The Improvement of Parturition Progress by High Intake of Dietary Fibre in Late Gestation is Associated with the Altered Gut Microbiome and Metabolome in Sows

Background: Gestational intake of dietary bre improves the parturition progress, which largely affects developmental outcomes of the offspring. Dietary bre can alter the gut microbiome and production of symbiotic metabolites, e.g. short chain fatty acids (SCFAs). We hypothesized that the improvement of parturition progress by dietary bre is associated with the symbiotic metabolites generated by the gut microbiome. Methods: Yorkshire sows were randomly given diet containing normal level of bre (NDF, 16.2% dietary bre, n = 20) or high level (HDF, 30.1%, n = 20) with other nutrients identical from days 90 of gestation to parturition. Faecal microbiome proled with 16S amplicon sequencing, SCFAs and metabolome in the faeces and plasma around parturition were compared between the dietary groups. Correlation analysis was conducted to further explore the potential associations between specic bacterial taxa and metabolites. Results: HDF signicantly improved the parturition progress, indicated by the shorter parturition duration. HDF increased abundance of the phyla Bacteroidetes and Synergistetes and multiple genera. Except for butyrate, SCFAs levels in the faeces and plasma of sows at parturition were increased in HDF group. The abundances of 15 and 12 metabolites in the faeces and plasma, respectively, markedly differ between HDF and NDF sows. These metabolites are involved in the bacterial metabolism of amino acids, bile acids, SCFAs and dietary bre. Correlation analysis also showed associations between specic taxa (genera Cellulosilytica and Lachnoclostridia) and metabolites (acetate and isobutyrate). Conclusions: The improvement of parturition process by high bre intake in late gestation is associated with altered gut microbiome, production of SCFAs and other metabolites, potentially serving for energy metabolism.


Background
Parturition progress is associated with birth and developmental outcomes of the offspring [1]. In polytocous animals, such as pigs, parturition process can be rather long. Prolonged parturition duration increases the number of stillbirth and adversely affects sow health postpartum. Around 75% of the stillbirth can be attributed to asphyxia, which affects last-born pigs even worse [1]. Dietary factors, besides length of gestation, can also affect parturition progress [2]. Fibrous diet during gestation has been shown to reduce the parturition length of sows with a range of 9-29% [3]. Our own study also demonstrated that diet enriched with inulin, a fermentable bre from roots, shortened parturition duration in sows [4]. Although several hypotheses have been proposed, including high bre intake reducing constipation or/and removing excess body fat, the underlying mechanisms of high bre intake during gestation on improving parturition process remains exclusive.
The gut microbiome is associated with various physiological aspects of host and largely affected by diet [5]. A plethora of studies have associated intake of dietary bre with the change of composition and function of gut microbiome [6][7][8]. The production of symbiotic metabolites by gut microbiome had been proposed to play roles in gut microbiome affecting host physiology [9]. Dietary bres, including nonstarch polysaccharides, oligosaccharides and resistant starches, are microbiota-accessible carbohydrates and fermented by gut microbiome to produce various symbiotic metabolites [9]. Short-chain fatty acids (SCFAs) are one major type of symbiotic metabolites [10], which decrease the pH in the gut lumen to suppress the bacterial growth, and serving as an energy substrate of the host [11]. SCFAs can also regulate metabolism and immunity not only based on the gut, but also the liver and peripheral tissues via the peripheral circulation [5]. To the best of our knowledge, whether the gut microbiome and symbiotic metabolites, such as SCFAs, are involved in the high bre diet-improved parturition progress is not known yet.
Against this background, we aimed to test whether the altered gut microbiome and -related metabolites are associated with the improved parturition progress by high intake of dietary bre, using sows as a model. The gut microbiome of sows given high level of dietary bre was pro led by 16S rRNA amplicon sequencing and compared with that of sows given normal level of dietary bre from days 90 (d 90) of gestation to parturition. Levels of SCFAs in faeces and plasma were quanti ed and metabolome were pro led by untargeted metabolomics. Abundance of these metabolites and bacterial taxa were correlated to further explore any potential association.

Materials And Methods
The animal experiment was approved by the Animal Care and Use Committee of the Sichuan Agricultural University (Permit No. DKY-B20121602), and was conducted at Tianfu Pig Farm, Giastar Group, Chengdu, China in accordance with the National Research Council's Guidelines for Care and Use of Laboratory Animals.

Animals and dietary intervention
Based on parity (4~6) and body weight (267.4 ± 16.8 kg), 40 Yorkshire sows were randomly allocated into 2 groups (20 per treatment) receiving two feedings different in bre content. The two feedings were formulated based on corn-soybean meal to meet or exceed the recommendation of NRC (2012) as shown in Table S1: Normal dietary bre diet (NDF, 16.2 % total dietary bre) and high dietary bre diet (HDF, 30.1% total dietary bre). In addition to dietary bre, other nutrients such as digestible energy and crude protein intake were similar via adjusting feed intake (3.0 kg/d in CON and 3.2 kg/d in HDF) from d 90 of gestation to parturition. The daily intake of dietary bre in HDF group was about twice as much as CON group (485 g/d in CON and 964 g/d in HDF). During d 90 to 110 of gestation, the diet was supplied once a day (08:00). On d 111 of gestation, sows were moved to the farrowing room and feeding frequency turned to twice a day (08:00 and 15:00) until parturition. Parturition duration was recorded accurately during delivery of the sows. The health care and immunization procedures of the sows and newborn piglets followed the regulation of the farm.

Sample Collection
Blood samples (10 mL, n=10) from ear vein were collected into sodium heparinized tubes on d 107 of gestation and within 1 h of parturition. Plasma was obtained by centrifuging at 3,000 × g for 15 min, and stored immediately at -20 °C for later analysis. Fresh faecal samples of sows were collected on d 110 of gestation and stored immediately in liquid nitrogen for later analysis.

Faecal Microbiomics
Genomic DNA was extracted from faecal samples using Mo Bio Power Faecal DNA Isolation Kit (Mo BIO, Carlsbad, CA, USA). The v4 hypervariable regions of 16S rRNA was ampli ed using primers 515F and 806, and the amplicon pyrosequencing was carried out on an Illumina HiSeq PE250 platform (Illumina, San Diego, CA, USA). Raw data was processed with RRR pipeline based UPARSE (Version 7.0.1001). Ribosomal Database Project classi er (Version 2.2) was used to assign taxonomic rank. OTUs were clustered at 97% sequence identity (Sequences with ≥ 97% similarity were assigned to the same OTUs).
OTU and sample information were imported into R [12] interfaced with R Studio for further analysis using the Phyloseq and VEGAN [13] packages. Bray-Curtis and unifrac dissimilarities were calculated to present the β-diversity and displayed by Non-metric Multi-dimensional Scaling (NMDS), and difference between the dietary groups was tested by Permutational Multivariate Analysis of Variance (PERMANOVA) test with permutation 1,000 times. Difference in α-diversity presented by Shannon index and differential abundance between the dietary groups of speci c genus or phylum were tested by the nonparametric Wilcoxon sum rank test. P values indicating signi cance of difference at genus level were further adjusted by a two-stage Benjamini and Hochberg (TSBH) step-up FDR-controlling procedure with type I error rate (α) set to 0.20 using the mt.rawp2adjp function in multcomp package [14]. P values of phyla were not adjusted as there were less than 20 tests conducted. Effect size was also calculated using equation, Effect size = , in which Z is the Wilcoxon Z and N is the number of samples.

Quanti cation of SCFAs in faeces and plasma
Concentrations of SCFAs (Acetate, Propionate, Butyrate, Isobutyrate, Valerate and Isovalerate) and sum of them (Total SCFAs) in faecal and plasma samples were determined as previously described [15] with minor modi cations. Brie y, supernatant obtained from faecal suspension (0.7 g faecal matter in 1.5 mL water) or plasma was mixed with 25% metaphosphoric acid and crotonic acid solution (210 mmol/L). The extract was further mixed with methanol (1: 3 dilution), ltered by 0.22-µm lter (Millipore, Bedford, MA, USA) before being manually applied onto a gas chromatographer with flame ionization detector (Varian CP-3800, manual injection, flame ionization detector, FID, 10 µL micro-injector) for quanti cation. Plasma was directly mixed with metaphosphoric acid and crotonic acid for SCFAs extraction. Parturition duration, SCFA concentrations were analysed in R using linear modelling against the dietary groups. Difference with P < 0.05 was regarded as signi cant.

Faecal and plasma metabolomics
Faecal and plasma samples were thawed and mixed with cold methanol/acetonitrile (1:1, v/v) to remove the protein. The supernatant was dried and re-dissolved in 50% acetonitrile before being applied onto a UHPLC (In nity 1290, Agilent, Santa Clara, CA, USA) coupled to a quadrupole time-of-ight mass spectrometer (qTOF MS, 6550, Agilent) for analysis in random order within the same type of samples.
ACQUITY BEH amide column (2.1 mm × 100 mm, 1.7 µm, Waters, Milford, MA, USA) was used as the separation column and the mobile phase was mixture of solvent A (25 mmol/L CH 3 COONH 4 and 25 mmol/L NH 4 OH) and B (acetonitrile). The 12 min elution gradient was starting with 95% solvent B (acetonitrile) for 0.5 min, decreasing to 60% of B in 6.5 min in linear fashion, further to 40% of B in 1 min, holding for 1 min, changing back to 95% of B in 0.1 min, then holding for 2.9 min for re-equilibration. To assist the chemical assignment, pooled sample was analysed on a TripleTOF MS (AB Sciex 6600, Framingham, MA, USA) to obtain MS and MS/MS data in an information dependent acquisition (IDA) mode. Obtained data was used for compound assignment against an in-house database established with available authentic standards.
Raw LC-MS data was converted into MzXML format using MSconvert (Version 3.0.6458, ProteoWizard, Palo Alto, CA, USA), then imported into XCMS (Version, Scripps Center, La Jolla, CA, USA) for data preprocessing. Processed data was annotated with chemical compound ID as aforementioned, then imported into R [12] integrated with R studio for abundance analysis. All the identi ed ions in the faeces or plasma were combined together, and applied to Principal Component Analysis (PCA). Each MS feature was tted to a linear model using lm function with treatment (NDF vs HDF) as the predictor. P-value for the signi cance of diet was generated by comparing these two models with anova function. P-values were further adjusted within each data set (faeces or plasma) by a two-staged TSBH FDR procedure as described above. Adjusted P value less than 0.05 was the threshold of signi cance.

Correlation Analysis of the Microbiomic and Metabolomic Data
To explore potential associations between the microbiomic compositions and abundance of metabolites in faeces or plasma, Spearman correlation was applied to the microbiomic data and individual SCFAs data, whilst unsupervised regularized Canonical Correlation Analysis (rCCA) with the shrinkage method was performed on the microbiomic and metabolomic data across the dietary groups using the mixOmics package. All analyses were conducted in R. Correlations were shown as heatmap with either Spearman's r or the rCCA similarity scores. Selected correlations with similarity score larger than 0.7 were further shown in scatter plots.

Parturition durition
The parturition durations are shown in Table 1. HDF sows had signi cantly shorter parturition duration (P < 0.05) and mean birth interval (P < 0.05), relative to NDF ones.

Faecal And Plasma Metabolomics
In total, 186 metabolites in positive mode and 170 in negative mode were annotated, whist 217 (positive) and 179 (negative) in plasma. Among these metabolites, 15 metabolites in faeces and 12 in plasma were found with signi cant difference (adjusted P < 0.05) between NDF and HDF sows. PCA score plots of faecal and plasma metabolome are shown as Fig. 2. Tendency of separation in both score plots indicates difference between the two groups in both faecal and plasma metabolome, and faecal NDF metabolome and plasma HDF metabolome was less diverse than their counterpart in the same sample type.
Information of identi ed metabolites, such as metabolite name, RT, m/z, adduct, abundance in NDF and HDF sows, fold change and adjusted P value, is listed in Table 3, respectively. While only nearly half of the faecal metabolites (7/15) showed higher abundance in HDF sows, levels of all identi ed plasma metabolites are higher in the HDF sows.

Correlation
Associations between abundance of speci c taxa and that of SCFAs by Spearman Correlation analysis, and that metabolites by rCCA in faeces and plasma are shown as heatmaps in Figs. 3, 4 and 5, respectively. Across the diet groups, multiple taxa were correlated with SCFAs in abundance in faeces and plasma. Of note, more correlations were found in plasma than in faeces and the correlations found are not identical in faecal and plasma samples. In faeces, four genera were found associated with acetate, two genera with propionate. Three genera were found associated with butyrate, six with isobutyrate ( Fig. 3A). In plasma, Anaerovibrio was correlated with acetate, while Sphaerochaeta, Sutterella and Bacteroides were correlated with propionate (Fig. 3B). The taxa, such as Cellulosilyticum, Lachnoclostridium, were found associated with intermediate metabolites related to energy metabolism, such as malic acid and 2-keto-gluconic acid as well as acetylcarnitine in plasma. Indole metabolites, such as norharmane in faeces (Fig. 4), imidazoleacetic acid in plasma (Fig. 5), were also correlated with bacterial taxa, including Anaerovibrio, which was also correlated with abundance of betaine.

Discussion
In this study, HDF diet improved the parturition progress, relative to NDF diet. HDF sows had a substantial deduction (24%) in the parturition duration and mean birth interval relative to NDF ones, which is in line with our previous study [16] and others [17].
For mammals, maternal uterine contractility directly determines fetal expulsion [18], and uterine contractility is an energy-demanding process, thus the parturition process would be affected by host energy status [19]. In human-being, glucose is of importance for the myometrium as energy source during parturition, and ATP production increases 2-to 3-fold during uterine contraction [20]. Our own study has shown that the increased energy intake in late gestation improved parturition process, using pigs as a model [21]. It has also been reported that pigs fed high-bre diet had a more constant level of blood glucose, indicating a more sustained energy supply during parturition [22]. Glucose can be produced by SCFAs, such as propionate and acetate through TCA cycle [23]. In addition to glucose, SCFAs can also be an important energy resource contributing to nearly 30% of the energy requirements of pigs [22,24], while 10% in human being and 80% in ruminants [23]. In this study, supportively, we did nd the markedly increased plasma concentrations of SCFAs in HDF sows at parturition. In the current study, acetate is the most abundant SCFA, followed by propionate and butyrate in both faeces and plasma, which is consistent with previous study [25]. The molar ratio of acetate, propionate and butyrate in faeces is 66:23:11 in sows in this study, a little different with that in human faeces, 60:20:20 [9]. Functionally, acetate can be used to produce propionate and butyrate [5], and involved in liver lipogenesis and lipolysis in adipose tissue [26]. Moroever, propionate and butyrate can regulate glucose homeostasis [5].
Of note, most of SCFAs, except for butyrate, showed higher levels in both faeces and plasma of HDF sows at parturition. In fact, ndings concerning the effect of dietary bre intake on butyrate level are inconsistent across different studies. The intake of inulin-type fructans by patients with type 2 diabetes showed unchanged level [27], whereas inulin intake by mice showed the increasing level of butyrate in faeces, but not in plasma [28]. Alfalfa-containing diet also increased butyrate level in the caecal digesta of pigs [29] and caecal levels of mucosal genes involved in SCFA sensing and absorption [29]. In our previous study, sows fed with dietary bre levels by guar gum and cellulose for the whole gestation had increased faecal level of butyrate [30]. Therefore, it is speculated that this inconsistency could be due to the heterogeneity in bre types and difference in species, and pathophysiological status of the subjects. In this study, moreover, butyrate failed to increase but valerate did. A previous study pointed out that butyrate and valerate have competitive metabolic pathways [31], this may be a reason for this result, but the precise mechanism is not su ciently clear. Regardless of diet, when it comes to parturition, the plasma concentrations of acetate and total SCFAs decreased, relative to d 107 of gestation. The lower levels of SCFAs at parturition further suggest the expenditure of SCFAs for energy supply, while HDF sows means the su cient intake of dietary bre for bacterial fermentation and constant production of SCFAs.
In addition to SCFAs, using non-targeted metabolomics analysis, we also found some intermediate metabolites were increased in HDF sows, which related to energy metabolism, such as betaine and dimethylglycine (DMG), which is in line with a study on pigs consumed high-bre rye [32]. Previous study has reported that betaine improves energy utilization, especially when energy intake is insu cient [33]. It should be noted that betaine is catabolized via a series of enzymatic reactions to donate methyl groups and DMG is an intermediate metabolite. The underlying mechanism is that betaine is catalysed by betaine homocysteine methyltransferase to form DMG [34]. In this study, the increased DMG in the plasma of sows fed HDF diet coincided with the higher betaine. In addition, the result of plasma metabolomics showed a higher concentration of propionic acid in HDF sows, which supported our SCFAs data. Furthermore, dietary bre can serve as the metabolism substrate for speci c bacteria producing secondary bile acids [35]. Our untargeted metabolomic analysis showed the decreased faecal level of 3b-Hydroxy-5-cholenoic acid, which is a monohydroxy bile acid, an intermediate of synthesis of lithocholate, chenodeoxycholate and cholate involving activities of gut bacteria [36]. Results from different studies on effect of consumption of dietary bre on bile acids are rather inconsistent. Wheat-bran bre decreased the faecal levels of lithocholate and deoxycholate in humans [37,38], whilst soluble beta-glucans increased faecal levels of primary and secondary bile acids [39].
Considering the crucial role of gut microbiome on fermentating dietary bre and production of SCFAs and other metabolites, the composition of gut microbiota was determined using 16S rRNA sequencing. We found HDF intake in the late gestation leaded to the comprehensive changes of gut microbiome, with lower abundance of Firmicutes, higher abundance of Bacteroidetes and lower ratio of Firmicutes to Bacteroidetes, which is in accordance with a previous study in rats [40]. Abundance of Phylum Synergistetes, digesting bre to produce acetic acid [31], also increased. Genera Turicibacter and Terrisporobacter showed decreased abundance in HDF sows, as reported in pigs and children consuming inulin-enriched diet [29,41,42]. The changes of multiple genera involving in bre degradation are positively correlated with the levels of speci c SCFAs in faeces and plasma, suggesting their roles in the production of SCFAs. The family Rikenellaceae and genus Cellulosilyticum degrades carbohydrates [43,44], and Cellulosilyticum was positively correlated with acetate and isobutyrate in faeces, but not in plasma in this study. Lachnoclostridium degrades complex polysaccharides to produce SCFAs [45], its positive correlation with acetate and isobutyrate was found. Alloprevotella and Pyramidobacter from the Synergistetes phylum produce acetate [45,46], but only a positive correlation between Pyramidobacter and faecal valerate was found in this study. Anaerovibrio was reported with lipolytic activity producing glycerol for propionate synthesis [31], but instead a strong correlation between the Anaerovibrio abundance and plasma acetate was observed. Nevertheless, these newly observed correlations suggest potential roles of these bacterial taxa in the production of speci c SCFAs.
Our correlations analyses showed the positive correlations between phyla Succinivibrio, Butyrivibrio and Isobutyrate and valuerate, butyrivibrio and butyrate in the faeces. Also we found the positive correlations between Succinivibrio, Cellulosilyticum and pseudouridine, which is a metabolite standing for cell turnover of the host. The family Prevotellaceae, genus Anaerovibrio and imidazoleacetic acid involving in bacterial metabolism of tryptophan, Lachnoclostridium and an intermediate of energy metabolism, 2-keto-Gluconic acid, were also observed by rCCA. Moreover, the strong positive correlation found between Anaerovibrio and plasma level of betaine suggested the role of gut microbiome in betaine metabolism. These correlations suggest new associations between bacterial taxa and host metabolites, revealing the potential mechanism of dietary bre intake on host physiology.

Conclusion
Our study showed that high intake of dietary bre in late gestation improved parturition process. This improvement may be associated with the altered gut microbiome and production of SCFAs, as well as other metabolites involving in energy metabolism and host physiology. However, the further investigations are needed, whether other mechanism remains about the improvement of parturition process by dietary bre, also bre types and gut microbiome differences across species need to be concerned.  α-diversity (A), β-diversity using Bray-Curtis distance (B) and unifrac distance (C) of the gut microbiome, and genera with differential abundance between HDF and NDF sows (D). Heatmap of correlation of the gut bacterial taxa and SCFAs in faeces (left) and plasma (right). Correlation coe cients were coloured according to the scale listed on the right.