Sample hydrolysis, extraction and quantification
The methodology for the extraction of lignans from food samples as well as of enterolignans from fermentation faecal samples was adapted from the work of Nørskov and Knudsen [46] as well as the work of Milder et al. [47] with minor modifications. The combination of solvent-based extraction and alkaline hydrolysis was used to release lignans from food or faecal matrix. On the other hand, enzymatic extraction with β-glucuronidase/sulfatase was performed in order to degrade the glycosidic bonds and to release the lignan aglycones. Solid Phase Extraction method was used to prepare samples for UHPLC-MS/MS analysis.
A UHPLC-MS/MS method was developed to obtain high resolution and signal accuracy. The deuterated standard Secoisolariciresinol-d6 (Seco-d6) was used as the reference standard. A multiple reaction monitoring (MRM) method was set up in negative ion mode for each analyte and quantification was established using the most intense MRM signal transition with 8-16 point calibration curves of pure analytical standards. The obtained MRM detection parameters are given in Table 1.
Table 1 Multiple reaction monitoring (MRM) detection parameters
Compound
|
MW
|
Parent m/z
|
fr
m/z
|
CV
|
CE
|
Equation
|
R2
|
LOD
|
LOQ
|
Seco
|
362.42
|
361.4
|
122*
|
30
|
30
|
y = 5.5627x
|
0.9989
|
0.025571
|
0.085237
|
Seco
|
362.42
|
361.4
|
346
|
|
17
|
|
|
|
|
Lari
|
360.4
|
359.4
|
329*
|
30
|
20
|
y = 20.963x
|
0.991
|
0.009145
|
0.030483
|
Lari
|
360.4
|
359.4
|
192
|
|
10
|
|
|
|
|
Pino
|
358.39
|
357.4
|
151*
|
30
|
17
|
y = 9.5455x + 261.14
|
0.9914
|
0.011945
|
0.039818
|
Pino
|
358.39
|
357.4
|
136
|
|
30
|
|
|
|
|
Mat
|
302.36
|
357.42
|
83*
|
30
|
23
|
y = 33.014x + 1235.7
|
0.9942
|
0.013987
|
0.046623
|
Mat
|
303.36
|
357.42
|
137
|
|
20
|
|
|
|
|
ED
|
298.33
|
301
|
253*
|
30
|
20
|
y = 67.712x + 5969.6
|
0.9912
|
0.111075
|
0.370249
|
ED
|
298.33
|
301
|
271
|
|
20
|
|
|
|
|
EL
|
298.33
|
297.3
|
107*
|
30
|
25
|
y = 105.71x + 17165
|
0.9984
|
0.019281
|
0.06427
|
EL
|
298.33
|
297.3
|
189
|
|
20
|
|
|
|
|
Seco d6
|
368.45
|
367
|
168*
|
30
|
25
|
y = 4.2692x - 60.534
|
0.9985
|
0.033394
|
0.111314
|
Seco d6
|
368.45
|
367
|
124
|
|
29
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
MW – Molecular weight; fr - fragment; * quantification fragment; CV – collision voltage; CE – collision energy; LOD – limit of detection; LOQ – limit of quantification. Calibration curves showed good linearity; LOD was calculated as a signal to noise ratio (S/N) equal to 3; LOQ was calculated as a signal to noise ratio (S/N) equal to 10
Determination of the lignan content in the oilseed mix
Obtained concentrations (mg/100 g of dry matter) of free and total (free plus bound) Seco, Lari, Pino and Mat in the oilseed mix are listed below in Table 2.
Table 2 Free and total (free plus bound) lignans in oilseed mix and baseline concentration in in vitro fermentation medium
|
|
Mean/SD
|
Seco
|
Lari
|
Mat
|
Pino
|
Total
|
Oilseed mix lignans
(mg/100 g)
|
Free
|
Mean
|
31.1
|
44.9
|
3.8
|
10.7
|
90.5
|
SD
|
9.2
|
19.1
|
0.4
|
2.3
|
30.9
|
Total
|
Mean
|
1467.6
|
188.9
|
132.2
|
172
|
1960.7
|
SD
|
194.5
|
4.3
|
23.6
|
0.5
|
222.9
|
Lignan baseline concentration in the fermentation medium (ng/ml)
|
Free
|
Mean
|
1480.95
|
2138.10
|
180.95
|
509.52
|
4309.52
|
SD
|
438.1
|
909.5
|
19.0
|
109.5
|
1471.4
|
Total
|
Mean
|
69885.71
|
8995.24
|
6295.24
|
8190.48
|
93366.67
|
SD
|
9261.9
|
204.8
|
1123.8
|
23.8
|
10614.3
|
As for the baseline in fermentation experiments, 0.5 g of oilseed mix was added to 105 ml of fermentation medium. Data are expressed as the mean ± SD (standard deviation)
The UHPLC-MS/MS analysis revealed that amongst free fractions, the quantities of Lari in the oilseed mix prevail (44.9 mg out of 90.5 mg per 100 g of mix), whereas amongst total lignans, the mean values for Seco are the highest (1467.6 mg out of 1960.7 mg per 100 g of mix).
Analysis of enterolignans in fermented samples
The results of batch culture fermentation experiments revealed differences in the concentration of individual lignans and enterolignans at different time points in the samples with added oilseed mix between young healthy (YD) and premenopausal (PD) donors (Table 3). Figure 2 and Figure 3 report the values of free lignans and enterolignans respectively, assessed during the fermentation at 3 time points, 0.2 (T0.2), 5 (T5) and 24 hours (T24).
Table 3 Concentration of lignans (Seco, Lari, Mat and Pino) and enterolignans (ED and EL) in the faecal samples from younger healthy donors and premenopausal donors at 0.2 (T0.2), 5 (T5) and 24 hours (T24) of fermentation
|
Younger Healthy Donors
|
|
T0.2
|
T5
|
T24
|
Seco
|
246±267.5
|
19.7±21.6
|
142.7±134.7
|
Lari
|
19.2±19.1
|
18±13.1
|
32.8±30
|
Mat
|
1140.9±1846.6
|
40.2±20.6
|
45.8±64.5
|
Pino
|
1987.8±2899.2
|
207.3±334.6
|
111.2±163.3
|
ED
|
233±227.5
|
38.7±49.6
|
160.4±166.7
|
EL
|
330.4±397.4
|
1748.9±2166.6
|
6739.2±10554.4
|
Total (lignans)
|
3393.9±2304.6
|
285.2±389.9
|
332.5±392.5
|
Total (enterolignans)
|
563.4±445.8
|
2034.05±2556.6
|
7071.7±10946.9
|
Premenopausal Donors
|
|
T0.2
|
T5
|
T24
|
Seco
|
30.52±6.6
|
4.5±1.82
|
81±121.2
|
Lari
|
27.81±8.9
|
43.3±35.8
|
51.3±38.1
|
Mat
|
51.3±14.6
|
528.4±580.9
|
743.1±1088.9
|
Pino
|
124.1±167.5
|
4.8±6.6
|
12±10.5
|
ED
|
247.13±475.23
|
176.6±183
|
12471.8±21453.5
|
EL
|
20809.5±55911.5
|
25619±43962.5*
|
258.9±226.7
|
Total (lignans)
|
233.66±197.5
|
580.9±625
|
887.3±1258.7
|
Total (enterolignans)
|
21056.6±56386.8
|
26199.9±44587.5
|
1146.2±1485.4
|
Data are expressed in ng/ml, as the mean ± SD (standard deviation). *p < 0.05, indicating a significant difference among samples collected after 0.2, 5 and 24 h of fermentation with the oilseed mix (Kruskal-Wallis test).
The average sum of the total individual lignans in the YD samples falls about 10 times during the first 5 h of fermentation, with a slight increase at 24 h. The concentration of Mat and Pino in this group decreased significantly, whereas Seco only slightly decreased and Lari increased. At the same time, the amounts of total enterolignans in the YD samples increased more than 10 times in 24 h, with EL expanding nearly 20 times but ED being slightly reduced.
In contrast, different dynamics were found in the fermentation samples from PDs: the total lignans increased almost 4 times in 24 h with only Pino falling about 10 times, while the levels of Seco, Mat and Lari expanded. Additionally, the amount of total enterolignans halved in 24 h, with ED levels expanding nearly 60 times, but EL dropping almost 100 times from 0.2 h to 24 h, suggesting the inability of faecal bacteria from PDs to convert ED to EL. It is also worth noting that all premenopausal participants entered the experiment with very different initial levels of enterolignans (particularly, ED), resulting in a high value of standard deviation.
Figure 2 shows the dynamics of Seco, Lari, Pino and Mat in the samples from YDs (a) and PDs (b) with added oilseed mix in 24-h fermentation experiments. While very different concentrations of lignans between YDs and PDs can be observed at T0.2, at T5 their levels are almost identical as well as at T24, except for Mat, whose level increased almost 500 times in the PD group.
Figure 3 illustrates the dynamics of ED and EL in YD and PD fermentation samples during the 24-h batch culture experiment. The ED levels are equally low in YD and PD samples at T0.2 and T5, while at T24, the value in the PD samples is almost 37 times higher than in the YD samples. As for EL, the average concentration is remarkably high in the PD samples at T0.2, falling slightly at T5 and experiencing a tremendous drop at T24. Conversely, the EL levels in the YD samples are insignificant at the beginning, slightly increase at T5 and are boosted at T24. Taken together, our results clearly demonstrate different dynamics of lignan transformation in different age groups of women: adding oilseed mix to YD samples leads to an overall drop in levels of dietary lignans and to the production of enterolignans. Contrary to this, in PDs, Mat levels increase after 24 h, with increased level of ED but decreased level of EL. These observations suggest that PDs may have a different, potentially altered intestinal microbiota configuration, with very poor conversion of dietary lignans and inability to efficiently convert ED or Mat to EL.
Impact of the oilseed mix on faecal-derived microbial communities from younger healthy and premenopausal women
The faecal microbial communities from YDs and PDs were profiled over time during the fermentation experiments, to assess whether the different dynamics of conversion of the lignans of the oilseed mix were potentially attributable to different layouts and trajectories of the gut microbiota. In parallel, for each woman, two other batch cultures were set up, inoculated with the well-known prebiotic compound inulin (Raftilose P95) as a positive control or without addition of any compound (i.e. the negative control), respectively. The 16S rRNA gene-based next-generation sequencing of all fermentation samples yielded a total of 1,809,764 high-quality reads, with an average of 33,514 ± 7,194 sequences per sample, binned into 2,623 amplicon sequence variants (ASVs) at 99% similarity.
No significant differences were observed in alpha diversity across the entire dataset, regardless of the origin of the faecal sample (PD vs. YD), experimental condition (oilseed mix vs. inulin vs. negative control) and time point (T0 vs. T5 vs. T24) (p value > 0.05, Wilcoxon test). However, it should be noted that the biodiversity of both PD and YD faecal-derived microbial communities tended to decrease over time in all experimental conditions, except for the negative controls where it was generally kept (Additional file 1: Fig. S1). With specific regard to the oilseed mix, this decreasing trend was already appreciable after 5 h of fermentation in PD samples while only after 24 h for YD samples.
The Principal Coordinates Analysis (PCoA) of inter-sample variation (i.e. beta diversity), based on unweighted (Fig. 4) and weighted (Additional file 2: Fig. S2) UniFrac distances, showed a significant separation between the faecal-derived microbial communities of PDs and YDs, regardless of the experimental condition and time point (p value < 1 × 10-4, permutation test with pseudo-F ratio). Within each group (PD and YD), the samples still segregated significantly according to both the experimental condition and the time point (p value < 0.001), suggesting a differential impact of supplements over time, likely related to the baseline microbial community. In particular, the PD faecal microbial ecosystem underwent an early modification of its overall structure, as evidenced by the shift at T5, and continued to evolve up to 24 h under all experimental conditions. The same was basically true for YD samples based on unweighted UniFrac distances, while a significant shift in the weighted UniFrac-based PCoA was observed only at T24, suggesting a dominant configuration likely to be more resilient to external perturbations compared to the PD one. Please, see Additional file 3: Table S1 for the results of adonis and ANOSIM statistics applied to weighted and unweighted UniFrac distance-based ordination.
At T0, a number of significant taxonomic differences were indeed observed between PDs and YDs. In particular, compared to YDs, the baseline faecal-derived microbial ecosystem of PDs showed increased relative abundance of some Bacteroidetes members, especially Bacteroides and Rikenellaceae, as well as Ruminococcaceae genera (p value ≤ 0.05, Wilcoxon test). On the other hand, Coriobacteriaceae, especially Collinsella, and Streptococcus were much more represented in the baseline microbiota of YDs vs. PDs (p value ≤ 0.05) (Additional file 4: Fig. S3).
With specific regard to the impact of oilseed mix on the faecal-derived microbial communities of PDs vs. YDs, we observed both common and unique microbial signatures of response (Fig. 5). Among the features shared between PDs and YDs, it is worth noting that 24 h of fermentation with the oilseed mix resulted in decreased proportions of Ruminococcaceae and various members of Lachnospiraceae, and increased amounts of Enterobacteriaceae (p value ≤ 0.1). Although in the absence of statistical significance, these variations were also observed in the control vessels (i.e. with the addition of inulin or without any supplement), suggesting that they could be related to the experiment per se. Nonetheless, the Enterobacteriaceae increase in the 24-h faecal-derived microbial communities of YDs in the presence of the oilseed mix was far greater than in PDs (p value = 0.012) (24-h relative abundance: YDs with the oilseed mix, 78.5%; PDs with the oilseed mix, 27.7%; YDs with inulin, 46.4%; YDs without addition, 48.2%; PDs with inulin, 21.5%; PDs without addition, 17.5%). Going down in the taxonomic classification, we found that this increase was largely attributable to Klebsiella, whose 24-h relative abundance was 51.3 vs. 1.7% in the YD and PD group, respectively. The mean values in the YD and PD controls were 6.2% (±8.7%, SD) and 1.7% (±1.3%), respectively. Unfortunately, the other enterobacteria were unclassified at the genus level. At the species level, Klebsiella ASVs were found to be variously assigned to K. pneumoniae and K. aerogenes, with the latter being overall more prevalent (i.e. present in 9/27 fermentation samples of YDs, of which 5 in the presence of oilseed mix, while only in 5/27 for PDs, of which 3 in the presence of oilseed mix). Furthermore, it is interesting to point out that the family Clostridiaceae increased in PD-derived microbial communities in the presence of oilseed mix but not in the YD-related ones, as well as the genus Collinsella, whose trend was exactly opposite (i.e. decreased) in YD-related samples (p value ≤ 0.2).