Restoring mitochondrial superoxide levels with elamipretide (MTP-131) protects db/db mice against progression of diabetic kidney disease

Exposure to chronic hyperglycemia because of diabetes mellitus can lead to development and progression of diabetic kidney disease (DKD). We recently reported that reduced superoxide production is associated with mitochondrial dysfunction in the kidneys of mouse models of type 1 DKD. We also demonstrated that humans with DKD have significantly reduced levels of mitochondrion-derived metabolites in their urine. Here we examined renal superoxide production in a type 2 diabetes animal model, the db/db mouse, and the role of a mitochondrial protectant, MTP-131 (also called elamipretide, SS-31, or Bendavia) in restoring renal superoxide production and ameliorating DKD. We found that 18-week-old db/db mice have reduced renal and cardiac superoxide levels, as measured by dihydroethidium oxidation, and increased levels of albuminuria, mesangial matrix accumulation, and urinary H2O2. Administration of MTP-131 significantly inhibited increases in albuminuria, urinary H2O2, and mesangial matrix accumulation in db/db mice and fully preserved levels of renal superoxide production in these mice. MTP-131 also reduced total renal lysocardiolipin and major lysocardiolipin subspecies and preserved lysocardiolipin acyltransferase 1 expression in db/db mice. These results indicate that, in type 2 diabetes, DKD is associated with reduced renal and cardiac superoxide levels and that MTP-131 protects against DKD and preserves physiological superoxide levels, possibly by regulating cardiolipin remodeling.

The steadily growing number of patients worldwide with diabetes is projected to increase the prevalence of DKD 4 (1). Exposure to chronic hyperglycemia modulates several pathological pathways, including intracellular signaling pathways, transcription factors, cytokines, chemokines, and growth factors, that lead to development and progression of DKD (2). Reactive oxygen species (ROS) play a crucial role in progression of DKD; however, the role of ROS, particularly that of superoxide (O 2 . ), in DKD is controversial. Under healthy conditions, normal levels of O 2 . and hydrogen peroxide (H 2 O 2 ) production have been shown to be critical in regulating several fundamental biological processes, including signal transduction, regulation of immune function, and cell growth (3). In mouse models of DKD, overexpression of cytosolic superoxide dismutase (SOD1) protects against mesangial expansion (4), and genetic ablation of SOD1 accelerates renal matrix deposition (5). Conversely, deletion of mitochondrial SOD (SOD2) does not promote susceptibility to DKD (6). In addition, we recently found that O 2 . production is reduced in the kidneys of mice with type 1 diabetes (6) and that enhancement of mitochondrial biogenesis and oxidative phosphorylation lead to increased O 2 . production but with beneficial effects on kidney function and structure. However, direct measurement of the O 2 . radical is difficult because of its short half-life (7,8). Dihydroethidium (DHE; hydroethidine (9)) is a cell-permeable compound that is rapidly oxidized in the presence of O 2 . but not H 2 O 2 or peroxynitrite (10). It subsequently interacts with DNA to emit a bright red color (11). An important caveat of this approach is that interpretation of DHE experiments requires special attention, as DHE is particularly prone to auto-oxidation in ambient oxygen (12). Two major oxidation products of DHE are ethidium (E ϩ ) and 2-hydroxyethidium (2-OH-E ϩ ).
Although in vitro studies using HPLC have suggested that the specific oxidation product of DHE by O 2 . is 2-OH-E ϩ (7,13,14), a recent report by Hall et al. (9) demonstrated the discrepancy between in vitro and in vivo results for DHE oxidation products. Using a novel in vivo live animal imaging method that utilized the difference of fluorescence lifetimes between the two DHE oxidation products (E ϩ , 6 ns; 2-OH-E ϩ , 12 ns), it was shown that the predominant oxidation product in vivo was E ϩ , not 2-OH-E ϩ (9). We recently applied a similar in vivo fluorescence lifetime measurement and confocal imaging to examine renal O 2 .
production in mouse models of type 1 diabetes (6). Surprisingly, O 2 . production was reduced not only in the kidneys but also in heart and liver tissues of type 1 diabetes compared with nondiabetic controls (6). The reduction in renal O 2 .
production was associated with reduced mitochondrial electron transport chain activity with diabetes or with in vivo inhibition of complex I activity. Interestingly, stimulation of mitochondrial electron transport chain activity led to an increase in renal O 2 . production and reduced evidence of DKD. In support of a similar pattern in human DKD, several metabolites related to mitochondrial function were reduced in the urine of patients with DKD, and kidney biopsies exhibited a reduction in mitochondrial biogenesis in patients with chronic kidney disease (15,16).
In this study, we used a murine model of type 2 diabetes (db/db) and applied confocal imaging of DHE oxidation to examine O 2 . production in the kidney. As mitochondrial function is reduced in DKD (6), we also evaluated the role of a novel mitochondrion-targeting peptide agent, MTP-131. MTP-131 is a mitochondrion-targeted tetrapeptide (H-D-Arg-Dmt-Lys-Phe-NH 2 ). It has been shown that MTP-131 has many renoprotective effects in various animal models of kidney disease and cell culture studies, including renal ischemic-reperfusion injury (17)(18)(19), unilateral ureteral obstruction (20), and hypoxia/reoxygenation-induced tubular injury (21). The mechanism of these effects has generally been ascribed to reduction of ROS and/or improvement in mitochondrial function and efficiency. We hypothesized that MTP-131 has a protective effect in the kidneys of diabetic db/db mice via optimization of renal O 2 .
production. We further examined whether MTP-131 had an impact on renal cardiolipins (CLs), a class of key phospholipids that play a major role in restoring normal mitochondrial complex activity (18).

Diabetic db/db mice have reduced renal O 2 . production
To evaluate O 2 . production in the kidneys, 18-week-old db/db and db/m mice underwent administration of DHE for 16 h prior to euthanasia, and kidneys were harvested (6); at this point, all unreacted DHE has been excreted (9). The DHE oxidation signal was prominent in the kidneys of db/m mice (Fig. 1A), but, as expected, diabetic db/db kidneys showed significantly reduced DHE oxidation in glomeruli and cortical tubules (Fig. 1, B-D), indicating reduced O 2 . production.

Metabolic characteristics with MTP-131 treatment
Body weight ( Fig. 2A), food intake ( Fig. 2B), and glycated hemoglobin (HbA1c) (Fig. 2C) were significantly increased in the db/db groups compared with the nondiabetic (db/m) groups. There were no differences between the db/db and db/dbϩB (Bendavia/MTP-131) groups during the study period and at the end of the study. Kidney weight per tibial length was increased in the db/db group; however, no difference between treated and untreated diabetic groups was observed (Fig. 2D). Perigonadal fat and liver weights per tibial length were also increased in the db/db group compared with the db/m group (Fig. 2, E and F). There were no significant differences, but a slight trend was noted of a reduction in liver and perigonadal fat weights by MTP-131 treatment. There were no statistical differences in heart weight between groups (Fig. 2G).

MTP-131 protects against progression of DKD
Representative findings of the glomeruli in PAS-stained sections are shown in Fig. 3A. Glomerular hypertrophy was observed in both diabetic db/db groups compared with the nondiabetic db/m groups (Fig. 3, A and B). MTP-131 treatment showed a slight improvement in glomerular size, but this was not significant (Fig. 3, A and B). However, mesangial matrix expansion was observed in the db/db group, and MTP-131 treatment significantly reduced mesangial matrix accumulation compared with the db/db group (Fig. 3, A and C). In the db/db group, the urinary albumin/creatinine ratio was significantly increased at 10, 14, and 18 weeks of age and significantly suppressed in the db/dbϩB group at 14 and 18 weeks of age (Fig. 3D). Urinary hydrogen peroxide/creatinine levels were markedly increased in the db/db group at 10 and 18 weeks age compared with the db/m groups and sig-

MTP-131 reduces perigonadal adipocyte size in db/db mice
We also examined the effect of MTP-131 on adipocyte size to verify whether systemic administration of MTP-131 affected other organs. Representative H&E-stained perigonadal adipose tissue images are shown in Fig. 4A. Interestingly, average adipocyte size was markedly increased in the db/db group and significantly reduced by MTP-131 treatment (Fig. 4, B and C).

MTP-131 preserved renal O 2 . production in db/db mice
We then examined the effect of MTP-131 on renal O 2 .
production (as measured by DHE oxidation using confocal microscopy) in db/db mice. Renal tubules showed intense ethidium staining in db/m mice, whereas glomeruli were quite faint (Fig. 5A). db/db mice showed a significant reduc-tion in renal DHE oxidation compared with the db/m group; however, MTP-131 significantly preserved renal O 2 . production in cortical tubules and glomeruli (Fig. 5, A-C). Interestingly, MTP-131 increased DHE oxidation in cortical tubules in the db/m group (Fig. 5C). We further examined DHE oxidation using multiphoton microscopy to construct renal 3D images. 3D DHE imaging demonstrated DHE fluorescence primarily in the cytosolic compartment of tubular cells, with reduction of DHE oxidation in db/db mice and restoration by MTP-131 treatment (Fig. 5D). As with the renal cortex, DHE oxidation was also significantly decreased in medullary regions of the db/db group and fully preserved by MTP-131 treatment (Fig. 5, E and F). We also measured the protein levels of the five oxidative phosphorylation system complexes I-V in renal cortical tissues; however, none of them showed significant differences among four groups of mice (Fig. S1).

MTP-131 protects against diabetic kidney disease
As it has been reported that MTP-131 plays protective roles in the heart (22,23), we also examined O 2 . production in the heart. Although the db/db group showed a significant reduction in DHE oxidation, MTP-131 did not maintain O 2 . production in heart tissues (Fig. 5, G and H).

MTP-131 protects against diabetic kidney disease
there was no difference in total CL content between the groups (Fig. 6A). However, total lysoCL levels were markedly increased in the db/db group and significantly reduced by treatment with MTP-131 (Fig. 6B). We examined whether the protein content in the kidney cortex affected the measurement of CL or lysoCL, but there was no difference between the groups (Fig. 6C).

MTP-131 protects against diabetic kidney disease
was decreased in the kidney cortex of MTP-131-treated db/db mice compared with db/m mice with or without MTP-131 (Fig.  7F). These observations clearly indicate that CL species are remodeled to those containing longer fatty acyl chains, as reported previously for diabetic mouse hearts (29), and that MTP-131 treatment protects normal regulation of CL synthesis and remodeling.

MTP-131 regulates mitochondrial fusion machinery in the kidneys of db/db mice
As CL plays important roles in mitochondrial dynamics (30), we measured relative expression of genes regulating mitochondrial fusion and fission. The expression of optic astrophy 1 (Opa1), which regulates mitochondrial fusion and crista structure in the inner mitochondrial membrane and contributes to ATP synthesis and apoptosis (31,32), was decreased in db/db mice versus the db/m control (Fig. S2A). In addition, the expression of mitofusin 1 (Mfn1), a mediator of mitochondrial fusion in the outer mitochondrial membrane (33,34), was up-regulated after MTP-131 treatment in db/db mice compared with the db/db mice with vehicle treatment (Fig. S2B). Another dynamin-related GTPase, mitofusion 2 (Mfn2), did not show a difference among the four groups of mice ( Fig. S2C. No significant changes were observed in the level of mediator (e.g. dynamin 1 (Dnm1)) or its receptors (e.g. mitochondrial fission 1 protein (Fis1)), involved in mitochondrial fission (Fig. S2, D and  E). These results further confirm that diabetic kidneys have reduced mitochondrial fusion and that MTP-1 treatment may protect mitochondrial fusion. We also assessed the mRNA abundance of several mitochondrial biogenesis transcription factors (mitochondrial transcription factor A (Tfam), peroxisome proliferator-activated receptor ␣ (Ppara), nuclear respiratory factor 1 (Nrf1), and nuclear respiratory factor 2 (Nrf2); Fig. S2, F-I) and found it to be unchanged.

Discussion
Our findings reveal that db/db mice, a well-accepted model of type 2 diabetes, experienced a significant reduction of renal O 2 . in association with increased albuminuria, elevated urine

MTP-131 protects against diabetic kidney disease
in both studies, frozen kidney sections were used, and DHE was applied under ambient oxygen conditions, where auto-oxidation may occur rapidly. Therefore, in this study, DHE was prepared under anaerobic conditions and immediately delivered intraperitoneally for in vivo measurement of O 2 . production.  (37). In fact, it has been reported that generation of NADPH oxidase (Nox)-derived ROS is increased in the kidneys of multiple models of DKD, with Nox4 likely the most important isoform that is up-regulated in the diabetic kidney (38). In addition, Nox4 can be a potent source of H 2 O 2 production, independent of O 2 . production (39). The increased H 2 O 2 is unlikely to originate from mitochondria, as we have reported that isolated mitochondria from diabetic mice have reduced H 2 O 2 production (6).
Interestingly, it has been reported that MTP-131 can directly scavenge ROS, including H 2 O 2 and free radicals; in addition, the Tyr and 2,6-dimethyltyrosine residues of the peptide play critical roles in its ROS-scavenging ability (19). However, a recent study was unable to detect any scavenging activity of MTP-131 (40). Thus, further work will be needed to determine the mechanism of reduction of urinary H 2 O 2 by MTP-131. MTP-131 can also inhibit high-fat diet-induced insulin resistance in muscle tissue (41) and islet cell apoptosis (42). In this study, although there was no difference in HbA1c levels or overall body weight between the db/db and db/dbϩB groups, MTP-131 treatment partially reduced visceral adipocyte size, which may reduce inflammation and insulin resistance.
The effects of MTP-131 are thought to be partly mediated by interaction with CL and inhibition of cytochrome c-mediated peroxidation (18). In this study, there was no detectable difference in total renal CL content between the control and db/db groups (Fig. 6A), indicating that total CL remains sustained despite severe hyperglycemia. In contrast to total CL content, there was a clear increase in renal total lysoCL content in db/db mice, which was significantly inhibited by MTP-131 treatment (Fig. 6B). It has been reported that an increase in lysoCL is associated with CL oxidation and suppression of CL oxidation by a mitochondrion-targeted nitroxide (XJB-5-131) prevented CL oxidation and hydrolysis (43), suggesting that the much smaller lysoCL pool is a more sensitive end point than the total CL pool (Fig. 6, A versus B).
CL is one of the most unsaturated lipids in the body, and the degree of unsaturation is maintained by means of a constant remodeling process (44). Pla 2 is a key enzyme that catalyzes hydrolysis of immature CL to lysoCL, which is involved in CL remodeling (25,26,28). In the remodeling process of CL, LCLAT1, also known as ALCAT1 or LYCAT, catalyzes acylation of lysoCL to produce CL (27,28,44). In this study, up-regulation of Pla 2 and reduction of LCLAT1 expression in the kidneys of db/db mice may lead to accumulation of lysoCL in

MTP-131 protects against diabetic kidney disease
db/db mice. In addition, our study uniquely demonstrates that MTP-131 can suppress Pla2 expression, preserve LCLAT1 expression, and normalize lysoCL content in the kidneys of db/db mice. This may prove to be a key mechanism by which MTP-131 provides renoprotection via regulation of CL remodeling pathway.
However, the roles of LCLAT1 in various animal disease models remain controversial. Li et al. (27) reported that LCLAT1-overexpressing mouse myoblast cells show a reduction of total CL contents compared with the vector control. In addition, they showed that LCLAT1 deficiency prevents onset of diet-induced obesity and improves mitochondrial complex I activity and lipid oxidation in the liver. Interestingly, LCLAT1 whole-body-deficient mice also show hyperactivity, hyperphagia, and hypermetabolism (27). However, these changes in phenotype may also directly affect mitochondrial function and systemic metabolic changes. It has been reported that LCLAT1 mRNA and protein expression are significantly decreased in heart tissues of patients with tetralogy of Fallot, a common form of cyanotic congenital heart defect; however, the mechanism underlying the regulation of LCLAT1 remains unknown (45). On the other hand, LCLAT1 expression is increased in lungs from patients with idiopathic pulmonary fibrosis and in murine models of lung fibrosis (46). It is noteworthy that LCLAT1 mRNA expression levels in peripheral blood mononuclear cells (PBMCs) positively correlates with improved lung function and survival rate in patients with idiopathic pulmonary fibrosis. In addition, overexpression of LCLAT1 attenuates fibrogenesis, and knockdown of LCLAT1 accentuates inflammation and injury in murine models of lung fibrosis, suggesting protective roles of lysocardiolipin acyltransferase 1. The latter data are consistent with the findings in this study, as there was a reduction in matrix accumulation in association with an increase in LCLAT1. The protective role of MTP-131 may be partly explained by stimulation of renal LCLAT1 expression.
Reduced content of immature CL and increased levels of lysoCL species in diabetic kidneys indicate enhanced CL hydro-lysis (i.e. up-regulated Pla 2 ) and loss of CL remodeling activity, such as LCLAT1 activity. However, long-chain mature unsaturated CLs were increased in diabetic kidneys and reduced by MTP-131 treatment. Accumulation of CL species containing longer fatty acyl chains (greater than 18 carbon atoms) has been reported as a detrimental factor in diabetic conditions (e.g. diabetic myocardium) (29). It has been reported that tetra 18:2 CL is a fully functional CL species, and increases in long chain CLs are associated with mitochondrial dysfunction (29,47,48). Although LCLAT1 has no acyl-chain specificity, we postulate that LCLAT1 tends to remodel lysoCL to more long-chain unsaturated mature CL in diabetic kidneys. We also found that MTP-131 might protect mitochondrial function by suppressing CL hydrolysis and reducing levels of long-chain (Ͼ18C) unsaturated CLs.
The profile of CL and lysoCL after remodeling in diabetic kidneys leads to mitochondrial dysfunction, likely because of increased basal uncoupling and H ϩ leakage. CL play an important role in regulating the uncoupling activity of uncoupling proteins (UCPs) and H ϩ translocation through the membrane (49). Modest depolarization of the mitochondrial inner membrane is known to attenuate mitochondrial ROS, such as O 2 .
generation (50). Normal CL could maintain proper activation of UCPs for mitochondrial depolarization to attenuate electron transport chain ROS production without disturbing ATP production (50). The mechanism underlying the role of lysoCL and long-chain unsaturated mature CL in diabetic kidneys might be related to enhanced activation of UCPs and leaking of H ϩ across the mitochondrial inner membrane, dissipating the H ϩ gradient generated by the respiratory chain. This reduced respiration could lead to reduced O 2 . production compared with normal conditions. MTP-131 treatment could reverse the CL remodeling in diabetic kidneys and preserve normal O 2 . levels.
The mechanisms of disordered CL remodeling and mitochondrial dysfunction in kidneys as well as MTP-131's potential roles in renoprotection through preserved CL remodeling and mitochondrial O 2 . levels in diabetes are summarized in Fig. 8.

MTP-131 protects against diabetic kidney disease
In conclusion, we found, for the first time, that renal O 2 . production is reduced in the kidney and heart tissues of db/db mice and that MTP-131 (Bendavia) can partly preserve renal O 2 . levels in the kidneys, lower elevated urinary albumin, and, importantly, delay expansion of the mesangial matrix, arguably the central lesion of DKD. MTP-131's effect to normalize lysoCL via restoration of LCLAT1 expression may contribute to renoprotective roles of MTP-131 in progression of DKD in db/db mice. These findings provide a better understanding of the pathogenesis of DKD in type 2 diabetes and support the concept that mitochondrion-protective agents may be developed into effective treatments for chronic kidney disease and DKD.

Animal study
Male db/db mice (BKS.Cg-Dock7 m ϩ/ϩ Lepr db /J strain) were purchased from The Jackson Laboratory (Bar Harbor, ME), and the corresponding heterozygote lean db/m mice were used as controls. Mice were given standard rodent chow and water ad libitum. MTP-131 (Bendavia) was provided by Stealth BioTherapeutics Inc. (Newton, MA). Six-week-old male nondiabetic db/m mice and diabetic db/db mice were randomly assigned to four groups (n ϭ 15/group): db/m mice treated with vehicle (db/m), db/m mice treated with MTP-131 (Bendavia) (db/mϩB), db/db mice treated with vehicle (db/db), and db/db mice treated with MTP-131 (db/dbϩB). Vehicle (PBS) or MTP-131 (3 mg/kg of body weight) was injected subcutaneously daily for 4 weeks and then infused subcutaneously via Alzet osmotic minipumps (Durect Corp., Cupertino, CA; 3 mg/kg/day) over an 8-week period. Osmotic minipumps were implanted subcutaneously in the back of all mice. As the life expectancy of the osmotic pump was 4 weeks, all pumps were replaced with new filled pumps when mice reached the age of 14 weeks. HbA1c was measured with a DCA Vantage Analyzer (Siemens, Malvern, PA). Urine albumin and creatinine were measured with the Mouse Albumin ELISA Quantitation Set (Bethyl Laboratories, Inc., Montgomery, TX) and the Creatinine Companion Kit (Exocell, Philadelphia, PA) according to the manufacturers' protocols. Urinary hydrogen peroxide levels were measured using a 2900D Biochemistry Analyzer (YSI Inc., Yellow Springs, OH). At 18 weeks of age, all mice were euthanized, and the organs were harvested as described previously (51). All animal studies were approved by the Institutional Animal Care and Use Committee of the University of California, San Diego.

Histology
Periodic acid-Schiff (PAS)-stained kidney sections were analyzed as described previously with slight modifications (51). To evaluate the glomerular size and mesangial matrix area, 15 randomly selected glomeruli per mouse were analyzed using i-solution software (Advanced Imaging Concepts, Princeton, NJ). To quantify perigonadal adipocyte size, H&E-stained sections were analyzed using ImageJ (National Institutes of Health; RRID: SCR_003070). Quantitative analysis was performed in a blinded manner.

Superoxide measurement
Superoxide generation was examined as described previously (6). DHE was purchased from Sigma-Aldrich (St. Louis, MO). Briefly, mice were administered two intraperitoneal doses of DHE (each 27 mg/kg) given 30 min apart. Sixteen hours after DHE injection, mice were anesthetized and perfused intracardially with cold PBS followed by 4% paraformaldehyde, and the kidneys were harvested. For confocal imaging, the right kidney was sectioned using a vibrating microtome (Leica Biosystems Inc., Buffalo Grove, IL), mounted on glass slides using Vectashield Antifade Mounting Medium (Vector Laboratories, Burlingame, CA), and imaged by A1R ϩ confocal microscopy (excitation, 561 nm; emission, 595 nm; Nikon Instruments Inc., Melville, NY). For multiphoton imaging, the kidney was cut into 3-mm thickness and placed onto a glass slide (Fisher Scientific, Pittsburgh, PA), washed gently with PBS overnight, and imaged using an A1R MP ϩ multiphoton confocal microscope (excitation, 800 nm; emission, 629 nm; Nikon). 3D images were reconstructed using NIS-Elements software (Nikon). DHE signal intensity was quantified by ImageJ. Briefly, average pixel intensity was quantified in 15 randomly selected cortical tubular lesions or glomerular lesions from each sample (n ϭ 3/group). The results were expressed as relative intensity compared with the db/m group. Quantitative analysis was performed in a blinded manner.

Cardiolipin and monolysocardiolipin quantification
For cardiolipin and monolysocardiolipin (lysoCL) analysis, kidney cortex tissues were snap-frozen in liquid nitrogen immediately after harvest. Lipids were extracted from frozen kidneys, and cardiolipins were analyzed by shotgun lipidomics as described previously (52).

RT-qPCR
Gene expression was assessed by RT-qPCR. Briefly, total RNA was isolated from kidney tissue using RNeasy Plus Universal Mini Kits (Qiagen, Germantown, MD) in accordance with the manufacturer's instructions. Total RNA (1.5 g) was reverse-transcribed to complementary DNA using the SuperScript TM VILO TM Kit (Thermo Fisher Scientific, Grand Island, NY). Real-time qPCR was performed with the gene-specific TaqMan qPCR primers (Thermo Fisher Scientific) according to the manufacturer's instructions in a realtime PCR machine (QuantStudio 3, Applied Biosystems by Thermo Fisher Scientific).

Immunoblotting
Kidney cortex lysates were prepared using radioimmunoprecipitation assay buffer (Cell Signaling Technology, Danvers, MA) according to the manufacturer's instructions. Total protein concentration was determined using the BCA Protein Assay Kit (Life Technologies). 30 g of protein per sample was loaded onto a 4%-12% BisTris gel and transferred to a nitrocellulose membrane. The blot was blocked with 5% nonfat dry milk and incubated with anti-LCLAT1 antibody (1:500, Aviva Systems Biology, San Diego, CA) at 4°C overnight. Following immunoblotting, ECL rabbit IgG and HRP-linked whole anti-MTP-131 protects against diabetic kidney disease body (GE Healthcare Life Sciences, Chicago, IL) were applied, and the signals were developed with SuperSignal TM West Pico Chemiluminescent Substrate (Life Technologies).

Statistics
Data are expressed as mean Ϯ S.E., with n values as indicated in the text or figure legends. Statistical significance was determined by Student's t test or one-way analysis of variance, with p Ͻ 0.05 considered significant.

Data availability
All data described are contained within the manuscript and Figs S1 and S2. Raw data used for generation of graphs are available upon request.