Kobe University Repository : Kernel

タイトルTit le

Hearing vulnerability after noise exposure in a mouse model of react iveoxygen species overproduct ion

著者Author(s)

Morioka, Shigefumi / Sakaguchi, Hirofumi / Yamaguchi, Taro / Ninoyu,Yuzuru / Mohri, Hiroaki / Nakamura, Takashi / Hisa, Yasuo / Ogita,Kiyokazu / Saito, Naoaki / Ueyama, Takehiko

掲載誌・巻号・ページCitat ion Journal of Neurochemistry,146(4):459-473

刊行日Issue date 2018-08

資源タイプResource Type Journal Art icle / 学術雑誌論文

版区分Resource Version author

権利Rights

© 2018 Internat ional Society for Neurochemistry. This is the peerreviewed version of the following art icle: [Journal of Neurochemistry,146(4):459-473, 2018], which has been published in final form atht tps://doi.org/10.1111/jnc.14451. This art icle may be used for non-commercial purposes in accordance with Wiley Terms and Condit ionsfor Use of Self-Archived Versions.

DOI 10.1111/jnc.14451

JaLCDOI

URL http://www.lib.kobe-u.ac.jp/handle_kernel/90005448

PDF issue: 2021-05-29

1

Hearing vulnerability after noise exposure in a mouse model of reactive oxygen 1

species overproduction 2

3

Shigefumi Morioka1,2, Hirofumi Sakaguchi2,*, Taro Yamaguchi3, Yuzuru Ninoyu1, Hiroaki Mohri1, 4

Takashi Nakamura1, Yasuo Hisa2,4, Kiyokazu Ogita3, Naoaki Saito1, Takehiko Ueyama1,* 5

6

1Laboratory of Molecular Pharmacology, Biosignal Research Center, Kobe University, Kobe 657-7

8501, Japan; 2Department of Otolaryngology-Head and Neck Surgery, Kyoto Prefectural 8

University of Medicine, Kyoto 602-8566, Japan; 3Laboratory of Pharmacology, Faculty of 9

Pharmaceutical Sciences, Setsunan University, Hirakata 573-0101, Osaka, Japan; 4Faculty of 10

Health and Medical Sciences, Kyoto Gakuen University, Kyoto 615-8577, Japan. 11

12

Running title: Hearing vulnerability after NE in NOX4-TG mice 13

Key words: antioxidant, Hsp47, hearing loss, NOX4, ROS, transgenic mouse model 14

15

*Corresponding authors: 16

Takehiko Ueyama 17

1-1 Rokkodai-cho, Nada-ku, Kobe 657-8501, Japan 18

Tel: +81-78-803-5962, Fax: +81-78-803-5971, E-mails: tueyama@kobe-u.ac.jp 19

Hirofumi Sakaguchi 20

465 Kajii-cho, Kawaramachi-Hirokoji, Kamigyo-ku, Kyoto 602-8566, Japan 21

Tel: +81-75-251-5603, Fax: +81-75-251-5604, E-mail: hiro-s@koto.kpu-m.ac.jp 22

2

Abstract 23

Previous studies have convincingly argued that reactive oxygen species (ROS) contribute to the 24

development of several major types of sensorineural hearing loss, such as noise-induced hearing 25

loss (NIHL), drug-induced hearing loss, and age-related hearing loss. However, the underlying 26

molecular mechanisms induced by ROS in these pathologies remain unclear. To resolve this issue, 27

we established an in vivo model of ROS overproduction by generating a transgenic (TG) mouse 28

line expressing the human NADPH oxidase 4 (NOX4, NOX4-TG mice), which is a constitutively 29

active ROS-producing enzyme that does not require stimulation or an activator. Overproduction 30

of ROS was detected at the cochlea of the inner ear in NOX4-TG mice, but they showed normal 31

hearing function under baseline conditions. However, they demonstrated hearing function 32

vulnerability, especially at high-frequency sounds, upon exposure to intense noise, which was 33

accompanied by loss of cochlear outer hair cells (OHCs). The vulnerability to loss of hearing 34

function and OHCs was rescued by treatment with the antioxidant Tempol. Additionally, we found 35

increased protein levels of the heat shock protein 47 (HSP47) in models using HEK293 cells, 36

including H2O2 treatment and cells with stable and transient expression of NOX4. Furthermore, 37

the upregulated levels of Hsp47 were observed in both the cochlea and heart of NOX4-TG mice. 38

Thus, antioxidant therapy is a promising approach for the treatment of NIHL. Hsp47 may be an 39

endogenous antioxidant factor, compensating for the chronic ROS overexposure in vivo, and 40

counteracting ROS-related hearing loss. 41

3

Abbreviations 42

4-HNE, 4-hydroxynoneal; Ab, antibody; ABR, auditory brainstem response; BCA, bicinchoninic 43

acid; CL, chemiluminescence, HBSS, Hank's balanced salt solution; HC, hair cell; H2O2, hydrogen 44

peroxide; Hsp47, heat shock protein 47; IB, immunoblotting; IHC, inner HC; IP, 45

immunoprecipitation; LB, spiral limbus; mAb, monoclonal antibody; NE, noise exposure; NIHL, 46

noise-induced hearing loss; NOX, NADPH oxidase; Nrf2, nuclear factor erythroid-2 related factor 47

2; OC, organ of Corti; OHC, outer hair cell; PFA, paraformaldehyde; polyA, polyadenylation; ROS, 48

reactive oxygen species; SG, spiral ganglion; SL, spiral ligament; LB, spiral limbus; SNHL, 49

sensorineural hearing loss; SOD, superoxide dismutase; SPL, sound pressure level; SV, stria 50

vascularis; Tempol (4-Hydroxy-2,2,6,6-tetramethylpiperidin-1-oxyl); TG mice, transgenic mice 51

4

Introduction 52

Reactive oxygen species (ROS), including superoxide anions, hydroxyl radicals, hydrogen 53

peroxide (H2O2), and singlet oxygen, are known to play key roles in numerous physiological and 54

pathological processes (Leto et al. 2009). Although appropriate levels of ROS are indispensable 55

for cell survival and differentiation (Leto et al. 2009), high levels induce oxidative stress in several 56

organs, including the inner ear, heart, brain, liver, and kidney, thereby triggering pathologies, such 57

as sensorineural hearing loss (SNHL), myocardial infarction, heart failure, neurodegenerative 58

disorders, and liver and renal fibrosis (Altenhofer et al. 2012, Wong & Ryan 2015, Yang et al. 59

2015a, Ma et al. 2017). 60

Hearing loss is one of the most common sensory deficits in humans, and about 90% of the 61

cases are accounted for by SNHL (Li et al. 2017). Most SNHL is caused by loss of hearing 62

sensitivity due to damage and/or cell death in the cochlea, an auditory detection organ (Wong & 63

Ryan 2015). The cochlea contains two types of hair cells (HCs), i.e., inner HCs (IHCs) and outer 64

HCs (OHCs). IHCs are responsible for detecting sounds and transmitting the acoustic information 65

to the brain. In contrast, OHCs are responsible for an active mechanical amplification process that 66

leads to high sensitivity and fine frequency resolution. The cochlea has a tonotopic gradient along 67

the length of the lumen in which the basal turn detects high-frequency sounds, while the apical 68

turn detects low-frequency sounds (Goutman et al. 2015, Wong & Ryan 2015). 69

Previous studies have convincingly argued that ROS contribute to the development of several 70

major types of SNHL, such as noise-induced hearing loss (NIHL), drug-induced hearing loss; 71

including the one caused by cisplatin; and age-related hearing loss (Wong & Ryan 2015, Yang et 72

al. 2015a). Among these pathologies, the ototoxic effect of oxidative stress has been most 73

frequently studied in NIHL. For example, biomarkers of lipid peroxidation, such as 4-74

hydroxynoneal (4-HNE) and malondialdehyde, are reportedly increased in the cochlear sensory 75

5

epithelium after noise exposure (NE) (Choi & Choi 2015, Le Prell et al. 2007, Fetoni et al. 2013). 76

In addition, antioxidants such as N-acetyl-L-cysteine (Le Prell et al. 2007, Coleman et al. 2007, 77

Choi et al. 2008) and 4-hydroxy-2,2,6,6-tetramethylpiperidine-n-oxyl (Tempol) (Murashita et al. 78

2006, Minami et al. 2007) have been reported to have a therapeutic effect on NIHL. Peri-traumatic 79

application of these drugs can decrease the permanent hearing threshold shift induced by NE. 80

Moreover, supplementation of antioxidants in animals and humans has been reported to prevent 81

progression of age-related hearing loss in some studies (Tavanai & Mohammadkhani 2017). 82

NADPH oxidase 3 (Nox3), one of the main NOXs in the inner ear, has been found to be 83

increased in the cochlea after administration of cisplatin in rats, which leads to ROS 84

overproduction and induces apoptosis in the cochlea, including OHC (Mukherjea et al. 2008, Kaur 85

et al. 2016). Subsequently, a small interfering RNA (siRNA) for Nox3 was reportedly effective in 86

reducing OHC damage and hearing function in rats (Mukherjea et al. 2010). On the contrary, Nox3 87

was proposed to be protective against NE based on a genome-wide association study, in which 88

Nox3 knockout (KO) mice were shown to be susceptible to NIHL (Lavinsky et al. 2015). In 89

addition, increased expression of Nox4 and Nox1 as well as Nox3 after administration of cisplatin 90

was reported in the mouse cochlea (Kim et al. 2010). Thus, ROS function and its underlying 91

molecular mechanism inducing SNHL may be different depending on the injury applied and the 92

Nox isoform involved. 93

To advance the understanding of the underlying molecular mechanisms in ROS-induced SNHL, 94

and especially in NIHL, we established an in vivo model for ROS overproduction by generating a 95

transgenic (TG) mouse line expressing human NOX4, which is a constitutively active ROS-96

producing enzyme (Leto et al. 2009). 97

6

Materials and methods 98

Plasmids 99

The human NOX4 in pcDNA3.1 was a kind gift from Dr. Thomas L. Leto (Boudreau et al. 100

2014). NOX4 with HindIII/NotI sites at the 5´ and 3´ ends was amplified by PCR, cloned into the 101

HindIII/NotI site of the p3xFLAG-CMV-10 vector (Sigma-Aldrich, St Louis, MO, USA), and 102

named 3xFLAG-NOX4. NOX4(1-305) in pcDNA3.1 was generated by insertion of a stop codon 103

immediately after Ser305 of NOX4 using the QuikChange Lightning Site-Directed Mutagenesis Kit 104

(Agilent Technologies, Santa Clara, CA, USA). In NOX4(1-305), the C-terminal, FAD-, and 105

NADPH-binding domains have been deleted, leading to loss of the ROS-generating capability 106

(Boudreau et al. 2014). 107

108

Generation of transgenic mice 109

This study was approved by the Institutional Animal Care and Use Committee and carried out 110

according to the Kobe University Animal Experimentation Regulation (24-04-08 and 26-03-05). 111

This study was not pre-registered. 112

Two PCR products, 3xFLAG-NOX4, with an XhoI site at the 5´ end, and the polyadenylation 113

(polyA) signal of the rabbit -globin, with an StuI site at the 3´ end, were amplified by PCR and 114

cloned into the XhoI/StuI sites of the pCAGGS vector, containing the CAG promoter, using the In-115

Fusion HDTM cloning kit (Takara Bio Inc., Tokyo, Japan). The final construct is shown is Fig. 1A. 116

After confirming the identity of the plasmid by sequencing and the protein expression of 117

3xFLAG-NOX4 in HEK293 cells by immunoblotting, its capabilities for ROS production were 118

examined. ROS production was found to be comparable between 3xFLAG-NOX4 and untagged 119

NOX4 (Fig. 1B). Thereafter, a purified fragment digested with SalI and StuI, which contained the 120

CAG promoter, 3xFLAG-NOX4, and polyA, was injected into fertilized eggs obtained at the 121

7

pronuclear stage from C57BL/6 mice (UNITECH, Kashiwa, Japan). Founder (F0) and first 122

generation (F1) mice were screened by PCR using the following primer pair: 5 ′ -123

CCTACAGCTCCTGGGCAACGTGTGCTGGT-3 ′ (located in the CAG promoter) and 5 ′ -124

AGAGGGAAAAAGATCTCAGTGGTAT-3′ (located in the polyA). We obtained four F0 and two 125

F1 mouse lines. The copy number of the transgene (CAG promoter; 3xFLAG-NOX4;polyA), as 126

integrated into the genome, was evaluated by Southern blotting; one line carried 12 and the other 127

carried 17 copies. The second and later generations of heterozygous NOX4-TG mice were used for 128

subsequent analyses. Offspring were genotyped by PCR using the above-mentioned primer pair. 129

All mice were identified by numbered ear tags. Mice were housed in specific pathogen-free 130

conditions using the individually ventilated cage system (Techniplast, Tokyo, Japan), and allowed 131

food and water ad libitum. The animal facility was maintained on a 14 h light and 10 h dark cycle 132

at 23 ± 2 °C and 50 ± 10% humidity. Mice from control group were always treated and assessed 133

first, followed by experimental group. No randomization was performed. Females were used in 134

the analyses unless otherwise indicated (mice younger than 1 week were not differentiated based 135

on sex). Age-matched wild type (WT) siblings were used as control. 136

137

Antibodies and chemicals 138

The following specific antibodies (Abs) were used (monoclonal unless indicated): 4-HNE 139

(HNE-J2; JaICA, Fukuroi, Japan; 1:100), heat shock protein 47 (HSP47) polyclonal (H-300; 140

RRID:AB_2185194, Santa Cruz Biotechnology; Santa Cruz, CA, USA; 1:50 for 141

immunohistochemistry and 1:200 for immunoblotting), and Hsp47 (EPR4217; 142

RRID:AB_10888995, Abcam; 1:200 for immunohistochemistry and 1:1000 for immunoblotting). 143

The anti-p22phox (#449, 1:200) Ab has been described previously (Ueyama et al. 2006). Magnetic 144

bead-conjugated anti-FLAG Ab (RRID:AB_2716804) and horseradish peroxidase (HRP)-145

8

conjugated anti-FLAG (1:2000, RRID:AB_2687989), anti-glyceraldehyde 3-phosphate 146

dehydrogenase (GAPDH; 1:2000, RRID:AB_10699462), anti-tubulin(1:2000, 147

RRID:AB_10695326), and anti-actin-(1:2000, PM053-7) Abs were obtained from MBL 148

International (Nagoya, Japan). Alexa Fluor 488-conjugated phalloidin (1:500, AB_2315147) and 149

HRP-conjugated secondary (1:10,000) Abs were obtained from Invitrogen (Carlsbad, CA, USA) 150

and Jackson Immuno Research Laboratories (West Grove, PA, USA), respectively. Validation data 151

for Abs is available from the companies. Tempol (4-Hydroxy-2,2,6,6-tetramethylpiperidin-1-oxyl), 152

an antioxidant metabolizing superoxide and H2O2 (Wilcox & Pearlman 2008), and diphenylene 153

iodonium (DPI) were obtained from Sigma-Aldrich. Chloral hydrate, H2O2, and HistoVT One 154

were obtained from Nacalai Tesque (Kyoto, Japan). 155

156

Cell culture and transfection 157

HEK293 cells (ATCC, Cat# CRL-1573, RRID:CVCL_0045) were maintained in Eagle’s 158

minimal essential medium (Wako Pure Chemical Industries, Osaka, Japan) containing 10% fetal 159

bovine serum (Nichirei Biosciences, Tokyo, Japan), 100 M nonessential amino acids (Wako), 160

100 units/mL penicillin, and 100 g/mL streptomycin at 37 °C in 5% CO2. Stable protein-161

expressing HEKFLAG-NOX4 and HEKcont cells were generated by transfection of the 3xFLAG-NOX4 162

or the control p3xFLAG-CMV-10 plasmids, respectively, by electroporation (NEPA21; NEPA 163

GENE, Ichikawa, Japan), followed by selection with the antibiotic G418 (0.5 mg/mL; Wako). 164

For transfection, HEK293 cells were seeded into 6-well dishes at 2.5 × 105 cells/well, 48 h 165

before. Plasmids were transfected into HEK293 cells using FuGENE6 (Promega, Madison, WI, 166

USA). Six hours after transfection, the cell culture medium was replaced with complete medium. 167

168

Sample preparation and immunoblotting 169

9

The membranous cochleae (two from each mouse) of postnatal day 7 (P7) or P46 pups, and 170

the heart of adult NOX4-TG and control mice were homogenized and lysed in lysis buffer 171

(Nakamura et al. 2017) with 0.1% Triton X and protease inhibitor cocktail (Nacalai Tesque). Forty-172

eight hours after transfection with plasmids or 24 h after treatment with H2O2, HEK293 cells were 173

lysed in the same lysis buffer as above. Each lysate was centrifuged at 13,000 g for 10 min at 4 °C 174

and supernatants were measured with the bicinchoninic acid (BCA) protein assay kit (Thermo 175

Fisher Scientific, Waltham, MA, USA). The supernatants were separated by SDS-PAGE and 176

transferred onto PVDF membranes. The membranes were incubated overnight at 4 °C with the 177

anti-FLAG, anti-HSP47, or anti-p22phox Ab, and then incubated for 30 min at 23 °C with a HRP-178

conjugated secondary Ab. Immunoreactivity was detected using the ECL detection system (Bio-179

Rad Laboratories, Hercules, CA, USA). For quantitative assessment of immunoreactive bands, 180

ImageJ software (National Institutes of Health) was used as previously described (Ueyama et al. 181

2016). Protein expression levels were normalized to those of tubulin, actin-, or GAPDH. 182

183

ROS production assay 184

Twenty-four hours after transfection, 2 × 105 HEK293 cells in Hank's balanced salt solution 185

(HBSS; Wako) were used for assays. For detection of overall ROS, we performed a 186

chemiluminescence (CL) assay. Cells were incubated in the presence of 200 M luminol and 10 187

units/mL HRP (Sigma-Aldrich) for 10 min, and CL signal was detected using a luminometer 188

(Mithras LB940; Berthold Detection Systems, GmbH), as previously described (Ueyama et al. 189

2015). 190

To measure ROS production in the cochleae of NOX4-TG mice, two membranous cochleae 191

were dissected from each NOX4-TG or control mouse at P7. Tissues of the organ of Corti (OC) 192

and the lateral wall of the cochlea, including the spiral ligament (SL) and stria vascularis (SV), 193

10

were collected in HBSS. After homogenization, lysates were centrifuged at 20,000 g for 15 min at 194

4 °C and precipitates containing NOX4 were resuspended in HBSS. After measuring the protein 195

concentration of the membrane pellets using the BCA protein assay kit, equal amounts of protein 196

were used for CL detection of ROS using the luminometer, and incubated for 10 min in the 197

presence of 10 M FAD, 500 M NADPH, 1 mM NaN2, 200 M luminol, and 10 units/mL of 198

HRP, as previously described (Ago et al. 2004, Ueyama et al. 2015). ROS production was inhibited 199

by a 10-min prior incubation with 0.1 M Tempol or 5 M DPI. Because this ROS assay was 200

performed in-vitro (cell-free), Tempol was used at the minimum concentration previously reported 201

(Bhattacharyya et al. 2008). 202

203

Immunohistochemistry 204

To examine surface preparations of the cochleae, dissected tissues were fixed with 4% 205

paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB, pH 7.4) and decalcified in 0.12 M EDTA 206

for 1 week at 4 °C (Ueyama et al. 2014). After permeabilization with phosphate-buffered saline 207

(PBS) containing 0.3% Triton X-100, fixed tissues were incubated with Alexa Fluor 488-208

conjugated phalloidin for 2 h at 23 °C. Stained tissues were mounted in Prolong anti-fade 209

(Invitrogen) with a coverslip and observed under an LSM700 confocal microscope (Carl Zeiss, 210

Jena, Germany). To evaluate OHC loss after NE, the percentages of remaining OHCs were 211

calculated separately at the apical, middle, and basal turns of the cochlea. 212

To analyze cross-sections of cochleae, tissues were isolated from P14 or adult mice and fixed 213

with 4% PFA in 0.1 M PB. Decalcified cochleae were embedded into paraffin blocks and cut into 214

6-m slices on a Leica RM2125 RTS manual rotary microtome (Leica Biosystems, Wetzlar, 215

Germany). Sections were immunostained after deparaffinization through a xylene and ethanol 216

graded series. For antigen unmasking, the slides were bathed in HistoVT One (Nacalai Tesque) for 217

11

20 min at 80 °C. Retrieved tissues were blocked in either 10% normal goat serum and 3% H2O2 218

(Nacalai Tesque) in PBS/0.03% Triton, or in 0.1% phenylhydrazine (Nacalai Tesque) in 219

PBS/0.03% Triton for 20 min at 23 °C. The tissues were incubated with primary Abs for 2 h at 220

23 °C in PBS/0.03% Triton, followed by MACH 2 Universal HRP-Polymer Detection (BIOCARE 221

Medical, Pacheco, CA, USA) for 30 min at 23 °C. 222

To analyze cryosections of hearts, adult mice were transcardially perfused with ice-cold 0.9% 223

saline solution, and subsequently with 4% PFA in 0.1 M PB. Hearts were dissected and post-fixed 224

overnight in the same fresh fixative. Twenty-μm cryosections were immunostained for HSP47, 225

followed by MACH 2 Universal HRP-Polymer Detection, as previously described (Nakamura et 226

al. 2017). Sections were visualized after staining with 3,3-diaminobenzidine tetrahydrochloride 227

(DAB, Sigma-Aldrich) and 0.02% H2O2 in Tris-buffered saline, pH 7.6. Slides were washed in 228

PBS/0.03% Triton and mounted in Entellan New (Merck Millipore, Billerica, MA, USA), 229

coverslipped, and photographed under a light microscope (Axioplan II; Carl Zeiss) equipped with 230

a DP26 camera (Olympus, Tokyo, Japan). 231

For the quantitative assessment of DAB intensity (evaluated using 4-HNE or Hsp47 Ab), we 232

used ImageJ software and the color deconvolution plugin for proper separation of the DAB color 233

spectra, as previously described (Varghese et al. 2014). Briefly, the region of interest (ROI) was 234

manually determined with the polygon or freehand tool, and then the deconvoluted image was 235

analyzed pixel-by-pixel. The color threshold for the positive area was defined in the range of 61–236

125/255, and the ratio of the positive to the total image area was calculated and presented as a 237

percentage of the control sample ratio. 238

239

Auditory brainstem response measurement and NE 240

For ABR measurements (Ueyama et al. 2016), mice were anesthetized with chloral hydrate 241

12

(500 mg/kg, i.p.) and stainless-steel needle electrodes were placed at the vertex and ventro-lateral 242

to the left and right ears. Electroencephalographic recording was performed with BioSigRP 243

Software and the TDT System 3 (Tucker-Davis Technologies, Alachua, FL, USA), in order to 244

generate a click or tone-burst stimulation at 4, 12, or 20 kHz (click, 4, 12, and 20 kHz for mice 245

under baseline conditions; 4, 12, and 20 kHz for mice in the NE experiment). ABR waveforms 246

were recorded for 12.8 ms at a sampling rate of 40,000 Hz with 505,000 Hz bandpass-filter settings. 247

Waveforms from 500 stimuli were averaged. ABR threshold was determined by decreasing the 248

sound intensity by 5-dB steps from 90 dB sound pressure level (SPL) and defined as the lowest 249

stimulus intensity that produced a reliable wave III on the ABR. When there was no response to 250

the stimulation at 90 dB, the threshold was considered as 100 dB. Blinded data analysis was 251

performed by two otologists or two scientists. ABR of control and NOX4-TG mice under baseline 252

conditions was measured at the age of 3, 5 (14.3–15.7 g), 7 and 24 (27.0–30.0 g) weeks. The 253

experiments were terminated if the mice showed signs of unbearable pain. 254

255

For NE experiments, 5-week-old control and NOX4-TG mice were anesthetized with chloral 256

hydrate and exposed to 110 dB SPL of octave-band noise, centered at 8 kHz, for 1 h inside a sound 257

chamber. These conditions of noise exposure were shown to cause a permanent threshold shift in 258

WT mice (Yamaguchi et al. 2017). Each animal was placed in a cage and the cage placed in the 259

sound chamber, which was fitted with a speaker (300 HT; FOSTEX, Akishima, Japan) driven by 260

a noise generator (SF-06; RION, Kokubunji, Japan) and power amplifier (DAD-M100proHT; 261

FLYING MOLE, Hamamatsu, Japan). To ensure the uniformity of the stimulus, we calibrated and 262

measured the sound levels with a sound-level meter (NL-26; RION), positioned at the level of the 263

animal’s head. The ABRs were measured immediately before NE. Then, ABR was sequentially 264

measured immediately after NE (day 0), and on days 1, 2, 5, 7, and 14 after NE. Hearing 265

deterioration due to NE was evaluated by measuring the ABR threshold shift, calculated by the 266

13

differences in ABR threshold before and after the NE test. In the antioxidant rescue experiment, 267

Tempol was injected intraperitoneally twice, immediately before and 24 h after NE, at a 268

concentration of 300 mg/kg, determined based on previous studies (Lahiani et al. 2016). During 269

the experiments, the mice were observed through a small window in the sound chamber, and the 270

experiments were terminated if the mice showed signs of unbearable pain. 271

272

Immunoprecipitation and mass spectrometry 273

HEKcont and HEKFLAG-NOX4 cells were homogenized in ice-cold buffer (20 mM Tris-HCl, pH 274

7.4, 120 mM NaCl, 1 mM EDTA, 5 mM EGTA, and 0.3% CHAPS) containing protease and 275

phosphatase inhibitor cocktails. Lysates were immunoprecipitated with anti-FLAG Ab-conjugated 276

magnetic beads. The immunoprecipitates were washed three times with the homogenization buffer, 277

separated by SDS-PAGE, and then analyzed by silver staining (Silver Stain MS Kit; Wako) or 278

immunoblotting. Subsequently, analysis of protein sequences by mass spectrometry was carried 279

out as described previously (Oshiro et al. 2007). Briefly, the silver-stained gel regions that seemed 280

to be different between the HEKcont and HEKFLAG-NOX4 samples were broadly excised and de-281

stained. The proteins in the gels were reduced and alkylated, followed by in-gel digestion with 282

trypsin in 25 mM ammonium bicarbonate for 15 h at 37 °C. The resulting peptides were then 283

subjected to liquid chromatography electrospray ionization mass spectrometry/mass spectrometry 284

(MS/MS) using the LCQ Advantage ion-trap mass spectrometer (Thermo Finnigan, Somerset, NJ, 285

USA). Protein identification, based on product ion mass lists, was performed by MASCOT 286

“MS/MS ion search.” 287

288

Statistical analysis 289

The exact number of experiments for each condition is provided in the figure legends. All data 290

14

are presented as the mean ± SEM. Two groups were compared using an unpaired, two-tailed 291

Student’s t-test. For comparisons of more than two groups, one-way or two-way ANOVA was 292

performed, followed by Bonferroni’s post-hoc test for pairwise group differences. Statistical 293

analyses were performed using Prism 6.0 software (GraphPad). Sample size calculation was not 294

performed. 295

15

Results 296

Generation of NOX4-TG mice 297

Low levels of Nox4 and Nox1 are detected in the mouse cochlea and these levels are increased 298

after administration of cisplatin in the mouse cochlea (Kim et al. 2010). NOX3 is one of the main 299

NOX isoforms in the inner ear; however, it is a multicomponent enzyme that requires at least four 300

other components (p22phox, Noxo1 or p47phox, Noxa1 or p67phox, Rac) in order to work functionally 301

in mouse cells (Ueyama et al. 2006, Leto et al. 2009). In addition, the effects of NOX3 after injury 302

are still controversial. For example, NOX3 is ototoxic after administration of cisplatin (Mukherjea 303

et al. 2008, Kaur et al. 2016), whereas it is protective after NE (Lavinsky et al. 2015). However, 304

NOX4 heterodimer with p22phox is the only constitutively active enzyme in the seven-membered 305

NOX family, which produces ROS without the need for a co-activator or stimulation (Leto et al. 306

2009). Therefore, to establish a mouse model with overproduction of ROS in the cochlea, we 307

decided to generate NOX4-TG mice. The expression levels of endogenous p22phox protein are 308

upregulated by introduction of exogenous NOX4 (Leto et al. 2009); hence we reconstituted NOX4-309

based ROS production by a single transfection of NOX4 expressing plasmid, without p22phox. By 310

using a luminol-based method and HEK293 cells expressing various NOX4 expression or mock 311

plasmids, we confirmed that untagged and 3xFLAG-tagged NOX4 at the N-terminus (3xFLAG-312

NOX4) showed comparable ROS-production capabilities (Fig. 1B), consistent with a previous 313

report (Martyn et al. 2006). In NOX4-TG mice, the expression of 3xFLAG-NOX4 was systemically 314

controlled by the CAG promoter (Fig. 1A). Hence, we confirmed the capability for ROS production 315

of two F1 NOX4-TG mouse lines obtained, by using primary fibroblasts and astrocytes (data not 316

shown). ROS production was higher in cells from mice carrying 17 copies of the transgene than 317

from mice carrying 12 copies, and thus further studies were performed using the mouse line 318

carrying 17 copies. NOX4-TG mice were fertile and did not show any gross body abnormalities. 319

16

320

Overproduction of ROS in the inner ear of NOX4-TG mice 321

Immunoblotting analysis using a FLAG Ab showed that FLAG-tagged NOX4 protein was 322

expressed in the cochlea and heart of NOX4-TG but not of control mice at P7 (Fig. 1C). We then 323

examined the in vivo function of FLAG-NOX4 in the cochlea of NOX4-TG mice at P7. A ROS 324

production assay using membranous cochleae showed that ROS production was higher in NOX4-325

TG than in control mice. As expected, the generation of ROS in NOX4-TG cochleae was 326

suppressed to control levels after incubation with the antioxidant Tempol (Wilcox & Pearlman 327

2008) or the NOX inhibitor DPI (Fig. 1D). Furthermore, although the morphology of the OC at 328

the basal turn of the cochlea in NOX4-TG mice showed no apparent abnormality, 4-HNE 329

immunoreactivity, a marker for lipid peroxidation (Zarkovic et al. 2017), was higher in the SL, SV, 330

and OHCs of the cochlea in NOX4-TG than in control mice (Fig. 1E). Taken together, these results 331

(ROS overproduction and lipid peroxidation in the cochlea) indicate that NOX4-TG mice can be 332

used as a model for ROS-overproduction in the cochlea. 333

334

Normal hearing function and cochlear morphology in NOX4-TG mice 335

We evaluated the hearing function of 3-, 5-, 7-, and 24-week-old NOX4-TG mice, bred under 336

normal noise conditions, by measuring the ABR to click and tone-burst stimuli, the latter at 337

frequencies of 4, 12, and 20 kHz (Fig. 3D). Unexpectedly, we did not find any significant 338

difference in the ABR between NOX4-TG and control mice regardless of stimulus or age, 339

indicating that there is no direct effect of ROS overproduction on hearing function in mice bred 340

under normal noise conditions. ABR data at 5 and 24 weeks are shown in Fig. 2A. No apparent 341

morphological differences were observed in the basal, middle, and apical turns of the cochleae 342

between 5-week-old NOX4-TG and control mice, with few HC loss detected in the entire area that 343

17

was examined (Fig. 2B). 344

345

Hearing vulnerability of NOX4-TG mice after NE 346

Next, we examined whether ROS overproduction in NOX4-TG mice affects their hearing 347

function after stress and accelerates the progress of NIHL. We exposed anesthetized, 5-week-old 348

NOX4-TG and control mice to intense noise with an intensity of 110 dB SPL for 1 h (Fig. 3D), 349

which is known to induce a permanent threshold shift in the hearing level of WT mice (Yamaguchi 350

et al. 2017). ABR threshold shifts (in dB SPL) were highest immediately after NE at day 0 and 351

gradually decreased in the 14-day period after the injury, being more remarkable when tested at 352

higher (12 and 20 kHz) than at low frequencies (4 kHz) (Fig. 3A). At day 0, the ABR threshold 353

shifts of NOX4-TG mice were comparable to those of control mice (Fig. 3A). However, at day 5, 354

the threshold shifts at high frequencies (12 and 20 kHz), but not at 4 kHz, became significantly 355

higher in NOX4-TG mice (49.0 ± 1.9 and 61.0 ± 3.3, for 12 and 20 kHz, respectively) than in 356

control mice (34.0 ± 3.4 and 44.0 ± 3.3, for 12 and 20 kHz, respectively; Fig. 3A). At 20 kHz at 357

days 7 and 14, the ABR threshold shifts were significantly higher in NOX4-TG mice (59.0 ± 2.4 358

at day 7 and 53.0 ± 5.1 at day 14) than in control mice (36.9 ± 3.8 at day 7 and 32.1 ± 5.3 at day 359

14) (Fig. 3A). 360

We also examined the morphology of the basal, middle, and apical turns of the cochlea at day 361

14. A few OHCs were missing in the basal and middle turns, but not in the apical turn, in control 362

mice, consistent with a permanent ABR threshold shift at high frequencies (Fig. 3B). The 363

remaining OHCs were significantly less at the basal turn of the cochlea in NOX4-TG mice (71.4 ± 364

5.4% in NOX4-TG mice and 93.3 ± 0.6% in control mice; Fig. 3B and C), in line with the 365

significantly increased ABR threshold, observed specifically at 20 kHz. We did not observe a loss 366

of IHCs in either NOX4-TG or control mice (Fig. 3C), consistent with previous studies reporting 367

18

that OHCs are mainly damaged in NIHL (Harding et al. 2005, Yang et al. 2015b, Yamaguchi et al. 368

2017). These results suggest that the greater OHC loss is one of the main causes of the increased 369

threshold shift in NOX4-TG mice in the NE experiment. 370

371

Rescue of hearing vulnerability in NOX4-TG mice by the antioxidant Tempol 372

To confirm that the exacerbated NIHL in NOX4-TG mice was caused by ROS, we applied the 373

antioxidant Tempol (300 mg/kg) twice intraperitoneally; immediately before and 24 h after NE 374

(Fig. 3D). Tempol functions by metabolizing superoxide to H2O2 (termed as superoxide dismutase 375

[SOD]-mimetic action) and by catalytically and stoichiometrically metabolizing H2O2 (catalase-376

like action) (Wilcox & Pearlman 2008). As expected, the NIHL observed in the NOX4-TG group 377

was rescued by treatment with Tempol. At day 1, the ABR threshold shifts were significantly lower 378

in the Tempol-treated NOX4-TG (NOX4-TG + Tempol; 48.3 ± 3.3 at 12 kHz and 53.3 ± 3.3 at 20 379

kHz) than in the untreated control group (66.6 ± 1.7 at 12 kHz and 65.0 ± 2.9 at 20 kHz). At day 380

14, the ABR threshold shifts in the NOX4-TG + Tempol group (15.0 ± 2.9 at 4 kHz, 25.0 ± 2.9 at 381

12 kHz, and 38.3 ± 3.3 at 20 kHz) were decreased to the levels observed in WT control mice (Fig. 382

4A, see 3A). Thus, the threshold shifts at high-frequencies (12 and 20 kHz), but not at 4 kHz, were 383

significantly lower in the group treated with Tempol than in the untreated group during the entire 384

14-day period (Fig. 4A). Loss of OHCs at day 14 was also decreased in the NOX4-TG + Tempol 385

group compared with the untreated control NOX4-TG group, most apparently in the basal turn of 386

the cochlea (Fig. 4B). The percentage of remaining OHCs in the basal cochlear turn was 387

significantly larger in the NOX4-TG + Tempol group (95.3 ± 1.8 %) than in the control NOX4-TG 388

group (78.3 ± 5.5 %; Fig. 4B, right). These results strongly suggest that the exacerbated NIHL 389

observed in NOX4-TG mice is due to the increased levels of ROS. 390

391

19

Increased HSP47 protein levels due to ROS 392

Although our results suggested the involvement of ROS in NIHL progression in NOX4-TG 393

mice, their hearing function was maintained under baseline conditions, suggesting chronic 394

activation of signaling pathways that protect against oxidative stress. To explore these putative 395

pathways, we attempted to identify proteins that bind to and function with NOX4, by means of 396

immunoprecipitation assays. We used lysates of HEK293 cells that were stably transfected with 397

3xFLAG-NOX4 (HEKFLAG-NOX4) or 3xFLAG-CMV(10) (HEKcont) and incubated then with 398

magnetic-bead-conjugated anti-FLAG monoclonal Ab. Immunoblotting analysis confirmed the 399

expression of 3xFLAG-NOX4 protein, as well as increased levels of the endogenous NOX4- 400

heterodimerizing p22phox protein, in HEKFLAG-NOX4 but not in HEKcont cells (Fig. 5A). Using mass 401

spectrometry, we then identified six proteins that potentially form a complex with NOX4; heat 402

shock protein 47 (HSP47, also called SERPINH1), calnexin, glucose-regulated protein 78, 403

apoptosis-inducing factor, nucleolin, and lipoprotein lipase. Among these, we focused on HSP47, 404

a collagen-specific molecular chaperone that has been reportedly involved in cell survival 405

(Kawasaki et al. 2015, Franco et al. 2016). Therefore, we investigated whether HSP47 can form a 406

complex with NOX4, using immunoprecipitation. Unexpectedly, we failed to detect HSP47 in 407

immunoprecipitates of HEKFLAG-NOX4 lysates, indicating that HSP47 does not bind to NOX4. 408

Instead, immunoblotting using whole cell lysates showed significantly increased HSP47 protein 409

levels in HEKFLAG-NOX4 cells than in HEKcont cells (Fig. 5B). Increased HSP47 protein levels were 410

also confirmed in cells transfected with untagged NOX4 or 3xFLAG-NOX4, but not in those 411

transfected with mock plasmid or NOX4(1-305) (Fig. 5C). Furthermore, HSP47 protein levels in 412

HEK293 cells increased after treatment with H2O2 in a dose-dependent manner; this increase was 413

statistically significant after treatment with 100 (P = 0.011) but not 10 (P = 0.680) M H2O2 (Fig. 414

5D). These results suggest that ROS produced by NOX4 increase HSP47 protein levels. 415

20

416

Increased Hsp47 protein levels in the cochlea of NOX4-TG mice 417

Finally, we examined Hsp47 protein levels in the cochlea of NOX4-TG mice. In the cochlea 418

of adult control mice, we detected Hsp47 immunoreactivity in OHCs, the SL, spiral limbus (LB), 419

and spiral ganglion (SG); the SV was weakly positive. The immunoreactivity of Hsp47 at these 420

regions was significantly more intense in NOX4-TG than in age-matched control mice (Fig. 6A). 421

To confirm the increased protein levels of Hsp47 in the cochlea of NOX4-TG mice, we performed 422

immunoblotting. Indeed, Hsp47 protein levels were significantly increased in NOX4-TG above 423

those in control mice, as early as by the first postnatal week (Fig. 6B). In addition, we found that 424

immunoreactivity of Hsp47 in the heart was also significantly stronger in adult NOX4-TG mice 425

than in age-matched control mice (Fig. 6C). This result was also confirmed by immunoblotting 426

analysis of Hsp47 levels in the heart (Fig. 6D). These results suggest that the NOX4-ROS-Hsp47 427

pathway is present and functions in many different tissues, organs, and cells. In contrast, we could 428

not detect any Hsp47 protein upregulation in the cochlea 2 days after NE by 429

immunohistochemistry in either WT or NOX4-TG mice at 2 weeks of age (data not shown). 430

21

Discussion 431

Establishment of a transgenic ROS overproduction mouse model 432

Gene-targeted mouse models are useful tools for studying the mechanism of ROS involvement 433

in various types of organ failure, including SNHL. Several mouse models with genetic 434

modifications of ROS metabolism have been reported to show increased cochlear damage. For 435

example, mice with deletion of superoxide dismutase 1 (Sod1), a major antioxidant molecule, show 436

increased vulnerability to age-related hearing loss (McFadden et al. 1999). Several KO mouse 437

lines with deficiencies in other major antioxidants or their regulators, such as Sod2 (Tuerdi et al. 438

2017), glutathione peroxidase 1 (Ohlemiller et al. 2000), and nuclear factor erythroid-2 related 439

factor 2 (Nrf2) (Honkura et al. 2016), show vulnerability to NIHL. Moreover, pejvakin KO mice, 440

which have disturbed peroxisome antioxidant activity, show a similar phenotype (Delmaghani et 441

al. 2015). Unlike these strains, this newly established NOX4-TG mouse line is a genetic model of 442

increased ROS production and is thus more suitable for studying the direct effects of ROS exposure 443

on various organs. The auditory phenotype that we observed in NOX4-TG mice is the accelerated 444

NIHL, which is limited to high-frequency sounds. This is consistent with the results from previous 445

defense-failure models, indicating that a common pathology is induced in the cochlea, either by 446

failure of defense against or overload of ROS. NOX4 is a constitutively active NOX, localized at 447

internal cell membranes, and mainly produces H2O2 (H2O2 and superoxide are approximately 90% 448

and 10%, respectively) (Leto et al. 2009, Takac et al. 2011, Nisimoto et al. 2014). In contrast, NOX 449

members, other than Nox4 and Duox-s, primarily produce superoxide and probably have a 450

different subcellular localization (Leto et al. 2009, Nisimoto et al. 2014, Ueyama et al. 2015). 451

Hence, although superoxide is spontaneously converted to H2O2 (Winterbourn 2017), the different 452

effects of ROS in NIHL observed in Nox3 KO mice (Lavinsky et al. 2015) and our NOX4-TG mice 453

could be explained by the different effect of H2O2, produced by NOX4, and of superoxide, 454

22

produced by NOX3. 455

456

Hearing phenotype in NOX4-TG mice under baseline conditions and after NE 457

Infusion of chemicals, which are known to generate H2O2, superoxide, or hydroxyl radicals, 458

into the perilymphatic space, cause a significant increase in the threshold of compound action 459

potentials of the cochlear nerve (Clerici & Yang 1996). Paraquat, for example, enhances 460

superoxide production in a reaction catalyzed by NADPH oxidase. When delivered into the 461

perilymph through the round window, paraquat induces hearing loss and HC loss (Bielefeld et al. 462

2005). Moreover, H2O2 treatment of cultured OCs isolated from P3 WT mice causes HC death in 463

a concentration-dependent manner (Baker & Staecker 2012). 464

NOX4-TG mice; however, have no baseline cochlear damage. One possible explanation for 465

this is the insufficient ROS overproduction. An early study reported that NE causes a permanent 466

hearing threshold shift by producing 4-fold higher ROS levels in the cochlea (Ohlemiller et al. 467

1999). However, a more recent report suggested that even lower levels of ROS can cause 468

irreversible damage to the cochlea. Furthermore, cisplatin treatment, which increases ROS 469

production by 1.8–2.2 folds, decreases cell viability in HEI-OC1 cells, an auditory cell line (Kim 470

et al. 2010, Chang et al. 2014). In cultured OCs, cisplatin treatment causes a 1.19-fold higher ROS 471

production, leading to increased loss of HCs (Kikkawa et al. 2014). Moreover, a 2.05-fold increase 472

in ROS induced by NE was reported to cause permanent hearing loss, as confirmed by ABR in 473

vivo (Wilson et al. 2014). These data suggest that the 1.7-fold higher levels of ROS seen in NOX4-474

TG mice should be sufficient to affect baseline hearing function. 475

Another plausible explanation of this contradiction is that chronic exposure to excessive ROS 476

in a genetically engineered model, such as NOX4-TG mice, may result in a reduced effect on 477

cochlear function as compared to acute exposure. Indeed, NOX4-TG mice show a hearing 478

23

vulnerability upon NE that is induced by ROS, as confirmed by the compensation observed after 479

the antioxidant treatment. Mouse models with antioxidant-defense failure reportedly have normal 480

hearing (Tuerdi et al. 2017, Honkura et al. 2016) or only subtle hearing loss (Ohlemiller et al. 2000, 481

Delmaghani et al. 2015) under baseline conditions, but show exacerbated NIHL and OHC loss 482

after NE. For example, mice lacking Nrf2, which is a transcription factor regulating the major 483

antioxidant response elements (Loboda et al. 2016), have normal hearing but increased 484

susceptibility to NE (Honkura et al. 2016). Interestingly, the expression of Nrf2-targeted genes is 485

not altered in Nrf2 KO mice, suggesting the presence of other antioxidant signaling molecules and 486

pathways activated in the chronic phase of ROS/antioxidant imbalance. Considering the above, 487

the hypothetical modified redox signaling(s), probably mediated by antioxidants and induced by 488

chronic ROS exposure, is insufficient to prevent the irreversible damage caused by an acute 489

increase of ROS. 490

With respect to the lesions, we found that OHC is mainly affected by the acutely increased 491

ROS levels; however, other regions, including the SL, SV, LB, and SG could also be involved in 492

the pathology of NOX4-TG mice. The SL and SV were also reported to be affected by NE (Sha & 493

Schacht 2017), and our 4-HNE immunostaining showed increased ROS production in these 494

structures. Indeed, although we detected a larger threshold shift in NOX4-TG mice at 12 kHz, 495

particularly at day 5, we observed a significantly higher OHC loss specifically in the basal turn of 496

the cochlea 14 days after NE. 497

498

Potential role of HSP47 in hearing protection 499

In the present study, we identified Hsp47 as a novel signaling molecule that is upregulated in 500

the cochlea upon NOX4-induced ROS stress. Hsp47 is a collagen-specific molecular chaperone 501

and plays an important role in the biosynthesis, structural assembly, and homeostasis of collagen 502

24

(Nagata et al. 1988, Ito & Nagata 2017). It was reported to have cytoprotective effects against 503

oxidative stress, since it could inhibit the H2O2-induced apoptosis in the lymphoma U937 cell line, 504

when added to the cells (Franco et al. 2016). Previous studies have shown that NOX4 is involved 505

in the development of fibrosis in several organs, including lung (Hecker et al. 2009), kidney 506

(Barnes & Gorin 2011), and liver (Jiang et al. 2012). Knockdown of Hsp47 by siRNA was shown 507

to decrease Nox4 levels both in vitro and in vivo in fibrosis models (Morry et al. 2015). These 508

findings support our hypothesis that both NOX4 and Hsp47 are involved in a common signaling 509

pathway, while, our data further demonstrate the opposite regulation of Hsp47 by NOX4 and 510

suggest that NOX4ROSHsp47 signaling is regulated in an interdependent manner. Altogether, 511

these data indicate that Hsp47 is an attractive candidate as a hearing-protective molecule against 512

chronic ROS overexposure. In contrast, Hsp47 protein levels were not increased in the cochlea 513

after NE, suggesting that upregulation of Hsp47 in vivo is not induced in the acute phase after ROS 514

exposure, consistent with a previous report (Gong et al. 2012). It is thus possible that sufficient 515

amounts of Hsp47, administered in the acute phase, may contribute to reducing cochlear damage. 516

Further studies are required to explore the functional mechanism and potential therapeutic effect 517

of HSP47 on NIHL. 518

In summary, NOX4-TG mice have normal hearing, but show hearing vulnerability after intense 519

NE, suggesting that antioxidant pathways are induced/activated in chronic ROS overexposure. 520

These results also suggest that ROS have different effects on functional and damaged tissue, 521

depending on chronic or acute exposure. Hsp47 protein levels are upregulated in the cochlea of 522

NOX4-TG mice, suggesting that HSP47 may be an endogenous antioxidant factor compensating 523

for chronic ROS overexposure. 524

25

Acknowledgement 525

We thank Dr. Ken-ichi Yoshino (Biosignal Research Center, Kobe University) for analysis of MS 526

experiments. This study was supported by grants from the JSPS KAKENHI program, #17H04042 527

to TU and #17H0432 to NS; and by a grant from the joint research program of the Biosignal 528

Research Center, Kobe University, #281005 to HS. 529

530

Author disclosure statement 531

No competing financial interests exist. 532

533

Author contributions 534

TU planned the project. SM, TU, HS, and TN performed the molecular and cellular biology 535

experiments. SM, YN, and HM performed the histological experiments and analysis. TY and KO 536

performed noise exposure experiments and analysis. SM, TU, HS, YH, and NS analyzed the data. 537

SM, HS, and TU wrote the manuscript. 538

26

References 539

Ago, T., Kitazono, T., Ooboshi, H. et al. (2004) Nox4 as the major catalytic component of an 540

endothelial NAD(P)H oxidase. Circulation, 109, 227-233. 541

Altenhofer, S., Kleikers, P. W., Radermacher, K. A., Scheurer, P., Rob Hermans, J. J., Schiffers, P., 542

Ho, H., Wingler, K. and Schmidt, H. H. (2012) The NOX toolbox: validating the role of 543

NADPH oxidases in physiology and disease. Cell Mol Life Sci, 69, 2327-2343. 544

Baker, K. and Staecker, H. (2012) Low dose oxidative stress induces mitochondrial damage in hair 545

cells. Anat Rec (Hoboken), 295, 1868-1876. 546

Barnes, J. L. and Gorin, Y. (2011) Myofibroblast differentiation during fibrosis: role of NAD(P)H 547

oxidases. Kidney Int, 79, 944-956. 548

Bhattacharyya, S., Dudeja, P. K. and Tobacman, J. K. (2008) Carrageenan-induced NFkappaB 549

activation depends on distinct pathways mediated by reactive oxygen species and Hsp27 550

or by Bcl10. Biochim Biophys Acta, 1780, 973-982. 551

Bielefeld, E. C., Hu, B. H., Harris, K. C. and Henderson, D. (2005) Damage and threshold shift 552

resulting from cochlear exposure to paraquat-generated superoxide. Hear Res, 207, 35-42. 553

Boudreau, H. E., Casterline, B. W., Burke, D. J. and Leto, T. L. (2014) Wild-type and mutant p53 554

differentially regulate NADPH oxidase 4 in TGF-beta-mediated migration of human lung 555

and breast epithelial cells. Br J Cancer, 110, 2569-2582. 556

Chang, J., Jung, H. H., Yang, J. Y., Lee, S., Choi, J., Im, G. J. and Chae, S. W. (2014) Protective 557

effect of metformin against cisplatin-induced ototoxicity in an auditory cell line. Journal 558

of the Association for Research in Otolaryngology : JARO, 15, 149-158. 559

Choi, C. H., Chen, K., Vasquez-Weldon, A., Jackson, R. L., Floyd, R. A. and Kopke, R. D. (2008) 560

Effectiveness of 4-hydroxy phenyl N-tert-butylnitrone (4-OHPBN) alone and in 561

combination with other antioxidant drugs in the treatment of acute acoustic trauma in 562

chinchilla. Free Radic Biol Med, 44, 1772-1784. 563

Choi, S. H. and Choi, C. H. (2015) Noise-induced neural degeneration and therapeutic effect of 564

antioxidant drugs. J Audiol Otol, 19, 111-119. 565

Clerici, W. J. and Yang, L. (1996) Direct effects of intraperilymphatic reactive oxygen species 566

generation on cochlear function. Hear Res, 101, 14-22. 567

Coleman, J. K., Kopke, R. D., Liu, J., Ge, X., Harper, E. A., Jones, G. E., Cater, T. L. and Jackson, 568

R. L. (2007) Pharmacological rescue of noise induced hearing loss using N-acetylcysteine 569

and acetyl-L-carnitine. Hear Res, 226, 104-113. 570

Delmaghani, S., Defourny, J., Aghaie, A. et al. (2015) Hypervulnerability to sound exposure 571

27

through impaired adaptive proliferation of peroxisomes. Cell, 163, 894-906. 572

Fetoni, A. R., De Bartolo, P., Eramo, S. L. et al. (2013) Noise-induced hearing loss (NIHL) as a 573

target of oxidative stress-mediated damage: cochlear and cortical responses after an 574

increase in antioxidant defense. J Neurosci, 33, 4011-4023. 575

Franco, L., Terrinca, J., Rodriguez, A. B., Espino, J. and Pariente, J. A. (2016) Extracellular heat 576

shock proteins protect U937 cells from H2O2-induced apoptotic cell death. Mol Cell 577

Biochem, 412, 19-26. 578

Gong, T. W., Fairfield, D. A., Fullarton, L., Dolan, D. F., Altschuler, R. A., Kohrman, D. C. and 579

Lomax, M. I. (2012) Induction of heat shock proteins by hyperthermia and noise 580

overstimulation in hsf1 -/- mice. Journal of the Association for Research in 581

Otolaryngology : JARO, 13, 29-37. 582

Goutman, J. D., Elgoyhen, A. B. and Gomez-Casati, M. E. (2015) Cochlear hair cells: The sound-583

sensing machines. FEBS Lett, 589, 3354-3361. 584

Harding, G. W., Bohne, B. A. and Vos, J. D. (2005) The effect of an age-related hearing loss gene 585

(Ahl) on noise-induced hearing loss and cochlear damage from low-frequency noise. Hear 586

Res, 204, 90-100. 587

Hecker, L., Vittal, R., Jones, T., Jagirdar, R., Luckhardt, T. R., Horowitz, J. C., Pennathur, S., 588

Martinez, F. J. and Thannickal, V. J. (2009) NADPH oxidase-4 mediates myofibroblast 589

activation and fibrogenic responses to lung injury. Nat Med, 15, 1077-1081. 590

Honkura, Y., Matsuo, H., Murakami, S. et al. (2016) NRF2 is a key target for prevention of noise-591

induced hearing loss by reducing oxidative damage of cochlea. Scientific reports, 6, 19329. 592

Ito, S. and Nagata, K. (2017) Biology of Hsp47 (Serpin H1), a collagen-specific molecular 593

chaperone. Semin Cell Dev Biol, 62, 142-151. 594

Jiang, J. X., Chen, X., Serizawa, N., Szyndralewiez, C., Page, P., Schroder, K., Brandes, R. P., 595

Devaraj, S. and Torok, N. J. (2012) Liver fibrosis and hepatocyte apoptosis are attenuated 596

by GKT137831, a novel NOX4/NOX1 inhibitor in vivo. Free Radic Biol Med, 53, 289-597

296. 598

Kaur, T., Borse, V., Sheth, S. et al. (2016) Adenosine A1 receptor protects against cisplatin 599

ototoxicity by suppressing the NOX3/STAT1 inflammatory pathway in the cochlea. J 600

Neurosci, 36, 3962-3977. 601

Kawasaki, K., Ushioda, R., Ito, S., Ikeda, K., Masago, Y. and Nagata, K. (2015) Deletion of the 602

collagen-specific molecular chaperone Hsp47 causes endoplasmic reticulum stress-603

mediated apoptosis of hepatic stellate cells. J Biol Chem, 290, 3639-3646. 604

28

Kikkawa, Y. S., Nakagawa, T., Taniguchi, M. and Ito, J. (2014) Hydrogen protects auditory hair 605

cells from cisplatin-induced free radicals. Neuroscience letters, 579, 125-129. 606

Kim, H. J., Lee, J. H., Kim, S. J. et al. (2010) Roles of NADPH oxidases in cisplatin-induced 607

reactive oxygen species generation and ototoxicity. J Neurosci, 30, 3933-3946. 608

Lahiani, A., Hidmi, A., Katzhendler, J., Yavin, E. and Lazarovici, P. (2016) Novel Synthetic 609

PEGylated Conjugate of alpha-Lipoic Acid and Tempol Reduces Cell Death in a Neuronal 610

PC12 Clonal Line Subjected to Ischemia. ACS Chem Neurosci, 7, 1452-1462. 611

Lavinsky, J., Crow, A. L., Pan, C. et al. (2015) Genome-wide association study identifies nox3 as 612

a critical gene for susceptibility to noise-induced hearing loss. PLoS Genet, 11, e1005094. 613

Le Prell, C. G., Yamashita, D., Minami, S. B., Yamasoba, T. and Miller, J. M. (2007) Mechanisms 614

of noise-induced hearing loss indicate multiple methods of prevention. Hear Res, 226, 22-615

43. 616

Leto, T. L., Morand, S., Hurt, D. and Ueyama, T. (2009) Targeting and regulation of reactive 617

oxygen species generation by Nox family NADPH oxidases. Antioxid Redox Signal, 11, 618

2607-2619. 619

Li, L., Chao, T., Brant, J., O'Malley, B., Jr., Tsourkas, A. and Li, D. (2017) Advances in nano-based 620

inner ear delivery systems for the treatment of sensorineural hearing loss. Adv Drug Deliv 621

Rev, 108, 2-12. 622

Loboda, A., Damulewicz, M., Pyza, E., Jozkowicz, A. and Dulak, J. (2016) Role of Nrf2/HO-1 623

system in development, oxidative stress response and diseases: an evolutionarily conserved 624

mechanism. Cell Mol Life Sci, 73, 3221-3247. 625

Ma, M. W., Wang, J., Zhang, Q., Wang, R., Dhandapani, K. M., Vadlamudi, R. K. and Brann, D. 626

W. (2017) NADPH oxidase in brain injury and neurodegenerative disorders. Mol 627

Neurodegener, 12, 7. 628

Martyn, K. D., Frederick, L. M., von Loehneysen, K., Dinauer, M. C. and Knaus, U. G. (2006) 629

Functional analysis of Nox4 reveals unique characteristics compared to other NADPH 630

oxidases. Cell Signal, 18, 69-82. 631

McFadden, S. L., Ding, D., Burkard, R. F., Jiang, H., Reaume, A. G., Flood, D. G. and Salvi, R. J. 632

(1999) Cu/Zn SOD deficiency potentiates hearing loss and cochlear pathology in aged 633

129,CD-1 mice. The Journal of comparative neurology, 413, 101-112. 634

Minami, S. B., Yamashita, D., Ogawa, K., Schacht, J. and Miller, J. M. (2007) Creatine and Tempol 635

attenuate noise-induced hearing loss. Brain research, 1148, 83-89. 636

Morry, J., Ngamcherdtrakul, W., Gu, S., Goodyear, S. M., Castro, D. J., Reda, M. M., Sangvanich, 637

29

T. and Yantasee, W. (2015) Dermal delivery of HSP47 siRNA with NOX4-modulating 638

mesoporous silica-based nanoparticles for treating fibrosis. Biomaterials, 66, 41-52. 639

Mukherjea, D., Jajoo, S., Kaur, T., Sheehan, K. E., Ramkumar, V. and Rybak, L. P. (2010) 640

Transtympanic administration of short interfering (si)RNA for the NOX3 isoform of 641

NADPH oxidase protects against cisplatin-induced hearing loss in the rat. Antioxid Redox 642

Signal, 13, 589-598. 643

Mukherjea, D., Jajoo, S., Whitworth, C., Bunch, J. R., Turner, J. G., Rybak, L. P. and Ramkumar, 644

V. (2008) Short interfering RNA against transient receptor potential vanilloid 1 attenuates 645

cisplatin-induced hearing loss in the rat. J Neurosci, 28, 13056-13065. 646

Murashita, H., Tabuchi, K., Hoshino, T., Tsuji, S. and Hara, A. (2006) The effects of tempol, 3-647

aminobenzamide and nitric oxide synthase inhibitors on acoustic injury of the mouse 648

cochlea. Hear Res, 214, 1-6. 649

Nagata, K., Saga, S. and Yamada, K. M. (1988) Characterization of a novel transformation-650

sensitive heat-shock protein (HSP47) that binds to collagen. Biochem Biophys Res 651

Commun, 153, 428-434. 652

Nakamura, T., Ueyama, T., Ninoyu, Y. et al. (2017) Novel role of Rac-Mid1 signaling in medial 653

cerebellar development. Development, 144, 1863-1875. 654

Nisimoto, Y., Diebold, B. A., Cosentino-Gomes, D. and Lambeth, J. D. (2014) Nox4: a hydrogen 655

peroxide-generating oxygen sensor. Biochemistry, 53, 5111-5120. 656

Ohlemiller, K. K., McFadden, S. L., Ding, D. L., Lear, P. M. and Ho, Y. S. (2000) Targeted mutation 657

of the gene for cellular glutathione peroxidase (Gpx1) increases noise-induced hearing loss 658

in mice. Journal of the Association for Research in Otolaryngology : JARO, 1, 243-254. 659

Ohlemiller, K. K., Wright, J. S. and Dugan, L. L. (1999) Early elevation of cochlear reactive 660

oxygen species following noise exposure. Audiol Neurootol., 4, 229-236. 661

Oshiro, N., Takahashi, R., Yoshino, K. et al. (2007) The proline-rich Akt substrate of 40 kDa 662

(PRAS40) is a physiological substrate of mammalian target of rapamycin complex 1. J Biol 663

Chem, 282, 20329-20339. 664

Sha, S. H. and Schacht, J. (2017) Emerging therapeutic interventions against noise-induced hearing 665

loss. Expert Opin Investig Drugs, 26, 85-96. 666

Takac, I., Schroder, K., Zhang, L., Lardy, B., Anilkumar, N., Lambeth, J. D., Shah, A. M., Morel, 667

F. and Brandes, R. P. (2011) The E-loop is involved in hydrogen peroxide formation by the 668

NADPH oxidase Nox4. J Biol Chem, 286, 13304-13313. 669

Tavanai, E. and Mohammadkhani, G. (2017) Role of antioxidants in prevention of age-related 670

30

hearing loss: a review of literature. Eur Arch Otorhinolaryngol, 274, 1821-1834. 671

Tuerdi, A., Kinoshita, M., Kamogashira, T., Fujimoto, C., Iwasaki, S., Shimizu, T. and Yamasoba, 672

T. (2017) Manganese superoxide dismutase influences the extent of noise-induced hearing 673

loss in mice. Neuroscience letters, 642, 123-128. 674

Ueyama, T., Geiszt, M. and Leto, T. L. (2006) Involvement of Rac1 in activation of 675

multicomponent Nox1- and Nox3-based NADPH oxidases. Mol Cell Biol, 26, 2160-2174. 676

Ueyama, T., Ninoyu, Y., Nishio, S. Y. et al. (2016) Constitutive activation of DIA1 (DIAPH1) via 677

C-terminal truncation causes human sensorineural hearing loss. EMBO Mol Med, 8, 1310-678

1324. 679

Ueyama, T., Sakaguchi, H., Nakamura, T. et al. (2014) Maintenance of stereocilia and apical 680

junctional complexes by Cdc42 in cochlear hair cells. J Cell Sci, 127, 2040-2052. 681

Ueyama, T., Sakuma, M., Ninoyu, Y., Hamada, T., Dupuy, C., Geiszt, M., Leto, T. L. and Saito, N. 682

(2015) The extracellular A-loop of dual oxidases affects the specificity of reactive oxygen 683

species release. J Biol Chem, 290, 6495-6506. 684

Varghese, F., Bukhari, A. B., Malhotra, R. and De, A. (2014) IHC Profiler: an open source plugin 685

for the quantitative evaluation and automated scoring of immunohistochemistry images of 686

human tissue samples. PloS one, 9, e96801. 687

Wilcox, C. S. and Pearlman, A. (2008) Chemistry and antihypertensive effects of tempol and other 688

nitroxides. Pharmacol Rev, 60, 418-469. 689

Wilson, T., Omelchenko, I., Foster, S., Zhang, Y., Shi, X. and Nuttall, A. L. (2014) JAK2/STAT3 690

inhibition attenuates noise-induced hearing loss. PloS one, 9, e108276. 691

Winterbourn, C. C. (2017) Biological Production, Detection, and Fate of Hydrogen Peroxide. 692

Antioxid Redox Signal. 693

Wong, A. C. and Ryan, A. F. (2015) Mechanisms of sensorineural cell damage, death and survival 694

in the cochlea. Front Aging Neurosci, 7, 58. 695

Yamaguchi, T., Yoneyama, M. and Ogita, K. (2017) Calpain inhibitor alleviates permanent hearing 696

loss induced by intense noise by preventing disruption of gap junction-mediated 697

intercellular communication in the cochlear spiral ligament. Eur J Pharmacol, 803, 187-698

194. 699

Yang, C. H., Schrepfer, T. and Schacht, J. (2015a) Age-related hearing impairment and the triad of 700

acquired hearing loss. Front Cell Neurosci, 9, 276. 701

Yang, W., Vethanayagam, R. R., Dong, Y., Cai, Q. and Hu, B. H. (2015b) Activation of the antigen 702

presentation function of mononuclear phagocyte populations associated with the basilar 703

31

membrane of the cochlea after acoustic overstimulation. Neuroscience, 303, 1-15. 704

Zarkovic, K., Jakovcevic, A. and Zarkovic, N. (2017) Contribution of the HNE-705

immunohistochemistry to modern pathological concepts of major human diseases. Free 706

Radic Biol Med, 111, 110-126. 707

32

Figure legends 708

Fig. 1. Generation of NOX4-TG mice: a model of ROS overproduction in the cochlea. 709

(A) Illustration of the linearized expression plasmid used for injection into fertilized eggs. The 710

cassette contains the CAG promoter, the 3xFLAG-tagged NOX4, and the rabbit -globin poly A 711

sequence. (B) ROS production, measured using a luminol-based chemiluminescence (CL) method, 712

in HEK293 cells transfected with mock, NOX4, 3xFLAG-tagged NOX4, or NOX4(1-305) plasmids. 713

No significant difference is detected between NOX4 and 3xFLAG-tagged NOX4 (n = 4; P = 714

0.2841; Student’s t-test). (C) Immunoblotting analysis using anti-FLAG and anti-actin-β (used as 715

loading control) antibodies. Expression of FLAG-tagged NOX4 protein is confirmed in the cochlea 716

and heart of NOX4-TG mice at P7 (right panels) but not in control mice (left panels). Multiple 717

bands marked by arrowheads, probably correspond to glycosylated forms of NOX4 (predicted MW 718

= 66.5 kDa). FLAG-NOX4 levels are much higher in the heart, even with much lower protein 719

loaded, as evident by actin-β bands. n = 4. (D) ROS production assessment in cochlear lysates 720

obtained from P7 mice, using a luminol-based CL method. Overproduction of ROS in the cochleae 721

of NOX4-TG mice is suppressed to control levels by Tempol or DPI; n = 4; P = 0.016 [control vs 722

TG], P = 0.010 [TG vs TG+Tempol], and P = 0.013 [TG vs TG+DPI]). *, P <0.05; one-way 723

ANOVA with Bonferroni’s post-hoc test. (E) Representative sections of cochleae from P14 control 724

and NOX4-TG mice stained with anti-4-HNE antibody. Rectangles in the upper panels (scale bars, 725

50 m) indicate the magnified areas shown in the lower panels (scale bars, 25 m). The asterisk 726

and black dot indicate Corti’s tunnel and Nuel’s space, respectively. Graph on the right is showing 727

the quantification of 4-HNE immunoreactivity in outer hair cells (OHCs; P = 0.045), spiral 728

ligament (SL; P = 0.001), and stria vascularis (SV; P = 0.045). n = 5; *, P <0.05; **, P <0.001; 729

Student’s t-test. 730

731

33

Fig. 2. Normal hearing and morphology of the organ of Corti (OC) in NOX4-TG mice. 732

(A) ABR thresholds (dB SPL) tested with clicks and pure-tone bursts at 4, 12, and 20 kHz in 5- 733

and 24-week-old control (n = 4) and NOX4-TG mice (n = 4 and 6 at 5 and 24 weeks, respectively). 734

No statistically significant difference is observed between control and NOX4-TG mice. (B) 735

Representative histochemical images of the OC at the three cochlear turns (apical, middle, and 736

basal) of control and NOX4-TG mice at 5 weeks of age. Sections were stained using Alexa Fluor 737

488-labeled phalloidin. No significant difference is observed between the two groups. IHC, inner 738

hair cell; OHC, outer hair cell. Scale bars, 25 m. 739

740

Fig. 3. NE causes vulnerability of hearing and OHC loss in NOX4-TG mice. 741

(A) Five-week-old mice were exposed to intense noise (110 dB) for 1 h. ABR threshold shifts (dB 742

SPL) between pre- and post-NE, were tested at 4, 12, and 20 kHz and monitored immediately after 743

NE (day 0) and for the next 14 days. A significant difference is observed at 12 kHz at day 5 (*, P 744

= 0.041) and at 20 kHz at days 5, 7, and 14 (*, P = 0.033, 0.022, and 0.028, respectively). n = 7 in 745

control and n = 5 in NOX4-TG mice; two-way ANOVA with Bonferroni’s post-hoc test. (B) 746

Representative histochemical images showing HCs at the three cochlear turns (apical, middle, and 747

basal) of the OC, in sections obtained from control and NOX4-TG mice at day 14 and stained using 748

Alexa Fluor 488-labeled phalloidin. OHC loss is most prominent in the basal turn, which is 749

responsive to high-frequency tones, and is more severe in NOX4-TG than in control mice. White 750

dots indicate missing OHCs. The quantification of the percentage of OHCs remaining in the apical, 751

middle, and basal turn is shown in C (n = 7 in control mice and 8 in NOX4-TG mice; **, P = 0.002 752

in the basal turn; Student’s t-test). Scale bars, 25 m. (D) Chart showing the timings of ABR 753

measurements under baseline conditions and in NE experiments. Bf, immediately before NE; Af, 754

immediately after NE. 755

34

756

Fig. 4. Tempol reduces the NE-induced hearing vulnerability and OHC loss in NOX4-TG 757

mice. 758

(A) Intense noise (110 dB) was applied to 5-week-old NOX4-TG mice for 1 h, with or without 759

combined pre- and post-treatment with Tempol (300 mg/kg, i.p.). ABR threshold shifts at 4, 12, 760

and 20 kHz were monitored immediately after NE (day 0) and for the next 14 days. At 12 kHz, P 761

= 0.029 at day 1, 0.025 at day 2, 0.021 at day 5, 0.016 at day 7, and 0.014 at day 14. At 20 kHz, P 762

= 0.026 at day 1, 0.010 at day 2, 0.026 at day 5, 0.005 at day 7, and 0.004 at day 14. n = 3; *, P 763

<0.05; **, P <0.01; ***, P <0.001; two-way ANOVA with Bonferroni’s post-hoc test. (B) 764

Representative sections stained using Alexa Fluor 488-labeled phalloidin, showing OHCs at the 765

basal turn of the OC at day 14 (left). White dots indicate missing OHCs. Graph on the right is 766

showing the percentage of OHCs remaining in the basal turn (n = 6 in control and 5 in NOX4-TG 767

mice; *, P = 0.031; Student’s t-test). Scale bars, 25 m. 768

769

Fig. 5. Upregulated protein levels of HSP47 by NOX4 and ROS in HEK293 cells. 770

(A) Immunoblotting analysis of FLAG-NOX4 and endogenous p22phox proteins in HEKFLAG-NOX4 771

cells using anti-FLAG and anti-p22phox antibodies, respectively, and anti-tubulin as a loading 772

control. Graph is showing the quantification of p22phox expression levels (n = 3; ***, P= 0.0007; 773

Student’s t-test). (B) Immunoprecipitation (IP) of HEKcont and HEKFLAG-NOX4 cell lysates, using 774

FLAG antibody-conjugated magnetic beads, followed by immunoblotting (IB) using an anti-775

HSP47 antibody. Binding of FLAG-NOX4 to HSP47 is not detected; however, significantly higher 776

levels of HSP47 protein are detected in HEKFLAG-NOX4 compared to HEKcont lysates. Quantification 777

of the immunoreactive bands in the first two lanes over GAPDH, used as a loading control, is 778

shown in the graph (n = 5; ***, P = 0.0009; Student’s t-test). (C) Immunoblotting analysis of 779

35

HEK293 cell lysates 48 hours after transfection with mock, NOX4, FLAG-NOX4, or NOX4(1-305) 780

plasmid. Blots were incubated with an anti-HSP47 antibody and actin-β, used as a loading control. 781

Graph is showing the quantification of HSP47 protein levels. HSP47 protein levels are 782

significantly increased in NOX4- and FLAG-NOX4-transfected HEK293 cells (P = 0.0034 and 783

0.0009, respectively, but are unchanged in mock- and NOX4(1-305)-transfected cells (P = 0.007 784

for NOX4 vs. NOX4(1-305) and P = 0.002 for FLAG-NOX4 vs. NOX4(1-305); n = 5; one-way 785

ANOVA with Bonferroni’s post-hoc test). **, P <0.01; ***, P <0.001. (D) Immunoblotting 786

analysis of HEK293 cell lysates after treatment with H2O2 for 24 h (10 M or 100 M), using anti-787

HSP47 and anti-actin-β antibodies, the latter used as a loading control. H2O2 increases HSP47 788

protein levels in a concentration-dependent manner (n = 3; *, P = 0.011; one-way ANOVA with 789

Bonferroni’s post-hoc test). 790

791

Fig. 6. Upregulation of Hsp47 protein levels in NOX4-TG mice. 792

(A) Representative sections of the cochleae from adult control and NOX4-TG mice stained with 793

an anti-HSP47 antibody. Rectangles in the upper panels (scale bars, 100 m) indicate the 794

magnified areas shown in the lower panels (scale bars, 25 m). Magnified images show sensory 795

regions of the OC, containing OHCs. The asterisk and black dot indicate Corti’s tunnel and Nuel’s 796

space, respectively. Graph is showing the quantification of Hsp47 immunoreactivity at OHCs (P 797

= 0.014), SL (P = 0.005), SV (P = 0.004), LB (P = 0.0006), and SG (P = 0.045). n = 3; *, P <0.05; 798

**, P <0.01; ***, P <0.001; Student’s t-test. (B) Immunoblotting analysis of Hsp47 protein levels 799

in membranous cochleae obtained from control and NOX4-TG mice at P4–P6, using an HSP47 800

antibody. Tubulin was used as a loading control. Graph is showing the quantification of the 801

immunoreactive bands (n = 3; *, P = 0.015; Student’s t-test). (C) Representative sections of the 802

heart from adult control and NOX4-TG mice, immunostained with an anti-HSP47 antibody. Graph 803

36

is showing the quantification of Hsp47 immunoreactivity (n = 4; *, P = 0.003; Student’s t-test). 804

Scale bars, 200 m. (D) Immunoblotting analysis of heart lysates obtained from adult control and 805

NOX4-TG mice, using an anti-HSP47 antibody and anti-GAPDH, used as a loading control. Graph 806

is showing the quantification of Hsp47 protein levels (n = 3; *, P = 0.017; Student’s t-test). 807

37

808

809

810

38

811

812

39

813

814

815

816

817

40

818