Plant-derived mitochondria-targeting cysteine-rich peptide modulates cellular bioenergetics

Mitochondria are attractive therapeutic targets for developing agents to delay age-related frailty and diseases. However, few promising leads have been identified from natural products. Previously, we identified roseltide rT1, a hyperstable 27-residue cysteine-rich peptide from Hibiscus sabdariffa, as a knottin-type neutrophil elastase inhibitor. Here, we show that roseltide rT1 is also a cell-penetrating, mitochondria-targeting peptide that increases ATP production. Results from flow cytometry, live-cell imaging, pulldown assays, and genetically-modified cell lines supported that roseltide rT1 enters cells via glycosaminoglycan-dependent endocytosis, and enters the mitochondria through TOM20, a mitochondrial protein import receptor. We further showed that roseltide rT1 increases cellular ATP production via mitochondrial membrane hyperpolarization. Using biotinylated roseltide rT1 for target identification and proteomic analysis, we showed that human mitochondrial membrane ATP synthase subunit O is an intramitochondrial target. Collectively, these data support our discovery that roseltide rT1 is a first-in-class mitochondria-targeting, cysteine-rich peptide with potentials to be developed into tools to further our understanding of mitochrondria-related diseases.

Mitochondria, the main organelles regulating energy production, play vital roles in biosynthesis, intracellular signaling, and innate immunity (1)(2)(3)(4)(5). Mitochondrial dysfunction is the root cause of degenerative diseases, aging, and attendant aging issues, such as frailty and chronic inflammatory diseases (6 -9). Emerging mitochondira-targeting strategies to slow the decline of mitochondrial functions showed promise in delaying the onset and reducing the severity of age-related diseases that increases the health span of aging populations (10 -13).
Our laboratory has a long-standing interest in cysteine-rich peptides (CRPs) from plants, a class of naturally-occurring hyperstable constrained peptides (32)(33)(34)(35). Plant CRPs of 2-6 kDa are highly under-represented as therapeutics or as leads for developing therapeutics in natural product research. Most CRPs are structurally compact and metabolically stable (32,34) due to their evolutionarily conserved disulfide scaffolds and signature intercysteine loops that are hypervariable in sequence and size (32)(33). Importantly, these loops often contain functional sequences that contribute to their biological activities (32,36).
To date, there is no report of a naturally occurring plantderived CRP with mitochondria-targeting properties. Previously, we identified roseltide rT1 from Hibiscus sabdariffa as a knottin-type neutrophil elastase inhibitor (32). Roseltide rT1 shares certain biophysical features similar to mitochondria-targeting molecules and Leu/Ile-rich helical peptides. It is positively-charged, has a Leu/Ile-rich sequence with 85% hydrophobic amino acids residues (Fig. S1), and four inter-cysteine loops that form a structure that resembles a four-leaf clover. Thus, roseltide rT1 could represent a novel class of naturally occurring mitochondria-targeting CRPs (mtCRP) with no sequence homology or structural similarity to known mitochondria-targeting helical peptides.
Herein, we show that roseltide rT1 is a cell-penetrating peptide that targets mitochondria and increases ATP production. Using interactomic analysis, we also identified ATP synthase This work was supported in part by Competitive Research Grant NRF-CRP8-2011-05 from the National Research Foundation in Singapore, Nanyang Technological University Internal Funding-Synzymes and Natural Products (SYNC), and AcRF Tier 3 Grant MOE2016-T3-1-003. The authors declare that they have no conflicts of interest with the contents of this article. This article contains Figs. S1-S13, Videos S1 and S2, and Datasets 1 and 2. 1 Both authors contributed equally to the results of this work. 2 To whom correspondence should be addressed: School of Biological Sciences, Nanyang Technological University, 60 Nanyang Dr., 637551 Singapore. E-mail: JPTam@ntu.edu.sg. 3 The abbreviations used are: TOM, translocase of the outer membrane; CRP, cysteine-rich peptide; mtCRP, mitochondria-targeting CRPs; Fmoc, N- (9- cro ARTICLE subunit O as the putative intramitochondrial binding partner of roseltide rT1. Our findings highlight the characterization of a first-in-class, hyperstable, plant-derived mtCRP, which represents a promising lead to increase the health span of aging populations.

Chemical synthesis and characterization of roseltide rT1
To avoid ambiguity from contaminants, particularly small molecules from plant extracts during isolation of native roseltide rT1, only the synthetic version of roseltide rT1 was used in the current work ( Fig. 1). Synthetic roseltide rT1 was prepared by stepwise solid-phase synthesis using Fmoc chemistry. Deprotection and trifluoroacetic acid (TFA) cleavage released the linear roseltide rT1 precursor from the resin support. The linear precursor was immediately subjected to oxidative folding in 0.1 M ammonium bicarbonate at pH 8.0 in a mixture of redox agents, cysteamine/cystamine and 10% dimethyl sulfoxide (DMSO) for 1 h at 4°C to give an overall yield of 50%. Further purification using reversed-phase (RP) high-performance LC (HPLC) resulted in a final peptide purity of Ͼ90%. Natural and synthetic roseltide rT1 were identical as determined by MALDI-TOF mass spectrometry (MS), co-elution by RP-HPLC, and overlay of their two-dimensional NOESY spectra (Figs. S2 and S3).

Cellular uptake of roseltide rT1
Roseltide rT1 is both positively charged and hydrophobic, properties commonly found in cell-penetrating peptides (37,38). To determine the cellular uptake of roseltide rT1, flow cytometry and live-cell confocal microscopy were used. Roseltide rT1, which does not contain a lysine, was site-specif-ically conjugated at its N terminus using cyanine 3 (Cy3)-Nhydroxysuccinimide (NHS) ester. Cy3-labeled roseltide rT1 (Cy3-rT1) was purified by RP-HPLC and its identity confirmed using MALDI-TOF MS (Fig. S4). Flow cytometry was used to quantitatively measure the cellular uptake of Cy3-rT1 in human lung fibroblasts (WI-38) and human umbilical vein endothelial cells (HUVEC-CS). Incubation with 1 M Cy3-rT1 in both cell lines resulted in increased fluorescence intensity of the cell population and reached a plateau in 1 h ( Fig. 2A). Similar results were obtained from lymphocytes (Jurkat) and monocytes (THP-1) (Fig. S5), suggesting that cellular uptake of Cy3-rT1 is not cell-type specific. Live-cell imaging via confocal microscopy was used to prevent artifacts in the localization of labeled peptides in fluorescence microscopy (37). Fig. 2B shows an orthogonal view of the Z-stacked live-cell images of HUVEC-CS cells after incubation with 1 M Cy3-rT1 for 15 min. The confocal images showed that Cy3-rT1 was internalized and distributed throughout the cell with no accumulation in the nucleus.

Cellular uptake of Cy3-rT1 is glycosaminoglycan-dependent
Roseltide rT1 contains a positively charged residue in loop 1 that could bind to negatively charged glycosaminoglycans. To determine whether glycosaminoglycan expression facilitates cellular uptake of roseltide rT1 at the extracellular matrix (39), we compared glycosaminoglycan-deficient mutant PgsA-745 cells with WT CHO-K1 cells as a control. Both cell lines were incubated with Cy3-rT1 for different durations of time, up to 30 min. Fig. 2C shows that CHO-K1 cells internalized Cy3-rT1 in a time-dependent manner, and the mean fluorescence intensity at different time points was significantly higher than that of Cy3-rT1-treated PgsA-745 cells (p Ͻ 0.05).

Mitochondria-targeting cysteine-rich peptide Endocytosis mediates cellular uptake of Cy3-rT1
To determine whether the mechanism of Cy3-rT1 cellular uptake is mediated by endocytosis, Cy3-rT1 was incubated with HUVEC-CS cells at 4°C for 1 h. Fig. 2D shows that Cy3-rT1 cellular uptake was substantially reduced compared with uptake at 37°C. These results suggested that the majority cellular uptake of Cy3-rT1 involves energy-dependent endocytosis. To further support the involvement of endocytosis in the cellular uptake of Cy3-rT1, HUVEC-CS cells were preincubated with different endocytosis inhibitors for 30 min before being incubated with Cy3-rT1 for 1 h. The results showed that the dynamin-dependent endocytosis inhibitor dynasore inhibited cellular uptake of Cy3-rT1, supporting the involvement of clathrin-mediated endocytosis, but not receptor-and caveolinmediated endocytosis (Fig. 2E).

Cy3-rT1 is delivered to the mitochondrial compartment
To determine the subcellular localization of Cy3-rT1, we used organelle-specific fluorescent trackers to carry out colo-calization experiments with live-cell confocal microscopy. Fig.  3A shows that Cy3-rT1 colocalized with MitoTracker Green FM, suggesting Cy3-rT1 escaped from endosomes and relocalized to the mitochondria. As a control, roseltide rT7, a hydrophilic rT1-homolog that was also isolated from H. sabdariffa, was fluorescently labeled with Cy3-NHS ester (Cy3-rT7). Confocal microscopic analysis showed that hydrophilic rT7 does not colocalize with MitoTracker Green FM. To further support colocalization experiments, N-terminal biotinylated roseltide rT1 (biotin-rT1) was incubated with HUVEC-CS cells for 1 h. The resultant mixture was then subjected to mitochondria isolation experiments, and immunoblot results showed that biotin-rT1 accumulated in the mitochondrial fraction (Fig. 3B).

Identification of TOM20 as an intracellular target of roseltide rT1 for mitochondrial localization
The TOM family is involved in the mitochondria-targeting properties of signal peptides, especially TOM20 (40). To determine whether Leu/Ile-rich roseltide rT1 could interact with

Mitochondria-targeting cysteine-rich peptide
TOM20, we modeled their interactions using rigid body docking on the ClusPro version 2.0 server. Fig. 4A shows the X-ray crystal structure of the cytosolic domain of rat mitochondrial protein import receptor TOM20 (PDB entry 2V1T, chain A) (41). Based on this crystal structure, the four helices in the cytosolic domain of TOM20 form a hydrophobic groove that was reported to be the binding site for the pre-sequence peptide of rat aldehyde dehydrogenase (41). Fig. 4A shows the predicted computer model interaction of the leucine-rich intercysteine loop 2 of roseltide rT1 with the hydrophobic groove of TOM20. In support of modeling results, a pulldown assay followed by immunoblot with a TOM20 antibody showed that biotin-rT1 interacts with TOM20 ( Fig. 4B, Fig. S6). To confirm these results in vitro, we established a TOM20-green fluorescent protein (GFP) stably transfected HEK293 cell line for colocalization experiments using live-cell confocal microscopy. The results showed that Cy3-rT1 colocalized with TOM20-GFP in vitro (Fig. 4C, Videos S1 and S2) and that mitochondria uncoupler carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP) did not inhibit the mitochondrial localization properties of Cy3-rT1 (Fig. 4D). Using CRISPR-CAS9 -mediated genome editing, we disrupted the mitochondrial protein expression of TOM20 in HepG2 cells (Fig. S7) and showed that Cy3-rT1 localization with MitoTracker Green FM was reduced compared with the control (Fig. 4E).

Effects of roseltide rT1 on mitochondrial function
Because roseltide rT1 targets mitochondria intracellularly, we examined its effect on mitochondrial function. Using Seahorse XF Cell Mito Stress assays, we determined the effects of roseltide rT1 on mitochondrial respiration. The results showed that roseltide rT1 does not significantly change the overall oxygen consumption rate of HepG2 cells. Treatment with roseltide rT1 showed a reduction in the FCCP-induced oxygen consumption rate, which is a mitochondria uncoupler that induces uncontrolled oxygen consumption (Fig. 5A). This suggested that roseltide rT1 affects mitochondria function. Simultaneously, roseltide rT1 did not affect the extracellular acidification rate (Fig. S8). To gain further insight into the mitochondrial effects of roseltide rT1, intact mitochondria were isolated and assayed with tetramethylrhodamine (TMRE) staining. The results demonstrated that roseltide rT1 induced hyperpolarization of the mitochondrial membrane in a dose-dependent manner, similar to mitochondrial hyperpolarization inducer oligomycin ( Fig. 5B, Fig. S9). Additionally, roseltide rT1 increased reactive oxygen species levels in a dose-dependent manner using 2Ј-7Ј-dichlorodihydrofluorescein diacetate (DCFH-DA) staining (Fig. 5C). Using an ATP bioluminescent assay, roseltide rT1 increased the mitochondrial ATP level, whereas the effect can be inhibited by the mitochondria-uncoupling agent (Fig. 5D). Using C2C12 and HepG2 cells, roseltide rT1 increased cellular ATP levels in a dose-dependent manner (Fig.  5, E and F). Roseltide rT7 was used as a negative control that showed no significant changes in cellular ATP levels in HepG2 cells (Fig. S10).

Roseltide rT1 interacts with mitochondrial ATP synthase machinery
To identify the binding partners of roseltide rT1, pulldown assays followed by liquid chromatography (LC)-MS/MS analysis was performed to search for potential binding targets related to roseltide rT1 mitochondrial functions. After subtracting proteins found in control experiments, a total of 202 proteins were uniquely identified in the roseltide rT1 pulldown samples from which 86 proteins were discovered with multiple peptides (Dataset 1). The proteins DNA-J heat shock protein family 40 member C3, ATP synthase subunit 5O (ATP5O), and centrosomal protein 112 were identified in all three experimental repeats (Dataset 1). Fig. 6A shows the peptide sequence of ATP5O identified by LC-MS/MS analysis (Dataset 2). To confirm LC-MS/MS results, a pulldown assay was performed followed by Western blotting. Western blotting results supported the interaction between roseltide rT1 and ATP5O (Fig. 6B, Fig.  S11). By measuring the reverse reaction of complex V as an ATPase, roseltide rT1 was not found to affect complex V activity (Fig. S12). Interestingly, incubation of roseltide rT1 in isolated intact mitochondria causes the dissociation of sirtuin 3 (SIRT3), an ATP5O-binding partner, from the ATP synthase complex (Fig. 6C). This suggests that roseltide rT1 can promote SIRT3-mediated deacetylation in the mitochondria.

Discussion
Mitochondria are important therapeutic targets for delaying the onset of aging and age-related diseases (1, 2, 6, 8, 9). Sub- Mitochondria-targeting cysteine-rich peptide stantial efforts have been made to design and develop novel mitochondria-targeting molecules that include small molecules and peptides (15, 19 -21, 25-27). Recently, there has been increasing interest in peptide-based therapeutics due to their ease of synthesis, tunability, and biocompatibility (42)(43)(44)(45)(46). Currently, the majority of mitochondria-targeting peptides designed are partly based on mitochondria-targeting sequences that are invariably rich in positively-charged and hydrophobic residues and adopt amphipathic ␣-helical conformations (21,27). The present study reports the first naturally occurring plant-derived mtCRP, which represents a novel class of nonhelical, hyperstable mitochondria-targeting compound with a four-loop "clover-like" structure that is formed by three disulfide bonds.
Roseltide rT1 possesses an evolutionarily conserved cystineknot disulfide connectivity and a four-looped structural-fold (32). Based on our previously reported in silico model, we speculated that the intercysteine loop 1 of roseltide rT1 is responsible for its inhibitory activities against human neutrophil elas-tase (32). Functional roles of the other loops, however, were not determined. To broaden our understanding of the functional spectrum of roseltide rT1, solid-phase peptide synthesis and chemical biology were utilized to determine its intracellular functions. Amine-reactive chemistry using NHS ester was performed to prepare fluorescent-and biotin-labeled roseltide rT1 (Cy3-rT1 and biotin-rT1, respectively). Cy3-rT1 was observed to bind and internalize into a variety of cell lines, including fibroblasts (WI-38), endothelial cells (HUVEC-CS), hepatocytes (Hep-G2), lymphocytes (Jurkat), and monocytes (THP-1). Additionally, internalization of Cy3-rT1 was significantly reduced in glycosaminoglycan-deficient cells (PgsA-745) compared with WT (CHO-K1). These findings suggested that the cell-penetrating properties of roseltide rT1 are not cell-type specific but dependent on the expression of glycosaminoglycans at the extracellular matrix. As glycosaminoglycans are highly negatively charged molecules (47), we speculated that the positively charged intercysteine loop 1 of roseltide rT1 was responsible for its binding. . Roseltide rT1 interacts with TOM20 for mitochondrial localization. A, modeling the interaction between roseltide rT1 and TOM20 (PDB entry 2V1T) using the ClusPro Version 2.0 server. B, Western blot analysis using a TOM20 antibody revealed that biotin-rT1 binds to TOM20. C, Cy3-rT1 colocalizes with TOM20-GFP in stably-transfected HEK293 cells according to live-cell confocal microscopy. D, co-incubation of 1 M Cy3-rT1 with 1 M mitochondria uncoupler FCCP, did not block the mitochondrial localization properties of Cy3-rT1 according to live-cell confocal microscopy. E, depletion of TOM20 in HepG2 cells by CRISPR/CAS9 reduced the mitochondrial localization of Cy3-rT1 according to live-cell confocal microscopy.

Mitochondria-targeting cysteine-rich peptide
Direct cell penetration and endocytosis are two primary mechanisms involved in the cellular uptake of cell-penetrating peptides (38,48). Our results showed that cellular uptake of Cy3-rT1 was temperature-dependent and sensitive to the endocytosis inhibitor dynasore, supporting our findings that Cy3-rT1 penetrates cells through dynamin-dependent endocytosis mechanisms. Upon cellular entry, Cy3-rT1 was shown to localize in the mitochondria using live-cell confocal micros-copy and subcellular fractionation. Eukaryotes have thousands of mitochondrial proteins that are mostly synthesized as precursors with cleavable N-terminal presequences (mitochondria-targeting sequences). An example is the presequence peptide of rat liver aldehyde dehydrogenase (49). These mitochondriatargeting sequences are often recognized by the first protein import receptor at the outer membrane of mitochondria (i.e. TOM20) (50). Previous studies demonstrated that leucine res- Figure 5. Effects of roseltide rT1 on mitochondrial function. A, oxygen consumption rate (OCR) profile of HepG2 cells after incubation with roseltide rT1 via Seahorse XFe96 Flux Analyzer. B, effects of roseltide rT1 on mitochondrial membrane potential using TMRE-stained, isolated, intact mitochondria from HepG2 cells. C, effects of roseltide rT1 on mitochondria reactive oxygen species levels (ROS) using DCFH-DA-stained, isolated, intact mitochondria from HepG2 cells. D, effects of roseltide rT1 on mitochondrial ATP content using isolated, intact mitochondria from HepG2 cells. n ϭ 3; *, p Ͻ 0.05 compared with blank. Effects of roseltide rT1 on cellular ATP level in C2C12 (E) and HepG2 (F) cells. n ϭ 3; *, p Ͻ 0.05 compared with control.

Mitochondria-targeting cysteine-rich peptide
idues of the pre-sequence peptide of rat aldehyde dehydrogenase are essential for binding to TOM20 (49). The consensus binding motif of TOM20 is 1X2X345 ( ϭ hydrophobic amino acids, X ϭ any amino acids) (41). Because the primary sequence of roseltide rT1 has Ͼ85% hydrophobic residues and is Leu/Ile-rich, TOM20 poses as its potential intracellular target. Rigid surface modeling of the protein-protein interactions between roseltide rT1 and TOM20 were performed in this study. The leucine-rich intercysteine loop 2 of roseltide rT1 was predicted to interact with the hydrophobic groove formed by the four ␣-helices (␣1-␣4) of TOM20. In support of our predictions, pulldown assays showed the interaction of roseltide rT1 with TOM20, and confocal microscopy revealed the colocalization of Cy3-rT1 with TOM20-GFP live cells. Moreover, depletion of TOM20 significantly reduced the localization of Cy3-rT1 in the mitochondria. Taken together, these findings suggest that roseltide rT1 interacts with TOM20 and accumulates in the mitochondria.
Mitochondria produce Ͼ90% of the total cellular energy available, and defects in energy metabolism are a common cause of aging and age-related diseases (51)(52)(53). Results from the Seahorse XF Cell Mito Stress assay showed that roseltide rT1 affects mitochondrial function. Interestingly, roseltide rT1 was resistant to FCCP (mitochondria uncoupler)-induced uncontrolled oxygen consumption. It also triggered mitochondria hyperpolarization and reactive oxygen species generation, suggesting mitochondrial oxygen-dependent ATP synthesis, resulting in an overall increase in cellular ATP production. These results were in contrast to the classical effects of chemical entities (e.g. hydrogen peroxide (54)) that affect FCCP-induced oxygen consumption rates.
Interatomic analysis of roseltide rT1 identified ATP5O as its putative intramitochondrial target. ATP5O, commonly known as oligomycin-sensitive conferral protein, is the ␦-subunit of human mitochondrial F 1 F 0 -ATP synthase (55). Located at the peripheral stalk, ATP5O plays a vital role in tethering F 1 and F 0 , Figure 6. Roseltide rT1 interacts with ATP5O. A, peptide sequence (underlined) identified from the full sequence of ATP5O using interactomic analysis by LC-MS/MS. B, Western blot analysis using an ATP5O antibody revealed that biotin-rT1 binds to ATP5O. C, incubation of roseltide rT1 (20 M) with isolated intact mitochondria from HepG2 cells disassociated SIRT3 from ATP synthase according to Western blot analysis following ATP synthase immunoprecipitation.

Mitochondria-targeting cysteine-rich peptide
and holding the F 1 subunits together (55). Roseltide rT1 showed no effects on the ATP hydrolysis activities of ATP synthase, which suggests that the observed mitochondrial effects of roseltide rT1 were not due to a direct effect on ATP synthase activity but an indirect effect related to the binding partners of ATP5O (Fig. S12).
Recently, Yang et al. (56) reported SIRT3 as a ATP5O-binding partner. 50% of the mitochondrial proteome was reported to be acetylated and SIRT3 was found to be an NAD ϩ -dependent protein deacetylase (56). Upon mitochondrial membrane depolarization by an uncoupler, SIRT3 dissociates from ATP5O and deacetylates electron transport chain complexes (56). This results in increased electron transport efficiency, restoration of the proton gradient, and promotion of oxidative metabolism (56). This phenomenon was suggested to be an acute response to stress and provides fast recovery of mitochondrial function (56). Because roseltide rT1 binds to ATP5O, have counteractive effects against FCCP, and increases ATP production, it is possible that roseltide rT1 facilitates the dissociation of SIRT3 from ATP5O. This unique mechanism results in proton gradient generation and increased ATP production. Thus, we proposed that the mitochondrial functions of roseltide rT1 are, at least in part, a result of its putative role as a binding partner of ATP5O (Fig. 6D). Future studies are warranted to explore the relationship between the functions of roseltide rT1 and mitochondrial acetylation.
In conclusion, this study reports the chemical synthesis of roseltide rT1 and demonstrated its energy-dependent endocytic cell-penetrating properties. Upon cellular entry, roseltide rT1 interacts with TOM20 and binds to ATP5O, resulting in an increased ATP production. Taken together, these findings highlight roseltide rT1 as a first-in-class naturally occurring mtCRP with the potentials to be developed into a mitochondrial therapeutic.

Materials
All chemicals and solvents were purchased from Sigma, and ThermoFisher Scientific, unless specified otherwise.

Extraction and purification of roseltide rT1
Dried calyces (1 kg) of H. sabdariffa were extracted with water and centrifuged at 9000 rpm for 10 min at 4°C (Beckman Coulter). The supernatant was filtered and the filtrate loaded onto a C18 flash column (Grace Davison, USA) and eluted with 60% ethanol, 0.01% TFA. The eluted fractions were loaded onto an SP-Sepharose resin column (GE Healthcare, UK) and eluted with 1 M NaCl (pH 3.0), followed by ultrafiltration (ViVaflow 200, 2000 molecular weight cut-off hydrostat). Further purification was done by RP-HPLC (Shimadzu, Japan). A linear gradient of mobile phase A (0.05% TFA in H 2 O) and mobile phase B (0.05% TFA in acetonitrile (ACN)) was used with the C18 column (250 ϫ 22 mm, 5 m, 300Å; Grace Davison, USA). MALDI-TOF MS was used to identify the presence of roseltide rT1 in the eluted fractions.

Site-specific N-terminal fluorescent labeling of roseltide rT1
Synthetic roseltide rT1 was fluorescently labeled with Cy3 using NHS ester (Lumiprobe, USA) in 100 mM phosphate buffer (pH 7.8). Fluorescent labeling was carried out at room temperature for 16 h, and Cy3-rT1 was then identified and purified by RP-HPLC and MALDI-TOF MS (Fig. S4).

Site-specific N-terminal biotinylation of roseltide rT1
Synthetic roseltide rT1 was biotinylated with EZ-Link NHS-LC-biotin in 100 mM phosphate buffer (pH 7.8). Biotinylation was carried out at room temperature for 2 h, and biotin-rT1 was identified and purified by MALDI-TOF MS and RP-HPLC (Fig.  S13).

NMR spectroscopy
Natural and synthetic rT1 for NMR spectroscopy was prepared by dissolving the lyophilized peptide in d 6 -DMSO containing at a final peptide concentration of 1 mM. All NMR spectra were collected at a sample temperature of 298 K on a Bruker AVANCE II 600 MHz NMR spectrometer equipped with four RF channels and a 5-mm Z-gradient TCI cryoprobe. Phasesensitive two-dimensional 1 H, 1 H-TOCSY and NOESY spectra were recorded with a spectral width of 12 ppm. For water suppression, excitation sculpting with gradients was applied to all NMR experiments. TOCSY and NOESY spectra were obtained with mixing times of 80 and 200 ms, respectively. All measurements were recorded with 2048 complex data points and zerofilled to 2048 ϫ 512 data matrices. Time domain data in both dimensions were multiplied by a 90°shifted squared sine bell window function prior to Fourier transformation. Baseline correction was applied with a fifth order polynomial. NMR data were acquired and processed by TopSpin (Bruker BioSpin). The NMR spectra were processed with NMRFAM-Sparky.

Mitochondria-targeting cysteine-rich peptide
CHO-K1 cells (deficient in xylosyltransferase) were cultured in DMEM/Ham's F-12 medium containing 15 mM HEPES, L-glutamine, 10% fetal bovine serum, and 100 units/ml of penicillin and streptomycin and grown in a 5% CO 2 humidified incubator at 37°C. HEK293 cells were stably transfected with a TOM20-GFP plasmid (kindly provided by Prof. Li Yu of Tsinghua University, Beijing, China) by electroporation and selected using 500 g/ml of G418 in DMEM supplemented with 10% fetal bovine serum and 100 units/ml of penicillin and streptomycin. To generate CRISPR/Cas9-engineered HepG2 cells with depleted TOM20 expression, the TOM20 CRISPR guide RNA sequence (5Ј-TAAGCTCCCAACAATTAGTC-3Ј) was cloned into the pSpCas9 BB-2A-Puro (PX459) version 2.0 vector by GenScript (USA). The genetically-modified cells were obtained by electroporation and selected using 10 g/ml of puromycin, DMEM supplemented with 10% fetal bovine serum and 100 units/ml of penicillin and streptomycin. TOM20 expression was verified by Western blotting following mitochondria isolation (Fig. S7).

Cellular uptake analyses by flow cytometry
To study the cellular uptake of Cy3-rT1 by flow cytometry, cells were incubated with 1 M Cy3-rT1 in serum-free medium at 37°C. Following incubation, cells were harvested and collected by centrifugation at 500 ϫ g for 5 min. To quench extracellular fluorescence, cells were treated with 150 g/ml of trypan blue, and the samples were analyzed by flow cytometry. A total of 10,000 cells were analyzed using a BD LSRFortessa TM X-20 flow cytometer. For temperature-dependent uptake studies, HUVEC-CS cells were incubated at 4°C for 30 min prior to incubation with Cy3-rT1 for 1 h at 4°C. For endocytosis inhibitor studies, HUVEC-CS cells were pretreated with endocytosis inhibitors, including dynasore (50 M), ethylisopropylamiloride (50 M), and nystatin (50 g/ml) for 30 min, followed by incubation with Cy3-rT1 for 1 h at 37°C.

Confocal microscopy analysis
To examine the intracellular distribution of Cy3-rT1, cells were seeded on an 8-well chamber slide (Ibidi, Germany). Prior to incubation with Cy3-rT1, cells were stained with Hoechst 333241. Cy3-rT1 was incubated on cells in phenol red-free and serum-free medium at 37°C. The slides were washed gently with phosphate-buffered saline (PBS) three times, and the medium was replaced prior to imaging. Slides were observed using a Zeiss LSM 710 confocal microscope. For certain experiments, cells were stained with MitoTracker Green FM (100 nM) for 15 min before incubating with Cy3-rT1 (1 M) for 15 min.

Mitochondria isolation using magnetically activated cell sorting
Mitochondria were isolated from HepG2 cell lysates using anti-TOM22 magnetic microbeads according to the manufacturer's instruction (Miltenyl Biotec, Germany). Briefly, mitochondria were labeled with anti-TOM22 microbeads in separation buffer for 1 h at 4°C. The suspension was then passed through a 30-m filter and loaded onto an LS column. After washing the column with separation buffer, intact mitochondria were eluted using elution buffer.

Localization of roseltide rT1 by subcellular fractionation
HUVEC-CS cells were incubated with 20 M biotin-rT1 in serum-free medium at 37°C. Following incubation, the cells were subjected to mitochondria isolation using magneticallyactivated cell sorting. The cytoplasmic and mitochondrial fractions were resolved using Novex TM 10 -20% Tricine gels (Life Technologies) at 100 V for 120 min.

Pulldown assay
Pulldown assay was performed using NeutrAvidin UltraLink Resin. Briefly, the resin was washed with PBS three times and incubated with biotin-rT1 or biotin alone (control) at room temperature with rotation for 2 h. Bovine serum albumin (BSA; 2%) in PBS was added to both tubes and incubated at room temperature with gentle end-over-end mixing for another 2 h. HUVEC-CS cell lysate (600 g) was added to each tube and allowed to incubate overnight at 4°C with rotation. After incubation, the resin was transferred to Pierce Spin columns and washed 10 times with PBS. Then, 6ϫ loading dye with ␤-mercaptoethanol was added to the resin and heated for 10 min at 85°C. The resultant mixture was centrifuged at 200 ϫ g for 1 min and resolved using 15% SDS-PAGE at 100 V for 120 min.

Immunoprecipitation of ATP synthase complex V
Isolated intact mitochondria were incubated with roseltide rT1 (20 M) for 15 min, then ATP synthase complex V was immunoprecipitated using an ATP Synthase Immunocapture kit (Abcam). Briefly, mitochondria were lysed using lauryl maltoside. After centrifugation at 21,000 ϫ g for 10 min at 4°C, the supernatant was incubated with solid beads overnight at 4°C with rotation. The beads were then washed with wash buffer twice and eluted with 50 l of 1% SDS. The eluted fraction was mixed with 6ϫ loading dye with ␤-mercaptoethanol and heated for 10 min at 85°C. The resultant mixture was centrifuged at 14,000 ϫ g for 1 min and resolved using 12% SDS-PAGE at 100 V for 120 min.

Sample preparation and in-gel digestion for LC-MS/MS
Experimental and control samples from pulldown assays were resolved by 15% SDS-PAGE at 100 V for 120 min. Each sample lane was cut and sliced into three fractions according to their molecular weight. The gel pieces were reduced with 10 mM DTT for 30 min at 60°C and alkylated with 55 mM iodoacetamide in the dark for 45 min at room temperature. The alkylated samples were subjected to tryptic digestion (Promega) at 37°C overnight. Tryptic peptides were then extracted with 5% acetic acid, 50% ACN buffer and evaporated by vacuum centrifuge (Thermo Electron).

LC-MS/MS
LC-MS/MS analysis was performed using an online Dionex UltiMate 3000 UHPLC system coupled with an Orbitrap Elite mass spectrometer (Thermo Scientific Inc., Germany). Tryptic peptides were dissolved in 0.1% formic acid solution and separated by a Acclaim PepMap RSL column (75 m inner diameter, 15 cm, 2 m particle size) using a 60-min gradient of mobile phase A and mobile phase B (0.1% formic acid in 90% ACN). Samples were sprayed through a Michrom Thermo Captive Spray nanoelectrospray ion source (Bruker-Michrom Inc.) with a source voltage of 1.5 kV. A full MS scan range was set at 350 -1600 m/z and resolution of 60,000 at 400 m/z. The Fourier transform-MS/MS scan range was set at 150 -2000 m/z, with a resolution of 15,000 at 400 m/z. The 10 most intense ions with a threshold of 500 counts were selected for high-energy collisional dissociation fragmentation and fragmented using 32% normalized collision energy with a maximum ion accumulation time of 120 ms. Parameters including an automatic gain control of 1Eϩ06 for the full MS scan and 2Eϩ05 for the MS/MS scan, active precursor ion charge state screening, and capillary temperature of 250°C were set for the experiment. Data acquisition was conducted in positive mode using LTQ Tune Plus software.

LC-MS/MS data analyses
Mascot generic format files were generated from the raw data using Proteome Discoverer 1.4.1.14 software prior to Mascot searching. All protein sequence database searches were performed on the in-house Mascot search engine (version 2.4.1; Matrix Science, UK) using the UniProt Knowledge Base human database along with the reverse sequences (downloaded on March 2016, including 3,47,018 sequences and 121,280,492 residues). Carbamidomethyl at cysteine was set as a static modification and methionine oxidation and asparagine/glutamine deamidation as dynamic modifications. Full trypsin digestion with a maximum of 2 missed and/or nonspecific cleavages was set as the digestion parameter. 10 ppm for precursor mass and 0.02 Da for fragment mass were set as mass tolerance parameters. The #13C value of 2 was set as other search parameter. Data with a significance threshold value of p Ͻ 0.05 and "Ignore Ions Score Below" value of 20 were extracted in *csv format. A target-decoy search strategy with the cutoff set to less than 1% false discovery rate and proteins identified with multiple peptides were applied for the selection of proteins for final analysis. The exponentially modified protein abundance index value for each identified protein was calculated by Mascot and used for label-free quantification.

Oxygen consumption and extracellular acidification rates
A Seahorse Bioscience Mito stress test kit was used to determine the oxygen consumption and extracellular acidification rates as per the manufacturer's instructions (Seahorse Bioscience). Briefly, HepG2 cells were treated with roseltide rT1, and extracellular oxygen levels and pH were measured in real time using an XFe96 Flux Analyzer (Seahorse Bioscience) by Cyprotex Discovery, Ltd. (UK).

Mitochondrial membrane potential
Mitochondrial membrane potential was determined using TMRE staining of isolated intact mitochondria. Briefly, 0.1 mg of isolated mitochondria from HepG2 cells were incubated with TMRE (200 nM) for 15 min. Then, mitochondria were incubated with roseltide rT1, and fluorescent intensity was measured using a microplate reader at 549 nm excitation and 575 nm emission wavelengths (Tecan Infinite 200 Pro, Switzerland).

Mitochondrial reactive oxygen species levels
Mitochondrial reactive oxygen species levels were determined using DCFH-DA staining of isolated intact mitochondria. Briefly, 0.1 mg of isolated mitochondria from HepG2 cells were incubated with DCFH-DA (10 M) for 10 min. Then, mitochondria were incubated with roseltide rT1, and fluorescent intensity was measured using a microplate reader at 485 nm excitation and 520 nm emission wavelengths (Tecan Infinite 200 Pro, Switzerland).

ATP bioluminescent assay
Cellular ATP levels were determined using a CellTiter-Glo luminescent cell viability assay as per the manufacturer's instructions (Promega). Briefly, C2C12 and HepG2 cells were incubated with roseltide rT1 for 4 h, and 100 l of CellTiter-Glo reagent was added to each well. Roseltide rT7 was used as negative control. For ATP measurement in isolated intact mitochondria from magnetically activated cell sorting, isolated mitochondria were incubated with roseltide rT1 with/without FCCP for 1 h, and 100 l of CellTiter-Glo reagent was added to each well. Following incubation, the luminescent intensity was measured using a microplate reader (Tecan Infinite 200 Pro, Switzerland).

Mitochondria isolation by reagent-based method
To examine the mitochondrial protein expression of TOM20 in HepG2 cells after CRISPR-CAS9-mediated genome editing, mitochondria were isolated using a commercial mitochondria isolation kit (reagent-based method) according to manufacturer's instructions (ThermoFisher).

ATP synthase (complex V) activity
The effects of roseltide rT1 on ATP synthase (complex V) enzyme activity were determined using a commercial ATP syn-Mitochondria-targeting cysteine-rich peptide thase specific activity microplate assay kit according to the manufacturer's instructions (Abcam).

In silico modeling
The in silico docking was performed using automatic protein-protein docking server ClusPro version 2.0 (57,58). To model the interactions between roseltide rT1 and TOM20, both the NMR structure of roseltide rT1 (PDB entry 5GSF) and the crystal structure of the cytosolic domain of rat TOM20 (PDB entry 2V1T) (59) were uploaded to the server. It uses a rigid body docking protocol, and the model was generated based on both electrostatic potentials and hydrophobic interactions.

Statistical analyses
Statistical comparisons were performed using GraphPad version 6.0d (USA). The data were analyzed by one-way analysis of variance followed by Newman-Keuls post hoc test. Data are expressed as the mean Ϯ S.E.M. p Ͻ 0.05 was considered statistically significant.