Plant based formula. Ingredients to compose the MD-based food formula were initially selected based on the radical scavenging activity. Both MWSE and WSE from moringa powder exhibited extremely high antioxidant activity: 204.4 ± 3.1 and 97.2 ± 1.8 mmol BHT 100 g− 1 of dry weight (DW), respectively. Among the other fresh ingredients, the highest (p < 0.05) activities were found for MWSE and WSE of pomegranate (10.1 ± 0.1 and 2.5 ± 0.1 mmol BHT 100 g− 1 fresh weight, [FW], respectively) and walnuts (8.6 ± 0.0 and 4.0 ± 0.1 mmol BHT 100 g− 1 FW, respectively). Based on this preliminary screening, moringa, pomegranate and walnuts were selected as basic ingredients for five formulas having a diverse fourth ingredient, namely apple, cabbage, broccoli, avocado or dates. All formulas had a ratio among ingredients of 3:3:2:2 (v/w/v/w). The antioxidant activities of all these formulas (30.9 ± 0.4 to 28.6 ± 0.6 BHT 100 g− 1 FW) did not significantly differ, but they were higher (p < 0.05) than those observed in single ingredients.
The protective effect towards oxidative stress was further assessed using Caco-2 cells. Compared to WSF, MWSF had higher cytotoxicity. Concentrations of 10 µg mL− 1 slightly decreased the Caco-2 cell viability, but concentrations up to 500 µg mL− 1 markedly affected the Caco-2 cell proliferation. This was mainly found for those MWSF from formulas having broccoli or avocado. Under culture conditions, Caco-2 cells developed morphological and functional features of enterocytes, including intercellular tight junctions. Treatment of Caco-2 cells with TNF-α markedly decreased TEER, but when Caco-2 cells were stimulated (basolateral compartment) by TNF-α and subsequently treated (apical compartment) with WSF and MWSF, the negative effect attenuated. This was mainly evident in the formulas comprising broccoli or avocado. The formula with broccoli showed the highest (p < 0.05) reduction of intracellular ROS. Compared to the negative control (DMEM), Caco-2 cells treated with H2O2 (50 µM) increased their catalase activity. The incubation with MWSF and WSF significantly counteracted the negative effects of H2O2. The presence of broccoli promoted the lowest (p < 0.05) catalase activity. GSH, a marker of oxidative stress, significantly increased in formulas with cabbage or broccoli. ELISA analysis revealed that Caco-2 cells subjected to an inflammatory stimulus, and treated with MWSF and WSF from the formula containing broccoli had the lowest (p < 0.05) levels of cytokines (IL-8).
Based on these results, the formula comprising moringa, pomegranate, walnuts and broccoli at a ratio 3:3:2:2 (v/w/v/w) was selected for manufacturing the MD-based food. In parallel, a placebo consisting of maltodextrin, caramel and curcumin colorants, chocolate aroma, guar gum and water was manufactured.
MD-based food. The chemical and nutritional composition of the MD-based food are reported in Table 1. Moisture and ash were 58.2 ± 1.8% and 1.16 ± 0.02%, respectively. The energetic value was 256 ± 26 Kcal and 1059 ± 106 KJ/100 g. The total content of carbohydrates was 4.5 ± 1.4%, with dietary fibers and vegetable proteins amounting to 6.4 ± 2.2 and 8.4 ± 1.0%, respectively. Fat was 21.30 ± 0.64%, with 2.25 ± 0.23% of total saturated fatty acids.
Table 1
Chemical and nutritional features of the MD-based food.
Parameter
|
Value
|
Humidity (%)
|
58.2 ± 1.8
|
Ashes (%)
|
1.163 ± 0.023
|
Total fat (%)
|
21.30 ± 0.64
|
Total saturated fatty acids (%)
|
2.25 ± 0.23
|
Energy value (Kcal/100g)
|
256 ± 26
|
Energy value (KJ/100g)
|
1059 ± 106
|
Carbohydrates (%)
|
4.5 ± 1.4
|
Galactose (%)
|
< 0.10
|
Glucose (%)
|
2.08 ± 0.31
|
Fructose (%)
|
2.24 ± 0.34
|
Sucrose (%)
|
< 0.10
|
Lactose (%)
|
< 0.10
|
Maltose (%)
|
< 0.10
|
Total sugars (%)
|
4.32
|
Total dietary fiber (%)
|
6.4 ± 2.2
|
Proteins (%)
|
8.4 ± 1.0
|
Salt (Sodium x 2.5) (%)
|
0.015
|
Fatty acids (g/100g of MD-based food fat)
|
Phenolics (µg/g DW of MD-based food)
|
Lauric acid (C12: 0)
|
0.02 ± 0.01
|
Gallic acid
|
23.91 ± 8.94
|
Myristic acid (C14: 0)
|
0.04 ± 0.01
|
Procyanidin B3
|
5.7 ± 0.49
|
Palmitic acid (C16: 0)
|
7.36 ± 0.15
|
Procyanidin B1
|
12.54 ± 0.38
|
Palmitoleic acid (C16: 1)
|
0.12 ± 0.01
|
Catechin
|
4.35 ± 1.68
|
Heptadecanoic acid (C17: 0)
|
0.07 ± 0.02
|
Chlorogenic acid
|
10.23 ± 0.40
|
Heptadecenoic acid (C17: 1)
|
0.02 ± 0.01
|
Epicatechin
|
7.99 ± 0.36
|
Stearic acid (C18: 0)
|
2.87 ± 0.06
|
Rutin
|
12.7 ± 1.44
|
Oleic acid (C18: 1)
|
14.08 ± 0.14
|
Ellagic acid
|
90.64 ± 43.16
|
Trans-Oleic acid (C18: 1)
|
0.03 ± 0.01
|
Epicatechin 3 gallate
|
0.93 ± 0.09
|
Linoleic acid (C18: 2)
|
61.5 ± 0.62
|
Caffeic acid
|
12.47 ± 0.12
|
Trans-Linoleic acid (C18: 2)
|
0.06 ± 0.02
|
Sinapic acid
|
3.5 ± 1.07
|
Linolenic acid (C18: 3)
|
13.34 ± 0.13
|
Isoquercetin
|
27.93 ± 6.87
|
Trans-Linolenic acid (C18: 3)
|
0.09 ± 0.03
|
Hyperoside
|
49.89 ± 12.86
|
Arachic acid (C20: 0)
|
0.10 ± 0.01
|
Phloridzin
|
5.34 ± 0.68
|
Eicosenoic acid (C20: 1)
|
0.23 ± 0.03
|
Quercetin
|
10.88 ± 0.5
|
Arachidonic acid (C20: 4)
|
0.01 ± 0.00
|
Kampferol
|
10.23 ± 0.17
|
Beenic acid (C22: 0)
|
0.03 ± 0.01
|
Luteolin
|
5.1 ± 0.07
|
Erucic acid (C22: 1)
|
0.01 ± 0.00
|
Naringenin
|
3.83 ± 0.21
|
Lignoceric acid (C24: 0)
|
0.02 ± 0.01
|
Glucosinolates (µM/100g DW of MD-based food)
|
Total monounsaturated fatty acids
|
3.08 ± 0.31
|
Glucobrassicapin
|
134.47 ± 1.44
|
Total polyunsaturated fatty acids
|
15.97 ± 1.60
|
Glucotropaelin
|
765.16 ± 109.27
|
|
|
Glucoerucin
|
78.37 ± 4.23
|
|
|
Glucobrassicin
|
48.23 ± 2.80
|
Monounsaturated and polyunsaturated fatty acids were 3.08 ± 0.31 and 15.97 ± 1.60%, respectively (Table 1). Oleic acid (C18: 1) was the predominant monounsaturated fatty acid (14.08 ± 0.14 g 100 g− 1 fat), with linoleic (C18: 2) (61.5 ± 0.62 g 100 g− 1 fat) and linolenic acids (C18: 3) (13.34 ± 0.13 g 100 g− 1 fat) prevailing within polysaturated fatty acids. Ellagic acid (90.64 ± 43.16 µg g− 1 DW) was the major free phenolic compound. Glucosinolates were glucobrassicapin (134.47 ± 1.44 µg g− 1 DW), glucotropaelin (62.94 ± 0.06 µg g− 1 DW), glucoerucin (78.37 ± 4.23 µg g− 1 DW) and glucobrassicin (48.23 ± 2.80 µg g− 1 DW).
Fecal donor. According to nutritional questionnaires, 26 out of 30 recruited individuals showed low MDA (average of the MED 4.8 ± 0.49). They represented the study cohort. Although with some interindividual variability, fecal samples showed the overall dominance of specific taxa at the lowest assigned taxonomic level. The list included Bacteroides (16% of average frequency of relative abundance), Faecalibacterium (11%), Lachnospiraceae (10%), Ruminococcaceae (7%) and Ruminococcus (7%). Blautia, Clostridia, Bifidobacterium, Prevotella, Roseburia and Akkermansia were present at < 5% (Supplementary Fig. 1a). The concentration of SCFAs, BCFA and VOCs agreed with literature data from individuals showing low MDA (Supplementary Table 4) [41][4]. The ward linkage hierarchical clustering analysis showed the stratification of the fecal samples according to the microbiota composition, SCFAs, BCFA and VOCs (Supplementary Fig. 1b). Three main clusters were formed, with the biggest comprising 11 out of 26 volunteers. The relative frequency of Clostridiales, Prevotella, Ruminococcaceae and Blautia, and the concentration of hexanoic acid butyl ester, pentanol heptanol, undecanol p-cresol and octadecane mainly determined the aggregation of this cluster. The fecal donor (TO07) was selected within this cluster. The lowest MDA associated with low levels of SCFAs [41], and decreased abundances of Lachnospira and Ruminococcus [4]. The fecal donor showed the lowest MDA score, and his/her microbiota and metabolome composition are shown in Supplementary Table 5.
MD-based food partially reshapes the microbiota. Beta diversity from unweighted unifrac distance matrix plotted into the PCoA graph did not stratify samples neither using 26S nor 16S rRNA genes on placebo or MD-based food intake (Supplementary Fig. 2a and b). On the contrary, colon tracts stratified allowing the separation of two clouds from AC, distinguished from TC and DC (Supplementary Fig. 2c). Based on 26S and 16S rRNA genes relative abundance matrices, PLS-DA allowed for discriminating sample clusters relying on the combination of canonical (exploratory) and regression (explanatory) approaches (Supplementary Fig. 3). Sample group clustering between MD-based food and placebo, both for prokaryotes and eukaryotes, was supported by high Pearson’s correlation values between components (0.79). At the same, PLS-DA cloud separation was supported when timing (before treatment, treatment and wash out) or colon tracts were used as metadata in the clustering analysis.
Once the relative percentages at different taxonomic levels were obtained, we used paired group nonparametric corrected statistics to explore differences between MD-based food and placebo. Differences were found for 18 statistically significant taxa. Twelve were at the genus level and 6 were from unclassified genera referring to higher taxonomic levels (Fig. 1). Feeding with MD-based food promoted increased abundances of Acidaminococcus, Parabacteroides, Proteus, Dialister, and unassigned Enterobacteriaceae and Burkholderiales. Bilophila, Holdemanella, Subdoligranulum, Faecalibacterium, Enterococcus, Alistipes, Holdemania, Ruminococcaceae UBA1819, Ruminococcaceae DTU089, and unassigned Oscillospiraceae, Rhizobiaceae and Enterobacterales decreased. By applying the same non-parametric test, eukaryotic taxa abundances from placebo and superfood groups were compared. Pichia genus abundance was higher (q < 0.05) in superfood samples than placebo, whereas the opposite was observed for Saccharomycetales order not assigned sub taxa, as evidenced by the boxplot interquartile ranges (data not shown). The statistically significant differences increased to 88 taxa when MD-based food and placebo were compared depending on the intestinal tract (Supplementary Table 6). Eighteen of these taxa were common between DC and TC, confirming the closeness between these tracts as observed with PCoA graph (Supplementary Fig. 2c).
With the aim of ascertaining the microbiota evolution in the Twin M-SHIME unit fed with MD-based food, we pooled the samples into three timeframes: the first including T2, T4 and T7 (namely the first week of intake); the second T9, T11 and T14 (second week of intake); and the third W2, W4 and W7 (wash out). Overall, 51 statistically significant changes were found, and 13 of them referred to the second week of MD-based food intake (Supplementary Table 7). Parabacteroides, Dialister, Enterococcus, [Eubacterium]_ventriosum_group, Proteus, Lactobacillus, Prevotellaceae_NK3B31_group, Eggerthella, Lachnospiraceae_UCG-010 and three unassigned groups belonging to Lachnospiraceae and Anaerovoracaceae families and to Burkholderiales order exhibited significant changes. Compared to the first week, Lactobacillus, Enterococcus and undetermined taxa resolved at the Burkholderiales order level increased in the second week of MD-based food administration. The reshape of the microbiota was lost in the luminal compartment after the wash out, but it persisted at the mucosal level.
MD-based food modulates the microbiome functional potential. To establish whether the MD-based food intake boosted beneficial activities, shotgun metagenomics was applied on a subset of samples only from the Twin-M-SHIME unit fed with MD-based food. On average, 26,828,286 (± 1,404,770) high quality reads were obtained for microbiome analyses. The variation of the taxonomic composition during the MD-based food administration was consistent with the results from 16S rRNA gene sequencing. The screening of HumaNn functional profiles allowed the identification of several microbial pathways, which were promoted through the MD-based food intake. Genes responsible for colanic acid biosynthesis increased their abundance both at lumen and mucosa levels (Fig. 2a). On the contrary, the abundance of genes involved in the L-valine pathway decreased (Fig. 2b). At the lumen level, the MD-based food intake boosted an increase of the abundance of several genes responsible for the carbohydrate metabolism (CAZy). They belong to glycosyl-transferase family, carbohydrate-binding modules and auxiliary activities. However, after the wash out, their abundance turned into values observed before treatment. In addition, CAZy gene patterns significantly differed between lumen and mucosa (Fig. 2c). An increase of the 6-phospho-beta-glucosidase gene, which is responsible for the synthesis of isothiocyanates from glucosinolates, was also detected both in lumen and mucosa samples (Fig. 2d). Finally, an increase of genes coding for butyryl-CoA:acetate-CoA transferase involved in butyrate biosynthesis was observed during treatment at the mucosal level, and persisted after the wash out (Fig. 2e).
MD-based food promptly increases the content of SCFAs. Metabolomics before and during feeding, and during wash out followed a statistical trend. Therefore, these and the following results only refer to some sampling times: before feeding (T0), at 7- and 14-days during feeding, and after 7 days of wash out (Table 2). Before treatment, the content of SCFAs did not significantly differ (p < 0.05) between the two Twin M-SHIME units. Compared to placebo, the intake of MD-based food promoted higher (p < 0.05) levels of SCFAs in all colon tracts both during feeding and after wash out. Compared to T0, the administration of MD-based food determined an increased content of SCFAs in all colon tracts until 7 days. However, this increase was temporary. After 14 days of intake, the concentration of acetic acid in all colon tracts did not significantly differ with respect to those observed before treatment. At the same time, propionic acid was even lower level in AC and TC (13.59 ± 1.30 and 11.31 ± 0.44 mM, respectively, p < 0.05). On the contrary, the concentration of butyric acid in DC remained significantly higher (5.99 ± 0.30 mM, p < 0.05). Except for butyric acid in TC and DC (6.36 ± 0.5 and 4.91 ± 0.24 mM, respectively), and acetic acid in DC (15.74 ± 1.21 mM), the levels of SCFAs after wash out were significantly lower than those observed at T0. The administration of placebo always displayed a decreasing trend out of SCFAs. The only exception was butyric acid, which never significantly varied during time.
Table 2
Concentration (mM) of short chain fatty acid (SCFAs) in the ascending (AC), transverse (TC) and descending (DC) colon tracts of the SHIME units fed with MD-based food (1) and placebo (2), before (T0) and during treatment (7 and 14 days) and after 7 days of wash out. Data are based on the average of three independent analyses. In the same column and for each compound, values with different superscript letters differ significantly (p < 0.05) based on one-way ANOVA and individual post hoc comparisons with Tukey–Kramer. At each timepoint, for each colon tract and specific SCFA, values in the same row (referring to SHIME units 1 and 2) with asterisks mark statistically significant results according to Student’s t-test (p < 0.05).
Short Chain Fatty Acids (mM)
|
|
Sample
|
AC1
|
AC2
|
TC1
|
TC2
|
DC1
|
DC2
|
Acetic acid
|
T0
|
26.08a
|
26.51a
|
22.85a
|
23.18a
|
17.92ac
|
18.01a
|
|
7 days
|
36.02b
|
25.08b*
|
32.35b
|
21.68a*
|
27.77b
|
16.22a*
|
|
14 days
|
24.32a
|
16.82c*
|
24.32a
|
14.54b*
|
18.62a
|
10.87b*
|
|
Wash out
|
19.56c
|
13.61d*
|
17.56c
|
11.66c*
|
15.74c
|
8.80b*
|
Propionic acid
|
T0
|
16.87a
|
16.76a
|
13.89a
|
13.88a
|
10.11a
|
9.92a
|
|
7 days
|
19.27b
|
14.78b*
|
16.87b
|
13.27a*
|
14.92b
|
9.08a*
|
|
14 days
|
13.59c
|
9.91c*
|
11.31c
|
8.90b*
|
10.01a
|
6.13b*
|
|
Wash out
|
10.46d
|
8.12d*
|
9.10d
|
7.21c*
|
8.18c
|
4.66b*
|
Butyric acid
|
T0
|
8.44a
|
8.76a
|
6.59a
|
6.32a
|
4.52a
|
4.69ab
|
|
7 days
|
11.98b
|
7.69b*
|
10.78b
|
5.99a*
|
8.93b
|
5.42a*
|
|
14 days
|
8.03ac
|
5.16dc*
|
7.23a
|
4.21a*
|
5.99c
|
3.63bd*
|
|
Wash out
|
6.50c
|
4.17d*
|
6.36a
|
3.25a*
|
4.91a
|
2.94cd*
|
MD-based food modulates the synthesis of VOCs. DAPC on metabolomic data demonstrated how the MD-based food administration led to a shift of VOCs with respect to placebo and T0. The shift was persistent also after the wash out (Fig. 3). Clouds referring to MD-based food intake and related wash out were plotted into the first and fourth quadrants, respectively. The other clouds were almost closely placed in the second quadrant. Variables that mainly contributed to this stratification were nonanoic acid, hexanol 2 ethyl, indole, acetic acid pentyl ester, methyl valerate, 2,6,10-trimethyl-1,5,9-undecatriene, 2(3H)-furanone, dihydro-5-pentyl-, butanoic acid, 2-methylpropyl ester, phenol, [1,1'-bicyclopentyl]-2-one, butanenitrile, 4-(methylthio)-, 1-hexanol, and n-propyl linoleate. When placebo and MD-based food were compared during treatment and wash out, the capability of MD-based food in modulating persistently the synthesis of VOCs was confirmed (Fig. 4). In detail, 27 out of 35 VOCs, which increased during MD-based food intake were also further detected. Eighteen compounds were esters derived from straight-chain saturated fatty and other carboxylic acids, and 9 were miscellaneous. Six (mainly alcohols and ketones) out of 9 VOCs, decreasing during MD-based food administration, kept this trend after wash out.
Correlation among SCFAs, microbiota and VOCs. We merged single normalized data matrices and run a Pearson’s correlation test to get putative negative/positive correlations among SCFAs, microbiota and VOCs (Supplementary Fig. 4). Only statistically significant (p < 0.01) correlations were reported. Looking for high correlation values (-0.7 < x < 0.7), it appeared that most of the positive correlations between VOCs and microbiota regarded Dialister, unassigned genera belonging to Enterobacteriaceae family, and Acidaminococcus, as the only genus showing a negative correlation with 2-dodecanol. The abundance of Acidaminococcus increased in agreement with MD-based food intake, and correlated with pentanenitrile, 5-(methylthio) and the derivative of the oxidative breakdown of linoleic acid, 1 octen 3ol. This latter positively correlated with the unassigned genera belonging to Enterobacteriaceae family and Acidaminococcus. Dialister positively correlated with several esters as well as with the unassigned genera from Enterobacteriaceae. Two-butanone positively correlated with Dialister and Acidaminococcus genera, while the abundance of Parabacteroides increased together with phenol. SCFAs positively correlated with 2(3H)-furanone, dihydro-5-pentyl-. Butyric acid also positively correlated with 1,6 octadien 3ol 3,7 dimethyl and pentanenitrile, 5-(methylthio).