Cardioprotection by the mitochondrial unfolded protein response requires ATF5

Abstract

The mitochondrial unfolded protein response (UPR^mt) is a cytoprotective signaling pathway triggered by mitochondrial dysfunction. UPR^mt activation upregulates chaperones, proteases, antioxidants, and glycolysis at the gene level to restore proteostasis and cell energetics. Activating transcription factor 5 (ATF5) is a proposed mediator of the mammalian UPR^mt. Herein, we hypothesized pharmacological UPR^mt activation may protect against cardiac ischemia-reperfusion (I/R) injury in an ATF5-dependent manner. Accordingly, in vivo administration of the UPR^mt inducers oligomycin or doxycycline 6 h before ex vivo I/R injury (perfused heart) was cardioprotective in wild-type but not global Atf5^−/− mice. Acute ex vivo UPR^mt activation was not cardioprotective, and loss of ATF5 did not impact baseline I/R injury without UPR^mt induction. In vivo UPR^mt induction significantly upregulated many known UPR^mt-linked genes (cardiac quantitative PCR and Western blot analysis), and RNA-Seq revealed an UPR^mt-induced ATF5-dependent gene set, which may contribute to cardioprotection. This is the first in vivo proof of a role for ATF5 in the mammalian UPR^mt and the first demonstration that UPR^mt is a cardioprotective drug target.

NEW & NOTEWORTHY Cardioprotection can be induced by drugs that activate the mitochondrial unfolded protein response (UPR^mt). UPR^mt protection is dependent on activating transcription factor 5 (ATF5). This is the first in vivo evidence for a role of ATF5 in the mammalian UPR^mt.

INTRODUCTION

Cardiac ischemia-reperfusion (I/R) injury is an underlying pathological mechanism of myocardial infarction, a major cause of morbidity and mortality worldwide. Mitochondria are key mediators of cardiac I/R injury and also a potential target for cardioprotective interventions (16, 24). An important concept in cytoprotection is hormesis, wherein small sublethal insults trigger signaling events that afford protection against subsequent larger injuries. Mitohormesis refers to such events at the mitochondrial level (36). An example is ischemic preconditioning (IPC), wherein small ischemic insults trigger mitochondrial reactive oxygen species (ROS) generation, signaling protection against pathological ROS during subsequent I/R (16).

The mitochondrial unfolded protein response (UPR^mt) is a mitohormetic response to proteotoxic stress in the organelle (39). Similar to the endoplasmic reticulum UPR (UPR^ER), UPR^mt signaling upregulates chaperones and proteases to restore proteostasis (20). Furthermore, the UPR^mt upregulates glycolysis (23) and downregulates respiratory chain subunit expression (22) to balance cell energetics and unburden the mitochondrial translation and folding machinery.

Significant insight to the UPR^mt has emerged from work in Caenorhabditis elegans, including the discovery of its central transcription factor ATFS-1 (activating transcription factor associated with stress 1) (14, 26). ATFS-1 contains both nuclear and mitochondrial target sequences, and under normal conditions is imported to mitochondria and degraded (14, 23). Mitochondrial proteotoxicity blunts ATFS-1 import, resulting in its nuclear translocation and target gene activation (22, 23).

Comparatively little is known about the mammalian UPR^mt, especially in heart. Activating transcription factor 5 (ATF5) is proposed as a mammalian ATFS-1 ortholog (14) and indeed can rescue UPR^mt signaling in C. elegans lacking ATFS-1 (10). However, although ATF5 has known roles in stress signaling (40), there is no in vivo evidence for its requirement in the mammalian UPR^mt.

Cardiac I/R injury causes bioenergetic crisis, protein misfolding, and oxidative stress (5), whereas the UPR^mt upregulates glycolysis, chaperones, and antioxidants (20). Thus, we hypothesized induction of an ATF5-dependent UPR^mt may be cardioprotective against I/R injury in mammals. This hypothesis was tested using the known UPR^mt inducers oligomycin and doxycycline (10, 17) in wild-type (WT) and global Atf5^−/− mice (6).

METHODS

Reagents, animals, drug treatments.

All reagents were from Sigma (St. Louis, MO) unless otherwise stated. All animal experiments were approved by the University of Rochester Committee on Animal Resources (Protocol No. UCAR-2014-036) in compliance with the National Institutes of Health’s Guide for the Use and Care of Animals. Atf5^-/+ mice were provided by Stavros Lomvardas on a C57BL/6(J/N) background (6) and backcrossed to WT C57BL/6J (JAX 000664) >3 generations. Mice were housed on a 12-h:12-h light-dark cycle with ad libitum food and water. Experimental mice aged 12–20 wk were of both sexes, and WT controls were age and sex matched, since high neonatal Atf5^−/− mortality made using littermates infeasible. PCR genotyping (tail clip) at 21 days employed three primers, with expected amplicons at 1,006 base pairs for WT and 439 base pairs for knockout (KO): WT: 5′-TCTGATTGGATGACTGAGCGG-3′, KO: 5′-GCAGCCTCTGTTCCACATACACTTC-3′, common: 5′-TCACTTGTGTTCCAAGTCCCC-3′.

To induce UPR^mt in vivo, oligomycin (500 µg/kg) or doxycycline (70 mg/kg) were injected intraperitoneally (12.5 µL/g body mass) in sterile saline with 0.1% DMSO. Controls received vehicle. Mice were returned to cages for 6 h before I/R injury. These drug doses elicited no mortality or other overt phenotypes.

I/R injury.

I/R injury was assessed using an ex vivo retrograde-perfused heart model (38). Anesthetized (tribromoethanol, 200 mg/kg ip) and heparinized (0.2 mL ip) mice were placed supine on a 37°C heat pad, the aorta cannulated (blunt 27-gauge needle), and the heart immediately transferred to the perfusion rig. Perfusion was at 4 mL/min, 37°C, with 0.22 µm filtered, 95% O₂-5% CO₂ gassed Krebs-Henseleit buffer comprising (in mM) 118 NaCl, 4.6 KCl, 1.2 MgSO₄, 25 NaHCO₃, 1.2 K₂HPO₄, 5 glucose, 0.1 palmitate-BSA, 0.2 pyruvate, and1.2 lactate. Left ventricular pressure was digitally recorded (DATAQ, Akron, OH) from a transducer-linked balloon. Pressure traces were analyzed using a custom MATLAB script (MathWorks, Natick, MA) to derive developed pressure, heart rate, and rate × pressure product (RPP). Following 25-min equilibration, hearts were subjected to 25 min global no-flow ischemia and then 60 min reperfusion. Post-I/R, hearts were transverse sliced (2 mm thick), stained with 1% tetrazolium chloride, formalin fixed, and imaged. Healthy (red) tissue versus infarcted (white) tissue was quantified by planimetry in MATLAB. In acute studies, doxycycline (20 µM) was dissolved in perfusion buffer with 0.25% DMSO, 0.22 µm filtered, and perfused throughout the procedure.

RNASeq, quantitative PCR, and Western blot analysis.

Following anesthesia (tribromoethanol, 200 mg/kg ip), hearts were excised, flushed with ice-cold saline, homogenized in TRIzol (Invitrogen, Carlsbad, CA), and snap frozen (liquid N₂). RNA was processed and purified using RNeasy Mini-kits (Qiagen, Hilden, Germany) with on-column DNase treatment. For RNA-Seq, library preparation and analysis were performed by the University of Rochester Functional Genomics Core (oligomycin experiments) or the Memorial Sloan Kettering Integrated Genomics Operation (doxycycline experiments). Total RNA was determined spectrophotometrically (NanoDrop, Wilmington, DE), and RNA quality was assessed with the Agilent Bioanalyzer (Agilent, Santa Clara, CA). Library construction and sequencing were performed from 200 ng RNA with the TruSeq-stranded mRNA preparation kit and a HiSeq 2500 v4 platform with cBot, according to manufacturer’s protocols (Illumina, San Diego, CA). Raw reads were demultiplexed using bcl2fastq.pl v1.8.4 conversion software (Illumina). Quality filtering and adapter removal were performed using Trimmomatic 0.32 (1), and processed reads were mapped to the mouse genome with STAR 2.4.2a (7).

For quantitative PCR (qPCR), cDNA was generated using an iScript cDNA synthesis kit (Bio-Rad, Hercules, CA) and used at 100 µg/well as a PCR template for target detection with iTaq universal SYBR green supermix (Bio-Rad) on a CFX connect real-time PCR system (Bio-Rad). Quantification cycle (Cq) of each RT-qPCR target was normalized to reference genes Actb or Hprt (ΔCq). Data are presented in treated versus untreated (vehicle) conditions, as relative ΔCq (i.e., ΔΔCq).

For Western blot analysis, samples of 50 µg of protein were run on 12% SDS-PAGE gels, wet-transferred to nitrocellulose, and visualized with Ponceau S. Blocking was performed with 10% milk and antibodies were incubated in 5% milk in TBS-T. Images were acquired using a KwikQuant Imager (Kindle Biosciences, Greenwich, CT) and analyzed in ImageJ. Since the enhanced chemiluinescence reagent emits blue light, densitometry and presented blot images were from the blue channel only, such that loss of background detail may have occurred due to removal of the red/green channels.

Statistics.

Comparisons between two groups employed two-tailed unpaired Student’s t-tests. Comparisons versus zero (qPCR) employed two-tailed one-sample Student’s t-tests. Univariate comparisons between greater than two groups used one-way ANOVA. Multivariate comparisons used two-way ANOVA. Differences were considered significant at P < 0.05. Data are reported as means ± SE.

RNA-Seq gene read counts were processed using R 3.4.4; Bioconductor 3.7; limma 3.34.9. One vehicle-treated Atf5^−/− sample was removed due to probable failure of the KO. Genes with <50 reads in all samples were filtered yielding 14,061 genes for analysis. The voom method (19) was used with limma empirical Bayes analysis methodology (28) to assess differential expression. Gene set testing was performed using ROAST (34) to account for correlation between genes. Full data are deposited in the sequence read archive (accession no. SRP150238).

RESULTS

Effects of ATF5 ablation on cardiac function and I/R injury response.

The complete data set is at DOI:10.6084/m9.figshare.7323929. Fig. 1A shows WT and Atf5^−/− genotyping. Consistent with previous reports (6, 33), a confounder in these studies was high neonatal mortality in Atf5^−/− mice (Fig. 1B) due to an olfactory neuron defect resulting in failure to suckle at birth. In ex vivo-perfused hearts, neither baseline function (Fig. 1C) nor post-I/R functional recovery or infarct size (Figs. 1, D–G) was different between genotypes or sexes. Further stratification by sex or genotype (not shown) did not reveal sex-specific genotype-dependent effects. Thus, for UPR^mt investigations, both sexes were used, and controls were age and sex matched and not always littermates. After weaning, adult Atf5^−/− mice were healthy but slightly smaller than WT and exhibited no phenotypes up to 2 yr of age.

Fig. 1.

Loss of activating transcription factor 5 (ATF5), baseline cardiac function, and ischemic-reperfusion injury. A: example PCR genotyping of Atf5, yielding amplicons at 1,006 base pair (bp) for wild type (WT), 439 bp for knockout, and both for heterozygotes. B: neonatal mortality (<21 days) for 93 litters (n = 769 mice) by genotype. Mortality is marked by white portion of bars, with fractional values above. C: baseline cardiac function [rate × pressure product (RPP)] of perfused hearts, by genotype and sex. bpm, beats/min; F, female; M, male. D: post-reperfusion functional recovery (RPP normalized to baseline) by genotype. Means ± SE, both sexes, n = 12 per genotype (WT: black, Atf5^+/−: green, Atf5^−/−: blue). E: infarct sizes of the hearts in D. Representative slices (E, top) and digitally parsed pseudocolor images (E, middle) show red healthy tissue and white infarct. Means ± SE shown to right of individual data. F: post-reperfusion functional recovery (data from D stratified by sex). Means ± SE, all genotypes sexes, n = 21 female (black), n = 15 male (gray). G: infarct data for hearts in F. No significant differences between any groups by 2-way ANOVA.

ATF5-dependent cardioprotection by oligomycin.

In vivo delivery of the UPR^mt inducer oligomycin 6 h before ex vivo I/R injury elicited significant improvement in ex vivo post-I/R functional recovery (oligomycin, 52 ± 4 vs. vehicle, 24 ± 4%) and reduction in infarct size (oligomycin, 39 ± 3 vs. vehicle, 64 ± 9%) in WT mice (Fig. 2A/B). However, such protection was absent in Atf5^−/− mice (functional recovery: oligomycin, 30 ± 4 vs. vehicle, 19 ± 5%; infarct; and oligomycin, 58 ± 6 vs. vehicle, 57 ± 8%) (Fig. 2, C and D). Oligomycin did not affect baseline cardiac function (Fig. 2E), consistent with previous reports of minimal effects on cardiac energetics in vivo at the same dose and in the same time frame (18).

Fig. 2.

Mitochondrial unfolded protein response induction by oligomycin is cardioprotective and requires activating transcription factor 5 (ATF5). A: post-reperfusion functional recovery [rate × pressure product (RPP), normalized to baseline] of hearts from wild-type (WT) mice given oligomycin (orange) or vehicle (black); n = 5. *P < 0.05 vs. matching time point. B: infarct sizes of hearts in A. Representative and pseudocolor images as per Fig. 1E. *P < 0.05 vs. vehicle. C: post-reperfusion functional recovery of hearts from Atf5^−/− mice given oligomycin (pink) or vehicle (blue); n = 5 per group. *P < 0.05 vs. matching time point. D: infarct sizes of hearts in C. E: baseline cardiac functional data for hearts in this cohort. V, vehicle; O, oligomycin; bpm, beats/min. All data are means ± SE.

ATF5-dependent cardioprotection by doxycycline.

In vivo delivery of the UPR^mt inducer doxycycline (17) 6 h before ex vivo I/R injury elicited significant improvement in ex vivo post-I/R functional recovery (doxycycline, 60 ± 8 vs. vehicle, 23 ± 4%) and reduction in infarct size (doxycycline 37 ± 3 vs. vehicle 69 ± 4%) (Fig. 3A and B). Doxycycline-induced cardioprotection was absent in Atf5^−/− mice (functional recovery: doxycycline, 42 ± 7 vs. vehicle, 34 ± 7%; infarct: doxycycline, 50 ± 2 vs. vehicle, 59 ± 4%) (Fig. 3, C and D). Doxycycline did not affect baseline cardiac function (Fig. 3E).

Fig. 3.

Mitochondrial unfolded protein response induction by doxycycline is cardioprotective and requires activating transcription factor 5 (ATF5). A: post-reperfusion functional recovery [rate × pressure product (RPP), normalized to baseline] of hearts from WT mice given doxycycline (red) or vehicle (black); n = 5 per group. *P < 0.05 vs. matching time point. B: infarct sizes of hearts in A. Representative and pseudocolor images as per Fig. 1E. *P < 0.05 vs. vehicle. C: post-reperfusion functional recovery of hearts from Atf5^−/− mice given doxycycline (purple) or vehicle (blue); n = 5 per group. D: infarct sizes of hearts in C. E: baseline cardiac functional data for hearts in this cohort. V, vehicle; D, doxycycline; bpm, beats/min. F: post-reperfusion functional recovery of WT hearts given acute doxycycline (red) or vehicle (black) ex vivo; n = 3. G: infarct sizes of the hearts in F. All data are means ± SE.

The ability of doxycycline to induce acute cardioprotection was also tested by direct delivery to perfused hearts, but in contrast to a previous report (30), no protection was observed (functional recovery: doxycycline, 40 ± 19 vs. vehicle, 34 ±9%; infarct: doxycycline, 66 ± 6 vs. vehicle, 49 ± 5%) (Fig. 3, F and G).

Genetic analysis of UPR^mt activation.

To confirm UPR^mt induction, expression of known UPR^mt target genes was analyzed by RT-qPCR on hearts from WT mice 6 h after oligomycin delivery. As expected, the UPR^mt-associated genes Hspd1, Clpp, and Lonp (13, 39) were all significantly induced by oligomycin, whereas Hspa5 (the chaperone BiP in the endoplasmic reticulum UPR) was not induced (Fig. 4A). To confirm induction of the UPR^mt at the protein level, Western blotting was performed, revealing that hearts from oligomycin-treated mice showed elevation in the level of HSP60, with no change in BiP (Fig. 4C).

Fig. 4.

Mitochondrial unfolded protein response (UPR^mt) gene expression. A: relative wild-type (WT) cardiac expression of selected genes 6 h after oligomycin treatment by RT-qPCR (−ΔΔCq, normalized to Hprt1 reference gene and compared with vehicle treatment). Hspa5 (BiP) is an endoplasmic reticulum UPR chaperone, Hspd1 (HSP60) is a UPR^mt chaperone, and Clpp and Lonp are UPR^mt proteases. Horizontal lines represent means. *P < 0.05 vs. vehicle, n = 5. B: primers used for quantitative PCR. D: Western blot normalized to protein loading (Ponceau-stained membrane) was used to quantify relative levels of indicated proteins in hearts 6 h after vehicle or oligomycin. Values for each protein are expressed relative to abundance in vehicle-treated hearts. Representative blots are shown, with one biological replicate (animal) per lane. Note that in lane 2 of the oligomycin group, the sample was lost during preparation so the lane was left blank. *P < 0.05 compared with vehicle-treatment, n = 5/4 per group. Data are means ± SE. C: RNA-Seq heat map showing relative expression of 69 genes identified as upregulated by oligomycin/doxycycline treatment in hearts from WT mice with a blunted response in hearts from Atf5^−/−; 48 of the 69 genes (left) are color-coded for annotation by functional category. V, vehicle; D, doxycycline.

To explore cardioprotective signaling by the UPR^mt→ATF5 axis in more detail, RNA-Seq was performed on cardiac mRNA from WT or Atf5^−/− mice treated with vehicle, oligomycin, or doxycycline. Data from both UPR^mt inducers were grouped for analysis, to identify common genes induced in WT (adjusted P < 0.2), with lower or no induction in Atf5^−/− (adjusted P < 0.4). Such analysis yielded 69 genes responding to UPR^mt induction in an ATF5-dependent manner (Fig. 4D). Known UPR^mt-induced genes (27) were poorly represented in this gene set, and unbiased pathway analysis did not yield insight to potential gene programs that may underlie cardioprotection. Nevertheless, guided annotation revealed 48 of 69 genes fell into seven broad categories with potential relevance to cardioprotection (Fig. 4D).

DISCUSSION

The key finding of this study is that pharmacological UPR^mt induction is cardioprotective against I/R in an ATF5-dependent manner. This is the first in vivo demonstration that ATF5 is necessary for the mammalian UPR^mt, and the first study to demonstrate a role for UPR^mt in cardioprotection by doxycycline (4).

Regarding the mechanism(s) of UPR^mt activation by oligomycin, its inhibition of mitochondrial ATP synthesis may impair ATP-dependent mitochondrial protein synthesis, and import (9, 25). Notably, protein turnover sits low in the hierarchy of ATP consuming processes in cells, such that mild energetic dysfunction may impair this process without affecting more important processes such as ion-homeostasis (2). Doxycycline disrupts mitochondrial ribosomes (29), such that overall both oligomycin and doxycycline likely induce a UPR^mt by generating a stoichiometric imbalance between nuclear and mitochondrial DNA encoded respiratory chain subunits (17).

The significant cardioprotection observed in hearts from UPR^mt-induced WT mice indicates mitochondrial proteotoxic stress is a hormetic protection mechanism, akin to other cardioprotective gene programs such as delayed preconditioning (35). UPR^mt-induced cardioprotection was absent in Atf5^−/− mice, yet these mice exhibited no detriment in baseline cardiac function or I/R injury, indicating loss of UPR^mt-induced protection is not simply due to greater stress sensitivity in Atf5^−/−.

The canonical UPR^mt gene program includes chaperones, proteases, antioxidants, and glycolysis (20, 22, 23), all of which are predicted to be beneficial in I/R injury. In particular, the mitochondrial protease LONP was recently shown to play a cardioprotective role in I/R injury (31). Although qPCR showed upregulation of several UPR^mt-associated genes (including LonP) in hearts from oligomycin-treated mice, and this was confirmed at the protein level for HSP60, RNA-Seq did not identify canonical UPR^mt target genes induced by oligomycin or doxycycline in an ATF5-dependent manner. This discrepancy is likely due to the low relative sensitivity of RNA-Seq versus qPCR.

RNA-Seq did identify 69 UPR^mt-induced ATF5-dependent genes, which fell into broad categories including metabolism, epigenetic regulation, and cell signaling (Fig. 4D). Among metabolic genes of interest, Pfkfb1 is a glycolytic activator, which could be beneficial under hypoxic conditions (11). Several modes of cardioprotection are known to require glycolytic enhancement, including ischemic preconditioning (21). The enzyme Gpt2 (alanine aminotransferase) facilitates glutamate anaplerosis to the Krebs’ cycle, which is important for fueling substrate-level phosphorylation by succinyl-CoA synthetase during ischemia (38). Stk11ip is an activator of AMP-dependent protein kinase (37), and Agap2 activates Akt (12), both established cardioprotective signals.

Although our data suggest doxycycline-induced cardioprotection requires ATF5, this tetracycline is also reported to confer cardioprotection by inhibiting matrix metalloproteinases (MMPs) (4, 32). There is no apparent connection between ATF5 and MMPs, suggesting UPR^mt→ATF5 signaling and MMP inhibition are parallel independent mechanisms of doxycycline cardioprotection. Recently, cardioprotection was reported upon acute tetracycline delivery to perfused hearts 30 min. before ischemia (30). We could not reproduce this result, possibly due to low doxycycline solubility in perfusion buffer, such that the necessary microfiltration (15) may limit availability. In addition, 30 min pre-treatment may be too short to induce a gene program such as the UPR^mt (3), suggesting acute tetracycline cardioprotection may be attributable to MMP-inhibition (4).

Although our data suggest ATF5 is a mammalian ortholog of C. elegans ATFS-1 in the UPR^mt, it remains unclear whether other features of the nematode UPR^mt are retained in mammals. Several nematode UPR^mt signaling proteins have unproven mammalian orthologs, including DVE-1 (SATB2) and UBL-5 (UBL5) (13). Furthermore, the nematode UPR^mt can signal cell nonautonomously, whereby a UPR^mt in one cell can trigger mitochondrial protection in another (8). Since our studies used global Atf5^−/− mice, we cannot exclude the possibility that pharmacological UPR^mt induction in vivo may have occurred in another tissue, signaling remote cardioprotection. Further studies using tissue-specific Atf5^−/− mice may therefore prove insightful.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grants R01-HL-127891 and R01-HL-071158.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.M.H., K.N., and P.S.B. conceived and designed research; Y.T.W. and K.-T.H. performed experiments; Y.T.W., Y.L., M.N.M., K.N., and P.S.B. analyzed data; Y.T.W., Y.L., M.N.M., C.M.H., K.N., and P.S.B. interpreted results of experiments; Y.L. prepared figures; M.N.M., P.S.B. drafted manuscript; C.M.H., K.N., and P.S.B. edited and revised manuscript; and Y.T.W., Y.L., M.N.M., K.-T.H., C.M.H., K.N., and P.S.B. approved final version of manuscript.