Rhamnose Displays an Anti-Obesity Effect Through Stimulation of Adipose Dopamine Receptors and Thermogenesis

The worldwide prevalence of obesity and its related diseases, such as type 2 diabetes and liver steatosis, has become a major threat to human health. Obesity is caused by an imbalance between energy intake and energy expenditure. Excessive energy intake, largely achieved by the consumption of sugar, is stored in white adipose tissue (WAT) in the form of triglycerides and leads to adipose hypertrophy and inflammation, which contribute to the development of systemic insulin resistance. Therefore, one common approach to restrict energy intake is to replace sugar with nonnutritive sweeteners without calories. Rhamnose is a naturally abundant deoxy sugar widely used as a nonnutritive sweetener to reduce calorie intake. Rhamnose is reported to reduce the synthesis of cholesterol and diacylglycerides (1); however, whether rhamnose could exert an anti-obesity effect is still unknown.

Besides WAT, mammals also possess brown adipose tissue (BAT) to dissipate chemical energy to generate heat (2). Meanwhile, cold exposure and β₃ adrenergic receptor (β₃AR) stimulation also induce white adipose browning to generate beige adipocytes. These beige adipocytes possess multilocular lipid droplets and a high content of mitochondria with uncoupling protein-1 (UCP1) to drive heat production, just like BAT adipocytes. Accumulated evidence suggests that activation of brown and beige thermogenesis could be a new strategy to combat obesity by increasing energy expenditure (3–6). Here, we sought to combine these two approaches by identifying rhamnose as a nonnutritive sweetener that simultaneously promotes thermogenesis.

All the mouse experiments in this study were performed following the guidelines established by the Animal Experiment Committee of Tongji University and in accordance with the guidelines of School of Medicine, Tongji University. All the animals were kept at room temperature (25°C) and in a humidity-controlled room with a 12-h light/dark cycle. A standard normal regular diet (RD) (5% fat; D12450 formula, Research Diet, New Brunswick, NJ) and sterile water were given ad libitum. All mice used in experiments throughout the study were male and exhibited normal health. All mice were used after 1 or 2 weeks of acclimatization after their arrival to the facility. The C57BL/6J and ob/ob mice (SPF grade) were purchased from Slack Laboratory Animal Co., Ltd. (Shanghai, China).

For studies with diet-induced obese mouse models, 6-week-old male C57BL/6J mice were given a high-fat diet (HFD; 60% fat, catalog D12492; Research Diet) for a total period of 9–10 weeks to generate mice with diet-induced obesity before other experiments were conducted. For rhamnose treatment in water, rhamnose was dissolved in water at a concentration of 0.25, 0.5, and 1 mg/mL and replaced every day. Body weight was measured every week for a total of 10 weeks. Fat mass and lean mass in mice with diet-induced obesity that were treated with rhamnose for 9 weeks were measured by nuclear magnetic resonance imaging. For rhamnose injection, 300 mg/kg rhamnose was injected intraperitoneally (i.p.) daily for 7 consecutive days. Saline was used as a control.

After rhamnose injection, mice were subjected 12 h later to the following experiments. For lentivirus local injection in adipose tissue, mice were anesthetized and a midline incision in the skin was made to expose subcutaneous WAT (scWAT). Both sides of scWAT were injected with either Lenti-shDRD1 or Lenti-shNS (4 × 10¹⁰ pfu). Ten days later, mice were subjected to the following experiments. For glucose tolerance testing, mice were fasted for 12 h and injected i.p. with glucose (1.5 g/kg for HFD-fed mice and 1 g/kg for RD-fed mice). For insulin tolerance test experiments, mice were fasted for 4 h and injected i.p. with insulin (1 unit/kg for HFD-fed mice and 0.75 unit/kg for RD-fed mice). Blood from anesthetized mice (4% isoflurane at euthanization) was collected by cardiac puncture, and blood from live mice was collected from a tail snip.

Six-week-old mice were fed an RD or HFD for indicated times. Oxygen consumption and food intake were measured by the Comprehensive Lab Animal Monitoring System (Columbus Instruments). Oxygen consumption data were analyzed using CaIR (version 1.1; https://calrapp.org/). Core body temperature and locomotor activity were monitored using the E-Mitters (StarrLife; Starr Life Sciences Corp.). The ER4500 Energizer/Receiver provided power and received measurement data back from implanted E-Mitters. Core body temperature and locomotor activity data were collected using VitalView software. The mice were acclimated to the system for 48 h before formal testing; they had free access to food and water throughout the process.

Primary beige adipocytes were induced by treating confluent stromal vascular fraction (SVF) of scWAT or immortalized brown and beige progenitor cells (BACs) with 0.5 mmol/L isobutylmethylxanthine, 125 nmol/L indomethacin, 2 μg/mL dexamethasone, 850 nmol/L insulin, 1 nmol/L triiodothyronine (T3), and 0.5 μmol/L rosiglitazone. Two days after induction, cells were switched to maintenance medium containing 10% FBS, 850 nmol/L insulin, 1 nmol/L T3, and 0.5 μmol/L rosiglitazone for another 4 days. HEK293T cells were obtained from the American Type Culture Collection, and immortalized BACs were provided by Junli Liu, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital, Shanghai, China. HEK293T cells and BACs were maintained in DMEM containing 10% FBS and combined penicillin and streptomycin (100 U/mL and 100 μg/mL, respectively). No commonly misidentified cell line was used in this study. All cell lines routinely tested negative for Mycoplasma contamination.

For detection of phosphorylation of PKA substrate, hormone-sensitive lipase (HSL), and AMPK in SVF-differentiated beige adipocytes by immunoblots, cells were pretreated with H89 (10 μmol/L), Compound C (10 μmol/L), L748337 (10 nmol/L), DRD2 antagonist (5 nmol/L), SKF83566 (1 μmol/L), or SCH23390 (1 μmol/L) for 1 h and then treated with rhamnose (3 μmol/L) for another 30 min before harvest. For detection of UCP1 in SVF-differentiated beige adipocytes by immunoblots, cells were pretreated with H89 (10 mmol/L), Compound C (10 mmol/L), L748337 (10 nmol/L), DRD2 antagonist (5 nmol/L), SKF83566 (1 mmol/L), or SCH23390 (1 mmol/L) for 1 h and then treated with rhamnose (3 μmol/L) for another 8 h before harvest. For detection of PKA kinase activity assay in BAC-differentiated beige adipocytes, cells were pretreated with the indicated drug for 1 h and then treated with rhamnose (3 μmol/L) for another 30 min before harvest. For detection of thermogenic gene expression in SVF-differentiated beige adipocytes by qPCR, cells were pretreated with the indicated drug for 1 h and then treated with rhamnose (3 μmol/L) for another 6 h before harvest. For lentivirus infection, cells were infected with either Lenti-shDRD1 or Lenti-shNS (1 × 10⁷ pfu) for 48 h and then treated with rhamnose (3 μmol/L) for the following experiments.

Primary adipocytes differentiated from BACs were treated with rhamnose (3 μmol/L) for 15 min and cell lysate were collected for a PKA kinase activity assay, following the standard protocol from the manufacture (ab139435; Abcam).

From MedChemExpress (catalog no.), we purchased H89 (HY-15979), Compound C (HY-13418A), L748337 (HY-103211), DRD2 antagonist (HY-129946), SKF83566 (HY-15979), SCH23390 (HY-15979), and rosiglitazone (HY-17386) were purchased from MedChemExpress. Insulin was purchased from Yeasen (catalog 40107ES25). From Sigma (catalog no.), we purchased rhamnose (R3875), T3 (T2877), 3-isobutyl-1-methylxanthine (I5879), indomethacin (PHR1247), dexamethasone (D4902), and glucose (D9434). GAPDH antibody (catalog sc32233) was purchased from Santa Cruz, and the following were purchased from Cell Signaling Technology (catalog no.): AMPK antibody (2532), pAMPK antibody (2531), pPKA substrate antibody (9621), HSL antibody (4107), pHSL antibody (4139), and horseradish peroxidase–conjugated goat anti–rabbit IgG antibody (7074).

Protein lysates from isolated tissues or cultured cells were extracted using radioimmunoprecipitation assay buffer with protease and phosphatase inhibitors, and a total of 20 μg of protein was separated by SDS–PAGE and transferred to polyvinylidene difluoride membranes (pore size: 0.22 μm; Millipore). Membranes were incubated with blocking solution (5% milk powder in Tris-buffered saline with Tween 20) for 1 h, then with primary antibody (in blocking solution) overnight at 4°C. After several washes in Tris-buffered saline with Tween 20, membranes were incubated with horseradish peroxidase–conjugated secondary antibodies for 1 h at room temperature in blocking solution. Membranes were incubated with enhanced chemiluminescence Western-blotting substrate and imaged in a Chemidoc XRS system or ChemiDOC (Bio-Rad Laboratories).

Total RNA (DNA was removed with DNA-free DNase Treatment & Removal Reagent) was extracted from tissue or cells using the RNAsimple Total RNA Kit (Tiangen, Shanghai, China) and underwent reverse transcription with a FastQuant RT kit (Tiangen). Real-time PCR was carried out using a SuperReal SYBR Green kit (Tiangen) and the Lightcycler 96 (Roche, Penzberg, Germany).

For lentivirus production, HEK293T cells were cotransfected with a control vector or a lentiviral plasmid carrying shRNA fragments, along with a lentiviral packaging mix (psPAX2 and the pMD2.G plasmid containing vesicular stomatitis virus G). The culture medium was collected at 48 h and 72 h after transfection, filtered through a 0.45-μm filter, and incubated overnight with polyethylene glycol 8000 before being concentrated by centrifugation (4,000g for 20 min at 4°C; Thermo). Lentivirus was titrated using flow cytometry (FACS) analysis.

Cellular oxygen consumption rate was measured using an XF24 Analyzer (Seahorse Bioscience). Primary beige adipocytes differentiated from SVF were stimulated with rhamnose (3 μmol/L) for 5 h. Then cells were incubated in a prewarmed assay medium (XF base medium with 10 mmol/L glucose, 1 mmol/L sodium pyruvate, 2 mmol/L l-glutamine) for 1 h without CO₂ before oxygen consumption analysis. The mitochondrial stress test used sequential injections of oligomycin (5 μmol/L), carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (1 μmol/L), and rotenone and antimycin A in combination (5 μmol/L). The parameters of basal respiration and maximal respiration were automatically calculated by WAVE software (Agilent). Values were normalized to protein levels.

Adipose tissues were fixed in 4% paraformaldehyde overnight at room temperature, followed by dehydration in 70% ethanol. Tissues were embedded in paraffin, sectioned at a thickness of 10 μm, and stained with hematoxylin and eosin, following the standard protocol. Quantification was performed using Image-Pro Plus 5.1 software (Media Cybernetics). Five to six sections were analyzed per mouse and all assessments were analyzed.

cAMP levels in adipocytes were detected and quantified by an ELISA kit (catalog KGE012B, R&D Systems) according to the manufacturer’s instructions.

All experiments were repeated at least three times. Data are reported as mean ± SEM and were analyzed using GraphPad Prism 8.0 (GraphPad Software Inc., San Diego, CA). An unpaired Student t test was a used to compare the two data sets. Two-way ANOVA was used when more than two data sets or groups were compared. The data from assessments of thermogenesis and indirect calorimetry were analyzed with ANCOVA using weight as a covariable, as reported (7). Statistical significance was accepted at P < 0.05.

All data generated or analyzed during this study are included in this article and its online supplementary files and are available upon request. No applicable resources were generated or analyzed during this study.

We assayed the potency of 13 widely used nonnutritive sweeteners, including rhamnose, aspartame, d-mannitol, d-sorbitol, isomalt, isosteviol, lactitol monohydrate, maltitol, meso-erythritol, mogroside V, saccharin sodium hydrate, sucralose, and xylitol, to promote browning in primary adipocytes differentiated from SVFs of subcutaneous WATscWAT and found that the mRNA levels of UCP1 were most highly promoted by rhamnose (Fig. 1A). Indeed, rhamnose treatment dose dependently increased the mRNA levels of thermogenic genes but did not alter the expression of the adipogenic gene Fabp4 (Fig. 1B). UCP1 protein abundance was increased dose and time dependently by rhamnose treatment in SVF-derived adipocytes (Fig. 1C and D). The cellular oxygen consumption rate showed increased maximum respiration capacity with rhamnose treatment, indicating an increase in lipolysis and fatty acid oxidation probably due to the increase of cAMP–PKA signaling (Fig. 1E). Indeed, mRNA levels of fatty acid oxidation genes, but not mitochondrial electron transport chain genes, were increased by rhamnose treatment in SVF-derived adipocytes (Supplementary Fig. 1A and B).

The increased thermogenic program in vitro by rhamnose treatment prompted us to investigate the effect of rhamnose on thermogenesis in vivo. Indeed, when mice were i.p. injected with rhamnose and subjected to whole-body indirect calorimetry analyses, an increase in VO₂ was produced without any influence on food intake and locomotive activity (Fig. 1F–H). mRNA levels of thermogenic genes were increased in scWAT, and protein levels of UCP1 in scWAT, but not BAT, were also upregulated (Fig. 1I and J and Supplementary Fig. 1C and D). When these mice were subjected to cold (10°C), core body temperature was also increased in rhamnose-treated mice compared with that of control-treated mice (Fig. 1K). Because of the glucose-sinking properties of beige thermogenesis, glucose tolerance tests showed improved glucose tolerance in rhamnose-injected mice compared with controls, although basal blood glucose levels remained unchanged (Fig. 1L and M). ITT results also showed a slightly improved insulin sensitivity in rhamnose-treated mice (Supplementary Fig. 1E and F). This prothermogenic effect was UCP1 dependent, because rhamnose injection in UCP1 knockout mice did not show any beneficial effects on energy expenditure and glucose homeostasis (Fig. 1N–P).

Thermogenesis is an attractive target for obesity treatment (8), which prompted us to further determine whether rhamnose might regulate the progress of obesity. We treated mice with HFD-induced obesity with different doses of rhamnose (0.25, 0.5, and 1 mg/mL) dissolved in drinking water for 9 consecutive weeks, and found that body weight was decreased in mice treated with 0.5 and 1 mg/mL rhamnose compared with control-treated mice (Fig. 2A). We chose mice treated with 0.5 mg/mL rhamnose for the following experiments.

Rhamnose-treated mice had less total fat mass with unchanged lean mass, compared with control mice (Fig. 2B). The weight of scWAT was lower, and histological analysis also showed smaller adipocytes in scWAT from treated mice than in control mice (Fig. 2C and D). Adipocytes in BAT also showed a trend of decreasing size (Supplementary Fig. 1G). Although food intake and activity were not changed, whole-body indirect calorimetry analyses revealed augmented VO₂ (Fig. 2E–G), indicating increased energy expenditure. Indeed, real-time PCR confirmed the increased expression of important thermogenic genes such as Ucp1, Cidea, Prdm16, and Cox8b in scWAT but not BAT (Fig. 2H and Supplementary Fig. 1H). Protein levels of UCP1 were also increased in scWAT, but not BAT, as indicated by immunoblot (Fig. 2I and Supplementary Fig. 1I). Glucose tolerance test results showed ameliorated glucose sensitivity in rhamnose-injected mice (Fig. 2J and K). We also assessed the anti-obesity effect of rhamnose in ob/ob mice by i.p. injection. Consistently, rhamnose administration exerted a rapid body weight–lowering effect with decreased scWAT weight (Fig. 2L and M). Although food intake and activity were not changed, VO₂ and expression of thermogenic genes were increased (Fig. 2N–Q). Improved glucose sensitivity and smaller adipocytes in scWAT were observed more frequently in rhamnose-injected ob/ob mice than in control-injected mice (Fig. 2R–T). These results demonstrate that rhamnose administration promotes adipose thermogenesis and prevents mice from obesity.

To decipher the mechanism of rhamnose’s effect, we first investigated PKA signaling during rhamnose treatment because this pathway is critical for the induction of thermogenesis (9). Rhamnose treatment increased phosphorylation of PKA substrate and HSL in SVF-derived adipocytes in a dose- and time-dependent manner (Fig. 3A and B). PKA activity was also increased by rhamnose treatment in a dose- and time-dependent manner (Fig. 3C and D).

We chose 3 μmol/L rhamnose-treated cells for the following experiments. Indeed, cAMP levels were also increased by rhamnose treatment (Supplementary Fig. 1J). When SVF-derived adipocytes were pretreated with PKA inhibitor H89, rhamnose was unable to stimulate phosphorylation of PKA substrate and HSL or PKA activity and expression of thermogenic genes and UCP1 protein (Fig. 3E–H). Thermogenic adipose tissue is enriched with sympathetic nerves, which release catecholamine to stimulate the expression of UCP1 and thermogenesis (10,11). β₃AR and downstream cAMP–PKA signaling are critical for catecholamine-stimulated lipolysis (through phosphorylation of HSL by PKA) and thermogenesis (through induction of UCP1 expression).

Having seen that rhamnose stimulated PKA signaling, we sought to investigate whether rhamnose targeted β₃AR. However, rhamnose’s effect on phosphorylation of PKA substrate and HSL, PKA activity, and expression of thermogenic genes and UCP1 protein remained unaltered by β₃AR selective antagonist L748337 cotreatment, indicating that rhamnose stimulated PKA signaling is independent of β₃AR (Fig. 3I–L). AMPK is also critical for the induction of thermogenesis, and it has been reported that rhamnose could stimulate phosphorylation of AMPK (12). However, in SVF-derived adipocytes, we were unable to detect the increase of AMPK phosphorylation by rhamnose, and expression of rhamnose-stimulated thermogenic genes and UCP1 protein remained unchanged with AMPK inhibitor Compound C cotreatment (Fig. 3M–O). These results indicated that AMPK does not participate in this setting.

To identify the G-protein coupled receptor responsible for rhamnose’s effect on PKA signaling, we used an online database (Swiss Target Prediction, https://www.swisstargetprediction.ch) (13) to predict the putative binding proteins for rhamnose, and found dopamine receptors as a candidate (Fig. 4A). Dopamine receptors, including dopamine receptor D1 (DRD1) and dopamine receptor D2 (DRD2), are abundantly expressed in adipose tissue (14,15), and it has been reported that DRD1 signaling stimulates cAMP production and activating DRD2 inhibits cAMP production (15). Indeed, DRD1 antagonist SKF83566, but not DRD2 antagonist, pretreatment abolished rhamnose’s effect on thermogenic gene expression (Fig. 4B and C). SKF83566 also abolished rhamnose-stimulated phosphorylation of PKA substrate and HSL, PKA activity, and UCP1 protein expression (Fig. 4D–F). Similar results were obtained with an alternative DRD1 antagonist, SCH23390 (Fig. 4G–J). To further confirm the importance of DRD1 for rhamnose, we infected primary adipocytes with a lentivirus expressing an shRNA targeting DRD1 (Lenti-shDRD1) to achieve ∼50% knockdown of DRD1 expression (Supplementary Fig. 1K) and then stimulated the cells with rhamnose. Rhamnose-stimulated phosphorylation of PKA substrate and HSL, PKA activity, and expression of thermogenic genes and UCP1 protein were decreased in lenti-shDRD1–infected cells compared with lenti-nonspecific shRNA (Lenti-shNS)–injected cells (Fig. 4K–N).

Next, we tried to investigate whether rhamnose functioned through DRD1 in vivo. We injected lenti-shDRD1 into the scWAT of mice locally to achieve a 50% knockdown of DRD1 expression in scWAT (Fig. 4O). After i.p. injection of rhamnose, knockdown of DRD1 impaired expression of rhamnose-induced thermogenic genes and UCP1 in scWAT when these mice were subjected to cold to boost thermogenesis (Fig. 4P and Q). Also, during cold exposure, the core body temperature was decreased in lenti-shDRD1–injected mice compared with lenti-shNS–injected mice after rhamnose injection (Fig. 4R). Glucose tolerance test results also showed deteriorated glucose sensitivity in lenti-shDRD1–injected mice after rhamnose injection (Fig. 4S and T). Together, these results confirmed the essential role of DRD1 in rhamnose-mediated adipose thermogenesis.

Despite their morphological and functional similarities, BAT and beige adipocytes are distinct types of fat cells with different developmental origins and depot locations (16). They possess different regulatory mechanisms within their unique tissue microenvironment (17). It seems that rhamnose specifically targeted adipose browning in scWAT with marginal effect in BAT. To understand the different effects of rhamnose on these two tissues, we detected the expression levels of DRD1 in scWAT and BAT. We found that DRD1 was more highly expressed in scWAT than in BAT (Supplementary Fig. 1H). The different expression levels of DRD1 in scWAT and BAT might explain our observation that rhamnose specifically targeted scWAT but not BAT. DRD1 expression in scWAT, but not BAT, was increased when mice were fed an HFD diet (Supplementary Fig. 1I and J). Importantly, scWAT, but not BAT, DRD1 mRNA levels were much higher in ob/ob mice (∼40 g) compared with HFD-fed mice (∼35 g) (Supplementary Fig. 1I and J), probably due to extreme hypertrophy of scWAT in ob/ob mice. This could explain the reason why rhamnose treatment resulted in a more dramatic body weight–lowering effect in ob/ob mice. Taken together, these results suggest rhamnose could be a useful tool against extreme obese conditions due to the obesity-induced DRD1 expression in scWAT.

Although there is one report saying that rhamnose may induce UCP1 expression in 3T3-L1 and HIB1B adipocytes (12), the link between rhamnose and thermogenesis and the contribution of rhamnose to anti-obesity treatment are still enigmatic. Here, by detecting cellular oxygen consumption rate and energy expenditure in mice, we provide direct evidence that rhamnose promotes thermogenesis. Furthermore, the lack of effect of rhamnose in UCP1 knockout mice directly links rhamnose’s induction of UCP1 with thermogenesis. To our knowledge, we are the first to identify DRD1 as a target of rhamnose. We clearly showed that rhamnose stimulated cAMP production and PKA activation through DRD1 in adipocytes. Importantly, knockdown of DRD1 expression in scWAT abolished rhamnose’s effect on thermogenesis in vivo. These results not only identified rhamnose as a prothermogenic reagent but also indicated its application in anti-obesity treatment.

Rhamnose has been widely used as a food sweetener to replace sucrose and is beneficial for human health by reducing energy intake. Here, we show that rhamnose also could promote energy expenditure by promoting UCP1-dependent thermogenesis. Thus, rhamnose has dual beneficial effects that make using it an intriguing strategy against obesity. Considering that its safety for human use has been tested, rhamnose, together with other beneficial protein or lipid molecules, might serve as a “molecular diet” to prevent the prevalence of obesity.

DRD1 and DRD2 contribute to the regulation of BAT thermogenesis. It has been reported that dopamine, with high similarity to catecholamine, modulates BAT thermogenesis in rodents (18). Additional studies have shown that dopamine and the DRD1 agonist SKF38393 increase BAT adipocyte thermogenesis in vitro (19), and activating DRD2 reduces BAT thermogenesis (20). However, whether dopamine-receptor signaling directly affects browning in WAT has not been studied, to our knowledge. Here, we reported that Rhamnose could activate cAMP signaling and stimulate adipose browning through DRD1. Although the binding of rhamnose with DRD1 still needs further characterization, the use of rhamnose as a DRD1 activator will greatly improve our understanding of the function of DRD1 in WAT and the regulation of adipose browning.

In the central nervous system, dopamine receptors are involved in the regulation of motor activity and several neurological disorders, such as schizophrenia, bipolar disorder, Parkinson disease, Alzheimer disease, and attention deficit hyperactivity disorder (21). For example, reduction in dopamine content in the nigrostriatal pathway is associated with the development of Parkinson disease, along with the degeneration of dopaminergic neurons in the substantia nigra region (22). Here, we identified rhamnose as a DRD1 modulator. Thus, rhamnose may not only serve as an anti-obesity drug but also a possible therapeutic target that might help slow the process of neurodegeneration, such as occurs in Parkinson’s disease, which will need further investigation.

Funding. This work was supported by the National Natural Science Foundation of China (grants 81922015, 82170862, and 82171405), the Shanghai scientific research project (grant 19ZR1461700), the Shuguang Program of Shanghai Education Development Foundation, the Shanghai Municipal Education Commission (grant 21SG21), the Shanghai Center for Brain Science and Brain-Inspired Technology (grant LG-QS-202205-10), and by the Fundamental Research Funds for the Central Universities.

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

Author Contributions. G.P. and B.L. conceived and directed the study. S.L., T.H., R.Z., Y.Z., W.Y., Z.W., C.S., J.L., S.H., B.L., and G.P. designed and performed experiments and analyzed the data. G.P., B.L., and S.L. wrote the manuscript with comments from all authors. All authors provided input and reviewed the manuscript. B.L. is the guarantor of this work and 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.