Rapid development of modern food industry drives people to change their dietary habits across the world. Trends in consumption of energy-dense diet containing highly-refined carbohydrates, fat and low dietary fiber are accompanied by a global obesity epidemics and related chronic metabolic diseases (Popkin et al., 2012). As for gestating women, diet-induced obesity or excessive gestational weight gain could not only result in remarkable influences on the maternal health and pregnancy outcomes, but also long-term effects on their offspring (Fraser et al., 2010; Lawlor et al., 2012). Gestation period is a key window for both mother and offspring with tremendous hormonal, metabolic and immunological changes occurring in the maternal body (Newbern and Freemark, 2011), during which microbiota remodeling is thought to be a positive response for mother to support a successful pregnancy (Mandal et al., 2016). Gut microbiota, living with an intimate relationship with its host, has been identified as major regulator in host nutrients metabolism, gastrointestinal health and immunologic functions (Hooper et al., 2012).

Previous studies have highlighted dietary fiber supplementation as an effective manipulation to improve diet-induced obesity and related metabolic abnormalities (Brownlee, 2011; Chen et al., 2015). As the major energy source for gut microbiota, dietary fiber is believed to have significant effects on the composition and diversity of microbiota (De Filippo et al., 2010; Brownlee, 2011; Heinritz et al., 2016). The awareness about the effect of microbiota on host metabolism and health has provided insights about the role of gut microbiota and their metabolites, short chain fatty acids (SCFA), in the link between dietary fiber and obesity and its related metabolic syndromes (Delzenne and Cani, 2011; den Besten et al., 2013a). However, due to the various physicochemical properties of dietary fiber, physiological effects of dietary fiber also vary greatly as reviewed by Hamaker and Tuncil (2014). Soluble dietary fiber, in particular, which is easily fermented, would have greater impact on bacterial metabolism compared with insoluble dietary fiber (Gråsten et al., 2002). Inulin-type fructans, a typical soluble dietary fiber, is a mixture of polymers and oligomers, which are composed of fructosyl units linked by β(2 → 1) glycosidic bonds. Due to this β-configuration, inulin is resistant to hydrolysis by digestive enzymes (Micka et al., 2017).

Given that there are few studies aiming to evaluate the microbial mechanism of soluble dietary fiber in improving maternal and neonatal health upon different diet types during pregnancy, the current research was undertaken to investigate effects of adding inulin to low- or high -fat diet, on the composition and metabolites of fecal microbiota from early gestation to perinatal period, as well as maternal and neonatal health parameters in a pregnant sow model. It is supposed to provide some microbial mechanistic insights into the application of inulin to a typical gestational diet characterized by high fat and energy intake for improving maternal and neonatal health.

Rodents have been used frequently as animal models for humans, however, the discrepancies in physiology and metabolism compared with humans are worth considering. In contrast, the similarities between humans and pigs in terms of anatomic, cardiovascular, gastrointestinal systems place the pig in a superior position over other non-primate models in many studies (Swindle et al., 2012). Above all, the pig is a human-sized omnivorous animal with comparable requirements for nutrients, therefore shows similarities to the human intestinal microbial ecosyste (Heinritz et al., 2013). This has driven the pig to be the alternative animal model for research into dietary modulation of microbiota–health interactions.

It is well-known that excessive intake of dietary fat is associated with being overweight or obese while dietary fiber has been extensively demonstrated to be an effective dietary intervention mean to reduce body weight gain and fat accumulation (Galisteo et al., 2005; Han et al., 2015, 2017; Palou et al., 2015). Although there is a scarcity of evidence on the effects of dietary fat or fiber in pregnant individuals, our findings that inulin attenuated the high fat diet-induced body weight gain and fat mass accumulation were in good agreement with previous studies conducted on non-pregnant individuals (Yan et al., 2013; Alligier et al., 2014; Cluny et al., 2015).

To evaluate the potential effects of lower gestational body weight and backfat gain induced by inulin supplementation, concentrations of circulating inflammatory indices and adipocytokines in both circulation and backfat tissues were investigated. Generally, obesity and related metabolic disorders are characterized by chronic or low-grade inflammation (Hotamisligil, 2006; Saltiel and Olefsky, 2017). Weight loss or fat mass reduction has been proven to result in reduction of inflammatory biomarkers (Richard et al., 2013; Barazzoni et al., 2014). Unsurprisingly, significantly decreased IL-6 and numerically increased IL-10 with inulin addition in the present study indicated an alleviative effect of inulin on maternal inflammatory status at perinatal period. The anti-inflammatory effect of inulin was also confirmed by Yasuda et al. (2009) and Lecerf et al. (2012) previously.

Leptin, adiponectin, and chemerin are adipocyte-secreted hormones with different physiological effects. Leptin regulates a wide variety of physiological processes, including feeding behavior, metabolic rate, sympathetic nerve activity, reproduction and immune response (Xiong et al., 2004; Brüll et al., 2017). Acute oral administration of specific SCFA in vivo, or incubating adipocyte cell line or adipose tissue with SCFAs in vitro have resulted in increased circulating leptin levels (Xiong et al., 2004) or leptin expression (Xiong et al., 2004; Al-Lahham et al., 2010). The mechanism under which SCFAs stimulate leptin production was due to the activation of G protein-coupled receptors (GPCRs) expression in adipose tissues by SCFAs which acts as the specific agonists of GPCRs (Xiong et al., 2004; Zaibi et al., 2010). In sharp contrast to those findings, there was consistent data showing that chronic HFD resulted in increased, while chronic dietary fiber addition caused decreased level of serum leptin in both non-pregnant (Artiss et al., 2006; Islam et al., 2011) and pregnant individuals (Vähämiko et al., 2013). The discrepancies among studies could be explained by the in vivo or in vitro, chronic or acute experimental backgrounds. Circulating leptin levels are directly associated with adipose tissue mass (Campfield et al., 1996; Ghantous et al., 2015). It is noteworthy that the decreased serum leptin level in chronic dietary manipulation in vivo studies were generally accompanied by distinctly less body weight gain when compared with high fat–fed group. As a consequence, despite the well-known leptin-stimulating effect in adipose tissue, circulating leptin level could still exert obvious reduction due to the less body weight gain and body fat accumulation with inulin addition combined with a high-fat diet. It should be mentioned that highly elevated leptin levels during pregnancy are associated with maternal leptin resistance, insulin resistance, metabolic disturbance, and increased risk of gestational diabetes and hypertension (Butte, 2000; Heerwagen et al., 2010; Vähämiko et al., 2010). Adiponection has anti-inflammatory and insulin-sensitizing effects. Plasma adiponectin concentration is negatively associated with adiposity (Nakamura et al., 2014). Although a growing body of evidence has indicated the positive association between blood adiponectin level and dietary fiber intake (Yannakoulia et al., 2008a,b; AlEssa et al., 2016), we only found a tendency of increase in adiponectin level in backfat tissue, but not in the circulating concentration. Chemerin, a recently identified adipokine that regulates adipocyte differentiation, has been positively linked to adiposity, insulin resistance, metabolic syndrome risk factors, and inflammatory markers (Lehrke et al., 2009; Ernst and Sinal, 2010; Sell et al., 2010; Weigert et al., 2010). Previous studies have indicated circulating levels of chemerin correlated with BMI, and weight loss was able to decrease the circulating chemerin level and adipose tissue chemerin expression in humans (Sell et al., 2010; Chakaroun et al., 2012). In line with these findings, circulating and backfat tissue chemerin levels in our study were significantly reduced by inulin addition, which may contribute to improved insulin sensitivity and subclinical inflammation beyond significant weight loss as indicated by Chakaroun et al. (2012).

As we know, fetal growth and development is primarily dependent upon the nutritional, hormonal and metabolic environment provided by the mother (Tzanetakou et al., 2011). Previous studies have indicated that maternal high fat feeding could cause obesity in dams and modulate the intrauterine environment to predispose offspring to obesity and related metabolic disorders (Shankar et al., 2008; White et al., 2009). Strikingly, along with the alleviation of maternal body weight gain and fat accumulation as well as changes in inflammatory markers and metabolic hormones, inulin addition resulted in lower average BMI and a more even BMI distribution of the neonatal offspring, and a decreased probability of relatively “obese” newborns.

Although multiple mechanisms have been suggested to involve in such positive regulation of DF in maternal body mass and metabolic status, the central role of gut microbiota in development of obesity and its associated inflammation and metabolic disturbance has come to the forefront (Bäckhed et al., 2004; Turnbaugh et al., 2006, 2009; Shen et al., 2013). An in-depth insight into the shift of microbiota composition with inulin supplementation during gestation would provide us a novel understanding about the beneficial effect of inulin on regulation of body weight, inflammatory status and adipose tissue derived hormones which have great impact on host metabolism.

Various factors including host genetics, diets, immunological status, and antibiotic use could shape the gut microbiota (Koren et al., 2012; Scott et al., 2013). Due to the preference for different energy sources, the gut microbiota can undergo dynamic population shifts with varied dietary manipulation even during pregnancy (Gohir et al., 2015; Kong et al., 2016).

Low microbial diversity is often associated with diseases such as inflammatory bowel disease (Manichanh et al., 2006) and obesity (Turnbaugh et al., 2009). In the present study, the significant increase in Simpson index with inulin addition, particularly when given with a high-fat diet was in good agreement with Li et al. (2016) who found Shannon's richness index was significantly decreased in high-fat feeding mice and restored under bamboo fiber addition. Therefore, the improved microbial diversity would partly explain the reduction in gestational weight and backfat gain as well as the improved inflammation response at perinatl period.

The abundant phyla in sow feces were in good agreement with previous studies on pigs (Kong et al., 2016; Yan et al., 2016) with Firmicutes and Bacteroidetes accounting for 75% or more of the microbial composition. Previous studies had demonstrated HFD could increase Firmicutes and decrease Bacteroidetes abundance (Turnbaugh et al., 2008). Increased ratio of Firmicutes to Bacteroidetes was indicated to be associated with obesity phenotype (Ley et al., 2005). In line with the tendency of increase in Firmicutes/Bacteroidetes, we did found notable increases in maternal weight and backfat gain with fat addition. Phylum Actinobacteria has been shown to be substantially more abundant in inflammatory bowel disease patients (Frank et al., 2007) and in colon and terminal ileum of obese Ossabaw minipigs (Pedersen et al., 2013) than their control counterparts. The tendency of increase of phylum Actinobacteria with high fat supplementation coincided with the excessive gestational weight gain induced by fat addition to some extent, also suggested the low-grade inflammatory state of sows under high fat treatment. Phylum Tenericutes, similarly, has been found to have significant higher abundance in obese Göttingen pig cecum (Pedersen et al., 2013) and diet-induced obese mice (Turnbaugh et al., 2008) compared to the controls. On the contrary, phylum Lentisphaerae was demonstrated to be higher in the healthy group than in humans with non-alcoholic fatty liver disease (NAFLD) (Jiang et al., 2015). The tendency of decrease in Tenericutes with inulin addition were partially in agreement with results found by Everard et al. (2014) who found probiotic yeast addition decreased 57% of the amount of Tenericutes. Moreover, significantly higher abundance of Lentisphaerae due to inulin addition in the present study would also be another possible microbial mechanism underlying the regulatory effects of inulin on host body weight and fat storage.

Oscillospira has been observed to relate to, even contribute to leanness or lower BMI and also negatively be associated with inflammatory diseases as reviewed by Konikoff and Gophna (2016). The potential underlying mechanism, is due to the fact that genus Oscillospira was able to degrade mammalian-derived glycans (such as fucose, sialic acids, and glucuronic acid) either from the host or from diet, leading to extra metabolic energy consumption by host to regenerate degraded glycoproteins that compose, for example, intestinal mucins (Kohl et al., 2014). Significant increased abundance of Oscillospira with inulin addition could explain the body weight- and fat storage-reducing effects of inulin at genus level. Genus Streptococcus was found to involve in the fermentation of sugars, yielding lactic acid as their predominant fermentation end product (van den Bogert et al., 2013). Species belonging to Phascolarctobacterium (for e.g., Phascolarctobacterium succinatutens) were demonstrated to be able to further metabolize succinate, a fermentation product of dietary fiber, to form propionate (Engels et al., 2016). The increases of these two genera well confirmed the nature of inulin as a dietary fiber to be utilized by specific bacteria. Compared with inulin, fat addition showed less dramatic effects on the modification of genus taxonomies. As the most well-known probiotics, Lactobacillus and Bifidobacterium have received great concern in previous studies. It should be mentioned that there were no significant increases for both of them with inulin addition in our study. This missing effect was in line with results reported by Paßlack et al. (2015) with 3% inulin addition in sows during gestation and lactation and results obtained by Mair et al. (2010) with 0.4% inulin addition in newly weaned piglets.

With regard to different gestation stages, our present results were inconsistent with Koren et al. (2012) and Kong et al. (2016) who found decreased OTUs in pregnant women or pigs, as well as Collado et al. (2008) who found increased total fecal cell counts from the first to third trimester in women. The discrepancy could be contributed by the genetic background of the experimental subjects and environmental deviation. Slight but not significant decline in Simpson index over the gestation aligned with the results obtained by Kong et al. (2016) and DiGiulio et al. (2015). However, significant increase of Chao 1 index on d110 of gestation, which is some controversial to previous results which found decreased Chao 1 index in sows (Kong et al., 2016) and similar Chao 1 index in women (DiGiulio et al., 2015) from early to late of gestation, was mostly caused by fat and inulin addition in the present study. This finding underscores the efficacy of different dietary treatments on modifying the gut microbiota.

At either phylum or genus level, results obtained on d110 of gestation showed striking differences from other gestation stages. Decreased relative abundance of Firmicutes, increased relative abundance of Bacteroidetes, as well as decreased Firmicutes/Bacteroidetes ratio on d110 of gestation was in line with results reported in the proximal colon content samples by Kong et al. (2016) who sampled on d45, d75, and d110 of gestation, however, Koren et al. (2012) and DiGiulio et al. (2015) found no dramatic change over the course of gestation in women. According to Jumpertz et al. (2011), increase in Firmicutes and a corresponding decrease in Bacteroidetes were associated with an increased energy harvest. Our results indicated some degree of increased energy expenditure on d110 of gestation. In fact, viewed from genus level, the decreased relative abundances of Clostridum, Ruminococcus, Turicibacter, Lactobacillus, Terrisporobacter, and Christensenellaceae_R-7_group, along with the increased abundances of Prevotella, Bacteroides, and YRC22 on d110 of gestation could well explain the decline in Firmicutes and increment in Bacteroidetes, respectively. Most species of Ruminococcus fall under Clostridium cluster IV and Clostridium cluster XIVa, which have been associated with obesity, weight gain and lipid storage (Turnbaugh et al., 2006; Nadal et al., 2009). The abrupt decline of Clostridum, more than 50% decline of Ruminococcus as well as decreases of other genera on d110 of gestation indicated that the sow were in catabolic metabolic state during the perinatal period. It is well demonstrated previously that body fat stores and protein reserves mobilized evidently at perinatal period due to reallocation of nutrients from the conceptus to the mammary tissue during which massive mammary growth occurs, and colostrum production is being initiated (Theil et al., 2014). Correspondingly, heat production also increases during this time (Theil, 2015). These physiological changes contribute to the shift of maternal metabolism from an anabolic to a catabolic state. The findings that specific microbial communities changed dramatically at perinatal period may provide the first evidence to reveal the microbial mechanism under which maternal catabolic state appeared around parturition.

Our results were in partially accordance with previous studies proving that gut microbiota changed greatly from first to third trimester of gestation with increased abundances of Proteobacteria and Actinobacteria (Koren et al., 2012; Gomez-Arango et al., 2016). Furthermore, the genus Escherichia belonging to Proteobacteria was only detected on d110 of gestation in the present study. Proteobacteria is a phylum known to encompass multiple pathogens and have pro-inflammatory properties (Mukhopadhya et al., 2012); in particular, some pathogenic Escherichia coli falling under the genus Escherichia are known to induce a series of inflammation responses. Moreover, a number of species from phylum Actinobacteria were associated with obesity (Turnbaugh et al., 2009). It is common recognized that the latter stage of gestation is characterized by a diabetogenic state in the mother to support the continuous supply of nutrients to the fetus, and an elevated levels of circulating proinflammatory cytokines (Mor and Cardenas, 2010). Increases of these two phyla provide us the insight into the microbial mechanism under which the mother performs some degree of inflammation during late gestation. Meanwhile, it confirms the previous conclusion that the gut microbial composition in the first trimester of pregnancy is similar to that of healthy, non-pregnant women (Nuriel-Ohayon et al., 2016), whereas that in the third trimester resembled an insulin resistance and greater inflammatory response–associated dysbiosis (Koren et al., 2012).

Fermentation of dietary fiber in the hindgut results in multiple groups of metabolites of which SCFAs are the major products. Acetic acid (C2), propionic acid (C3), and butyric acid (C4) are the most abundant, representing 90–95% of the SCFA present in the colon (Ríos-Covián et al., 2016). During the intestinal absorption process, butyrate, mainly acting as the energy source, is metabolized by the colonocytes whilst the rest will be transported to the liver and be used in different biosynthetic routes. Propionate will mainly act as a precursor for gluconeogenesis (Roy et al., 2006), acetate and butyrate will be mostly introduced into the lipid biosynthesis (den Besten et al., 2013b; Ríos-Covián et al., 2016). In addition to being as substrates, SCFA can act as signaling molecules to be sensed by specific G protein –coupled receptors (GPRs) and involve in the regulation of lipid and glucose metabolism (den Besten et al., 2013a). All these contribute to the central role of SCFA in the diet–gut microbiome-host metabolism axis. Due to the fact that acetate is produced by most enteric bacteria as a product of carbohydrate fermentation, it is the most abundant fecal VFA and makes up more than 50% of the total detected in feces in our study which is in good agreement with Louis et al. (2007). The linear decreases of individual and total VFA concentrations from d30 to d90 of gestation were partially in line with those reported by Kong et al. (2016) who found acetate and total SCFAs have a tendency to decrease as the pregnancy progressed, however, contradictory results with remarkable increases on d110 of gestation were also found in our study. This might imply a greater amount of energy loss in feces at perinatal period as suggested by Koren et al. (2012) who did find significant increase in stool energy content in T3 compared with T1 trimester. The possible explanation for the reduced metabolite absorption would be the potentially decreased GIT contractility. Studies had suggested the progressively rises in pregnancy hormones such as progesterone and abdominal compression from the enlarging gravid uterus during gestation would lead to decreased GIT contractility (Body and Christie, 2016). In comparison to a growing body of evidence indicating that high fiber–LFD are characterized by the presence of higher amounts of fecal SCFA than diets with lower fiber content (De Filippo et al., 2010; Cuervo et al., 2013; Heinritz et al., 2016), our results align with those obtained by Paßlack et al. (2015) who did not detect any significant effects of inulin inclusion on total SCFA and its constituents in feces of sows during gestation and lactation. Furthermore, Nyman (2002) reviewed that inulin-type fructans generally does not lead to a significant increase of fecal SCFA or to a change in molar proportions of acetate, propionate, and butyrate, which can be explained by their very efficient colonic absorption in humans with <5% being excreted in feces (Raninen et al., 2011). It should be mentioned here that the lacked effects of inulin addition on the fecal VFA concentrations would question whether stool SCFA output is a suitable proxy for luminal SCFA production as introduced by Vogt and Wolever (2003). Previously, fecal SCFA concentrations had been suggested to be more suitable to reflect the overall net production and absorption of SCFAs in the GIT, rather than the production or the absorption of SCFAs per se (Vogt and Wolever, 2003; Montoya et al., 2016). To evaluate the potential production of VFAs, abundances of VFA producing-related genera were studied. Despite only the butyrate-producing genus Eubacterium-hallii-group was significantly increased by inulin addition, an overwhelming majority of presented genera were numerically increased or showed tendency to increase. With respect to the effect of gestation stage, propionate-producing genera and butyrate-producing genera showed the opposite way with the former increasing and the latter decreasing from d90 of gestation to perinatal period. The mechanism underlying the discrepancy between the abundance changes of propionate- and butyrate-producing genera is not clear yet, but may be related to the fate of these two VFAs in the host metabolism at perinatal period.

In conclusion, the present study suggested that improved microbial dysbiosis triggered by soluble dietary fiber inulin addition during gestation would be the potential mechanism underlying the positive effect of inulin on gestational weight gain regulation, maternal anti-inflammation and adipose tissue derived hormone production, as well as the lower neonatal average BMI and the lower risk of relatively “obese” newborns. Dramatical changes of energy metabolism- and inflammation-related microbial communities and their metabolites VFAs indicated a catabolic and inflammatory state under which the mother was undergoing at perinatal period. Given that surprisingly little understanding about the maternal microbial changes and corresponding metabolic improvement induced by soluble DF supplementation during gestation exists at present, special focus should be given when further studies are to be conducted. Application of soluble DF inulin in gestational diet would be a fruitful area for prevention of diet induced excess gestational weight gain and related inflammation status and metabolic disturbances with remodeled microbial ecosystem.

LC, JL, YLin, SX, BF, ZF, and DW contributed to the experimental design. PZhou, YZ, YLi, and TG performed the animal experiments. PZhou, PZhang, and CJ executed the lab analysis. PZhou and YZ performed the statistical analysis and wrote the manuscript. JW and DW revised the manuscript. All authors read and approved the final manuscript.