Cold-Inducible Klf9 Regulates Thermogenesis of Brown and Beige Fat

Obesity results from a chronic imbalance between energy intake and energy expenditure, which is closely associated with many diseases, including cardiovascular diseases, type 2 diabetes, and nonalcoholic fatty liver disease. Traditionally, adipocytes are divided into two types: unilocular white adipocytes and brown adipocytes. White adipose tissue (WAT) is essential for triglyceride storage and endocrine signaling, while brown adipose tissue (BAT) dissipates energy to generate heat through uncoupled respiration mediated by Ucp1 (1–3). Recent studies have identified another type of thermogenic adipocytes, namely, beige cells. Beige adipocytes reside with white adipocytes and emerge in response to cold exposure or β-adrenergic receptor agonists (4,5).

PGC1α is a central regulator in brown fat thermogenesis and is highly expressed in BAT. PGC1α expression in BAT is strongly induced by cold stress and β-adrenergic signals, linking the physiological activator of brown fat thermogenesis and the transcriptional machinery in brown adipocytes (6,7). Genetic ablation of Pgc1α results in reduced capacity for cold-induced thermogenesis in vivo and in a blunted response to cAMP signaling in brown fat cells (8,9).

Krüppel-like factor 9 (KLF9) (also called basic transcription element binding protein 1), a member of the Krüppel-like family of zinc-finger domain transcription factors, plays a key role in development (10–12). Interestingly, a human genetic study (genome-wide association study [GWAS]) indicated that Klf9 is associated with BMI (13). However, how KLF9 regulates obesity remains unclear. Furthermore, we recently reported that KLF9 promotes hepatic gluconeogenesis and hyperglycemia (14). Klf9 is ubiquitously expressed in many tissues (15). Nevertheless, whether and how fat KLF9 regulates energy metabolism remains unexplored. In the current study, we reveal the physiological function of KLF9 in adipose tissues.

Studies involving human specimens were approved by the ethics committees of Huashan Hospital, Fudan University, and Shihezi University School of Medicine. Human omental adipose tissues specimens were collected after informed consent was obtained, and the study was approved by the institutional review board of Huashan Hospital (no. 2015-145). All animal experiments were approved by the Institute of Basic Medical Sciences and Peking Union Medical College. All animal experiments were conducted under protocols approved by the Institutional Animal Care Use & Welfare Research Committee, the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences and Peking Union Medical College (ACUC-A01-2014-033 and ACUC2011A02-293).

Male mice were used during this study. Global Klf9 mutant mice were obtained from The Jackson Laboratory (no. 012909). Klf9 transgenic mice were generated at Beijing Biocytogen Co., Ltd. The full-length coding sequence of mouse Klf9 was amplified from hepatic RNA by PCR. The 5.4-kb adiponectin promoter was kindly provided by Dr. Philipp E. Scherer (Department of Internal Medicine, University of Texas Southwestern Medical Center), which was inserted into pBluescript vector. The Klf9 cDNA was inserted into vector containing the 5.4-kb adiponectin promoter. Mouse oocytes were injected with this construct at Beijing Biocytogen Co., Ltd. Global Ucp1 mutant mice were kindly provided by Dr. Yifu Qiu (Institute of Molecular Medicine, Peking University). C57BL/6J, ob/ob, and db/db mice were purchased from the Model Animal Research Center of Nanjing University (Nanjing, China) and were housed and maintained in 12-h light and dark photoperiods. For DIO studies, 4-week-old male mice were fed on a high-fat diet (HFD) (D12492; Research Diets) for 3 months.

The glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs) were performed as previously described (14,16).

Body weight was measured weekly. Body temperature was measured by rectal thermometer. Body composition (fat and lean mass) was determined by MRI (Echomri.Combo-700). For metabolic studies, male mice were housed individually in metabolic cages (Columbus Instruments) with free access to food and water. Oxygen consumption rate (OCR) was monitored for 48 h. Activity monitoring was performed simultaneously with metabolic measurements.

Glucose uptake of brown adipose tissue were determined by positron emission tomography/computed tomography (PET/CT) as previously described (17).

For hematoxylin-eosin (H-E) staining, Oil Red staining, and UCP1 immunohistochemistry, inguinal WAT (iWAT) and epididymal WAT (eWAT) and BAT tissues were treated as previously described (17). For transmission electron microscopy, BAT sections were treated as previously described (17).

Isolation of brown fat stromal vascular fraction (SVF) and differentiation of primary brown preadipocytes were performed as previously described (17), with minor modifications. Briefly, the digested brown adipose tissue was filtered through a 60-mesh nylon screen and centrifuged (1,000 × rpm) for 10 min to collect the preadipocytes.

For determination of cellular oxygen consumption, isolated brown preadipocytes were plated in an XF24-well microplate (Seahorse Bioscience) and differentiated into mature brown adipocytes, followed by OCR measurement at 37°C with an XF24 analyzer (Seahorse Bioscience) in accordance with the manufacturer’s instructions. Oligomycin (2 μmol/L), carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone (FCCP) (2 μmol/L), and rotenone/antimycin (0.5 μmol/L) were delivered to detect the uncoupled respiration, maximal respiration, and nonmitochondrial respiration, respectively. Oxygen consumption of brown fat tissues was measured as previously described (18).

Chromatin immunoprecipitation (ChIP) assay was performed as previously described (14,17). Antibodies specific for KLF9 (ab227920; Abcam) or unspecific IgG (sc-2027; Santa Cruz Biotechnology) was used for ChIP assay. The purified DNA was used to amplify the KLF9 regulatory element on the mouse Pgc1α promoter by a real-time PCR reaction. Primers directed at upstream or downstream of the binding site were used as a negative control. The sequences of primers are shown in Supplementary Table 1.

Immunoblotting was performed with the following primary antibodies: KLF9 (A7196; ABclonal), PGC1α (AB3242; Millipore), UCP1 (ab10983; Abcam), and β-Tubulin (AC021; ABclonal). Target protein bands were quantified with ImageJ software.

Total RNA from either the mouse adipose tissue or the primary adipocytes was extracted using the TRIzol-based method (Invitrogen). Real-time PCR was performed as previously described (14). The sequences of primers are shown in Supplementary Table 1.

Metabolites were measured as previously described (14).

The quantitative data are represented as the mean ± the SEM of three independent experiments. In most of the cases for in vivo experiments in mice, an n = 5 was the minimum amount used. A two-tailed, unpaired Student t test was used for pairwise comparison of genotypes or treatments. One-way ANOVA and two-way ANOVA were used in comparison of three or more groups, as indicated in the figure legends and otherwise. Analysis was performed using Microsoft Excel and/or GraphPad Prism. P < 0.05 was considered significant, as indicated in the figure legends.

The data sets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. No applicable resources were generated during the current study.

To identify the key signaling molecules mediating the beiging and thermogenesis induced by β3-adrenergic agonist in fat, we first performed mRNA microarray analysis of subcutaneous iWAT from normal C57 mice intraperitoneally injected with β3-adrenergic agonist CL 316,243 or saline (control). As expected, preliminary analysis of the microarray data indicated that genes involved in energy metabolism, glucose and lipid metabolism, are induced by CL 316,243 compound (Supplementary Fig. 1A–C). Notably, we observed that KLF9, a transcription factor, is also induced by CL 316,243 (Supplementary Fig. 1A). Our real-time PCR and Western blotting analyses confirmed CL 316,243–mediated increases in Klf9 expression in BAT and iWAT (Fig. 1A and B). We also examined Klf9 expression levels in different fat pads (including BAT, iWAT, and eWAT). Klf9 is highly expressed in BAT, with lower expression levels in iWAT and eWAT (Supplementary Fig. 1D).

Cold exposure also induces adaptive thermogenesis in brown and beige fat via stimulating norepinephrine secretion. For exploration of the potential roles of KLF9 in cold-induced thermogenesis, mice were maintained at 4°C. As a result, Klf9 mRNA and protein levels in both BAT and subcutaneous iWAT were elevated after cold exposure compared with those at room temperature. Moreover, the expression of thermogenic genes including Pgc1α and Ucp1 was also induced by cold exposure (Fig. 1C and D).

We next examined KLF9 expression in BAT and iWAT from db/db and HFD-fed mice. Western blotting analysis showed that KLF9 protein abundance is lower in db/db diabetic mice and mice with HFD-induced obesity than in their respective control db/m and chow diet–fed mice (Fig. 1E and F).

Moreover, a recent human genetic study (GWAS) suggested that a single nucleotide polymorphism in the KLF9 promoter region (rs11142387) is associated with BMI (13). Thus, we tested whether KLF9 expression in adipose tissue samples was associated with BMI. Omental fat tissue biopsies were collected from 39 obese individuals and 11 normal weight individuals. Regression analysis revealed a reverse association between KLF9 mRNA levels in fat and BMI (r² = 0.5139, P < 0.0001) (Fig. 1G). These results implicate Klf9 in obesity and thermogenesis of brown and beige fat, and decreased adipose KLF9 expression may impair systemic energy homeostasis.

We recently reported that hepatic KLF9 stimulates gluconeogenesis via induction of PGC1α (14), and PGC1α has been shown to be a master regulator of brown fat thermogenesis (19). Thus, we hypothesized that KLF9 might regulate fat thermogenesis via PGC1α. For testing this hypothesis, primary adipocytes differentiated from the SVF from the brown fat pads of wild-type (WT) mice were transduced with an adenovirus expressing Klf9 (Ad-Klf9). The overexpression of Klf9 in adipocytes induced thermogenic genes, including Ucp1, Dio2, and Pgc1α (Supplementary Fig. 2A). Based on these preliminary data, we generated fat-specific Klf9 transgenic mice (Supplementary Fig. 2B), driven by the 5.4-kb adiponectin promoter (20). We selected two independent founder lines, tg-1 and tg-2, with tg-1 having a higher level of Klf9 expression in brown fat (Fig. 2A).

As a result, under a normal chow diet, the transgenic mice gained less body weight as they aged, though they consumed an amount of food similar to that of controls (Fig. 2B–D and Supplementary Fig. 2C). eWAT, iWAT, and interscapular BAT from transgenic mice were markedly smaller and weighed less than those from control mice (Fig. 2E and Supplementary Fig. 2D). MRI examination confirmed less fat mass in transgenic mice, while the lean mass remained unaltered compared with that from control mice (Fig. 2F). Consistently, histological examination (H-E staining) of these fat depots revealed a reduction in size of the lipid droplets and adipocytes in eWAT, iWAT, and interscapular BAT of Klf9 transgenic mice (Fig. 2G and Supplementary Fig. 2E–G). Transmission electron microscopy reveal that Klf9 transgene increased the number of mitochondria in brown adipocytes but did not markedly affect mitochondrial structure (Supplementary Fig. 2H).

To test whether the Klf9 transgene affects energy expenditure, we monitored gas exchange and activity levels by housing mice in metabolic cages. Considering that the explanation of causality in metabolic cage analysis is clearer in groups of mice with similar body weights (21,22), we used body weight–matched 3-month-old Klf9 transgenic mice and controls. The Klf9 transgenic mice showed no changes in locomotor activity compared with controls, while energy expenditure, measured by respiratory oxygen consumption and carbon dioxide (CO₂) production, was higher in transgenic mice during both day and night cycles (Fig. 2H–K). Moreover, Klf9 transgene promoted OCR in BAT of mice (Fig. 2L). When subjected to acute cold exposure, these transgenic mice showed cold resistance (Fig. 2M).

We previously demonstrated that KLF9 activates Pgc1α gene transcription through direct binding to its promoter region in hepatocytes (14). As expected, ChIP assays using mouse BAT extracts confirmed that the Klf9 transgene enhanced Klf9 protein binding to the Pgc1α gene promoter region (Fig. 2N). Correspondingly, the Klf9 transgene increased the expression of genes involved in thermogenesis, fatty acid oxidation, and mitochondrial energy metabolism in iWAT, interscapular BAT, and eWAT (Fig. 2O–Q and Supplementary Fig. 2I). Immunohistochemical analysis also revealed more intense UCP1 immunoreactivity in BAT of Klf9 transgenic mice (Fig. 2R), while UCP1 protein levels in iWAT and eWAT of these old mice were below the limit of detection (Supplementary Fig. 2J). However, Klf9 transgene did not influence the expression of Zfp423 (Supplementary Fig. 2K), a critical factor maintaining white adipocyte identity through suppression of thermogenic gene program (23).

We also examined glucose and lipid metabolism in Klf9 transgenic mice. Blood glucose levels in transgenic mice were markedly lower than those in control mice under short-term fasting conditions (Supplementary Fig. 2L). The GTTs indicated that the Klf9 transgene enhanced glucose tolerance, and the ITTs indicated improved insulin sensitivity in transgenic mice (Supplementary Fig. 2M and N). Finally, biochemical analysis revealed a significant decrease in serum triglycerides, free fatty acids (FFAs), and hepatic triglyceride levels in the transgenic mice (Supplementary Fig. 2O–Q).

We also examined these mice under an HFD condition and obtained results similar to those above (Fig. 3). Klf9 transgenic mice are resistant to HFD-induced obesity (Fig. 3A and B). The transgenic mice weighed less at 9 weeks after HFD feeding. The sizes of three different fat depots in transgenic mice were smaller compared with those in control mice (Fig. 3C). MRI analysis also confirmed reduced fat mass in Klf9 transgenic mice (Fig. 3D). Likewise, the sizes of the lipid droplets and adipocytes in BAT and WAT of Klf9 transgenic mice were smaller than those in control mice (Fig. 3E–H). Consistent with the increased expression of genes involved in the thermogenesis of BAT and iWAT (Fig. 3I–L and Supplementary Fig. 3A), transgenic mice had decreased serum FFA, serum triglyceride, and hepatic triglyceride levels (Fig. 3M–O). Notably, after 3 months of HFD feeding, transgenic mice had lower fasting blood glucose levels and improved glucose tolerance and insulin sensitivity compared with control littermates (Fig. 3P–S). These results suggest that Klf9 overexpression in fat protects against HFD-induced obesity and improves glucose metabolism.

For exploration of whether the effects of KLF9 on fat thermogenesis are cell autonomous, the SVF isolated from the brown fat pads from Klf9 transgenic mice was induced into adipogenic differentiation. Real-time PCR and Western blotting analyses confirmed that the Klf9 transgene in primary adipocytes also enhanced the expression of Pgc1α and its target gene (Fig. 4A and B). Upon using a Seahorse XF-24 Extracellular Flux Analyzer, we observed that basal mitochondrial respiration and maximal mitochondrial respiratory capacity increased in Klf9 transgenic adipocytes, as did oligomycin-dependent uncoupled cellular respiration (Fig. 4C and D).

Since Pgc1α is a direct target gene of Klf9 (14) (Fig. 2N), we next explored whether PGC1α is required for a KLF9-induced thermogenic program. Indeed, knockdown of Pgc1α gene expression largely abolished the stimulatory effects on thermogenic genes (Supplementary Fig. 4A).

To test whether the ability of KLF9 to stimulate thermogenesis in primary brown adipocytes is Ucp1 dependent, we treated brown adipocytes derived from Ucp1 knockout (KO) mice and WT mice with Ad-Klf9 and Ad-GFP (control). Although the thermogenic gene expression program induced by Ad-Klf9 was similar in Ucp1 KO adipocytes and WT cells, the ability of KLF9 to increase cellular respiration was lost in Ucp1 KO cells (Fig. 4E–H). These data suggest that Ucp1 is required for Klf9-mediated stimulation of thermogenesis in fat cells.

For further study of KLF9 physiological function in fat cells, the SVF isolated from brown fat from Klf9 KO mice and WT mice were differentiated into mature adipocytes. Notably, depletion of Klf9 did not affect morphological differentiation or change the expression of adipocyte general markers (Supplementary Fig. 4B and C). However, the expression of thermogenic and mitochondrial genes was decreased in Klf9-mutant cells (Fig. 4I and J). Oxygen consumption experiments also revealed decreased total respiration and uncoupled respiration (Fig. 4K and L). These data indicate that KLF9 is required for the maintenance of brown adipocyte identity and function. However, KLF9 is not required for adipogenesis per se.

To further examine KLF9 function in energy metabolism in vivo, we first employed global Klf9 mutant mice (11). Western blotting analysis confirmed the lack of expression of KLF9 in BAT and iWAT of these KO mice (Fig. 5A). On a chow diet, the body weight of mice began to diverge at 15 weeks of age, with mutant mice gaining more weight than control mice, although mutant mice and control mice had similar food intake per day (Fig. 5B and C and Supplementary Fig. 5A). MRI examination confirmed that mutant mice had more fat mass than control mice (Fig. 5D). We dissected and weighed different fat pads and organs (including the liver, kidney, and spleen) and found that the individual fat pads of the mutant mice were markedly larger than those of control mice (Fig. 5E and Supplementary Fig. 5B). Histological analysis (H-E staining) revealed adipocyte hypertrophy in BAT and WAT in Klf9-mutant mice (Fig. 5F and Supplementary Fig. 5C–E). Notably, many unilocular fat cells appeared in the BAT of these mutant mice, indicating that the Klf9 mutation induces a brown-to-white fat switch (Fig. 5F). Transmission electron microscopy revealed that Klf9 deficiency decreased the number of mitochondria in brown adipocytes (Supplementary Fig. 5F).

Metabolic cage experiments using body weight–matched 3-month-old Klf9 mutant mice and controls suggested that the Klf9 mutation did not affect locomotor activity, although it decreased energy expenditure, as measured by respiratory oxygen consumption and CO₂ production (Fig. 5G–J). Furthermore, Klf9 deficiency decreased OCR of BAT of mice (Fig. 5K). At room temperature, Klf9 mutant mice had core body temperatures comparable with those of WT mice. However, when subjected to acute cold exposure, Klf9 mutant mice displayed cold intolerance at 4°C (Fig. 5L). Furthermore, we analyzed the in vivo BAT function of Klf9 mutant mice. PET-CT imaging data suggested a marked decrease in [¹⁸F]-fluorodeoxyglucose uptake in the BAT of Klf9 mutant mice (Fig. 5M).

ChIP assays confirmed that the endogenous KLF9 proteins in BAT of WT mice bind to the Pgc1α gene promoter region, while Klf9 mutation abolished these effects (Fig. 5N). As expected, Klf9 deficiency led to a decrease in PGC1α and its downstream target genes (Fig. 5O–Q and Supplementary Fig. 5G). Klf9 mutant mice had elevated serum and hepatic triglyceride levels. Moreover, serum FFA levels were also higher in Klf9 mutant mice (Supplementary Fig. 5H–J). Collectively, these data suggest that loss of KLF9 function impairs whole-body metabolism.

Since both cold stress and β-adrenergic agonists stimulate fat Klf9 expression, we explored whether KLF9 functions in thermogenesis induced by a β-adrenergic agonist. We first observed reduced size and mass of BAT, eWAT, and iWAT of WT mice following 4 days of β-adrenergic agonist treatment. The subcutaneous iWAT of CL 316,243–treated control mice appeared browner than that of vehicle-treated control mice (Fig. 6A). Histological analysis (H-E staining) also revealed reduced lipid droplet size in BAT and eWAT of CL 316,243–treated control mice (Fig. 6B and C). Furthermore, CL 316,243 treatment of control mice led to the occurrence of abundant UCP1-positive beige adipocytes with multilocular lipid droplets, as revealed by immunohistochemical staining (Fig. 6D). We further analyzed in vivo thermogenic fat function by measuring whole-animal oxygen consumption before and after injection of CL 316,243. CL 316,243 injection immediately stimulated oxygen consumption; however, Klf9 mutation significantly impaired the compound-stimulated oxygen consumption, clearly suggesting that KLF9 is required for the thermogenic function of brown and beige fat in vivo (Fig. 6E and F).

Molecular mechanism studies suggest that CL 316,243 injection or cold exposure promotes KLF9 proteins binding to Pgc1α gene promoter (Fig. 6G and H). As expected, treatment with the β3-adrenergic agonist CL 316,243 induced thermogenic gene expression in BAT and iWAT of control mice (Fig. 6I–K). However, these effects were largely impaired in global Klf9 KO mice (Fig. 6I–K). These data indicate that KLF9 is also required for adipose thermogenesis induced by a β-adrenergic agonist.

Based on these data, we proposed a model of cold exposure induction of the thermogenesis of brown and beige fat (Fig. 6L). In response to cold exposure, sympathetic neurons secrete catecholamines, which bind to β-adrenoreceptors, leading to activation of adenylyl cyclase and increased cAMP and PKA activity, thereby stimulating Klf9 expression in adipocytes. Furthermore, induced KLF9 promotes browning of WAT and thermogenesis of brown and beige fat via a direct binding to Pgc1α gene promoter to activate its transcription (Fig. 6L).

Previously, a human genetic study (GWAS) suggested that the rs11142387 single nucleotide polymorphism located in the promoter region of KLF9 is associated with obesity in the East Asian population. However, the nature of KLF9 affecting human BMI was not explored. In the current study, we show that fat KLF9 regulates energy metabolism by stimulating the expression of Pgc1α, a master regulator of oxidative phosphorylation and thermogenesis. Global Klf9-deficient mice exhibited obesity; conversely, Klf9 transgenic mice gained less weight and were resistant to obesity induced by an HFD. Of note, although our in vitro and in vivo data indicate KLF9 stimulates fat cell thermogenesis, we cannot rule out the possibility that the Klf9 deficiency in other tissues also contributes to the obesity in global Klf9-deficient mice. Of note, in our study the body weight of WT mice at 21–57 weeks of age on a chow diet is higher than that in other studies in the field. The possible reason is that fat in the chow diet in this study accounts for 13.8% of total calories (cat. no. 1010010; Xietong Organism, Jiangsu, China), which is a little higher than that in regular diet from Research Diets (10% of total calories from fat, cat. no. D12450J).

Moreover, we observed a reverse association between Klf9 mRNA levels in fat and human BMI. These data confirm its physiologically thermogenic role and conservation of KLF9 function between different species. Further studies are required to confirm whether KLF9 also correlates with the browning of human omental fat. In addition, a previous study suggests that Klf9 regulates 3T3-L1 adipocyte differentiation (24). However, in the current study we show that Klf9 deficiency did not affect differentiation of primary brown adipocytes. A possible explanation is that other genes might compensate for the long-term Klf9 deficiency in brown fat, while KLF9 is required for differentiation of white adipocytes. Further studies are required for clarification of this discrepancy.

KLF9 lies upstream of Pgc1α, and they share similar functions in hepatic gluconeogenesis and brown fat thermogenesis (19,25). Both Klf9- and Pgc1α–deficient mice have abnormal brown fat, with abundant accumulation of large lipid droplets, indicating impaired thermogenic function. Moreover, both of these two mouse lines are sensitive to cold exposure (our present data and data previously published [8,26]). However, KLF9 has a function distinct from that of Pgc1α. Our data suggest that Klf9-deficient mice display obesity compared with control mice. However, Lin et al. (8) reported that Pgc1α–deficient mice are lean and weigh 10–15% less than control mice at 2 months of age due to the profound hyperactivity of null mice resulting from lesions in the striatal region of the brain that controls movement. However, this hyperactivity phenomenon was not observed in Klf9-deficient mice. Additionally, Zucker et al. (27) reported that Klf9 increases ROS levels in various types of cells and mouse tissues. Klf9-deficient mice display resistance to bleomycin-induced oxidative stress and pulmonary fibrosis (27). In contrast, PGC1α decreases ROS levels and protects neural cells from oxidative stressor–mediated death. Pgc1α–null mice are much more sensitive to the neurodegenerative effects of oxidative stress (28). Further studies are required to clarify the molecular mechanism underlying the differences in function between KLF9 and PGC1α.

In summary, the results of the current study suggest that KLF9 is a critical regulator of thermogenesis of fat.

Funding. This work was supported by the National Key Research and Development Program of China (2018YFA0800601), the National Natural Science Foundation of China (grants 81730024, 81825004, 81670749, and 81471079), and the Natural Science Foundation of Ningxia (2020AAC03422).

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

Author Contributions. Y.C., H.F., Yu.Z., and J.Z.L. contributed to designing research studies. H.F., Yu.Z., J. Zhang, and Q.Y. contributed to conducting experiments. H.F., Yu.Z., J. Zhang, Q.Y., Y.S., Q.S., J.L., Y.G., X.W., L.Z., and Yi.Z. contributed to acquiring data. H.F., Yu.Z., J. Zhang, Q.Y., Y.S., Q.S., P.L., J. Zhao, and Q.C. contributed to analyzing data. J. Zhao, Q.C., and J.Z.L. contributed to providing reagents. Y.C. wrote the manuscript. Y.C. 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.