Generation of the HEK293-CISD2 reporter cell line

A CISD2 bacterial artificial chromosome (BAC) reporter clone was constructed from a 102-kb human BAC clone (CTD-2303J4, Invitrogen, San Diego, CA, USA, #96012) that contained the intact gene of human CISD2 in its native chromosomal setting, together with its flanking upstream and downstream regions. This human BAC clone carries the entire 23.8-kb genomic sequence of the CISD2 coding region, a 31.3-kb upstream region and a 46.9-kb downstream region (Additional file 1: Fig. S1A). To generate the CISD2 BAC reporter clone, IRES-Luc-pA, namely an internal ribosome entry site (IRES), luciferase (Luc) and a polyA signal (pA), was inserted into exon 2 of the CISD2 gene (Additional file 1: Fig. S1B) by in vivo recombineering-based method in Escherichia coli [34]. To establish the HEK293-CISD2 reporter cell line, the linearized CISD2 BAC reporter construct and pCI-neo plasmid (Promega, Mannheim, Germany, #E1841) were co-electroporated (250 V with capacitance 500 μF) into the HEK293 cells and selected with 1 μg/mL puromycin (Invitrogen, #A11138-03) for 12–14 days in growth medium (DMEM; Gibco, Carlsbad, CA, USA, #11965) supplemented with 10% fetal bovine serum, 1% glutamine/penicillin/streptomycin, 1% non-essential amino acid, and 1 mM sodium pyruvate. The correct clones for the HEK293-CISD2 reporter system were confirmed by PCR and luciferase reporter assay.

Luciferase reporter assay

The HEK293-CISD2 BAC reporter cells were seeded on a 96-well plate at 2 × 104 cells/well. After 24 h of seeding, the HEK293-CISD2 BAC reporter cells were treated with different doses of hesperetin (and various other compounds individually) for 24 h. After treatment with each compound, luciferase activity was assayed using a ONE-GloTM Luciferase Assay System kit (Promega, #E6120) following the manufacturer’s instructions. The intensity of luminescence was monitored using an Infinite 200 Microplate Reader (Tecan Group Ltd., Männedorf, Switzerland).

The noble herb compound library

The herb compound library, which contains 780 samples, was established to systematically screen for CISD2 activators from the 60 noble herbs (Additional file 1: Table S1) that were described in a traditional Chinese medicine book, The Divine Husbandman’s Herbal Foundation Canon (神農本草經, Shén Nóng Běn Cǎo Jīng, written between 200 and 250CE) [35]. Briefly, 50 g of each herb was crushed and extracted with 200 mL of 60% ethanol shaking at 25 °C, this was repeated three times. The extracted solution (600 mL) was partitioned using CH2Cl2 (600 mL), and then dried on a rotary evaporator. The CH2Cl2 layer and the 60% ethanol layer were further separated into five fractions using a silica gel-based high performance liquid chromatography (HPLC) column or a C18 HPLC column, respectively. The silica gel column was eluted using 100 mL of 100% CH2Cl2, CH2Cl2/methanol (95/5), CH2Cl2/methanol (9/1), CH2Cl2/methanol (8/2) and 100% methanol in series to give the C1–C5 fractions. The C18 column was eluted using 200 mL of ddH2O, 30% methanol (aq), 60% methanol (aq), 90% methanol (aq), and 100% methanol, in series, to give the M1–M5 fractions. Thus, for each herb there are 13 fractions including the crude extract, the CH2Cl2 layer, the 60% ethanol layer, C1–C5 fractions, and M1–M5 fractions. Each fraction was dried using a rotary evaporator and stored at − 20 °C.

Identification of Cisd2 activators

Initially, sophoricoside and genistein from Sophora japonica (Additional file 1: Table S1) were identified as Cisd2 activators. Five fractions (CH2Cl2 layer, C2, C3, C5 and M3) from Sophora japonica gave Z-scores > 2; and pure sophoricoside and genistein were obtained from the precipitate of the CH2Cl2 layer of Sophora japonica via a bioassay-guided purification. The precipitate of the CH2Cl2 layer was dissolved in methanol and purified by semi-preparative reverse phase HPLC (Cosmosil C18 ARII, 10 × 250 mm) under the following conditions: isocratic running 24% acetonitrile/H2O for 20 min, then to 60% acetonitrile/H2O in 5 min, and hold for 5 min; flow rate at 4 mL/min; monitoring at 200 to 400 nm range of a diode array detector. The structures of sophoricoside and genistein were confirmed by NMR [36, 37] and MS/MS data. Subsequently, based on the chemical structures of sophricoside and genistein, seven structural analogs of these flavonoids, namely baicalin, formononetin, Kaemferol-3-O-rhamnoside, medicarpin, puerarin, rutin, and hesperetin were selected and examined for their ability to enhance CISD2 expression using the HEK293-CISD2 reporter cell assay (Additional file 1: Fig. S1C, D). Finally, hesperetin (a single compound with > 98% purity) was identified as a promising Cisd2 activator that is able to enhance Cisd2 expression both in vitro and in vivo (Additional file 1: Fig. S1E–G).

Analysis of hesperetin and its conjugated metabolites

The levels of hesperetin, hesperetin-7-O-beta-d-glucuronide (H7G) and hesperetin-7-O-sulfate (H7S) in the serum and tissues of the mice were quantified by LC–MS/MS. The mice were fed the dietary hesperetin supplemented food ad libitum for 4 months until the day of sacrifice. To synchronize the food intake, the mice were fasted for 6 h (2 p.m. to 8 p.m.) and then fed the hesperetin supplemented food for 2 h (8 p.m. to 10 p.m.). Sample preparation for the LC–MS/MS assay consisted of 50 µL of mouse serum or tissue homogenates (liver, cardiac muscle or skeletal muscles) being mixed with 50 µL of 250 ng/mL of hesperetin-d3 (TORONTO Research Chemicals INC., Toronto, Canada, #H289502). Hesperetin-d3 is an internal standard of labelled hesperetin in which the three hydrogens were replaced by deuterium. The mixture was vortexed and then centrifuged at 15,000×g for 20 min in a Beckman Coulter Microfuge 22R Centrifuge at room temperature. The supernatant was transferred to a clean tube and finally 15 µL of the supernatant was injected onto the LC–MS/MS system.

LC–MS/MS analysis

The chromatographic system consisted of an Agilent 1200 series LC system and an Agilent ZORBAX Eclipse XDB-C8 column (5 µm, 3.0 × 150 mm) interfaced with a MDS Sciex API4000 tandem mass spectrometer. The MS/MS ion transitions monitored were m/z 300/9/163.9, 477.0/301.0, 380.9/301.0 and 303.9/163.8 for hesperetin, H7G, H7S and hesperetin-d3, respectively. A gradient HPLC method was employed for separation. The mobile phase A consisted of 10 mM ammonium acetate aqueous solution containing 0.1% formic acid, while the mobile phase B consisted of acetonitrile. The gradient profile was as follows (min/%B): 0.0–0.5/10, 0.5–1.2/60, 1.2–3.4/80 and 3.5–5.0/10. The flow rate was set at 1.5 mL/min into the mass spectrometer with the remainder being split off to waste. The retention times of hesperetin, H7G, H7S, and hesperetin-d3 were 2.46, 2.07, 2.16, and 2.44 min, respectively.

Mice and hesperetin treatment

The CISD2 reporter transgenic (TG) mice were generated as previously described [38]. Briefly, the linearized CISD2 BAC reporter construct, which carries luciferase as the reporter and driven by the human CISD2 promoter, was microinjected into the pronuclei of fertilized eggs obtained from C57BL/6 mice. The Cisd2 mcKO mice, which carry a Cisd2 KO background specifically in the skeletal and cardiac muscles, were generated as previously described [22]. Briefly, mice carrying the Cisd2 floxed allele (Cisd2 f/f) were bred with transgenic mice carrying the muscle creatine kinase-Cre (MCK-Cre; JAX006475). After two generations of breeding, Cisd2 mcKO (Cisd2f/f;MCK-Cre) mice were obtained. All the mice used in this study are males. All mice have a pure or congenic C57BL/6 background and were housed in a specific pathogen-free facility with a 12–12 h light–dark cycle at constant temperature (20–22 °C). For the dietary hesperetin treatment, old wild-type (WT) mice (19.5 mo to 23.5 mo of age) were provided with a diet (AIN-93G Growth Purified Diet, TestDiet, St. Louis, MO, USA; Additional file 1: Table S2) containing the vehicle (Veh) (3.04% propylene glycol [w/w]; Sigma-Aldrich, Munich, Germany, 16033) with or without hesperetin (0.07% [w/w]; Sigma-Aldrich H4125; purity (HPLC area %) > 95%; 100 mg/kg/day) for 3 to 6 months. After these treatments, the mice were sacrificed using carbon dioxide (CO2) inhalation as the method of euthanasia. All animal protocols were approved by the Institutional Animal Care and Use Committee of Chang Gung Memorial Hospital (No. 2017103002 and 2017030901) and National Yang Ming Chiao Tung University (No. 1040104r). The animal protocol was designed to respect the associated guidelines and the 3R principles (Replacement, Reduction and Refinement) according to the “Animal Protection Act” of Taiwan.

In vivo imaging system (IVIS) analysis

For the in vivo luciferase assay, the luciferase activity in CISD2 reporter TG mice was measured before and after dietary hesperetin treatment (100 mg/kg/day provided in food) using an In Vivo Bioluminescence Imaging System (IVIS) (IVIS 50 System, Xenogen Corp., Alameda, CA, USA). The CISD2 reporter TG mice were injected intraperitoneally with the substrate d-luciferin (150 mg/kg in PBS) and then anesthetized using 2.5% isoflurane in IVIS 50 System for image acquisition. The luminescent intensity at the mouse ventral view was analyzed by living image software 3.2 (IVIS 50 Imaging System, Xenogen Corp.). The bioluminescent signal is presented as mean photons/second/centimeter2/steradian (photon/s/cm2/sr).

Serum biochemical and complete blood count (CBC) analyses

Whole blood samples were collected from the facial vein or by cardiac puncture at sacrifice. Serum alanine aminotransferase, aspartate aminotransferase, blood urea nitrogen, creatinine, creatine kinase-MB, total cholesterol, triglyceride, Ca2+, Mg2+, Na+, K+, Cl− levels were monitored by Fuji Dri-Chem 4000i (Fujifilm, Tokyo, Japan). Whole blood samples were collected from the facial vein using an EDTA (final concentration 5 mM) coated tube. The CBC was analyzed using a hematology analyzer (model ProCyte Dx, IDEXX, Columbus, OH, USA).

Whole body composition analysis

Mouse body lean and fat volumes were measured using a micro-CT scanner (SkyScan 1076, Bruker, Kontich, Belgium). The quantitative results for lean, fat, and visceral fat in the whole body of the mice were analyzed using the three-dimensional structure obtained from the micro-CT and software SkyScan 1076 (Bruker).

Whole-body metabolic rate

A TSE Calorimetry Module of the LabMaster System (TSE Systems GmbH, Homburg, Germany) was used to monitor the oxygen consumption rate (VO2), carbon dioxide production rate (VCO2), and energy expenditure (EE) of mice. Individual mice were acclimated for 72 h and then assayed for another 48 h using a 12–12 h light-dark cycle (lights on at 8:00 a.m.) with ad libitum access to food and water. The whole-body metabolic rate of each mouse was measured by indirect calorimetry and then corrected according to lean mass, which was calculated as follows: lean mass = lean volume × 1.06 g/cm3 muscle density [39].

Oral glucose tolerance test and insulin tolerance test

For the oral glucose tolerance test, the mice were orally administrated with glucose water (1.5 mg/g) after a 6 h fasting (9:00 a.m. to 15:00 p.m.). Blood samples were collected at the indicated time points [6]. The blood glucose levels were measured using OneTouch Ultra glucose test strips and a SureStep Brand Meter (LifeScan, Milpitas, CA, USA). Serum insulin levels were determined by a mouse insulin ELISA kit (Mercodia, Uppsala, Sweden, #10-1249-01). For the insulin tolerance test, the mice were examined after a 2 h fasting (9:00 a.m. to 11:00 a.m.) and an intraperitoneal injection of insulin (0.75 U/kg) (Actrapid human regular insulin, Novo Nordisk, Bagsværd, Denmark).

Rotarod trials

Rotarod trials were used to examine the motor coordination, balance and exhaustion resistance of the mice and were conducted using a Rotarod instrument (RT-01, Singa Technology Corporation, Taipei, Taiwan). The mice were placed on a rotarod running at different speeds for the same duration (5 min). Mice were pre-trained three times (5 rpm for 5 min) before the tests. In the test phase, the rotating speed was set at 10, 20 and 30 rpm (the speed up rate was 1 rpm/s). The time of falling was automatically recorded by an infrared sensor at the bottom of the instrument [40].

Transthoracic echocardiography

Cardiac functions were assessed using a VisualSonics VeVo 2100 Imaging System (VisualSonics, Toronto, Ontario, Canada). Male mice were anesthetized with 1% isoflurane in 95% O2. Body temperature was maintained and monitored at 36 °C to 37 °C on a heated pad (TC-1000, CWE Inc., Ardmore, PA, USA). Cardiac function was assessed using a high-frequency 30–50 MHz probe, as described previously [41]. Data analysis was carried out using VisualSonics software. The personnel responsible for data acquisition were blinded to the animal groupings.

Electrocardiography (ECG)

Functional testing of the mice’s hearts using ECG was performed as described previously [24]. The mice were maintained on a 12:12 h dark–light cycle with lights switched on at 6:00 am. All procedures took place during the light phase. Anesthesia was initially induced by placing the mice for 3–5 min in a chamber filled with 3% volume-to-volume isoflurane (Aesica Pharmaceuticals, Hertfordshire, UK). The mice were then positioned on a warm pad (ALA Scientific Instruments, New York, NY, USA) that maintained their temperature during ECG recording. The mice were able to breath freely through a nose cone. Anesthesia was maintained by inhalation of 1.5% isoflurane. Continuous 5-min ECGs were obtained using subcutaneous electrodes attached to the four limbs and recorded via a PowerLab data acquisition system (model ML866, ADInstruments, Colorado Springs, CO, USA) and Animal Bio Amp (model ML136, ADInstruments). The ECG analysis was performed in an unbiased fashion with 1500 beats being analyzed using LabChart 7 Pro version 7.3.1 (ADInstruments). Detection and analysis of QTc interval, QRS intervals, Tpeak-Tend intervals were set to Mouse ECG parameters. The values obtained were compared statistically by the Mann–Whitney U test, and a p < 0.05 was accepted as significant.

Western blotting

Skeletal muscle (femoris and gastrocnemius) and cardiac muscle tissue samples were homogenized using a MagNA Lyser (Roche, Basel, Switzerland) in RIPA buffer (50 mM Tris at pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.5% Sodium deoxycholate, 0.1% SDS with complete protease inhibitor and phosphatase inhibitor cocktails [Roche, #04693124001]) and then denatured in 2% SDS sample buffer (50 mM Tris at pH 6.8, 100 mM Dithiothreitol, 2% SDS and 10% glycerol) for 15 min at 100 °C. Total protein lysate was separated by SDS-polyacrylamide gel electrophoresis (Bio-Rad, Hercules, CA, USA) and then electro-transferred to a polyvinylidene fluoride transfer membrane (PerkinElmer, Waltham, MA, USA, #NEF1002001PK). The membrane was blocked with 5% (w/v) non-fat dried milk in TBST buffer (25 mM Tris at pH 7.5, 137 mM NaCl, 2.7 mM KCl and 0.1% Tween-20 [v/v]) for 1 h at room temperature, and then incubated with a primary antibody for 14–16 h at 4 °C. The membrane was then washed three times with TBST buffer before probing with an appropriate secondary antibody for 1 h at room temperature; washing and then detection by ECL (Thermo Fischer Scientific, Waltham, MA, USA, #34580). The following antibodies were used: Cisd2, Gapdh (Millipore, Burlington, MA, USA, #MAB374), Anti-Rabbit IgG HRP Linked (Sigma-Aldrich, #NA934) and Anti-Mouse IgG HRP Linked (Sigma-Aldrich, #NA931).

Histopathology and transmission electron microscopy (TEM)

Mouse skeletal muscle (femoris and gastrocnemius) and cardiac muscle tissue samples were harvested and then fixed with 10% formalin for 14–16 h at 4 °C. The samples were processed using a tissue processor (STP120, MICROM, Walldorf, Germany) and embedded in paraffin. H&E, Masson’s trichrome and Sirius Red staining of tissue sections (3–4 μm) were carried out by standard protocols [22]. The TEM was performed as described previously [24]. In brief, mouse skeletal muscle (gastrocnemius) and cardiac muscle tissues were fixed in a TEM fix buffer (1.5% glutaraldehyde and 1.5% paraformaldehyde in 0.1 M cacodylate buffer at pH 7.3), post-fixed in 1% OsO4 and 1.5% potassium hexanoferrate and then tissues were washed in cacodylate and 0.2 M sodium maleate buffers (pH 6.0) followed by block-stained with 1% uranyl acetate. Following dehydration, the skeletal muscle (gastrocnemius) tissue and the cardiac muscle tissue were embedded in Epon (EMS, Hatfield, PA, USA, #14120) and sectioned for TEM analysis.

Tissue reactive oxygen species (ROS) and reactive nitrogen species (RNS) levels

ROS and RNS levels were assayed in the skeletal muscle and cardiac muscle tissue lysates by an in vitro ROS/RNS Assay Kit for quantification of ROS and RNS levels following the manufacturer’s instructions (Cell Biolabs, San Diego, CA, USA, #STA-347). The fluorescence intensity of 2ʹ,7ʹ-dichlorodihydrofluorescein (DCF) was monitored using an Infinite 200 Microplate Reader (Tecan Group Ltd.).

Tissue RNA isolation, RNA sequencing, and pathway analysis

Total RNA was isolated from skeletal muscle (gastrocnemius), cardiac muscle, and liver tissue using TRI Reagent (Sigma-Aldrich, #T9424) and phenol/chloroform extraction. The quality of the total RNA was examined using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA); samples with an RNA Integrity Number higher than 8 were subjected to RNA sequencing. The RNA sequencing (RNA-seq) was conducted by the Genome Research Center at National Yang Ming Chiao Tung University. The dataset was generated to a depth of at least 20 million reads for each sample by single-end sequencing. After mapping, the unique gene reads were analyzed as RPKM (reads per kilobase of exon model per million reads) to assess gene expression. A total of 6404 and 6231 genes were retained after filtering to identify expressed genes in the cardiac muscle and skeletal muscle (gastrocnemius) tissues (minimal counts in RPKM > 4 detected in at least 50% of samples), respectively. The p-values of the gene expressions were adjusted using the Benjamini–Hochberg method. Differentially expressed genes (DEGs) were identified using a false discovery rate (FDR) cut-off threshold as indicated in the figure legends. DEGs reversed by hesperetin were analyzed using the following criteria: (1) 26-month WT-Veh vs. 3-month WT, FDR < 0.1; (2) 26-month WT-hesperetin vs. 26-month WT-Veh, p < 0.05, and reversing of 26-month WT-Veh vs. 3-month WT; (3) 26-month WT-hesperetin vs. 3-month WT, p > 0.05. The DEGs from the RPKM was loaded into the EZinfo software package for principal component analysis (PCA, EZinfo 3.0.3 software, Umetrics, Umeå, Sweden). Gene Ontology (GO) functional characterization was performed using the online tools PANTHER (www.pantherdb.org) and Mouse Genome Informatics (MGI) GO term finder (www.informatics.jax.org). The values of the RPKM were transformed into z-scores and these scores were used to generate heatmaps using Multi Experiment Viewer 4.9 software (mev.tm4.org).

Statistical analysis

The data are presented as mean ± SD or mean ± SEM, as described in the figure legends. Comparisons between two groups were carried out using an unpaired two-tailed Student’s t test. Comparisons among groups greater than two were carried out using either one-way or two-way ANOVA with Bonferroni multiple comparison test as indicated in the figure legends. The survival rates of the mice were compared using a log-rank (Mantel–Cox) test; power analysis (SPSS Statistics 26.0, IBM Corp, Armonk, NY, USA) revealed that a sample size of 47 animals (including the no-treatment, Veh and hesperetin groups) has a power of 0.9608. When analyzing statistical differences among groups, p < 0.05 was considered significant using the software Graphpad Prism 6.0 (GraphPad Software, San Diego, CA, USA).