Beneficial metabolic transformations and prebiotic potential of hemp bran and its alcalase hydrolysate, after colonic fermentation in a gut model

Changes in bacterial alpha and beta diversities

The microbiota diversity indices were analyzed to study the impact of HPBA on microbial population, to assess population’s stability during fermentation, and to compare its microbiota to that of other bioreactors (Figure ^S1). The baseline of value was compared to the endpoints of fermentation of different treatments. It is undisputable that a part of the effect of reduction in richness (Observed OTUs) was derived by the passage from in vivo to in vitro condition, but the focus must be set on the different trend that other alpha diversity indices had. For example, abundance (Chao 1) for HBPA was significantly higher at the end of fermentation (p < 0.05), while a not significant reduction was seen for HB or FOS. Surges in evenness (Shannon) were seen for HB (p > 0.05) and HPBA (p < 0.05), but no changes were seen in dominance (Simpson) (p > 0.05), while oppositely, FOS decreased in evenness (p > 0.05) and raised in dominance (p < 0.05). This output indicates a different performance of HPBA or minorly HB in respect to FOS and is well explained by the trend of dominance that tells that for FOS some taxa overcame others, reducing the uniform distribution of bacterial groups in the microbiota. This effect was already observed and could be justified by the ability of FOS to foster Bifidobacteriaceae and make them dominant over the microbiota^8,13. HBPA and minorly HB instead had an effect with a wider range of bioactivity on more bacterial targets; that higher biodiversity could be seen as an added value on its prebiotic potential.

When the bacterial diversity between samples (beta diversity) was examined with Bray–Curtis analysis, the fecal samples was set distant to the BL, and the BL distant to the end point cases, as demonstrated by principal coordinate analysis (PCoA) based on an unweighted (qualitative) phylogenetic UniFrac distance matrix. This feature confirms that shifts occurred during the experiments. Additionally, the four cases at the end point were relatively distant one to each other. This feature confirms that different shifts occurred from the BL on. So far, the study of biodiversity indicated the ability to keep an eubiosis conditions by fermentations of both the hempseed bran samples, with generally a higher capacity of HBPA in respect to HB. Considering that HBPA should have a higher availability of shorter fiber chains and more unbound saccharides due to the action of alcalase treatment², that result could indicate that HBPA is generally more appetible for colonic fermentation than HB.

Changes in taxa abundances at the phylum level

The total sequence reads used in this study were classified into eight phyla and one unassigned (Table ^S2). In any tested sample, the core microbiota was represented by five taxa: three with a relative abundance higher than 10% (Firmicutes, Bacteroidetes, and Actinobacteria) and two lower than 3% (Proteobacteria and Verrucomicrobia). Anyhow, just Firmicutes, Bacteroidetes, and Proteobacteria underwent significant changes in comparison to the baseline (p < 0.05).

As a general parameter for microbiota eubiosis we chose the famous ratio Firmicutes/Bacteroidetes (F/B), and we considered the differences from the baseline to the end point. Within this ratio a value over two is usually referred to microbiota dysbiosis^11,14. The fecal samples at the baseline had a F/B of 1.66 and this eubiosis condition was maintained by HBPA (1.55), by HB (1.62) and FOS (0.73), although significantly just for the latter (p > 0.05). These results indicate that during the time of fermentation, HB and HBPA did not perturb the colon core microbiota of healthy donors but was able to provide a substrate that meet the energetic expenditure of the microbiota, keeping an eubiosis condition.

Changes in taxa abundances at the species level

A dataset of significant OTUs changes relative to the family level is reported in Table ^S3. Anyhow, we focused the discussion on results obtained at the specie taxonomic level, where 113 OTUs were constructed and assigned to microbial taxa (cutoffs 0.001%). Of these, 113 were identified at the baseline, while 106, 102, 96, and 100 were identified at the endpoint of fermentations with HPBA, HB, FOS, and the blank control, respectively. Then a dataset of 41 microbial OTUs was selected and tested for ANOVA group comparison in respect to the baseline (p > 0.05). Among these, 31 variables were significant and their Log₂ fold changes in respect to the baseline were compared by post-hoc test (Table ​(Table1).1). The 41 OTUs selected were those that recorded shifts after fermentation and that from literature are susceptible to the effect of prebiotic or fiber substrates. We have included even three OTUs of Archea relative to QC of the experiments (previously discussed).

The first group of OTUs included beneficial or commensal bacteria that usually respond to prebiotics. In this group, three Bifidobacterium were picked showing increases on the substrates and reduction on the blank control. HB and HBPA fostered Bif. bifidum, but just the latter did it significantly, making this taxon grew up to the 3.30% of relative abundance (p < 0.05). Besides, FOS fostered even Bif. adolescentis (p < 0.05). Among Bacteroides, five OTUs were chosen and except B. fragilis were all significant (p < 0.05). B. thetaiotaomicron and B. uniformis were the most abundant in HBPA, HB, and FOS bioreactors at the endpoint, the first recorded top shift for HBPA reaching 7.88% of total abundance. Parabacteroides distasonis was found rich and significantly increased after fermentation with HBPA, HB, and FOS (p < 0.05), but not in the blank control (p > 0.05). From the class of Lactobacillales, significant shifts (p < 0.05) were seen for two Enterococcus and two Lactobacillus OTUs, that augmented with the substrates and decreased in the blank control. Interestingly, while En. durans was largely fostered by both HBPA, HB, and FOS, En. faecalis just by HBPA and HB and reduced by FOS. Lactobacillus mucosae and Lb. plantarum were represented in very low amounts at the baseline and were intensively fostered by both substrates. For example, the first reached the top quantity of 0.06% with HBPA, while the second that of 0.24% with FOS. Faecalibacterium prausnitzii and Akkermansia muciniphila were more abundant after substrates fermentation and less in the blank control, although not all significantly (p < 0.05).

From our results, even at the depth of the species level, it was possible to highlight the prebiotic potential of HB and on larger extend of HBPA that, similarly to FOS, fostered several taxa of beneficial bacteria. In particular, the surges in these taxa were relative to: (i) three species of health associated and SCFAs-producer Bifidobacterium⁴; (ii) MCFAs- and sphingolipids-producer B. thetaiotaomicron, succinate-producer P. distasonis^15,16, and (iii) competitive excluders Lactobacillales, as Lb. plantarum and E. durans^17,18. Moreover, HBPA showed to be able to foster beneficial SCFAs-producer F. prausnitzii⁵ and fiber-degrading B. caccae¹⁹. In comparison to HB, the better performance obtained by HBPA are due the action of alcalase that gives a product with a higher rate proteins/peptides with MW around or lower than 15 kDa, and an increased percentage of soluble proteins (10%).² It is reported, for example that Lactobacillaceae likes peptides more than proteins, and prefers to ferment low molecular weight peptides than proteins²⁰.

A second list of bacterial taxa, that changed in abundance at the endpoint, was that of opportunistic species. Bilophila wadsworthia, Desulfovibrio (pathobiontic, highly proteolytic and sulphate producers) and Escherichia albertii, (close relative to pathogenic species) were reduced by HBPA, HB, and FOS while increased in the blank control (p < 0.05). In particular HPBA performed better than HB and FOS in the containment of Bil. wadsworthia and Desulfovibrio. In details, considering these three taxa, HBPA was stronger than HB but with no significant differences in the reductions (p > 0.05), except for that relative to Bil. wadsworthia (p < 0.05). The ability to counteract opportunistic and enteropathogenic microbes is an essential feature of a prebiotic compounds. Particularly, these species are involved in dysbiosis of the microbiota and pathogenesis^21,22 and were reduced in a similar study on prebiotics and vegetal fibers⁵. Thus, HBPA had superior performances than HB in limiting the development of opportunistic microbes. This evidence may contribute to explain the beneficial effects of hydrolyzed proteins. Indeed, the modulation of gut microbiota usually results from unabsorbed sugars, resistant starch, and fibers, but indigestible proteins and bioactive peptides have been proven beneficial too, such as hydrolysed proteins from soy²³. Additionally, this feature could in part attributable to the increased release of peptides with higher antioxidant capacity when HB underwent the alcalase treatment, as well as an higher content of bioactive peptides. In fact, Setti et al.² found that HBPA in comparison to HB has an in vitro antioxidant activity up to 10 times stronger. Additionally, Samaei et al.¹⁰ identified on HBPA 47 bioactive peptides, that for the most are short sequences of a few amino acids, potentially resistant to the gastrointestinal conditions.

The third list regards to those taxa that usually respond to vegetal fibers. The performances of HBPA that deserve merit of notion are the significant reduction of Ruminococcus gnavus and R. torques, as well as that of Colinsella aerofaciens and Eggerthella lenta, similarly to FOS (p < 0.05). The two Ruminococcus are culprits of dysbiosis associated to intestinal syndromes and are effective responders to fiber diet regime²⁴. In contrast to FOS, HPBA and minorly HB were able to increase the quantity of Oscillospira and Sutterella. For HBPA the surges were significant reaching 11.90%, and 1.94% of relative abundance, respectively (p < 0.05). Despite the role in gut microbiota of Oscillospira remains enigmatic, as a member of Ruminococcaceae should be implicated in fiber degradation²⁵ and could explain the reduced abundance observed in HPBA for same family member F. prausnitzii. Sutterella, in the past indicated as an opportunistic species has been recently reconsidered for its ability to degrade plant-based pectins and similar compounds in in vitro systems likely the MICODE²⁶.

qPCR prebiotic index

qPCR Prebiotic Index (qPI) was recently introduced⁹ as a revised method based on an updated Prebiotic Index equation, originally elaborated on 24 h controlled batch culture condition with 1% w/v addition of prebiotic by Palframan et al.²⁷. Considering the results (Fig. 1), we found out that the fermented substrate with the best prebiotic activity was FOS after 18 h, and the runner-up was HBPA after 24 h. In comparison to FOS 18 h, HBPA 18 h scored 1.44-fold lower values. The blank control scored for any time points lower values than any HBPA, HB, and FOS cases (all significant, but one) and reached the lowest value of the dataset at the endpoint (26.24-fold lower than FOS 18 h).

Figure 1

qPCR Prebiotic Index (qPI) of colonic fermentations on the substrates HBPA, HB, FOS, and on a blank control, at different time points. ^abcdDifferent letters indicate statistical significance by Tukey’s honestly significant differences (HSD) test (p < 0.05). Marker = mean; box = mean ± S.D.; whiskers = min and max; dots = outliers; asterisks = extremes. HB Hempseed bran; HBPA HB protein extract hydrolyzed by alcalase; FOS fructooligosaccharides; Blank Blank control.

So far, the qPI of HBPA leans to reach high level later than the FOS. Anyhow, even at the earlier time points qPI of HBPA was higher than the blank control. Thus, the comparable prebiotic index of HBPA could be mostly due to its high portion of soluble fibers. Similarly to FOS it is known that soluble fibers are excellent substrates for production of SCFAs in the large intestine¹⁵.

Changes in main microbial metabolites related to prebiotic potential

To analyze the main changes in volatile microbial metabolites related to prebiotic potential, we have considered the shift in loads from the baseline to the endpoint (24 h) of fermentations of 10 selected VOCs (ANOVA p < 0.05) with renowned bioactivity in humans (short and medium chain organic acids and aromatic compounds) as follows: (a) each single compound was normalized (mean centering method) within its dataset, which included cases from HB, HBPA, FOS, and the blank control at different time points; (b) the baseline dataset (Table ^S4) was then subtracted to the endpoint dataset; (c) post-hoc analysis was done to compare the sample productions of a single molecule (Tukey’s HSD, p < 0.05). The first set of compounds is relative to low organic acids, such as Acetic, Propanoic, Butanoic, Pentanoic, and Hexanoic acids that are beneficial compounds essential for the host, the mucosa, and the colon microbiota itself (Fig. 2). The second set is relative to compounds related to proteolytic fermentation and/or detrimental for the host, such as Indole, Phenol, p-Cresol, Benzaldehyde, and Phenol, 2,4-bis(1,1-dimethylethyl)- (2,4-DTBP) (Fig. 3).

From the results shown in Fig. 2, organic acid concentration was increased with HB, HBPA, and FOS, while no changes nor production of any of them was recorded in the blank control. Starting from small amounts detected at the baseline (< 0.010 mg/kg for Acetic, 0.012 mg/kg for Propanoic, 0.101 mg/kg for Butanoic), the capacity to produce Acetic, Propanoic, and Butanoic acids was generally (considering the means of every time points) stronger for FOS than for HB or HBPA (p < 0.05). In particular, FOS fermentation accounted for 2.25-, 3.37-, and 4.87-folds more than HBPA, respectively for these three compounds. A reduction in Acetic, Propanoic, and Butanoic acids abundances is linked to dysbiosis of the colon microbiota and a reduced intestinal cell homeostasis⁴. The prebiotic activity of HBPA is linked to its capacity to foster Lactobacillus spp., Bifidobacterium ssp., and Enterococcus spp. that metabolize the fibers and produce low organic acids. On the opposite, starting from little amounts at the baseline (< 0.010 mg/kg either for Pentanoic and Hexanoic acids), the surge of Pentanoic and Hexanoic acids was stronger for HBPA than FOS (p < 0.05). In details, HBPA fermentation accounted for 1.27- and 2.08-folds more, respectively of these two compounds. Besides, HBPA fermentation released its top abundances at the endpoint, while FOS was able to release those compounds earlier, reaching the top at the intermediate time point (18 h), except for Pentanoic acid. Pentanoic and Hexanoic acids are medium chain fatty acids (MCFAs) are protective on glucose homeostasis and against insulin resistance and are important metabolic biomarkers of dysbiosis and intestinal bowel disease (IBD)^14,28,29. The increased abundance in MCFAs observed in this study could be due to the ability of HBPA to foster Bifidobacteriaceae and commensals Clostridium group IV, or Bacteroides spp. Actually, MCFA production by these three bacterial groups happened during fiber fermentation³⁰.

The ability of HBPA to liberate once fermented more SCFAs than HB could be due to the higher availability of lower MW peptides/proteins and to the higher fermentation preference of these substrates by Lactobacillales and Bifidobacteriaceae. Similarly, the more of these species were fostered and the more was the ability to elongate MCFAs from lactate production via reverse β-oxidation³¹.

The second set comprised VOCs that had a different trend for the substrates than the blank control (Fig. 3). Indole abundances increased with HB, HBPA, and FOS but decreased in BC. Oppositely, Phenol, p-Cresol, Benzaldehyde, and 2,4-(DTBP) were reduced with HBPA and partially with HB, while increased with BC. HBPA was able to produce 2.08-fold more Indole than FOS, and to reduce 1.54- and 1.48-fold more Phenol and 2,4-(DTBP) than FOS, respectively (p < 0.05). Indole is a tryptophan catabolite, deriving from degradation of the proteinaceous portion of the food⁵ by commensal Escherichia coli. Indole is also suggested to have beneficial effects, such as the attenuation of inflammation indicators on HCT-8 cells at the concentration of 1mM³². Otherwise, its accumulation as bacterial products (Clostridium spp. and Escherichia spp.) could result toxic for the host, because if it is not microbially degraded in beneficial derivates (e.g. Indole propionic acid) is metabolized into Indoxyl sulphate in the liver that, as the prototype of protein–bound uremic toxins³³, provokes chronic kidney disease and vascular disease^5,34. Despite, the dose of indole to generate such detrimental effect is undefined, a study finds that cattle injected with 0.2 g/kg of body weight after 72 h had diarrhoea, haemolysis, haemoglobinuria, and microscopic lesions of haemoglobinuric nephrosis³⁵.

Similarly, Phenol and p-Cresol are derived from proteolytic fermentation and have been shown to damage epithelial barrier function in vitro and can be potentially carcinogenic⁵.

From the results shown in Fig. 3, FOS, HB, and HBPA fermentations indicated increases in Indole content in respect to the baseline, although significant just for HBPA (p < 0.05) and a reduction in metabolites (phenols) related to animal fat and protein degradation. This scenario was opposite for fermentation with the BC.

These compounds were more abundant at the baseline, as derived by fecal samples of omnivores. Their reductions are in line with the results obtained from the microbiota, indicating an increase in those taxa specialized in plant-based fibers fermentation. When comparing the reduction of these detrimental VOCs of HBPA to that of HB, the better action of HBPA could be attributed to higher bioactivity of alcalase treated HB. Indeed, a previous study demonstrated that HBPA low MW (Molecular Weight) proteins or peptides have an antioxidant activity higher than the high MW proteins or peptides of HB².

Volatilome analysis through SPME GC/MS

The SPME GC-MS analyses were conducted on 32 duplicated cases (n = 64). With NIST 11 MSMS library and the NIST MS Search program 2.0 (NIST, Gaithersburg, MD, USA) 125 molecules with more than 80% of similarity were identified, of which 77 were relatively quantified at the baseline and 113 during and after colonic fermentations. The whole volatilome was produced from a dataset of 93 significant VOCs (ANOVA at p < 0.05) and presented as a quantification heatmap (Figure ^S2). Afterwards, this dataset was separated and super-normalized by chemical classes of VOCs, i.e., organic acids, main detrimental aromatic VOCs, aldehydes, ketones, alcohols, and others (alkenes, alkanes, amines, sulphurates). Organic acids VOCs and detrimental aromatic VOCs were just previously discussed, as main microbial metabolites related to prebiotic activity, while, from each dataset of the other classes, multivariate analyses, such as untargeted Principal Component Analysis (PCA) and targeted MANOVA (p < 0.01) was achieved to address the specific contributes to VOCs production by the independent variables^3,8,9. Super-normalization of the dataset was essential to unveil the effect of those compounds that are less volatile than others and could be underrepresented, as well as to avoid comparing one chemical class to another^3,8,9.

A PCA of 27 statistically significant alcohols distributed cases on the plot, separating fermentation with HBPA, HB, FOS, and BC from each other and from the baseline (Fig. 4A). From our results, the main descriptors of fermentation with HBPA were mainly complex terpenoid alcohols (p < 0.01), such as 4-Terpineol, Beta-Linalool, Cuminol, Eucalyptol, Borneol, and 1,8-Menthadien-4-ol, mainly produced at the intermediate and late time points (p < 0.01) while those for FOS were 1-Dodecanol, Propanol, 4-methyl, 3-Buten-1-ol, 3-methyl, and Ethyl alcohol mainly produced at the intermediate time point (p < 0.01) (Tables ^S5, ^S6). The main descriptor of alcohol production from BC samples remained Isopropyl alcohol (p > 0.01). The colon microbiota produces different alcohols during fermentation of dietary polysaccharides. Terpineol, Beta-Linalool, Cuminol, Eucalyptol, and Borneol, that are major terpenoids found in hemp seed with anti-oxidant and anti-inflammatory activity, were increased after lactobacilli fermentation of HB³.

Considering aldehydes, we found 16 significant VOCs that by PCA discriminated the HBPA and HB from the controls and from the baseline (Fig. 4B). The main descriptor of fermentation with FOS was 2-Hexenal (p < 0.01), while that for HBPA were Heptanal and 2-Octenal, (E), principally produced at the early time point and at the endpoint, respectively (p < 0.01) (Tables ^S5, ^S6). Lastly, the main descriptor of BC was Benzeneacetaldheyde, that was present at the baseline, but absent after fermentation with the substrates. Aldehydes are a result of microbial fermentation and lipid oxidation. Certain aldehydes are health-promoters, like 2-Octenal, (E) that was reported to limit the growth of several intestinal pathogens at a very low concentration³⁶, while most are detrimental, being cytotoxic at a low threshold, such as Benzeneacetaldheyde³⁶.

Considering Ketones, 16 significant VOCs were able to discriminate by PCA the substrates from each other and from the baseline (Fig. 4C). Descriptors of fermentation with HBPA were p-Menthone (77.00%) and Acetophenone (81.00%), majorly produced at the endpoint (77.04% and 51.83%, respectively) (p < 0.01) (Tables ^S5, ^S6). The main descriptor of fermentation with FOS was 2,3-Butadione (68.93%), that of HB and BC were 2-Heptanone and Acetone, respectively but not significantly (Table ^S5). During colonic fermentation, many ketones are produced; considering their bioactive attributes, some are desirable, such as Acetophenone that acts as antimicrobial to different Gram-negative bacteria, and its N-substitute derivates have been proposed as a therapeutic approach in diabetes³⁷. In our experimental condition, Acetophenone is probably derived from the bacterial deconjugation of polyphenols, as Lactobacillales³⁸, which was increased by hydrolytic process in HBPA.

A PCA of 22 statistically significant VOCs (alkenes, alkanes, amines, and sulphurates) distributed cases on the plot, separating the substrates from each other and from the baseline (Fig. 4D). The main descriptor of fermentation with FOS was Ethyl Acetate (p < 0.01), while those for HBPA were Caryophillene and D-Limonene, that for HB was Eicosane, while the baseline was described by Aniline. These VOCs were discriminated significantly just for the category of substrates (p < 0.01) (Tables ^S5), but no significant differences were detected for the category of time (p > 0.05), except for Aniline that was reduced significantly on a time basis (p < 0.01) (Table ^S6). Caryophillene and D-Limonene are potent health-related terpenes³ and the features observed indicate that the descriptors of HBPA were not subject to fermentation and thus their bioactivity was preserved from the food matrix. Aniline is instead a carcinogen derived from benzenoid pollutants³⁶, and its reduction by fermentation with HBPA is a positive feature.

Materials

Chemicals, solvents, and enzymes for batch culture fermentation were of the highest analytical grade and were purchased from Sigma-Aldrich (St. Louis, MO, USA), Merck (Darmstadt, Germany), and Carlo Erba Reagents (CEDEX, Val de Reuil, FR), unless otherwise stated. Reagents for molecular biology and kits for DNA extraction or purifications were purchased from Thermo Fisher Scientific (USA).

Experimental sample and controls

Experimental HB was previously prepared and characterized^2,3. HB, the byproduct remaining after mechanical pressing of hemp seeds and subsequent grinding and sieving, was supplied by a local company (Hemp Positive World, Cesena, Italy). The original hemp variety was Futura 75. This hemp variety is registered for fiber production with a THC content < 0.2% and does not require legal permission to collect. To prepare HB samples for fermentation, 50 g of bran were resuspended in 300 mL of distilled water, sterilized (121 °C and 100 kPa for 20 min) (Vapor Matic 770, ASAL Srl, Milan, Italy) in 500 mL Corning–Pyrex bottles (Corning. NY, USA), aseptically poured on sterilized metal vessels, and stored at − 80 °C. HBPA was produced by enzymatic digestion of chemically extracted HB protein isolate, as described by Setti et al.², extracting protein from HB following published procedure⁴⁴. Briefly, the protein isolate was dissolved in deionized water (1:8, w/v) and hydrolyzed for 2 h with Alcalase (2%, v/v) at 50 °C and pH 8.0. The enzyme was then heat inactivated at 85 °C for 15 min. After cooling down to room temperature, the solution was centrifuged at 14,000 × g for 10 min, and the supernatant was collected and stored at − 80 °C.

Both HB and HPBA solutions were freeze dried with a Savant freeze-dryer Lyolab 3000 apparatus (Thermo Fisher Scientific, USA), and the powder was used to test the prebiotic potential^2,3.

Fecal batch-culture fermentation and samples collection

Colonic fermentations were conducted for 24 h in independent vessels on 1% (w/v) of HB, on 1% (w/v) of HBPA, on 1% (w/v) of fructo-oligosaccharides (FOS) from chicory (positive control), and on a blank control (BC) (negative control), using an in vitro gut model, MICODE (Multi-Unit in vitro Colon Model), obtained by the assembly of Minibio Reactors (Applikon Biotechnology BV, Delft, NL) and controlled by Lucullus PIMS software (Applikon Biotechnology BV, NL)^8,9. The preparation of the experiments were made according to published procedures^5,8,9,41,42. In details, bioreactors were autoclaved at 121 °C and 100 kPa for 15 min and once cooled aseptically filled with 90 mL of anaerobic pre-sterilized basal nutrient medium. Basal medium (BM) contained (per liter): 2 g peptone, 2 g yeast extract, 0.1 g NaCl, 0.04 g K₂HPO₄, 0.04 g KH₂PO₄, 0.01 g MgSO₄·7H₂O, 0.01 g CaCl₂·6H₂O, 2 g NaHCO₃, 2 mL Tween 80, 0.05 g Hemin dissolved in 1 mL of 4 M-NaOH, 10 mL vitamin K, 0.5 g L-cysteine HCl, and 0.5 g bile salts (sodium glycocholate and sodium taurocholate)^41,42. The medium was adjusted to pH 7.0 before autoclaving and 2 mL of 0.025% (w/v) resazurin solution were added afterwards once the media was cooled^41,42. Fermentation vessels were filled aseptically with 90 mL of BM and the bioreactor headplates were mounted, including previously sterilized and calibrated sensors, i.e. pH and DO₂ (Dissolved Oxygen) sensors. Anaerobic condition (0.0–0.1% w/v of DO₂) in each bioreactor was obtained in about 30 min flushing with filtered O₂-free N₂ through the mounted-in sparger of Minibio reactors (Applikon Biotechnology BV, NL), and was constantly kept over the experiment. Temperature was set at 37 °C and stirring at 300 rpm, while pH was adjusted to 6.75 and kept throughout the experiment with the automatic addition of filtered NaOH or HCI (0.5 M) to mimic the conditions located in the distal region of the human large intestine^8,9. Once the exact environmental settings were reached, the four vessels were aseptically injected with 10 mL of fecal slurry (10% w/v of human feces to a final concentration of 1%, w/v) and then three of them independently with 1 g of HB, HBPA or FOS (to a final concentration of 1%, w/v), while the fourth vessel was set as blank control (BC, basal medium and 1% fecal slurry only). Batch cultures were run under these controlled conditions for a period of 26.55 h during which samples were collected at 4 time points (Baseline, 6, 18, and 24 h). The baseline (BL) was defined on the first pH changes detected by Lucullus (1 read/10 s) via the pH Sensors of MICODE^8,9. For this work, the BL was set after 2.55 ± 0.11 h. Sampling was performed with a dedicated double syringe filtered system (Applikon Biotechnology BV, NL) connected to a float drawing from the bottom of the vessels without perturbing or interacting with the bioreactor’s ecosystem. To guarantee a close control, monitoring and recording of fermentation parameters the software Lucullus 3.1 (PIMS, Applikon Biotechnology BV, NL) was used. This also allowed to keep the stability of all settings during the experiment^8,9. Fermentations were conducted in duplicate independent experiments, using for each a new pool of feces from the same three healthy donors^8,9.

The freeze-dried samples were directly fermented in the colon with no gastric phase digestion, as the nature of prebiotic is to reach the colon to feed the microbiota without being affected by host’s enzymes⁴⁵.

Sequence data analysis

Reads were de-multiplexed based on Illumina indexing system. Sequences were analyzed using QIIME 1.5.0⁴⁷. After filtering based on read quality and length (minimum quality = 25 and minimum length = 200), Operational Taxonomic Units (OTUs) defined by a 97% of similarity were picked using the Uclust v1.2.22 q method⁴⁸ and the representative sequences were submitted to the RDP classifier²¹ to obtain the taxonomy assignment and the relative abundance of each OTU using the Greengenes 16S rRNA gene database⁴⁹. Alpha- and beta-diversity analyses were performed using QIIME 1.5.0^8,46.

Enumeration of bacterial groups for the qPI (Quantitative Prebiotic Index)

Changes in Eubacteria kingdom, Lactobacillales order, Bifidobacteriaceae, Enterobacteriaceae families, and Clostridium group I were also assessed by qPCR and SYBR Green I chemistry^9,13,50, targeting small fragments of monocopies, or multicopy genes by degenerated or specific primer pairs, previously amplified by high-fidelity DNA polymerase (Invitrogen Platinum SuperFi II DNA Polymerase, Thermo Fischer Scientific, USA). Extraction of bacterial DNA was obtained with Pure Link Microbiome DNA Purification Kit (Invitrogen, USA). Genetic standards were prepared from relative PCR amplicons of the target bacterial species, using GeneJet Genomic DNA purification kit (Thermo Fisher Scientific, USA), as described previously^8,51. For each of the targets, general qPCR reactions were set as follows: a holding stage at 98 °C for 6 min, and a cycling stage made of 95 °C for 20 s and 60 °C for 60 s, repeated for 45 times, followed by melting curve analysis^3,9. Quantifications were made by a RotorGene 6000 (Qiagen, Hilden, Germany) with a five-point standard of the given amplicon, separately. Reactions were prepared with 1 ng of DNA, 2 × Power-Up SYBR Green (Thermo Fisher Scientific, USA) and 250 nM of each primer (Eurofins Genomics, Ebersberg, Germany)^3,9. Details of primer pairs for PCR and qPCR are supplied as Table ^S1. All results were expressed as mean values obtained from triplicates from two independent experiments.

The Prebiotic Index was revised from the original equation elaborated by Palframan et al.²⁷ introducing substitution on bacterial taxa, the molecular approach based on quicker qPCR, data normalization, sextuplicate values, and significant differences⁹. Analogously to the original method, we employed an equation based on quantification values expressed as Log₁₀ cell/mL, and similar conditions applied in fermentation (24 h controlled batch with 1% w/v of prebiotic fiber). So far, the new equation for the Prebiotic Index is based on qPCR data (qPCR Prebiotic Index—qPI) as follows:

.

Volatilome analysis

Volatile organic compound (VOCs) evaluation was carried out on an Agilent 7890A Gas Chromatograph (Agilent Technologies, Santa Clara, CA, USA) coupled to an Agilent Technologies 5975 mass spectrometer operating in the electron impact mode (ionization voltage of 70 eV) equipped with a Chrompack CP-Wax 52 CB capillary column (50 m length, 0.32 mm ID) (Chrompack, Middelburg, The Netherlands)^3,8,9. The solid phase micro-extraction (SPME) GC-MS protocol and the identification of volatile compounds were done according to previous reports, with minor modifications^8,52–54. Briefly, 3 mL of vessel content or fecal slurry were placed into 10 mL glass vials and added to 10 μL of the internal standard (4-methyl-2-pentanol) to a final concentration of 4 mg/L. Samples were then equilibrated for 10 min at 45 °C. SPME fiber, coated with carboxen-polydimethylsiloxane (85 μm), was exposed to each sample for 40 min. Preconditioning, absorption, and desorption phases of SPME–GC analysis, and all data-processing procedures were carried out according to previous publications^8,53,54. Briefly, before each head space sampling, the fiber was exposed to the GC inlet for 10 min for thermal desorption at 250 °C in a blank sample. The samples were then equilibrated for 10 min at 40 °C. The SPME fiber was exposed to each sample for 40 min, and finally the fiber was inserted into the injection port of the GC for a 10 min sample desorption. The temperature program was: 50 °C for 1 min, then programmed at 1.5 °C/min to 65 °C, and finally at 3.5 °C/min to 220 °C, which was maintained for 25 min. Injector, interface, and ion source temperatures were 250, 250, and 230 °C, respectively. Injections were carried out in split-less mode and helium (3 mL/min) was used as a carrier gas. Identification of molecules was carried out by searching mass spectra in the available databases (NIST 11 MSMS library and the NIST MS Search program 2.0 (NIST, Gaithersburg, MD, USA). Each VOC was relatively quantified in percentage (LOD = 0.001 mg/kg)^3,8,9,55. Besides, in samples prior to in vitro colonic fermentation (baseline) (Table ^S3) the main microbial metabolites related to prebiotic activity were also absolutely quantified in mg/Kg (LOQ = 0.03 mg/kg and LOD = 0.01 mg/kg)^3,8,9,53,54. For these latter compounds, samples at the endpoint (24 h) were compared to the baseline and values were expressed as shifts. All results were expressed as normalized mean values obtained from duplicates in two independent experiments^8,9.

This research was partially funded by Alma Idea from the University of Bologna.