Participants in the clinical trial were recruited from December 2015 to April 2017 in the Nutrition Department of Yunnan First People’s Hospital in Kunming, China. Considering the influence of daily diet on gut microbiota, we make sure participants had relatively similar diet structure by questionnaires that included: (1) the number of midnight snack: hardly eat, 1–2 times/week, 3–4 times/week, or almost eat every day; (2) type of night snack: fruit, fried food, sweet cakes, dairy products, starch food, fried food and starch food, or cakes and fruit; (3) control the type of diet: controlling staple food quantity, increasing vegetable quantity, prohibiting sweet cakes and sugary drinks, prohibiting fat meat and animal viscera, prohibiting fried food, not controlling diet, or controlling staple food quantity, fat meat and animal viscera; (4) Beverage type: none, carbonated beverage, juice, milk tea and coffee tea, or tea; (5) Weekly alcohol intake. Of the 90 registered obese adults whose diet structure is roughly as follows: (1) the number of midnight snack: 1–2 times/week; (2) type of night snack: fried food; (3) control the type of diet: controlling staple food quantity; (4) Beverage type: carbonated beverage; (5) almost no alcohol consumption.) screened, 69 were included. According to a simple randomization procedure (computer random number), the participants were randomly assigned to one of the two treatment groups. The subjects included men and non-pregnant women aged between 25 and 35 years with a body mass index (BMI) ≥28 kg/m².

Subjects are excluded if they are taking drugs that may affect weight changes, namely, antidiabetic drugs, lipid-lowering drugs, or antifungal drugs, or if they have endocrine, cardiovascular, thyroid, or chronic liver disease. If they underwent surgery to lose weight, took probiotics or antibiotics within one month, or if their weight changed more than 5% within three months, then subjects were also excluded.

This is a 5-week double-blind, placebo-controlled trial. The subjects were randomly assigned to one of the two groups and received 10 g of lotus root starch placebo or KF per day (5 g lotus root starch placebo or 150 ml of warm water plus KF 15 min before breakfast and dinner) ( Figure 1 ). By emphasizing the intervention personnel and participants, blindness and balance are strictly maintained. Also, all investigators, staff, and participants were unaware of the outcome measurement and test results. This study was approved by the Ethics Committee of Kunming University of Science and Technology School of Medicine. All volunteers provided written informed consent before the study.

Height was measured at the beginning of the study, and weight and waist circumference were measured at weeks 0 and 5. At baseline and week 5, the subjects received a DXA scan (Hologic QDR 4500, Hologic, Inc., Bedford, MA, USA) to measure fat mass and body fat percentage. Blood samples were collected at recruitment (week 0) and at the end of the trial (week 5). Serum glucose, HbA1c, serum total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), triglycerides (TG) and liver function indicators (aspartate aminotransferase) (AST) and an automatic biochemical analyzer was used to measure alanine aminotransferase (ALT). Stool samples were collected at recruitment (week 0) and at the end of the trial (week 5). After collecting the stool sample, it was transported to the laboratory in an ice box and stored at −80°C for microbiota analysis.

To analyze the microbial content, the QIAamp DNA Stool Mini Kit (Qiagen, USA) was used to extract total genomic DNA from stool samples according to the manufacturer’s instructions. Forward primer F-5’CCTACGGGRSGCAGCAG3’ and reverse primer R-5’GGACTACVVGGGTATCTAATC3’ were used to amplify the bacterial 16S rRNA gene V3–V4 region. The PCR method and DNA purification were performed according to the Illumina Miseq 16S metagenomic sequencing library preparation protocol. For the first round of PCR, the PCR mix (25 µl) contained 12.5 µl 2× KAPA HiFiHotStartReadyMix (KAPA Biosystem, Woburn, MA, USA), 2.5 µl microbial genomic DNA (5 ng/µl) and 5 µl amplicon for PCR, respectively with Forward (1 μM) and reverse (1 μM) primers. The PCR protocol followed was: 95°C initial denaturation for 3 min, then 25 cycles of amplification, 95°C denaturation for 30 s, 55°C annealing for 30 s, 72°C extension for 30 s, and finally 72°C extension for 5 min. Gel electrophoresis was used to observe the PCR product to confirm the size of the amplicon (630 bp). In the second round, the PCR products were purified using AMPure XP beads. Index PCR used Nextera XT Index Kit to connect dual index and Illumina sequencing adapters. Index PCR was performed on all purified DNA using MJ Research PTC-200P thermal cycler (MJ Research, Waltham, USA). PCR mix (50 μl) contained 25 μl 2× KAPA HiFiHotStartReadyMix, 5 μl Nextera XT Index 1 Primers (N7XX) (1 μM), 5 μl Nextera XT Index 2 Primers (S5XX) (1 μM), 10 μl PCR pure water, and 5 μl purified DNA. PCR conditions are: initial denaturation at 95°C for 3 min, then denaturation at 95°C for 30 s, annealing at 55°C for 30 s, extension at 72°C for 30 s, 8 cycles of amplification, and finally extension at 72°C for 5 min. Gel electrophoresis was used to visualize the indexed PCR amplicons to confirm the amplification of the appropriate size product (630 bp). Then AMPure XP beads were used to purify the indexed PCR products. The purified index PCR product was quantified on a NanoDrop ND2000 spectrophotometer (Thermo Scientific, Wilmington, DE), diluted to a working concentration of 20 ng/μl with 10 mM Tris pH 8.5, and combined equimolar amounts (100 ng). After separating the products by electrophoresis on a 2% agarose gel, a Zymo clean Gel DNA Recovery Kit (Zymo Research Corp., Irvine, CA, USA) was used to recover and concentrate the PCR products of the correct size. The final combined DNA concentration was measured by using the Invitrogen Qubit platform (52.4 ng/μl).

The large number of 16S rRNA gene reads generated by the Illumina Miseq sequencer were initially quality trimmed using Illumina standard software tools. The 16S rRNA readings were processed according to the previously described QIIME pipeline (Caporaso et al., 2010). In short, 97% homology with USEARCH and NCBI’s 16SMicrobial dataset (https://github.com/mtruglio/QIIME_utilities) and taxonomy were used to cluster sequences into operational taxa (OTU).

The intestinal microbial analysis pipeline was based on Qiime1 (http://qiime.org/) and R (https://www.r-project.org/). The data are expressed as the mean ± SD. For cross-sectional analyses of baseline characteristics and comparison of the relative abundance of each taxonomic group, differences were indicated using the Paired two-tailed Student’s t-test. P-value less than 0.05 was considered as statistical significance.

Based on the RDP classifier version 2.2 algorithm, the Greengene database was used to annotate OTU classification information. The OTU abundance data are normalized using the standard sequence number corresponding to the sample with the least sequence. Check the relative proportions of each OTU at the phylum, class, order, family, genus, and species level. QIIME versions 1.7.0 and R 3.4.1 were used to analyze the diversity of Alpha (within the community) and beta (between the communities). For beta diversity, unweighted (considering the presence or absence of each species) UniFrac was used to generate principal coordinate analysis (PCoA) plots. The unweighted paired group method that used arithmetic mean (UPGMA) clustering was used as a hierarchical clustering method to explain the distance matrix using average links. A Spearman correlation analysis was performed to determine the correlation. A p-value of ≤0.05 is considered statistically significant.

Ninety subjects began the study. A total of 38 KF subjects and 31 control subjects completed the study for a total study retention rate of 76.7%. Before the intervention, there were no significant differences in the obesity index and blood parameters between the KF and placebo groups ( Table 1 ). The consumption of KF and placebo did not cause any adverse effects ( Table 1 ). Body mass index (BMI), fat mass, percentage body fat (PBF), triglyceride (TG), glycated hemoglobin A1c (HbA1c), aspartate aminotransferase (AST), and alanine aminotransferase (ALT) were significantly reduced in KF consumers ( Table 1 ) (p <0.05 or p <0.01). There were no significant differences in the control group ( Table 1 ).

Gut microbial alpha diversity metrics, which take into consideration of both richness and evenness, as indicated by the observed species, Chao1 index, Simpson index, and Shannon index. The observed species and Chao1 index significantly increased in the AKF group compared to the BKF group (p = 0.0235 and p = 0.0044) ( Figures 2A, B ). However, the observed species, Chao1 index, Simpson index, and Shannon index significantly decreased in the AP compared to the BP group (p = 0.0104, p = 0.0102, p = 0.0317, and p = 0.0274) ( Figures 2A–D ). These results suggested that KF treatment significantly increased gut microbial alpha diversity.

In order to better understand the richness shared between each group, a Venn diagram showing the overlap between the groups was developed. The analysis showed that only 449 of the total richness of 1,743 OTUs were shared among all samples ( Figure 2E ). The BP group had 674 OTUs, the AP group had 576 OTUs, the BKF group had 946 OTUs, and the AKF group had 1,014 OTUs ( Figure 2E ). In order to measure the degree of similarity between microbial communities, a principal coordinate analysis was used to calculate β diversity ( Figure 2F ). The principal coordinate analysis demonstrated that samples after KF treatment tender to cluster together and away from the samples before treatment. These data demonstrated that KF treatment remarkably improved richness and diversity of intestinal microbiota.

Gut microbiota composition significantly changed before and after treatment with KF.

At the phylum level, the relative abundance of Actinobacteria and Acidobacteria were significantly increased in the AKF group compared to the BKF group (p <0.05) ( Figure 3A ).

At the family level, the relative abundance of Lachnospiraceae, Bacillaceae, Aerococcaceae, Solirubrobacteraceae, 288-2 and RB41 significantly elevated after treatment with KF, while there was relative abundance of Sporolactobacillaceae significantly reduced after treatment with KF (p <0.05) ( Figure 3B and Table S1 ). Furthermore, placebo treatment reduced the relative abundance of Bifidobacteriaceae and Leuconostocaceae in the AP group compared to the BP group (p <0.05) ( Figure 3B and Table S1 ).

At the genus level, the relative abundance of Roseburia, Lachnoclostridium, Erysipelotrichaceae UCG-003, Lachnospiraceae NK4A136 group, Lachnospiraceae UCG-004, Fusicatenibacter, Lachnospiraceae UCG-003, Clostridium sensu stricto 13, Howardella, Intestinimonas, Bacillus, Holdemania, Epulopiscium, Candidatus, Soleaferrea, Solobacterium, Solirubrobacter, and Abiotrophia significantly increased after KF treatment, while Lactococcus, Eubacterium brachy group, Sporolactobacillus, Geobacillus, and Silanimonas significantly reduced ( Figure 3C and Table S1 ). In addition, the relative of abundance of Dorea significantly reduced in the AP group compared to BP group (p <0.05) ( Figure 3C and Table S1 ).

At the species level, the relative abundance of Roseburia inulinivorans, Clostridium perfringens, Intestinimonas butyriciproducens, Actinomyces graevenitzii, Clostridium sp. ATCC 29733, and Solobacterium moorei significantly elevated in the AKF group compared to the BKF group, while the relative abundance of the Bacteroides fragilis, Lactococcus garvieae, Bacteroides coprophilus, Bacteroides ovatus, Bacteroides caccae, Bacteroidales bacterium ph8, Bacteroides thetaiotaomicron, and Geobacillus stearothermophilus significantly reduced in the AKF group compared to the BKF group (p <0.05) ( Figures 4A–N ). Moreover, the relative abundance of B. fragilis and B. caccae significantly increased in the AP group compared to BP group (p <0.05) ( Figures 4A–N ).

Altogether, KF treatment can increase some beneficial bacteria and decrease some harmful bacteria.

At the phylum level, five phyla were significantly correlated with obesity index and blood parameters (p <0.05 or p <0.01) ( Figure 5A ). Proteobacteria was related to obesity. For blood parameters, Proteobacteria was significantly and positively correlated with serum AST and ALT content (p <0.01).

At the family level, thirteen families were significantly correlated with obesity index and blood parameters (p <0.05 or p <0.01) ( Figure 5B ). Prevotellaceae was related to obesity. For blood parameters, Prevotellaceae was significantly and positively correlated with serum FBG content (p <0.01).

At the genus level, sixteen genera were significantly correlated with obesity index and blood parameters (p <0.05 or p <0.01) ( Figure 5C ). Erysipelotrichaceae UCG.003 and Lachnospiraceae NK4A136 were related to KF treatment. For blood parameters, Erysipelotrichaceae UCG.003 was significantly and negatively correlated with serum FBG content (p <0.01). Lachnospiraceae NK4A136 group was significantly and negatively correlated with serum ALT content (p <0.01).

At the species level, fifteen species significantly correlated with obesity index and blood parameters (p <0.05 or p <0.01) ( Figure 5D ). Bacteroides ovatus were related to KF treatment. For obesity index, B. ovatus abundance was significantly and negatively correlated with BMI and WHR (p <0.01). For blood parameters, C. perfringens were significantly and negatively correlated with serum TC and LDL content (p <0.01).

Altogether, these bacteria can influence obesity index and blood parameters.

KF intervention could reduce obesity indexes and blood parameters, regulate the diversity and composition of the gut microbiota, accelerate growth of good bacteria, and inhibit growth of maleficent bacteria. In addition, changes in the microbiota have a strong correlation with obesity index and blood parameters. These findings support that the anti-obesity and other effects of KF may be mediated through its selective regulation of the microbiota. The duration of trial treatment is short, which maybe a limitation of our study. Then we would focus on KF and make further investigation.