Potential interaction between timing of infant complementary feeding and breastfeeding duration in determination of early childhood gut microbiota composition and BMI

Summary

Background:

Introducing complementary foods other than breastmilk or formula acutely changes the infant gut microbiota composition. However, it is unknown whether the timing of introduction to complementary foods (early vs. late) in infancy is associated with early childhood gut microbiota and BMI, and if these associations depend on breastfeeding duration.

Objective:

Our primary objective was to investigate whether timing of infant complentary feeding with solid foods is associated with early childhood gut microbiota composition and BMI-z, and whether these associations differ by duration of breastfeeding.

Methods:

We used data from a Canadian pre-birth cohort followed till age 5 years. We examined timing of introduction to solid foods with the gut microbiota, determined by 16S rRNA gene sequencing of stool collected at 5 years of age, and age-and-sex specific BMI-z. We conducted analyses before and after stratifying by breastfeeding duration, and adjusted for delivery mode, gestational age and birth weight.

Results:

Of the 392 children in the analysis, 109 (27.8%) had early (≤4 months) solids. The association between early (vs later) solids and BMI-z at 5 years was modified by breastfeeding status at 4 months (P = .06). Among children breastfed >4 months, early (vs later) solids were associated with differential relative abundance of 6 bacterial taxa, including lower Roseburia, and 0.30 higher BMI-z (95% CI: 0.05, 0.55) at 5 years. In children breastfed <4 months, early solids were associated with differential relative abundance of 9 taxa, but not with child BMI-z.

Conclusions:

Early (vs. later) introduction to solid foods in infancy is associated with altered gut microbiota composition and BMI in early childhood, however these associations differ by duration of breastfeeding.

1 |. INTRODUCTION

The prevalence of early childhood overweight and obesity continues to rise, with an estimated 26% of US children ages 2–5 years living with overweight or obesity in 2015 and 2016.¹ Beyond portending adolescent and adult obesity,^2,3 having a greater body mass index (BMI) in childhood is associated with higher future risk of dysipidemia, hypertension, type 2 diabetes, and atherosclerosis.⁴ Given these statistics, and the intractable challenge of treating obesity once it manifests, there is need to identify modifiable early-life determinants of childhood overweight and obesity.

Several lines of evidence indicate that timing of introduction of complementary foods (ie, introducing foods other than breastmilk or formula) may play a role in the development of childhood overweight and obesity. A meta-analysis of observational studies found that complementary feeding before 4 months of age was associated with a 20% and 30% greater risk of childhood overweight and obesity, respectively.⁵ More recently, the Prevention and Incidence of Asthma and Mite Allergy (PIAMA) birth cohort study found that the association of early complementary foods with childhood overweight and obesity was modified by breastfeeding status at the time of solid food introduction.⁶ This finding is consistent with the Project Viva birth cohort, which reported that children who were not breastfeeding and given complementary solid foods before 4 months were 6.3 times more likely to be obese at 3 years of age.⁷ Another study by Butte et al, following infants from birth to 24 months of age, found that early complementary foods were associated with increased adiposity in infants who were breastfed for at least 4 months.⁸

It has been postulated that oxidative stress induced by early complementary feeding influences differential growth and metabolism of gut microbiota, eventually leading to immune-mediated and metabolic conditions such as obesity.⁹ Indeed, differences in the gut microbiota composition have been previously associated¹⁰ with both infant diet¹¹ and later obesity^12–14 in longitudinal studies. Yet, to our knowledge, no studies in humans have reported on the association of timing of introduction to complementary foods with measures of gut microbiota and adiposity in childhood. Identifying gut microbiota features associated with early complementary feeding could at once elucidate the underlying mechanism linking early introduction of complementary foods with childhood obesity, among other conditions associated with early food introduction, and provide an intervention target for prevention.

Given this premise and the aforementioned literature gaps, the primary objective of this study was to examine associations of early vs later introduction of complementary foods with the composition and diversity of gut microbiota in childhood. We also evaluated associations between timing of infant complementary feeding with child BMI z score (BMI-z). We examined these associations before and after stratifying on breastfeeding duration to confirm findings from previous studies on this topic. We hypothesized that early complementary food introduction is associated with altered early childhood gut microbiota composition and greater BMI-z, and that these associations are modified by whether the infant is on breastmilk when complementary foods are introduced.

2 |. RESEARCH DESIGN AND METHODS

2.1 |. Study population

We conducted our study within the prospective Canadian birth cohort, Genetics of Glucose regulation in Gestation and Growth (Gen3G). The Gen3G cohort was designed to elucidate the biological, environmental, genetic and epigenetic determinants of glucose regulation during pregnancy and the impact on offspring development.¹⁵ We recruited expecting mothers during their first trimester of pregnancy, enrolling them if they were at least 18 years of age with a singleton pregnancy and did not have pre-pregnancy diabetes based on medical history and screening during the first trimester blood sampling. We followed mothers throughout pregnancy and collected data and samples at delivery. We further followed up mother-child pairs at 3 and 5 years post-delivery to collect anthropometric data, questionnaires and biologic samples.

For this study, we asked the mother to collect stool samples from herself and her child at their 5-year follow-up exam for microbiota analyses, restricting to those that had not taken antibiotics in the last 3 months. Study participants provided written informed consent prior to enrolment in accordance with the Declaration of Helsinki. All study protocols were approved by the ethics review board from the Centre Hospitalier Universitaire de Sherbrooke (REB # 07–027-A1).

3 |. METHODS

3.1 |. Exposures—Infant feeding variables

Mothers self-reported the age at which they first introduced complementary foods to their children, whether they ever breastfed, how long they breastfed their child in months and if and when they introduced formula at both the 3-year visit and the 5-year visit. We defined early introduction to complementary foods as infants given any solid foods at or before 4 months of age, based on prior studies and clinical recommendations.^5,16–18 We categorized breastfeeding duration as any breastfeeding for at least 4 months vs breastfeeding for less than 4 months. We used data from the 3-year visit or data from the 5-year visit if data from the 3-year visit were not available.

3.2 |. Covariates—Early infant exposures and maternal characteristics

From electronic medical records, we abstracted data on delivery mode (vaginal vs Caesarean section), gestational age, birth weight, child sex and diet (breastfeeding vs formula) during hospital stay. Maternal smoking, pre-pregnancy body mass index (BMI; kg/m²), parity and ethnicity were determined by staff-administered questionnaires at the first prenatal study visit. If pre-pregnancy BMI was missing, we substituted with maternal BMI during the first trimester, which we calculated after direct measurement of weight (kg) and height (m) using a calibrated electronic scale and wall stadiometer, respectively.

3.3 |. Child anthropometry

Child weight (kg) and height (m) at 5 years, measured using a calibrated electronic scale and a wall stadiometer, respectively, were used to calculate child BMI (kg/m²). We used the World Health Organization’s AnthroPlus software to calculate sex- and age-specific child BMI z score.¹⁹

3.4 |. 16S rRNA sequencing

Full sample collection and sequencing methods are described in the Supplement (eMethods and Figures S1–S3 in Data S1). Briefly, mothers collected fresh stool from their children (n = 181) at 5 years of age using the OMNIgene•Gut OM-200 collection tube from DNA Genotek (Canada). We performed multiplex sequencing of the 16S V4 rRNA region using the Illumina MiSeq Personal Sequencing platform according to the protocol described by Kozich et al.²⁰ Sequences were processed into amplicon sequence variants (ASVs) using the DADA2 and phyloseq R packages according to a recently published workflow and package author recommendations.^21–23 We de-replicated and pooled forward and reverse reads before denoising and merging sequence pairs using default parameters.²² We identified and removed chimeras from merged sequences using the pooled option (86% of merged reads non-chimeric). This resulted in a final set of ASVs containing 78% of the raw reads (83% of filtered reads, as seen in Figure S3 in Data S1). We also sequenced technical controls with our samples to confirm the quality of our sequencing data (Figure S1A,B in Data S1).

3.5 |. ASV taxonomic assignment and phylogenetic tree generation

For taxonomic assignment of ASVs, we used the “assignTaxonomy” function with the HITdb v.1.00 16S rRNA sequence database designed for classification of human intestinal taxa.²⁴ Using recommendations described in a recently published workflow, we then aligned ASVs using the R package DECIPHER to construct a preliminary neighbour-joining tree, which we then used to calculate a Generalized time-reversible with Gamma rate variation (GTR) maximum likelihood phylogenetic tree using the R package phangorn.^21,25,26 The tree was rooted at the mid-point using phangorn.²⁵ We then combined the sample metadata, ASVs, taxonomy and phylogenetic tree using the R package phyloseq.²³

4 |. STATISTICAL ANALYSIS

4.1 |. Microbial community differential abundance analysis

We assessed the differential abundance of taxa in children using beta-binomial count regression models from the R package corncob, which accounts for within-sample taxa correlation and varying sequencing depth.²⁷ We filtered taxa by prevalence (≥10% of all samples) and mean relative abundance (≥20th percentile) before testing to prevent comparisons of rare taxa.

4.2 |. Diversity analyses

To examine microbial alpha diversity, we estimated ASV richness and Shannon diversity using the R packages Breakaway and phyloseq, respectively.^23,28,29 We tested for differences in Shannon diversity using generalized linear regression models (GLMs), and we used betta, a three-part hierarchical model, for estimating differences in ASV richness.²⁹ For beta diversity (pairwise community composition), we estimated weighted UniFrac distances using the R package phyloseq,²³ and tested for differences using permutational multivariate analysis of variance (PERMANOVA) with 9999 permutations.³⁰ Further description of diversity analyses can be found the Supplemental Methods.

4.3 |. Child BMI z score regression models

For all 392 children with measurement of height and weight at the 5-year follow-up visit, along with complete data on pertinent covariates, we used GLMs to estimate the differences in child BMI z score associated with timing of introduction to complementary foods (≤4 months vs >4 months of age) before and after adjustment for covariates (described below).

4.4 |. Multi-variable models

For all anthropometric and microbiota outcomes, we first ran univariable models for timing of introduction to complementary foods (≤4 months vs >4 months of age). We then ran a multi-variable model that included adjustment for breastfeeding duration (breastfed ≥4 months vs <4 months of age), delivery mode, gestational age and birth weight. We included breastfeeding duration, delivery mode, gestational age and birth weight as potential confounders because they were associated with our exposure and outcome, but not on the causal pathway. To evaluate if breastfeeding modified the association of timing of solids with outcomes, we conducted all univariable and multi-variable analyses before and after stratification by breastfeeding status at 4 months.

We defined statistical significance as P < .05 for the outcomes BMI-z, alpha diversity and beta diversity, and P < .10 for tests of interaction. We used a two-sided false discovery rate (FDR) adjusted P < .05 for comparisons of differential abundance of microbial taxa to account for multiple testing.

5 |. RESULTS

5.1 |. Participant characteristics

Of the 392 children with complete data on timing of introduction to complementary foods, breastfeeding duration and child BMI-z at 5 years, 109 (27.8%) were introduced to complementary foods ≤4 months of age. On average, children introduced to complementary foods early stopped any breastfeeding earlier and had a slightly higher proportion of mothers who smoked during pregnancy. Table 1 shows the characteristics of mother-child pairs according to timing of complementary food introduction and breastfeeding status at 4 months of age. Of the 141 children who were not breastfed at least 4 months, 57 never initiated breastfeeding. Baseline characteristics for the 180 children with complete data on the gut microbiota at 5 years of age, timing of complementary foods and breastfeeding duration are shown in Table S1 in Data S1.

TABLE 1.

Gen3G maternal and child characteristics by timing of complementary food introduction, stratified by breastfeeding duration (child n = 392)

	Breastfed <4 mo			Breastfed ≥4 mo
	Introduction to complementary foods ≤4 mo (n = 49)	Introduction to complementary foods >4 mo (n = 92)	P	Introduction to complementary foods ≤4 mo (n = 60)	Introduction to complementary foods >4 mo (n = 191)	P
Pre-pregnancy BMI, kg/m2, median [IQR]	23.95 [21.20, 26.57]	23.10 [20.85, 27.60]	.63	23.35 [20.88, 28.67]	22.80 [20.78, 25.80]	.44
Maternal delivery age, median [IQR]	28.97 [23.89, 31.61]	30.02 [26.58, 32.17]	.058	28.97 [26.12, 31.66]	29.75 [27.20, 32.47]	.23
Smoked during pregnancy, n, %	9 (18.75)	7 (7.61)	.092	4 (6.67)	7 (3.76)	.56
Primiparous, n, %	22 (44.90)	43 (46.74)	.98	32 (53.33)	95 (49.74)	.74
Delivery mode, n, %			.50			.99
C-Section	11 (22.45)	15 (16.30)		10 (16.67)	30 (15.71)
Vaginal	38 (77.55)	77 (83.70)		50 (83.33)	161 (84.29)
Female sex, n (%)	19 (38.78)	52 (56.52)	.067	27 (45.00)	94 (49.21)	.67
White ethnicity, n, %	47 (95.92)	88 (95.65)	.99	56 (93.33)	186 (97.89)	.18
Gestational age, week, median [IQR]	39.20 [38.00, 40.10]	39.50 [38.50, 40.20]	.19	39.35 [38.20, 40.30]	39.30 [38.60, 40.20]	.39
Birth weight, g, mean (SD)	3385.04 (460.82)	3323.04 (531.69)	.49	3423.62 (517.89)	3420.18 (438.43)	.96
Breastfed ever, n, %			.64			NA
No	18 (36.73)	39 (42.39)		0 (0.00)	0 (0.00)
Yes	31 (63.27)	53 (57.61)		60 (100.00)	191 (100.00)
Breastfeeding duration, months, median [IQR]	0.50 [0.00, 2.00]	0.38 [0.00, 2.00]	.70	8.00 [6.00, 11.00]	10.00 [7.00, 13.00]	<.01
Introduction to complementary foods, months, median [IQR]	4.00 [3.00, 4.00]	6.00 [5.00, 6.00]	<.001	4.00 [3.88, 4.00]	6.00 [5.50, 6.00]	<.01

5.2 |. BMI Z score

In Figure 1, we show estimated BMI-z means according to timing of complementary food introduction. Overall, in multi-variable-adjusted models, early (vs late) introduction to foods was not associated with differences in BMI-z (beta = 0.13, 95% CI: −0.07, 0.32). However, there was suggestive evidence of effect measure modification by breastfeeding status at 4 months on the association of timing of complementary foods with BMI-z (P for interaction = 0.06). Among children who were breastfeeding at 4 months, early introduction to other foods was associated with a 0.30 (95% CI: 0.05, 0.55) higher BMI-z at age 5. Among those not breastfeeding at 4 months, early introduction to other foods was non-significantly associated with BMI-z at 5 years of age (−0.14 [95% CI: −0.46, 0.19]).

FIGURE 1.

Adjusted means (with 95% confidence intervals) of BMI-z at 5 years of age for children introduced to complementary foods earlier in infancy (≤4 months, in blue) vs later in infancy (>4 months; referent group, in red), before (top panels) and after (bottom panels) stratification by breastfeeding status at 4 months of age (P for interaction = .06). Multi-variable models adjusted for delivery mode, gestational age, birthweight and breastfeeding status at 4 months (only in the full cohort model, with breastfeeding ≥4 months as the referent group, in yellow). *=P < .05

5.3 |. Gut microbiota alpha diversity

Figure 2 illustrates the mean and 95% confidence interval for the estimated difference in alpha diversity of stool collected at 5 years of age from children introduced to complementary foods at or before 4 months (early) vs later (reference), before and after stratification by breastfeeding duration. There was no evidence that breastfeeding status at 4 months modified the association of timing of complementary foods and Shannon diversity (P for interaction = 0.96) or ASV richness (P for interaction = 0.89). After multivariable adjustment, early introduction to complementary foods was not associated with Shannon diversity before stratification by breastfeeding status (beta = 0.14, 95% CI: −0.04, 0.33). Early introduction to complementary foods was also not associated with gut microbiota Shannon diversity among children who were not breastfeeding at 4 months (beta = 0.09, 95% CI: −0.25, 0.42), nor among children still breastfeeding at 4 months (beta = 0.16, 95% CI: −0.06, 0.39). Early introduction to complementary foods was not associated with differences in ASV richness in the full cohort (beta = −1.8, 95% CI: −19.4, 15.8) or within breastfeeding strata (breastfed ≥4 months beta = −2.2, 95% CI: −26.0, 21.6; breastfed <4 months beta = −6.7, 95% CI: −29.6, 16.2).

FIGURE 2.

Adjusted means (with 95% confidence intervals) of gut microbiota alpha diversity (measured by Shannon diversity) in 5-year old children that were introduced to complementary foods earlier in infancy (≤4 months) vs later in infancy (>4 months; referent group), before (top panels) and after (bottom panels) stratification by breastfeeding status at 4 months of age (P for interaction = 0.96). Multivariable models adjusted for delivery mode, gestational age, birthweight and breastfeeding status at 4 months (only in the full cohort model, with breastfeeding ≥4 months as the referent group, in yellow). For Shannon diversity models, we categorized birth weight and gestational age at their median value to allow model convergence.

5.4 |. Gut microbiota beta diversity

The timing of complementary food introduction was not associated with overall microbial community structure nor in either of the breastfeeding strata at 5 years of age (P for interaction = 0.98), as determined visually by Weighted UniFrac PCoA plots (Figure S2 in Data S1) and statistically [(a) PERMANOVA for overall cohort comparison, P value = .45; (b) among children still breastfeeding at 4 months (P = .37); (c) among children not breastfeeding at 4 months (P = .94)].

5.5 |. Gut microbiota composition

Timing of complementary food introduction was associated with differences in the log odds of relative abundance of several gut microbiota ASVs before (univariable associations shown in Figure S5.1 and S5.2 in Data S1) and after multi-variable adjustment at 5 years of age (Figure 3). After multivariable adjustment, early vs late introduction of complementary foods was significantly associated with the differential abundance of 7 taxa (Figure 3 and Table S2 in Data S1, FDR-adjusted P < .05). Taxa tested were present in at least 10% of all samples (≥18 children). Children with early introduction to complementary foods had higher mean abundance of Lachnospiraceae Blautia (sp. unknown), Porphyromonadaceae Barnesiella intestinihominis, Streptococcaceae Lactococcus lactis, Veillonellaceae Veillonella (sp. unknown), and Lachnospiraceae Anaerostipes (sp. unknown), while later introduction to complementary foods was associated with higher abundance of Lachnospiraceae (genus and sp. unknown) and Porphyromonadaceae Barnesiella intestinihominis. After multivariable adjustment, and independent of timing of complementary foods, breastfeeding duration (≥4 months vs <4 months) was associated with the differential abundance of 8 taxa (Figure 3 and Table S2 in Data S1 in 7 taxa , FDR-adjusted P < .05). Children who were breastfed less than 4 months had higher mean abundance of Porphyromonadaceae Barnesiella intestinihominis and Lachnospiraceae Blautia faecis while children breastfed a longer duration had higher abundance of Lachnospiraceae Lachnoclostridium (sp. unknown), Lachnospiraceae Lachnoclostridium scindens, Lachnospiraceae Blautia (sp. unknown), two Melainabacteriaceae Melainabacter (sp. unknown) ASVs and Erysipelotrichaceae Erysipelatoclostridium (sp. unknown).

FIGURE 3.

Estimated differences in the log odds of relative abundance of gut microbial amplicon sequence variants (ASVs) in 5-year old children that were not breastfeeding at 4 months of age (<4 months) vs those still breastfeeding (≥4 months; referent group); and that were introduced to complementary foods (CF) earlier in infancy (≤4 months) vs later in infancy (>4 months; referent group). Model further adjusted for delivery mode, gestational age, and birthweight. The dashed line indicates the chosen FDR threshold of 0.05 for statistical significance, at which level all ASVs in colour were significant

Given the evidence of effect measure modification by breastfeeding duration on the association of timing of complementary foods with BMI-z, we next present associations of early introduction of foods with gut microbiota composition according to strata of breastfeeding duration. Among children who were not breastfed at 4 months, the timing of complementary food introduction was significantly associated with the differential abundance of 9 taxa after multi-variable adjustment (Figure 4 and Table S3 in Data S1, FDR-adjusted P < .05): children with early introduction to foods had higher mean abundance of Ruminococcaceae Ruminococcus bromii, Lachnospiraceae (genus and sp. unknown) and Prevotellaceae Para-prevotella clara, while later introduction to foods was associated with higher abundance of Rikenellaceae Alistipes (sp. unknown), Sutterellaceae Parasutterella excrementihominis, Lachnospiraceae Dorea (sp. unknown), Lachnospiraceae Lachnoclostridium (sp. unknown), Sutterellaceae Sutterella (sp. unknown) and Bifidobacteriaceae Bifidobacterium animalis. Among children who were breastfeeding at 4 months, the timing of food introduction was significantly associated with the differential abundance of 6 taxa after multi-variable adjustment (Figure 3, FDR-adjusted P < .05): children with early introduction to foods had higher mean abundance of Lachnospiraceae Blautia (sp. unknown), Streptococcaceae Lactococcus lactis and Veillonellaceae Veillonella (sp. unknown), while later introduction to foods was associated with higher abundance of Clostridiaceae (genus and sp. unknown), Lachnospiraceae Roseburia (sp. unknown) and Porphyromonadaceae Barnesiella intestinihominis.

FIGURE 4.

Estimated differences in the log odds of relative abundance of gut microbial amplicon sequence variants (ASVs) in 5-year old children that were introduced to complementary foods (CF) earlier in infancy (≤4 months) vs later in infancy (>4 months; referent group) after stratification by breastfeeding status at 4 months of age. Models adjusted for delivery mode, gestational age, and birthweight. The dashed line indicates the chosen FDR threshold of 0.05 for statistical significance, at which level all ASVs in colour were significant

6 |. DISCUSSION

In this Canadian pre-birth cohort of mother-child dyads, early introduction to complementary foods in infancy, while breastfeeding or not, was associated with differential abundance of several metabolically active microbial taxa in the gut microbiota of children at 5 years of age. Early infant complementary feeding was also associated with higher childhood BMI-z, but only among children that were breastfed when early complementary foods were introduced. We summarized our main results in Figure 5.

FIGURE 5.

A graphical summary of gut microbiota composition (differential relative abundance), gut microbiota diversity (Shannon diversity) and BMI z score results in 5-year old children that were introduced to complementary foods (CF) earlier in infancy (≤4 months) vs later in infancy (>4 months; referent group), after adjustment for delivery mode, gestational age and birthweight

To our knowledge, our study is the first to report on the association of timing of infant introduction to complementary foods with the gut microbiota composition in childhood. The European-based INFABIO project (n = 605) demonstrated that the infant microbiota composition changes from pre-weaning (6 weeks of age) to post-weaning (4 weeks after introduction of solids).¹¹ The TEDDY birth cohort study reported that solid foods and breast milk explained the greatest amount of variation in the gut microbiota of infants at 7–10 months (the time in which there was the greatest variation in exposure to solids and breastfeeding).³¹ However, similar to the INFABIO project, the TEDDY birth cohort did not examine whether the timing of infant introduction to solids was prospectively associated with changes to the gut microbiota composition. We recently found that early complementary feeding is associated with the infant gut microbiota at 3 and 12 months of age.³² Our current findings suggest that the impact of early complementary foods on the gut microbiota composition may extend into childhood.

With respect to our findings on child BMI, the PIAMA and Project Viva birth cohorts also found that breastfeeding duration was an effect measure modifier of the association between the timing of infant solids and childhood overweight and obesity.^6,7 However, in these cohorts, there was a stronger association among non-breastfed infants, whereas we found no association of early introduction to solids with BMI-z among children who were not breastfed at the time of solids introduction. The average childhood BMI-z for formula-fed infants in our study was still higher than the average BMI-z for breastfed infants, suggesting that early introduction to solids may simply not add to the effect of formula (vs. breastmilk) consumption on BMI-z.

Solid foods are nutritionally distinct from breastmilk in that they do not contain human milk oligosaccharides which are the preferred substrate for infant gut microbiota. Like solids, formula does not contain human milk oligosaccharides. This may explain why among children that were no longer breastfed, early introduction of solids was not associated with higher BMI-z. Diets low in indigestible carbohydrates such as human milk oligosaccharides increase oxidative stress and alter microbiota abundance in mice,³³ supporting a mechanism by which substituting early infant solids for breast milk early may alter childhood BMI-z. Our study adds to the literature on this topic by assessing the timing of complementary foods with differences in childhood gut microbiota composition, which may be one of the mechanisms explaining the relationship between early complementary food introduction and childhood obesity,^6,31,34 among other immune-mediated conditions that have been associated with early introduction to solids (eg, celiac disease,³⁵ type 1 diabetes³⁶ and asthma³⁷). We found that early introduction to complementary foods in infancy was associated with differential abundance of several potentially important gut bacteria, but the specific taxa (discussed below) depended on the duration of breastfeeding.

Several of the microbial taxa associated with early introduction of complementary foods in our study may have clinical implications for paediatric nutrition. Among children still breastfeeding at 4 months of age, we found that later introduction of food was associated with higher abundance of Roseburia (sp. unknown). Roseburia species have been found in breastmilk³⁸ and are sensitive to changes in diet^39,40; thus, introduction of complementary solid foods early could allow other microbes to displace them. This is potentially important because individuals with obesity have lower relative abundance of Roseburia species.⁴⁰ Roseburia are efficient butyrate producers,⁴¹ and murine studies have found that butyrate can mediate inflammatory responses through activation of intestinal epithelial cell receptors such as GPR41 and GPR43, suggesting that disruption of these pathways could play a role in the chronic inflammation seen in obesity.⁴² As such, our findings are consistent with the hypothesis that Roseburia may play a role in the association between early complementary foods and child BMI among infants on breast milk when foods were introduced.

There were several intriguing associations between timing of complementary foods and microbiota composition among infants not on breast milk at 4 months that may also have implications for paediatric nutrition. Ruminococcus bromii, an efficient digester of resistant starches, was higher among children breastfed less than 4 months and introduced to complementary foods early.⁴³ In diets high in resistant starch (eg, certain formulas), R. bromii may out-compete other commensals incapable of extracting energy as efficiently, which could contribute to altered metabolism and disruption of the gut microbial community.⁴³ Further, children not breastfeeding at 4 months and given solids earlier had lower abundance of Bifidobacterium animalis. Bifidobacterium are among the most prevalent genera in the infant microbiota, but decrease as the infant ages and weaning is initiated.^31,34,44 They help maintain mucosal health and feed neighbouring microbiota by producing acetate and B-vitamins through fermentation of otherwise inaccessible oligosaccharides commonly found in human breastmilk.^45–48 Supplementation of a Bifidobacterium animalis strain in a randomized clinical trial of pre-term infants was found to reduce faecal calprotectin and intestinal permeability, demonstrating that higher abundance of this species may be beneficial for infant gut health.⁴⁹ The metabolic by-products of some Bifidobacterium species can also be used by butyrate-producing microbes like Roseburia, thus allowing Bifidobacterium species to indirectly influence butyrate production.⁵⁰

Taken collectively, some of our findings on the association of timing of complementary foods with microbiota composition are compatible with the effect measure modification we observed between early foods and BMI-z. We did not find evidence that early complementary foods were associated with microbial alpha or beta diversity, suggesting that overall microbial community structure may not be modified. Future studies, with longitudinal measurement of the gut microbiota in infancy, are needed to specifically test the hypothesis that changes to the microbiome composition and diversity mediate the observed breastfeeding-specific associations of early introduction to solids in infancy with child obesity and other health outcomes.

6.1 |. Strengths and limitations

Our study has distinct strengths. Gen3G is a well-characterized birth cohort that has collected a diverse array of covariate data, affording us the opportunity to evaluate effect measure modification by breastfeeding duration and to rigorously adjust associations for potential confounding factors. Associations in our study were robust to adjustment for measured confounders. We also sequenced technical controls with our samples to confirm the quality of our sequencing data.

Our study also has some limitations. First, maternal recall of timing of complementary food introduction and breastfeeding duration may have contributed to exposure misclassification. However, the mothers recall of these exposures at the 3 and 5 year follow up visits was quite reliable (Kappa statistic of 0.87 and 0.71, respectively), and we expect misclassification to be non-differential with respect to study outcomes, thus biasing observed associations towards the null. We also did not ask mothers about the type of complementary foods introduced. Another limitation was the lack of stool collection before 5 years of age, which precluded us from characterizing more proximal changes in the gut microbiota. Gut microbiota are highly dynamic over the first years of life, becoming more adult-like around age 3 years.⁵¹ As such, some of the associations with microbiota signatures observed in our study will differ from studies³² that assess microbiota closer to the time of complementary feeding. And while we estimated microbial community membership from a single stool sample, an adult study showed that one sample alone provides long-term information regarding microbial composition.⁵² As 16S rRNA gene sequencing on the MiSeq platform requires PCR amplification, our choice of primers and hypervariable region may have introduced bias into our quantification of microbial relative abundances.⁵³ Yet while our ability to detect specific taxa may have been reduced compared to shotgun metagenomic sequencing, we have no reason to believe that any differences in microbial quantification as a result of primer or hypervariable region choice would be related to our exposure or outcome. Finally, although we controlled for delivery mode, birth weight and gestational age, we cannot exclude the possibility of residual confounding by factors such as socioeconomic status, for which we had limited data.

7 |. CONCLUSION

Our study addresses an important gap in knowledge by jointly examining the timing of introduction to infant complementary foods and breastfeeding duration in relation to the child gut microbiota, showing that timing of introduction to foods other than breast milk or formula may impact the gut microbiota composition out to at least 5 years of age. Specifically, we found that earlier introduction to complementary foods (at or before 4 months of age) was associated with differential abundance of several metabolically active bacteria, but the specific taxa we identified differed by whether the infant was breastfed more or less than 4 months of age. We also found that early infant complementary feeding was associated with higher childhood BMI-z, but only among children that were still breastfeeding when early complementary foods were introduced. The clinical implications of these results are that the effect of early introduction of complementary foods on paediatric nutrition and gut microbiota composition may depend on whether the infant is breast fed or formula fed when foods are introduced. Larger prospective birth cohort studies with repeated measures of the gut microbiota and microbiota-derived metabolites are needed to further elucidate the role of altered gut microbiota composition with health outcomes that have been associated with early infant introduction to solid foods.

Abbreviations:

ASV

amplicon sequence variant

BMI

body mass index

BMI-z

BMI z-score

FDR

false discovery rate

Gen3G

genetics of glucose regulation in gestation and growth

GLM

generalized linear regression model

PCoA

principal component analysis

PERMANOVA

permutational multivariate analysis of variance