Results

In vivo SAR study of glabridin and its derivatives identifies anti-obesity components of glabridin-backbone and the final optimized compound HSG4112
We performed a SAR study to first overcome the chemical instability of glabridin, whose structure is shown in Fig. 1a. Glabridin’s low stability can be attributed to the pyranobenzene structure in ring A, which is labile under acidic conditions or light, and the resorcinol structure shown in ring B, which is labile under basic conditions. Accordingly, we validated the anti-obesity effect of each modified component by orally administering the compounds for 4 to 6 weeks to male HFD-induced obese C57BL/6J mice fed with HFD for 11 weeks prior to the administration. We used this direct phenotypic screening method instead of in vitro screening to take into account the divergent pathways and multifaceted network of signals needed for ameliorating obesity.  
We modified the pyranobenzene structure by hydrogenating the double bond between the carbon atoms at 3’’ and 4’’ in ring B to create 3’’,4’’-dihydro-glabridin, using the hydrogenation reaction from the previously reported protocol [15]. 3’’,4’’-dihydro-glabridin induced greater body weight reduction in HFD-induced obese mice than glabridin (Fig. 1b), demonstrating that the double bond of the pyranobenzene group is not required for the weight-reducing effect of glabridin. Given the improvement in chemical stability and efficacy, we performed hydrogenation for all following synthetic derivatives.
Next, we modified the resorcinol structure while retaining the -oxy backbone of glabridin. We etherified C-2’, C’4, or both carbons by attaching methoxy groups, using the typical methylation (MeI, K2CO3 in acetone) process and column separation of the resulting mixture. The etherification allowed us to test whether the hydroxy groups are necessary for the efficacy, and if so, to determine which of them is critical and which of them can be modified to increase stability. We found that hydroxy-to-methoxy modification at C-2’ (R1) induced markedly lower weight loss effects than glabridin (Fig. 1c). This suggests that the C-2’ hydroxy group in ring A is a pharmacophore for weight reduction. Surprisingly, while glabridin led to 13.2% weight loss after 5-week administration, hydroxy-to-methoxy modification at C-4’ (R2) led to 25.5% weight loss, which is approximately two-fold greater in percentage. We report for the first time that the attachment of a methoxy group at C-4’ remarkably improves the weight-reducing action of glabridin.
Given this beneficial effect from hydroxy-to-methoxy modification at C-4’, we tested whether a further chain elongation of the C-4’ alkoxy group would have any differential effects. In terms of weight-reducing efficacy, we found that the attaching an ethoxy group at C-4’ surpassed the attachment of methoxy or all other tested alkoxy groups (Fig. 1d). This compound, with optimized chain length, was termed (R)-HSG4112 (Hydrogenated Synthetic Glabridin 4112) by the authors. Further lengthening of the alkoxy substituents had diminished yet retained weight-reducing effects.
Lastly, since the chiral syntheses of glabridin derivatives present a serious challenge in methodology and productivity [10], we synthesized racemic HSG4112 and tested its effect on body weight. Glabridin in nature exists in (R)-form, and the effect of (S)-glabridin and its derivatives is unknown and predictably null; for most cases of small molecular drugs, only one enantiomer is pharmacologically active while the other enantiomer is either inactive or toxic [21,22]. Surprisingly, we found that at equivalent dose levels, the (S)-isomer surpassed both the (R)-isomer and racemic HSG4112 in body weight reduction (Fig. 1e). This presents a remarkable discovery of a glabridin derivative with both enantiomers active and with the more potent enantiomer being the unnatural, synthetic (S)-form. Given the similar pharmacological effect of both enantiomers and the inefficient and elaborate protocol needed for the synthesis of chiral HSG4112, racemic HSG4112 – simply termed HSG4112 – was chosen as our most optimized compound.
In order to confirm the increased chemical stability of HSG4112 relative to glabridin, both compounds were placed in acidic and basic conditions in MeOH and their decomposition rates were measured using HPLC (Fig 1f, g). HSG4112 proved to be dramatically more stable in both conditions. Expectedly, improvement on the pyranobenzene and resorcinol parts of glabridin led to a significant increase in stability.
 
HSG4112 fully reverses adiposity in HFD-induced obese mice in a dose-dependent manner
After the discovery of HSG4112, we aimed to characterize its full preclinical efficacy in the same experimental setup as the SAR study, using HFD-induced obese mice given different doses of HSG4112 (10, 30, and 100 mg·kg-1) with the addition of a pair-fed group. The pair-fed group was given the amount of feed the HSG4112-100 mg·kg-1 group consumed the day before. 6-week administration of HSG4112 at 10, 30, and 100 mg·kg-1 dose led to significant dose-dependent body weight reduction by 4.0 g (8.3%), 10 g (21%), and 19 g (40%), respectively, compared to the 48.1 g body weight of HFD-induced obese mice administered with only the vehicle (hereinafter vehicle group) (Fig. 2a). Expectedly, the plasma concentrations of HSG4112 in the HSG4112-30 mg·kg-1 and -100 mg·kg-1 groups at the end of the 6-week administration period showed dose-proportionality (Supplementary Fig. 2). At 100 mg·kg-1 dose, the body weight of HFD- induced obese mice was completely normalized to 29.2 g, equivalent to that of normal chow-fed mice (hereinafter normal group). The body weight of the pair-fed group was 40.9 g; under the assumption that the pair-fed group fully represents the reduction of food intake in the HSG4112-100 mg·kg-1 group, reduced food intake accounts for 37.8% (-7.2 g) of the weight loss induced by HSG4112 while enhanced energy expenditure accounts for 62.2% (-11.7 g). The mean daily food intake was significantly reduced in all HSG4112-treated groups compared to 5.47 g of daily HFD consumed by the vehicle group (Fig. 2b); HSG4112-10, 30, and 100 mg·kg-1 groups had 3.67, 3.74, and 2.97 g of daily food intake, respectively. However, no statistically significant difference in food intake reduction was observed between the groups treated with different doses of HSG4112, suggesting incomplete dose-proportionality. Representative images from the normal, vehicle, and HSG4112-100 mg·kg-1 groups are shown in Fig. 2c.
Next, we investigated the effect of HSG4112 on additional parameters of obesity: adipose tissue and muscle mass, adipocyte size, and blood hormone and glucose levels.  HSG4112 dose-dependently normalized body-weight-relative gastrocnemius muscle weight, while only the HSG4112-100 mg·kg-1 group showed significant decrease in absolute gastrocnemius muscle weight (Fig. 2d). Significant and dose-dependent reduction of both relative and absolute white adipose tissue (WAT) masses in abdominal fat – periepididymal and perirenal adipose tissues – were observed (Fig. 2e, f) in addition to the similar degree of effects in individual adipocyte size (Fig. 2g, h). Additionally, leptin and insulin are key hormones in mediating glucose homeostasis, and their blood levels are indicative markers of adiposity and negative predictors for future weight gain [23,24]. HSG4112 significantly decreased serum leptin, insulin, and glucose concentration to a normal level (Fig. 2i-k) in a dose-dependent manner. In summary, dose-dependent amelioration of adiposity and full normalization of adiposity in the HSG4112-100 mg·kg-1 group were observed. Given the immense beneficial effect of HSG4112 at 100 mg·kg-1, this dosage was used for treatment groups in the following analyses (hereinafter HSG4112 group). 
 
HSG4112 markedly enhances energy expenditure in HFD-induced obese mice
In order to confirm in vivo the energy expenditure-enhancing effect of HSG4112, we placed HFD-induced obese mice treated with HSG4112 or vehicle for 4 weeks into open-circuit indirect calorimetry cages [25] to measure respiration, movement, and energy consumption for two consecutive days. HSG4112 significantly increased the overall oxygen consumption rate and carbon dioxide production rate during both light and dark hours (Fig. 3a, b). The respiratory exchange ratio (RER), which is an indicator of predominant substrate utilization for energy usage as either carbohydrate or fat, did not change significantly (Fig. 3c). HSG4112 also significantly increased the overall energy expenditure (Fig. 3d) in both light and dark hours – physically active and inactive hours –without increasing physical activity (Fig. 3e), suggesting increased basal metabolism rate. Dorsal surface temperature, measured by infrared thermography, was not affected by HSG4112 administration (data not shown). Overall, the above results show that HSG4112 increases energy expenditure in vivo.
 
HSG4112 regulates metabolic gene expression towards increased energy expenditure in liver and muscles
To further gauge the effects of HSG4112 on obesity, we investigated the expression levels of genes related to energy metabolism, leptin and insulin signaling, and inflammation. A total of 68 genes were selected through literary search [26–28], and we performed qRT-PCR on mRNA extracted from the hypothalamus, liver, gastrocnemius muscle, and interscapular tissue (BAT) of the HSG4112 group at terminal sacrifice after 6 weeks of treatment. All genes with significant difference in their expression level are shown in Fig. 4a-d; their list and primer sequences are available in Supplementary Table 1.
Amongst the muscle, liver, hypothalamus, and BAT, expressions of the selected genes were most affected by HSG4112 in the muscles and liver. In muscles, genes related to fatty acid oxidation (ACOX1, PRKAA2) were significantly upregulated, while genes related to glucose transport and metabolism (SLC2A4, PDK4) were normalized after HSG4112 treatment (Fig. 4a). In the liver, genes related to glucose metabolism (FGF21, PEPCK), insulin signaling (FOXO1, HNFA4), and lipid metabolism (CREB3L3, APOA5, SCD1) were induced by HSG4112 to either be normalized or increase towards the direction of increased energy expenditure (Fig. 4b). Relatively few changes were observed in the hypothalamus and BAT. In the hypothalamus, metabolism enhancing TTR expression was upregulated, and the PTPN1 level denoting insulin resistance was downregulated (Fig. 4c), while NPY, AGRP, POMC, CART levels denoting leptin resistance showed a normalizing trend (Supplementary Fig. 3). In BAT, the LEP level was robustly normalized (Fig. 4d). The mRNA level of UCP1, which is one of the proteins well-known for enhancing energy expenditure through futile cycling of mitochondrial potential and consequent thermogenesis in BAT [29], was not influenced by HSG4112 in all tissues. PGC-1ɑ, the master regulator of mitochondrial biogenesis [30], was consistently upregulated in the BAT, liver, and muscles, suggesting a possible role of this gene or its pathway in the mechanism of action of HSG4112. Inflammation-related genes (CCL2, NFE2L2) also had significant change in expression in the liver and muscles.
The majority of the affected genes suggested enhancement of energy expenditure as HSG4112’s mode of action. Therefore, we further investigated whether HSG4112 acts through two well-known enhancers of energy expenditure – AMPK and UCP1 – by measuring their protein levels at an early timepoint of the 11-day administration, when body weight just begins to decrease. We found that phosphorylated hypothalamic AMPK markedly decreased to approximately half-fold in the HSG4112 group (Supplementary Fig. 4); decrease in hypothalamic phospho-AMPK signifies activation of peripheral AMPK signaling and increased energy metabolism [31]. Consistent with the transcriptome data, HSG4112 had no effect on the UCP1 protein level in BAT (Supplementary Fig. 4), confirming that HSG4112’s mode of action does not converge with UCP1 activation.
 
HSG4112 ameliorates features of fatty liver in HFD-induced obese mice
After observing the robust effect of HSG4112 on adiposity, we examined parameters associated with fatty liver, such as liver weight, liver histology, and blood markers of liver injury. This approach was based on the knowledge that obesity is often accompanied by and causal to the development of nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis (NASH), whose prevalence is dramatically growing worldwide without any approved pharmaco-therapeutic drug up to the current date. Fatty liver disease is characterized by varying degrees of hepatic steatosis, lobular inflammation, cell ballooning, and fibrosis [32].
At 100 mg·kg-1 dose, HSG4112 normalized the absolute liver weight and histology in HFD-induced obese mice (Fig. 5a, b). In terms of the NAFLD activity score (NAS), HSG4112 fully reduced the steatosis and inflammation score, and the concurrent overall NAS of HFD-induced obese mice, down to the normal group’s level (Fig. 5c-e). Fibrosis was not significantly induced in the vehicle group (Fig. 5f), and hepatocyte ballooning was not induced at all in all groups (data not shown). Together, HSG4112 improved liver histology mainly through the resolution of inflammation and steatosis.
In terms of blood parameters, treatment of HSG4112 led to reducing trends in serum triglyceride (Fig. 5g) and HDL cholesterol levels (Fig. 5i), and significantly reduced and normalized LDL cholesterol (Fig. 5h) and total cholesterol levels (Fig. 5j); serum triglyceride and HDL cholesterol levels in murine models are challenging to interpret and to apply to fatty liver status [33]. The serum AST level showed a decreasing trend (Fig. 5k) and the ALT level was significantly reduced (Fig. 5l). In sum, HSG4112 normalized most of the relevant blood parameters, indicating robust amelioration of fatty liver in HFD-induced obese mice.