EGCG and Catechin Relative to Green Tea Extract Differentially Modulate the Gut Microbial Metabolome and Liver Metabolome To Prevent Obesity In Mice Fed A High-Fat Diet

Abstract

Green tea extract (GTE) alleviates obesity, in part, by modulating gut microbial composition and metabolism. However, direct evidence regarding the catechin-specific bioactivities that are responsible for these benefits remain unclear. The present study therefore investigated dietary supplementation of GTE, epigallocatechin gallate (EGCG), or (+)-catechin (CAT) in male C57BL6/J mice that were fed a high-fat (HF) diet to establish the independent contributions of EGCG and CAT relative to GTE to restore microbial and host metabolism. We hypothesized that EGCG would regulate the gut microbial metabolome and host liver metabolome more similar to GTE than CAT to explain their previously observed differential effects on cardiometabolic health. To test this, we assessed metabolic and phenolic shifts in liver and fecal samples during dietary HF-induced obesity. Ten fecal metabolites and ten liver metabolites (VIP > 2) primarily contributed to the differences in the metabolome among different interventions. In fecal samples, nine metabolic pathways (e.g., tricarboxcylic acid cycle and tyrosine metabolism) were differentially altered between the GTE and CAT interventions, whereas three pathways differed between GTE and EGCG interventions, suggesting differential benefits of GTE and its distinctive bioactive components on gut microbial metabolism. Likewise, hepatic glycolysis / gluconeogenesis metabolic pathways were significantly altered between GTE and EGCG interventions, while only hepatic tyrosine metabolism was altered between CAT and GTE interventions. Thus, our findings support that purified catechins relative to GTE uniquely contribute to regulating host and microbial metabolic pathways such as central energy metabolism to protect against metabolic dysfunction leading to obesity.

1. Introduction

The World Health Organization reports that 39% of adults worldwide are classified as overweight and 13% as obese [1]. The global prevalence of obesity has also nearly tripled from 1975 to 2016, which is concerning because obesity is a leading risk factor for certain metabolic diseases (e.g., diabetes, cardiovascular diseases), musculoskeletal disorders, and cancers [1]. Epidemiological studies suggest that the gut microbiome composition influences obesity risk [2], an association that is likely mediated through the gut metabolome. Indeed, gut microbiota composition in obese persons differs from that of healthy persons [3]. Not only does the gut microbiota affect host metabolism and human health [4–7], but the altered community structure of gut bacteria influences the abundance of gut microbial metabolites [8] that potentially regulate obesity-associated disease risk.

Green tea is the most commonly consumed prepared beverages in the world [9]. Its polyphenolic catechins, including epigallocatechin gallate (EGCG), have been reported to reduce body weight, dietary fat absorption, circulating lipids (triglycerides, free fatty acids, cholesterol), as well as glucose and insulin in obese rodent models [10, 11]. Catechin-rich green tea extract (GTE) is also shown to inhibit obesity in association with altered gut microbiota composition and metabolism [12–16]. However, it is unclear whether catechins provide unique structure/function benefits on obesity risk. While the benefits of GTE are often attributed to EGCG because of its greatest abundance in green tea [17], lesser abundant green tea catechins and/or catechin metabolites of host or microbial origin may also contribute to altering gut microbial and host metabolism [18]. We reported that GTE compared with EGCG or (+)-catechin (CAT) more greatly protected against high-fat (HF)-induced obesity, whereas GTE and EGCG more favorably attenuated insulin resistance. Gut microbial diversity that was otherwise lowered in HF mice was maintained by GTE and CAT, but not EGCG. Further, analysis of genomic sequences indicated that the predicted microbial metabolic functions were more similar between GTE and CAT [19].

To further define the mechanism by which GTE and its catechins alleviate obesity risk, the present study investigated the independent benefits of EGCG and CAT relative to GTE on gut microbial metabolism and host liver metabolism in obesity. We used archived fecal and liver samples from our earlier study [19] to test the hypothesis that the previously observed differential anti-obesity effects mediated along the gut-liver axis by GTE, EGCG, and CAT are attributed to their unique influence on gut microbial and hepatic metabolic signatures. This hypothesis was tested using our well-established targeted mass spectrometry (MS)-based metabolomics workflow [20, 21] to discover the metabolic differences of gut microbes and at the liver in obese mice fed a HF diet containing either GTE, EGCG or CAT. We also applied a targeted MS approach to assess the major constitutes of GTE and its associated metabolites in fecal and liver samples. Our results indicated that the microbial metabolic responses corresponded to treatment-specific phenotypic responses on HF-induced obesity. Outcomes of this metabolomics investigation are therefore expected to advance an understanding of the connections between the metabolic benefits of GTE/EGCG/CAT to alleviate obesity risk.

2. Materials and methods

2.1. Study design

The study design has been detailed by our group previously [19], and all procedures were approved by The Ohio State University Institutional Animal Care and Use Committee (Protocol #2012A00000156-R2). Briefly, male C57BL/6J mice (n= 50; 5 weeks old; Jackson Laboratory) were acclimated to the environmentally controlled facility for one week, and randomized (n = 10/group) to receive a low-fat (LF) diet, a high-fat (HF) diet, or the HF diet containing powdered GTE (HF+GTE), EGCG (HF+EGCG), or CAT (HF+CAT) for 8 weeks. The composition of the LF (#D12450J, Research Diets) and HF diets (#D12492; Research Diets) was detailed previously, and contained 10% and 60% energy from fat, respectively [22]. The basal HF diet used in these studies was based on reports that it induces obesity within 8 weeks [22, 23] and increases NFκB activation in liver [19, 24–26] and adipose tissue [23]. Female mice were not included due to their known resistance to diet-induced obesity and insulin resistance [27].

GTE containing ~30% w/w total catechins (45.0% EGCG; 31.6% epigallocatechin; 12.1% epicatechin gallate; 9.4% epicatechin; and 1.9% CAT) and 5.1% caffeine (w/w) was obtained from Unilever BestFoods. The HF diet was formulated to contain GTE at 2% (w/w) in the present study based on reports that it alleviates HF-induced increases in body mass, adiposity, and insulin resistance, and decreases hepatic and adipose NFκB activation in HF-diet-fed rodents [19, 23, 25, 26]. In a 75 kg human, and based on the typical catechin content (~180 mg/serving) of fresh brewed tea [28], dietary GTE at 2% corresponds to a human dose of 24 mg/kg of total catechins. GTE at 2% also corresponds to 10 servings per day in humans based on the presence of 105 mg of EGCG in a cup of brewed green tea and humans consuming 2000 kcal/d [29]. Intakes of green tea at high levels are common in certain regions, including that described in observational studies suggesting that ≥10 servings per day among Japanese men is inversely related to liver injury and metabolic abnormalities [30].

EGCG (>94%, Teavigo^®, Taiyo International) was supplemented at 0.3% (w/w) in the HF diet, to match the abundance of EGCG in the HF-GTE diet. Likewise, based on EGCG representing ~50% of the total catechin content of green tea, and EGCG (0.3%) provided in the present study equivalent to the EGCG content of GTE, the human equivalent dose is 12 mg/kg of EGCG. Different from EGCG, CAT is the least abundant (0.01%, w/w) catechin in HF+GTE diet. It is also substantially divergent in chemical structure from EGCG in that it contains fewer reactive hydroxyl groups and is devoid of a gallate moiety. To directly compare their anti-obesity effects, CAT (≥98%) purchased from Sigma was supplemented at 0.3% (w/w) to match the EGCG content of the HF+EGCG diet, which corresponds to 12 mg/kg of the human equivalent dose.

After 8 weeks of dietary intervention, fecal samples were collected and then all mice were sacrificed in the fasted state (6–8 h). Terminal body masses of mice were obtained, and whole blood collected from the retroobital sinus was centrifuged (2000g; 15 min) to isolate serum. The measurements of serum glucose and insulin, calculated insulin resistance (homeostatic model assessment of insulin resistance; HOMA-IR), and nonalcoholic steatohepatitis (NASH) activity score were reported previously [31]. Adiposity was determined gravimetrically by weighing excised adipose tissue at the time of sacrifice [19]. Liver was also collected, and its triglyceride content was reported previously. All biospecimens were stored at −80°C (~6-mo) until MS-metabolomomic analyses of the present study were conducted.

2.2. Metabolic Profiling of Fecal and Liver Samples

LC-MS grade solvents for metabolite extraction were purchased from Fisher Scientific (Pittsburgh, PA, USA). Livers samples (~50 mg) or fecal samples (~10 mg) were weighed and homogenized in 400 uL 1:4 (v:v) water:methanol on dry ice. For fecal samples, after homogenization, 800 uL 1:4 (v:v) water:methanol was added and incubated for 30 min on dry ice. The mixture was centrifuged at 22,000 g for 5 min, and the soluble extract was transferred into a new tube and placed on dry ice. The resulting pellet was resuspended in 500 uL solvent with 1:4 (v:v) water:methanol. The mixture was again incubated for 15 min on dry ice and centrifuged at 22,000 g for 5 min. The resulting extract was combined with the initial extract in the new tube. Similarly, for liver samples, 800 uL 1:4 (v:v) water: methanol was added to the homogenized sample and incubated for 30 min on dry ice. The mixture was centrifuged at 22,000 g for 5 min, and the soluble extract was transferred into a new tube and placed on dry ice, the pellet was resuspended with 1500 uL 1:4 (v:v) water:methanol again to collect and combine the soluble extract. The soluble exact samples from feces and livers were completely dried under nitrogen gas using a Speedvac at 30°C. The samples were reconstituted with 250 μL 50% acetonitrile. After 2 min vortex, the samples were centrifuged for 5 min at 22,000 g, and 150 μL supernatant was transferred into LC vials for MS analysis.

The targeted metabolomics method, which was established by our group, was published previously [20, 21]. Briefly, a Thermo Scientific Ultimate 3000 HPLC system coupled with a Thermo Scientific TSQ Quantiva triple quadruple mass spectrometer was used, and the electrospray ionization (ESI) source was equipped with the mass spectrometer. Authentic metabolite standards (n = 221) were purchased from Sigma (Saint Louis, MO, USA) and IROA Technologies (Boston, MA, USA). External standards were used to confirm metabolite retention times and selected reaction monitoring (SRM) transitions on our HPLC-MS/MS in both positive and negative ionization modes in biological samples. The method is well-established and validated monthly using the same standards following the same procedures as MS method setup. Pooled samples, injected between every 10 biological samples, were used as quality control to monitor instrument stability. Xcaliber 4.0 software was used to control the LC-MS/MS system, and the targeted analysis was in SRM mode.

For liver and gut microbial metabolites analysis, a hydrophilic interaction chromatography (HILIC) column (Waters Corporation, Milford, MA, USA) with 2.1 × 150 mm, amide 2.5 μm (column that is packed with 2.5 μm Ethylene Bridged Hybrid [BEH] particles and Trifunctional Amide as ligand type) was used to separate metabolites. HPLC separation was performed isocratically (0.30 mL/min) with the auto-sampler thermostatted to 4°C and the column compartment to 40°C. Mobile phase A was 10:90 acetonitrile:water containing 5 mM ammonium acetate and 0.2% acetic acid, while mobile phase B was 90:10 acetonitrile:water containing 5 mM ammonium acetate and 0.2% acetic acid. The total run time for each injection was 20 min with the following gradient program: 0–2 minutes, 70% B; 2–5 minutes, 30% B; 5–9 minutes, 30% B; 9–11 minutes, 70% B; 11–20 minutes, 70% B).

2.3. HPLC–MS/MS analysis of polyphenol compounds

Liver samples (~50 mg) or fecal samples (~20 mg) were weighed and homogenized in 500 uL 10% ascorbic acid and 0.1% EDTA in 0.4M NaH₂PO₄ (pH = 3.6), and extracted as we described [32]. Samples were reconstituted with 150 μL 15% acetonitrile. After 2 min vortex, the samples were centrifuged for 5 min at 22,000 g, and 100 μL supernatant was transferred into LC vials for MS analysis. The same HPLC-ESI-MS system as described above was used to detect polyphenol compounds. Phenolic standards (n = 17) were purchased from Cayman Chemical (Ann Arbor, MI, USA) and an XTerra Shield RP18 Column (Waters Corporation, Milford, MA, USA) with 3.9 × 100 mm, amide 3.5 μm was used. HPLC separation was performed at 0.90 mL/min with the auto-sampler at 4°C and the column compartment at 40°C. Mobile phase A was 10% acetonitrile, 90% H₂O and 0.1% formic acid, while mobile phase B was 30% acetonitrile, 70% H₂O and 0.1% formic acid. The separation gradient was: 0 minute, 1% B; 8 minutes, 99% B; 10 minutes, 1% B; 12 minutes, 1% B. The detailed MS detection parameters of these phenolic compounds are listed in Table 1.

Table 1.

LC-MS detection parameters for the target phenolics

Compound	Retention Time (min)	RT Window (min)	Polarity	Precursor (m/z)	Product (m/z)	Collision Energy (V)	RF Lens (V)
Epigallocatechin gallate	5.48	2	Negative	457.183	125.04	39.376	0
Epigallocatechin gallate	5.48	2	Negative	457.183	168.857	17.23	0
5-(3’-Hydroxyphenyl)-gamma-valerolactone	5.50	11	Negative	191.2	41	40	106.382
5-(3’-Hydroxyphenyl)-gamma-valerolactone	5.50	11	Negative	191.2	99	20	106.382
m-Coumaric Acid	6.69	2	Negative	163.19	91.2	25.472	57.191
m-Coumaric Acid	6.69	2	Negative	163.19	119	16.522	57.191
Epicatechin gallate	6.84	2	Negative	441.061	245.026	26.534	0
Epicatechin gallate	6.84	2	Negative	441.061	258.958	20.82	0
Quercetin 3-D-galactoside	6.95	2	Negative	463.141	271	43	113
Quercetin 3-D-galactoside	6.95	2	Negative	463.141	299.915	26.584	113
Hesperetin	7.21	2	Positive	303.183	153.071	26.028	75.73
Hesperetin	7.21	2	Positive	303.183	177.125	17.938	75.73
Quercetin	8.17	2	Negative	301.048	151.125	24	117
Quercetin	8.17	2	Negative	301.048	179.08	18	117
Homovanillic acid sulfate	2.81	2	Negative	261.03	137.04	23.601	69.056
Homovanillic acid sulfate	2.81	2	Negative	261.03	181.054	16.624	69.056
Epigallocatechin	3.17	2	Negative	305.1	125.054	22.185	0
Epigallocatechin	3.17	2	Negative	305.1	179.071	15.916	0
3-Hydroxyphenylacetic acid	3.79	2	Negative	151.152	65.111	24.865	35.933
3-Hydroxyphenylacetic acid	3.79	2	Negative	151.152	107.111	10.253	35.933
5-(3’,4’-Dihydroxyphenyl)-valerolactone	4.07	2	Negative	207.152	122.111	19.607	0
5-(3’,4’-Dihydroxyphenyl)-valerolactone	4.07	2	Negative	207.152	161.04	23.551	0
Gallic acid	1.8	2	Negative	169.03	69.2	25.118	79.191
Gallic acid	1.8	2	Negative	169.03	79.183	16.674	79.191
3,4-Dihydroxyphenol ethanol	2.07	2	Negative	153.152	95.111	20.365	59.416
3,4-Dihydroxyphenol ethanol	2.07	2	Negative	153.152	123.111	14.197	59.416
3,4-Dihydroxyphenylacetic acid	2.46	2	Negative	167.122	123.111	10.253	39.64
3,4-Dihydroxyphenylacetic acid	2.46	2	Negative	167.122	152	15.107	39.64
3,4-Dihydroxybenzoic acid	2.76	2	Negative	153.333	91	25.624	61.393
3,4-Dihydroxybenzoic acid	2.76	2	Negative	153.333	109.125	15.562	61.393
Procyanidin B2	4.55	2	Negative	577.222	407.125	25.017	106.382
Procyanidin B2	4.55	2	Negative	577.222	425.125	15.461	106.382
Caffeic Acid	4.59	2	Negative	179.152	107.111	22	72
Caffeic Acid	4.59	2	Negative	179.152	135.054	15.258	72

2.4. Microbiome Analysis

The diversity and composition analysis of the gut microbiota were performed and the detailed method was reported in our previous study of these same mice [19]. Existing data from the previously conducted microbiome analysis was used in the present study for novel hypothesis testing that examined the association between gut community structure and gut or hepatic metabolic profiling.

2.5. Statistical analysis

Quan Browser in Xcaliber 4.0 software (Thermo Fisher Scientific) was used to process the raw LC-MS/MS data based on the retention time and SRM information, and the result was manually integrated and exported to Excel file [33, 34]. Phenotypic and metabolic health data (means ± SEM) were analyzed using GraphPad Prism V9.0. Group-wise differences were determined using 1-way ANOVA followed by Newman Keuls post-test when a significant effect was detected. Data having unequal variances were log-transformed to achieve equal variances, and p<0.05 was considered statistically significant. In MetaboAnalyst 5.0 (http://www.metaboanalyst.ca/), Partial Least Squares - Discriminant Analysis (PLS-DA) were applied to compare the metabolic profiles, and MetaboAnalyst Pathway Analysis was used to compare the metabolic pathways among different groups (with pathway impact>0.2 and −log(p)>5 used as cutoff values) [35]. Data were normalized by the weight of fecal or liver samples, log-transformed and auto-scaled to achieve the normal distribution.

3. Results

3.1. Anti-obesity effects of GTE and its catechins

In our parent study [19], we reported that HF mice had greater body mass and adiposity compared with LF mice. Without affecting energy intake [19], GTE fully protected against HF-induced obesity, while EGCG and CAT only partly protected protected against obesity. Although EGCG and CAT prevented HF-induced obesity to a similar extent, these treatments differentially affected HOMA-IR, serum glucose, and liver triglyceride [19]. HF mice had increased HOMA-IR, but only GTE and EGCG decreased HOMA-IR to levels not different from LF mice, whereas CAT only partly decreased HOMA-IR. Similarly, serum glucose and liver triglyceride were increased in HF mice. While GTE and EGCG fully inhibited HF-mediated increases in serum glucose and triglyceride, the effects of CAT were less pronounced. As reported in our parent study [19], histological evidence of HF-induced NASH, specifically liver steatosis was attenuated markedly in the HF+GTE group compared with HF controls.The composite NASH activity score that was otherwise increased in the HF group was attenuated in the HF +GTE, HF+EGCG and HF+CAT groups [19]. To understand these differential treatment effects between GTE and purified catechins, we hypothesized in this present study that regulation of the gut microbiome/metabolome and liver metabolome would be differentially affected by GTE, EGCG, and CAT.

3.2. GTE and purified catechins significantly regulate the gut microbial metabolome

In the present study, our targeted metabolomics workflow (n = 221 metabolites spanning diverse metabolic pathways) detected more than 100 metabolites in fecal samples of all five experimental groups (Table S1). Partial least squares-discriminant analysis (PLS-DA) permitted a visualization of the similarity/differences among the detected metabolic profiles in fecal samples of all sample groups. As shown in Figure 1A, treatment groups were clearly separated from LF and HF control groups, suggesting that specific catechins or intact GTE can significantly alter a variety of gut microbial metabolites. The different colored ovals represent the 95% confidence intervals of the separation. For PLS-DA component 1, the R-square was 0.70 and the Q-square was 0.61; while the R-square and the Q-square values for component 2 were 0.96 and 0.92, respectively. These values suggested a robust classification ability of the applied PLS-DA models to differentiate between the dietary treatments based on fecal metabolic profiles. Figure 1B shows the top 15 metabolites that primarily contributed to the separation in Figure 1A based on their variance importance projection (VIP) scores. Furthermore, several endogenous metabolites, namely glucosamine-6-phosphate, indole-3-acetaldehyde, 3-hydroxyanthranilate, glutathione, cyclic GMP, and homocysteine were considered as the primary contributing factors for the separation with VIP scores >2 (Figure 1B). Figure 1B also shows that homocysteine, glutathione, and indole-3-acetaldehyde were only detected in LF and HF groups, and were commonly downregulated in the GTE, EGCG, or CAT groups to the level below our detection threshold.

Fig. 1.

PLS-DA analysis and metabolic pathway analysis of fecal metabolic profiles from different groups. Each group contains 10 mice replicates. A. 2D score plot demonstrates the difference in gut microbial metabolic profiles among five study groups. B. Top 15 metabolites that contribute to the separation between five groups. C. 2D score plot demonstrates the difference in gut microbial metabolic profiles among GTE intervention (HF+GTE), EGCG intervention (HF+EGCG), and CAT intervention (HF+CAT) groups. D. Top 15 metabolites that contribute to the separation among HF+GTE, HF+EGCG and HF+CAT. The VIP score in x axis quantified the contribution. The color on the right represents the relative amount of each metabolites in each diet group. E. Major fecal metabolic pathways altered between HF+GTE and HF+EGCG. F. Major fecal metabolic pathways altered between HF+GTE and HF+CAT. The x-axis is the pathway impact value, and y-axis is the statistical significance of the impact. The dot size corresponds to x-axis, and the dot color corresponds to y-axis. a, tricarboxylic acid cycle (TCA cycle); b, Tyrosine metabolism; c, Glycine serine and threonine metabolism; d, Tryptophan metabolism; e, Glutamine and glutamate metabolism; f, Phenylalanine metabolism; g, Alanine aspartate and glutamate metabolism; h, Phenylalanine, tyrosine and tryptophan biosynthesis; i, Biotin metabolism

3.3. EGCG-induced changes in gut microbe metabolic profiles are more similar to the effects of GTE than CAT

The overall PLS-DA analysis that permitted visualization of differentiation between green tea treatments groups and the HF/LF groups was unable to fully visualize separation between the GTE, EGCG and CAT treatments (Figure 1A). Therefore, PLS-DA was again performed to specifically compare the gut microbial metabolic profiles among the HF+GTE, HF+CAT and HF+EGCG groups to inform detailed comparisons of their metabolic implications. Figure 1C clearly shows that EGCG treatment shifted the microbial metabolome away from the GTE treatment, while CAT shifted the microbial metabolome even further. For PLS-DA component 1, the R-square was 0.75 and the Q-square was 0.61, while for component 2, the R-square was 0.95 and the Q-square was 0.88. These values suggest that the fecal metabolic profiles between the three treatment groups are quite different, and that the treatment effect of EGCG on the gut microbial metabolome is more similar to that of the GTE treatment. Among the top 15 metabolites shown in Figure 1D, ten metabolites with a VIP score >2 were primarily contributing to the separation observed in Figure 2C. The color bar illustrates that 5-hydroxyindoleacetate, 3-methoxy-4-hydroxymandelate, and N-acetylserotonin were more abundant in HF+GTE; 3-methyl-2-oxindole was detected at the highest abundance in HF+EGCG; malate, threonine, adipic acid, nicotinamide, and pterin were most abundant in HF+CAT group.

Fig. 2.

PLS-DA analysis of phenolics from three treatment groups from fecal sample analysis. Each group contains 10 mice replicates. A. 2D score plot demonstrates the difference in phenolic profiles among GTE intervention (HF+GTE), EGCG intervention (HF+EGCG), and CAT intervention (HF+CAT) groups. B. Top 15 phenolics that contribute to the separation among HF+GTE, HF+EGCG and HF+CAT. The VIP score in x axis quantified the contribution. The color on the right represents the relative amount of each phenolic in each diet group.

We next performed metabolic pathway analysis to evaluate the impact of GTE, EGCG and CAT treatments on metabolic pathways of the gut microbial metabolome. All detected metabolites were included to construct metabolic networks towards the evaluation of the dietary treatments to influence metabolic pathways. Figures 1E–F and Figure S1 show the difference in major metabolic pathways of the fecal metabolome between HF+GTE and HF+EGCG, between HF+GTE and HF+CAT, and between HF+EGCG and HF+CAT, respectively. With a predefined threshold of pathway impact >0.2 and −log(p) >5, three metabolic pathways were significantly different between the HF+GTE and HF+EGCG groups: tricarboxylic acid cycle (TCA cycle, upregulated in HF+EGCG vs. HF+GTE), tyrosine metabolism (upregulated in HF+EGCG vs. HF+GTE), and glycine serine and threonine metabolism (downregulated in HF+EGCG vs. HF+GTE). Six additional metabolic pathways (i.e. 9 in total) were also altered between GTE and CAT. Specifically, HF+GTE compared to HF+CAT had an upregulation in tryptophan metabolism; D-glutamine and D-glutamate metabolism; phenylalanine metabolism; phenylalanine, tyrosine and tryptophan biosynthesis. Whereas, HF+CAT downregulated alanine aspartate and glutamate metabolism, and biotin metabolism. Thus, these metabolomics analyses suggest that microbial metabolic profiles and pathways affected by GTE are similar to those affected by EGCG but relatively different than those affected by CAT.

3.4. EGCG-induced changes in gut microbial phenolic profiles are more similar to GTE than CAT

GTE is a complex extract of multiple phenolic compounds whereas EGCG and CAT were provided as isolated compounds. Thus, clear potential exists for group-wise treatment effects on metabolic pathway regulation to be explained by the differential presence of polyphenols and/or polyphenol metabolites (and non-polyphenol compounds in the case of GTE treatment). PLS-DA analysis was able to differentiate phenolic profiles of fecal samples from the GTE-, EGCG-, and CAT-based treatment groups. Not surprisingly, the score plot (Figure 2A) showed that EGCG shifted the phenolic profile away from GTE, and CAT shifted the profile even further (R²=0.96 and Q²=0.95 for component 1, and R²=0.98 and Q²=0.97 for component 2). The phenolic profile differences between HF+EGCG and HF+CAT are probably due to different microbial metabolites formed from different precusors. Thus, phenolic profiles in the HF+GTE, HF+EGCG, and HF+CAT groups are clearly separated, which is consistent with the observed gut microbial metabolic profiles shown in Figure 1C. Figure 2B shows the top 15 polyphenols/polyphenol metabolites that were responsible for the group-wise separations. Seven phenolics were detected with VIP score higher than 1. As the color bar indicated, epigallocatechin, quercetin 3-D-galactoside, hesperetin, and homovanillic acid sulfate were only detected in GTE, while gallic acid, epicatechin gallate (ECG), and EGCG were present in both HF+GTE and HF+EGCG samples. Furthermore, fecal EGCG was more abundant in HF+EGCG group, while gallic acid and epicatechin gallate were more abundant in the HF+GTE group. None of these 7 phenolics were detected in the HF+CAT group, and 3,4-dihydroxybenzoic acid was the phenolic metabolite in the HF+CAT group with the highest VIP score of 0.88.

3.5. Microbial metabolic changes by GTE, EGCG, and CAT are associated with changes in microbial composition

As we reported, microbial composition in HF+GTE was more comparable to HF+CAT rather than that of HF+EGCG [19]. Measures of α-diversity, as represented by Chao1 richness and Shannon diversity index, indicated that HF controls had lower species diversity and richness compared with LF controls. Both measures of α-diversity in HF+GTE and HF+CAT, but not in HF+EGCG, were increased relative to the HF group and not different from the LF group [19]. For the microbial populations from all phylum, order, and genus levels, PLS-DA analysis in Figure 3A (Component 1, R²=0.58, Q²=0.48; Component 2, R²=0.70, Q²=0.56) demonstrated that the microbial profile of HF group separated from the LF group, and HF+GTE, HF+EGCG, and HF+CAT groups also shifted away in different directions from the HF group. Furthermore, the HF+GTE and HF+CAT groups were not well-separated, while the HF+EGCG group was separated from the HF+GTE and HF+CAT groups.

Fig. 3.

PLSDA analysis of microbial diversity and correlation between microbial populations and microbial metabolites. Each group contains 10 mice replicates. A. 2D score plot demonstrates the difference in microbial profiles among five treatment groups. B. 2D score plot demonstrates the difference in microbial profiles among GTE intervention (HF+GTE), EGCG intervention (HF+EGCG), and CAT intervention (HF+CAT) groups. C. Top 15 microbial populations that contribute to the separation among HF+GTE, HF+EGCG and HF+CAT based on the VIP analysis. The VIP score in x axis quantified the contribution. The color on the right represents the relative amount of each metabolites in each diet group. D. Correlation between microbial populations and microbial metabolites. The microbial diversity data were extracted from our parent publication [19].The detailed correlation matrix is in Table S3. Pair-wise correlations that do not meet the correlation cuttoff (≥0.5 or ≤−0.5) were set to be 0.

To define the microbiota structures that were affected by GTE and purified catechins, PLS-DA analysis was repeated using only the microbiota profiles in the HF+GTE, HF+EGCG, and HF+CAT groups (Component 1, R²=0.64, Q² =−0.42; Component 2, R²=0.73, Q²=−0.75). The microbial profiles of all three groups were well-separated (Figure 3B). The top 15 bacterial genera are shown in Figure 3C, and eleven had a VIP score higher than 1 that were contributing significantly to the group-wise separation. The color bar indicated that Ruminiclostridium and Coriobacteriales were most abundant in the HF+GTE group and least abundant in the HF+EGCG group. Akkermansia, Bacteroides, Betaproteobacteriales, Anaeroplasmatales, Parasutterella, Proteobacteria, Verrucomicrobia, and Verrucomicrobiales were all detected at the highest relative abundances in the HF+EGCG group. Although EGCG decreased the Firmicutes:Bacteroidetes ratio (Figure S2), it did not alter microbial populations similar to CAT, indicating that the microbiota profiles altered by CAT are more similar to those affected by GTE.

To understand the associations between the changes of gut microbial composition and microbial metabolism, correlation analysis was performed to explore microbe-metabolite relationships after different interventions in HF+GTE, HF+EGCG and HF+CAT groups. Pearson correlation coefficients were calculated (Figure 3D). When we focused the analysis on the top 15 microbial metabolites shown in Figure 1D, we identified 3 microbe-metabolite pairs with correlation coefficients >0.5 or <−0.5. Specifically, Erysipelotrichales positively correlated with nicotinamide, while Lachnospiraceae NK4A136 group negatively correlated with biotin. In addition, Akkermansia, which had the highest relative abundance in HF+EGCG group, positively correlated with n-acetylserotonin. Furthermore, four other microbial metabolites not belonging to the top 15 listed in Figure 1D are also listed to consider other potential biological impacts between the microbiome-metabolome data.

3.6. EGCG shifts the hepatic metabolome more similar to GTE than CAT in obese mice

PLS-DA analysis of the hepatic metabolic profiles (Table S2) in HF+GTE, HF+EGCG and HF+CAT groups showed that the hepatic metabolome of HF+EGCG was partly overlapped with both HF+GTE and HF+CAT groups (Figure 4A), but HF+CAT was separated from HF+GTE with only minor overlap (R² = 0.61 and Q²= 0.21 for component 1; R²= 0.79 and the Q²= 0.30 for component 2). Overall, there were eight metabolites with VIP score higher than 2 (listed in Figure 4B) that primarily contribute to the group-wise metabolome differences at the liver. Our findings revealed two common metabolites among the most abundant microbial metabolites (Figure 1D) and liver metabolites (Figure 4B). These were malate (VIP=1.9) and nicotinamide (VIP=1.72), and both were most abundant in HF+CAT in liver and fecal samples compared with GTE and EGCG treatment groups.

Fig. 4.

PLSDA analysis of liver metabolites and phenolics in three green tea component treatment groups. Each group contains 10 mice replicates. A. 2D score plot demonstrates the difference in liver metabolic profiles among GTE intervention (HF+GTE), EGCG intervention (HF+EGCG), and CAT intervention (HF+CAT) groups. B. Top 15 metabolites that contribute to the separation among HF+GTE, HF+EGCG and HF+CAT based on the VIP analysis. The VIP score in x axis quantified the contribution. The color on the right represents the relative amount of each metabolites in each diet group. C. Major liver metabolic pathways altered between HF+GTE and HF+EGCG. D. Major liver metabolic pathways altered between HF+GTE and HF+CAT. The x-axis is the pathway impact value, and y-axis is the statistical significance of the impact. The dot size corresponds to x-axis, and the dot color corresponds to y-axis. E. 2D score plot demonstrates the difference in liver phenolic profiles among HF+GTE, HF+EGCG and HF+CAT groups. For component 1, the R square was 0.76 and the Q square was 0.68. For component 2, the R square was 0.88 and the Q square was 0.77. F. Top 15 phenolics that contribute to the separation among HF+GTE, HF+EGCG and HF+CAT. The VIP score in x axis quantified the contribution. The color on the right represents the relative amount of each metabolites in each diet group.

We also performed metabolic pathway impact analysis to assess treatment-wise differences between liver metabolomes. As shown in Fig. 4C and 4D, with a threshold of pathway impact>0.2 and −log(p)>5, our results indicated that only hepatic glycolysis / gluconeogenesis metabolic pathways were significantly altered between HF+GTE and HF+EGCG (p = 0.005) groups, while only hepatic tyrosine metabolism was altered between HF+CAT and HF+GTE (p = 0.002). This is consistent with the metabolic pathways analysis of feces showing that metabolic effects altered by GTE are closer to those of EGCG than CAT. Tyrosine metabolism was also reported as an altered metabolic pathway in fecal results (Fig 1F).

We next hypothesized that the different hepatic metabolic profiles in HF+GTE, HF+CAT, and HF+EGCG groups would be associated with their unique liver phenolic profiles. As shown in Fig. 4E, PLS-DA analysis of liver phenolic profiles revealed poor separatation among the HF+GTE, HF+EGCG, and HF+CAT groups. The plot area of HF+EGCG is in the middle of HF+GTE and HF+CAT, which indicates that the phenolic profile from liver samples of HF+GTE is closer to HF+EGCG compared to HF+CAT. The separation pattern was consistent with that of polar metabolic profiles of the liver. Hepatic phenolic profiles in HF+GTE, HF+EGCG, and HF+CAT exhibited higher similarity than fecal phenolic profiles. Fig. 4F listed the top phenolics that contributes to the separation, with 4 having a VIP score higher than 1 (homovanillic acid, quercetin, quercetin 3-D-galactoside, and EGCG). All of these phenolics were highest in HF+GTE and not detected in HF+CAT. This is consistent with phenolic profiles of HF+EGCG being closer to HF+GTE (Fig. 4E).

3.7. Metabolic changes in the gut and liver are associated with mouse phenotypic/metabolic parameters

Correlation analyses of fecal / liver metabolites and the phenotypic parameters of mice were performed to define whether altered metabolic profiles are related to obesity and liver health. Heatmaps were generated to depict the correlations between metabolites identified herein and previously reported metabolic phenotypes [19]. Many fecal metabolites (e.g., malate, 3-methyl-2-oxindole, nicotinamide, and N-acetylserotonin) were correlated (P<0.05) with several phenotypic parameters (Figure 5A). Nicotinamide was highly correlated with serum insulin, HOMA-IR, final body weight, total adipose mass, liver triglyceride, and serum glucose (r = 0.50 to 0.73), while N-acetylserotonin was negatively correlated with all phenotypic parameters except liver malondialdehyde (r = −0.50 to −0.72). Relatively fewer liver metabolites were associated with mouse phenotypic responses. Azelaic acid and cyclic CMP were moderately negatively correlated with several phenotypic parameters such as liver malondialdehyde, serum insulin, HOMA-IR, body mass and adiposity (Figure 5B).

Fig. 5.

Correlation between the significantly altered metabolites from fecal samples (A) or liver (B) and host metabolic phenotypes. Pairewise associations between variables were determinded by Person’s correlation analysis. The metabolic phenotypic data were extracted from our parent publication [19]. The heatmanp depicts the strength of relationships with statistical significant positive (green) or negative (pink) correlations. $ indicates P<0.05 and # indicates P<0.005.

4. Discussion

Findings of this study demonstrate that, in agreement with our hypothesis, the differential anti-obesity effects that are mediated along the gut-liver axis by GTE, EGCG, and CAT are attributed to their unique influence on gut microbial and hepatic metabolic signatures. This is supported by our evidence indicating that EGCG-induced metabolic regulation of gut microbes is significantly different than the CAT-induced regulation, and is also different than the effect of GTE (Figure 1). Our hypothesis was further supported by outcomes of the analysis of microbial diversity (Figure 3) and the liver metabolome (Figure 4). Importantly, our results indicated that the metabolic changes in the gut and liver are associated with the changes in mouse phenotypic parameters (Figure 5). Outcomes of correlation analyses suggested that the altered metabolites contributing to the changes in metabolic profiles in GTE, EGCG, CAT groups are likely to impact host metabolic health via several different biological pathways. These include regulated NAD⁺ metabolism and other related central energy metabolism (e.g., TCA and glycolysis), and regulating metabolites such as N-acetylserotonin that have antioxidant activity. Taken together, our study suggests that catechins and GTE differentially influence obesity risk, and that bioactive polyphenols regulate metabolic responses at the gut and liver in a compound-dependent manner. Thus, these novel findings help to identify the physiological/metabolic responses by GTE, CAT, and EGCG that reduce obesity, and provide strong foundation for future study to examine organ-specific bioactivities of purified catechins (such as CAT) on cardiometabolic health.

Absorption of tea polyphenols is well-established to be poor [17]. Therefore, most polyphenolic components bypass small intestinal absorption and are directed to the lower gut where they undergo microbial biotransformation, which could alter gut microbial diversity and metabolism [13–16]. To understand the influence of bacterial populations and their metabolic responses to polyphenol-rich treatments, multi-omics workflows were used to investigate the relationships between gut microbial composition and their metabolic profiles following GTE/EGCG/CAT supplementation in HF mice. This permited identification of several correlations that can form the basis of advanced hypothesis testing in future study. For instance, we observed that higher abundances of gallic acid, epicatechin gallate, EGCG, and epigallocatechin resulting from GTE treatment may help to restore the HF-induced loss of Akkermansia and increased the abundance of N-acetylserotonin. This result is consistent with others who reported decreased Akkermansia in obese and type 2 diabetic mice, whereas administration of Akkermansia alleviated HF diet-induced intestinal inflammation [36]. However, it is worth noting that while HF+GTE and HF+CAT groups had similar microbial populations, HF+GTE and HF+EGCG groups have similar metabolic and phenolic profiles, and similar anti-obesity potentials. Our analysis therefore suggests that EGCG in GTE plays an important role in the gut with an anti-obesity potential modulated by metabolic regulation independent from its impact on microbial population. These findings also highlight the importance of combined study of gut microbial populations and their metabolic function to infer more meaningful conclusions.

Treatment-wise comparisons of gut microbial metabolic profiles also suggested that EGCG-induced metabolic changes of gut microbes are more similar to those of GTE compared to CAT. For example, malate and nicotinamide were limitedly abundant in HF+GTE and HF+EGCG, which is consistent with evidence that EGCG in a diet-induced obese mouse model increased the activity of malate dehydrogenase, an enzyme that catalyzes malate to oxaloacetate by reducing NAD⁺ to NADH [37]. As an important component of NAD⁺, nicotinamide can be produced during NAD hydrolysis by glycohydrolase in the intestinal luman, where the involvement of gut microbes has also been documented [38, 39]. While NAD⁺ is a well-known essential metabolic co-factor that is central to energy metabolism, NAD⁺ metabolism has gained interest as a target for metabolic health [40]. The two NAD⁺ degradation products, nicotinamide riboside and nicotinamide mononucleotide, were noted for their pharmacological efficacies in myocardial and cerebral ischemia and diabetes mellitus. In our study, GTE and EGCG significantly decreased elevated nicotinamide concentration in HF group, while CAT did not significantly impact its signal intensity (data not shown). This suggests that GTE and EGCG, but not CAT, may regulate central metabolism related pathways to influence metabolic outcomes of the treated mice. Meanwhile, N-acetylserotonin is an intermediate in the metabolic pathways of neurotransmitters melatonin and indoleamine, and was reported to be a microbial metabolite of tryptophan metabolism in the gut that exhibits antioxidant activity [41] to reduce lipopolysaccharide-induced lipid peroxidation. In our analysis, N-acetylserotonin was detected in LF mice but not in HF mice, whereas GTE and EGCG (but not CAT) fully or partially restored N-acetylserotonin abundance. This supports the concept that GTE and EGCG may also regulate obesity and liver health via the modulation of lipid peroxidation as we reported previously [19].

Although the biological effects mediated by EGCG were more similar to those of GTE in many parts of our analysis, our results indicated that CAT tested at the equivalent dose of EGCG can induce unique metabolic responses in HF mice. These unique differential benefits of CAT were somewhat surprising, and support the concept that polyphenols exhibit structure/function dependent bioactivities that differentially regulate obesity pathogenesis; this warrants additional detailed study. Indeed, in the gut microbial metabolome, CAT uniquely regulated the gut metabolism, which was evidenced by its differential metabolic profile compared to GTE and EGCG (Figure 1). CAT specifically downregulated alanine aspartate and glutamate metabolism, and biotin metabolism among several highlighted pathways. Many of these altered pathways were consistent with our prior bioinformatic predicitons [19]. Moreover, our PLS-DA results indicated that the hepatic metabolic profiles modulated by GTE and CAT are also largely different. Among the altered hepatic metabolites, myo-inositol is an important component of inositolphosphoglycan, which is the putative mediator of insulin action that could help to alleviate IR and obesity [42, 43]. It is also widely reported that EGCG regulates IR [44], and inhibit obesity, metabolic disease, and fatty liver disease by decreasing lipid absorption, attenuating inflammation, and regulating the NF-κB and STAT3 signaling pathways [22, 45]. It is therefore reasonable that EGCG and GTE had similar effects to decrease HOMA-IR and liver triglyceride, while CAT had a more modest effect on these pathologic responses [19]. Since EGCG has known benefits on obesity, metabolic disease, and nonalcoholic fatty liver disease [22, 45], and the metabolic profiles of HF+EGCG were closer to those of HF+GTE, the findings again suggest that EGCG in GTE can partially mimic the health benefits of GTE. On the other hand, even though the presence of CAT in GTE is quite low, it is more abundant in other foods and thus still warrants attention. Indeed, our findings support a critical need to study diverse polyphenols in different dietary pattens to understand their anti-obesity mechanisms.

Our study have several strengths, which included systematic assessments of metabolic shifts in liver and fecal samples during dietary HF-induced obesity that were integrated with phenotypic characteristics to provide novel evidence to explain anti-obesity effects of GTE, EGCG and CAT. We also acknowledge several limitations, including our study solely of male mice which precludes broader generalization to female mice. Also, we deliberately did not match the total catechin quantity between all treatments to their original composition in GTE. For instance, EGCG at 0.3% was selected to match its content in GTE as it cannot be overlooked that EGCG in GTE accounts for ~50% of the total catechins present in GTE. Although CAT in GTE is quite low, its study at a dose equivalent to EGCG provided the clearest evidence to date that these catechins uniquely regulate microbial and hepatic metabolism compared with EGCG. This approach permitted a functional characterization of CAT and EGCG while having GTE as a reference foodstuff that is rich in catechins. Future studies should consider additional treatments to match EGCG and CAT to the total catechin content of GTE and/or studying CAT at its typical abundance in GTE. It is also recognized that the dose of GTE (2%) used in the present study corresponds to a high intake of green tea in humans (~10 servings/d), which may preclude effective research translation. However, high intakes are common in regions where green tea is widely consumed (e.g. Asia), with epidemiological evidence indicating that ≥10 servings/d are associated with cardiometabolic benefits [46]. As an alternative to high volumes of green tea, GTE-rich confections have been developed as a low-energy snack food delivery system with acceptable sensory characteristics to achieve high green tea catechin intakes, which would be expected to facilitate research translation to human health [47]. Lastly, we also recognize that our targeted analysis of phenolics only considered those species that were present biologically in unconjugated or aglycone forms. Future studies may consider pre-treatment of samples to hydrolyze phase II metabolites (e.g., glucuronides, sulfates). It is possible that the assessment of total phenolics (i.e., the sum of aglycones and conjugated compounds) may differentially influence group-wise separations if xenobiotic metabolism differs between treatment groups.

5. Conclusion

Although GTE is well-reported to alleviate obesity by modulating host and gut microbial metabolism, direct evidence concerning the specific bioactive components, and microbial and host metabolites, responsible for these effects remain limited. Here, we demonstrated that EGCG regulated the fecal and liver metabolome in a manner more similar to GTE than CAT, whereas CAT at a dose equivalent to EGCG uniquely modulated the fecal/liver metabolome. When green tea phenolics reach the liver after interacting with gut microbes, a reduced abundance is expected. Indeed, our data also indicated that more pronounced metabolic changes at the gut level are contributing to the anti-obesity effect compared to metabolic regulations at the liver level. Still, the accumulated phenolics at the liver may contribute to the observed anti-obesity effects by altering the host metabolism. The exact mechanism and specific site of activity, however, requires future studies. Take together, our study suggests that EGCG and GTE exhibited similar anti-obesity effect in the metabolic analysis and the mechanism is not entirely dependent on the gut microbial composition. On the other hand, our findings support detailed investigation of CAT as a promising phenolic that could be used as a potential direct treatment or co-therapy with other phenolics to best alleviate obesity risk.