Obesity during pregnancy results in maternal intestinal inflammation, placental hypoxia, and alters fetal glucose metabolism at mid-gestation

We investigated whether diet-induced changes in the maternal intestinal microbiota were associated with changes in bacterial metabolites and their receptors, intestinal inflammation, and placental inflammation at mid-gestation (E14.5) in female mice fed a control (17% kcal fat, n = 7) or a high-fat diet (HFD 60% kcal fat, n = 9; ad libitum) before and during pregnancy. Maternal diet-induced obesity (mDIO) resulted in a reduction in maternal fecal short-chain fatty acid producing Lachnospiraceae, lower cecal butyrate, intestinal antimicrobial peptide levels, and intestinal SCFA receptor Ffar3, Ffar2 and Hcar2 transcript levels. mDIO increased maternal intestinal pro-inflammatory NFκB activity, colonic CD3+ T cell number, and placental inflammation. Maternal obesity was associated with placental hypoxia, increased angiogenesis, and increased transcript levels of glucose and amino acid transporters. Maternal and fetal markers of gluconeogenic capacity were decreased in pregnancies complicated by obesity. We show that mDIO impairs bacterial metabolite signaling pathways in the mother at mid-gestation, which was associated with significant structural changes in placental blood vessels, likely as a result of placental hypoxia. It is likely that maternal intestinal changes contribute to adverse maternal and placental adaptations that, via alterations in fetal hepatic glucose handling, may impart increased risk of metabolic dysfunction in offspring.

. High-fat fed dams were heavier than controls at conception and throughout gestation and displayed increased adiposity, hyperglycemia, hyperleptinemia, hyperinsulinemia, and insulin resistance at E14.5. mDIO reduces intestinal butyrate and β-defensin 3 and is associated with changes in maternal intestinal barrier proteins. As mDIO was associated with a reduction in the relative abundance of intestinal taxa involved in the production of SCFA, we investigated whether this corresponded to reduced cecal SCFA levels. Consistent with a decrease in the relative abundance of butyrate producing taxa Clostridiales, Lachnospiraceae, and Ruminococcaceae, cecal butyrate levels were lower in high-fat fed dams compared to controls (Fig. 3A). Acetate, propionate, isobutyrate, isovalerate, pentanoate, and lactate cecal levels were not different between groups (data not shown). Maternal DIO altered intestinal transcript levels of SCFA receptors, Ffar3 (p interaction = 0.0199) and Hcar2 (p interaction = 0.0061) and reduced levels of Ffar2 (p diet = 0.0019) (Fig. 3B). Reductions in SCFAs receptor levels appeared to be predominantly located in the ileum and colon. A reduction in Ffar2 is consistent with a decrease in intestinal antimicrobial peptide, β defensin-3 levels (Fig. 3C) as others have shown that Ffar2 −/− mice display lower levels of antimicrobial peptides 28 .
Butyrate has been shown to promote mucin 2 production in the colon [29][30][31] and consistent with this, mRNA levels of the major secreted colonic Muc2 were decreased, (Fig. 3E) although Muc1 and Muc4 were similar between groups (Fig. 3E). Colonic Muc5ac levels were significantly higher in mDIO mothers compared to controls (Fig. 3E). Butyrate has been shown to mediate intestinal barrier function through transcriptional regulation of tight junction protein claudin-1 and by the induction of occludin and ZO-1 redistribution in the cellular membrane 32 . Maternal DIO reduced claudin (Cldn1 p diet = 0.047) and Zo-1 (p diet = 0.0013) transcript levels (although only different in the duodenum), while occludin (Ocln) levels were unchanged (Fig. 3D). These observations are consistent with those that we have previously observed in the maternal intestine at term (E18.5) gestation 20 , but , family (f), or genus (g) level classification within pregnant control (n = 5) and high-fat (n = 8) females. (B) Principle Coordinate Analysis using the Bray-Curtis dissimilarity metric showed a significant separation of intestinal microbial communities as a result of diet but not pregnancy in control (n = 5) and high-fat (n = 8) females. (C) Relative abundance of Clostridiales, Ruminococcus, Lachnospira, and Lactobacillus in each female prior to pregnancy in control (maroon, n = 5) and high-fat (green, n = 8) and at four time points during gestation; E0.5, E6.5, E10.5, and E14.5 in control (teal, n = 5) and high-fat (beige, n = 8) females. unlike our previously published data, we do not observe changes in maternal circulating levels of endotoxin with mDIO at E14.5 (Fig. 3F), suggesting that at midgestation, the level of maternal intestinal barrier compromise is not great enough to elicit increases in maternal serum endotoxin levels. mDIO is associated with increased intestinal NFκB activity and colonic CD3 + t cells. Compromised intestinal barrier function is often accompanied by local inflammation and dysfunction of innate immune responses 33 . Maternal intestinal NFκB activity was significantly higher with mDIO (p diet = 0.0075, Fig. 4A, control n = 6, mDIO n = 7) and thus we investigated key factors known to orchestrate intestinal pro-inflammatory signaling cascades in response to altered intestinal bacterial profiles 34,35 during obesity. Consistent with our previous data in pregnant mice at E18.5, we found modest reductions in transcript levels of Tlr2 (p diet = 0.0500) and Tlr4 (p interaction = 0.0080) in the maternal intestine at E14.5 with mDIO (Fig. 4B). Downstream signaling component Traf6 was unchanged and Nfkb (p diet = 0.0608), and Il6 (p diet = 0.0230), transcript levels were modestly reduced with mDIO (Fig. 4B). Il10 transcript levels were reduced in the jejunum with mDIO (p interaction = 0.0096) (Fig. 4B). These observations are consistent with our observation that maternal serum levels of cytokines and chemokines (Fig. 4C) were similar between groups.
SCFAs are known to promote intestinal homeostasis through the immunomodulation of immune cell activation 10 , through G-protein-coupled receptor activation to regulate T cell differentiation to dampen inflammation 14 . We observed a modest increase in CD3 + T cell number in the maternal colon with mDIO (p = 0.0555) which was associated with a modest increase in Cd3ε transcript levels (p = 0.0589; Fig. 4C). mDIO promotes pro-inflammatory TLR signaling and induces a pro-inflammatory macrophage phenotype in the placenta. To determine whether mDIO was associated with altered inflammatory signaling in the placenta, we investigated transcript levels of key pattern recognition receptors, their downstream signaling components, and pro-inflammatory cytokines. mDIO resulted in an increase in both male and female placental Tlr2 levels (p diet = 0.0247), although Tlr4 mRNA was unchanged (Fig. 5A). Both Traf6 (p diet = 0.0272) and Nfkb ((p diet = 0.0253, p sex = 0.0643) mRNA levels were increased in mDIO placentae and this increase was more pronounced in placental tissue associated with female fetuses (Fig. 5A). Tnf, Il6 and Il10 mRNA levels were significantly increased with mDIO (Fig. 5A), while their receptor levels Tnfr1a, Il6ra, and Il10ra were similar between groups (data not shown).
To identify the cell types involved in pro-inflammatory signaling in the placenta with mDIO, we investigated transcript levels of cellular markers and chemo-attractants known to regulate pro-inflammatory macrophage function. Maternal DIO resulted in an increase in transcript levels of macrophage maturity marker F4/80 (Adgre1) (p diet = 0.0372) and monocyte chemoattractant protein 1 (Ccl2) (p diet = 0.0173), as well as a modest increase in macrophage colony stimulating factor 1 (Csf1) (p diet = 0.0575) and Csf2r (p diet = 0.0497). Arginase 1 (Arg1) mRNA levels, typically expressed by immunomodulatory macrophages involved in tissue remodeling and repair, were modestly reduced with mDIO (p diet = 0.0625), although inducible nitric oxide synthase (Nos2) mRNA, expressed by pro-inflammatory macrophages, was unchanged, and the ratio of Nos2 to Arg1 was similar between groups (Fig. 5B). The number of F4/80 immunopositive macrophages was similar between groups (Fig. 5). mDIO promotes placental hypoxia and increases nutrient transporter transcript levels. Hypoxia 36 and increased vascularization 20,37 are previously reported placental outcomes as a result of mDIO. In the present study, carbonic anhydrase IX, a marker of hypoxia, was elevated in both male and female mDIO placentae (p diet and p sex = 0.001 and p interaction = 0.04; Fig. 6A), where males showed modestly higher levels compared to females (Fig. 6A). Immunoreactive protein levels of pro-angiogenic vascular endothelial growth factor (VEGF) were elevated in both male and female mDIO placentae, (Fig. 6B). Consistent with a hypoxic, pro-angiogenic microenvironment, immunoreactive protein levels of platelet endothelial cell adhesion molecule 1 (CD31), an endothelial cell marker, were higher in both male and female mDIO placentae (Fig. 6C). We also investigated blood vessel maturity by immunolocalizing α-smooth muscle actin (α-SMA), a marker of placental pericytes. Interestingly, despite an apparent increase in placental CD31, both male and female placentae showed a reduction

Impacts of mDIO on maternal and fetal metabolic function.
Maternal adiposity is associated with impaired glucose homeostasis 38 , and obese women show both peripheral and hepatic insulin resistance, impaired insulin mediated hepatic glucose disposal 39 , and hyperlipidaemia 40 . We thus investigated the impacts of mDIO on principal enzymes known to regulate hepatic gluconeogenesis. Although maternal hepatic peroxisome proliferator activated receptor gamma 1 alpha (Pgc1a), pyruvate carboxylase (Pc), and glucose-6-phosphatase (G6pc) www.nature.com/scientificreports www.nature.com/scientificreports/ mRNA levels were not altered by mDIO, we found a significant reduction in transcript levels of cytosolic phosphoenolpyruvate carboxykinase (Pepck) and hepatic nuclear factor 4 alpha (Hnf4a), a known transcriptional activator of PEPCK in the maternal liver (Fig. 7A). The nuclear receptors liver X receptor (LXR)α and LXRβ are sensors of cholesterol metabolism and lipid biosynthesis and are known regulators of inflammatory cytokines, and suppress hepatic glucose production by inhibiting genes involved in gluconeogenesis including G6Pase (G6pc) and PEPCK (Pepck) 41,42 . LXRs are stimulated by fatty acid-induction of PPAR and by insulin 41 and thus we investigated whether reduced Pepck mRNA levels were the result of LXR gene induction. Both LXRα (Nr1h3) and LXRβ (Nr1h2) transcript levels were unchanged with mDIO ( Fig. 7B).
Maternal obesity has known impacts on offspring metabolism 43 which show a higher risk of insulin resistance 44,45 and fatty liver disease 46,47 , suggesting that the developing liver is vulnerable to mDIO exposure. We found that both male and female hepatic pyruvate carboxylase (Pc) and PEPCK (Pepck) transcripts were significantly reduced in fetuses exposed to maternal obesity (Fig. 8A). This may be due to LXR-mediated suppression as both LXRα (Nr1h3) and LXRβ (Nr1h2) transcripts in male and female high-fat fetuses were increased (Fig. 8B). Transcript levels of Pgc1α, G6pc, and Hnf4α were similar between groups (Fig. 8C).

Discussion
We have found that mDIO resulted in a reduction in the relative abundance of maternal SCFA producing gut microbial taxa, as well as corresponding decreases in maternal intestinal butyrate and SCFA receptor levels at mid-gestation in mice. mDIO resulted in increased maternal intestinal pro-inflammatory NFκB activity and an increased number of colonic CD3 + T cells, without associated changes in circulating soluble inflammatory factors. Furthermore, mDIO increased placental transcript levels of components of the TLR4 signaling  www.nature.com/scientificreports www.nature.com/scientificreports/ pathway, inflammatory cytokines, macrophage cell markers, chemo-attractants, and factors known to regulate TNF-producing macrophage function. Placental inflammation was accompanied by hypoxia, increased vessel density and decreased maturity, both of which may impact nutrient transport. These maternal and placental changes were accompanied by altered fetal hepatic glucose handling, namely decreased gluconeogenic capacity.
We 19,20 and others 17,48 have shown that pregnancy is associated with a shift in the types and abundance of commensal bacteria in the maternal intestine. In the current study, maternal obesity significantly changed the abundance of Firmicutes, Bacteroidetes, Verrucomicrobia and Tenericutes phyla, similar to observations in male mouse models of obesity 12,13 , and in obese humans 49,50 . Metagenomic analyses in obese male mice, showed an increase in the representation of genes encoding proteins responsible for the synthesis of SCFAs from the fermentation of dietary carbohydrates 12 , utilized by the host for hepatic lipogenesis 11 . Whether this also applies to maternal intestinal microbiota during pregnancy is not clear. At E18.5, we showed that the gut microbiota of high fat fed pregnant mice was significantly enriched in genes involved in fatty acid and sulfur-containing amino acid metabolism, and glycolysis and gluconeogenesis compared to control fed pregnant mice 19 .
SCFAs produced by anaerobic bacteria play a key role in regulating intestinal barrier integrity 51,52 but despite the widespread interest in SCFAs and GPRs and their role in obesity and inflammation, our understanding of their function during pregnancy remains unclear. In our study, obese dams had a decreased relative abundance of a number of bacterial genera belonging to the butyrate-producing families, and with this we observed a significant reduction in maternal caecal butyrate levels and decreased transcripts of SCFA receptors Ffar3, Ffar2 and Hcar2 and the mucosal protein Muc2. The extent to which these changes result in changes in maternal intestinal epithelial function is still unknown, although in other work we have observed an obesity-induced loss of maternal barrier integrity at E18.5 20 . Consistent with this observation, we show here that mDIO resulted in reductions in genes encoding tight junction proteins claudin-1 and ZO-1 in the maternal intestine, suggesting that during mDIO, SCFAs and their receptors may have lost their protective effects on maternal barrier function 51,52 . SCFA also regulate the production of intestinal antimicrobials via a GPR43-dependant pathway 28 . Here and in our previous work 19,20 we show maternal obesity can disrupt this homeostasis via reductions in microbial communities that produce SCFAs, as well as their receptors potentially negatively affecting barrier proteins, mucosal proteins, and antimicrobial peptides (β-defensin 3).
SCFAs produced by anaerobic bacteria also play a key role in regulating intestinal immunity 53 . Colonic Treg cells regulate intestinal homeostasis and control inflammation by limiting proliferation of effector CD4 + T cells. We show that maternal obesity resulted in a modest, but significant increase in the number of colonic CD3 + T cells and intestinal NFκB activity in all sections of the maternal intestine. This did not appear to be associated with increased transcript levels of either TLR2 or TLR4 signaling pathways and was also not associated with overall maternal serum inflammatory status. There are contrasting reports regarding maternal obesity and circulating markers of inflammation 54 fueled by differences in species and in the timing of measures. We have previously reported that mDIO resulted in increased maternal serum LPS and TNF levels 20 at E18.5 in mice, without changes in other pro-inflammatory cytokines. Similar results have been reported in term gestation obese women 55 . It is possible that maternal obesity-induced inflammatory status changes over the course of pregnancy, but this must be more thoroughly investigated.
Although we did not observe maternal systemic inflammation in mDIO dams we did observe increased markers of placental inflammation. This is consistent with studies in humans [56][57][58] , non-human-primates 59 , and ovine placentae 60 . Our data are limited to transcript levels, and do not describe immune cell specific changes, although we did investigate immune cell markers. We found that transcript levels of macrophage recruitment factor Ccl2 and macrophage maturation marker Adgre1 were elevated by mDIO, along with increases in macrophage activating cytokine expression (Csf1) and modest elevation in myeloid specific receptors Csf1r and Csf2rα. While our samples are highly enriched for placental tissue, we cannot exclude the possibility that changes in the abundance of macrophage-related transcripts were derived from changes to decidual macrophage populations from remaining decidual tissue, where tissue resident macrophages reside. Our ongoing studies will determine immunophenotypic changes to specific cells using more targeted methods including flow cytometry.
Alternatively activated macrophages are key regulators of normal placental angiogenesis 61 and placental inflammation is strongly associated with impaired tissue function 60,62 . mDIO was associated with placental hypoxia and increased VEGF and CD31 immunostaining. Other studies show that obese human placenta have increased non-branching angiogenesis with increased VEGF immunostaining in non-villous cytotrophoblast and endothelial cells of intermediate and terminal villi, but decreased staining in syncytiotrophoblast 63 . Maternal obesity has been linked to placental vascular dysfunction 59,64,65 but the independent effects of obesity on placental vascular development still remain to be fully clarified. Indeed, others have reported decreased vessel density in the context of mDIO 66,67 using other surrogate endothelial markers. Placental hypervascularization has been described to occur in the context of hyperglycemia and hyperinsulinemia 68,69 , both of which are present in mDIO dams, but whether this is causative is unclear. Hypoxia is also a robust angiogenic stimulus capable of inducing changes to placental vasculature 42,70 . As such, further studies examining the relative contribution of these factors to placental development, and the etiology of placental hypoxia in the context of mDIO are warranted.
Placental adaptations to hypoxia likely fuel downstream changes in nutrient transport and our data are consistent with that of others 25,71 , showing that mDIO increased glucose (Slc2a1, Slc2a3) and amino acid (Slc38a2) transporter expression. This increase was accompanied by changes in markers of fetal glucose handling, where we observed a decrease in fetal gluconeogenic capacity that appears to be driven by increased fetal hepatic Nr1h3 transcripts, and corresponding reductions in Pepck and pyruvate carboxylase transcript levels. LXRs are known regulators of adipogenesis, but also directly inhibit hepatic gluconeogenesis, including Pepck expression 42 . Fetal exposure to a maternal low protein diet in mice similarly resulted in reduced LXRα and LXRβ mRNA levels with corresponding changes in LXR gene promoter methylation 72 , an effect that appears to persist into early postnatal life in rats 73 . Postnatal modifications in LXR regulation are proposed to remove LXR's brake on gluconeogenesis (2019) 9:17621 | https://doi.org/10.1038/s41598-019-54098-x www.nature.com/scientificreports www.nature.com/scientificreports/ and result in postnatal glucose dysregulation 73 . Although the investigation of metabolism in offspring was beyond the scope of the current study, we have previously shown that a maternal high fat diet results in postnatal offspring insulin resistance and obesity 74 . The present study extends these observations and suggests that already at midgestation, fetuses are likely making critical hepatic adaptations in response to mDIO.
Accompanying these changes in fetal gluconeogenic capacity, we observed similar changes in maternal gluconeogenic capacity in dams that were fed a high fat diet, although these effects were far more modest. Seminal work in rats has shown that maternal gluconeogenic enzymes, their activity and the rate of in vivo gluconeogenesis increases with pregnancy 75 , thus maintaining a high maternal to fetal glucose ratio, facilitating feto-placental glucose transfer. Our observed reduction in maternal hepatic Hnf4α suggest perhaps that elevated maternal insulin levels, due to mDIO, may be acting through Pgc1α / Hnf4α pathways to reduce transcription of gluconeogenic target genes including Pepck. Although we did not observe changes in Pgc1α transcript levels, compromised activity of PGC-1α and HNF4 α have been associated with the development of insulin resistance and Type 2 diabetes [76][77][78] . It is unknown whether these changes in transcript levels of hepatic gluconeogenic regulators in the maternal liver persist postpartum, but it is interesting to note that maternal obesity during pregnancy is a risk factor for gestational diabetes 79,80 and increased risk of postpartum Type 2 diabetes 80 . Future studies should set out to investigate not only the impacts of mDIO on the fetus and offspring but also on maternal metabolism in the postpartum period.
We recognize that our study is not without limitations. Due to the design of the study, we have only measured transcript levels of key markers, enzymes and signaling molecules. Although most of these are transcriptionally regulated, we are unable to comment on protein levels at the present time. Future studies should probe deeper into the impacts of mDIO on maternal hepatic glucose regulation and our fetal liver measures could be extended with measures of hepatic pyruvate, glycerol, and lipid content. We have previously published that mDIO results in reduced maternal intestinal barrier function 20 but despite this, full investigation of maternal intestinal permeability is required to understand whether gut barrier function changes with advancing gestational age. Finally, a thorough investigation of maternal immunophenotyping using flow cytometry would improve our understanding of mDIO-induced changes in maternal immunity and its relationship to intestinal and placental inflammation.

Conclusion
In conclusion, we show that maternal diet-induced obesity (mDIO) resulted in a reduction in the relative abundance of maternal short-chain fatty acid producing microbial genera, and corresponding decreases in maternal intestinal levels of the antimicrobial beta defensin, butyrate, and SCFA receptor levels at mid-gestation in mice. We found that mDIO increased maternal intestinal pro-inflammatory NFκB activity and increased number of colonic CD3 + T cells. These changes were accompanied by increased placental transcript levels of components of the TLR4 signaling pathway, inflammatory cytokines, macrophage cell markers, chemo-attractants, and factors known to regulate TNF-producing macrophages. Placental inflammation was associated with hypoxia, increased vessel density but decreased maturity, both of which may impact placental nutrient transport. These maternal and placental changes were accompanied by altered fetal hepatic gluconeogenic capacity. Maternal intestinal microbial metabolites and their receptors likely play a major role in mediating the impacts of maternal diet induced obesity. These maternal intestinal changes could contribute to adverse maternal and placental adaptations that, via alterations in fetal hepatic glucose handling, impart increased risk of metabolic dysfunction in the offspring.

Animal model. All animal procedures for this study were approved by the McMaster University Animal
Research Ethics Board (Animal Utilization Protocol 12-10-38) in accordance with the guidelines of the Canadian Council of Animal Care. Female C57BL/6 J mice were obtained from the Jackson Laboratory (Bar Harbor, ME, Strain 000664), maintained in standard mouse cages in the same room with a constant temperature of 25 °C and a 12-h light, 12-h dark cycle, and fed a control standard diet (17% kcal fat, 29% kcal protein, 54% kcal CHO, 3 kcal/g; Harlan 8640 Teklad 22/5 Rodent Diet) and provided water ad libitum. Following a 1-week acclimation period, 8-week-old females were randomly assigned to 1 of 2 dietary interventions: control diet (Control n = 10; 17% kcal fat, 29% kcal protein, 54% kcal carbohydrate, 3.40 kcal/g; Harlan 8640 Teklad 22/5 Rodent Diet) or a high-fat diet (High-fat n = 10, 60% kcal fat, 20% kcal protein, 20% kcal carbohydrate, 5.24 kcal/g, Research Diets Inc., D12492). Females were housed two per cage with ad libitum food and water and were maintained on their respective diets for a total of 6 weeks. Body weight and caloric intake were recorded weekly.
After 6 weeks of dietary intervention, females were co-housed with control fed 8-week-old C57BL/6 J males (Jackson Laboratory, Bar Harbor, ME, Strain 000664) overnight and pregnancy confirmed by the observation of a vaginal plug and designated as embryonic (E) day 0.5. Pregnant females were maintained on their respective diets and water ad libitum. Throughout gestation maternal weight and caloric intake were recorded and maternal fecal pellets were collected at four gestational time points; E0.5, E6.5, E10.5 and E14.5 ( Supplementary Fig. 1). Fecal pellets were stored at −80 °C until DNA extraction and sequencing was performed. Pregnant dams were sacrificed at late-mid gestational timepoint E14.5, at which time the murine placenta is fully developed. At E14.5, pregnant dams were fasted for 6 hours. Maternal body weight was recorded. Maternal whole blood glucose was measured using a commercial glucometer (Precision Xtra, Abbott Diabetes Inc.), a blood sample was collected, serum isolated, and stored at −80 °C until analyzed. Pregnant dams were sacrificed by cervical dislocation and gonadal fat pads were weighed as a proxy measure of maternal adiposity. Maternal cecal contents were collected, and sections of the maternal duodenum, jejunum, ileum, and colon were collected, flash frozen in liquid nitrogen and stored at −80 °C or fixed in 10% formalin. Placentae and fetuses were dissected, weights were recorded and whole placentae from half of each litter (randomly selected) were snap-frozen in liquid nitrogen or fixed in 4% paraformaldehyde. Fetal tails and livers were collected (to determine fetal sex and to investigate markers of fetal metabolic function respectively) and flash frozen in liquid nitrogen and stored at −80 °C until analyzed. All Scientific RepoRtS | (2019) 9:17621 | https://doi.org/10.1038/s41598-019-54098-x www.nature.com/scientificreports www.nature.com/scientificreports/ sample sizes in the study represent the dam as a biological replicate, with one female and one male fetus randomly selected from each litter for each tissue of interest.
Maternal microbiota profiling: DNA extraction and 16S rRNA gene sequencing. Genomic DNA was extracted from fecal samples (control n = 5, high-fat n = 8) as previously described 19 , with minor modifications: 0.2 g of fecal material was mechanically lysed in 800 μl of 200 mM NaPO 4 monobasic (pH 8) and 100 μl guanidinium thiocyanate-ethylenediaminetetraacetic-sarkosyl with 0.2 g of 0.1 mm glass beads (MoBio Laboratories Inc., Carlsbad, CA, USA). Additional enzymatic lysis was performed by adding 50 μl lysozyme (100 mg/ml) and 10 μl RNase A (10 mg/ml) and incubated for 1 hour at 37 °C, followed by the addition of 25 μl 25% sodium dodecyl sulfate, 25 μl proteinase K and 62.5 μl 5 M NaCl and was incubated for 1 hour at 65 °C. The resulting solution was centrifuged (12,000 g for 5 minutes). The supernatant was removed and combined with an equal volume of phenol:chloroform:isoamyl alcohol (25:24:1) in a new micro-centrifuge tube. The sample was vortexed and centrifuged (12,000 g for 10 minutes). The aqueous phase was removed and combined with 200 μl DNA binding buffer (Zymo Research, Irvine, CA, USA). This mixture was passed through a Clean and Concentrator-25 DNA column (Zymo Research, Irvine, CA, USA) according to kit directions, washed, and elution was performed with ultrapure water.
PCR amplification of the variable 3 (V3) region of the bacterial 16 S rRNA gene was performed on extracted DNA from each sample independently, as previously described 19 . Briefly, each reaction contained 5 pmol of primer, 200 mM of dNTPs, 1.5 μl 50 mM MgCl 2 , 2 μl of 10 mg/ml bovine serum albumin (irradiated with a transilluminator to eliminate contaminating DNA) and 0.25 μl Taq DNA polymerase (Life Technologies, Canada) for a total reaction volume of 50 μl. 341 F and 518 R rRNA gene primers were modified to include adapter sequences specific to the Illumina technology and 6-base pair barcodes were used to allow multiplexing of samples as described previously 81 . DNA products of the PCR amplification were subsequently sequenced using the Illumina MiSeq platform (2 × 150 bp) at the Farncombe Genomics Facility (McMaster University, Hamilton ON, Canada).
Sequence processing and data analyses. As previously described 20,82 , the sl1p pipeline was used to process resultant FASTQ files 83 . In this pipeline, Cutadapt 84 was used to trim primers and low quality bases, PANDASeq 85 to align paired-end reads, AbundantOTU+ 86 to group reads into Operational Taxonomic Units (OTUs) based on 97% similarity, and the RDP Classifier 87 as implemented in Quantitative Insights into Microbial Ecology (QIIME) 88 against the Feb 4 2011 release of the Greengenes reference database 89 to assign a taxonomy to each OTU. Any OTU not assigned to the bacterial domain was removed, as was any OTU to which only 1 sequence was assigned. Chimeric sequences were identified and removed with the usearch61 algorithm implemented in QIIME 90,91 . This processing resulted in a total of 7214096 reads (mean 11986 reads per sample; range: 60669-155790) and 2423 OTUs. All OTUs with a summed relative abundance of <1% across all samples were excluded. The data set used for analysis contained 3943882 reads (mean 110986 reads per sample; range 60669-155790) and 127 OTUs.
Analyses of these data were completed using R (v. Maternal cecal short chain fatty acids (SCFAs). Maternal cecal SCFA levels were quantified as described previously 20 with minor modifications, by the McMaster Regional Centre of Mass Spectrometry (MRCMS). Thirty mg of frozen cecal contents was acidified in a weight equivalent volume of 3.7% HCl. Ten μl of internal standards were added to each sample and SCFA extraction was performed by adding 500 μl of diethyl ether to each sample, and the resultant solution was vortexed for 15 minutes. After vortexing, 800 μl of diethyl ether cecal extract was transferred to a clean 1.5 ml centrifuge tube. A 60 μl aliquot of each diethyl ether cecal extract was derivatized with 20 μl of N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide (MTBSTFA; Sigma-Aldrich) in a chromatographic vial containing an insert. The resultant organic extract-MTBSTFA mixture was incubated at ambient temperature for 1 hour and SCFA levels were quantified using a gas chromatograph (Agilent 6890 N GC), coupled to a mass spectrometer detector (Agilent 5973 N Mass Selective Detector).
Reading was performed using a plate reader at a wavelength of 450 nm (BioTek ® , Oakville, Canada). www.nature.com/scientificreports www.nature.com/scientificreports/ was performed by the McMaster Immunology Research Centre John Mayberry Histology Facility with high pH antigen retrieval on the Leica Bond RX Rabbit protocol (1:150, Abcam, ab16669, RRID:AB_443425) and CD+ 3Tcells were counted in entire section, by an experimenter blinded to the study groups.
RNA extraction and cDNA synthesis. Total RNA was extracted from frozen maternal intestinal sections, liver, placentae, and fetal liver using TRIzol ® -Reagent following the manufacturer's instructions. Briefly, 45 mg of frozen tissue was homogenized in 900 μl TRIzol ® -Reagent (Ambion, Life Technologies) and 0.2 g of glass beads (Sigma-Aldrich) and centrifuged at 12000 g for 10 minutes at 4 °C. The supernatant was removed and added to 300 μl chloroform. The solution was thoroughly mixed, incubated at ambient temperature for 3 minutes, and centrifuged at 12000 g for 10 minutes at 4 °C. The aqueous phase was removed, combined with 500 μl of isopropanol. The resultant solution was mixed thoroughly, incubated for 20 minutes at ambient temperature, and were centrifuged at 12000 g for 10 minutes at 4 °C. Total RNA was purified using RNeasy columns (Qiagen RNeasy ® Mini Kit) according to the manufacturer's instructions. RNA was eluted from the RNeasy column following a 15-minute incubation with 20 μl of RNase free water heated to 55 °C. RNA quantity and quality were assessed using a NanoDrop 2000/2000c spectrophotometer (Thermo Scientific). For each sample, RNA concentration and integrity was >1.7. All RNA samples were stored at −80 °C until first strand complementary DNA (cDNA) synthesis. Following extraction, 2 μg of RNA was used for cDNA synthesis using the SuperScript ® VILO ™ cDNA synthesis kit (Life Technologies) according to the manufacturer's instructions. Each reaction consisted of 4 μl of 5X VILO ™ Reaction Mix, 2 μl of 10X SuperScript ® Enzyme Mix, 4 μl of 250 ng/μl RNA and 12 μl of ultrapure water for a total reaction volume of 20 μl. The reactions were incubated in a thermocycler (Bio-Rad Laboratories, C1000 Touch) under the following cycling conditions: 25 °C for 10 minutes, 42 °C for 1 hour, and 85 °C for 5 minutes. Complementary DNA was stored at −20 °C until qPCR assays were performed.
Quantitative polymerase chain reaction (qPCR). To quantify transcript levels, quantitative PCR assays were performed using the LightCycler 480 II (Roche Diagnostics, Canada) and LightCycler 480 SYBR Green I Master Mix (Roche Diagnostics, Canada, 04887352001). Primers were designed using Primer-BLAST software available at NCBI (website: blast.ncbi.nlm.nih.gov) and manufactured by Life Technologies. Primer sequences are listed in Supplemental Table 2. The following PCR cycling conditions were used for each assay: enzyme activation at 95 °C for 5 minutes, amplification of the gene product through 40 successive cycles of 95 °C for 10 seconds, 60 °C for 10 seconds, and 72 °C for 10 seconds. Specificity of primers was tested by dissociation analysis and only primer sets producing a single peak were used. Each plate for a gene of interest contained a 10-fold serial dilution standard curve, generated using pooled cDNA. Each sample and standard curve point was run in triplicate. The crossing point (Cp) of each well was determined via the second derivative maximum method using LightCycler 480 Software, Release 1.5.1.62 (Roche Diagnostics, Canada). An arbitrary concentration was assigned to each well based on the standard curve Cp values by the software. The geometric mean of 2 or 3 housekeeping genes was used as reference gene value for each sample; β-Actin (Actb), Cyclophilin (Ppia) and Hypoxanthine phosphoribosyltransferase (Hprt), β2 microglobulin (B2M), and Non-POU domain containing octamer (Nono). Relative mRNA levels for each sample were determined by dividing the geometric mean of the triplicate for each sample by that sample's reference gene value.