Catalase Deletion Promotes PreDiabetic Phenotype in Mice

Abstract

Hydrogen peroxide is produced endogenously and can be toxic to living organisms by inducing oxidative stress and cell damage. However, it has also been identified as a signal transduction molecule. By metabolizing hydrogen peroxide, catalase protects cells and tissues against oxidative damage and may also influence signal transduction mechanisms. Studies suggest that acatalasemic individuals (i.e., those with very low catalase activity) have a higher risk for the development of diabetes. We now report catalase knockout (Cat^-/-) mice, when fed a normal (6.5% lipid) chow, exhibit an obese phenotype that manifests as an increase in body weight that becomes more pronounced with age. The mice demonstrate altered hepatic and muscle lipid deposition, as well as increases in serum and hepatic triglycerides (TGs), and increased hepatic transcription and protein expression of PPARγ. Liver morphology revealed steatosis with inflammation. Cat^-/- mice also exhibited pancreatic morphological changes that correlated with impaired glucose tolerance and increased fasting serum insulin levels, conditions consistent with pre-diabetic status. RNA-seq analyses revealed a differential expression of pathways and genes in Cat^-/- mice, many of which are related to metabolic syndrome, diabetes, and obesity, such as Pparg and Cidec.

In conclusion, the results of the present study show mice devoid of catalase develop an obese, pre-diabetic phenotype and provide compelling evidence for catalase (or its products) being integral in metabolic regulation.

Introduction

Alterations in redox balance are found in numerous pathologies. These often occur as a result of increased reactive oxygen species (ROS) or decreased antioxidant defense systems. Hydrogen peroxide, an important ROS, is produced endogenously and historically has been perceived to be deleterious by promoting oxidant damage to cells [1]. Well-known for its role in metabolizing hydrogen peroxide to water and oxygen, catalase prevents cellular oxidant damage by removing hydrogen peroxide. In the absence of catalase, hydrogen peroxide can be metabolized by other antioxidant enzymes, such as glutathione peroxidase (GPX), or it can form hydroxyl radicals through Fenton reactions [2]. These hydroxyl radicals can initiate a free radical cascade and thereby damage numerous macromolecules, such as lipids (peroxidation), and promote amino acid oxidation. Catalase, normally localized to the peroxisome where beta oxidation reactions produce hydrogen peroxide [3], functions as an antioxidant by catalyzing the conversion of hydrogen peroxide to oxygen and water. Several studies have emphasized the correlation between oxidative stress, obesity, and insulin resistance in type 2 diabetes [4-7]. In addition to its well-established role as a oxidant, hydrogen peroxide can act as a physiological signal transduction molecule [8]. Catalase, by influencing cellular levels of hydrogen peroxide, would be anticipated to also regulate hydrogen peroxide-dependent signal transduction pathways.

Acatalasemia is a rare genetic deficiency in humans that involves severe reductions in catalase activity. Given the important role played by catalase in detoxifying hydrogen peroxide, it would be predicted that catalase-deficient populations (acatalasemic and hypocatalasemic) would suffer from overt oxidative damage-related disease; however, this has not proven to be the case. Acatalasemia in humans has been considered to be an asymptomatic disorder [9] and catalase deficient mice are described as being phenotypically normal [10]. Nevertheless, more recent epidemiological studies show acatalasemic human subjects to be at an increased risk for developing numerous pathologies including altered lipid and carbohydrate metabolism [11] and type 2 diabetes [12, 13].

Ho and colleagues created global catalase knockout (Cat^-/-) mice by gene targeting of intron 4 and exon 5 [10]. These mice have a diminished rate of removal of hydrogen peroxide and thus are more susceptible to oxidant tissue injury [10]. Catalase-deficient mice, while appearing to be phenotypically normal [10, 11], are more vulnerable to induction of diabetes by alloxan [14] and to renal injury associated with streptozotocin-induced diabetes [13]. This is thought to be caused by accumulating ROS damaging pancreatic beta cells of the islets of Langerhans and thereby reducing insulin secretion [15, 16]. Low levels of protective antioxidants in beta cells [17, 18] would also be a factor contributing to the beta cell vulnerability. Several lines of evidence support a role for hydrogen peroxide in ROS-induced beta cell toxicity. For example, sequestration of hydrogen peroxide by alpha lipoic acid attenuates the toxicity [19, 20], while reduced expression of catalase mRNA and protein may promote toxicity [21-23] by allowing the accumulation of hydrogen peroxide and thereby enhancing ROS-induced cellular damage.

Genes involved in metabolism, such as transcription factors, can also be influenced by cellular hydrogen peroxide. Peroxisome proliferator-activated receptors (PPARs) are transcription factors that play an important role in obesity and insulin sensitivity [24]. Agonists of PPARs, thiazolidinediones (TZDs), have been used to treat diabetic insulin resistance [24, 25]. Although the exact mechanism remains to be established with certainty, TZDs act on PPARγ which alters transcription of numerous genes, ultimately leading to decreased insulin resistance and increased adipocyte lipid storage. Thiazolidinediones also attenuate ROS-induced beta cell damage through a mechanism involving induction of catalase [26]. In addition to influencing the effects of insulin, PPARγ activation increases hepatic lipogenesis, and blood and tissue lipid accumulation [27-29].

The present study uses the Cat^-/- mouse model to examine the role of catalase in lipid dysfunction and metabolic regulation. It was predicted that catalase deficiency would adversely affect lipid trafficking and glucose homeostasis by promoting the accumulation of hydrogen peroxide and thereby eliciting oxidant damage and/or amplified signal transduction pathways.

Materials and Methods

Animals

All procedures involving animals were approved by the institutional Animal Care and Use Committee of the University of Colorado and were performed in accordance with published National Institute of Health guidelines. The catalase knockout (Cat^-/-) mice were generated as previously described in the C57BL/6 background [10] and obtained from Dr. Ho (Wayne State, MI).

Male C57BL/6 wild-type (WT) and Cat^-/- mice were housed in groups (3-5 per group) with water and standard chow (Teklad 2920x: 6.5% fat, 19.1% protein, 47.0% carbohydrate) diet available ad libitum. Mice were weighed weekly and all but one Cat^-/- mouse survived the duration of the study. Necropsy revealed no overt cause of death for the non-surviving mouse. Upon completion of the study, mice were euthanized by CO₂ followed by decapitation. Tissues were excised, weighed and fixed, or flash frozen for biochemical analysis.

Protein expression and carbonylation

Frozen liver tissues were homogenized on ice in lysis buffer containing phosphatase and protease inhibitors. Proteins were prepared for SDS PAGE as previously described [30]. Protein concentrations were determined using a BCA Protein Assay (Pierce, Rockford, IL). Proteins (24 μg) were separated using a 10% SDS gel and subsequently transferred to a PVDF membrane. Membranes were blocked for 60 min with Tris-buffered saline solution containing 1% Tween-20 (TBST) and 5% non-fat dry milk, and probed overnight with rabbit primary antibodies (catalase: 1:5000, Sigma; PPARγ: 1:1000, Novus Biologicals, CO). A horseradish peroxidase-conjugated goat secondary antibody (Jackson Immune Research, PA) was then applied and membranes developed using ECL-Plus Reagent (GE Healthcare, UK). Chemiluminescence was visualized using either film or a Storm 860 scanner (Molecular Dynamics, Sunnyvale, CA). Biotin hydrazide detection of carbonylated proteins was conducted as previously described [31]. Briefly, liver homogenates were treated with 5mM biotin hydrazide (BH; Pierce, Rockford, IL) for 2 h in the dark (corresponding untreated controls were incubated in dark without addition of BH). Proteins were then loaded onto a 10% SDS gel and then transferred to a PVDF membrane. A rabbit anti-biotin antibody (GeneTex, Irvine, CA) was applied and imaged using the Storm 860 scanner (Molecular Dynamics, Sunnyvale, CA). Densitometric analyses were performed using ImageJ software (NIH, USA) and normalized to total protein content (Ponceau).

Biochemical analyses

The liver was removed en bloc from 12-15 mo old mice and flash frozen in liquid nitrogen. One lobe was excised, weighed and homogenized in buffer. Liver triglycerides were measured in a 2:1 chloroform:methanol extract of liver homogenate using a triglyceride-SL kit (Seikesui, Japan). Serum was derived from blood drawn from the inferior vena cava or submandibular face vein. Triglycerides and alanine transferase (ALT) were measured using commercially available kits (triglyceride-SL, Alanine Aminotransferase-SL assay, Seikesui Assay Kit, Japan). Fasting serum insulin was collected from mice at 5 mo old, before the body weights were different and 8 mo old, at which time the weights were significantly increased (Fig. 1C). Serum insulin was determined using a commercially-available kit for mouse insulin (Crystal Chem, IL, USA).

Figure 1. Confirmation of Cat^-/- and changes in oxidative stress, body weight, and tissue weight in Cat^-/-.

A . Liver homogenates from 12 mo old wild-type (WT) and catalase knockout (Cat^-/-) mice were probed with anti-catalase antibody. Representative immunoblots are shown. B. Oxidative stress in liver homogenates was estimated by the extent of protein carbonylation. Carbonylated proteins were identified using an anti-biotin antibody in Western blot analyses (upper figure) and subjected to densitometric analysis (lower figure). Data were normalized to untreated controls and expressed as a percentage of WT. Data are presented as the mean and associated SEM from 3 mice. * P < 0.05. Student's unpaired t-test, compared to WT. C. Body weight was measured in WT and Cat^-/- mice, fed a normal chow (6.5%) diet, over the period of 1.5-8 mo of age. Data are presented as the mean and associated SEM from 6 mice. * P < 0.05, Student's unpaired t test, compared to WT mice of same age. White adipose tissue (WAT) was harvested from 12 mo old Cat^-/- and WT mice and prepared for histological (paraffin embedded and stained with hematoxylin and eosin) analysis. D. Representative images of adipose tissue histology revealed hypertrophic and irregular adipocytes (noted by *) and crown structures (shown by an arrow).

Glucose Tolerance Test (GTT)

Nine mo old WT and Cat^-/- mice were fasted overnight with water available ad libitum. Mice were anesthetized by isofluorane inhalation. The tail was then nicked to allow initial collection (t = 0) of blood from the tail vein for measurement of fasting blood glucose. Fifteen % (w/v) D-glucose, prepared by dissolution in tap water and passage of the solution through a sterile filter, was administered intraperitoneally (1.5 g/kg). Blood was then sampled from the tail vein 15, 30, 60 and 120 min thereafter and glucose levels were determined using a glucometer (OneTouch Ultra, One Touch). The area under the curve (AUC) of the blood glucose concentration–time curve was calculated using Graphpad Prizm V 6.0 (La Jolla, CA).

mRNA determination (RNA-seq)

The liver was excised from 2 mo old mice (n=3) and flash frozen in liquid nitrogen. RNA was isolated the same day using a commercially-available kit (Qiagen, RNeasy kit). RNA-seq was run on two of the samples from each genotype and through biostatistical analysis detailed below. Data presented is combined from two animals. Mice of this age were selected to allow determination of RNA changes that might underlie latter physiological changes.

Quality control

Total RNA quality was determined spectrophotometrically (Nanodrop, Thermo Fisher Scientific Inc., USA) using the A₂₆₀/A₂₈₀ and A₂₆₀/A₂₃₀ ratios. RNA integrity was verified by the relative abundance of 18S and 28S subunits of ribosomal RNA (Agilent 2100 Bioanalyzer).

Library preparation

RNA samples (n=2) were prepared using the llumina TruSeq RNA Library Preparation Kit v2. Messenger RNA was purified from 500ng of total RNA using oligo-dT beads and sheared by incubation at 94°C. Following first-strand synthesis with random primers, second strand synthesis was performed with dUTP for generating strand-specific sequencing libraries. The cDNA library was then end-repaired and A-tailed, adapters were ligated and second-strand digestion was performed using uracil DNA glycosylase. Indexed libraries were quantified by qRT-PCR using a commercially-available kit (KAPA Biosystems, Wilmington, MA). The insert size distribution was determined using the Agilent Bioanalyzer High Sensitivity DNA Chip. Samples with a concentration of ≥ 0.5 ng/μl were used for sequencing.

Flow cell preparation and sequencing

Sample concentrations were normalized to 2 nM and loaded onto rapid or high-output flow cells (Illumina, San Diego, CA) at a concentration that yielded 150-250 million passing filter clusters per lane. Samples were sequenced using 75 bp single or paired-end sequencing (Illumina HiSeq 2500) according to Illumina protocols. Data generated during sequencing runs were transferred to the Yale Center for Genomic Analysis (YCGA) high-performance computing cluster. A positive control (prepared bacteriophage Phi X library provided by Illumina) was spiked into every lane at a concentration of 0.3% to monitor sequencing quality in real time.

Data analysis and storage

Signal intensities were converted to individual base calls during a run using Real Time Analysis (RTA) software (Illumina). Primary sample analysis de-multiplexing and alignment to the human genome was performed using CASAVA 1.8.2 software suite (Illumina, San Diego, CA). The raw RNA-sequencing reads were first assessed for quality using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). FASTX-toolkit [32] was then used to trim off adaptor sequences and low quality bases. Reads passing the quality control were aligned to the Ensembl annotation of mouse genome reference sequence GRCm38 by TopHat 2 [33] and Bowtie [34] alignment engines. Reads mapped to multiple locations were discarded from further analysis. Raw gene counts (gene expression levels) were established by HTSeq package [35] with the default union-counting mode, and normalized using the TMM method in edgeR package [36]. A negative binomial generalized linear model implemented in edgeR [37-40] was used to determine differentially-expressed genes with multiple experimental factors accounted for. Genes from Cat^-/- mice with fold-change (> 2 or < 0.5) and false discovery rate (FDR)-controlled (P < 0.05) were considered differentially expressed relative to WT mice.

Histology

Liver and pancreas were excised en bloc from 12-15 mo old mice and weighed. Visceral and subcutaneous white adipose tissue (WAT) was crudely collected from each mouse and weighed. Sections of liver lobes, and adipose and muscle tissue, and the entire pancreas were collected, fixed in 10% neutral buffered formalin and embedded in paraffin for histological examination. Slides were cut and prepared for hematoxylin and eosin (H&E) staining by the UCD Research Histology Shared Resource. Brunt scoring for liver damage was adapted for use in the mouse as previously described [41, 42]. Briefly, the extent of steatosis, the presence of inflammatory cells and foci, features of hepatocyte injury, markers of tissue response, and Picrosirius Red-detectable collagen were determined by cell morphology and staining by a trained histopathologist; the extent of each was quantitated by the histopathologist blinded to the origin of the slides. Histologic images were captured on an Olympus BX51 microscope equipped with a 4MP Macrofire digital camera (Optronics) using the PictureFrame Application 2.3 (Optronics). Cross-polarized light was used to enhance visualization of Picrosirius Red-stained images as previously described. Briefly, slides were imaged with polarized light in a tiling fashion across the whole tissue and analysis of the images by the 3I program slidebook (3I, Denver, Colorado). Islet areas (% of WT) were determined from H&E-stained histological images using ImageJ software (NIH, USA).

Statistical analyses

AUC differences were identified by Student's unpaired t-test or two-way ANOVA. Biochemical and histological comparisons between Cat^-/- and WT mice were performed using a Student's unpaired t-test. Body weight differences were determined by two-way ANOVA with differences at each time point being analyzed using a Student's unpaired t-test. Statistical comparisons were conducted using Graphpad Prizm v 6.0 (La Jolla, CA). Statistical significance was defined as P < 0.05.

Results

Catalase knockout, oxidative stress

The Cat^-/- knockout genotype was confirmed by tail snip genotyping (data not shown), the absence of hepatic expression of catalase protein (Fig. 1A) and RNA-seq analyses (Table 1). RTPCR was not conducted to verify RNA-seq analysis results due to altered internal control genes in Cat^-/- mice. While mRNA for catalase was shown to be repressed, the mRNA for other peroxide-metabolizing enzymes, including glutathione peroxidase (Gpx) and peroxiredoxin (Prx), were unchanged in the Cat^-/- mice (data not shown). Protein carbonylation, a marker of the action of ROS on proteins, was observed in Cat^-/- mice (Fig. 1B).

Table 1. RNA transcripts most altered in catalase knockout (Cat^-/-) mice.

Genome-wide gene transcription profiling was obtained by RNA-sequencing of hepatic RNA obtained from two 2 mo old mice. RNA-seq was run on two samples from each genotype then, through biostatistical analysis. Data presented is combined from two animals. Genes with false discovery rate (FDR)-controlled P < 0.05 and fold-change (FC, >2 or <0.5) in Cat^-/- samples relative to WT samples were arbitrarily defined as being differentially-expressed. The transcripts in Cat^-/- mice that differed most from WT mice is ranked in order of statistical significance. Data were trimmed to include only known transcripts. The five most relevant genes that were up-regulated or down-regulated are listed.

Up-regulated genes
Gene name	Gene	Log2FC	FC	FDR
Cell Death-Inducing DFFA-Like Effector C	Cidec	6.949	124	2.1E-73
Transmembrane protein 28	Tmem28	6.889	119	3.0E-55
Oxysterol Binding Protein-Like 3	Osbpl3	5.053	33	1.6E-65
Complement factor D adipsin	Cfd	4.973	31	3.4E-12
Monoacylglycerol acyltransferase	Mogat1	4.715	26	5.6E-40
Down-regulated genes
Gene name	Gene	log2FC	FC	FDR
Catalase	Cat	-4.507	0.044	1.3E-96
Insulin-Like Growth Factor Binding Protein 1	Igfbp1	-4.493	0.044	1.7E-78
Protein phosphatase 1, regulatory (inhibitor) subunit 3G	Ppp1r3g	-3.947	0.065	4.7E-48
Chemokine (C-C motif) ligand 21A	Ccl21a	-3.153	0.11	8.8E-24
C-type lectin domain family 2, member h	Clec2h	-3.084	0.12	6.2E-24

Catalase knockout mice fed standard chow diet exhibited increased body weight relative to their corresponding WT controls beginning after 6 mo of age (Fig. 1C). In spite of the increased body weight, the WAT weight-to-total body weight ratio in Cat^-/- mice was not different from WT mice (Table 2). Nevertheless, histological analysis revealed adipose tissue from Cat^-/- mice to contain hypertrophic adipocytes as well as an increased number of macrophages surrounding some of the adipocytes (crown-like structures) relative to their WT counterparts (Fig. 1D).

Table 2. Pathophysiological changes in catalase knockout mice.

Biochemical and morphological parameters were measured in tissues obtained from wild-type (WT) and catalase knockout (Cat^-/-) mice. Liver and white adipose weights are expressed as the ratio of tissue weight-to-body weight (BW) as well as individually. Liver tissue was prepared for biochemical (triglyceride, alanine aminotransferase (ALT)) analyses. Data represent the mean and associated SEM from the number of mice indicated in parentheses.

Parameter	Age (mo) a	WT	Cat-/-
Body weight (BW) (g)	12-15	38.2 ± 1.6 (9)	48.6 ± 2.3 * (9)
White adipose tissue weight (g)	12-15	2.7 ± 0.7 (3)	2.7 ± 0.4 (3)
Liver weight (g)	12-15	1.7 ± 0.1 (9)	2.6 ± 0.3 * (9)
White adipose tissue weight (% BW)	12-15	7.0 ± 1.5 (3)	6.1 ± 0.9 (3)
Liver weight (% BW)	12-15	4.3 ± 0.2 (9)	5.7 ± 0.4 * (9)
Serum triglycerides (mg/ml)	12	0.70 ± 0.03 (3)	0.86 ± 0.04 * (3)
Hepatic triglycerides (mM/g tissue)	12	5.3 ± 1.1 (3)	13.0 ± 2.0 * (3)
Alanine aminotransferase (ALT) (U/l)	12	114 ± 20 (3)	362 ± 116 (2)

^aAge of mice in which parameter was measured.

^*P<0.05, Student's unpaired t-test, compared to WT.

Significant hepatic changes were induced by the absence of catalase. The liver weight of Cat^-/- mice was increased relative to WT controls (Table 2). Morphologically, the livers of Cat^-/- mice had a lighter yellow color than WT control (data not shown) and histological analysis revealed lipid accumulation in the livers of Cat^-/- mice (Fig. 2A) consistent with steatosis. Cat^-/- mice also had a higher incidence of more advanced macrosteatosis (10-33%) and microsteatosis (83%) (Table 3). Inflammation (manifesting as foci of inflammatory cells and the presence of lipogranulomas) and liver cell injury (ballooning degeneration) were not observed in the livers of WT mice but were present in at least 33% of the livers of Cat^-/- mice (Table 3). Liver fibrosis was not observed in any animals. Elevated levels of triglycerides were observed in the serum and liver of Cat^-/- mice relative to their WT counterparts (Table 2). A trend for increased serum ALT was also observed in the 12 mo old Cat^-/- mice but it failed to reach significance due to large variability (P = 0.2) (Table 2). Expression of the transcription factor, PPARγ, was increased in the livers of Cat^-/- mice (Fig. 2B).

Figure 2. Influence of catalase knockout on liver tissue.

Liver was extracted from 12 mo old wild-type (WT) and catalase knockout (Cat^-/-) mice and prepared for histological (paraffin embedded and stained with hematoxylin and eosin) analysis. A. Liver histology. Lipid accumulations are identified by arrows (macrosteatosis) and circle (microsteatosis) (representative images shown). Histological orientation shown: portal triad (PT) and central vein (CV) 40x. B. Liver PPARγ expression was measured by immunoblot of liver homogenates (n=3) (representative blot shown; upper panel) and normalized to total protein (Ponceau stain) with densitometric analysis shown in density units (du) (lower panel). Data are expressed as mean ± SEM from 3 mice. * P < 0.05, Student's unpaired t-test, compared to WT.

Table 3. Liver damage associated with catalase knockout mice (shortened).

Male wild-type (WT) and catalase knockout (Cat^-/-) mice were sacrificed at >12mo of age. The liver was harvested, weighed, formalin fixed and then embedded in paraffin blocks and prepared for histological analysis using hematoxylin and eosin staining. Tissue damage was analyzed using the Brunt scoring method adapted for mice [41, 42]. Data represent the proportion of mice in each group meeting the pathological criterion.

Pathology	# mice showing pathology
	WT	Cat-/-
Macrosteatosis a
<10%	6/7	3/6
10-33%	1/7	3/6
>33%	0/6	1/6
Microsteatosis b
any	1/7	5/6
Inflammation c
foci of inflammatory cells	0/7	3/6
lipogranulomas	0/7	3/6
foamy macrophages	0/7	4/6
Liver cell injury d
ballooning degeneration	0/7	2/6
Fibrosis e	0/7	0/6

^aVesicles larger than hepatocyte nucleus.

^bHepatocyte filled with small vesicles, central nucleus.

^cLooked for lobular inflammation, foci of inflammatory cells, lipogranulomas, portal inflammation, foamy macrophages.

^dLooked for ballooning, acidophil bodies, pigmented macrophages, megamitochondria.

^eDetermined by Picrosirius Red staining, imaging with polarized light in a tiling fashion across the whole tissue and analysis of the images by the 3I program Slidebook (3I, Denver, Colorado).

There were a total of 1,102 up-regulated transcripts and 1,351 down-regulated transcripts in the livers of Cat^-/- mice. The five most up-regulated transcripts include Cell Death-Inducing DFFA-Like Effector C (Cidec, 124-fold), Transmembrane protein 28 (Tmem28, 119-fold), Oxysterol Binding Protein-Like 3 (Osbpl3, 33-fold), Complement factor D, (corresponding to human adipsin) (Cfd, 31-fold), and monoacylglycerol acyltransferase (Mogat1, 26-fold). Because of the indication of lipid dysregulation seen in histology and up-regulated transcripts, we evaluated genes previously reported to be involved in obesity [33]. These were found to be differentially expressed in Cat^-/- mice (Table 4). Leptin (Lepr 0.37-fold change) and insulin signaling/sensitivity (Irs2 0.29-fold change) were decreased while Irs3 was increased (3.1-fold change). Transcripts involved in free fatty acid metabolism, specifically Adra1a and Lipg (3.2- and 2.4-fold change, respectively), were increased while Adrb2 and Lpl were decreased (0.4- and 0.3-fold, respectively). Transcripts involved in lipid biosynthesis, such as Fasn (2.7-fold change) and Scd1-4 (2.0-2.5-fold change), were also increased in Cat^-/- mice. Transcripts involved in lipid transport, specifically importing genes Cd36 (4.9-fold change) and Slc2a4 (4.4-fold change), were increased.

Table 4. Differentially-expressed liver transcripts in catalase knockout mice.

Genes involved in metabolic dysfunction were identified through a literature search. Genome-wide gene transcription profiling was obtained by RNA-sequencing of hepatic RNA from 2 mo old mice (n=3). Genes with false discovery rate (FDR)-controlled P < 0.05 and fold-change (FC) >2 or < 0.5 in Cat^-/- samples relative to WT samples were arbitrarily defined as being differentially-expressed.

Category	Gene	Gene symbol	Log2FC	FC	FDR
Obesity	Leptin receptor	Lepr	-1.431	0.37	4.62E-08
Free fatty acid metabolism	Beta2 adrenoceptors	Adrb2	-1.289	0.41	2.28E-03
	alpha1A adrenoceptors	Adra1a	1.657	3.2	6.36E-11
	Lipases	Lipg	1.251	2.4	1.25E-05
		Lpl	-1.749	0.30	3.88E-15
Insulin sensitivity	Peroxisome proliferator-activated receptor gamma (PPARγ)	Pparg	2.273	4.8	1.66E-15
	Insulin receptor substrate	Irs 2	-1.804	0.29	7.81E-15
		Irs 3	1.633	3.1	2.90E-02
Lipid metabolism	Fatty acid translocase	Cd36	2.279	4.9	1.48E-18
	Apolipoprotein A 4	Apoa4	1.614	3.1	1.42E-09
Biosynthesis	3-Hydroxy-3-Methylglutaryl-CoA Reductase	Hmgcr	-1.122	0.46	1.77E-04
	Fatty Acid Synthase	Fasn	1.423	2.7	3.71E-12
	Stearoyl-CoA desaturase-1	Scd1	1.332	2.5	3.75E-10
	Stearoyl-CoA desaturase-2	Scd2	1.213	2.3	5.52E-07
	Stearoyl-CoA desaturase-3	Scd3	1.104	2.1	5.88E-07
	Stearoyl-CoA desaturase-4	Scd4	1.254	2.4	4.38E-09
Growth and differentiation	Ribosomal Protein S6 Kinase	Rps6ka1	1.023	2.0	3.32E-06
Glucose transport	Solute Carrier Family 2 (Facilitated Glucose Transporter), Member 4 (glut4)	Slc2a4	2.128	4.4	2.40E-07

Fasting blood glucose levels in Cat^-/- mice were comparable to WT mice (Table 5). Nevertheless, Cat^-/- mice showed impaired tolerance to a bolus dose of glucose such that higher blood glucose levels than WT mice occurred at 15 and 30 min, and with a trend at 60 min (P = 0.06) (Fig. 3A). Relatedly, the AUC for the glucose tolerance test curve was greater for Cat^-/- mice than WT mice (Fig. 3B). Histological analysis of the pancreas showed a larger islet area in Cat^-/- mice than in WT mice (Fig. 3C, D). Higher fasting insulin was measured in the serum of the Cat^-/- mice (Table 5). Intermuscular adipose tissue was found in the gastrocnemius skeletal muscle of Cat^-/- mice but not WT mice (Fig. 3E).

Table 5. Glucose and insulin changes in catalase knockout mice.

Fasting blood glucose and serum insulin were measured in wild-type (WT) and catalase knockout (Cat^-/-) mice. Insulin levels were measured in 5 (before weight differences) and 8 (after weight differences) mo old WT and Cat^-/- mice. Data represent the mean and associated SEM from the number of mice indicated in parentheses.

Parameter	Age (mo) a	WT	Cat-/-
Fasting blood glucose (mg/dl)	12	157 ± 8 (3)	135 ± 16 (5)
Fasting serum insulin (pmol/l)	5	65.9 ± 0.3 (4)	67.1 ± 0.9 (4)
	8	69 ± 13 (3)	475 ± 158 * (3)

^aAge of mice in which parameter was measured.

^*P<0.05, Student's unpaired t-test, compared to WT.

Figure 3. Changes in glucose tolerance, fasting insulin, and pancreatic and muscle histology in catalase knockout mice.

A A glucose tolerance test (GTT) was performed in 9 mo old wild-type (WT) and catalase knockout (Cat^-/-) mice (n=3-5). B. The area under the curve (AUC) for the GTT was calculated for mice of both genotypes (n=3-5). Data are presented as mean and associated SEM. * P < 0.05, Student's unpaired t-test, compared to WT. Tissues were collected from 12-15 mo old mice, embedded in paraffin, sectioned and stained with hematoxylin and eosin. Images of pancreas (C) and gastrocnemius skeletal muscle (E) are shown at 40x magnification with insets at 200x magnification. Arrows indicate islets (C) or intermuscular adipose tissue (E). The areas of the islets (n=3) were determined using Image J software and are shown in (D). * P < 0.05, Student's unpaired t-test, compared to WT.

Discussion

The genesis of this study emanated from our observation that catalase knockout (Cat^-/-) mice gained weight at a faster rate later in their life (i.e., after 6 mo of age) while maintained on a normal low fat chow diet. For example, at 8 mo of age, Cat^-/- mice weighed 23% more than wild-type mice. This is less than the weight gain observed in two established mouse models of obesity, ob/ob and db/db, in which the increases are approximately 200 and 60% (respectively) more than wild-type [43-45].

Initial experiments revealed the weight gain in Cat^-/- mice to be unrelated to the accumulation of white adipose tissue. However, the increase in intermuscular adipose tissue observed in Cat^-/- mice may have contributed to the weight gain. Although the amount of white adipose tissue did not change in Cat^-/- mice, the nature of this tissue differed from wild-type mice in containing hypertrophic and irregular adipocytes, and increased numbers of crown-like structures. Such ultrastructural changes, when taken together with the observed elevated levels of serum triglycerides, suggest that an absence of catalase promotes alterations in lipid mobilization and utilization that favor excess circulating lipids. This is of particular interest as the mice were fed a standard chow diet which is low fat, but high carbohydrate. Notably, crown-like structures, reflective of macrophages surrounding dying or dead adipocytes, are commonly observed in the adipose tissue of ob/ob and db/db mice and are thought to be associated with insulin resistance [46]. A recent study showed higher levels of plasma free fatty acids and triglycerides in Cat^-/- mice [47], reaffirming a potential role for catalase in lipid mobilization. The same study also revealed increased numbers of crown-like structures in white adipose tissue in Cat^-/- mice, but only under conditions of a high fat diet.

Catalase deficiency significantly impacted the liver and its function. Specifically, the liver of Cat^-/- mice weighed more, contained higher levels of triglycerides, and exhibited histopathological changes associated with hepatosteatosis, including microsteatosis and macrosteatosis. The hepatocyte injury (ballooning degeneration) and inflammation is indicative of the development of steatohepatitis in the Cat^-/- mice, i.e., a progression from non-alcoholic fatty liver disease (NAFLD) to non-alcoholic steatohepatitis (NASH). These data lend further support to the notion that loss of catalase adversely affects lipid homeostasis. Hepatic mRNA expression analyses demonstrated changes that would be predicted to affect fatty acid handling by the body, including increases in adipsin (Cfd), cell death inducing DFFA-like effector C (Cidec), oxysterol binding protein-like 3 (Osbpl3), protein phosphatase 1 regulatory subunit 3G (Ppp1r3g), monoacylglycerol acyltransferase (Mogat1), fatty acid translocase (Cd36), fatty acid synthase (Fasn), endothelial lipase (Lipg), stearoyl-coA desaturases (Scd1, Scd2, Scd3, Scd4), alpha_1A adrenoceptors (Adra1a) and decreases in mRNA expression of lipoprotein lipase (Lpl), beta₂ adrenoceptors (Adrb2) and 3-hydroxy-3-methylglutaryl-coA reductase (Hmgcr).

Macrophage accumulation in white adipose tissue has been correlated with insulin resistance and hyperinsulinemia in mice [48]. Under fasting conditions, the Cat^-/- mice exhibited elevated levels of insulin (hyperinsulinemia) that were comparable to those observed previously in db/db mice [49]. Histological analysis of the pancreas of Cat^-/- mice revealed larger islets of Langerhans, a pathophysiological change previously documented in obese (ob/ob) and diabetic (db/db) mice [50]. Increases in beta cell mass and higher levels of fasting plasma insulin have also been noted in obese h-IAPP transgenic mice [51]. Fasting blood glucose levels in Cat^-/- mice were not different from wild-type mice, suggesting that the Cat^-/- mice were not diabetic but were more resistant to insulin than their wild-type counterparts. Intermuscular adipose tissue has been shown to positively correlate to insulin resistance and higher fasting insulin [52, 53]. Increases in pancreatic beta cell mass and insulin secretion have been proposed to represent pathophysiological adaptions to chronic insulin resistance and are similar what is thought to occur during metabolic syndrome [51] in humans.

The reduced capacity of Cat^-/- mice to buffer glucose (manifesting as elevated peak blood glucose levels and an increased area under the glucose tolerance test curve in response to a bolus administration) indicates that the absence of catalase adversely affects glucose homeostasis and renders mice in a pre-diabetic state. Changes in transcriptional expression for facilitated glucose transporter 4 (Glut4) and protein phosphatase 1 regulatory subunit 3G would be expected to contribute to these effects. These results in Cat^-/- mice align with those in hypocatalasemic human subjects with diminished catalase activity who have a higher risk (≈13%) of developing diabetes [54] and in acatalasemic mutant mice that are more vulnerable to induction of diabetes by alloxan [14]. RNA-Seq experiments in liver tissue from Cat^-/- mice revealed changes that would influence the actions of insulin, including down-regulation of insulin-like growth factor binding protein 1 (Igfbp1) and insulin receptor substrate 2 (Irs2), and up-regulation of insulin receptor substrate 3 (Irs3). IRS2 has previously been shown to control glucose homeostasis and be involved in insulin resistance [55] as Irs2^-/- mice develop type 2 diabetes despite compensatory changes in IRS3 [56]. Insulin inhibits insulin-like growth factor binding protein 1 gene expression in the liver [57]. Accordingly, the elevated levels of serum insulin in the Cat^-/- mice would be predicted to decrease Igfbp1 mRNA.

Peroxisome proliferator-activated receptors (PPARs) are transcription factors that play an important role in obesity and insulin sensitivity [24]. Agonists of PPARγ have been used clinically to reduce insulin resistance and treat type 2 diabetes [58]; A common contention is that they do so by increasing the quantity of adipose tissue which then acts as a “sink” for the excess glucose [59]. Liver transcripts for and protein levels of PPARγ were elevated in Cat^-/- mice. PPARγ promotes elevated hepatic triglyceride levels and fat accumulation, and steatosis by increasing expression of genes that promote lipogenesis, lipid transport and beta-oxidation, such as adipsin, cd36, fsp27 (cidec) and mogat1 [27, 60-62]. Hence, many of the transcript changes observed in the liver of Cat^-/- mice could involve PPARγ.

The effects of catalase absence on lipid and glucose homeostasis would be predicted to relate to the accumulation of hydrogen peroxide and subsequent amplification of hydrogen peroxide-dependent signal transduction pathways or oxidant damage. Ho and colleagues, the originators of the mice used in the present study, showed increased hydrogen peroxide production by tissue from Cat^-/- mice [10]. Park and colleagues demonstrated elevated nitrotyrosine expression in epididymal white adipose tissue of Cat^-/- mice [47]. Reactive oxygen species are abundantly produced in obesity and metabolic syndrome [5] and adipose tissue proteins of obese patients have been found to be more carbonylated (a marker of oxidative stress) [63]. Elevated levels of carbonylated proteins in the liver of Cat^-/- mice indicate oxidative stress was occurring in the absence of catalase.

Hydrogen peroxide plays an important role in insulin-producing cells, insulin signaling, and insulin resistance. It alters adipokine signaling pathways, and genes involved in metabolic function, such as transcription factors, inflammation markers, and insulin receptor signaling molecules, and can be toxic to beta cells of the islets of Langerhans [17, 64-67] Beta cells have relatively low levels of antioxidant enzymes [68] and catalase is important for protecting these cells from the toxic effects of ROS [5]. Consistent with such a protective role is the demonstration that acatalasemic mice are more vulnerable to induction of diabetes by alloxan, a treatment that induces ROS formation in beta cells [14]. In the present study, the absence of catalase did not appear to be toxic to beta cells because the islets of Langerhans were increased (in size and number) in Cat^-/- mice, as were serum levels of insulin.

The present study has demonstrated the absence of catalase leads to insulin resistance in mice. Twin studies in humans have strongly implicated diet as a root cause of insulin resistance [69]. For example, the consumption of fructose, a commonly-used sweetener and food additive, is strongly linked to insulin resistance [70]. This same sugar has been shown to lower hepatic catalase mRNA and activity in experimental animals [21, 71]. These data support the intriguing possibility that modulation of catalase by fructose could contribute to the development of insulin resistance induced by this sugar.

In summary, the results of the present study show increased oxidative stress and increased body weight in Cat^-/- mice. There is excess lipid accumulation in adipocytes and hepatocytes, as well as in skeletal muscle. Excess lipid in the liver is accompanied by inflammation and damage, and transcriptional changes in genes involved in lipid synthesis (Fasn) and transport (Cd36), and insulin sensitivity (Irs2, Pparg). Glucose intolerance occurred in Cat^-/- mice with pancreatic changes manifesting as more numerous enlarged islets of Langerhans and higher fasting serum insulin levels. Together, these results provide compelling evidence for catalase (or its products) being involved in insulin resistance and development of a pre-diabetic state.