Metabolically Healthy Obese Children and the Role of the Gut Microbiota

Background The term “metabolically healthy obese (MHO)” denotes a hale and salutary status, yet this connotation has not been validated in children, and may, in fact, be a misnomer. As pertains to obesity, the gut microbiota has garnered attention as conceivably a nosogenic or, on the other hand, protective participator. Objective This study explored the characteristics of the fecal microbiota of obese Chinese children and adolescents of disparate metabolic status, and the associations between their gut microbiota and circulating proinammatory factors, such as IL-6 and TNF-α, and a cytokine up-regulator and mediator, leptin. Results Based on weight and metabolic status, the 86 Chinese children (ages 5-15 years) were divided into three groups: metabolically healthy obese (MHO, n=42), metabolic unhealthy obesity (MUO, n=23), and healthy normal weight controls (Con, n=21). In the MUO subjects, the phylum Tenericutes, as well as the alpha and beta diversity, were signicantly reduced compared with the controls. Furthermore, Phylum Synergistetes and genus Bacteroides were more prevalent in the MHO population compared with controls. For the MHO subgroup, Spearman’s correlation analysis revealed that serum IL-6 positively correlated with genus Paraprevotella, and leptin correlated positively with genus Phascolarctobacterium and negatively with genus Dialister (all p<0.05). Conclusion microsystem prevails and some metabolism-related bacteria associates obesity-


Introduction
The global epidemic of childhood obesity, and the accompanying rise in the prevalence of endocrine, metabolic, and cardiovascular comorbidities, is perhaps the most impactful and ubiquitous public health disorder of the modern world [1]. In the context of this pandemic, a distinct subgroup of youth with obesity who are devoid of metabolic disturbances-so-called "metabolically healthy obese" (MHO) -have been identi ed. Obesity notwithstanding, by de nition MHO children retain a favorable metabolic pro le, with preserved insulin sensitivity along with normal blood pressure, glucose homeostasis, lipids, and liver enzymes. Moreover, their hormonal, in ammation, and immune pro les are seemingly impervious to obesity [2]. First described in obese adults, the MHO phenotype has also been extensively studied in young people with obesity [2]. Arguably, MHO may be a transitional stage to the far more common, more high-risk, conventional cardio-metabolic obese phenotype. Regardless of the aforesaid normal biochemical characteristics of MHO, the risk for cardiovascular disease persists since the MHO phenotype may be unstable, thereby transitory. [3,4].
Among the non-genetic factors associated with obesity, the gut microbiota has garnered attention as an obesity regulator given the robust correlations in animal studies between gut microbiota and body weight.
Obese individuals, whether adults or children, have increased abundance in Firmicutes in concert with decreased in Bacteroidetes [5,6]. The distinctive gut bacterial ora prevalent in obese subjects is recognized as promoting an unhealthy metabolic obese (MUO) phenotype with attendant comorbidities, such as increased endotoxemia, intestinal and systemic in ammation, as well as insulin resistance. An altered gut microbiota has been implicated in obesity and type 2 diabetes mellitus (T2DM) insofar as a decrement in certain species and gene richness have been linked to adiposity, dyslipidemia, and insulin resistance [7]. Hence, the clinical repercussions aside, it is plausible that differences in the gut microbiota could dictate whether an obese child is metabolically t (MHO) or not (MUO) [8,9].
Firstly, this study examined the metabolic heterogeneity of obese children as it relates to the composition of the bacterial ora of the gut. And, as a secondary end point, identify metabolic-speci c bacteria which associate with serum in ammatory factors incriminated in obesity comorbidities.

Study participants
Based on weight status, the metabolically stable cohort (MH) subjects (n=63) were subdivided as MHO (n=42) or Con (n=21). Recall that, by de nition, the MH refers to a merger of the Con and MHO cohorts ∴ MH= Con and MHO.
The age of the 86 participates ranged from 5.5 to 14.3 years, with a mean of 9.76 ±1.93 years. There were 65 obese children, of whom 23 were MUO and 42 were MHO. The BMI of other 23 children were normal. Age, BMI, BMI-Z, WHtR, SBP, TG and LDL-c in the MUO group were signi cantly higher than the MH group, and HDL-c in the in the MUO group were signi cantly lower than the Con and MHO children (all p<0.05, Table 1).
The BMI, BMI-Z, WHR, WHtR, SBP, DBP, TG and LDL-c were signi cantly higher in the MHO group than the Con children, and HDL-c in the MHO group were signi cantly lower than the Con group (all p <0.05). There was no statistical difference in age, gender, FPG and fasting TC between MHO and Con (all p>0.05, Table   1).

Microbiota Pro les in different metabolic status subjects
A total of 918,578 sequencing reads were obtained from 86 fecal samples, with an average value of 10,681 counts per sample. We identi ed an overall of 146 OTUs, among which 136 OTU with ≥2 counts, and they were grouped in 9 phylum and 38 families.
(1)Abundance pro ling in different metabolic status subjects Grouping OTUs at phylum level, and applying the Mann-Whitney U test on the relative abundances of phyla for the two groups, the relative abundances of phylum Tenericutes was more prevalent in the MH subjects (recall that MH=Con and MHO)compared to the MUO group (p = 0.006, Table S1 and Figure 1a). On OTUs at the genera level, by Mann-Whitney U-test, including all the genera (merging small taxa with counts<10), we identi ed that genera Anaerostipes, Alistipes, Desulfovibrio,Fusobacterium, Gemmiger, Odoribacter, Oscillospira and Parabacteroides were more prevalent in the two metabolically healthy cohorts(MH) versus MUO children, yet the genus Dorea was more prevalent in MUO (p < 0.05; Figure 1b, Table 2).
(2) Alpha-and beta-diversity in different metabolic status subjects To assess the overall differences of microbial community structures in MH and MUO subjects, we measured ecological parameters based on alpha-diversity. The alpha-diversity analysis showed signi cantly higher diversity in MH in comparison to MUO participants (p<0.05, Figure2 a,b, Table S2).
To determine the differences between microbial community pro les in MH and MUO subjects, we calculated beta-diversity. By Distance method Bray-Curtis dissimilarities PCoA analysis, the gut microbiota samples from MH were clustered together and separated partly from the MUO group. Upon analysis, metabolism explained 19.8% of the variance in microbiota composition (P = 0.038, Figure 2c, Table S3).
(3)Bacterial taxa differences in different metabolic status subjects We next used LEfSe analysis to identify bacteria in which the relative abundance was signi cantly increased or decreased in each phenotypic category. The MH subjects had members of the phylum Tenericutes, class Deltaproteobacteria, Mollicutes, order Desulfovibrionales, RF39, family Christensenellaceae, Odoribacteraceae, Porphyromonadaceae, Ruminococcaceae, genera Anaerostipes, Oscillospira, Odoribacter, Gemmiger, Parabacteroides, Alistipes, that were signi cantly higher than MUO subjects. Furthermore, the MUO subjects had members of the genus Fusobacterium that were signi cantly higher than the MH subjects (all p<0.05, Figure 3a, b).

Microbiota Pro les in obese subjects with different metabolic status
(1)Abundance pro ling in obese subjects with different metabolic status Grouping OTUs at phylum level, and applying the Mann-Whitney U test on the relative abundances of phyla for the MHO and MUO groups, the relative abundance of phylum Tenericutes was more prevalent in the MHO group compared to the MUO group (p = 0.027, Table 3 and Figure 1c). On OTUs at the genera level, by Mann-Whitney U analysis, including all the genera (merging small taxa with counts<10), we identi ed that genera Desulfovibrio, Parabacteroides and Gemmiger were more prevalent in MHO subjects compared to MUO subjects (p = 0.027, 0.040 and 0.047, respectively; Figure  1d).

(2) Alpha-and beta-diversity between MHO and MUO subjects
Regarding alpha-diversity, in both the MHO and MUO group, the analysis exposed signi cantly higher diversity in MHO subjects versus MUO participants (all p <0.05, Figure 2d,e, Table S2).
Regarding beta-diversity, by an unweighted-UniFrac method, the MHO group was lower than the MUO group (p=0.021, Table S3).

Microbiota Pro les in MHO and Con subjects with different weight status
(1)Abundance pro ling in metabolically healthy subjects with different weight status Grouping OTUs at phylum level, the relative abundances of phylum Synergistetes was more prevalent in the MHO group compared to the Con group (p < 0.05, Figure 1e, Table 4). On OTUs at the genera level, including all the genera (merging small taxa with counts<10), genera Anaerotruncus, Bacteroides, Adlercreutzia and Pyramidobacter were more prevalent in MHO subjects versus MUO subjects (p < 0.05; Figure 1f).
(2) Alpha-and beta-diversity between different weight status Regarding alpha-diversity, the Shannon diversity index, Observed OTUs, Faith's phylogenetic diversity and Pielou's evenness based on OTU distribution did not reveal any signi cant difference between MHO and Con (all p >0.05, Table S2); also, beta-diversity did not differ signi cantly between these two groups.
Importantly, none of the comparisons were signi cantly different (all p>0.05) after correction for multiple testing (Table S3).
(3) Bacterial Taxa Differences in MHO and Con subjects of different weight status LEfSe analysis showed MHO subjects had members of the phylum Synergistetes, class Synergistia, order Synergistales, Erysipetotrichales, family Dethiosulfovibrionaceae, genus Pyramidobacter were signi cantly higher than the Con-, however, the latter had members of the family Bacteroidaceae, genus Anaerotruncus that were signi cantly higher (all p<0.05, Figure 3e, f).

Detecting Microbial Biomarkers in different metabolic status
Discriminant analysis (DA) based on univariate ANOVAs, Fisher's coe cient and leave-one-out classi cation were performed to de ne a model based on the capability of OTUs to discriminate the three groups of study participants (MHO, MUO and Con).
A DA showed that 58.1% of the original grouped subjects were correctly classi ed, and the canonical discriminant plot revealed partly separation among groups ( Figure 4 and Table S4). Furthermore, applying a cross-validation (CV) test, we found that 54.7% of cases were correctly classi ed, attesting to the capability of the entire OTUs set to discriminate the three groups (Table S4).
For MUO subjects, Spearman's correlation analysis revealed that IL-6 positively correlated with genus Lactococcus, TNF-α positively correlated with phylum Bacteroidetes, negatively correlated with genus Citrobacter. Leptin positively correlated with genus Eubacterium and negatively correlated with genus Faecalibacterium and Lachnospira (all p<0.05, Table S5).

Metabolic Pathway Predictions
A total of 15 KEGG pathways were generated using the composition of the fecal microbiota based on PICRUSt2 in the MH cohorts versus MUO subjects ( Figure 5, Table S6). In the comparison between MHO and MUO subjects, we obtained 3 differential pathways ( Figure 5, Table S7). Moreover, 11 differential metabolic patterns differentially expressed resulted in the comparison between MHO versus Con ( Figure  5, Table S8).

Discussion
Recognized for decades, there is wide-ranging heterogeneity among obese individuals as to their risk for developing metabolic dysfunction and attendant complications [10]. Also well-established, and which may contribute to this metabolic heterogeneity, is the observation those with central obesity are more prone to developing T2DM and cardiovascular disease than those with peripheral obesity [11]. In this study, to indirectly address the issue of fat distribution, we found there were no signi cant differences in WHR and WHtR between the two obese cohorts, MHO vs MUO.
A chronic low-grade in ammation, triggered by nutrient surplus, is a constituent of obesity. Adiposeoriginated metabolic in ammation develops pari passu with insulin resistance and, as such, is a key element in the metabolic syndrome [12]. In this study, we found there were no signi cant differences in serum IL-6, TNF-α and leptin between MHO and MUO subjects. It stands to reason that, besides these cytokines, other biochemical factors likely contribute to the metabolic diverseness in obese subjects. Or, perhaps, the concentrations of circulating compounds-such as those abovementioned-poorly re ect those found in extracellular or intracellular tissues.
Evidence can be adduced that the gut microbiota is involved in the aetiology of obesity and obesityrelated complications such as nonalcoholic fatty liver disease, insulin resistance and T2DM [13,14]. These disorders are characterized by alterations in the diversity of the gut microbiota, and the relative abundance of certain genera. And bacteria-generated metabolites, translocated from the gut across a disrupted intestinal barrier, can affect several metabolic organs, such as the liver and adipose, thereby contributing to systemic metabolic in ammation [15].
Recently, several animal studies concluded that an optimal healthy-like gut microbiota may bestow a more propitious obese phenotype [16,17]. For instance, the abundance of Bacteroidetes and Tenericutes were closely aligned with bile acid metabolism and obesity-related in ammation in a murine model of the metabolic syndrome [18]. In our study, we corroborate this nding: reduced abundance of Tenericutes in the MUO group compared with the metabolically healthy groups (MHO and Con). Similarly, in high fat diet-induced obese mice, β-glucan favorably increases bacteria that generate butyrate (such as Anaerostipes), thereby mitigating hepatic stress and intestinal atrophy [19]. In another study, gammaaminobutyric acid enriched rice bran ameliorated the metabolic syndrome (insulin resistance, lipids) in dietary-induced obese rats by enhancing Anaerostipes production of two salutary short-chain fatty acids, C2 butyrate and C3 propionate, and manifest both in the intestine and circulation. Finally, another bene t was a signi cant upturn in serum leptin and glucagon-like peptide-1 [20]. We also observed more abundance of Anaerostipes in the MH cohort, as well as the alpha and beta diversity. These results buttress the notion of dysbiosis in the gut microbiota of MUO individuals.
To characterize the gut microbiota in obese children of different metabolic status, we further analyze the MHO and the MUO subgroups. The abundance of Tenericutes was signi cantly reduced in the MUO group compared with the MH children, indicating that Tenericutes is related to the metabolic state, and the bacterial imbalance is independent of weight. Previously reported, the abundance of Parabacteroides was signi cantly decreased in obese subjects with metabolic syndrome [6], and nonalcoholic fatty liver disease [21], and negatively correlated with weight gain and leptin plasma levels [22]. And germane to our ndings, both genera Gemmiger [30] and Parabacteroides [23] are gut bacteria negatively associated with obesity and disturbed host metabolism. In accordance, we found that that the fecal abundance of these bacteria was signi cantly higher in the MHO group compared with MUO.
The genera Parabacteroides are short-chain fatty acids (SCFAs)-producing bacteria. SCFAs are low molecular weight molecules produced from fermentation of dietary ber or polysaccharides by gut microbiota. Absorbed by the intestinal epithelium into the blood, they can beget physiological disorders in the host, such as deranged lipid metabolism and intestinal environment imbalances [24,25]. Furthermore, alpha and beta diversity were signi cantly higher in MH subjects compared with the MUO group, again supporting the notion of dysbiosis in the unhealthy MUO population.
Notwithstanding that the gut microbiota obese individuals with metabolic syndrome may indeed unhealthy, is the gut microbiota of the MHO population really healthy? We compared the characteristic of gut microbiota in the MH population of different weights. Even though there was no signi cant difference in alpha and beta diversity, the relative abundances of phylum Synergistetes and genus Bacteroides were elevated in the MHO group compared to the Con children. Based on a metagenomic approach and bioinformatics analysis in obese adults, it is plausible that an abundance of the microbiota taxa Bacteroides could portent the evolution to T2DM [26].
Alterations in gut ecology can propel in ammatory pathways in several tissues, resulting in glucose intolerance and CVD [27,28]. In rodents, both the dysregulation of the tandem microbiota-host metabolism of bile acids and also the bacterial production of lipopolysaccharides (i.e., endotoxemia) can beget derangements in glucose homeostasis [29,30]. Herein, we found that, depending on the metabolic status, the serum levels of classic proin ammatory factors IL-6, TNF-α and leptin were related to the abundance of various fecal bacteria. Notably, in MHO children, serum leptin correlated positively with genus Phascolarctobacterium and negatively with Dialister -the latter genera observed with low abundance in obese children [31]. And, relevant to our ndings, it is noteworthy that Phascolarctobacterium purportedly is a biomarker for adult T2DM [26]. As illustrated in our MHO children and the above-cited studies in humans, the gut microbiota is a marquee player in preserving normal metabolism despite obesity or, perhaps, an ephemeral protective ora destined to change with transition to MUO.

Conclusion
In aggregate, the MUO population had lower alpha-and beta-diversity, and lower abundance of Tenericutes, which were independent of weight, inferring a robust inter-relationship between gut bacterial ecology and host metabolic state. In the MHO population, phylum Synergistetes and genus Bacteroides and Phascolarctobacterium were more prevalent, and the abundance of some metabolism-related bacteria correlated with circulating proin ammatory factors, suggesting that dysbiosis of gut microbiota was already extant in the MHO children, conceivably a compensatory or remedial response to a surfeit of nutrients.

Study population
This study was approved by the Ethics Committee of the Fuzhou Children's Hospital of Fujian Medical University and, in all cases, informed consent was obtained.
The cross-sectional study consisted of participants managed by Fuzhou Children's Hospital of Fujian Medical University from September 2017 to March 2018. This study was limited to participants who met the following criteria: (a) ages between 5 to 15 years old, and (b) residence of Fujian province.
The exclusion criteria were as follows: any endocrine disorder, history of antibiotic therapy in the past 3 months prior to the enrollment, chronic gastrointestinal illness or use of gastro-intestinal-related medication, or diarrheal disease (World Health Organization de nition) in the past one month.

Clinical assessment
Height and weight were measured by trained nurses. BMI-Z scores were calculated based on reference values of Li Hui et al [32]. At the end of normal expiration, waist and hip circumference were measured to the nearest 0.5 cm using standard technique with nonelastic tape. Waist circumference was measured at a point midway between the lower border of the ribs and the iliac crest, and hip circumference was measured at the widest part of the hip. A waist-to-hip ratio (WHR) was calculated by waist circumference (cm) divided by hip circumference (cm) and a waist-to-height ratio (WHtR) by waist circumference (cm) divided by height (cm).

Laboratory examination
All participants maintained their usual dietary pattern at least 3 days before blood sampling. After 12 h of fasting, 10 ml venous blood was drawn by registered nurses. All blood samples were stored at −80℃, and analyzed within two weeks of sampling. Serum IL-6 was measured using a commercial ELISA kit (Abcam, UK), with an 4.4% inter-assay coe cient of variation (CV). Serum TNF-α levels was measured using a commercial ELISA kit (Abcam, UK), with inter-assay and intra-assay CVs of 3.3% and 9%, respectively, and serum leptin assayed using a commercial ELISA kit (Abcam, UK), with inter-assay and intra-assay CVs of 2.4% and 2.7%, respectively. Fasting plasma glucose (FPG) and plasma lipids, including total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-c) and low density lipoprotein cholesterol (LDL-c), were assayed by standard methods using speci c reagents (Beckman Coulter AU5800, USA). Fasting insulin (INS) was determined by a chemiluminescent immunoassay (IMMULITE 2000, Siemens Healthcare Diagnostics Products Limited, Germany). Fecal samples were collected and processed as previously described [33].

De nition of metabolic unhealthy
Metabolic syndrome parameters were applied according to 2019 Expert Committees [34], and MUO was de ned by the presence of at least one of the following metabolic traits: (1) FPG ≥ 5.6 mmol/L; (2) systolic blood pressure ≥ 90th percentile for gender and age; (3) fasting HDL-C< 1.03 mmol/L; and (4) fasting TG ≥ 1.7 mmol/L.

Genomic DNA extraction and Library Construction
The microbial community DNA was extracted and quanti ed as previously described [33]. Variable regions V3-V4 of bacterial 16s rRNA gene were ampli ed with degenerate PCR primers [33]. Libraries were quali ed by the Agilent 2100 bioanalyzer (Agilent, USA). The validated libraries were used for sequencing on Illumina MiSeq platform (BGI, Shenzhen, China) following the standard pipeline of Illumina, and generating 2 × 300bp paired-end reads.

Statistical analysis
Statistical analyses of clinical data were performed using the Statistical Package for the Social Sciences software version 23.0 (SPSS Inc. Chicago, IL, USA). The normality of the data was tested by Kolmogorov-Smirnov test. Data are expressed as mean ± SD. Comparisons of the results were assessed using independent samples t test, Mann-Whitney U test and Kruskal-Wallis test, depending on the type of data distribution (e.g., non parametric). Comparison of rates between two groups was by chi-square. A value of P < 0.05 was deemed statistically signi cant.
Statistical analysis of 16s rRNA sequencing data were performed on alpha-and beta-diversity measurements, which was done by software QIIME2(v2019.7) [35]. Kruskal-Wallis Test was adopted for two groups comparison. Linear discriminant analysis Effect Size (LEfSe) Analysis was assessed by software LEFSE [36]. To predict metagenome functional content from 16S rRNA gene surveys, Picrust2 [37] have been applied to obtain the KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways, and STAMP [38] was used to analyze the differential pathways.

Declarations
Ethics approval and consent to participate This study was reviewed and approved by the Ethics Committee of Fuzhou Children's Hospital of Fujian Medical University, and was conducted in agreement with the Declaration of Helsinki Principles. Informed consent was obtained from all individual participants included in the study.

Consent for publication
Informed consent for publication was obtained from all individual participants included in the study.

Availability of data and materials
The original contributions presented in the study are publicly available. The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2017) in National Genomics Data Center (Nucleic Acids Res 2020), Beijing Institute of Genomics (China National Center for Bioinformation), Chinese Academy of Sciences, under accession number CRA003010 that are publicly accessible at https://bigd.big.ac.cn/gsa.

Competing interests
The authors declare that they have no competing interests. Author Contributions XY drafted the initial manuscript; RMC conceptualized and designed the study, and reviewed and revised the manuscript; KL. M revised the manuscript; YZ and XHY collected cases; XQL did the laboratory testing.      Table S2. Principal coordinates analysis (PCoA) plot of MH and MUO groups (e). The plots show the rst two principal coordinates (axes) for PCoA using Bray-Curtis Distance method.