Remodeling of the gut microbiome during Ramadan-associated intermittent fasting

Young male adult cohort

We prospectively recruited male volunteers who expressed the intention to observe fasting in the month of Ramadan, according to the Islamic law. The entire fasting period lasted for 30 d and was performed during Ramadan, and the intermittent fasting period of each day was from dawn to sunset (which was approximately 16 h in this study). Volunteers were recruited from 2 independent locations (Najiaying town in Yuxi city and the Chenggong district in Kunming city), but both are located in Yunnan province, China. Through reviewing the medical history, only healthy individuals were selected. Exclusion criteria were 1) obese [BMI (in kg/m²) ≥30] and 2) self-reported diabetes, hypertension, arthritis, chronic respiratory disorders, or any gastrointestinal disorders. Prescription of antibiotics in the month prior to the start of the study was used as another exclusion criterion. In total, 30 volunteers were selected. The stool of each participant was collected on the first day of the 2016 Ramadan, at an intermediate time point (day 15), and on the final day (day 30) of this religious fasting period. To avoid environmental contamination, all volunteers were kindly requested to 1) place a clean newspaper to catch the stool, 2) wash their hands before sampling their own stool, and 3) collect the stool in a clean, screw-top container by using the spoon that came with the container and then close the container by screwing the lid shut. Following collection, stool samples were transported on ice (using a polystyrene icebox) to the laboratory and immediately stored at –80°C until further processing. All samples spent no longer than 2 h on ice. All participants provided informed consent, and the study was approved by the relevant ethical committee.

Middle-aged cohort

Participants were recruited from Xining city, which is located in Qinghai province in China. Volunteers who, upon review of their medical history, showed earlier diagnoses of metabolic disorder, cardiovascular disease, or chronic inflammatory disorders were excluded, as were those with a current acute inflammatory disorder. In addition, volunteers employing antibiotic use in the month prior to the study were excluded as well. Following review of the medical history, 37 healthy participants were selected and enrolled in this cohort. Twenty-seven of the selected participants subjected themselves to intermittent fasting, and the remaining 10 served as nonfasting controls. The stool samples for each participant were collected at T1 (start of the 2018 Ramadan), T2 (the end of the 2018 Ramadan), and T3 (1 mo following the end of Ramadan fasting). Environmental contamination for stool sampling was controlled by the same measures as described for the young population above. Following production, samples were transported in a Styrofoam icebox to the laboratory and immediately stored at –80°C. All samples spent no longer than 3 h on ice. During the study, 25 volunteers in the intermittent fasting group fully complied with the investigation protocol, but 2 volunteers failed to provide their fecal samples at T3. All volunteers in the nonfasting control group fully complied with the study protocol. Of note, all volunteers, including those in the unfasted group, were asked to provide their fecal and blood samples at each time point: before, the end, and 1 mo after the cessation of the 2018 Ramadan fasting. All participants provided informed consent, and the study was approved by the relevant ethical committee.

In this cohort, we also measured selected physiologic parameters. Changes in fat mass and fat percentage in these middle-aged participants during fasting and body composition indices were measured by a body composition analyzer (SH-900G; Novogene) at all 3 time points. For blood parameter analysis, following overnight fasting, blood was collected and subsequently sent to Zhejiang DIAN Diagnostics, and the following blood parameters were determined: aspartate aminotransferase, alanine aminotransferase, direct bilirubin, indirect bilirubin, total bilirubin, γ-glutamyltransferase, glucose, hemoglobin A1c, triglycerides, HDL cholesterol, LDL cholesterol, total cholesterol, creatinine, urea, and uric acid.

In addition, information on food intake was collected. To this end, we used a modified FFQ to assess dietary intake 1 mo before fasting, during fasting, and a month after the cessation of the fasting. There were 5 frequency categories in FFQ ranging from “never eaten” to “eating daily,” “eating one time per week,” “eating one time per month,” and “eating one time per year.” Participants reported their food intake frequency through a recall interview after intermittent fasting. In this study, we did not include any food types that were only consumed on a yearly basis. For dietary components with a self-reported consumption frequency of once per week or more, calorie intake was calculated according to the corresponding frequency. Finally, derived weekly energy and nutrient intake was calculated in accordance with the China Food Composition Database.

Ethics

The protocol of this study was approved by the ethical committee of the medical faculty at Kunming University of Science and Technology (protocol 2017JC002; KMUST-MEC-017) and the medical ethical committee of Northwest Minzu University (XBMZ-YX-2,016,001). This study is registered at www.chictr.org.cn (ChiCTR2000034646). All volunteers provided written informed consent prior to the study.

Outcome

The primary outcome was the change in gut microbiota composition after a 30-d Ramadan-associated intermittent fasting compared with baseline. To this end, gut microbiota were analyzed for diversity, composition, and taxonomic abundance. Secondary outcome variables were changes in body composition, blood parameters, and food intake.

Next-generation sequencing and data processing

Fecal DNA was extracted from the stool sample using the QIAamp DNA Stool Mini Kit (Qiagen), according to previously described procedures (10). To minimize the impact of reagent microbiome (kitome), fecal DNA was prepared using the same batch of kits and reagents for each cohort. In addition, a blank sterile microcentrifuge tube was used as a negative control to characterize kitome influences during DNA extraction. DNA quality was monitored on 1% agarose gels (Supplementary Figure 1). As a control for bacterial environmental contamination during DNA library preparation, DNA-free water (blank library) was used. The quality of the library was assessed using the Qubit@ 2.0 Fluorometer (Thermo Scientific). Next-generation sequencing of the V3–4 region of the bacterial 16S ribosomal RNA gene was performed by the NovoGene Company. Data processing and analysis were performed according to routine procedures, which have been described elsewhere in detail (11–14).

Statistical analysis

Gut microbiota diversity was estimated by calculating the Shannon and Simpson indices using alpha-diversity.py in QIIME (15), and the Wilcoxon signed-rank test was used to determine statistical significance between groups. Principal coordinates analysis (PCA) of Bray–Curtis distance was performed using the “vegan” package in R programming language (version 4.0.2) for establishing the shift of the gut microbiota composition after Ramadan-associated intermittent fasting. Multivariate data analysis according to analysis of similarities (ANOSIM) was applied to determine gut microbial shifts. A P value less than 0.05 was considered statistically significant. To identify bacterial taxa whose sequences were differentially abundant between groups, linear discriminant analysis (LDA) coupled with effect size measurements (LEfSe) analysis was applied with the significance level of 0.05 and the logarithmic LDA score threshold equal to 4 (http://huttenhower.sph.harvard.edu/galaxy). Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was performed to identify functional genes in the sampled microbial community based on the Kyoto Encyclopedia of Genes and Genomes pathway database (16). Predicted metagenomics pathways were tested for potential significance by Student t test using the function t.test() in R. Differences in operational taxonomic unit (OTU) number and taxonomic relative abundance were analyzed by the Wilcoxon signed-rank test using the function wilcox.test() in R. Spearman correlation between gut microbiota and host markers was calculated using the function cor.test() in R. A P value of less than 0.05 was considered statistically significant.

Availability of data and materials

The data sets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Participants, sample collection, and study execution

Ramadan-associated intermittent fasting is from dawn to sunset (∼16 h in the present study) and is typically adhered to for a consecutive 30 d. In this study, participants were kindly requested to restrain from overeating or binge eating at each meal during the Ramadan month. We first prospectively enrolled 30 healthy nonobese young men (average age 19 y) who expressed the intention to fast for 30 d during Ramadan. These volunteers were from Kunming city of Yunnan province in China (Table 1; Supplementary Figure 2), where white rice or rice noodles are staple foods and the mainstay of calorie intake. During the study, stool samples from each participant were collected at 3 time points to allow assessment of individual-specific dynamic changes in gut microbiome (Figure 1A). In addition, changes in selected metabolic parameters were assessed. We observed that 30 d of intermittent fasting caused a statistically significant loss of 3.45% in body weight in this young adult cohort (Figure 1B), which is comparable to what was observed by Stekovic et al. (17) after 4 wk of strict alternate-day fasting in a volunteer cohort. In addition, we included a middle-aged (on average 40 y old; Table 2) cohort from Xining city of Qinghai province in China who adhered to 30 d of Ramadan fasting in 2018 and compared the results with a nonfasting matched control group (participants matched by age, body weight, and BMI), which was sampled at the same time points as the fasting group (Table 2; Supplementary Figure 3). In this area, wheat flour is the most important staple food and calorie source. Fecal and blood samples were collected before (T1), after (T2), and 1 mo after the cessation of fasting (T3) (Figure 2A). Also, in this cohort, the Ramadan-inspired intermittent fasting provoked statistically significant weight loss (Supplementary Figure 4A), which was driven by a reduction in body fat (Table 3). It should be noted that the body fat mass of 3 of the fasting volunteers was much higher than that of the nonfasting volunteers at T1, thus leading to a higher average level of body fat in the Ramadan group. This effect was not statistically significant (Supplementary Figure 4B). In this cohort, we also analyzed blood enzymes and observed that especially liver enzymes improved during intermittent fasting, especially compared with nonfasting controls, in line with the beneficial effects associated with intermittent fasting in general (Table 3) (18). We thus concluded that our study cohorts would allow us to make meaningful statements on intermittent fasting–provoked changes in the microbiome and relate effects to changes in physiologic parameters.

Intermittent fasting and microbiome diversity

Taxonomic diversity is an important characteristic of the human microbiome (19), and generally speaking, more diverse microbiomes are considered healthier (20). An often-used measure to quantify diversity in microbiomes is the Shannon–Weaver index (21), and thus we endeavored to determine the effects of intermittent fasting on the ecological diversity of the gut microbiota by quantifying this index in the samples collected. To this end, fecal bacterial compositions were analyzed by 16S ribosomal RNA gene sequencing. Following OTU-calling, it was found that in the young cohort, the diversity of gut microbiota, as indicated by the Shannon–Weaver index and other indices, was statistically significantly increased following intermittent fasting (Figure 1C; Supplementary Table 1). In the middle-aged cohort, the Shannon–Weaver index showed a slight upward trend, both compared with controls as well as compared with the pre–intermittent fasting state (Figure 2B), but this effect was not statistically significant (Supplementary Table 2), probably because of age-related factors such as immunosenescence and differences in nutrient absorption. These changes in apparent gut microbiome diversity provided a first indication that intermittent fasting affects the composition of the gut microbiome.

Intermittent fasting remodels the gut microbiome

The notion that the gut microbiome is remodeled during intermittent fasting was confirmed by Bray–Curtis PCA of overall gut microbiota structure after intermitting fasting (the young cohort: ANOSIM test, P < 0.001; Figure 1D; the middle-aged cohort: ANOSIM test, P < 0.001; Figure 2C). Interestingly, the gut microbial community showed a significant trend (ANOSIM test, P < 0.001) of return toward baseline conditions after the discontinuation of fasting (Figure 2C). Further confidence in these data was bolstered by the observation that Bray–Curtis PCA of the non–intermittent fasting control group of the middle-aged cohort showed that microbiome composition did not change in the study period (Supplementary Figure 5), also in agreement with the notion that gut microbiomes tend to be stable when lifestyles are not changed (22). We thus concluded that intermittent fasting provoked reversible alterations in the composition of the gut microbiome.

Intermittent fasting increases butyric acid–producing Lachnospiraceae

To further characterize the alternative microbiota taxa abundance associated with intermittent fasting that accounts for the greatest differences before and after fasting, we performed LEfSe analysis (23). As shown in Figure 3A, multiple taxonomic differences were found when intermittent fasting in the young adult cohort was contrasted to results from the prefasting microbiome. Among the effects, especially upregulation of OTUs belonging to the phylum Firmicutes was evident, which was driven, among other changes, by an increase in OTUs mapping to the order Clostridiales (Supplementary Figure 6A; Supplementary Tables 3–4). Of note, the family Lachnospiraceae did not reach a significance threshold in LEfSe analysis of this cohort. Conversely, OTUs belonging to the phylum Bacteriodes and especially the family Prevotellaceae were reduced following intermittent fasting. These changes appeared more related to the intermittent fasting per se rather than the intermittent fasting–provoked reduction in BMI in general, as no correlation was found between gut microbiota composition and BMI before fasting (Figure 3B). However, we did detect that following fasting, there was a positive correlation between BMI and the abundance of OTUs belonging to the phylum Proteobacteria, whereas after fasting, a negative correlation between BMI and abundance of the class Negativicutes and order Selenomonadales became apparent (Figure 3C).

In the middle-aged cohort, almost no taxa were changed in nonfasting control participants during the study period (Supplementary Figure 7). LEfSe analysis of the gut microbiome of the intermittent fasting participants of this middle-aged cohort, however, revealed major shifts in microbiome composition, which were largely similar to those observed in the younger cohort (Figure 4A). Again, the abundance of the order Clostridiales was significantly increased after fasting (Figure 4B), with this effect being dependent on increased abundance of the Lachnospiraceae and Ruminococcaceae families (Supplementary Tables 5–6). Concentrations of the former family quickly returned to baseline following cessation of fasting, but increased concentrations of Ruminococcaceae were more resilient in this respect (Supplementary Figure 6B; Supplementary Figure 8). Also mirroring the effects seen in the younger cohort, we observed in the intermittent fasting group of the middle-aged cohort decreased abundance of the Prevotellaceae. In this cohort, a correlation between the concentration of serum urea and the abundance of OTUs belonging to the family Lachnospiraceae and genus Agathobacter was detectable at the end of fasting (Figure 4C) but not detectable before fasting (Supplementary Figure 9A) as well as following cessation of fasting (Supplementary Figure 9B). Functional analysis using PICRUSTs showed 29 pathways that were present and 60 that were absent in the young cohort (Supplementary Figure 10A), as well as 14 pathways that were present and 24 that were absent in the middle-aged cohort at the end of fasting when contrasted to pathways present at the start of fasting (Supplementary Figure 10B). Among other observations, the results are in agreement with an increase in mucin degradation-associated metabolic functions in both cohorts during Ramadan. To assess the levels of calorie restriction associated with the effects of intermittent fasting, we estimated the calorie intake via FFQs over the whole course of study. We found that energy intake during the month of fasting was significantly reduced compared with that before or 1 mo after intermittent fasting or with that of unfasted controls (Figure 4D), which fits well with previously published data in this respect (10).