Sustained Antidiabetic Effects of a Berberine-Containing Chinese Herbal Medicine Through Regulation of Hepatic Gene Expression

The study was approved by the animal experimentation ethics committee of The Chinese University of Hong Kong. The animal experiments were conducted in accordance with the Animals (Control of Experiments) Ordinance of the Department of Health of the Hong Kong Special Administrative Region.

On the basis of a literature search, C. chinensis Franch, Astragalus membranaceus, and Lonicera japonica are popular herbs that have been used in diabetes for thousands of years. One of the most popular formulae, referred to as JCU (Supplementary Fig. 1), contains Rhizoma coptidis (JCU-1), Radix astragali (JCU-2), and Flos lonicerae (JCU-3) at a ratio of 1:1.5:6. JCU-1, known as huanglian in Chinese pinyin, is the dried root and stem of C. chinensis Franch; JCU-2, known as huangqi in Chinese pinyin, is the dried root of A. membranaceus (Fisch.) Bunge var. mongholicus (Bunge) Hsiao; and JCU-3, known as jinyinhua in Chinese pinyin, is the dried flower of L. japonica Thunb. All three herbs were authenticated with high-performance liquid chromatography (HPLC) analysis using phytochemical markers (17), which demonstrated 3.6% berberine in JCU-1, 0.040% astragaloside IV in JCU-2, and 1.5% chlorogenic acid and 0.10% luteolin in JCU-3.

According to Chinese pharmacopoeia, the raw ingredients of JCU include 60 g JCU-1, 90 g JCU-2, and 360 g JCU-3. We used validated methods of aqueous extraction to reduce the volume of JCU-2 and JCU-3 (SIPO patent no: 200510097793.4 200610137944.9 200610016490.X). The final formula was made up of 60 g crude JCU-1 in fine powder, 60 g crude JCU-3 in fine powder admixed with 29.73 g aqueous extracts from 90 g crude JCU-2, and 135.15 g aqueous extracts from 300 g crude JCU-3.

The HPLC analysis was performed on a serial system (Agilent 1100; Agilent Technologies, Santa Clara, CA) with the Agilent G1365D MWD using validated methods in terms of linearity, limits of detection, quantification, reproducibility, and recovery. Using chemical references from the National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China) with 98% of purity, these assays were confirmed to be accurate, reproducible, and sensitive.

In brief, samples were separated on a reverse-phase analytical column (Zorbax XDB-C8, 4.6 × 150 mm, 5μm; Agilent Technologies). The mobile phase was acetonitrile and 0.1% aqueous acetic acid, and the flow rate was 0.1 mL/min. Chemical profiles of the composite formula JCU and its water extract were analyzed by HPLC. The berberine concentration was 0.42% in JCU powder and 1.39% in JCU-1 water extract.

Five parental pairs of Zucker rats were obtained from Monash University Animal Services (Melbourne, Australia) and transferred to the Laboratory Animal Services Center at The Chinese University of Hong Kong. The ZDF rats inherit obesity through a simple recessive gene (fa) of leptin receptor defect (18). The animals had free access to water and were fed a standard laboratory rat diet (5001 Rodent Diet; LabDiet, St. Louis, MO) containing 12% of energy from fat (4.5% fat by weight), 60% of energy from carbohydrate, and 28% of energy from protein (23% protein by weight).

We conducted three sets of animal experiments to study the effects of JCU on blood glucose levels. In a single-dose study (Supplementary Fig. 2A), nine male ZDF rats (aged 5 months) were given a single oral dose of JCU (4.0 g/kg) and observed up to 1 week posttreatment. Fasting and 2-h blood glucose levels during oral glucose tolerance test (OGTT) and insulin tolerance test (ITT) were measured on days 0, 1, and 7.

Thirteen male ZDF rats (aged 10 months) orally received either JCU (4.0 g/kg, n = 7) or vehicle (water) (10 mL/kg, n = 6) for 2 weeks and were observed up to 2 months posttreatment. OGTT, ITT, and blood tests for renal, liver, and lipid parameters were performed at week 0 and 2 with repeat OGTT and ITT at 1 and 2 months posttreatment when the animals were killed for histological examination of liver and pancreatic tissues (Supplementary Fig. 2B). Western blotting, microarray, and quantitative PCR analysis were applied to the liver tissues to examine expression profiles.

Twelve male ZDF rats (aged 7 months) were matched for body weight (range, 430–470 g) and treated with either JCU (4.0 g/kg, n = 7) or vehicle (10 mL/kg, n = 5) followed by a 12-month posttreatment observation period. OGTT, ITT, and biochemical tests, including fasting insulin, were performed at month 0 and month 1. During the posttreatment period, OGTT was performed monthly for 12 months followed by killing of animals for histological examination of liver tissues (Supplementary Fig. 2C).

Serum insulin concentrations were measured using rat insulin ELISA kits (Mercodia, Uppsala, Sweden). Blood glucose during OGTT (glucose powder, 2.5 g/kg) and ITT (0.5 units/kg i.p.) were measured by a blood glucose meter (Onetouch Ultra; LifeScan, Milpitas, CA). Lipid, liver, and renal function parameters were measured using methods previously reported (19).

Tissue specimens from pancreas and liver were obtained and fixed in neutral formaldehyde and embedded in paraffin. For immunofluorescence microscopy, pancreas slides were stained with mouse anti-insulin antibody (1:100 dilution; Zymed Laboratories, South San Francisco, CA) and counterstained with DAPI.

Fresh liver proteins were obtained from male ZDF rats at 2 months after 2-week treatment with JCU (n = 7) or vehicle (n = 6). As previously reported (20), Western blot was performed to detect signaling molecules using the following primary antibodies: AMPK-α, phosphorylated (p)AMPK, acetyl-CoA carboxylase (ACC), pACC, Akt, pAkt, hydroxymethylglutaryl CoA reductase (HMGCR), sterol regulatory element–binding protein (SREBP)1, SREBP2, cytochrome c oxidase (CCO), and β-actin (all at dilution 1:1,000 and obtained from Cell Signaling Technology, Danvers, MA). Signals were then quantitated by densitometry and corrected for the β-actin signal.

Fresh liver specimens were collected from the ZDF rats at 2 months after 2-week treatment with either JCU (n = 7) or vehicle (n = 6). The miRNA samples from liver specimens were extracted using mirVana miRNA Isolation Kit (Applied Biosystems Inc., Carlsbad, CA) and thereafter reverse transcripted to cDNA using the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems Inc.). The miR-U87 was used as internal quantitative control. The relative expression levels of miRNAs in the vehicle-treated ZDF rats were normalized as 1.

Fresh liver specimens were obtained from male wild-type Zucker normal rats (n = 3) and ZDF rats at 2 months after 2-week treatment with either JCU (n = 3) or vehicle (n = 3). Total RNA from rats’ livers was extracted using RNeasy Mini Kit with DNase digestion (QIAGEN, Valencia, CA). An equal amount of total RNA prepared from each animal within the same treatment group was pooled for analysis. The total RNA sample was sent to the Li Ka Shing Institute of Health Sciences Core Laboratory for reverse transcription, labeling, microarray hybridization, washing, and scanning using the Affymetrix GeneChip Gene 1.0 ST Array System. GeneSpring GX2 10.0 software was used to reconstruct global normalization and gene network with ≥1.5-fold change as cutoff point for differential expression.

The mRNA sample was thereafter estimated via quantitative real-time PCR using TaqMan 20× Universal PCR Master Mix (Applied Biosystems Inc.). For mRNA amplification, primers were synthesized by Invitrogen (Cergy-Pontoise, France) and listed as follows: 1) insulin-like growth factor binding protein (IGFBP)1, forward 5′-GAAGCTTTTCTCATCTCCATACATGT-3′; reverse 5′-AAGGCCCCTACCTCAGAC TGA-3′, and 2) cytochrome P450 family 7 (CYP7a1), forward 5′-ATGACACGCTCTCCACCTTTG A-3′; reverse, 5′-AGCTCTTGGCCAGCACTCTGT-3′. Relative mRNA expression was quantified by the comparative cycle threshold (Ct) method and expressed as 2^−ΔΔCt.

Data are expressed as mean ± SD. SPSS 10.0.7 for Windows 2000 (SPSS, Chicago, IL) was used to perform statistical analysis. Student t test and paired t test were used to detect between-group and within-group difference, respectively. A two-tailed P value <0.05 was statistically significant.

Since decoction is the traditional method of preparing TCM, bioactive compounds are more likely to reside in water extracts. HPLC analysis indicated that the 5- and 16-month-old preparations of JCU water extracts had almost identical chemical peaks with comparable berberine concentration of 11.86 and 12.83%, respectively (Supplementary Fig. 1C). These data suggest that there was little chemical decomposition in the JCU formula during a 1.5-year study period using berberine as a reference.

The fa/fa genetic defect for the leptin receptor (OBR) was detected by PCR (Supplementary Fig. 3A). Genotyping of 35 obese and 548 lean rats, aged between 4 and 12 weeks, confirmed that the obese animals were all homozygous for the mutation (fa/fa), whereas the lean rats were either heterozygous (fa/FA, lean) or homozygous wild type (FA/FA, normal). Compared with male Zucker normal and lean rats, male ZDF rats exhibited overt obesity (Supplementary Fig. 3B and C), glucose intolerance, and insulin resistance from 12 weeks onward (Supplementary Fig. 3D and E) and developed overt diabetes from 18 weeks onward (Supplementary Fig. 3F). In this study, normal blood glucose was defined as a fasting blood glucose <5.6 mmol/L and 2-h OGTT blood glucose <7.8 mmol/L. Impaired fasting glycemia was defined as a fasting blood glucose of 6.1–6.9 mmol/L, impaired glucose tolerance was defined as 2-h OGTT blood glucose of 7.8–11.0 mmol/L, and diabetes was defined as 2-h postload glucose ≥11.0 mmol/L during OGTT. We used male ZDF rats of at least age 20 weeks with diabetes and insulin resistance in all experiments.

We validated these animals by treating eight male 20-week-old ZDF rats with metformin (single dose, 50 mg/kg) for 1 week (Supplementary Fig. 3G), which reduced posttreatment blood glucose levels during OGTT (paired t test, all P ≤ 0.021). The area under the curve (AUC) of blood glucose fell from 25.5 ± 6.0 to 17.2 ± 1.7 mmol/L (P = 0.036), whereas body weight increased from 447.8 ± 41.3 to 454.4 ± 39.2 g (paired t test, P = 0.004).

In the single-dose study, JCU treatment improved glucose tolerance (Fig. 1A and B) and insulin sensitivity (Fig. 1C and D) in nine male ZDF rats, with reduced AUC for blood glucose during OGTT (Fig. 1B) and ITT (Fig. 1D). The blood glucose–lowering effects persisted for at least 1 week after the single-dose intervention was discontinued (Fig. 1B–D). During the 1-week posttreatment period, eight of nine rats treated with a single dose of JCU had 25% sustained reduction of AUC blood glucose during OGTT (paired two-tailed t test, P < 0.001).

Prolonged treatment with JCU for 2 weeks and 1 month confirmed these sustained beneficial effects. During 2-week treatment, a significant blood glucose–lowering effect was observed 2 h after the oral gavage of JCU and sustained during an extended 3-h OGTT (Fig. 1E and F) and 2-h ITT (Fig. 1G and H). The reduction in AUC of blood glucose during OGTT and ITT persisted for 2 months after stopping 2-week treatment (Fig. 1I and J).

During the 1-month treatment period, blood glucose levels were lower in JCU-treated ZDF rats during OGTT and ITT than in vehicle-treated animals (Fig. 1K and L and Supplementary Table 1). After treatment discontinuation, fasting and 2-h blood glucose levels during OGTT were monitored monthly for up to 390 days (Fig. 1K and L). Fasting (Fig. 1K) and 2-h (Fig. 1L) blood glucose levels were lower in the JCU-treated than in the vehicle-treated group throughout the observation period. By day 360, three of five control rats and one of seven rats in the JCU-treated group died. In the surviving rats, all six JCU-treated animals remained nondiabetic with a fasting blood glucose <7 mmol/L and 2-h blood glucose <11 mmol/L during OGTT, compared with only one of the two surviving rats in the vehicle-treated group without diabetes.

After 2-week treatment, blood triglyceride decreased in the JCU-treated rats and increased in the vehicle-treated rats with significant between-group difference. Blood creatinine and urea levels tended to decrease in the JCU-treated rats, albeit not significantly. In the 1-month experiment, these trend differences persisted but did not reach statistical significance. In both the 2-week and 1-month experiments, treatment with JCU significantly decreased liver enzyme alanine aminotransferase (ALT) levels by ∼50% (Table 1). After 1-month treatment, despite a lower fasting blood glucose, fasting serum insulin concentrations were similar between the JCU-treated (43.9 ± 22.3 pmol/L) and control (37.8 ± 9.1 pmol/L) groups (P = 0.600).

Figure 2A–D shows the histopathological changes of liver specimens obtained at 2 months (Fig. 2A and B) and 12 months (Fig. 2C and D) after discontinuation of 2-week and 1-month treatment, respectively. In ZDF rats treated with JCU for 2 weeks (Fig. 2A) and 1 month (Fig. 2C), there was normalization of liver tissues. In the group treated with vehicle for 2 weeks (Fig. 2B) and 1 month (Fig. 2D), there were foci of fatty change and diffuse hepatocyte ballooning degeneration.

Figure 2E–H shows the histopathological features on hematoxylin-eosin staining (Fig. 2E and F) and immunostaining of pancreatic islets identified by insulin (Fig. 2G and H) in animals killed at 12 months after 1-month treatment. Both vehicle-treated and JCU-treated animals showed similar islet cytoarchitecture (Fig. 2E and F) and insulin reactivity (Fig. 2G and H).

After 2-week treatment, rats were killed at 2 months for protein studies of the liver sections. Compared with the vehicle-treated rats, JCU-treated animals showed upregulated hepatic expression levels of pAMPK and pAkt (Fig. 3A), with increased pAMPK-to-AMPK ratio and pAkt-to-AKT ratio compatible with persistent activation of AMPK and insulin signaling pathway, despite treatment discontinuation for 2 months. The JCU treatment also stimulated expression level of the mitochondrial CCO (Fig. 3B) and repressed the protein expression levels of pACC, HMGCR, and SREBP1, consistent with activation of hepatic AMPK signaling pathway and a switch from ATP consumption (e.g., fatty acid and cholesterol synthesis) to ATP production (e.g., fatty acid and glucose oxidation) pathways.

We examined a panel of seven miRNA markers implicated in hepatic insulin action (13) in the JCU- or vehicle-treated rats at 2 months after 2-week experiment. Compared with the vehicle-treated rats, the JCU-treated animals had decreased hepatic expression levels of six miRNA markers (miR-1, miR-21, miR-29a, miR-29b, miR-122, and miR-150), reaching significance for miR-29b with 50% reduction, and a 1.5-fold increase in miR-29c expression (Fig. 3C).

We applied a microarray assay that detected 27,342 hepatic genes to total RNA samples extracted from fresh liver specimens of wild-type Zucker normal rats (n = 3) and ZDF rats killed at 2 months after 2-week treatment with either JCU (n = 3) or vehicle (n = 3). Compared with wild-type normal rats, vehicle-treated animals demonstrated a ≥1.5-fold change in 313 genes and a ≥2-fold change in 76 genes. In the 313 genes, 247 genes were annotated, and among them, 49 genes were related to lipid or glucose metabolisms (lipid, 39 genes; fatty acid, 14 genes; cholesterol, 10 genes; bile acid, 5 genes; and glucose, 4 genes). Analysis using a standard DAVID tool (http://david.abcc.ncifcrf.gov) and χ² tests demonstrated that P values of the different interested terms (Supplementary Table 2A) and the union of all genes annotated by these terms (Supplementary Table 2B) related to lipid metabolism were <0.0001. Compared with vehicle, JCU treatment corrected 33 (10.5%) of 313 genes with >1.5-fold change in expression and 6 (7.9%) of 76 genes with >2-fold change in expression. A total of 8 of 33 corrected hepatic genes were related to gluconeogenesis and/or lipid metabolism (Table 2).

Of these 33 genes, 26 annotated genes were subjected to the GeneSpring GX 10.0 software for gene network reconstruction and biological pathway analysis, which suggested a potential central regulating role of IGFBP1 and CYP7a1 (Supplementary Fig. 4A). The former plays a pivotal role in cell cycle, whereas the latter is the rate-limiting enzyme in hepatic efflux of cholesterol through conversion to bile acids. Quantitative real-time PCR confirmed that the JCU treatment significantly upregulated hepatic expression of IGFBP1 (10.9-fold) and CYP7a1 (3.8-fold) compared with vehicle (Supplementary Fig. 4B).

We performed enrichment analysis by examining existing annotations of the 33 genes using the DAVID tool, which showed that 5 genes (Cyp7a1, G6pdx, Me1, Rdh2, and Por) were associated with the term NADP (Benjamini adjusted P = 0.0025). The latter is an essential substrate for fatty acid synthesis and other pathways, including cytochrome P450–related pathways. Among these five genes, four were related to lipid and/or glucose metabolism, whereas Por interacts with different P450 enzymes to complete the electron transfer chain in energy production, resulting in reduced fatty acid synthesis, amelioration of fatty liver, and improved liver function.

Using validated animal models and molecular tools, we reported for the first time the sustained antidiabetic effects of a berberine-containing TCM in ZDF rats through complex regulation of genome expression. The sustained antidiabetic effects of JCU included improvement of hyperglycemia, amelioration of insulin resistance, normalization of elevated liver enzyme, and hepatocyte ballooning degeneration. These sustained metabolic and histopathological changes after treatment discontinuation were associated with changes in gene and protein expression implicated in energy metabolism and cell cycle, in part mediated via changes in miRNA (Fig. 4). On an miRNA level, JCU downregulated miR29-b, which may target Akt and Btg2 to, respectively, influence glucose uptake and CYP7a1 transcription, with increased glucose uptake and lipid catabolism. On a protein level, JCU activated AMPK and Akt with reduced expression of ACC, SREBP1, and HMGCR and increased expression of mitochondrial CCO, favoring lipid oxidation. On an mRNA level, JCU corrected a panel of gene expression to favor lipid catabolism, lipid oxidation, and cell survival, with IGFBP1 and CYP7a1 playing possible linking roles.

In this experiment, we used metformin to validate the ZDF rat model but did not use it as a comparative agent in the mechanistic study since during the 12-month period after 1-month treatment, JCU but not metformin (50 mg/kg body wt) showed sustained favorable effects (data not shown). These observations were in line with TCM clinical practice, in which treatment is usually given for 2 weeks rather than a prolonged period. In our pilot studies (data not shown), the multicomponent formula of JCU stimulated glucose uptake in rat skeletal muscle L6 cells and lowered blood glucose levels in ZDF rats more effectively than the single compound of berberine. Other authors have also reported molecular mechanisms underlying the synergistic effects of TCM formula using cell cultures and animal studies (21). On the basis of these principles, we treated the animals with JCU for 2 weeks and examined the sustained blood glucose–lowering effects using a multipronged strategy.

The sustained, coordinated, and multilayered changes in genome and protein expression by JCU treatment suggested a resetting of energy homeostasis with predominant energy dissipation and catabolism with improved cell survival. These genomic changes were consistent with the phenotypic changes, including weight reduction, improved insulin sensitivity, amelioration of liver fat accumulation and hepatocyte degeneration with increased cellular regeneration. To the best of our knowledge, such complex beneficial effects have not been reported with any conventional Western medicine containing a single chemical except for thiazolidinediones, with beneficial effects on glucolipotoxicity, durability of glycemic control (22), and cancer growth (23).

Apart from animal studies that demonstrate antidiabetic (24) and antihypertensive effects of TCM (25), human studies also report sustained therapeutic effects of TCM and acupuncture in patients with irritable bowel syndrome (26), low back pain (27), and cancer associated with chemotherapy-induced hepatotoxicity (28). Despite these clinical benefits, there have been few comprehensive studies that examine the molecular mechanisms underlying the pluripotent effects of natural medicine.

Several lines of evidence suggested that the biological effects of plant medicines might be mediated through alteration of gene expression. On the basis of this premise, we discovered that 2-week treatment with JCU changed the expression levels of several miRNAs, with repression of upregulated hepatic expression of miR29-b. Several studies report association of insulin resistance with upregulation of miR29 and consequent repression of insulin-stimulated glucose uptake (14). Since the rat miR29-b had 272 predicted target genes, including Akt, changes in this single miRNA might have diverse biological effects (29). On data mining (microRNA.org), we discovered that one of these targets might be Btg2, which is a transcription coregulator. It is interesting that Btg2 was also implied to regulate transcription of CYP7a1, a key enzyme for conversion from cholesterol to bile acids, which was also upregulated by JCU. Thus, apart from modulating Akt expression, miR29-b might interact with Btg2 to increase CYP7a1 expression to promote cholesterol catabolism (Supplementary Fig. 4A).

We further discovered that 33 genes, mainly implicated in energy metabolism and cell cycle, were changed by JCU treatment. Gene network analysis suggested the linking roles of IGFBP1 and CYP7a1 in a molecular network of genes that regulate cholesterol catabolism, bile acid biosynthesis, and cell cycle. These inferred gene-gene networks are compatible with changes in hepatic genes implicated in gluconeogenesis, lipogenesis, cytochrome P450 pathway, and glucose utilization in ZDF rats with prediabetes, diabetes, and late-stage diabetes (30).

The insulin/IGF-I/IGFBP pathway constitutes a complex network to modulate energy metabolism and cellular growth (31), in part mediated through the PI3K/Akt/mammalian target of rapamycin pathway (32–34). In agreement with other reports (35), our vehicle-treated ZDF rats had reduced hepatic IGFBP1 mRNA, which was normalized after 2-week treatment with JCU. Other researchers have reported spontaneous apoptosis in livers of IGFBP1-deficient mice (36). In humans, low fasting serum IGFBP1 concentration was associated with increased liver fat and reduced hepatic insulin sensitivity, independent of obesity (37). In human hepatocytes, endoplasmic reticulum stress induced IGFBP1 expression as an adaptive response (38). Collectively, these data suggest that JCU treatment increased IGFBP1 expression, possibly through increased insulin signaling, and in concert with other genes, to promote liver repair from damage due to fatty degeneration.

In this study, fasting serum insulin concentrations were similar between the JCU-treated and control groups after 1-month treatment. In addition, immunohistological examination demonstrated similar islet size, shape, and number of islets with similar insulin reactive cells between the two groups. Although our study concentrated mainly on hepatic mechanism, since we did not measure glucose-stimulated insulin levels, we could not exclude possible effects of JCU on insulin secretion.

The reduction in SREBP1c protein level might be correlated to the inhibition from AMPK activation since AMPK is a regulator of SREBP1. In mice, activation of SREBP1 and inhibition of hepatic AMPK accounted for ethanol-induced fatty liver (39,40), whereas hepatic activation of AMPK and suppression of SREBP1 protected against hepatic steatosis, hyperlipidemia, and attenuated atherosclerosis in diet-induced insulin resistance (41). In this study, the joint inhibition of SREBP1 and activation of Akt and AMPK may be particularly relevant to the sustained antidiabetic actions of the TCM formula.

Because of small numbers of ZDF rats in each group, we were not able to demonstrate a beneficial effect of JCU on plasma triglyceride level, despite its sustained blood glucose–lowering effects (Table 1). Although Western blot assays showed altered protein levels of ACC, HMGCR, and CCO with JCU treatment, this was not accompanied by significant changes in gene expression. Although altered protein expression level might have occurred posttranscription, the small number of samples and large variability of mRNA expression might reduce the study power to detect differences between expression for each gene. Alternatively, use of a P value cutoff for differential expression and gene network analysis may help identify differentially expressed genes with statistical significance but that do not have a large expression fold change.

Our study also shows that long-term treatment with JCU might prevent body weight gain in ZDF rats (Supplementary Table 1). In db/db mice, berberine treatment reduced body weight and improved triglyceride level (7). Given the importance of abnormal fat metabolism in insulin resistance, it will be useful to further explore the effects of berberine-containing JCU on fat mass and metabolism.

Using scientifically validated animal models and state-of-the-art methodologies, we have confirmed the pluripotent effects of a long-tested TCM in altering gene expression, in part through changes in miRNA, to explain its sustained beneficial effects on glucose metabolism, fatty liver, and cellular repair.

This study was partially supported by a grant from the Hong Kong Jockey Club Charities Trust (JCICM-P2-05 [CUHK]), the Hong Kong Foundation for Research and Development in Diabetes, The Chinese University of Hong Kong Focused Investment, and the Liao Wun Yuk Diabetes Memorial Research Fund. The funding bodies had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

No potential conflicts of interest relevant to this article were reported.

H.-L.Z. designed the experiment, analyzed data, and wrote the manuscript. Y.S. generated and analyzed data and wrote the manuscript. C.-F.Q. and H.-X.X. designed the experiment and generated data. K.Y.Y., R.K.K.L., and H.-M.L. analyzed data. S.K.W.T. contributed to discussion. H.K.T.W., X.Z., J.J.S., L.H., J.G., and L.-Z.L. generated data. P.C.Y.T. designed the experiment. J.C.N.C. designed the experiment, contributed to discussion, and edited the manuscript. H.-L.Z. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors are grateful to Dr. Tony James and Lik Wong Lam at The Chinese University of Hong Kong Laboratory Animal Services Centre for their assistance in breeding the Zucker rats. The authors thank Professor Wai-Yee Chan, director of the School of Biomedical Sciences, Faculty of Medicine, The Chinese University of Hong Kong, for his invaluable advice.