Advancements in the past decade in next-generation sequencing and associated bioinformatics analyses have facilitated a more in-depth study of the human gut ‘microbiome’; a word coined to describe the overall community of micro-organisms in the gastrointestinal tract(Reference Aagaard, Petrosino and Keitel1). Links between the microbiome and many physiological conditions of the associated host have been made(Reference Serino, S and Trabelsi2–Reference Majnik and Lane4). The various components contributing and modulating the microbiome are yet to be truly defined; however, environmental factors such as lifestyle and diet have come to the fore(Reference Hemalatha5,Reference Rothschild, Weissbrod and Barkan6) .
Diet and dietary patterns have been shown to rapidly alter microbial diversity and in turn influence host physiology(Reference Laitinen, Collado and Isolauri7,Reference Sheflin, Melby and Carbonero8) . In non-pregnant cohorts, the dietary macronutrients fat and fibre have most commonly been demonstrated to be able to cause a shift in microbial diversity, with fibre consumption associated with beneficial effects(Reference Chu, Antony and Ma9–Reference Valdes, Walter and Segal11).
With respect to dietary patterns, the Mediterranean diet, the Western diet, low-fat and high-fibre diets have been examined in greatest detail, with some research showing Western diet to influence the gut microbiome more considerably than BMI(Reference Sheflin, Melby and Carbonero8,Reference Davis, Yadav and Barrow12) . Diets high in fibre have been shown to have the ability to increase the relevant abundance of SCFA-producing bacteria(Reference LeBlanc, Chain and Martín13). This is in contrast to diets rich in animal fats, high in saturated fat and protein, which have been shown to have a negative impact(Reference Singh, Chang and Yan14). The blueprint for the optimal gut microbiome is still unknown, but the negative association of decreased diversity is commonly observed. Decreased diversity is linked to a phenomenon called dysbiosis (a disruption of normal gut microbiota); diversity is involved in the survival and adaptability of any ecosystem, the microbiome being no exception(Reference Cardinale, Palmer and Collins15). Furthermore, diets such as the Western diet are associated with decreased microbial diversity(Reference Davis, Yadav and Barrow12,Reference Wu, Chen and Hoffmann16) .
Diversity is the method used to assess the gut microbiome. α Diversity (also described as the intra-personal variation) is the individual’s diversity in the microbiota. It has been suggested that a higher α diversity correlates with a healthier microbiome(Reference Kennedy, Naeem and Howe17,Reference Manichanh, Rigottier-Gois and Bonnaud18) . As for many ecosystems, a high species diversity is linked with greater resistance to dysbiosis (disruption of microbiota composition from outside normal ranges) and an overall health within the host(Reference Keesing, Belden and Daszak19).
β Diversity on the other hand describes the interpersonal variation of microbial composition and can be based on collapsing all microbial data to a single coordinate point and measuring the distance (using various metrics, e.g. Bray-Curtis, unweighted and weighted UniFrac, Euclidean) between this point and another, usually another participant, person or collection site.
In pregnancy, the gut microbiome is thought to be dynamic with a change seen in first trimester diversity compared with that of the third trimester(Reference Koren, Goodrich and Cullender20). Mode of delivery, pre-term birth, breast-feeding and maternal diet have been identified as important factors that directly influence the composition of the neonatal gut microbiota(Reference Chu, Valentine and Seferovic21). Likewise, the presence of furry pets in the home has been shown to influence the composition of the gut microbiota of newborns(Reference Tapiainen, Paalanne and Tejesvi22).
There is limited literature examining the association between maternal macronutrient and micronutrient intake and infant and maternal gut microbiome. Without this knowledge, it is impossible to develop a therapeutic use of dietary manipulation to modulate the microbiome and in turn lead to improvements in infant and maternal health.
The aim of this systematic review was to summarise current evidence relating to the association between maternal diet in pregnancy and both the maternal and neonatal gut microbiome.
Methods
Protocol and registration
The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) Statement for reporting on systematic reviews was followed(Reference Liberati, Altman and Tetzlaff23). A search checklist of items to include methods, strategy, study selection process, a risk of bias tool and summary measures was used and reported.
Eligibility criteria
To be included in the review, the studies had to be observational or cross-sectional in design, subjects needed to be pregnant women and/or infants within the first 6 weeks postpartum. The study needed to include a formal dietary analysis during pregnancy and use a culture-independent sampling technique to assess the gut microbiome. Studies had to include details of ages, ethnicities and demographic characteristics of the women/infants. Studies that evaluated the effect of dietary supplementation or probiotic use only, without formal dietary assessment, were excluded, as book chapters, online abstracts and conference proceedings were not included. Articles had to be published in English, and no time restrictions was imposed.
Outcomes
The main outcomes examined in this review were the maternal or neonatal gut microbiome composition and diversity, as assessed by culture-independent sequencing techniques. These outcomes are expressed as microbial diversity in terms of both α (intra-individual variation) and β (inter-individual variation) diversity and relative abundance of specific microbes. Indices such as Shannon’s index, whole-tree phylogenetic diversity and Simpson’s index, which measure diversity within microbial communities or Unifrac distances, Bayesian models or principal component analysis, which measure diversity between microbial communities were included.
Information sources
The following five electronic databases were searched; MEDLINE (PubMed), Cochrane Library, Web of Science, CINAHL and Ovid. The last search was conducted on 7 October 2019.
Search
Search terms are as follows: human; antenatal; pregnant; pregnancy; maternal; microbiome; microbial; microbiota; microbe; gut bacteria; gut microbiome; nutrient; diet; nutrition; dietary.
Search terms were identified by initial scoping searches and then adjusted depending on the electronic database searched, to better match the key words and indexing terms of each database, and align with MeSH (Medical Subject Headings) terms.
Study selection
Summary measures
It was not possible to carry out a summary analysis or meta-analysis for this systematic review due to heterogeneity across the included studies. This included differences in stage of pregnancy of participants, the stool sample analysed, the dietary assessment tool used and the method of microbiota analysis. An overall description of individual results is therefore provided in the Results section, separated into two sections: maternal gut microbiota and neonatal gut microbiota.
Results
Identified articles were added to a reference manager software package (EndNote version 7.7.1), and duplicates removed. A new file was created minus the duplicates. Studies were then screened based on the study title. Papers were then excluded based on reading an abstract and its fitting of the defined population, comparison, intervention and outcome (PICO) terms. Abstracts were reviewed independently by two researchers (S. E. M and E. C. O’B.), and two individual spreadsheets were created with researchers’ final included abstracts. Full papers of said abstracts were reviewed independently by two researchers (S. E. M. and E. C. O’B.) and both parties selected final papers. Disagreements were resolved by a third party (F. M. M.). A flow chart created based on the PRISMA guidelines can be seen below (Fig. 1).
Fig. 1. Flow diagram of study selection. Flow diagram depicting each stage of the study identification process. PICO, population, comparison, intervention and outcome.
Study characteristics
The study characteristics are described in Table 1.
Table 1. Summary of results
NoMIC, Norwegian microflora study; RCT, randomised controlled trial; SPRING, Study of PRobiotics IN Gestational diabetes; Ow, overweight; Ob, obese; GDM, gestational diabetes mellitus; N/A, not applicable; V, vegetarian; C, control; DSQ, Dietary Screener Questionnaire; rRNA, ribosomal RNA; IDQ, Index of Dietary Quality; PcoA, principal component analysis; LefSe, linear discriminant analysis effect size; PD, phylogenetic diversity; OTU, operational taxonomic unit; PERMANOVA, permutational multivariate ANOVA; ACE, abundance-based coverage estimator; FDR, false discovery rate; C-section; Caesarean section; TEI, total energy intake.
Risk of bias in individual studies
The seven studies were assessed for risk of bias using the 2016 ROBINS-I (‘Risk Of Bias In Non-randomised Studies – of Interventions’) assessment tool(Reference Sterne, Hernan and Reeves31). The ROBINS-I consists of an assessment and a scoring algorithm that ranks studies with little, moderate or severe bias, on contact with the Cochrane Group; this was agreed to be the most suitable risk of bias tool. Three researchers (S. E. M., E. C. O’B. and D. F. B.) independently assessed the included articles.
Risk of bias assessment
All studies were subject to a varying level of bias due to the observational nature of the analysis and potential confounders. Four studies were found to be at serious risk of bias in at least one domain, with three studies at moderate risk of bias (Table 2). No study was judged to be at a critical risk of bias in any domain. Therefore, the seven studies were included in this review(Reference Chu, Antony and Ma24–Reference Laitinen and Mokkala30).
Table 2. ROBINS-I (Risk Of Bias In Non-randomised Studies – of Interventions) risk of bias results
Maternal diet and the maternal gut microbiota
The association between maternal diet and the maternal gut microbiome composition in pregnancy was investigated in five studies. All five studies reported that the maternal gut microbiome in pregnancy is influenced by maternal diet to varying degrees. In addition, specific macronutrients are associated with distinct bacterial compositions and relative abundances and can modulate, either positively or negatively, the diversity of the gut microbiome.
Three studies identified an association between dietary fat intake and gut microbiome composition(Reference Mandal, Godfrey and McDonald25,Reference Roytio, Mokkala and Vahlberg26,Reference Barrett, Gomez-Arango and Wilkinson28) . Two of these studies reported a negative correlation between α diversity and intakes of cholesterol(Reference Mandal, Godfrey and McDonald25), total fat and SFA(Reference Roytio, Mokkala and Vahlberg26). The third study(Reference Barrett, Gomez-Arango and Wilkinson28) reported a difference in β diversity, although α diversity did not differ. Furthermore, microbial composition differed by type of fat. Intakes of cholesterol and MUFA were associated with relative increases in Proteobacteria composition(Reference Mandal, Godfrey and McDonald25). In contrast, SFA intake was linked to relative decreases in this phylum and also negatively associated with the genus Roseburia (rho = −0·4, P = 0·038)(Reference Barrett, Gomez-Arango and Wilkinson28). The study by Barrett et al. (Reference Barrett, Gomez-Arango and Wilkinson28) compared the effect of a vegetarian diet v. omnivorous diet in early pregnancy on the maternal microbiome composition. Barrett et al. (Reference Barrett, Gomez-Arango and Wilkinson28) reported that women on the vegetarian diet had a higher intake of PUFA, of which, linoleic acid positively correlated with Holdemania (rho = 0·51, P = 0·006) and Roseburia (rho = 0·40, P = 0·04) abundance, but negatively with Collinsella (rho = −0·50, P = 0·009).
Four studies reported results on dietary carbohydrate intake and gut microbiome composition(Reference Roytio, Mokkala and Vahlberg26,Reference Barrett, Gomez-Arango and Wilkinson28–Reference Laitinen and Mokkala30) . Each of these studies reported that higher dietary fibre intakes were positively associated with increased gut microbiota diversity and richness. Moreover, similar associations between dietary fibre intake and relative abundance of specific bacteria were reported in three of these papers(Reference Barrett, Gomez-Arango and Wilkinson28–Reference Laitinen and Mokkala30). Higher fibre intakes were positively associated with increased relative abundances of Holdemania, Roseburia, Lachnospira and Coprococcus. In contrast, dietary fibre intake was negatively associated with relative Collinsella (Actinobacteria) and Sutterella (Proteobacteria) abundances.
The study by Mandal et al. (Reference Mandal, Godfrey and McDonald25) reported increased dietary intakes of fat-soluble vitamins, such as vitamin D and retinol are inversely correlated with α diversity. Vitamin D showed the strongest associations for both measures. For Shannon’s diversity, only vitamin D was significantly associated (−5·1 % change in diversity per unit increase in vitamin D intake, P < 0·001). The authors report that associations between dietary components and β diversity did not show any effects (UniFrac (weighted and unweighted; data not shown)). Furthermore, multiple regression modelling was used to assess associations between microbial composition and one standard deviation of nutrient intake for several dietary components. Vitamin D was associated with relative increases in Actinobacteria and Proteobacteria. Retinol was also associated with relative increases in Proteobacteria composition. Conversely, protein and vitamin E correlated with relative decreases in Proteobacteria.
Protein intake was collected and examined by all studies; however, significant findings were not seen(Reference Chu, Antony and Ma24–Reference Barrett, Gomez-Arango and Wilkinson28,Reference Sterne, Hernan and Reeves31) .
Maternal diet and the neonatal gut microbiome
Two studies investigated the effect of maternal diet in pregnancy on the neonatal gut microbiome. Both studies reported that maternal diet in pregnancy is associated with distinct changes in the neonatal gut microbiome.
Chu et al. (Reference Chu, Antony and Ma24) identified an association between maternal dietary fat intake and distinct changes in the neonatal gut microbiota, at birth and 4–6 weeks of age. Participants were grouped by extremes of dietary fat intake (1 sd greater or less than the cohort mean), to produce a high-fat maternal diet group (n 13, 43·1 % fat intake) and low-fat group (n 13, 24·4 % fat intake). Significant differences in neonatal microbiome clusters were detected between groups (principal component analysis unweighted UniFrac: P = 0·04). There was an inverse association between high-fat maternal diet and relative abundance of Bacteroides in neonatal stool at delivery, persisting at 6 weeks, whereas Enterococcus abundance was higher in the high-fat group at delivery only.
The study by Lundgren et al. (Reference Lundgren, Madan and Emond27) found that associations between maternal diet and the gut microbiome composition of infant stool samples differed by mode of delivery. Three distinct genera clusters were identified in vaginally born infants (cluster 1: Bifidobacterium; cluster 2: Streptococcus and Clostridium and cluster 3: Bacteroides). Through multinomial logistic regression, the odds of falling within cluster 2 were 2·73 times higher with each additional fruit serving per d. Furthermore, maternal fruit intake was negatively associated with the Bifidobacterium group. The clusters differed in infants delivered by Caesarean section (cluster 1: Bifidobacterium; cluster 2: high Clostridium, low Streptococcus and low Ruminococcus; cluster 3: high Enterobacteriaceae, Ruminococcus and Lachnospiraceae). In this sub-group, the analysis found a 2·36 increase in odds of being in a high Clostridium-low Streptococcus cluster with every increase of dairy portion. Maternal fish intake was positively associated with the Streptococcus genus in both groups of infants. In addition, red meat consumption was positively associated with the Bifidobacterium genus for the Caesarean section group. Likewise, the association between maternal alternative Mediterranean diet score differed slightly by mode of delivery, with positive associations existing with the Enterobacteriaceae family and the genus Streptococcus in the vaginally born group. In the Caesarean section group, a negative association was observed. Taking premature infants out of the analysis did not change results.
Discussion
Main findings in this study
Pregnancy is a unique time point during which improvement to the health of the woman can also benefit the immediate and long-term health of the child. Manipulating the gut microbiome during pregnancy may be beneficial to the health of both mother and baby(Reference Chu, Valentine and Seferovic21). Indeed, each of the studies included in this review demonstrates the important influence of maternal diet in pregnancy in modulating the gut microbiome of mother and infant, both beneficially and detrimentally. They provide evidence that diet quality, determined by factors including amount of fibre, fat, fat-soluble vitamins, fruit and vegetables, and fish and meat consumed, is associated with distinct gut microbiota profiles and diversity of the gut microbiota. Interestingly, the findings from Lundgren et al. (Reference Lundgren, Madan and Emond27) demonstrate that the influence of maternal diet on gut microbiota profiles differ by delivery mode.
The findings from this review align with those of the prevailing literature. Recent studies have shown the influence of diet and the gut–brain axis in the prenatal period, with the gut microbiome potentially playing a role in neurodevelopment(Reference Borre, O’Keeffe and Clarke32). In addition, diet has been shown to change the composition and metabolism of gut microbes(Reference Riaz Rajoka, Shi and Mehwish33). Fibre and to a lesser degree fat have been identified as important modulators of the human gut microbiome(Reference Rothschild, Weissbrod and Barkan6,Reference Valdes, Walter and Segal11) . It is estimated that approximately 20–60 g of undigested carbohydrate reaches the large intestine (the area with the highest density of gut microbes) daily(Reference Silvester, Englyst and Cummings34). This is larger than the amount of fat and protein that reach the colon, which are both readily digested in the upper gastrointestinal tract(Reference Scott, Gratz and Sheridan10), and thus are more likely to impact on the small intestinal microbiota. In high-fat diets (>35 % of total energy intake), a greater proportion of fat will reach the colon and it is hypothesised that this causes reduction of bacteria usually used for carbohydrate degradation, causing a shift in the microbiome as a whole(Reference Rowland, Gibson and Heinken35). In contrast, high-fibre diets (>25 g/d(36)) are associated with greater relative abundances of SCFA-producing bacteria (such as Holdemania and Roseburia) and relative depletion of lactate producers (such as Collinsella), with the former considered directly associated with beneficial metabolic profiles(Reference den Besten, van Eunen and Groen37).
In addition, probiotics have emerged as another promising means by which to manipulate the maternal gut microbiota with a view to improve health and clinical outcomes(Reference Valdes, Walter and Segal11). However, the research behind their use in pregnancy has not shown clear reduction of adverse outcomes such as preterm birth or secondary outcomes such as gestational diabetes or reduction in glucose level(Reference Lindsay, Brennan and Kennelly38,Reference Jarde, Lewis-Mikhael and Moayyedi39) . Jarde et al. (Reference Jarde, Lewis-Mikhael and Moayyedi39) conducted a systematic review with nineteen studies which found no definitive link between probiotic supplementation and improved clinical sequela. Likewise, Lindsay et al. (Reference Lindsay, Brennan and Kennelly38) examined the effect of probiotic supplementation on several important clinical outcomes including birth weight and fasting glucose, with no reported difference in those parameters. Further clarity is required regarding the clinical benefits of probiotic supplementation use during pregnancy. Hence, dietary manipulation of the maternal (and neonatal) gut microbiota may offer more readily available opportunities in the immediate term for improving the health of mother and child.
Environmental determinants have been demonstrated as important mediators of the human gut microbiota, including the shared home environment. Factors such as having other children at home, or having furry pet animals, have been shown to directly influence the composition of the maternal and neonatal gut microbiota(Reference Rothschild, Weissbrod and Barkan6,Reference Tun, Konya and Takaro40) . None of the studies in this review explored these variables.
Significant heterogeneity pervades multiple domains of the studies included in this review. Consequently, the findings of this review should be interpreted with caution and considered in the context of the wider literature. Four of the five studies focusing on maternal gut outcomes studied a cohort of women with overweight and obesity. Although this could be considered a representative sample in the context of rising overweight and obesity rates, a comprehensive well-designed study examining normal-weight and overweight/obese women in pregnancy, nutrients and the microbiome must be conducted first for comparison. BMI was self-reported by participants in the study by Lundgren et al. (Reference Lundgren, Madan and Emond27). It has been shown that self-reported BMI underestimates actual BMI in pregnancy(Reference Natamba, Sanchez and Gelaye41).
In addition, the method of dietary assessment varied considerably across the studies. Five studies assess diet by FFQ, one by 3-d food diaries, and one by Index of Diet Quality. Roytio et al. (Reference Roytio, Mokkala and Vahlberg26) used 3-d food diaries as well as providing participants with oral and written instruction and a portion picture booklet. This would allow for a more accurate correlation between diet and the microbiome. Of the five studies that employed FFQ, there were differences in the period of time assessed (from 4 to 16 weeks) and the time point in pregnancy it was administered (two in first trimester, two in second trimester and one in third). As pregnancy progresses, diet may vary considerably due to increased early satiety, reflux and constipation. There is also potential for misclassification of food groups using FFQ. In the Willett FFQ used in Lundgren et al. (Reference Lundgren, Madan and Emond27), fruit and fruit juices are both in the fruit food group. Fruit juices contain high amounts of free sugar and lower amounts of fibre, and therefore the effect on the gut microbiota could be considerably different(Reference Koutsos, Tuohy and Lovegrove42). Likewise, differences in the temperature at which collected stool samples were stored and the time point at which they were collected across the studies could influence the comparability of the results.
A major strength of this systematic review is the techniques used in the search strategy and the analysis of bias. The PRISMA guidelines recommended by the Cochrane Group were used(Reference Liberati, Altman and Tetzlaff23).
Another strength of this review is that all seven studies used culture-independent analytical techniques. The use of culture-specific sampling technique is now seen as a major risk of bias in the microbiological research. The benefit of culture-independent analytical techniques is that all microbial species present in the microbiome can be identified and therefore analysed(Reference Aagaard, Petrosino and Keitel1).
Future directions of studies
The examination of detailed dietary data in pregnancy and its influence on the microbiome must be conducted in detail in a cohort representative of a normal obstetric population. Without this, findings from subgroups are difficult to interpret. Dietary analysis should be conducted in a systematic manner. Food diaries most accurately capture intake within the last week and therefore may be most appropriate compared with FFQ that capture intake in the last few months. With this said, there is emerging evidence to suggest that long-term food patterns have a stronger role in the metabolism and composition of the human gut microbiome than short-term dietary changes(Reference Wu, Chen and Hoffmann16). Therefore, perhaps both FFQ and food diaries methodologies should be used for each analysis.
Conclusion
In summary, this review demonstrates the important influence of maternal diet in pregnancy in modulating the gut microbiome of mother and infant, both beneficially and detrimentally. The findings provide evidence that diet quality, determined by factors including amount of fibre, fat, fat-soluble vitamins, fruit and vegetables, and fish and meat consumed, is associated with distinct gut microbiota profiles and diversity of the gut microbiota. However, confidence in the quality of this evidence is limited due to methodological limitations within the studies, and variability between studies. Pregnancy is a unique time point during which benefits to the health of the mother can also benefit that of the child. Hence, further high-quality research is required in this area to elucidate the relationship between diet quality and the gut microbiota of mother and child.
Acknowledgements
This publication has emanated from research supported in part by a research grant from Science Foundation Ireland (SFI) under grant nos 12/RC/2273 and 16/SP/3827 and by a research grant from Alimentary Health Ltd.
R. S., E. F. M., D. V. S., P. D. C. and F. M. M. designed the research; S. E. M., A. A. G., R. L. M., E. C. O’B. and D. F. B. conducted the research; S. E. M., E. C. O’B. and D. F. B. analysed the data; S. E. M., E. C. O’B. and D. F. B. wrote the paper; F. M. M. had primary responsibility for final content. All authors read and approved the final manuscript.
Contents are the authors’ own view. E. F. M. is Technical Director at Alimentary Health Group. The authors have no other disclosures to declare.
Open access
