475, Virtis, Gardiner, NJ), as described previously ( 34 ). Total-cell lysates (0.1 ml) composed of 10 µl of cell lysate containing ATGL and CGI-58, 80 µl of cell lysate containing either empty vector or Plin5 or Plin3 from cells expressing the constructs described in supplementary Table III were added to 0.1 ml of the substrate and then incubated in a water bath at 37°C for 30 min. The reaction was terminated by adding 3.25 ml of methanol/chloroform/heptane (10:9:7) and 1 ml of 0.1 M potassium carbonate, 0.1 M boric acid , pH 10.5. After centrifugation at 800 g for 20 min, the radioactivity in 0.5 ml of the upper phase was determined by liquid scintillation counting as described previously ( 15 ).

Cell culture
CHO-K1 cells were obtained from ATCC (Manassas, VA). Cells were grown in Ham's F-12 medium supplemented with 10% fetal calf serum, 2 mmol/liter L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C with 5% CO 2 and 95% humidity. The constructs ATGL-CFP, CGI-58-YFP, Plin5-YFP, and Plin3-GFP were introduced into the CHO-K1 cells according to the manufacturer's instructions. All constructs have been validated previously ( 14,15 ) Echocardiography Cardiac function was assessed by echocardiography using a Vevo 2100 high-resolution imaging system (Visual Sonics, Toronto, Canada) with an integrated rail system and a 55 MHz transducer (MS550D). Measurements were performed on mice anesthetized with 1% isofl urane. Mice were shaved and placed in a supine position on a warming pad. M-mode frames were recorded from the parasternal short axis, and Doppler measurements were recorded from the apical view. Absolute wall thickness (AWT) and relative wall thickness (RWT) were calculated as: (PWTd + AWTd) and (PWTd + AWTd) / EDD, where PWTd is the end diastolic posterior wall thickness, AWTd is the end diastolic anterior wall thickness, and EDD is the end diastolic diameter. Ejection fraction was calculated as: (EDV Ϫ ESV) / EDV × 100%, where EDV is the end diastolic volume and ESV is the end systolic volume. EDV and ESV were calculated as: ((7.0 / (2.4 + EDD)) × EDD 3 and ((7.0 / (2.4 + ESD)) × ESD 3 , respectively, where ESD is the end systolic diameter. Myocardial performance index (MPI) was calculated as (isovolumetric contraction time + isovolumetric relaxation time) / ejection time.

Intraventricular catheterization
Animals were anesthetized with inhaled 2.5% isofl urane and ventilated at a tidal volume of 0.30 ml at 80 breaths/minute (Harvard Apparatus Inspira, Holliston, MA). Anesthesia was confi rmed by toe pinch, and the heart was exposed with a midline sternotomy. A high-fi delity manometer-tipped catheter (SPR-1000 model, Millar, Houston, TX) was rapidly inserted through the apex into the left ventricle (LV), and pressure was recorded for 1-2 min, after which the terminal necropsy procedure was performed.

Tail-cuff blood pressure and heart rate measurement
Systolic blood pressure and heart rates were measured on unanesthetized mice using a Hatteras Instruments SC1000 Dual-Channel Blood Pressure Analysis System (Cary, NC). Mice were warmed to increase blood fl ow in the tail. Blood pressure was Mitochondria (0.25 mg protein ) were maintained in 0.5 ml solution consisting of 100 mM KCl, 50 mM MOPS, 1 mM EGTA, and 0.5 mg fatty acid-free albumin at pH 7.0 and 37°C. State 3 (ADP-stimulated) and state 4 (ADP-limited) respiration were measured with pyruvate + malate (10 mM and 5 mM, respectively) (glycolytic substrates), palmitoyl-CoA (10 mM) (fatty acid substrate), and then succinate plus rotenone (10 mM and 7.5 µM, respectively) was used to assess respiration through complex II of the ETC exclusively. State 4 respiration was measured ± oligomycin. Respiratory control ratios (RCR; state 3 divided by state 4) refl ect the control of oxygen consumption by phosphorylation ("coupling"). The ADP/O ratios (number of ADP molecules added for each oxygen atom consumed) are an index of the effi ciency of oxidative phosphorylation. These indexes were calculated as described ( 36 ).

ROS production
ROS content was estimated by measuring malondiadehyde using a lipid peroxidation assay kit according to the manufacturer's instructions (Oxford Biomedical Research, MI).

Enzyme activity assays
The enzyme activities of citrate synthase (CS), medium-chain acyl-CoA dehydrogenase (MCAD), and aconitase were assayed in tissue homogenates and normalized to milligram of wet tissue weight. To obtain homogenized tissue, 10-20 mg of tissue was combined with 150 l buffer containing 0.1 M Tris-HCl and 15 mM tricarballylic acid (pH 7.8). Tissue was homogenized in a Bullet blender (Next Advance, Averill Park, NY). For mitochondrial enzymatic measurements, mitochondria were isolated as described above and diluted to 1 mg of protein per milliliter of homogenate.
For aconitase activity, 10 l of myocardial tissue homogenate or 20 l of mitochondrial homogenate was added to cuvettes containing 490 l of aconitase reaction mix (0.083% chloroform, 1.67 mM sodium citrate, 26.7 mM triethanolamine, 0.5 mM NADP+, 0.5 mM MgCl 2 , pH 7.4), and the increase in absorbance at 340 nm was measured over 5 min. Activity was then determined by multiplying the slope to the molar concentration constant for NADPH (6.7 M Ϫ 1 ). For CS activity, 0.5 l of myocardial tissue homogenate or 2 l of mitochondrial homogenate was added to cuvettes containing 500 l of CS reaction mix (0.1 M Tris-HCl, 1.25 mM 5,5 ′dithiobis[2-nitrobenzoic acid], pH 8). Then, 25 l of 50 mM oxaloacetate and 5 mM acetyl-CoA were added to the reaction mixture, and the increase in absorbance at 412 nm was measured over 5 min. Activity was then determined by multiplying the slope to the molar concentration constant for 5,5 ′ -dithiobis[2nitrobenzoic acid] (13,600 M -1 ).
For MCAD activity, 5 l of myocardial tissue homogenate or 10 l of mitochondrial homogenate was added to cuvettes containing 500 l of 100 mM KH2PO4, 1 mM EDTA, 0.5 mM sodium tetrathionate, and 200 M ferrocenium hexafl uorophosphate, pH 7.4. Then, 50 l of 0.5 mM octanoyl-CoA was added to the reaction mixture, and the decrease in absorbance at 300 nm was measured over 5 min. Activity was then determined by multiplying the slope to the molar concentration constant for ferrocenium hexafl uorophosphate (4,300 M Ϫ 1 ).

Triglyceride hydrolase activities assays
Tissues were surgically removed and washed in ice-cold PBS and 1 mM EDTA. Tissues were homogenized on ice in lysis buffer A (0.25 M sucrose, 1 mM EDTA, 1 mM dithiothreitol, complete proteases cocktail inhibitor, pH 7.0; Roche Diagnostic) using a glass homogenizer. The infranatants were obtained after centrifugation at 20,000 g at 4°C for 90 min. Cells were harvested by and intracellular TAG in MHC-Plin5 hearts ( Fig. 2A -C ). TAG was present in larger and more numerous LDs ( Fig.  2C ). Lipid fat cake composition assayed by thin-layer chromatography showed that TAG constitutes the major lipid species in heart and confi rmed that it were strikingly elevated in MHC-Plin5 mice (supplementary Fig. I). While fasting increased the number and the LD droplet size in WT mice cardiac myocytes as previously reported ( 37 ), larger and more numerous LDs were chronically present in MHC-Plin5 cardiac myocytes, irrespective of the nutritional status of the mice. We confi rmed by histocytochemistry and Western blots that Plin5 was present at the lipid surface in cardiac muscles and in the fat-cake fraction from fed MHC-Plin5 mice (supplementary Fig. IIA, B).
Next, we checked whether the LD accumulation was due to the increased presence of Plin5 and not due to a decreased capacity to hydrolyze TAG in MHC-Plin5 cardiac myocytes. Measurements of total heart hydrolase activity performed on whole-tissue extracts were not affected by Plin5 overexpression (4.6 ± 0.54 nmol FA released/ hour/milligram protein for WT and 5.19 ± 0.7 nmol FA released/hour/milligram protein for MHC-Plin5 mice, values are ± SEM, n = 4). In addition, RNA expression levels for ATGL and for CGI-58 were found unchanged (supplementary Fig. IVA). ATGL protein expression was not found different in whole cardiac tissue lysates between genotypes (supplementary Fig. IVB). In addition, ATGL protein expression was measured in fat-cake and supernatant fractions (supplementary Fig. IVC). In both fractions, an antibody against mouse ATGL recognized a doublet appearing at 54 kDa and at 50 kDa, respectively. The 54 kDa measured on three consecutive days, each analysis consisting of a 5 min preliminary acclimation phase followed by 10 min of data recording. Systolic blood pressure and heart rates were recorded each minute for the 10 min period. Reported values are for all values obtained over the last two consecutive days.

Statistical analysis
All values are expressed as means ± SE. Statistical signifi cance was tested using either one-way ANOVA or a two-tailed Student t -test (GraphPad Software).

Myocardial TAG content is chronically increased in MHC-Plin5
Mice with cardiac myocyte-targeted Plin5 expression (MHC-Plin5) are viable and survive to adulthood. Hearts from MHC-Plin5 mice showed a signifi cant increase in both Plin5 mRNA levels ( Fig. 1A ) and Plin5 protein expression ( Fig. 1B, C ). Overexpression of Plin5 was myocardium specifi c, as Plin5 protein expression in other tissues, such as liver and skeletal, was unchanged ( Fig. 1B ). Plin5 overexpression did not affect Plin3 protein or Plin2-4 RNA expression, but it did markedly increase Plin2 and Plin4 protein levels in cellular extracts ( Fig. 1A, B ). Additionally, Plin4 was found increased in Plin4 supernatant as well as fat-cake fractions (supplementary Fig. III), suggesting a role of Plin5 to stabilize LDs in the heart. This was further confi rmed by our histomorphological analysis and TAG measurements showing a marked increase in LDs

Plin5 overexpression induces a gene expression pattern indicative of downregulation of FA utilization, upregulation of glucose utilization, and activation of the Nrf2 antioxidative pathway
To identify cellular pathways likely to be affected by overexpression of Plin5 in the heart, a microarray analysis of gene expression in myocardium from three-monthold mice was performed, and results are reported in Tables 1 , 2 , and supplementary Table I . Gene ontology analysis showed that genes downregulated in MHC-Plin5 myocardium were substantially enriched in mitochondrial bioenergetic processes and lipid metabolism ( P < 0.01), whereas genes upregulated in MHC-Plin5 myocardium were enriched in oxidative stress-induced gene expression via Nrf2 ( P < 0.005), especially genes involved in gluthatione metabolism ( Table 2 and supplementary  Table I ). Further pathway analysis showed that many differentially expressed genes in the MHC-Plin5 hearts encoded enzymes that are involved in mitochondria ␤ -oxidation, oxidative phosphorylation, and fatty acid metabolism and that most of these genes were previously reported to be PPAR ␣ target genes ( 38 ). In contrast, a shift in energetic utilization toward glucose is suggested by the regulation of key regulatory genes promoting glycolysis with decreased expression for 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 1 ( Pfkfb1 ) and increased expression of glucokinase ( Gck ), hexokinase ( Hk1 ), and glucose transporter 1 ( Slc2a1 ) (supplementary Table I), although measured cardiac glycogen stores remained unchanged (amount of glucose was 0.88 ± 0.45 mg/g wet tissue in MHC-Plin5 hearts versus 0.92 ± 0.98, n = 4 in control WT). Importantly, the expression of a number of genes previously involved with cytoskeleton reorganization in MHC-Plin5 hearts was found increased, as well as a small subset of genes reported to be associated with early heart failure conditions ( Table 2 ). band was similar in the supernatant of both WT and MHC-Plin5 supernatant samples; the 50 kDa band was increased in the MHC-Plin5 supernatant and MHC-Plin5 fat-cake samples; and the 54 kDa band was found increased in the WT fat-cake samples. Differences in detergent content of the buffer used for whole extract as opposed to fractionation of fat cake might explained these differences in ATGL protein (see Material and Methods). CGI-58 protein expression was found increased in both supernatant and fat-cake fractions (supplementary Fig. IVD).
Mitochondria in MHC-Plin5 appeared to always be distributed in very tight cluster around LDs and in some cases, the enlarged mitochondria in the MHC-Plin5 tissue displayed further abnormalities in their internal structure, such as loss of cristae and formation of inner membrane vesicles ( Fig. 3A and supplementary Fig. V). Morphometric analysis of mitochondria from cardiac muscles from WT and MHC-Plin5 size revealed a signifi cant increase in mitochondrial size in cardiac muscle from the MHC-Plin5 mice ( Fig. 3B ). In contrast, the relative amount of mitochondrial DNA in cardiac muscle from the WT and MHC-Plin5 mice was unchanged as estimated by the determination of the ratio of ␣ -actin/cytochrome b (32.1 ± 3.6 for WT and 28.7 ± 3.8 for MHC-Plin5 mice, values are ± SEM, n = 4), thus suggesting similar number of mitochondria in the hearts of MHC-Plin5 mice. To further characterize the consequences of Plin5 overexpression on mitochondria morphology, cardiac mitochondrial subpopulations were isolated. Mitochondrial size was found to be larger by fl ow cytometry for both subsarcolemma mitochondria (SSM) and interfi brillar mitochondria (IFM) in the MHC-Plin5 hearts than in WT ( Fig. 3C ), consistent with the electron microscopy analysis ( Figs. 2B, 3A , and  supplementary Fig. IV). The membrane potential of SSM and IFM was lower with high expression of Plin5 in both SSM and IFM ( Fig. 3C ).

Mild impaired cardiac mitochondrial respiration in MHC-Plin5
We found that heart Plin5 overexpression prompted impairment of cardiac mitochondrial function. We prepared heart homogenates from 3-month-old mice and found a small but signifi cant decline in the activity of the mitochondrial enzymes medium chain acyl-CoA dehydrogenase (MCAD) (Fatty acid ␤ -oxidation) of -18% and aconitase of -21% or citrate synthase (tricarboxylic acid cycle) of -23% in MHC-Plin5 hearts ( Fig. 5A ). Results from measurements of mitochondrial enzyme activities in isolated mitochondria subpopulations confi rmed the decrease of citrate synthase, aconitase and MCAD previously found in whole cell extracts from MHC Plin5 hearts ( Fig. 5B ).
Studies in isolated mitochondria revealed a lower yield of SSM and IFM from MHC-Plin5 hearts than from WT hearts (20.9 ± 1.13 versus 10.2 ± 1.4 mg mitochondria protein per gram wet mass for WT and MHC-Plin5 SSM, Our follow-up analysis of mRNA expression on a smaller subset of genes by quantitative real-time PCR analysis revealed a reduced expression of PPAR ␣ target genes in MHC-Plin5 hearts, including PGC1-␣ ( PGC1-␣ , Ϫ 30%), long-chain acyl-CoA dehydrogenase ( Acadl , Ϫ 40%), medium-chain acyl-CoA dehydrogenase ( Acadm , Ϫ 60%), acyl-CoA synthetase long-chain family member 6 ( ACSL6 , Ϫ 40%), acetyl-CoA acyltransferase 2 ( ACAA2 , Ϫ 65%), acyl-CoA synthetase medium-chain family member 5 ( Acsm5 , Ϫ 65%), uncoupling protein 3 ( UCP3 , Ϫ 70%), stearoyl-CoA desaturase 4 ( scd4 , Ϫ 70%), and pyruvate dehydrogenase kinase 4 ( Pdhk4 , Ϫ 70%), compared with the levels found in WT mice ( Fig. 4A ). We confi rmed the upregulation of the oxidative stress pathway via Nrf2 by measuring increased expression of two Nrf2-targeted genes, glutathione S-transferase A1 ( Gsta1 , +1,600%) and activating transcription factor 4 ( ATF4 , +298%), by RT-PCR and the increased presence of Nrf2 protein in MHC-Plin5 nuclear extracts by Western blot ( Fig. 4B ). Intracellular accumulation of ROS, estimated by measuring malondiadehyde content, was found increased in MHC-Plin5 cardiac tissue homogenates compared with WT (13.2 ± from an increased number of mitochondria still associated to the LD fraction that remained in the supernatant during mitochondria isolation. Indeed, we found increased presence of a mitochondria marker voltage-dependent respectively, P < 0.001; 16.2 ± 1 versus 12.1 ± 1 mg mitochondria protein per gram wet mass for WT and MHC-Plin5 IFM, respectively, P < 0.001). It is likely that this lower yield of mitochondria obtained from MHC-Plin5 hearts result ed Enriched biological processes of genes were differentially expressed in myocardium of MCH-Plin5 versus WT control mice. F.C., fold change. Enriched biological processes of genes were differentially expressed in myocardium of MCH-Plin5 versus WT control mice. F.C., fold change. ratios compared with WT (5.14 ± 0.15 for MHC-Plin5 versus 4.79 ± 0.1 for WT, P < 0.05). Echocardiography demonstrated a LV hypertrophy with thickening of ventricular walls ( Fig. 6A ), but none of the parameters related to heart function were affected in MHC-Plin5 hearts ( Fig. 6B ). Direct measurement of LV pressure found no differences between WT and MHC-Plin5 mice for peak LV pressure, peak positive or negative dP/dt, or end diastolic pressure, providing further support for no effect on cardiac systolic or diastolic function ( Fig. 6B ).

Perilipin 5 inhibits TAG hydrolase activity in a reconstituted in vitro system
To test for the ability for Plin5 to inhibit TAG hydrolyase activity, we used a reconstituted in vitro system in which we measured the ability of Plin5 to prevent the hydrolytic activity of ATGL in the presence of its colipase CGI-58 toward an exogenous substrate containing phospholipids and triolein. As control, we used Plin3, another exchangeable perilipin protein. Presence of Plin5, but not Plin3, statistically decreased the ability of ATGL and CGI-58 to hydrolyze TAG ( Ϫ 40%) ( Fig. 7 ).

DISCUSSION
The relationship between cardiac LDs and heart function is yet to be fully understood. A major fi nding of this investigation is that Plin5 stabilizes cardiac LDs, and we confi rmed that Plin5 is a major player in cardiac LDs. Plin5 overexpression resulted in cardiac steatosis but did not lead to a major heart dysfunction phenotype despite differences in gene expression, suggesting a shift in energy substrate utilization, noticeable structural mitochondrial abnormalities, modest decrease in mitochondrial function, and left concentric ventricular hypertrophy. An interesting feature of our results is that Plin5 overexpression elicits an induction of the Nrf2 antioxidative pathway, suggesting a possible compensation by this pathway against the maladaptive cardiac response to increased cardiac Plin5-coated LDs.
Members of the perilipin protein family are the proteome signature of LDs. Four out of the fi ve perilipin members (Plin2-5) have been found expressed in mammalian cardiac muscle ( 15 ). Among them, Plin5 must play a distinctive function as it is selectively expressed in oxidative tissues. Plin5, along with Plin3 and Plin4, has been termed an exchangeable perilipin because it can be found in the cytosol at discrete locations in the absence of FA. However, upon initiation of LD biogenesis by FA uptake, Plin5 is found bound to LD surface ( 39,40 ). Most importantly, Plin5 has been shown to increase FA-mediated TAG accumulation in several cell culture models, and we suggested that Plin5, like adipocyte-specifi c Plin1, inhibits LD hydrolysis in cultured cells ( 23 ). We demonstrate here that Plin5 exerts the same property in vivo when constitutively expressed in heart. Interestingly, by overexpressing Plin5, protein levels of Plin2 and Plin4 are also found increased in absence of noticeable change in mRNA levels. Reciprocally, absence of Plin5 was reported to lead to anion channels (VDAC) in MHC-Plin5 supernatant (supplementary Fig. VI).
Mitochondrial respiration was signifi cantly altered in MHC-Plin5 SSM and IFM mitochondria ( Fig. 5C and supplementary Table II). State 3 respiration was decreased in MHC-Plin5 SSM mitochondria with pyruvate + malate as substrate ( Ϫ 30%), but no statistical difference was observed in IFM mitochondria. State 3 respiration was also decreased with palmitoylcarnitine used as a substrate in SSM mitochondria but was not statistically different, whereas in IFM mitochondria it decreased signifi cantly ( Ϫ 25%) . Independent of the substrate used for the experiment, levels of state 4 respiration were similar among all three groups except in presence of succinate and rotenone. RCR (state 3/state 4 ratio) values were found signifi cantly different in MHC-Plin5 SSM mitochondria in presence of malate + pyruvate, indicating some defect in mitochondria coupling (2.4 ± 0.42 for WT and 1.64 ± 0.12 for MHC-Plin5 mice, values are ± SEM, n = 12). ADP/O ratio was normal (data not shown), indicating that overexpression of Plin5 did not affect the effi ciency of phosphorylation.

Overexpression of Plin5 leads to increased left ventricular mass and remodeling but not to heart failure
We next characterized the cardiac function of fourmonth-old mice. Hearts from MHC-Plin5 were enlarged with significantly elevated heart-weight to tibia-length ATGL gene and total protein expression. In our hands, ATGL protein content in cardiac fat-cake fraction is difficult to quantify because of proteolysis occurring during the fat-cake preparation and possibly due to a lack of detergents in the fractionation buffer. In addition, it was shown that heart depletion in cardiac LDs in Plin5 knockout mice could be overcome by the use of a lipase inhibitor ( 24 ). The combined results from these transgenic mice models and from past studies performed in vitro ( 11,(14)(15)(16)(17) suggest that Plin5 is a negative regulator of cardiac LD hydrolysis. The exact mechanism(s) by which Plin5 inhibits LD hydrolysis remain to be determined. Two mechanisms can be hypothesized. The coordinated increase of perilipin proteins at LD surface may be suffi cient to impair access of ATGL to its TAG substrate. Indeed, presence of Plin2 was shown to inhibit ATGL binding to LD surface in cell culture systems ( 43 ), whereas absence of Plin2 in mice was associated with decreased LDs ( 44 ). Alternatively, decreased Plin2 protein in the heart ( 24 ). The positive relationship between Plin2 and Plin5 is unexpected and differs from the one observed between adipose Plin1 and Plin2: presence of Plin1 was reported to decrease Plin2 protein expression, whereas absence of Plin1 promoted increased presence of Plin2 in the LD ( 29,41 ). A possible explanation for this "off-target effect" is that Plin5, by stabilizing LDs, increases Plin2 and Plin4 protein expression via a posttranslational mechanism. Indeed, Plin2 has been previously found to have posttranslation regulation by the ubiquitin/proteasome pathway at times when Plin2 is unable to bind to LD surface ( 42 ). Existence for such a regulation is yet to be demonstrated for Plin4. Hence, these experiments demonstrate that Plin5 plays an important role in promoting and stabilizing cardiac LDs by increasing the levels of other perilipin proteins.
Cardiac Plin5 overexpression leads to increased LDs without alteration in total heart hydrolase activity or in between LDs and mitochondria. Therefore, it remains a possibility that Plin5 overexpression may also directly alter mitochondria function, which in turn would result in cardiac LD accumulation. To precisely dissect the individual role of Plin5 on these organelles is diffi cult as both of these organelles share common regulatory pathways. Future work is needed to investigate the relationships among LDs, mitochondria, and Plin5.
Due to the ability of Plin5 overexpression to inhibit LD hydrolysis in vivo and in vitro, we predicted that the phenotype of MHC-Plin5 hearts may resemble the one recently described for the ATGL-knockout mice, leading to severe cardiac steatosis, decreased PPAR ␣ gene-targeted activities, and decreased number of mitochondria and mitochondria, function followed by heart failure and early death ( 22 ). MHC-Plin5 hearts show some adaptive features to LD overload but do not experience heart failure. Unexpectedly, chronic Plin5 expression led to oxidative stress and triggered an upregulation of the Nrf2 antioxidative pathway with specifi c increases in gene expression involved in gluthatione metabolism. Nrf2 is a pleiotropic protein Plin5 may act as a unique ATGL negative regulator by providing either scaffolding for ATGL or CGI-58, as suggested in vitro ( 14,16 ). ATGL-or CGI-58-Plin5 binding may then alter the interactions between ATGL and CGI-58, an established positive regulator of ATGL ( 45 ). In support of the latter, we demonstrated here that Plin5 inhibits TAG hydrolase in a reconstituted in vitro system and that this property is not shared by Plin3, another exchangeable perilipin protein. In addition, the presence of CGI-58 is increased in the MHC-Plin5 fat-cake fraction. It was shown that Plin1 regulated ATGL activity via its binding to CGI-58 in adipose cells ( 46 ); by analogy, a similar regulation can be hypothesized for Plin5 regulation of ATGL. Although strong evidence from the existing literature and the data presented here support a role for Plin5 to regulate ATGL activity in vitro and in vivo ( 11,14,23 ), we cannot defi nitively attribute all observed consequences of Plin5 overexpression to the regulation of ATGL activity by Plin5, as overexpression of a protein might induces unexpected offtarget effects. More specifi c to Plin5, we and others have previously found that Plin5 elicits a close physical proximity that binds a cis -acting enhancer sequence known as the antioxidant response element (ARE) to control the basal and inducible expression of a battery of antioxidant genes and other cryoprotective phase II detoxifying enzymes ( 47 ). The cardioprotective role of Nrf2 in heart was recently investigated in a mouse model of heart failure ( 48 ). Nrf2targeted gene expression was induced following the fi rst two weeks of transverse aortic constriction (TAC) in control mice, whereas Nrf2 absence led to faster induction of heart failure after surgery, an indication that the transient activation of the Nrf2 antioxidative pathway is a compensatory mechanism against heart failure ( 47 ). Therefore, MHC-Plin5 hearts may be protected against progression toward heart failure due to the activation of the Nrf2 antioxidative pathway. Interestingly, mice lacking Plin5 were found to have increased cardiac ROS but developed heart dysfunction that can be preventable with an antioxidant treatment ( 24 ). Results from both transgenic mice models suggest a novel and intriguing relationship between Plin5coated LDs and Nrf2 signaling that remains to be investigated. ROS are known to be implicated in the induction of cardiac hypertrophy in various pathologic states, and cardiac hypertrophy is observed in transgenic mice with Plin5 overexpression or loss of function. It is hypothesized that cardiac ROS production in Plin5 knockout mice might result from a surplus fl ux of FA into the mitochondria; in contrast, it may be hypothesized that ROS production observed in cardiac plin5 overexpression might originate from endoplasmic reticulum stress due to excess formation of LDs or from mitochondria structural alteration due to LD close proximity. Further studies are needed to confi rm ROS cellular origin in these two transgenic models.
The combination of coexisting risk factors, e.g., obesity, insulin resistance, hyperlipidemia, hypertension, and type 2 diabetes, makes it diffi cult to attribute a direct cause and effect relationship to myocardial lipid deposition. Here, the direct manipulation of the LD compartment with overexpression of Plin5, a structural LD protein, led to an Fig. 7. Plin5 inhibits TAG hydrolase activity in an in vitro reconstituted system. CHO-K1 cells were transfected with ATGL-CFP and CGI-YFP plasmids, with Plin5 alone, or with Plin3 alone. Cellular extracts (20 l) overexpressing ATGL and CGI-58 were incubated with 80 l of cellular extracts overexpressing either Plin5 or Plin3 and 100 l of a radioactive exogenous substrate. After 1 h of incubation, released radiolabeled NEFA were extracted and dpm was measured with a liquid scintillation counter. Means are errors bars ± SEM; n = 3 for empty GFP vector, n = 3 for Plin5-YFP, n = 3 for Plin3-GFP; * P < 0.05 for Plin5-YFP versus empty GFP vector or versus Plin3-GFP.