Decorin Induces Mitophagy in Breast Carcinoma Cells via Peroxisome Proliferator-activated Receptor γ Coactivator-1α (PGC-1α) and Mitostatin*

Background: Decorin functions as a soluble tumor repressor via binding receptor-tyrosine kinases, such as Met, to curb rampant tumor neovascularization. Results: Decorin evokes tumor cell mitophagy through dynamic co-regulation of PGC-1α and mitostatin via physical interactions between PGC-1α and mitostatin Conclusion: Decorin requires mitostatin to evoke mitophagy as the underlying basis for angiogenic attenuation. Significance: We have identified mitostatin as a novel mitophagic effector. Tumor cell mitochondria are key biosynthetic hubs that provide macromolecules for cancer progression and angiogenesis. Soluble decorin protein core, hereafter referred to as decorin, potently attenuated mitochondrial respiratory complexes and mitochondrial DNA (mtDNA) in MDA-MB-231 breast carcinoma cells. We found a rapid and dynamic interplay between peroxisome proliferator-activated receptor γ coactivator-1α (PGC-1α) and the decorin-induced tumor suppressor gene, mitostatin. This interaction stabilized mitostatin mRNA with concurrent accumulation of mitostatin protein. In contrast, siRNA-mediated abrogation of PGC-1α-blocked decorin-evoked stabilization of mitostatin. Mechanistically, PGC-1α bound MITOSTATIN mRNA to achieve rapid stabilization. These processes were orchestrated by the decorin/Met axis, as blocking the Met-tyrosine kinase or knockdown of Met abrogated these responses. Furthermore, depletion of mitostatin blocked decorin- or rapamycin-evoked mitophagy, increased vascular endothelial growth factor A (VEGFA) production, and compromised decorin-evoked VEGFA suppression. Collectively, our findings underscore the complexity of PGC-1α-mediated mitochondrial homeostasis and establish mitostatin as a key regulator of tumor cell mitophagy and angiostasis.

"guardian from the matrix" (9). Decorin has been involved in the control of various biological processes encompassing collagen fibrillogenesis, wound healing, myogenesis, stem cell biology, and fibrosis (10 -16). Initially identified as a cell growth inhibitor via blockage of transforming growth factor ␤ (TGF␤) (17,18), soluble decorin is emerging as a potent pan-receptortyrosine kinase inhibitor targeting EGF receptor, Met, IGF-I receptor, VEGF receptor 2, and PDGF receptor (19 -29). Indeed, a decorin fusion protein linked to a wound-targeting peptide enhances wound healing and reduces scar formation via abrogation of TGF␤1/2 signaling (30). Interestingly, recent findings now suggest a much broader role for decorin insofar as modulating the biophysical properties of tendons and ligaments (14), orchestrating a critical signaling events during myogenesis (16), and regulating the innate immune receptors, Toll-like receptors 2/4, during inflammation (31).
Consistent with the proclivity of decorin to induce tumor suppressor genes (PEG3, CDKN1A), decorin induces mitostatin, a mitochondrial protein with oncostatic activity (55). Upon induction, mitostatin displays several hallmarks of a classical tumor suppressor gene such as inhibiting tumor cell migration, growth, and proliferation and simultaneously triggering proapoptotic pathways (55,56). Furthermore, mitostatin is absent in ϳ35% of human prostate carcinomas (56), whereas decreased expression is associated with advanced cancer stages (55). Thus, we hypothesized that decorin could compromise tumor mitochondria as the underlying mechanistic basis for suppressed tumor angiogenesis under normoxia. We discovered that decorin evoked mitostatin production via the Met receptor, thereby triggering a signaling cascade leading to a mitostatin-dependent mitophagy associated with a negative feedback control on VEGFA transcription, thus indirectly attenuating tumor angiogenesis. We further discovered a novel interaction between PGC-1␣ and mitostatin, and this interaction led to stabilization of mitostatin mRNA and concurrent accumulation of mitostatin protein.

EXPERIMENTAL PROCEDURES
Cells and Materials-Triple-negative breast carcinoma cells and cervical carcinoma cells, MDA-MB-231 and HeLa, respectively, were procured from American Type Culture Collection Bioresource Center (ATCC) and maintained in Dulbecco's modified Eagle's medium (DMEM) from Sigma and supplemented with Hyclone 5% bovine calf serum (BCS) from Thermo Scientific (Waltham, MA). Dulbecco's phosphate buffer saline was purchased from Corning (Tewksbury, MA). The anti-mitostatin affinity rabbit purified antibody was described before (55). The rabbit antibodies against PGC-1␣ and voltage-dependent anion channel (VDAC) were from Abcam Inc. (Boston, MA). Rabbit polyclonal antibodies targeting GAPDH and GFP were purchased from Cell Signaling (Beverly, MA) and Invitrogen, respectively. Antibodies detecting VEGFA and Met were from Santa Cruz Biotechnology (Santa Cruz, CA). The mouse monoclonal antibody mixture (mAb) detecting mitochondrial oxidative phosphorylation components came from MitoSciences (Abcam Inc.). This mixture contains five different monoclonal antibodies, existing in a previously optimized premix, directed against specific subunits comprising the various electron transport chain complexes. Therefore, the monoclonal antibody responsible for detecting complex I reacts with the NDUFB8 subunit. The antibody detecting complex II is against the subunit encompassing the iron-sulfur region of succinate dehydrogenase, and the antibody recognizing complex III is against the UQCRC2 subunit (contains core protein 2 of complex III). Finally, the antibodies responsible for detecting complex IV are directed against the MTCO2 subunit, whereas the complex V monoclonal antibody has been directed against the ATP5A subunit. HRP-conjugated donkey anti-rabbit and sheep anti-mouse were purchased from Millipore (Billerica, MA). Goat anti-rabbit (Alexa-Fluor 488) antibodies were purchased from Invitrogen. The small molecule inhibitor SU11274 was from EMD (Millipore). Rapamycin was from Sigma. Lipofectamine 2000, Lipofectamine RNAiMAX, and the 5,5Ј,6,6Ј-tetrachloro-1,1Ј,3,3Ј tetraethylbenzimidazol carboncyanine iodide (JC-1) mitochondrial potential dye were procured from Invitrogen. Recombinant human HGF, actinomycin D, and cycloheximide came from Calbiochem. Protein A-Sepharose magnetic beads were purchased from GE Healthcare. SuperSignal West Pico chemiluminescence substrate was from Thermo Scientific. Decorin purification was described previously (57) and found free of contaminants (54).
Immunofluorescence Imaging, and Live Cell Confocal Microscopy-Typically, ϳ5 ϫ 10 4 MDA-MB-231 cells were grown on poly-L-lysine (50 g/ml) and 0.2% gelatin-coated 4-chamber slides (Nunc, Thermo Scientific) and exposed to decorin (100 nM) as necessary for the given analysis. Cells were washed with Dulbecco's phosphate buffer saline and subsequently fixed for 30 min in 4% paraformaldehyde at 4°C. Cells were then blocked in PBS, 5% BSA, incubated with various antibodies for 1 h, washed twice in PBS, and then incubated for 1 h with the appropriate secondary antibodies (goat anti-rabbit IgG Alexa-Fluor 594). Nuclei were visualized with DAPI (Vector Laboratories, Burlingame, CA). Images were subsequently acquired with a 63 ϫ 1.3 oil-immersion objective installed on a LEICA DM5500B microscope programmed with Leica Application Suite, Advanced Fluorescence v1.8 software from Leica Microsystems, Inc. (Frankfurt, Germany). Fluorescence imaging of MDA-MB-231 Su9-GFP cells was carried out as described above for immunofluorescence microscopy with identical parameters maintained for the treatment groups examined. Images were quantified with ImageJ (National Institutes of Health) software programmed with a special macro specifically designed to measure mitochondrial morphology variables as described previously (58). For live cell confocal microscopy, 3.5-cm inset dishes were coated with poly-L-lysine and 0.2% gelatin as before and seeded with ϳ5 ϫ 10 4 MDA-MB-231 Su9-GFP cells and visualized with a 63ϫ 1.3 oil-immersion objective using a Zeiss LSM-780 confocal laser scanning microscope and analyzed using the Zeiss LSM-780 software. All collected images were analyzed using ImageJ and Adobe Photoshop CS5.5 (Adobe Systems, Carlsbad, CA), and movies were processed and assembled with Adobe Premiere Pro CS5.5 (Adobe Systems). Three-dimensional surface blots were created with ImageJ software as described (33).
Transient siRNA-mediated Knockdown-Transient knockdown of Met, PGC-1␣, and mitostatin in MDA-MB-231 was achieved via transfection of three separate and validated siRNAs specific for either Met (sc-29397), PGC-1␣ (sc-38884), or mitostatin (sc-95954), all from Santa Cruz Biotechnology. Scrambled siRNA (siScr, sc-37007) served as a control for all siRNA experiments presented herein. The following protocol was used subsequent to protein (radioimmunoprecipitation extraction), RNA isolation (TRIzol reagent, Invitrogen), or mtDNA isolation via RNAzol B (from Molecular Research Center, Inc. (Cincinnati, OH)) from variably treated samples for further analysis. As such, 6-well plates were seeded with ϳ2 ϫ 10 5 MDA-MB-231 cells followed by overnight incubation at 37°C/5% CO 2 until cultures were ϳ70% confluent. Targeting or scrambled siRNA duplex (80 and 20 pM, respectively) was added to diluted Lipofectamine 2000 RNAiMAX (Invitrogen) in 1% BCS-DMEM. The transfection mix was applied for 6 h at 37°C. 5% CO 2 whereupon additional full serum (5% BCS-DMEM) media was added. The cells were then allowed to incubate overnight in the same culture conditions. The media were changed, and the transfection was carried out for an additional 48 h. Verification of RNAi-mediated knockdown of the target protein was determined via immunoblotting or quantitative real-time polymerase chain reaction (qPCR).
Quantitative Real-time PCR Analysis-Gene expression or mtDNA content analysis by qPCR was carried out on subconfluent 6-well plates seeded with ϳ2 ϫ 10 5 of MDA-MB-231 cells were treated variably depending on experimental parameters in full serum (5% BCS DMEM) media. After incubation, cells were directly lysed in either 1 ml of TRIzol reagent (Invitrogen) or 1 ml of RNAzol B (Molecular Research Center) to extract total RNA or for later purification of mtDNA (see below). Subsequently, for gene expression analyses only, ϳ1 g of total RNA was annealed with oligo(dT 18 -20 ) primers, and cDNA was synthesized with SuperScript Reverse Transcriptase II (SSRT II, Invitrogen). PCR amplicons representing target genes and the endogenous housekeeping gene, ACTB, were amplified in quadruplicate, independent reactions with the Brilliant SYBR Green Master Mix II reagent (Agilent Technologies, Cedar Creek, TX). All samples were run on the Roche LightCycler 480-II Real Time PCR platform (Roche Applied Sciences), and cycle number (Ct) was recorded for each independent reaction. -Fold change determinations were made utilizing the Comparative Ct method. ⌬Ct values represent normalized gene expression levels to ACTB. ⌬⌬Ct values were then calculated and represent the experimental cDNA (for example, those samples were treated with 100 nM decorin) minus the corresponding gene levels (⌬Ct values) of the calibrator sample (i.e. control samples). Last, -fold changes were calculated using the double ⌬Ct method 2 Ϫ⌬⌬CT Ϯ S.E. Data presented herein represent at least three independent trials run in quadruplicate for each gene of interest examined.
RNA Immunoprecipitation (RIP)-RIP followed by qPCR of precipitated RNA was employed to investigate the occupancy of PGC-1␣ binding directly to MITOSTATIN mRNA in the presence of decorin or in the presence of SU11274 and decorin in MDA-MB-231 cells. The RIP protocol was executed according to the manufacturer's instructions enclosed with the Magna RIP kit (Millipore). Briefly, two confluent (ϳ90%) 10-cm dishes of MDA-MB-231 per experimental condition (totaling ϳ16 ϫ 10 6 cells) were lysed in RIP lysis buffer on ice after washes in PBS and stored at Ϫ80°C until further use. Magnetic beads were prepared by with initial PBS washes followed by incubation at room temperature for 30 min with primary antibody raised against PGC-1␣ (5 g of total antibody used per immunoprecipitation). Extensive washes were performed before incubation of absorbed magnetic beads with previously collected cell lysates. Incubation of conjugated beads with lysate took place overnight at 4°C with end-over-end rotation. The beads were thoroughly washed and digested with proteinase K (45 min at 55°C) to disengage PGC-1␣ containing ribonucleoprotein complexes. RNA from immunopurified PGC-1␣-positive ribonucleoproteins were harvested via a canonical phenol chloroform isoamyl extraction and further precipitated via ethanol. Immunoprecipitated RNA from PGC-1␣ (ribonucleoproteins) was then subjected to cDNA synthesis and qPCR analysis as described above.
mtDNA Isolation-Analysis of mtDNA was performed in MDA-MB-231 cells grown in a six-well plate. Isolation of mtDNA was done according to a modified protocol derived from Tom Getty (Michigan State University). Briefly, after treatment according to experimental conditions, confluent (ϳ90%) MDA-MB-231 cells were lysed in 1 ml of RNAzol B and subjected to a chloroform extraction. A polyacryl carrier (Molecular Research Center) was utilized to facilitate precipitation of the DNA in conjunction with an ethanol extraction. After purification of DNA samples (containing both mtDNA and genomic DNA), 5 ng of purified DNA was used per qPCR reaction, and mtDNA content was measured using primers specific for NADH dehydrogenase subunit 1 (ND1) with Ct values normalized to the genomic marker, LPL (lipoprotein lipase). Reported -fold changes Ϯ S.E. were calculated via the ⌬⌬Ct method as described above.
Immunoblotting and Immunoprecipitation-After each treatment as described herein, MDA-MB-231 cells were lysed in radioimmunoprecipitation assay buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA/EGTA/sodium vanadate, 10 mM ␤-glycerophosphate, and various protease inhibitors including 1 mM phenylmethanesulfonyl fluoride and 10 g/ml leupeptin/tosylphenylalanyl chloromethyl ketone/aprotinin each) for 20 min on ice and separated on SDS/PAGE. For immunoprecipitation experiments, protein A-Sepharose magnetic beads (GE Healthcare) were co-incubated with antibodies overnight at 4°C. The next day the beads were washed extensively, and the lysates were added to the beads and incubated overnight at 4°C with rotation. After extensive washing, the beads were boiled in reducing buffer, and supernatants were separated by SDS/ PAGE. Proteins were then transferred to nitrocellulose membranes (Bio-Rad), immunoreacted with the indicated primary antibodies, subsequently developed with enhanced chemiluminescence (Thermo Scientific), and detected using an ImageQuant LAS-4000 (GE Healthcare).
Measuring Mitochondrial Membrane Potential-At least two individual assays were performed in MDA-MB-231 cells using the mitochondrial dye JC-1. JC-1 accumulates in actively respiring mitochondria forming "J-aggregates," which emits a red/orange fluorescence at ϳ590 nm. However, during times of low mitochondrial membrane potential (i.e. depolarization), JC-1 exists as a monomer and emits a green fluorescence at ϳ525 nm. As such, MDA-MB-231 cells were grown in fourchambered glass slides coated with poly-L-lysine (50 g/ml) and 0.2% gelatin for 24 h in 5% BCS-DMEM. Cells were then treated with 100 nM decorin for 6 h. One chamber was treated with carbonylcyanide 4-triflouromethoxy phenylhydrazone (FCCP) 10 min before staining. Each chamber was incubated with JC-1 (7.5 M) for the last 10 or 20 min of the experiment. Cells were washed twice with PBS and imaged live using a Leica DM5500B microscope. All the images were procured using the same settings.
Quantification and Statistical Analysis-Immunoblots were quantified by densitometry using ImageJ software. Gene expression and mtDNA analysis were determined as described above and reported as -fold changes ϮS.E. All experiments contained herein were carried out with a minimum of three independent trials performed in triplicate. Results are expressed as the mean Ϯ S.E. Statistical analysis was performed with SigmaStat for Windows Version 3.10 (Systat Software, Inc, Port Richmond, CA). Significance of differences was determined by unpaired Student's t test, and data were considered significant when p Ͻ 0.05. For quantification of immunofluorescence studies, fluorescence intensity and three-dimensional surface plots were quantified by measuring pixels with ImageJ software.

RESULTS
Decorin Requires Met for Suppression of Oxidative Phosphorylation-Decorin transcriptionally suppresses several critical oncogenes under normoxia including Myc and HIF-1␣ (42), which play instrumental roles in the reprogramming of metabolism, as it pertains to the dichotomy of aerobic glycolysis and mitochondrial respiration (4, 59 -61). Thus, we determined the effect of soluble decorin protein core (62) on mitochondrial respiratory chain complexes (OXPHOS) in triple-negative breast carcinoma MDA-MB-231 cells. Utilizing a mixture of monoclonal antibodies targeting each OXPHOS subunit, we found a marked suppression of complexes II, III, IV, and V in response to decorin (Fig. 1A). Quantification of three independent experiments revealed a reduction of 40 -60% relative to control (Fig. 1B) after 4 h, and this effect was long lasting (Fig. 1C).
Next, we investigated the involvement of the HGF/Met signaling axis in the regulation of OXPHOS. Exogenous HGF increased the amount of all complexes (Fig. 1D) by ϳ2-fold over basal state (p Ͻ 0.001; Fig. 1E), and Met was required, as depletion of Met via siRNA (Fig. 1F) abrogated this response for complex V (Fig. 1, F and G), complex III ( Fig. 1, F and H), and complex IV (Fig. 1, F and J). Intriguingly, silencing of Met did not prevent suppression in response to decorin of complex II (Fig. 1, F and I). Therefore, it is possible that a differential pathway exists for complex II degradation (as complex II is also a key enzyme for the citric acid cycle) and/or decorin integrates signaling over multiple receptor-tyrosine kinases (e.g. EGF receptor). Thus, decorin suppresses the mitochondrial respiratory chain complexes via a Met-dependent pathway under normoxic and nutrient-rich conditions.
Decorin Evokes a Met-dependent Induction of Mitostatin-We hypothesized that decorin-mediated induction of mitostatin could suppress OXPHOS components by promoting mitochondrial turnover/degradation. Decorin induced mitostatin protein levels by ϳ2.5-fold (p Ͻ 0.001; Fig. 2A) with a concurrent relocation of mitostatin epitopes to large (2-3 m) perinuclear vacuoles, reminiscent of autophagosomes (white arrows, Fig. 2B). Importantly, these experiments were performed in nutrient-rich conditions. Incubation with HGF depressed mitostatin protein levels (p Ͻ 0.01, Fig. 2C), implicating Met in the negative control of mitostatin. Preincubation with HGF followed by decorin resulted not only in a complete block of decorin-evoked induction of mitostatin (p Ͻ 0.01, Fig.  2D) but also in lowering mitostatin below control levels, analogous to HGF alone (Fig. 2D). Depletion of endogenous Met via siRNA ( Fig. 2E) blocked the ability of decorin to evoke mitostatin as compared with either siMet alone or scrambled siRNA (siScr) (Fig. 2E). Thus, these novel findings indicate a dependence on Met for decorin to induce mitostatin in MDA-MB-231 cells, a process that is either blocked and/or subdued by the natural agonist, HGF.
PGC-1␣ Is Required for Decorin-evoked Mitostatin Induction-Next, we investigated whether decorin-mediated induction of mitostatin was due to suppression of mitochondrial biogenesis factors. It is well established that PGC-1␣ serves as the primary node in the transcriptional network for mitochondrial biogenesis (63). PGC-1␣ regulates loci encoding nuclear mitochondrial genes, TCA cycle enzymes, outer and inner membrane mitochondrial transporters, and TFAM (mitochondrial transcription factor A), the master regulator of mtDNA expression (63). Decorin significantly decreased PGC-1␣ mRNA (data not shown) and protein (p Ͻ 0.001, data not shown) in MDA-MB-231 and HeLa cells. In contrast, HGF substantially increased PGC-1␣ protein levels (data not shown). Exogenous decorin, after transient depletion of Met, resulted in no further suppression of PGC-1␣ relative to either decorin (p ϭ 0.474) or siScr (data not shown). Knockdown of Met alone decreased basal levels of PGC-1␣ (p Ͻ 0.01, data not shown), implicating Met as a necessary receptor-tyrosine kinase for steady state PGC-1␣.
Next, we tested whether a functional relationship existed between PGC-1␣ suppression and mitostatin induction as a mechanism for decorin-evoked suppression of OXPHOS. Unexpectedly, depletion of PGC-1␣ resulted in a concomitant loss of basal mitostatin relative to siScr and abrogated decorinmediated induction of mitostatin (p Ͻ 0.001, Fig. 3A). Paradoxically, the addition of decorin in the presence of PGC-1␣ siRNA resulted in further suppression of mitostatin (p ϭ 0.024, Fig.  3A), whereas loss of PGC-1␣ did not affect HGF-mediated reduction of mitostatin (Fig. 3A).
Following this observation, time-course experiments performed at steady state levels revealed a dynamic regulation of PGC-1␣ (Fig. 3B). As early as 15 min, decorin evoked a significant increase in PGC-1␣ levels (Ͼ3-fold) with maximal amounts occurring at 30 min; however, at later time points, PGC-1␣ returned to base line (Fig. 3B). Decorin evoked a steady increase of mitostatin that paralleled the co-regulation of PGC-1␣ up to 30 min (Fig. 3B). However, as PGC-1␣ returned to base line, mitostatin levels continued to increase and eventually plateaued at ϳ120 min (Fig. 3B). Immunofluorescence studies utilizing antibodies specific for PGC-1␣ corroborated the transient increase in PGC-1␣ at 15 and 30 min with subse-quent decline at 120 min (Fig. 3C). Importantly, induction of PGC-1␣ at 30 min was dependent upon positive Met signaling, as preincubation with the small molecule inhibitor, SU11274, blocked PGC-1␣ induction (Fig. 3D). Furthermore, confocal microscopy revealed rapid nuclear accumulation of PGC-1␣ with sustained nuclear occupancy (Ͼ2-fold over control) and lasting up to 60 min (Fig. 3E).
As a working model, we envision two distinct mechanisms of action accounting for this positive regulatory function of PGC-1␣. The first possibility is a post-translational mechanism involving direct protein-protein interactions between PGC-1␣ and mitostatin leading to accumulation of mitostatin. The second possibility would contemplate a post-transcriptional control whereby PGC-1␣ would stabilize mitostatin by directly binding to its mRNA via the C-terminal RNA recognition motif (RRM) found in PGC-1␣ (63,64). To test the first model, as discussed above, we preincubated the cells with cycloheximide to block new protein synthesis and then added decorin at various time points. We found a time-dependent increase in both PGC-1␣ and mitostatin, albeit with delayed kinetics due to the non-steady state context (Fig. 3F). The rate of PGC-1␣ induction was analogous to steady state conditions, with maximal PGC-1␣ levels occurring at 60 min (Ͼ22-fold) and remaining high for up to 90 min (p Ͻ 0.01, Fig. 3F). Although mitostatin displayed slower kinetics, it did increase concomitantly with the increase in PGC-1␣ at 60 min with plateauing at 90 min (Fig. 3F). These data indicate a dynamic regulation of PGC-1␣ and mitostatin under steady state and non-steady state conditions. Next, we tested whether there was any physical association between PGC-1␣ and mitostatin. Using co-immunoprecipitation with anti-PGC-1␣ and immunoblotting with anti-mitostatin, we found a constitutive association between PGC-1␣ and mitostatin, and this binding was increased by decorin (Fig. 3G,  left panel). Reciprocal co-immunoprecipitation experiments yielded analogous results (Fig. 3G, right panel), in contrast with a negative rabbit IgG control. Furthermore, quantification of the immunoprecipitated material after decorin stimulation revealed a significant (p Ͻ 0.01) increase in bound proteins (Fig.  3G, bottom). Collectively, decorin enhanced already preexisting binding between PGC-1␣ and mitostatin that might mechanistically underlie the requirement of PGC-1␣ for basal and decorin-induced mitostatin.
PGC-1␣ Stabilizes MITOSTATIN mRNA for Rapid Decorinevoked Induction-Next, we evaluated the second possibility of post-transcriptional control via PGC-1␣ and its C-terminal RRM. Cells were preincubated with the transcription inhibitor actinomycin D (ActD) to determine stability of PPARGC1A and MITOSTATIN mRNAs. Both mRNAs increased in a dose-dependent manner with respect to decorin, reaching ϳ6 and ϳ2-fold at 100 nM, respectively (p Ͻ 0.001, Fig. 4A). Timecourse experiments showed rapid stabilization of PPARGC1A and MITOSTATIN mRNAs starting as early as 5 min (p Ͻ 0.001, Fig. 4B). Importantly, PPARGC1A mRNA levels peaked dramatically (Ͼ7-fold) at 30 min and then declined to ϳ2-fold above control (p Ͻ 0.001, Fig. 4B). Concurrent with the sudden burst in PPARGC1A mRNA stabilization, MITOSTATIN mRNA was also stabilized at an increased rate at 30 min (Ͼ4fold) and remained elevated up to 60 min (p Ͻ 0.001, Fig. 4B).
To gain insight into whether PGC-1␣ would directly contribute to stabilizing MITOSTATIN mRNA, we performed additional mRNA stability assays under various conditions. First, we reproduced the decorin-evoked stabilization of PPARGC1A and MITOSTATIN mRNA after 30 min with each species stabilizing to a similar extent relative to the ActD-only treated control (p Ͻ 0.001, Fig. 4D). We then proceeded to deplete PGC-1␣ with targeting siRNA. After verification of PPARGC1A knockdown (Fig. 4C), we discovered that the loss of PGC-1␣ completely abrogated the ability of decorin to stabilize MITOSTATIN mRNA (p Ͻ 0.001, Fig. 4D). Protein methylation of RRMs contained within RNA-binding proteins catalyzed by protein arginine methyltransferases (PRMTs) is required for RNA-substrate interaction (63,65). As such, siRNA-mediated depletion of PRMT1, the isoform specifically responsible for arginine methylation of PGC-1␣ (64), phenocopied the effect after depletion of PGC-1␣ (p Ͻ 0.001, Fig. 4C) insofar as abrogating decorin-evoked stability of MITOSTA-TIN mRNA (Fig. 4D). Additionally, and consistent with previous reports (63) suggesting that PGC-1␣ binds and stabilizes its own mRNA, loss of PRMT1 also precluded stability of PPARGC1A (Fig. 4D). These data reinforce the concept that decorin promotes direct PGC-1␣ binding to MITOSTATIN mRNA as the molecular basis for rapid stabilization.
Consistent with the dependence of Met on decorin-mediated mitostatin induction, siRNA-mediated depletion of Met (Fig.  4C) resulted in a total block of PPARGC1A and MITOSTATIN mRNA stabilization (Fig. 4D). Surprisingly, this rapid stabilization for both mRNA species was entirely dependent on Met signaling as co-incubation with SU11274 completely inhibited stabilization of PPARGC1A and MITOSTATIN mRNA (Fig. 4D).
Parallel findings concerning the requirement of PGC-1␣ in mediating decorin-evoked MITOSTATIN mRNA stability were made in an independent breast carcinoma cell line, T47D. After verifying depletion of PPARGC1A mRNA under non-steady state conditions (Fig. 4E), we found that decorin stabilizes both PPARGC1A and MITOSTATIN transcripts (Fig. 4F). However, loss of mitostatin-prevented stabilization of MITOSTATIN mRNA, akin to the triple negative breast carcinoma cell model (cf. Fig. 4D).
Finally, we investigated the link between involvement of the C-terminal RRM within PGC-1␣ and the potential for direct binding to the MITOSTATIN mRNA via this protein module. To this end, cells were stably transfected with a PGC-1␣ construct harboring a truncated RNA recognition motif known as  FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8
Collectively, we provide a novel model whereby decorin requires Met signaling to stabilize PPARGC1A and MITOSTATIN mRNA. Depletion of PGC-1␣, prevention of PGC-1␣ methylation, or utilization of a PGC-1␣ truncation mutant deficient in the RRM all resulted in total loss of decorin-evoked mRNA stability.
Stabilized PPARGC1A and MITOSTATIN mRNAs Reflect Increased Mitostatin and Are Dependent on PGC-1␣-We determined whether the stabilized PPARGC1A and MITOSTATIN mRNAs yielded a corresponding increase at the protein level. Preincubation with ActD followed by stimulation with decorin revealed a significant increase of PGC-1␣ and mitostatin (p Ͻ 0.01, Fig. 5A). Analogous to the effect of PGC-1␣ depletion on mRNA stability, PGC-1␣ loss abrogated the ability of decorin to evoke stabilization of mitostatin protein (Fig. 5B). These data are in line with the requirement of PGC-1␣ for mitostatin induction during steady state. Next, we utilized the HA-PGC-1␣-⌬RRM construct to recapitulate the necessity of the RRM domain in mRNA stabilization. Unlike empty vector, expression of HA-PGC-1␣-⌬RRM completely blocked decorinevoked induction of mitostatin (p Ͻ 0.01, Fig. 5C), in full agreement with the mRNA data. These findings indicate a biologically relevant output for stabilized PPARGC1A and MITO-STATIN mRNAs and provide support for a key role of PGC-1␣ in this pathway and for RRM requirement to stabilize mitostatin mRNA.
PGC-1␣ Directly Binds MITOSTATIN mRNA via a Met-dependent Pathway-Having established a biological and functional link between PGC-1␣ and mitostatin, we evaluated the direct binding of PGC-1␣ to MITOSTATIN mRNA by per-forming RIP. Immunoprecipitation of PGC-1␣-positive ribonucleoprotein complexes followed by qPCR analysis revealed that, after a 30-min decorin treatment, PGC-1␣ protein was enriched Ͼ90-fold on MITOSTATIN mRNA as compared with negative rabbit IgG-G2 control (p Ͻ 0.001, Fig. 5D). Importantly, preincubation with SU11274 completely blocked PGC-1␣ binding (p Ͻ 0.001, Fig. 5D) to MITOSTATIN mRNA. In contrast, decorin failed to promote PGC-1␣ binding the Parkinson protein 7 (PARK7) mRNA, which encodes DJ-1 and functions as a positive regulator of androgen receptor signaling (66) (Fig. 5E), confirming specificity of the RIP as well as PGC-1␣ target selectivity. We further validated our RIP experiments via semiquantitative PCR to confirm the identity of mitostatin from our input and immunoprecipitated RNA. The MITOSTATIN primers used for analysis gave a 120-bp amplicon. PCR analysis of input and RIP samples revealed a specific 120-bp band representing the presence of the MITOSTATIN amplicon (Fig. 5F) and thereby further complementing the specificity of PGC-1␣ immunoprecipitation. Thus, we demonstrate a novel role for Met-dependent signaling to rapidly promote PGC-1␣ binding to MITOSTATIN mRNA under the influence of decorin. We have, therefore, discovered a novel regulatory function for PGC-1␣ to maintain a basal threshold of mitostatin as well as being required for decorin-mediated induction vis à vis the dynamic co-regulation between PGC-1␣ and mitostatin.
Decorin Disrupts Mitochondria Membrane Potential and Requires Mitostatin to Evoke Mitophagy-Based on decorin suppression of OXPHOS components and re-localization of mitostatin into large perinuclear vacuoles (cf. Fig. 2B), we tested whether decorin could evoke mitophagy in tumor cells via mitostatin. Early signaling events leading to a mitophagic response are initiated immediately after loss of mitochondrial membrane potential, ⌬m (67, 68). Thus, we assessed the effects of decorin on ⌬m using the voltage-sensitive dye JC-1. JC-1 is a lipophilic cationic dye that accumulates in the mitochondrial inner membrane in response to the ⌬m. At low ⌬m, JC-1 monomers show green fluorescence but at high levels of accumulation resulting from a high ⌬m leads to formation of JC-1 aggregates that shifts the JC-1 emission spectrum to red fluorescence (69). Therefore, red to green fluorescence ratio has been used as a semiquantitative indicator of ⌬m. After the addition of decorin for 6 h and utilizing live cell  FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8 JOURNAL OF BIOLOGICAL CHEMISTRY 4959 microscopy, we observed a significant loss of ⌬m (Fig. 6A) relative to control cells, similar to what was observed when depolarization was induced by the addition of the uncoupler FCCP (Fig. 6A). Quantification of two independent experiments revealed a significant decorin-mediated loss of ⌬m comparable to that of FCCP (p Ͻ 0.001, Fig. 6B).

Decorin Evokes Mitostatin-dependent Mitophagy
To further investigate decorin-evoked tumor cell mitophagy, we stably transfected MDA-MB-231 cells with GFP fused to the mitochondrial matrix protein, Su9, a subunit of the F0-ATPase (70). Decorin evoked mitochondrial fragmentation and aggregation into large autophagosomes vis à vis control (insets, Fig.  6C). Importantly, the decorin-evoked changes in mitochondrial morphology were identical to those evoked by rapamycin, an inhibitor of the mTORC1 nutrient sensor signaling complex and established autophagic inducer (71) (Fig. 6C). Depletion of mitostatin rendered the cells unresponsive to both decorin and rapamycin. Indeed, loss of mitostatin completely abrogated tumor cell mitophagy, as shown by the preservation and persistence of the mitochondrial network and characteristic tubular morphology while under either decorin or rapamycin (insets, Fig. 6C). Unlike control cells (supplemental Movie 1), decorin-treated cells revealed progressive fragmentation of the mitochondrial network culminating into collapse and aggregation in autophagosomes as visualized in real time via live cell confocal microscopy (supplemental Movie 2). Similar changes to mitochondrial structure were achieved with rapamycin (supplemental Movie 3), indicating that decorin causes effects similar to blocking the PI3K/Akt/mTOR pathway.
Quantification of mitochondrial morphology (58) revealed a significant increase in total mitochondrial number evoked by both decorin and rapamycin (Fig. 6D). Moreover, relative mitochondrial surface area and length were significantly decreased by both decorin and rapamycin (p Ͻ 0.001, Fig. 6, E and F, respectively), consistent with mitochondrial turnover and fragmentation. We also determined mitochondrial form factor, a measurement of "roundness" achieved by calculating the reciprocal of mitochondrial circularity; the latter becomes closer to 1 as the mitochondrial become more circular and serves as a valid prognosticator of mitophagy (58). Measurement of this parameter revealed that both decorin and rapamycin had form factor values closer to 1 vis à vis control mitochondria (p Ͻ 0.001, Fig.  6G). No significant difference existed between decorin-and rapamycin-treated cohorts (p ϭ 0.611), indicating that decorin evokes mitophagy similarly to rapamycin. Su9-GFP cells depleted of mitostatin via siRNA knockdown were resistant to both decorin and rapamycin treatments insofar as the mitochondrial network persisted as quantified for the analyzed parameters of mitochondrial number (Fig. 6D), occupied area (Fig. 6E), and mitochondrial length (Fig. 6F) were similar to controls. Furthermore, mitochondrial form factor returned to near-basal levels for both decorin and rapamycin (Fig. 6G). Thus, our findings support a role for decorin in evoking tumor cell mitophagy in a mitostatin-dependent manner.
Decorin Requires Mitostatin to Suppress OXPHOS Components and mtDNA-Utilizing a siRNA targeting mitostatin, we found that its loss completely abrogated decorin-mediated suppression of OXPHOS complexes II, III, and V (Fig. 7A). Interestingly, complex IV was sensitive to mitostatin loss alone (Fig. 7A). As a functional output for decorin-evoked mitophagy, we found a significant suppression of Su9-GFP (p Ͻ 0.001, Fig. 7B), which was blocked upon depletion of mitostatin. Similarly, the VDAC, a protein located on the outer mitochondrial membrane, was markedly suppressed by decorin (p Ͻ 0.01, Fig. 7C), and this effect was blunted by mitostatin knockdown.
Next, we evaluated mtDNA content in the presence of decorin and found a highly significant suppression of mtDNA (p Ͻ 0.001, Fig. 7E). Furthermore, validation of mitostatin suppression via targeting siRNA (Fig. 7D) mimicked the effects of mitostatin loss for Su9-GFP and VDAC insofar as mitostatin knockdown prevented decorin-mediated suppression of mtDNA (p Ͻ 0.001, Fig. 7E). Moreover, decorin suppressed both steady state and non-steady state mRNA levels of the TFAM, a major regulatory component of the mtDNA genetic program (Fig. 7F) (72). Depression of TFAM mRNA at the non-steady state was blocked by SU11274 (Fig. 7F). Importantly, we recapitulated our findings concerning the requirement of mitostatin in decorin-evoked mitophagy after validated knockdown of MITOSTATIN (Fig. 7G) in T47D cells prevented decorin-mediated mtDNA suppression (Fig. 7H) when compared with decorin stimulation alone (Fig. 7H).
Next, we determined if mitophagic induction vis à vis mtDNA reduction was dependent on Beclin 1 as decorin did not modulate Beclin 1 levels in MDA-MB-231 cells (Fig. 7I), as recently shown to occur for other transformed cell lines (73). After knockdown of Beclin 1 (cf. Fig. 7D), decorin still suppressed mtDNA similarly to decorin alone (p Ͻ 0.001, Fig. 7J). Treatment with rapamycin or Hanks' balanced salt solution (HBSS) to simulate nutrient deprivation reduced mtDNA to levels comparable to decorin alone (p Ͻ 0.001, Fig. 7J). Loss of mitostatin also abrogated the ability of both rapamycin and HBSS to induce mtDNA degradation (Fig. 7J). Collectively, these results corroborate the imaging studies using Su9-GFPexpressing cells where tumor cell mitophagy induced by  FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8 decorin or canonical agents such as rapamycin or HBSS depends strongly on mitostatin and is Beclin 1-independent.
Decorin Concurrently Induces Mitostatin and Suppresses mtDNA in Vivo-To translate our in vitro findings into a physiologically relevant setting, we established MDA-MB-231(GFPϩ) triple-negative orthotopic breast carcinoma xenografts. As such, introduction of ϳ3 ϫ 10 6 MDA-231(GFPϩ) cells into the mammary fat pads of SCID mice permitted, once the tumors became palpable, daily systemic administration of decorin protein core (10 mg/kg) via intraperitoneal injections. End point analysis of the breast carcinoma xenografts revealed a potent induction of MITOSTATIN mRNA (Fig. 9A) and protein, as confirmed by immunohistochemical staining (Fig. 9B) after systemic delivery of human recombinant decorin protein core (the same batch used for the current in vitro studies). Importantly, decorin suppressed mtDNA in the tumor xenografts (Fig. 9C). Collectively, these data support a role for decorin-mediated mitophagy within an established breast carcinoma xenograft model.  FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8

DISCUSSION
Mitophagy is a conserved homeostatic process by which unnecessary or compromised mitochondria are subjected to selective turnover and lysosomal-mediated degradation (74,75). Mitochondria, via oxidative phosphorylation, supply large amounts of ATP in quiescent, terminally differentiated cells. As molecular oxygen serves as the ultimate electron acceptor and due to the intrinsic nature of the OXPHOS process, mitochondria are the primary source of highly reactive oxygen species. Stringent safeguards that monitor mitochondrial integrity exist to circumvent the mutagenizing effects reactive oxygen species exert on subcellular structures, proteins, lipids, and nucleic acids. In the context of tumorigenesis, mitophagy is oncosuppressive to ensure mitochondrial quality control and preclude the propagation of damaged mitochondrial and genomic DNA (74). Decorin suppresses oncogenic signaling and curtails tumorigenic growth and angiogenesis (7,26,76) by triggering proteolysis of ␤-catenin and Myc (33) and inducing tumor suppressor genes such as p21 WAF1 , Peg3, and mitostatin (54). Furthermore, decorin antagonizes HIF1A expression leading to an attenuation of the angiogenic network under normoxia (42). As Myc and HIF-1␣ control opposing effects for metabolic reprogramming (1), recent research indicates a complex regulatory cooperation controlling mitochondrial biogenesis (59) that is inherently advantageous for neoplastic growth.  In this study we demonstrate for the first time that a soluble small leucine-rich proteoglycan member induces mitochondrial turnover and degradation in breast carcinoma cells as the underlying link for angiogenic suppression by relying on a novel mitochondrial localized tumor suppressor, mitostatin. We demonstrate that decorin evokes a protracted suppression of mitochondrial OXPHOS complexes, and this process is mediated by the decorin/Met axis. Notably, we were unable to detect complex I (NADH dehydrogenase) due to low abundance within our lysates. Complex I functions as an important regulator of the electron transport chain insofar as removing two electrons from reduced NADH via transfer to ubiquinone for shuttling. In concert with electron transfer from NADH, complex I simultaneously translocates four proton from the mitochondrial matrix into the intermembrane space. This begins the formation of a proton motive force and underlies the fundamental principle of oxidative phosphorylation for mass ATP production (74). Importantly, decorin significantly represses downstream complexes, although we cannot state conclusively the abrogation of this initial step. However, in the context of tumor mitochondrial physiology, suppression of downstream complexes is more consistent with inhibiting biosynthetic pathways necessary for tumorigenic growth. Interestingly, HGF/Met signaling augments glycolytic flux (77); however, we show a significant enhancement of OXPHOS components, potentially to support the cell with increased biosynthetic abilities consummate with the pro-proliferative and pro-angiogenic nature of HGF/Met signaling. These data substantiate a novel anti-tumorigenic role for decorin to compromise mitochondrial capacity (and potentially function) through the concerted suppression of multiple key components of OXPHOS (Fig. 10). This function is critical as metabolic intermediates are shunted to support the shift to anabolic metabolism (1).
Consistent with reports utilizing urothelial and prostate carcinoma cells (55), decorin induces mitostatin in basal breast carcinoma cells where it rapidly accumulates, beginning at 30 min and remains elevated for up to 4 h. Importantly, an antag-onistic relationship existed for regulation of mitostatin via the HGF/Met axis. Image analysis revealed a surprising and unexpected subcellular localization of mitostatin; that is, after decorin treatment; mitostatin localized to large (2-3 m) perinuclear structures reminiscent of autophagosomes. This provides tantalizing possibilities consistent with mitostatin localization to mitochondria that decorin evokes mitophagy in a mitostatin-dependent fashion, as the basis for OXPHOS suppression.
Mitochondrial homeostasis proceeds via a PGC-1␣-mediated transcriptional network (63). As decorin repressed PGC-1␣ mRNA and protein via Met, we hypothesized that an antagonistic relationship might exist between mitostatin and PGC-1␣ in controlling mitochondrial turnover. Unexpectedly, siRNA-mediated knockdown of PGC-1␣ did not lead to a derepression of mitostatin. In contrast, PGC-1␣ served as a positive regulator to maintain not only basal levels but also the decorinevoked induction of mitostatin. Paradoxically, depletion of PGC-1␣ was permissive for further mitostatin loss. These data indicate that other factors are involved in the decorin-evoked induction of mitostatin that are no longer active (or present) due to PGC-1␣ depletion.
We surmised two possible roles for PGC-1␣ to function as a positive regulator of mitostatin, either via direct protein-protein interactions or through a post-transcriptional pathway involving PGC-1␣ binding to MITOSTATIN mRNA via the C-terminal RNA recognition motif. We further refined our analyses to unravel the mechanistic relationship between PGC-1␣ and mitostatin. Reciprocal co-immunoprecipitation experiments revealed a novel protein complex containing PGC-1␣ and mitostatin, with enhanced binding in the presence of decorin. Regardless of decorin involvement, this new association of PGC-1␣ in complex with mitostatin has important regulatory ramifications for mitochondrial homeostasis and cellular bioenergetics. We demonstrated that knockdown of PGC-1␣ resulted in a loss of basal levels of mitostatin. Functionally, this might result in loss of mitostatin protein stability, coincident with increased mitostatin proteolysis via the 26 S proteasome, as basal PGC-1␣⅐mitostatin complexes exist constitutively and may aid in mitostatin protection.
Mechanistically, the underlying functional requirement of PGC-1␣ to maintain basal and induced levels of mitostatin involves post-transcriptional and post-translational, direct protein-protein regulatory events. As such, suppression of PGC-1␣ results in the complete abrogation of PPARGC1A and MITOSTATIN mRNA stability. These data are critical for the mechanistic role PGC-1␣ plays in regulating MITOSTATIN stability at both the mRNA and protein level in response to decorin (Fig. 10).
These lines of evidence converge to provide a causal relationship for PGC-1␣ to stabilize mitostatin mRNA and protein in response to matrix-derived signaling cues. Importantly, decorin requires Met signaling to induce stabilization in a PGC-1␣-dependent manner. An emerging mechanism of action for decorin is to act as a partial agonist to elicit downstream biological effects before receptor internalization and lysosomal degradation, as found previously for decorin/EGF receptor interactions to evoke thrombospondin-1 secretion (43). This could be mediated via differential phosphorylation of the unstructured cytoplasmic domains for combinatorial recruitment of signaling effectors.
The functional output and biological activity of mitostatin are general regulators of mitophagy for angiogenic suppression in tumor cells (Fig. 10). Decorin suppresses several mitochondrial markers including OXPHOS complexes, mtDNA, Su9, and VDAC in a mitostatin-dependent manner, constituting a mitostatin-dependent mitophagy program that is independent of Beclin 1. Moreover, loss of mitostatin also prevents rapamycin-induced mitophagy, thereby rendering mitostatin a master regulator of this critical cellular homeostatic pathway. Importantly, one of the known effector functions of mitostatin is to induce mitochondrial fragmentation (55), a requisite "first step" for mitochondrial autophagy before inclusion into mature LC3-positive autophagosomes (74). Interestingly, mitochondrial fragmentation appears to be Drp1a-independent (55). Mechanistically, mitostatin might induce mitophagy via suppression of mitofusion-1/2 activity or through modulation of the PINK1/Parkin signaling axis (74) where mitostatin might act as a receptor to recruit the Parkin E3 ubiquitin ligase, as ubiquitin signaling serves as the impetus for mitophagic signaling (74). It is also plausible that mitostatin, as a component of Mfn2-dependent mitochondrial-endoplasmic reticulum tethering complex, could evoke mitophagy via Mfn2-dependent fragmentation (78).
Decorin-mediated suppression of VEGFA depends on the capacity of decorin to induce mitophagy in a mitostatin-dependent manner. Loss of mitostatin severely compromises the ability of decorin to stunt VEGFA expression and secretion. Mitochondrial turnover and degradation directly interfere with mitochondrial reactive oxygen species generation and reactive oxygen species-mediated activation of the HIF-1␣/VEGFA pathway that would otherwise stimulate tumor angiogenesis under normoxia. Furthermore, this represents a novel feedback inhibitory control loop mediated by PGC-1␣ that could stunt quantity of mitochondria and ultimately compromise the ana-bolic metabolism necessary for tumorigenic growth and angiogenesis. Clearly, these results indicate complex regulatory circuits for PGC-1␣ to maintain proper mitochondrial integrity and homeostasis.
In conclusion, our data assign a regulatory paradigm for a decorin-inducible tumor suppressor gene to evoke mitochondrial autophagy in basal breast carcinoma cells. This induction is mediated through rapid stabilization of mitostatin in a PGC-1␣-dependent manner. Our data have broad implications for mitostatin-evoked mitophagy as a novel pathway amenable to therapeutic intervention in tumorigenesis and other pathobiological states where mitochondrial pathology is predominant.