Liquid chromatography-tandem mass spectrometry (LC-MS) grade methanol (MeOH) and acetonitrile (ACN) were purchased from Fisher Scientific (Loughborough, UK). Formic acid was obtained from TCI (Shanghai, China). 2-Amino-3-(2-chlorophenyl) propionic acid (internal standard) and ammonium formate were obtained from Sigma Aldrich (Shanghai, China). Hematoxylin plus Eosin dye kit (DH0006) was purchased from Leagene (Beijing, China). Ultrapure water was generated using a Milli Q system (Millipore, Bedford, USA).

MHP is mainly consisted of mulberry leaf aqueous extract, Hippophae protein peptide and L-arabinose. In the present study, MHP (batch No. 20220623) was manufactured in a Good Manufacturing Practice (GMP) pilot plant following specific quality assurance instructions for the nutritional ingredients, heavy metals and microorganisms (see Supplementary Table S1 in Supplementary Materials ).

Male Sprague-Dawley rats (5-6 weeks old) were supplied by Guangdong Medical Laboratory Animal Center (Guangzhou, China). Animal study protocols were approved by the Institutional Animal Care and Use Committee of the Guangdong Pharmaceutical University. All animal studies were conducted in accordance with the ARRIVE (Animal research: reporting of in vivo experiments) guidelines (23). All animals were maintained under specific pathogen-free conditions, housed with 12 h light/dark cycles at controlled 22-24°C and 60-65% humidity. After acclimatization on a normal chow diet for 7 days, the rats were randomly divided into control group and high calorie diet group. As previously described (22, 24), the control group was fed with standard chow and the high calorie diet group was fed with a high-fat diet plus HFCS-sweetened drink. Two weeks later, an elimination about 20% of obesity-resistant rats was performed and the calorie diet group was divided into HFF group, MHP-H and MHP-L group (n=8 for each group). Except the rats of normal group, all the other rats were fed with a high-fat high-fructose (HFF) diet, and the rats of MHP-H and -L group were supplemented with MHP at a dose of 5.25 g/kg/day and 0.875 g/kg/day by gavage respectively.

The intervention lasted for 8 weeks. Body weight of each rat was measured twice per week throughout the experiment. In addition, food and drink intake were monitored, and the energy intake and energy efficiency (body weight gain(g)/total energy intake (kcal)) of each group were calculated.

At the end of the experiment, all animals were anaesthetized with pentobarbital sodium after an overnight (12h) fast. Blood and tissues were harvested as previously described (25). Blood was collected for serum metabolomics analysis. The inguinal white adipose tissue (iWAT) was also collected and weighted. Part of iWAT was fixed with 4% PFA for histological examination, while the rest was snap frozen in liquid nitrogen. Samples were frozen at -80°C until analyzed for metabolomics (plasma) or transcriptomics (iWAT tissue).

Body composition analysis was carried out as previously described using an NMR analyzer Minispec LF90II body composition analyzer (Bruker Optics, Billerica, MA, USA) 1 day prior to sacrifice (25). The adiposity, the ratio of lean mass to body weight, and the ratio of fluid mass to body weight were calculated.

The untargeted metabolomics analysis was performed as described (26). All the samples were thawed at 4°C. To 100 µL of serum, 400 µL of methanol was added and vortexed for 1 min. The mixture was then centrifuged for 10 min at 12,000 rpm and 4°C, and the supernatant was transferred, concentrated and dried in vacuum. Next, the dry samples were redissolved with 150 µL of 80% methanol solution containing 2-chlorolphenylalanine (4 ppm) and filtered through 0.22 μm membrane prior to LC-MS detection. A small aliquot of each sample was pooled together as the quality control sample.

The LC analysis was performed on a Vanquish UHPLC System (Thermo Fisher Scientific, USA) using an ACQUITY UPLC^® HSS T3 (2.1×100 mm, 1.8 µm) column (Waters, Milford, USA). The column was maintained at 40 °C. The flow rate and injection volume were set at 0.3 mL/min and 2 μL, respectively. For LC-ESI (+)-MS analysis, acetonitrile containing 0.1% formic acid (B1) and water containing 0.1% formic acid (A1) were used as mobile phase, while for LC-ESI (–)-MS analysis, pure acetonitrile (B2) and 5 mM ammonium formate (A2) were used. Chromatographic separation was carried out under the following gradient: 0~1 min, 8% B1(B2); 1~8 min, 8%~98% B1(B2); 8~10 min, 98% B1(B2); 10~10.1 min, 98%~8% B1(B2); and 10.1~12 min, 8% B1(B2). Simultaneous MS and data-dependent MS/MS acquisition of metabolites mass spectrometry data were performed on Orbitrap Exploris120 with ESI ion source (Thermo Fisher Scientific, USA). The sheath gas pressure and auxiliary gas were 40 and 10 arb, respectively. The spray voltage was 3.50 kV and -2.50 kV for ESI (+) and ESI(-), respectively. The capillary temperature was 325 °C. MS¹ scan range of m/z was from 100 to1000 and MS¹ resolving power was set at 60000 FWHM. The number of data dependent scans per cycle was set at 4. The MS/MS resolving power was set at 15000 FWHM. The normalized collision energy was 30 eV.

The raw data were first converted to mzXML format by MSConvert in ProteoWizard software package (v3.0.8789) and processed using R XCMS (v3.12.0) for feature detection, retention time correction and alignment (27, 28). The metabolites were identified by accuracy mass and MS/MS data which were matched with HMDB (http://www.hmdb.ca), KEGG (https://www.genome.jp/kegg/), mzcloud (https://www.mzcloud.org) and the in-house metabolites database of Panomix Biomedical Tech Co., Ltd. (Shuzhou, China).

The metabolic pathway and function enrichment analysis was proceeded as in previous studies (25, 29). PCA (principal component analysis) and OPLS-DA (orthogonal partial least squares discriminant analysis) were applied to discriminate the groups by R ropls (v1.22.0) package (30). In order to screen the differential metabolites (DMs), p value < 0.05 and VIP (variable projection importance) value > 1 were set by fault as the threshold. Using the generated DMs, pathway enrichment analysis was performed using the online metabolomics tool of Biodeep Platform (http://www.biodeep.cn/) to visualize and interpretate the data.

RNA-Seq was performed as reported previously (22). Briefly, RNA from the iWAT was extracted using Trizol reagent kit (Invitrogen, USA) following the manufacturer’s instruction. The cDNA library was constructed and sequenced on the Illumina (Novaseq 6000) by Gene Denovo Biotechnology Co. (Guangzhou, China). After the raw data were filtered, the clean reads were used for identification the differentially expressed genes (DEGs) by DESeq2 package. The level of mRNA FPKM (fragments per kilobase of exon model per million fragments mapped reads) value was also calculated. Genes with p value < 0.05 and |Fold change|> 1.5 were categorized as DEGs, and then used for gene ontology (GO) function analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. Gene Set Enrichment Analysis (GSEA) was implemented on the Java GSEA platform while gene sets with NES absolute value >1, p value <0.05 and FDR value < 0.25 were considered statistically significant. All the raw data generated by RNA-Seq were submitted to the GenBank’s Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra/PRJNA1063764).

All data are expressed as means ± standard deviation (SD). A Student’s t-test or one-way analysis of variance (ANOVA) was carried out. p value < 0.05 was considered statistically significant. R (https://www.rproject.org) and Graphpad Prism 6.0 software (GraphPad, CA, USA) were used for graphics.

In order to address the efficacy of MHP for weight loss, a serial of experiments were performed on a HFF-induced rat model of obesity ( Figure 1A ). As shown in Figures 1B, C , the body weight of the HFF group rats was significantly increased compared with normal rats, but 8-week intervention with a high dose of MHP ameliorated the body weight gain caused by HFF diet.

Since obesity is associated with excessive fat accumulation, the lean mass and fluid mass of each group rats were also measured. As expected, a high dose of MHP treatment significantly decreased the adiposity and increased the lean index, while there was no difference in the ratio of fluid to body weight ( Figures 1D–F ). In order to investigate whether the increase of body weight or WAT in rats was related to food, drink or total energy intake, we measured the calory intake and calculated the energy efficiency. It was found that there was no difference in total energy intake between the HFF group and MHP group, suggesting that the weight loss was not due to a reduced total energy intake and that MHP did not inhibit the rat appetite ( Figures 1G–J ).

The metabolic profiles of animals are dynamically changing, which can reflect the physiologic or pathologic conditions. Although it is complicated, metabolomics analysis can present some useful hints for mechanistic investigation (31). To study the underlying mechanisms for MHP efficacy, we performed a LC-MS based metabolomic profiling of serum collected from all the three groups of rats. Considering their efficacy, the MHP-H group was chosen for metabolomics analysis. Representative total ion chromatograms of LC-MS profiling of Cont, HFF and MHP groups of rats under positive ion mode are shown in Figure 2A . Mass spectrometry-based metabolomics studies require quality control to obtain reliable and high-quality metabolomics data (32). For this, the PCA score plots were used for both positive and negative modes ( Supplementary Figures S1A, B ). The PCA (R²X = 0.525) and PLS-DA score plots (R²X =0.505, R²Y =0.999, and Q^{2 =} 0.936) generated by multivariate statistical analysis based on acquired data indicated that the metabolome of the three groups of rats were distinct from each other ( Figures 2B, C ).

The volcano plot of the metabolites represented by MS data also showed obvious difference between the control and HFF group ( Figure 2D ), so do the HFF and MHP groups ( Figure 2G ). In order to maximize the discrimination between every two groups and screen out the differential metabolites, OPLS-DA analysis was carried out. As a result, in the OPLS-DA score plot of Cont versus HFF group, clear difference was presented while the cumulative R²X, R²Y, Q² value was 0.488, 0.998 and 0.762, respectively ( Figure 2E ). Similarly, HFF group was also different from MHP group in the OPLS-DA score plot with the cumulative R²X, R²Y, Q² values being 0.479, 0.98, and 0.745 respectively ( Figure 2H ). R²X and R²Y are known to be the cumulative model variation in X and Y, while Q² is the cumulative predicted variation. In general, the values of these parameters that are approaching 1.0 indicate that a model is stable and has predictive reliability. A permutation test with 200 iterations confirmed that the constructed OPLS-DA model was valid and not over-fitted, as the original R² and Q² values to the right were significantly higher than the corresponding permutated values to the left (33). The value of R² was 0.99 and Q² was 0.09 in Cont versus HFF group ( Figure 2F ), while R² was 0.93 and Q² was 0.0 between HFF and MHP group ( Figure 2I ).

Similarly, the LC-MS data under negative ion mode of metabolic profiles were acquired and analyzed. Representative total ion chromatograms of Cont, HFF and MHP groups were shown in Figure 3A . The PCA (R²X =0.533) and PLS-DA (R²X =0.226, R²Y =0.995 and Q^{2 =} 0.914) score plots showed clear differences among the Cont, HFF and MHP groups ( Figures 3B, C ). The volcano plots originated from ESI(-)-MS data also showed distinct difference between Cont and HFF groups ( Figure 3D ) as well as between HFF and MHP groups ( Figure 3G ). Following the aforementioned procedure, the OPLS-DA score plot of Cont versus HFF group also showed clear difference with the cumulative R²X at 0.248, R²Y at 0.999, and Q2 at 0.876 ( Figure 3E ). The cumulative R²X, R²Y and Q² value were 0.231, 0.996, and 0.735 respectively between HFF and MHP groups ( Figure 3H ). Moreover, a permutation test confirmed that the present OPLS-DA model was valid and not over-fitted. The detailed parameters were as the following: R² was 0.97 and Q² was 0.03 in Cont versus HFF group ( Figure 3F ), while R² was 0.98 and Q² was 0.03 between HFF and MHP groups ( Figure 3I ).

Metabolome reflects the holistic condition of an organism, presenting a global manifestation by a list of metabolites which are inextricably linked in terms of origination and metabolic pathways involved. Taking VIP > 1 and p < 0.05 as threshold values, a total of 73 serum small molecules were identified as the DMs between the control and HFF group rats when the ESI(+) and ESI(-) data were pooled together ( Supplementary Table S2 in Supplementary Materials ). As shown in Figure 4A , the DMs clustered into two distinct groups in the heatmap, which showed that the serum metabolome was greatly altered after HFF diet consumption.

Individually, each metabolite involves in specific biochemical pathway and plays an important role in the delicate life process. For example, when compared with Cont group, the contents of oleic acid and uridine in serum of rats were considerably upregulated (fold change, FC >3.0) while α-tocotrienol was significantly downregulated (FC <0.5) in HFF group. In fact, oleic acid can induce oxidative damage and steatosis in hepatocytes both in vitro and in vivo (34). Uridine is one metabolite positively associated with obesity, and is also a potential biomarker of insulin resistance (35, 36). Tocotrienols possess powerful antioxidant, anticancer, and cholesterol-lowering properties (37), and tocotrienols supplementation not only inhibits body weight gain but also attenuates oxidation stress in HFD-treated mice (38). On the whole, however, the significantly changed DMs should be taking into account for a living organism. To this end, Cytoscape, MetaboAnalyst, MetaCore and other programs can generate algorithmic networks which can present more insights than the blunt results analyzed individually (25, 39). Thus, in order to illustrate and visualize the underlying metabolic stress induced by HFF diet, the above DMs were further analyzed using MetaboAnalyst 5.0 (www.metaboanalyst.ca) to interpret the apparently independent changes into altered canonical pathways and metabolic regulation networks. As shown in Figure 4B , pathway enrichment results suggest that the changes of metabolic profile in rats were related principally to (1) PPAR (peroxisome proliferator-activated receptors) signaling pathway, (2) linoleic acid metabolism, (3) renal cell carcinoma, and (4) other pathways. The metabolic network generated with DMs is visualized by Cytoscape and shown as Figure 4C . PPARs, including PPARα, PPARγ and PPARβ/δ, are fatty acid-activated transcription factors of the nuclear hormone receptor superfamily that regulate energy metabolism. PPAR signaling pathway is critical to coordinate many cellular events during normal and stress conditions (40).

Likewise, using VIP > 1 and p < 0.05 as the thresholds, OPLS-DA generated 76 serum DMs between the HFF challenged rats and MHP supplemented rats based on the integrated ESI(+) and ESI(-) data ( Supplementary Table S3 in Supplementary Materials ). The DMs can also be clustered into two distinct groups in the heatmap ( Figure 5A ), which demonstrates that MHP intervention can ameliorate the metabolism stress induced by HFF consumption. Among them, some individual metabolites are interesting. For instance, compared with HFF group, 6-hydroxyhexanoic acid was remarkably upregulated in MHP group (FC>3.0), which can significantly reduce high-fat diet induced weight gain and improve glucose intolerance and insulin resistance in mice (41). While the content of uric acid was significantly downregulated in MHP group (FC <0.5), and recent studies have confirmed that high level of serum uric acid is associated with obesity and metabolic syndrome (42). However, an algorithmically constructed metabolic network presents more insight than a serial of the individual metabolites (25). Therefore, the above DMs were also used to perform pathway enrichment analysis by MetaboAnalyst 5.0. As shown in Figure 5B , the changes of metabolic profile in rats can be related principally to (1) PPAR signaling, (2) renal cell carcinoma, (3) central carbon metabolism in cancer, and (4) other pathways. The metabolic network generated with DMs between HFF and MHP groups was shown as Figure 5C . Unexpectedly, in both Figures 4B and 5B , PPAR signaling pathway seems to be the most impacted pathway, suggesting that this pathway is critical to coordinate the whole metabolism. Although more evidence may be needed, these results suggest that the efficacy of MHP is closely related to its regulation on the PPAR signaling pathway, which is worthy of further investigations.

Abnormal WAT is a phenotypic indicator of over-weight or obesity. In the present study, HFF diet intake led to an enlarged iWAT ( Figure 6A ) and an increase in the ratio of iWAT weight to body weight ( Figure 6B ). Therefore, we expect that MHP supplementation may attenuate excessive lipid deposition. Indeed, pathohistological examination of iWAT sections, both with H&E staining and Masson’s trichrome staining, showed that the rats in MHP-H group had less hypertrophic adipocytes as compared with HFF group ( Figures 6C, D ) and collagenous fibrosis ( Figures 6E, F ). These results confirmed the efficacy of MHP on weight loss.

As metabolome changes should originate from its upstream effects, including proteome and transcriptome changes, thus we performed RNA-Seq analysis of iWAT to verify the influence of HFF on the transcriptomic profiles in white adipose tissue ( Figure 6G ). As shown in Figure 6H, I , an obvious difference was found between the control and HFF groups as well as between the HFF and MHP-H groups. By setting p < 0.05 and |Fold change| > 2.0 as thresholds, a total of 5095 DEGs and 7907 DEGs were screened out between control and HFF groups and between HFF and MHP groups, respectively (see Supplementary Figure S2 in Supplementary Materials ). Furthermore, KEGG pathway enrichment analysis of these DEGs showed that several pathways, including CAMP signaling, glycolipid metabolism, regulation of lipolysis in adipocytes, PPAR signaling, insulin resistance and MAPK signaling, were significantly changed by HFF diet ( Figure 6J ), confirming that HFF significantly exerted influence on carbohydrate, lipid and energy metabolism of WAT as shown in our previous report on hepatic transcriptome (25). As expected, the high-dose MHP supplementation relieved this situation ( Figure 6K ) – nearly all involved pathways were redressed after MHP intake, suggesting that MHP could attenuate white adipose tissue transcriptome stress in HFF-diet-induced obese rats.

Storage of extra calories is an inborn ability of white adipose tissue. In the case of overnutrition, white adipose tissue usually experiences two scenarios, i.e., hyperplasia and hypertrophy. Adipocyte hyperplasia is commonly regarded as a healthy adaptation whereas adipocyte hypertrophy is associated with increased hypoxia which is regarded as an unhealthy state (43). In general, adipocyte hypertrophy is achieved in mature adipocytes via an increase in lipid accumulation or lipogenesis (i.e., triglyceride synthesis), which is synergetic with a decreased lipolysis (i.e., triglyceride breakdown). In contrast, adipocyte hyperplasia relies on a complicated adipogenesis process (44). The nuclear hormone receptor PPARγ is regarded as the master regulator of adipogenesis as it is indispensable for adipocyte differentiation. Mechanistically, TGF-β signals through SMAD3 and directly inhibits PPARγ-C/EBPα complex formation and, in consequence, adipogenesis (45).

TGF-β and PPARγ signaling pathways play a crucial regulatory role in adipocyte hyperplasia and hypertrophy, which may be a key target for the treatment of obesity ( Figure 7A ). Take PPARγ for instance, some of its ligands have been developed as clinically effective antidiabetic drugs (46, 47). Therefore, GSEA analysis on TGF-β signaling pathway and PPARγ signaling pathway in adipose tissue transcriptome were conducted. As compared to the control group, the HFF group exhibited a significant downregulation trend, which was reversed after MHP treatment ( Figures 7B, C ). The heatmap of genes related to adipocyte hyperplasia and hypertrophy reflected an obvious difference between the Cont and HFF groups, while, after MHP treatment, mRNA levels of differential genes tended to be normal ( Figure 7D ). As a transcription factor, PPARγ activates expression of various genes involved in fat deposition, such as Lpl, Fatp1, CD36, Fabp4, and Plin, by binding a functional peroxisome proliferator response element (PPRE) on their promoter in adipose tissue (48). In this study, MHP significantly suppressed the expression of Pparγ and its downstream target genes Fabp4 and Plin1 induced by HFF diet, suggesting that MHP may improve fat deposition by regulating the PPARγ pathway ( Figure 7E ).

In consideration of their structures and functions, the TGF-β superfamily ligands can be divided into two subfamilies: TGF-βs and BMPs (49). In our study, Tgfβ3, the key gene involved in adipocyte proliferation, exhibited a significant upregulation trend following MHP treatment ( Figure 7E ), indicating that MHP may improve the state of adipocyte hypertrophy by inhibiting TGFβ3-mediated adipocyte proliferation. Several genes involved in adipogenesis and lipid accumulation, for example, Smad1/2 and Bmp2/4, were robustly upregulated in the adipocytes under the HFF diet, but there was a significant inhibition of Bmp2/4 expression upon MHP intervention ( Figure 7E ). BMPs are pleiotropic proteins that regulate processes like cell-fate determination, proliferation, apoptosis and differentiation (50). BMP2 stimulates adipogenesis in 3T3-L1 cells by induction of PPARγ via mainly Smad-1/5/8 and p38 MAPK pathway (51). BMP4 promotes MSCs commitment to the adipocyte lineage and induces adipogenesis in a dose-dependent manner (52). Therefore, our findings suggest that MHP may exert its effects by modulating TGF-β and PPARγ signaling pathways, thereby promoting adipocyte proliferation and inhibiting adipogenesis. Consequently, this could alleviate adipocyte hypertrophy and improve the unhealthy obese conditions.

Obesogenic adipogenesis is a pathological process which is closely related to ECM (extracellular matrix) (53). In physiological states, a balance between synthesis, deposit and degradation of ECM components exists. However, diet-induced obesity is usually accompanied with ongoing ECM remodeling, which likely results from an excess synthesis of fibrillar components, such as collagens I, III and VI, and an impaired degradation of these proteins (54). Excessive ECM deposition in adipose tissue is a hallmark of fibrosis, which is a major contributor to adipose tissue dysfunction (55). As described, Masson’s trichrome staining adipose tissue sections showed a higher content of collagen fiber in HFF-diet challenged rats as compared to the control group, but MHP reversed its progression ( Figure 6C ). Thus, we reasoned that the health benefit of MHP may also be due to its improvement on the adipose dysfunction ( Figure 8A ). In other words, MHP intervention may reduce ECM deposition as well as fibroblasts of adipocyte, consequently improving adipose tissue fibrosis challenged by HFF diet.

Since ECM remodeling is a complicated and dynamically changing process, we performed mRNA expression enrichment analysis of the gene set of extracellular matrix organization ( Figure 8C ). As expected, HFF diet changed the expression of ECM organization pathway genes whereas MHP intervention reversed this condition. As the matrix metalloproteinases (MMP) are regarded as marker proteins involved in adipose ECM remodeling, mRNA expression levels of MMP genes were also compared. As shown in Figure 8G , significant upregulation expression of MMP-2 and MMP-14 genes was found in HFF group rats, while it was normal in MHP group. It has been shown that MMP-2 is a key regulator of adipocyte differentiation, and high-expression of MMP-2 is necessary for adipocytes hyperplasia when responding to excess energy from HFF diet (56). Also, MMP14 is found dramatically upregulated in adipose tissue of obese mice induced by high-fat diet (57). Therefore, the efficacy of MHP is likely due to its protection against adipose tissue dysfunction.

Crosslinking of collagen, the main component of connective tissue surrounding adipocytes, is known to affect adipose remodeling, which is crucial for maintaining function and metabolic homeostasis of adipose tissue (58). In fact, increased interstitial fibrosis in WAT may lessen the flexibility of ECM and decrease the plasticity of adipose tissue as obesity progresses, which ultimately leads to adipose tissue dysfunction (55). Through GSEA analysis of the iWAT transcriptome, we found the trend back of crosslinking of collagen fibrils pathway after MHP intervention ( Figure 8B ). Similar results were also found in collagen formation pathway and assembly of collagen fibrils and other multimeric structures pathway ( Supplementary Figures S3A, B ).

In addition to ECM remodeling, adipose fibrosis also contributes greatly to adipose tissue dysfunction. We found that the signaling by FGFR1 pathway, which is associated with adipose tissue fibrosis, was markedly downregulated in the HFF group ( Figure 8D ), so did the downstream signaling of activated FGFR1 pathway ( Supplementary Figure S3C ). Interestingly, these two pathways were significantly upregulated after MHP treatment ( Figure 8D ). Further, the heatmap of the DEGs from signaling by FGFR1 pathway and crosslinking of collagen fibrils pathway suggests that MHP was able to improve HFF diet-induced obese adipose tissue fibrosis at the transcriptomic level ( Figure 8F ). As shown in Figure 8G and Supplementary Figure S3D , the critical genes involved in these two pathways in the HFF group, including Fgf1, Fgfr1, Col1α1, Col1α3, Col6α2, Lox and Loxl1/2/3/4 showed a marked upregulation trend as compared with the control group, but showed a substantial downregulation trend after MHP intervention.

Fgf1 is unique among the Fgf family in its ability to activate all four Fgfrs and their isoforms (59). It is reported that adipose Fgf1 is dramatically induced in HFD-fed mice (60) and expression of Fgf1 and Fgfr1 in the adipose tissue of HFD-fed and ob/ob mice is significantly elevated when compared with normal diet-fed and lean control mice (61). In line with these reports, the expression of Fgf1 and Fgfr1 was significantly increased in the iWAT of HFF-induced obese rats but was normal after MHP treatment ( Figure 8G ). Collagen type I, III, and VI are highly expressed in adipose tissue (62) and diet induced obesity is found to promote the expressions of Col1α1, Col3α1 and Col6α3 in eWAT of C57/B6J mice (58). Moreover, Col3α, Col6α, and Lox gene expressions are downregulated by metformin and resveratrol, which are shown to decrease collagen deposition in adipose tissue induced by HFD (63). In fact, Lox (Lysyl oxidase), which catalyzes the formation of allysine and hydroxylysine, is crucial for collagen fiber crosslinking and thus for fibrosis development (64). Berberine is also found to suppress the expression of Hif-1α and Lox in eWAT at the state of obesity induced by HFD (65). Therefore, the efficacy of MHP is potentially associated with its protection against adipose tissue fibrosis.

As aforementioned, the adipocyte hypertrophy is usually along with an insufficient local blood supply and a subsequent locally hypoxic microenvironment in adipose tissue (66). Therefore, the genes related to hypoxia response may have some expression alterations in our case. For this, GSEA analysis of the iWAT transcriptome on the response to hypoxia pathway was carried out ( Figure 8E ). As expected, supplementation of MHP resulted in a strong downregulation trend for the key genes related to hypoxia, including Hif-1α, Vegfr, Egfr and Igf1 ( Figure 8G , Supplementary Figure S3D ). Interestingly, an elevated expression of Hif-1α, the hypoxia-inducible factor-1 α, in obese adipose tissue is reported to trigger a potent pro-fibrotic transcriptional program, and inhibition or knockout Hif-1α can ameliorate the negative aspects of the obesity-associated fat pad expansion and adipose tissue fibrosis (66, 67).

Rapid tissue expansion during the development of obesity causes the neo-vasculature to struggle to keep up, and the mRNA expression of vascular endothelial growth factor (Vegf), a well-known Hif-1α target gene, is markedly elevated in HFD fed mice (55). On the other hand, hyperplastic expansion is preceded by angiogenesis but angiogenesis is often insufficient and occurs after hypertrophic adipocyte growth, which can result in hypoxia and tissue dysfunction (68, 69). Previous research has revealed that HFD increases expression of Egfr and its ligand amphiregulin in adipose tissue macrophages (ATMs). Furthermore, Egfr deletion in ATMs leads to an inhibition of resident ATM proliferation and monocyte infiltration into adipose tissue, which results in decreased obesity and insulin resistance (70). As described above, the expression of hypoxia-related genes was markedly increased in the HFF group, but it was normal after MHP intervention ( Figure 8G ), suggesting that the efficacy of MHP is related to its improvement on hypoxia condition of the hypertrophic adipocytes.