The participants were recruited using posters at Theragen Bio (Seongnam, Gyeonggi, South Korea). The eligible subjects were healthy men and women between 20 and 65 years of age. Individuals who consumed pro-, pre-, or antibiotics within 3 months before the study were excluded. All participants provided written informed consent, and the study protocol was approved by the Institutional Review Board (IRB) of Theragen Bio (IRB No. 700062-20180905-JR-005-01).

Saengshik (Eromplus Cellfood Goodmorning Saengshik Premium) was provided by Erom Co., Ltd. (Seongnam, Gyeonggi, South Korea) and prepared by mixing freeze-dried and powdered ingredients and transferring them into plastic packaging. The formula contained grains, beans, vegetables, fruits, mushrooms, seaweed, and other diverse raw materials (Supplementary Table 1). The nutrient contents were measured at SGS Korea Co., Ltd. (Uiwang-si, Gyeonggi-do, Korea) and are summarized in Supplementary Table 2.

This study was a single-group trial conducted at Theragen Bio from October 2018 to February 2019. On day 1 of subject arrival, they were provided enough Saengshik powder in 40-g packages and enough for 8 weeks of consumption. Subjects were instructed to consume 40 g of Saengshik powder with 100–150 mL of water daily over the 8-week treatment period. Although not assigned a time to eat Saengshik, subjects were asked to report whether they ate Saengshik daily. After 8 weeks, subjects that failed consume Saengshik more than once were excluded from the analysis due to poor compliance. Additionally, all participants were instructed to maintain their regular diet patterns and avoid any pro- and prebiotic supplements during the study period. Fecal samples from the participants were collected using a commercial fecal sampling kit (OMNIgene GUT; OMR-200; DNA Genotek, ON, Canada) at two fixed time points (at baseline and the study endpoint) for gut microbiota-profiling analysis (Supplementary Figure 1). The instructions specified in the stool sampling kit were explained, with each subject instructed to collect a specific amount of stool.

Habitual dietary intake was assessed at baseline using the food frequency questionnaire (FFQ) method, which was developed and validated by the Korea Centers for Disease Control and Prevention (23). The FFQ included 103 food or beverage items with three serving sizes and nine intake-frequency questions, with the survey used to assess the items consumed over the previous year. The validity and reproducibility of the FFQ has previously been verified (23). According to the food intake measured by the FFQ, the amount of calories and nutrients consumed daily by each individual was calculated and corrected to a level of 1000 kcal for use in creating subgroups based on nutrient intake.

The metagenomic DNA of each stool sample was extracted with a QIAcube instrument using a QIAamp DNA stool mini kit (Qiagen, Hilden, Germany) according to manufacturer instructions along with an additional bead-beating procedure. Briefly, 250 aliquots of stool samples were transferred to a tube together with 1.2 mL of ASL lysis buffer and 0.3 g of sterile 0.1 mm zirconia beads (BioSpec, Bartlesville, OK, USA) and homogenized twice at 30 Hz for 1 min on the TissueLyser system (Qiagen) (24). The third to fourth hypervariable region (V3–V4) of the 16S rRNA gene from the extracted DNA sample was amplified and sequenced using the Illumina MiSeq 2 ×300 System (Illumina, San Diego, CA, USA).

The DADA2 pipeline (25) of the QIIME2 package (v.2019.1; https://qiime2.org) (26) was used to generate unique sequence variants by filtering low-quality sequences and chimeras. Taxonomy was assigned using a pre-trained Naïve Bayes classifier based on the Greengenes 13_8 (27) 99% operational taxonomic units (OTUs), which had been trimmed to include the V3–V4 region. Chimeric reads were filtered from the sequencing data using UCHIME software (28). After excluding samples with low sequencing depths (<10,000 reads), the gut microbial compositions were analyzed in 66 paired samples (29 men and 37 women) and compared before and after Saengshik supplementation. The relative abundance of bacterial genera was calculated from sequencing data at baseline, with this data used to establish the enterotype of the subject.

Statistical analysis was performed using GraphPad Prism software (v.8.0.1; GraphPad Software Inc., San Diego, CA, USA) and R studio (v.1.1.447) (29). Differences in genera and phyla between before and after Saengshik supplementation were assessed using a Wilcox test in R studio using the “wilcox.test (paired=TRUE)” function from the “stats” R package. Significant changes in bacterial taxa between groups were calculated using a two-tailed Mann–Whitney U test (GraphPad Prism v.8.0). To cluster the gut microbial community of each subject, enterotype assignment was performed based on a published tutorial (http://enterotype.embl.de/enterotypes.html) (22). Briefly, the 66 samples were clustered using the PAM (partitioning around medoids) clustering algorithm in the R package “cluster,” and the optimal number of clusters was chosen based on the Calinski–Harabasz index using the “clusterSim” R-package. The result of enterotype clustering was visualized in a principal component analysis plot using the R package “ade4.” The level of statistical significance was set at P < 0.05.

A total of 102 individuals who were assessed for eligibility were included in this study, with 11 participants withdrawing before the end of the study. After an 8-week treatment period, we performed 16S rRNA sequencing of fecal samples before and after intervention from 78 participants that completed the study. In total, 12 subjects (15.4%) withdrew from the data analysis owing to an insufficient number of sequencing reads (<10,000 reads) and low compliance. Therefore, the gut microbial composition of 66 participants at baseline was compared with that at post-intervention, as presented in the schematic procedure of the intervention study (Figure 1). Table 1 shows the daily energy and nutrient intake of participants at baseline. At baseline, participants demonstrated a lower intake of energy, carbohydrates, proteins, fats, vitamin B₂, folate, calcium, phosphorus, potassium, iron, and zinc than that of an average Korean, although intake of vitamins A and C were higher than that of the average Korean. Daily intake of vitamin B₁ and niacin were similar to the Korean average (Table 1). The baseline characteristics of the participants are shown in Table 2. The 66 participants comprised 29 men and 37 women, with a mean age of 40.89 ± 9.94 years.

To evaluate the effect of Saengshik administration on gut microbiota profiles, we performed 16S rRNA sequencing analysis of stool samples collected at baseline and post-intervention. First, we evaluated β-diversity by calculating the weighted and unweighted unique fraction metric (UniFrac) distances, finding that the overall gut microbial community differed significantly between baseline and post-intervention (weighted F = 6.365, P = 0.001; unweighted F = 7.489, P = 0.001; PERMANOVA) (Figures 2A,B). To measure differences in diversity and evenness between baseline and post-intervention samples, we calculated Pielou's evenness (pielou_e), Shannon, Faith's phylogenetic diversity (Faith's PD), and observed OTU indices, which revealed significant decreases in α-diversity following intervention as compared with that at baseline (pielou_e, P = 0.0125; Shannon, P < 0.0001; Faith's PD, P < 0.0001; observed OTUs, P < 0.0001; Wilcoxon signed-rank test) (Figure 2C).

To detect the specific bacterial taxa that were significantly affected by Saengshik consumption, we compared the composition of gut microbiota at the phylum and genus levels before and after intervention. Figure 3A shows the relative abundance (%) of each bacterial phylum of participants as different color bars. The distribution of Bacteroidetes increased after Saengshik consumption, whereas that of Verrucomicrobia, Proteobacteria, and Actinobacteria decreased, with statistically significant changes observed in the relative abundance of all phyla [Bacteroidetes (P = 0.0002), Proteobacteria (P < 0.0001), Actinobacteria (P < 0.0001), and Verucomicrobia (P = 0.0093)] (Figure 3B). At the genus level, we identified 23 genera showing significant differences in abundance between baseline and post-intervention (P < 0.01, Wilcoxon signed-rank test), with Figure 3C showing fold changes for each genus. The relative abundance of Faecalibacterium, Bacteroides, and Bifidobacterium significantly increased after Saengshik consumption, whereas that of Alistipes, Megamnas, Odoribacter, Oscillospira, Anaerotruncus, Sutterella, Dialister, Butyricicoccus, Bilophila, Phascolarctobacterium, (Eubacterium), Fusobacterium, Veillonella, Holdemania, Butyricimonas, Akkermansia, Streptococcus, Collinsella, Weissella, and Clostridium significantly decreased (Figure 3C). Collectively, these data demonstrate that daily intake of a plant-based food powder altered gut microbial communities and decreased gut microbial diversity. Furthermore, we identified specific bacterial phyla and genera that were significantly affected by Saengshik consumption.

The human gut microbiome reportedly comprises two or three strains of major indigenous bacteria, which are referred to as the enterotype. Enterotypes reportedly reflect the unique gut microbial characteristics of individuals and are unaffected by body mass index (BMI), sex, or race. To verify whether enterotypes can affect compositional changes in gut microbiota after Seangshik consumption, we classified participant enterotypes and compared changes in gut microbiota among enterotypes. Following identification of a Calinski–Harabasz index of k = 2 as an optimal number of clusters, we confirmed that 66 participants were separated into two distinct clusters (Figures 4A,B), of which 43 belonged to the Bacteroides-rich enterotype (B-type) and 23 to the Prevotella-rich enterotype (P-type). The relative abundance of Bacteroides was significantly greater in B-type than that in P-type (P < 0.0001), whereas Prevotella was more abundant in P-type than in B-type (P < 0.0001) (Supplementary Figure 2). Additionally, we observed no statistically significant differences in age, height, weight, or BMI between individuals characterized as B- or P-type (all P > 0.05) (Supplementary Table 3). We then examined whether changes in the relative abundance of genera Figure 3C differed significantly according to enterotype before and after Saengshik intake. Figure 4C shows that the relative abundance of Bacteroides (P = 0.028) was significantly higher in B-type individuals relative to P-type individuals after Seangshik consumption, whereas the relative abundance of Eggerthella (P = 0.010) and Fusobacterium (P = 0.021) decreased significantly decreased in B-type relative to P-type individuals. However, Dialister (P = 0.020), Anaerostipes (P = 0.022), and Catenibacterium (P = 0.035) decreased significantly in P-type relative to B-type individuals after Seangshik consumption (Figure 4C). These observations suggest that enterotypes affected the response of gut microbial composition to Saengshik consumption.

To investigate whether habitual dietary patterns affect changes in gut microbial composition after Saengshik consumption, we compared changes in gut microbiota among subgroups classified by habitual dietary patterns. Figure 5A shows the results of clustering each subject based on the amount of micronutrient intake per 1000 kcal/day at baseline and according to FFQ data. We specified these two subgroups as dietary pattern 1 (DP1; n = 31) and DP2 (n = 35). There were no significant differences in age, height, weight, or BMI between DP1and DP2 (all P > 0.05) (Supplementary Table 4). Those in DP1consumed higher amounts of carbohydrates, fiber, vitamin C, and vitamin E, whereas those in the DP2consumed higher amounts of protein, fat, cholesterol, vitamin B₁, retinol, carotene, and zinc (all P < 0.05) (Supplementary Table 5). We then determined whether changes in the relative abundance of genera (Figure 3C) differed significantly according to individual dietary pattern in each subgroup and compared these changes before and after Saengshik intake. The results showed that the relative abundance of Faecalibacterium (P = 0.043), Clostridium (P = 0.015), and Coprobacillus (P = 0.036) was significantly higher in DP2 than in DP1 after Seangshik consumption (Figure 5B). Moreover, the relative abundance of Oscillospira decreased significantly in DP2 as compared with that in DP1 (P = 0.023); however, Veillonella decreased significantly in DP1 as compared with that in DP2 (P = 0.037) (Figure 5B). These results indicate that compositional changes in gut microbiota after Saengshik consumption were differentially affected by individual dietary patterns.

To determine whether enterotype and habitual dietary patterns can collectively influence changes in gut microbial genera after Saengshik consumption, we compared changes in 23 genera (Figure 3C) between subjects with different dietary patterns in the same enterotype. In the B-type, 20 and 23 individuals belonged to the DP1 and DP2 clusters, respectively, whereas in the P-type, 11 and 12 individuals belonged to each respective cluster (Supplementary Figure 3). Notably, three of 29 genera were affected by enterotypes and habitual dietary patterns. After intervention and among B-type individuals, the relative abundance of Faecalibacterium (P = 0.018, Mann–Whitney U test) and Veillonella (P = 0.005, Mann–Whitney U test) increased in DP2 individuals but not in DP1 individuals. There was no difference in changes in the abundance of these two genera between DP1 and DP2 P-type individuals (Figures 6A,B). Among P-type individuals, DP2 showed a larger decrease in the relative abundance of Bilophila than DP1 individuals (P = 0.042, Mann–Whitney U test) (Figure 6C). Collectively, these results indicate that Saengshik consumption changed gut microbial communities, which were influenced by not only individual habitual dietary patterns but also enterotype.

Previous studies revealed that plant-rich diets are related to increased bacterial diversity because of their capability to provide a source of microbiota-accessible carbohydrates (30, 31). Compared with plant-rich diets, a Western-style diet is characterized by high intake of saturated fat and simple carbohydrates, whereas low intake of fiber is associated with reduced bacterial diversity (32). Additionally, microbial diversity is inversely associated with chronic diseases and metabolic dysfunction (33, 34). Therefore, we expected that Saengshik intake would improve gut microbial diversity; however, contrary to our expectation, we observed significant decreases in gut microbial diversity after 8 weeks of Saengshik consumption relative to that at baseline. Given that the average daily fiber intake of Korean adults is 24.3 g (35), we speculated that the amount of fiber contained in the product used in this study (Supplementary Table 2) would not contribute to increasing the diversity of intestinal microbes. Multiple in vivo and in vitro studies have revealed that phenolic compounds and flavonoids exhibit antibacterial activity against several pathogenic bacteria (36, 37). However, discrepancies in the antibacterial properties of even the same phenolic compounds have been reported among bacterial strains (38). Therefore, the antimicrobial activity of various phenolic compounds and flavonoids present in Saengshik might selectively affect different strains of gut microbiota, which might have contributed to the reduced diversity observed in the present study. Another possibility is that the various food components consumed as complex combinations in daily meals during the intervention period might have had an interactive effect (39), although this factor was not ruled out in the present study. Furthermore, the present study was conducted in healthy adults who already showed relatively high gut microbial diversity. The possibility of individuals with low bacterial diversity being more responsive to Saengshik intake should be considered. Additionally, some interventions or cross-sectional studies report contradictory findings, including cases in which changes in diversity were not observed (40–42) or those in which diversity decreased after a plant-based diet (43, 44). Thus, to confirm the effect of Saengshik intake on gut microbial diversity, well-controlled and designed clinical trials are needed for individuals with decreased microbial diversity.

We found that the population of Faecalibacterium increased significantly after Saengshik consumption, which might be related to positive effects on health. This result is consistent with the correlation between Faecalibacterium and plant-food intake revealed reported in previous studies (40, 45). Faecalibacterium can utilize non-digestible carbohydrates (dietary fiber) from host mucus and concomitantly produce butyrate, which is associated with several beneficial health properties (46). This might be a factor related to the beneficial effects of Saengshik, including its anti-obesity and anti-inflammatory effects. Moreover, recent studies demonstrated an inverse association between Faecalibacterium and the risk of diseases, such as obesity (47), inflammatory colitis (48), and diabetes (49). In the present study, we also showed that the abundance of Akkermansia and Collinsella decreased after Seangshik consumption. Akkermansia exhibits probiotic potential that has been linked to maintaining intestinal barrier integrity, anti-inflammatory properties, and immune function (50). By contrast, other studies reported that Akkermansia spp. exacerbated gut inflammation in an animal infection model based on their ability to degrade mucins (51). Furthermore, elevated Collinsella abundance is reportedly positively associated with increased levels of insulin, cholesterol, and triglycerides (52) and negatively correlated with vegetable intake (53, 54). In line with previous studies, the present results suggest that a lower abundance of Collinsella might be mediated by Saengshik consumption. Nevertheless, the reduced abundance of beneficial microbes, including Butyricimonas, Megamonas, Phascolarctobacterium, and Butirycicoccus, should be verified through further studies. Collectively, although it is difficult to conclude that the changes in intestinal microflora caused by Saengshik would have a positive effect on host health, these findings show a possible link between gut microbiota, Saengshik consumption, and physiological properties.

Saengshik contains many micronutrients and compounds, including phenolics and flavonoids. Studies suggest that the gut microbiome-altering effects of Saengshik might originate from a variety of these ingredients. Indeed, several clinical trials report that polyphenol consumption can modulate gut microbial composition, anthropometric variables, and clinical markers (55). In the present study, we found that Saengshik consumption increased the abundance of Bacteroidetes, Bifidobacterium, and Bacteroides. Consistent with these findings, an intervention study reported increases in these microbes following intake of polyphenol-rich foods, such as red wine, freeze-dried cranberry powder, fruits, and vegetables (15, 56–58). Moreover, Bacteroides and Bifidobacterium are negatively correlated with various clinical markers, such as tumor necrosis factor-α, cholesterol, blood pressure, and lipopolysaccharide (55) levels. These results support our findings that Saengshik might be involved in beneficial changes in biological markers. In addition to being a good source of flavonoids and phenolic compounds, Saengshik is high in folate. The serving size of Saengshik (40 g) used in this study contained folate at a level of 83% of the recommended dietary allowance for Korean adults (400 μg) (59); therefore, the benefit of a sufficient folate intake could be expected. Mammals cannot synthesize their own folate de novo; therefore, they need to obtain folate via the diet and gut microbes, such as Lactobacillus and Bifidobacterium (60). Mice with folate deficiency reportedly show a significant increase in the Firmicutes:Bacteroidetes ratio (61), and cross-talk between gut microbiota and folate can modulate the regeneration of intestinal epithelial cells through stem cells by epigenetic modification (62). Although it remains unclear whether dietary folate intake is directly attributable to the composition and metabolism of gut microbiota (63), the consumption of sufficient folate (i.e., >3-fold greater than the habitual intake of the study participants) might contribute to microbial changes.

A notable finding in the present study was the inter-individual variability in response to the intervention, which was similar to results from clinical trials investigating the effect of diet on gut microbiota. Here, we showed that B-type individuals in the DP2 group showed a stronger increase in Faecalibacterium levels than those with DP1 group. Faecalibacterium reportedly plays a crucial role in maintaining human health owing to its production of anti-inflammatory metabolites or short-chain fatty acids. Indeed, Faecalibacterium is less abundant in patients with many diseases relative to healthy populations (47–49). Additionally, we showed that Veillonella abundance increased in individuals with a B-type enterotype and those with unhealthy dietary habits, whereas it decreased in individuals belonging to the B-type and DP1 groups. A recent study on healthy non-obese adults reported that increased abundance of Veillonella was negatively associated with saturated fatty acid consumption and body fat percentage (64). Furthermore, Veillonella abundance increases in women with high fiber intake, with its abundance associated with improved insulin sensitivity (65). Based on these findings, we suggest that adults with “less healthy” diet patterns and a Bacteroides-dominant enterotype might experience greater Saengshik-mediated positive health benefits. Moreover, these results suggest that further investigation of the role of habitual dietary patterns or enterotypes in evaluating dietary intervention is warranted.

This study has several limitations. First, given to the descriptive nature of the study and the absence of related clinical outcomes, we cannot provide insights into the beneficial properties of changed bacterial taxa related to Saengshik consumption. Second, the study lacked use of a placebo or blank control and presented a small sample size. Third, given that we did not collect 3-day or 24-h diet records of participants, we cannot completely rule out the possibility that differences in the usual dietary intake before and after intervention could have affected the outcomes. Finally, pre-selected participants with dysbiosis or low bacterial diversity rather than heathy individuals might be more suitable for identifying the effect of Saengshik on gut microbiota. Therefore, well-designed and controlled clinical trials with a larger sample size are necessary to determine the biological effect of Saengshik on the gut microbial community.

Despite these limitations, this represents the first study providing evidence that Saengshik consumption by healthy adults can influence gut microbial composition and diversity. Furthermore, the observed compositional changes were influenced by not only baseline gut microbial composition (enterotype) but also habitual dietary intake. We believe that this study underlines the need to investigate both baseline gut microbiota composition and diet patterns to allow the design of personalized interventions for diet-induced changes in gut microbiota and human health.