Black Raspberry Extract Shifted Gut Microbe Diversity and Their Metabolic Landscape in a Human Colonic Model

Abstract

Human gut microbiota is critical for human health, as their dysbiosis could lead to various diseases such as irritable bowel syndrome and obesity. Black raspberry (BRB) has been increasingly studied recently for its impact on gut microbiota as a rich source of phytochemicals (e.g., anthocyanin). To investigate the effect of BRB extract on the gut microbiota composition and their metabolism, an in-vitro human colonic model (HCM) was utilized to study the direct interaction between BRB and gut microbiome. Conditions (e.g., pH, temperature, anaerobic environment) in HCM were closely monitored and maintained to simulate the human intestinal system. Fresh fecal samples donated by three young healthy volunteers were used for gut microbiota inoculation in the HCM. 16S ribosomal DNA sequencing and liquid-chromatography mass spectrometry (LC/MS) based metabolomics were performed to study the impact of BRB on gut microbiota characteristics and their metabolism (fatty acids, polar metabolites, and phenolic compounds). Our data suggested that BRB intervention modulated gut microbiota at the genus level in ascending, transverse, and descending colons. Relative abundance of Enterococcus was commonly decreased in all colon sections, while modulations of other bacteria genera were mostly location-dependent. Meanwhile, significant changes in the metabolic profile of gut microbiota related to fatty acids, endogenous polar metabolites, and phenolic compounds were detected, in which arginine and proline metabolism, lysine degradation, and aminoacyl-tRNA biosynthesis were mostly regulated. Moreover, we identified several significant associations between altered microbial populations and changes in microbial metabolites. In summary, our study revealed the impact of BRB intervention on gut microbiota composition and metabolism change, which may exert physiological change to host metabolism and host health.

Graphical Abstract

1. Introduction

It is now well-known that our large intestine, which is a primary fermentation site in the body, contains up to 10¹² microorganisms with an estimation of 100-fold more bacterial genes than human genes (1, 2). These microorganisms consist of the microbial community which is commonly referred as gut microbiota. The interactions between gut microbiota and host health are dynamic and complicated, and dysregulation of gut microflora is often associated with adverse health outcomes such as obesity, irritable bowel syndrome, and even colorectal cancer (3). Increasing evidence indicates that diet is a key element to regulate gut microbial community. Recently, multiple studies reported the effect of black raspberries (BRB) on gut microflora (4–6). BRB is a rich source of bioactive phytochemicals such as anthocyanin and ellagic acid. BRB also contains a wide range of nutrients such as fiber, folic acid, calcium, and vitamin C(4). It is known to increase beneficial gut microbe species such as Akkermensia muciniphila (4–6) and inhibit pathogenic microbes like H. pylori (7). Also, BRB has promoted short-chain fatty acid (SCFA) production by gut microbes to exert beneficial effects to human health (5).

Despite these progresses, there is a knowledge gap in how BRB intervention leads to the systematic changes of microbial metabolism rather than just the frequently reported alternation of SCFAs; as well as in the detailed description of the interplay between the modification of microbial composition and the changes of a variety of metabolites. To study the relationship between BRB intervention and gut microbes, preclinical and clinical models have been utilized previously (4–8). While these studies have provided interesting knowledge relating to BRB impact on gut microbes, the observations were a mixed effect of diet-gut microbes-host interactions. In order to specifically investigate the gut microbe response to diet (BRB), we utilized a recently developed, in vitro three-stage human colonic model (HCM) (9) to detect locational specific changes of gut microbial population and gut microbial metabolism. In addition, HCM allows us to study host-free regulation on gut microbes resulting from dietary interventions and create a gut environment that can be constantly monitored and sampled without any ethic concerns. We hypothesized that BRB intervention could boost growth of beneficial gut microbes, and modulate their metabolite productions that may have potential impact to human health. A multi-omics approach combining 16S rRNA gene sequencing with large-scale metabolomics analyses of endogenous microbial metabolites (e.g., fatty acids, amino acids, etc.) was performed. Both longitudinal and locational gut microbial population modulations, and the microbial metabolic changes were carefully surveyed within our HCM system to reveal the BRB induced eco-system shift and chemical landscape reforming.

2. Materials and Methods

2.1. Chemicals

Short-chain and medium-chain fatty acid standards including formic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, valeric acid, hexanoic acid, isocaproic acid, and n-heptanoic acid were purchased from Sigma–Aldrich (St. Louis, MO, USA). Phenolic compound standards including protocatechuic acid, 3,4-Dihydroxyphenyl ethanol, 3,4-dihydroxyphenylacetic acid, gallic acid, caffeic acid, isoferulic acid, urolithin A, trans-resveratrol, equol, daidzein, homovanillic acid sulfate, genistein, naringenin, phloretin, kaempferol, epicatechin, catechin, enterolactone, quercetin, hesperetin, epigallocatechin, isorhamnetin, myricetin, chlorogenic acid, roseoflavin, phlorizin, epicatechin gallate, epigallocatechin gallate, quercetin 3-D-galacoside, and rutin were purchased from Cayman company (Ann Arbor, MI, USA). The stable isotope-labeled amino acid mix (20 AA U-13C, 97–99%; U-15N, 97–99%) was purchased from Cambridge Isotope Laboratories (Tewksbury, MA, USA). HPLC-MS-grade acetonitrile, ammonium acetate, formic acid, and acetic acid were purchased from Fisher Scientific (Pittsburgh, PA, USA).

2.2. Gut bacteria isolation and HCM inoculation

Fresh fecal samples were donated by three male healthy volunteers with a mean age of 22.7 who do not have record of antibiotics usage for the six months before the sample collection (the protocol approval has been obtained by the Institute Review Board and informed consent was obtained from each individual). Samples were immediately transferred into anaerobic chamber (Coy Lab, Grass Lake, MI, USA) and mix thoroughly with autoclaved phosphate-buffered saline (PBS) at 1:150 (w:v) ratio. After standing for 5 minutes, supernatant, with a final concentration of 20% (v:v), was mixed with Gifu Anaerobic Broth (Himedia, West Chester, PA, USA.). After 24 hours incubation at 37°C, 3mL of gut bacteria culture was added in each vessel of the HCM system as starting culture.

2.3. BRB Extract Sample Preparation

BRB liquid extract was purchased from BerriHealth company (Corvallis, OR, USA). Sample was kept in a 4°C refrigerator. BRB supplement concentration was determined by a dose-dependent test. Original concentration of serial dilution used in dose-dependent study was based on BRB liquid extract product instruction (3mL BRB extract in at least 59.15mL water per day for daily supplement). Drop plate results indicated that 0.507% BRB extract in medium promoted the abundance of mixed gut bacteria (Fig. S1). As a result, we used 5.07 mL BRB extract in 1L medium as feeding medium in our intervention study, which is equivalent to 2.25 g powder extract / day and is comparable to some recent clinical studies (10–12).

2.4. Human Colonic Model (HCM)

HCM was constructed to resemble Macfarlane’s three-stage compound continuous culture system (13) and SHIME system (14) with modifications. The application of HCM in studying nutrition intervention on gut microbiota has been demonstrated in our earlier study (9). HCM is an in-vitro model consisting of three compartments simulating ascending, transverse, and descending colons (Fig. S2). Gut bacteria from healthy individuals was inoculated in each vessel and vessels were automatically controlled by computer software to maintain pH and temperature. Anaerobic environment was achieved by continuous nitrogen flushing through the entire system. Fresh medium was transferred into the system three times per day to mimic regular meal intakes. The entire experiment included 14 days of equilibrium phase to stabilize the system, 7 days of pre-treatment phase, 14 days of treatment phase, and 7 days of wash out phase. HCM samples were collected on a daily basis from each colon compartment and immediately stored at −80°C freezer for further metagenomics and metabolomics analysis. In addition, plate counting was done daily in order to keep track of gut bacteria count and the stability of HCM system.

2.5. Microbiome Analysis

Representative samples were selected according to drop plate results and processed after the entire sample collection procedure. DNA was extracted following an established protocol provided by Wizard Genomic DNA Purification Kit (Promega, Madison, WI, USA) (15–17). After chemically cell lysis, RNA and protein were precipitated and removed from each sample. DNA was precipitated and washed by ethanol to increase purity. Then, extracted DNA samples were used for library preparation following a previously reported protocol (9). Briefly, GoTaq^® Hot Start Colorless Master Mix (Promega, Madison, WI, USA) was used for V4 region amplification and each DNA sample was individually labeled by a 12-base Golay barcode. After PCR amplification, SequalPrep Normalization Plate kit (Thermo Fisher, Waltham, MA, USA) was used for amplicon purification and KAPA Library Quantification Kit Illumina Platforms (Kapa Biosystems, Wilmington, MA, USA) was used for quantification. Sequencing was achieved on Illumina MiSeq platform using a read length up to 2 × 250 bp at Miami University Center for Bioinformatics and Functional Genomics.

2.6. Sample preparation and Liquid-chromatography Mass Spectrometry (LC/MS)

Short-chain and medium-chain fatty acids, endogenous polar metabolites, and phenolic compounds were extracted and analyzed with LC/MS-based metabolomics platform according to their different chemical properties. For short-chain and medium-chain fatty acid, extraction protocol was reported previously with modification (18). HCM sample was mixed with acetonitrile (ACN), 3-Nitrophenylhydrazine hydrochloride (3NPH·HCl) solution, and N-(3-dimethylaminopropyl)-N’-ethylcarbodiimide (EDC·HCl) – 6% (v:v) pyridine solution (1:1:1:1, v/v/v/v). After 30 minutes of 40°C water bath, mixture was cooled on ice and 1.92mL of 10% (v/v) ACN was added. For endogenous polar metabolites, the extraction was performed following a previously reported protocol (19, 20). Briefly, 1mL HCM sample was washed with three rounds of phosphate buffer saline (PBS) and centrifuging. After adding 250 μL methanol and 50 μL ¹³C¹⁵N-labeled internal standard mixture to the bacteria pellet, vigorously vortex the mix and cool in −20 °C fridge for 20 minutes. After incubation, 150 μL supernatant was collected and loaded into a Speedvac system to dry. 250 μL of 50% (v/v) aqueous ACN was added to dissolve the pellet from Speedvac. The phenolic compounds were extracted following an established protocol with modification (21). Briefly, 150 μL sample was mixed with 1.5mL 80% (v/v) methanol and sonicated for 30 minutes. After 10 minutes of centrifuge, supernatant was collected and filtered for LC-MS analysis. All extracted samples were transferred into LC vials and loaded onto sample tray in Vanquish UHPLC System (ThermoFisher, Waltham, MA, USA). Different columns were used for different types of compound separation: Acquity UPLC CSH C18 Column (1.7um, 2.1×100mm, Water, Milford, MA, USA) for short-chain and medium-chain fatty acids, Xbridge BEH Amide (2.5um, 2.1×150mm, Waters, Milford, MA, USA) for endogenous polar metabolites, and XTERRA RP 18 (3.5um, 3.9×100mm, Waters, Milford, MA, USA) for phenolic compound analysis. Mass spectra were generated by Hybrid Quadrupole Orbitrap Q Exactive^™ mass spectrometer (ThermoFisher, Waltham, MA, USA). Parameters set up for each group of compound detection were listed in Table 1. Mobile phases, flow rate and gradient, MS method, resolution used, etc. were described in this table. Both positive and negative modes analyses were performed. In addition, pooled quality control (QC) samples for each type of metabolites were injected every ten biological samples to ensure the instrument stability. All metabolites were semi-quantitatively measured by LC/MS with comparable peak intensity but not absolute concentration.

Table 1:

Liquid chromatography mass spectrometry parameters of free fatty acids, polar metabolites, and phenolic compounds analysis.

Parameter	Short-chain and Medium-chain Fatty Acid	Endogenous Polar Metabolites	Phenolic Compound
Mobile Phase	A: 100% H2O, 0.01% formic acid B: 100% ACN, 0.01% formic acid	A: 10% ACN, 90% H2O, 5mM ammonium acetate, 0.2% acetic acid B: 90% ACN, 10% H2O, 5mM ammonium acetate, 0.2% acetic acid	A: 89.9% H2O, 10% ACN, 0.1% formic acid B: 69.9% H2O, 30% ACN, 0.1% formic acid
Flow Gradient	0–2min: 15% B; 2.1–11min: 55% B; 11.1–12min: 100% B; 12.2–15min: 15% B	0–0.1 min: 70% B; 5–9 min: 30% B; 1120min: 70%	0–1min: 10% B; 9 min: 20% B; 14 min: 45% B; 20 min: 55% B; 21–22 min: 100% B; 23–25 min: 10%
Flow Rate	0.35ml/min	0.3ml/min	0.5ml/min
Column Temperature	40°C	40C	40C
Method	Target-SIM, PRM	Full Scan	Full Scan, ddms2
Resolution	70,000	70,000	70,000
AGC target	3e6	3e6	1e5
Maximum IT	200ms	200ms	50ms
Scan Range	60–900 m/z	60–900 m/z	80–1200 m/z
Spray Voltage	2.75kv	2.75kv	2.75kv

2.7. Bioinformatics and Statistical Analysis

DNA sequences generated from the MiSeq run were analyzed using QIIME2 software(22) and were clustered into 97% identity using an operational taxonomic unit (OTU) picking protocol against the Greengenes reference database (23). The taxonomy assignments for OTUs were based on the Greengenes reference sequence using gg138 v4 human stool classifier. Resulting feature tables were converted to relative abundance tables. Unclassified taxonomy was excluded. Final abundance results were imported to MetaboAnalyst 4.0 (http://metaboanalyst.ca) (24) and PRISM for figure generation. The Quanbrowser module of Xcalibur 4.0 was used to manually process short-chain and medium-chain fatty acid profiling data based on in-house database containing nine fatty acid standards (formic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, valeric acid, hexanoic acid, isocaproic acid and n-heptanoic acid). Peak area data was filtered with coefficient variance (cv) of QC (<0.25) and normalized with bacteria number. MS-generated raw files for endogenous polar metabolites and phenolic compounds were processed by Compound Discover 3.1(ThermoFisher, Waltham, MA, USA). Exported data was manually exanimated and normalized by bacteria count for quality control purpose, then uploaded onto MetaboAnalyst 4.0 (24) for further statistical analysis. Partial least-squares discriminant analysis (PLS-DA) and variable importance of projection (VIP) scores were assessed to distinguish endogenous polar metabolite difference between each group. Heatmaps were generated to visualize the changes of phenolic compounds through experimental timeline.

3. Results

3.1. Microbial characteristics in HCM

To reveal longitudinal and locational changes of gut microbial population in each colon section in response to BRB intervention, we first examined their relative abundance generated by 16S rRNA sequencing and bioinformatic analysis by QIIME 2. In order to compare microbial composition before and after BRB intervention, we analyzed microbial composition at phylum and genus level from pre-treatment, treatment, and wash out stages. Among all colon sections, the most dominant phylum from the inoculated healthy people’s gut microbiota was Proteobacteria, which accounted for 87.6%, 79.1%, and 64.5% on average in ascending, transverse, and descending colons, respectively (Fig. S3). Other bacterial phyla that appeared in the human gut included Firmicutes, Actinobacteria, Bacteroidetes, Verrucomicrobi, and Chloroflexi. There was no significant change in gut microbiota composition at the phylum level corresponding to BRB intervention; at the genus level, however, significant changes among different experimental phases in each colon compartment were observed. Clear separation in the principal component analysis (PCA) score plot of gut microbiota at the genus level indicated an overall shifted bacteria genera composition after BRB intervention in all three colon sections (Fig. S4). More than 30 genera were identified from HCM samples in each vessel. The top 10 most abundant genera were shown in Fig. 1A–C. Dominant genera that were commonly existed in ascending, transverse, and descending colons were Citrobacter, Escherichia, Klebsiella, and Phascolactobacterium. Among all identified genera, three from ascending colon, six from the transverse colon, and four from the descending colon were significantly modulated by BRB intervention (Fig. 1D). Different colored boxes indicated different colon sections, respectively. Some bacteria genera were significantly impacted in more than one colon section. For example, the relative abundance of Enterococcus was significantly reduced (p<0.01) with BRB supplementation in all three colon sections, and the relative abundance of Citrobacter was significantly decreased (p<0.01) in the transverse and descending colons with BRB supplementation. For bacteria genera that were modulated in a single colon section, Eggerthella and Clostridiales were significantly lower (p<0.05) in ascending colon corresponding to BRB supplementation. This decreasing trend was later reversed in the washout phase, indicating a BRB-dependent effect. In transverse colon, in addition to Enterococcus and Citrobactor, Pseudomonas was also significantly reduced (p<0.05) with BRB supplementation, while this trend was not reversed after the seven-day washout phase without BRB supplementation, which may suggest a longer recovery time needed for these groups of bacteria. Meanwhile, Klebsiella, Tissierellaceae, and Clostridium significantly increased (p<0.05) corresponding to BRB intervention, among which, only the abundance of Tissierellaceae was reversed in the washout phase. In the descending colon, BRB supplementation led to a significant decrease (p<0.05) in the relative abundance of Clostriaceae, while a significant increase (p<0.05) in the relative abundance of Escherichia was observed.

Figure 1:

Bar graph of microbial composition change of (A) ascending, (B) transverse, and (C) descending colons at genus level during pre-treatment, treatment, and wash out phases. (D) The relative abundance of the significantly modulated bacteria genera in ascending (blue box), transverse (red boxes), and descending colon (green box). (E) PCA plot of locational comparison among ascending, transverse, and descending colons bacterial composition at genus level during treatment phase. Different bacteria genera were color-coded, and due to the large number of genera, only top ten genera with the highest abundance were individually labeled. Significance was assessed by t-test (✱: p-value<0.05; ✱✱: p-value <0.01; ✱✱✱: p-value<0.005; ✱✱✱✱: p-value<0.001). P1, pre-treatment phase; P2, treatment phase; P3, wash out phase.

To further entail the locational differences of BRB-dependent gut microbial modulation in the HCM model, we compared bacterial composition in different colon sections during the same experimental period. The gut bacterial compositions at genus level from the treatment phase in ascending, transverse, and descending colons were compared (Fig. 1E). PCA plot indicated a clear separation among three colon sections, which demonstrated that gut microbiota composition at the genus level was distinctively different in three colons during the BRB treatment phase. This separation can potentially be explained by the delayed expose of BRB contents at the transverse and descending colons, and may be attributed to the nature of the HCM design that mimics the continuous culturing environment from one colon compartment to another. As BRB supplementation was sequentially consumed by gut microbiota inoculated in three colon sections of HCM with different pH values, nature selections were most likely to take place, and the collective effects of fermentation stages and pH difference eventually resulted in the different bacterial composition among three colon sections.

3.2. Microbial metabolite regulation in HCM

As major metabolites produced by gut microbiota, seven short-chain and medium-chain fatty acids were semi-quantitatively measured after excluding two with QC CV > 0.25 (Fig. S5). As summarized in Fig. 2, four out of seven fatty acids were significantly modulated (p<0.05) by BRB intervention. Isovaleric acid, hexanoic acid and isocaproic acid productions were significantly promoted (p<0.05) in ascending colon. In transverse colon, hexanoic acid and n-heptanoic acid significantly increased (p<0.005) in relative abundance. In descending colon, same as transverse colon, the production of n-heptanoic acid by gut microbes was significantly enhanced (p<0.005). Taken together, BRB supplementation boosted the production of various short-chain and medium-chain fatty acids from gut microbes in all three colon sections of our HCM system.

Figure 2:

Short-chain and medium-chain fatty acid regulation in ascending, transverse, and descending colon from pre-treatment and treatment phase. Difference between pre-treatment phase and treatment phase was assessed by t-test (✱: p-value<0.05; ✱✱: p-value <0.01; ✱✱✱: p-value<0.001).

In addition to targeted analysis of fatty acids, an untargeted metabolomics approach was performed to characterize the endogenous polar metabolic profile of gut microbiota in our HCM model. Using our LC/MS system, 1839 and 2711 metabolic features were detected in negative and positive mode, respectively. After annotation, filtering and normalization, a total of 169 endogenous polar metabolites were identified and further analyzed. PLS-DA analyses of metabolic profiles for each colon, and the variable importance in projection (VIP) plots of top 10 metabolites with a VIP score >1.5 were presented (Fig. 3A–C). These results revealed the clear differences of the endogenous polar metabolites between pre-treatment and treatment phases, as well as highlighted those metabolites primarily contributed to these differences. It is interesting to note that, in VIP score plots, all metabolites that contributed most to distinct metabolic profiles showed an increasing trend in their relative abundance with BRB supplementation. Additional statistical analysis with t-test (p<0.05) and fold change (FC>2) identified a total of 32 metabolites that were modulated significantly after BRB intervention in three colon sections (Table S1). These endogenous polar metabolites were closely associated with arginine and proline metabolism, lysine degradation, and aminoacyl-tRNA biosynthesis (Fig. 4). From these pathway analyses, it is noticed that BRB supplementation mostly impacted gut bacteria amino acid metabolism. Furthermore, in order to compare the locational difference of endogenous polar metabolite modulation by BRB intervention, we compared polar metabolite profile from the treatment stage among ascending, transverse, and descending colons. Clear separation in the PLS-DA plot presented distinctive metabolic profiles among different colon sections (Fig. 3D). Five of the top 10 metabolites that contributed most to this difference included eicosenoic acid, linoleic acid, gamma-homolinolenic acid, guanine, and agmatine, which are involved in alpha-linolenic acid and purine metabolism.

Figure 3:

Partial least squares discriminant analysis (PLS-DA) and loading plots of mass spectrometry-based metabolomics results in (A) ascending colon, (B) transverse colon, and (C) descending colon in pre-treatment and treatment phase, and (D) locational comparison of ascending, transverse, and descending colons during treatment phase. Top 10 modulated metabolites were displayed in VIP plots.

Figure 4:

Scatter plot showed representative pathways that have been regulated after BRB intervention in ascending, transverse, and descending colons. Pathways that have −log(p) above 0.5 are labeled in lower case letters as: (a) Arginine and proline metabolism; (b) Lysine degradation; (c) Aminoacyl-tRNA biosynthesis; (d) Histidine metabolism; (e) Beta-Alanine metabolism; (f) Nicotinate and nicotinamide metabolism; (g) Butanoate metabolism.

As BRB is a rich source of phenolic compounds such as anthocyanin and ellagic acid (4), we also analyzed the metabolic change of phenolic compounds in gut microbiota. Nine phenolics (gallic acid, naringenin, phlorizin, phloretin, protocatechuic acid, quercetin, rutin, genistein, caffeic acid) from all three colon sections were detected in our HCM samples. Overall, the changed abundance of phenolic compound between treatment and wash out phases was shown in heatmaps for all three compartments (Fig. 5), in which eight out of nine identified phenolic compounds were significantly changed during our experiment. In ascending colon, other than genistein and gallic acid, all other detected phenolic compounds were significantly modulated (p<0.05). In transverse colon, gallic acid, naringenin, phlorizin, phloretin, rutin, and caffeic acid were significantly modulated (p<0.05). There was no significant change in descending colon for phenolic compounds. Overall, our data suggested that BRB supplementation leads to significant regulation in short-chain and medium-chain fatty acids, polar metabolites, and phenolic metabolism in gut microbes.

Figure 5:

Heatmaps of mass spectrometry-based detection of phenolics in (A) ascending colon, (B) transverse colon, and (C) descending colon in treatment and wash out phase. All 9 identified phenolics were displayed in the heatmap. (D) Clustered scatter plots of significantly modulated phenolics in ascending (blue box) and transverse (red box) colons from treatment phase to wash out phase. Difference between treatment phase and wash out phase was assessed by t-test (✱: p-value<0.05; ✱✱: p-value <0.01; ✱✱✱: p-value<0.005; ✱✱✱✱: p-value<0.001. P2, treatment phase; P3, wash out phase)

3.3. Relationship between the modulated gut microbiota and changes in gut metabolism

To understand the possible connection between the BRB-induced modulation of microbial population and their metabolism, Spearman correlation analyses were performed using top 10 endogenous polar metabolites (based on VIP score rank from PLS-DA analysis) and the significantly modulated (ρ<−0.8 or ρ>0.8) bacteria genera in each colon section (Fig. 6). The analyses revealed that in ascending colon, decrease in Clostridailes was negatively correlated (ρ<−0.8) with the increase of homogentisic acid, vernolic acid, bis (methylbenzylidene) sorbitol, and eicosenoic acid. In the transverse colon, Enterocuccos is negatively correlated (ρ<−0.8) to all discriminating metabolites except gentisic acid. Also, Citrobactor was negatively correlated (ρ<−0.8) with homogentisic acid, phenylacetate, monostearate, and icosanoate. On the other hand, icosanoate was positively correlated (ρ>0.8) with Klebsiella and Tissierellaceae. Klebsiella was also positively related (ρ>0.8) to homogentisic acid, eicosenoic acid, monostearate, and gamma-homolinolenic acid. In descending colon, only Enterococcus was negatively correlated (ρ<−0.8) with tetraglyme, Bis(methylbenzylidene)sorbitol, and δ-Valerolactam. In addition, the correlation analysis of gut bacteria and short-chain and medium-chain fatty acid, as well as phenolic compounds were performed and summarized. As shown in Fig. S6, Citrobacter and Klebsiella in transverse colon and Enterococcus in transverse and descending colons were strongly correlated (ρ<−0.8 or ρ>0.8) with n-hepatic acid. For phenolic compounds, Enterococcus was strongly correlated (ρ<−0.8 or ρ>0.8) with quercetin, rutin, and phlorizin in ascending colon. In the transverse colon, Klebsiella was strongly correlated (ρ<−0.8 or ρ>0.8) with naringenin, phloretin, and phlorizin (Fig. S7). In summary, these results revealed the potentially important relationships between the significantly changed metabolites/phenolics in the HCM system and the changes of microbial composition induced by BRB supplementation, which could lead to the generation of new study hypotheses that may focus on the particular microbe-chemical connections.

Figure 6:

Correlation analysis of significantly modulated gut microbes and endogenous polar metabolites with top 10 VIP scores in (A) ascending colon, (B) transverse colon, (C) descending colon. Significance was calculated by Spearman correlation analysis (✱: ρ<−0.8 or ρ>0.8).

4. Discussion

In this study, we utilized a three-stage, automated, computer-controlled in vitro system to investigate BRB intervention on gut microbiota donated by healthy subjects. This model allows us to observe gut microbial population and metabolism changes induced by dietary intervention without complicated host interactions, and allowed for simplified interpretation of nutrient-gut microbe-metabolites relationships. In addition, this model enables us to investigate locational differences of microbiota in the colon after BRB intervention. As previous evidences have indicated that incidence and pathogenesis of colorectal cancer exhibited difference depending on the location of the tumor (25–27), with our HCM system, we showed the probability of elucidating both longitudinal and locational differences in gut microbial modulation by BRB supplementation, and potential to assist further investigation of diet - gut microbe - cancer relationships within different colon sections.

As mentioned in the result section, BRB intervention significantly changed certain bacteria genera, such as Enterococcus and Citrobacter, in healthy gut microbiota. Previous studies indicated that BRB intervention was associated with the increased abundance of Akkermensia on healthy animal models (4–6). In our study, Akkermensia was at relatively low abundance but identified in all three colon sections. Furthermore, in transverse and descending colons, it showed a slightly increasing trend without a statistical significance after BRB intervention (Table S2). In addition, BRB supplementation resulted in a significant decrease of Enterococcus in all three colon sections in our study. Enterococcus is a group of Gram-positive, lactic-producing bacteria, which are relevant to human health for their role in associating with infections (28). There are two opportunistic pathogen species belonging to Enterococcus genus that contribute most to enterococcal infections: E. faecalis and E. faecium. At a high abundance level, they are more likely to compromise the host’s immunity, and they can develop resistance to common antibiotics (29, 30). Therefore, BRB supplementation may be used to improve host health status by decreasing the quantity of these bacteria.

While investigating the modulation effects of BRB on microbial metabolism, we are the first to report a large-scale metabolomics analysis for endogenous bacterial metabolism modulation with BRB intervention in an in vitro model. Compared to earlier studies, the increasing trend of isovaleric acid in ascending colon in our study is consistent with the previous finding of raspberry anthocyanin consumption in a mouse model (31). In addition, previous clinical studies illustrated that administration of short-chain fatty acids is positively associated with improvement in ulcerative colitis, obesity, metabolic syndrome, bowel disorders, and cancer (32–36). Medium branched-chain fatty acids like isovaleric acid and isocaproic acid are degradation products from amino acid fermentation gut bacteria. Thus, it is likely that BRB supplementation enhanced amino acid metabolism in our study as pathway analysis of endogenous polar metabolites suggested. For other berries like strawberry and blueberry, previous studies have reported microbial and human metabolome changes in response to berry polyphenol consumption (e.g. anthocyanin, flavanol, ellagitannin) (37). For example, homogentisic acid is considered as a metabolite of gut bacteria from anthocyanin digestion (38), which is consistent with our finding of enhanced production of homogentisic acid after BRB supplementation. Also, phenylacetic acid was previously reported from flavonoid degradation in microbial cell culture (39). We identified an increased abundance of phenylacetaldehyde in the transverse colon, which was the reduced form of phenylacetic acid, and could be hydrolyzed and oxidized to phenylacetic acid later with the assistance of host enzymes. Also, multiple amino acids (increase in histamine and L-histidine in descending colon) were significantly modulated in our study. Endogenous amino acids can either be metabolites from certain gut bacteria (e.g. Klebsiella pneumoniae and Clostridium bifermentans) or excreted to serve as energy fuel for amino acid fermentation bacteria in the digestive tract (40). Also, we have found pathways involving amino acids like arginine and proline metabolism, lysine degradation, aminoacyl-tRNA biosynthesis, histidine metabolism, and beta-alanine metabolism were significantly regulated based on metabolomics analysis. For example, lysine, as one of the essential amino acids, can be utilized in butyrate production(41), which as previously mentioned, butyrate could exert more beneficial health effects such as reinforcing gut barrier and reducing oxidative stress in the colon(42). In addition, aminoacyl-tRNA biosynthesis was also reported to be regulated in our previous GTE treatment on healthy patients utilizing the HCM system (9). Modulation in aminoacyl-tRNA biosynthesis indicated BRB modulations on the bacterial translation process since aminoacyl-tRNA biosynthesis could generate aminoacyl-transfer ribonucleic acids (aa-tRNA) for transferring amino acids to the ribosome for translation purposes (43). Overall, increasing in microbial catabolites of phenolics and certain amino acids could have a further impact on host health.

5. Conclusion and future direction

In this study, we demonstrated the utilization of an HCM model for a nutrition intervention study that aimed to understand the BRB extracts - human gut microbiota – microbial metabolites relationships. Through this three-stage, automated, computer-controlled in vitro system, we investigated the impact of BRB intervention on the gut bacteria population and their metabolism. 16S rRNA sequencing revealed changes in gut bacteria relative abundance, certain bacteria genera, and bacterial metabolites including short-chain and medium-chain fatty acids, endogenous polar metabolites, and phenolic compounds, corresponding to BRB supplementation. Overall, HCM appears to be a suitable tool to conduct nutrition intervention on gut microbiota studies as demonstrated in this BRB experiment. It is acknowledged that certain limitations exist in using this HCM model for gut microbiota studies. Since it is a host-free model, we focused on the direct impact on gut bacteria and their metabolism from BRB supplementation. We were able to exclude other factors from the host that could influence microbial change. However, HCM cannot provide a perfect simulation of various regulations and signaling happened in the human gut. Intrinsic limitations of HCM includes lack of mucin layer so that not all bacteria species inoculated in these vessels can survive throughout the experiment. Therefore, we will continue our future studies using in vivo animal models and clinical studies to validate the findings discovered here, before practical dietary recommendations are provided.

Highlights:

• An in vitro study of gut microbial metabolic responses to black raspberry

• Development and application of a novel human colonic model

• Investigation of microbial response to dietary components

• Multi-omics approaches links microbes to their metabolites