The relationship between infant feeding types, gut microbiome, intestinal inflammation, and neurodevelopment in a neonatal piglet model

Abstract

Background

The influences of nutrition on the infant's developing gastrointestinal (GI) microbiome, intestinal tract, and brain is unclear. Human milk (HM) is associated with beneficial immune and cognitive development compared with infant formula (IF). This study used a neonatal piglet model to determine the effects of infant feeding exposures (HM vs IF) on the GI microbiome, intestinal inflammation, and brain oligodendrocyte maturation.

Methods

Six pairs of piglets received HM or IF for 28 days. Fecal samples were collected weekly and GI regions (jejunum, ileum, and colon) and brains were harvested at necropsy. Fecal microbiome composition was determined by 16S ribosomal RNA (16S rRNA) sequencing. Intestinal inflammation was assessed via quantification of intestinal interleukin (IL)‐1β, IL‐8, IL‐10, tumor necrosis factor (TNF)‐α, and fecal calprotectin. Neurodevelopment was evaluated by quantifying mature and immature oligodendrocytes in gray and white matter.

Results

Bacterial community composition differed between feeding groups (P < 0.002) and over time (P = 0.001). Various highly abundant genera were associated with changes over time (P < 0.05) and only Escherichia‐Shigella was associated with feeding group by time interactions (P < 0.05). No differences were found in intestinal inflammatory markers between feeding types, but mature oligodendrocytes in white matter were higher in HM‐fed piglets (P = 0.004). Various intestinal inflammatory markers and relatively highly abundant genera were significantly associated.

Conclusion

Piglet fecal bacterial compositions differed by feeding group and over time, with several relatively highly abundant genera associated with intestinal inflammatory markers. Additionally, HM may support proper white matter development. Future research should investigate mechanisms underlying these relationships.

CLINICAL RELEVANCY STATEMENT

Early life nutrition is crucial for and influential on both short‐term and long‐term biological development. Both the World Health Organization (WHO) and American Academy of Pediatrics (AAP) recommend exclusive human milk (HM) feeding for the first six months of life because of its nutritive and nonnutritive, bioactive impacts. This pilot study investigated how HM and infant formula (IF) differ in their impact on the developing gastrointestinal (GI) tract, microbiome, and brain.

INTRODUCTION

Infant gastrointestinal (GI) microbiome colonization and composition are important for overall development, ¹ , ² but underlying mechanisms are not well‐understood. The GI microbiome supports development and regulation of the immune system. ³ Microbiome colonization induces GI‐associated lymphoid tissue development and expansion, ⁴ which is associated with a low‐grade, regulated inflammatory response. ⁵ This state promotes intestinal maturation ⁶ and trains both the innate and adaptive immune systems. ⁷ , ⁸ Furthermore, concurrent development of the infant GI microbiome, intestinal maturation, and brain is important for establishing the gut–brain axis. ⁹

Early life nutrition impacts GI microbiome composition, ¹⁰ , ¹¹ , ¹² and human milk (HM) is beneficial for development and immune protection of newborns. ¹³ This study explores the relationship between HM feeding on GI microbiome composition and beneficial effects. Specifically, HM is positively associated with various cognitive and GI microbiome developmental outcomes. ¹⁴ Although bovine‐based infant formula (IF) supports growth needs, ¹⁵ there is controversy regarding physiologic manifestations in infants lacking HM bioactive molecules. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸

How HM and IF influence infant development, and the role of the gut–brain axis therein, is not well‐characterized. Magnetic resonance imaging (MRI) studies suggest that HM‐fed infants have greater white matter volume compared with IF‐fed counterparts. ¹⁹ Oligodendrocytes are glial cells that ensheathe the neuronal axons with myelin to improve signal conduction in the central nervous system; this forms a significant structural component of white matter. Quantifying immature and mature oligodendrocytes can assess neurodevelopment in animal models. Although low‐grade inflammation stimulates GI development, exaggerated intestinal inflammation may reduce GI barrier function and lead to systemic inflammation with negative impacts on neurodevelopment. ⁹ This pilot study aims to determine effects of HM and IF on the GI microbiome composition, intestinal inflammation, brain oligodendrocyte maturation, and assess interactions between these variables. We hypothesized that HM feeding would modify the GI microbiome, reduce intestinal inflammation, and improve neurodevelopment compared with IF‐fed neonatal piglets.

MATERIALS AND METHODS

Piglet model

Six pairs of 2‐day‐old Yorkshire‐Duroc male neonatal piglets were reared in individual housing. Piglets received either donated unpasteurized HM or IF for 28 days. Piglets were fed and monitored three times daily, and feedings were calculated (263 kcal/kg per day ²⁰ ) based on morning weights. Frozen donated milk was thawed at 4°C overnight, pooled, stirred, separated into 2–3 L batches, and frozen at −80°C. Before feeding, milk was thawed at 4 C overnight and warmed to 42°C–45°C. Milk batches were pooled from at least three donors; characteristics and nutrition composition are shown in Supporting Information S1: Table S1A and B, respectively. Bovine‐based IF (Similac NeoSure, Abbott) was prepared daily. Published HM compositions were used to estimate energy and protein contents. ²¹ Additional whey protein isolate, vitamins, and minerals were added to both feeding groups (Supporting Information S1: Table S2) to be isocaloric and isonitrogenous and meet nutrition needs of growing piglets. ²⁰ After 28 days, piglets were fed in the morning, anesthetized with tiletamine/zolazepam and xylazine (Dechra Veterinary Products), and euthanized with Fatal Plus (Vortech Pharmaceutical Ltd). GI tracts sections (jejunum, ileum, and colon) were snap frozen and stored at −80°C. Brains were harvested and postfixed in 10% neutral‐buffered formalin for 48 h at 4°C, rinsed, and stored in phosphate‐buffered saline (PBS) with 0.1% sodium azide. This study was approved by the University of Idaho Institutional Animal Care and Use Committee (IACUC‐2020‐56/2023‐53).

DNA processing, extraction, and library preparation

Fecal samples from all piglets were collected at baseline (week 0), weekly (weeks 1–4), and necropsy (week 5). Using wooden spatulas, fecal samples were collected from piglet housing within 24 h of defecation and stored in sterile cryocontainers at −80°C. For extraction of DNA, we utilized the Quick‐DNA Fecal/Soil Microbe Miniprep Kit (Zymo Research) following manufacturer's instructions. Whole‐cell positive controls (Catalog No. D6310, D6300, D6331, D6310; Zymo Research) and negative controls (elution buffer and sterile water; Catalog No. R0582; Thermo Fisher Scientific) were also extracted. Sequencing libraries were constructed based on a dual indexing strategy and a two‐step polymerase chain reaction (PCR1 and PCR2). Primers targeted the V3–V4 region of the 16S ribosomal RNA (rRNA) gene; permutations of primers 319F (5′‐ACTCCTACGGGAGGCAGCAG‐3′) and 806 R (5′‐GGACTACHVGGGTWTCTAAT‐3′) were used for PCR1. ²² See Supporting Information: Table S3 and Supporting Information: Methods for details.

Sequence data processing

Demultiplexing of sequences was conducted using dbcAmplicons (https://github.com/msettles/dbcAmplicons) using default parameters. Quality control, trimming, filtering, merging of forward and reverse reads, chimera removal, identification of amplicon sequence variants (ASVs), and taxonomic assignment of the ASVs using the SILVA database (version 138.1) ²³ were performed with R software (version 4.3.1) ²⁴ using the DADA2 pipeline. ²⁵ Sequence read contaminants were determined from negative controls using the R package decontam ²⁶ (version 1.22.0, function=isContaminant, method=prevalence, threshold=0.1) and excluded from further analysis. The DADA2‐silva–derived taxonomy was edited to replace “NA” classifications with the next highest classification as described by Pace et al. ²⁷ Relative abundance of taxa was calculated, and β diversity (ie, a measure of how similar or dissimilar bacterial community profiles are between samples) was compared between HM and IF feeding groups, and over time using both principal coordinates analysis (PCoA) and nonmetric multidimensional scaling (NMDS) based on Jaccard (presence/absence) and Bray–Curtis (relative abundance) distances. The 10 most significant genera driving NMDS ordinations based on Bray–Curtis distances were determined using the envfit function of the vegan package ²⁸ (version 2.6.4). Other R packages used in analysis include phyloseq ²⁹ (version 1.46.0), tidyverse ³⁰ (version 2.0.0), ggplot ³¹ (version 3.5.0), and fantaxtic ³² (version 0.2.0).

Intestinal inflammation

Protein extracts from frozen segments (~0.2 g) of the GI tract (jejunum, ileum, and colon) were obtained by homogenization with glass beads using the Tissue Lyser II (30 Hz, 2 min, repeated 3–4 times; Qiagen) in Tissue Protein Extraction Reagent (TPER; 1 ml TPER/0.15 g tissue; Thermo Fisher Scientific) with Halt Protease Inhibitor Cocktail (HALT; 10 μL HALT/1.0 ml TPER; Thermo Fisher Scientific). Total protein concentrations were quantified in triplicate using the Pierce BCA protein assay quantification kit (bovine serum albumin as protein standard; Thermo Fisher Scientific). Concentrations of the proinflammatory cytokines interleukin (IL)−1β, IL‐8, and tumor necrosis factor (TNF)‐α and the anti‐inflammatory cytokine IL‐10 were quantified using enzyme‐linked immunosorbent assay (ELISA; Porcine Quantikine ELISA; R&D Systems, Inc) kits according to manufacturer's instructions. All cytokines were measured in triplicate using undiluted samples and values were normalized to the total protein concentration. Fecal calprotectin concentrations were measured using the Porcine Calprotectin ELISA Kit (MyBioSource) with modifications as previously outlined (see Supporting Information: Methods). ³³

Oligodendrocyte maturation

Piglet brain tissue was cryoprotected in a 10–30% sucrose gradient, embedded in optimal cutting temperature compound (Fisher Scientific), and sectioned at 20 μm prior to immunohistochemical assays for immature oligodendrocytes (platelet‐derived growth factor receptor α [PDGFRα]) and mature oligodendrocytes (adenomatous polyposis coli [APC]) and imaging. Tissues were blocked with 10% donkey serum (Thermo Fisher Scientific), 0.1% Triton X100 (Sigma‐Aldrich) prior to overnight incubation at 4°C with primary antibodies in 1× PBS + 0.1% Triton‐X (PBS‐T). Dilution of primary antibodies was 1:500 for mouse anti‐APC (OP80‐100UG; Calbiochem) and 1:250 for rabbit anti‐PDGFRα (AB203491; Abcam). Tissue was rinsed with PBS and incubated with secondary antibodies (1:500 antimouse 488 and antirabbit 568; Abcam Alexa) for 2 h at room temperature. Sections were counterstained with 4′,6‐diamidino‐2‐phenylindole (DAPI; Sigma‐Aldrich) and cover‐slipped with Fluoromount‐G (Invitrogen). Images were acquired on Keyence BZ‐X810 using BZ‐X800 viewer software to define areas of capture that encompassed the entirety of right prefrontal cortical segment. Images were acquired at ×20 magnification and stitched using Image Stitch, BZ‐X800 Analyzer, and WideImage Viewer. Pyramidal tiff files were exported to QuPath ³⁴ to capture 1X magnified images (0.05 mm² per field of view) for each section which were quantified with ImageJ. ³⁵ Oligodendrocyte maturation was assessed by quantifying and comparing cell counts of PDGFRα and APC in both gray and white matter between groups.

Statistical and exploratory analyses

Statistical analyses were run using R software (version 4.3.1). ²⁴ The vegan ²⁸ (version 2.6.4) adonis2 function was used for permutational multivariate analysis of variance (PERMANOVA) using default parameters to determine effects of feeding group and time (weeks) on β diversity. Beta dispersion was assessed using the betadisper function. Generalized linear mixed effect models were applied to determine the effects of feeding group and time, and feeding group by time interactions using the glmmTMB package ³⁶ (version 3.1.4) (Formula: Bacteria ~ Feeding × week + (1|Pig_ID)). The model was fit using a zero‐inflated model with a β family distribution. The emmeans package ³⁷ (version 1.10.1) was used for post hoc analyses and each test was controlled for multiple comparisons with the Tukey Honest Significant Difference method. ³⁸

Normality was assessed for intestinal inflammatory markers by visualizing qqplots. A multivariate analysis of variance (MANOVA in R ²⁴ ) was used to determine differences in cytokine concentrations across GI locations and feeding groups, and Bonferroni P value corrections were used to determine significance. A linear mixed effect model was used to determine the effects of feeding group, time, and the interaction of feeding group by time on fecal calprotectin (formula: fecal calprotectin ~ feeding × week + (1|Pig_ID)) using the lme4 package ³⁹ (version 1.1.35.2). Spearman rank correlations were conducted to assess relationships between intestinal inflammatory markers and the 20 most relatively abundant genera using the corrplot package ⁴⁰ (version 0.92).

For brain analysis, total cell counts of APC+ and PDGFRα+ oligodendrocytes in both gray and white matter were quantified in uniform view fields and compared between groups. Linear mixed effect models (lme4 package, ³⁹ version 1.1.35.2) were used to assess the effect of feeding, sections per piglet, and brain region on oligodendrocyte type counts, as deemed necessary for the model fit (formula: count of cell type (PDGFRα or APC) in brain matter (gray and white separately) ~ diet + (1|Animal_ID/Section)). Assumptions for the model were confirmed by visualizing qqplots. The warning “isSingular” appeared and it was determined that the individual sections within the individual piglets did not add variance to the model; thus, the model was simplified (formula: APC cells in white matter ~ diet). Residual plots confirmed the model was appropriate for statistical inference.

Statistical significance was determined at P < 0.05. Because of the exploratory nature of this study, global P value adjustments were not made.

RESULTS

All samples collected and included in the study are shown in Supporting Information: Figure S1. From six pairs of piglets, 56 fecal samples were collected for microbiome analysis (HM: n = 29; IF: n = 27), 18 fecal samples underwent calprotectin analysis after microbiome analysis (HM: n = 9; IF: n = 9), cytokines were analyzed in all intestinal sections, and 6–9 sections from six brains were utilized for brain analysis (HM: n = 3; IF: n = 3).

GI microbiome

Average relative abundances of bacterial communities of the overall 20 most relatively abundant genera (organized by phyla) across all time points are included in Figure 1. Although variation in bacterial memberships is visually apparent between feeding groups, there was no statistically significant difference (Figure 1A). Relative abundances of bacterial genera changed over time. Notably, Ruminococcus peaked at week 3 (0.090 ± 0.076) before decreasing in both groups (week 4: 0.072 ± 0.081; week 5: 0.046 ± 0.058; P < 0.05; Figure 2A). Variations were observed in Streptococcus over time with trends similar in both groups. At baseline, Streptococcus was present at high relative abundance (0.099 ± 0.091) before dropping significantly (0.004 ± 0.005) in both groups by the end of the study (Figures 1B and 2B; P < 0.001). Escherichia‐Shigella was the only genera with significant feeding group by time interactions (between HM at baseline and IF at necropsy, P < 0.05; and between IF at week 2 and HM at week 4, P < 0.05; Supporting Information: Table S4).

Figure 1.

Average relative abundance bar charts of the 20 most relatively abundant genera (categorized and colored by phyla) found in fecal samples from: (A) human milk (HM)‐fed and infant formula (IF)‐fed piglets across the entire study period; and (B) HM‐fed and IF‐fed piglets by each study week. (See Supporting Information: Figure S2 for relative abundance patterns of individual HM‐fed and IF‐fed piglets throughout the study.).

Figure 2.

Relative abundances of various genera across sampling time points are shown. Purple‐filled boxes signify the presence of time main effects, whereas red‐filled and blue‐filled boxes signify the presence of feeding group by time interactions. Significant differences in genera relative abundances determined by generalized linear mixed effect models are included for (A) Ruminococcus (main effect of time between: week 0 and week 2, week 0 and week 3, week 0 and week 4, and week 1 and week 3); (B) Streptococcus (main effect of time between: week 0 and week 1, week 0 and week 2, week 0 and week 3, week 0 and week 4, week 0 and week 5); and (C) Escherichia‐Shigella (main effect of time between: human milk (HM; maroon)‐fed piglets at week 0 and HM at week 3, HM at week 0 and week 4, HM at week 0 and week 5, infant formula (IF; blue)‐fed piglets at week 2 and week 5; feeding group by time interactions between HM at week 0 and IF at week 5, IF at week 2 and HM at week 4). BL, baseline or week 0; Nec., necropsy or week 5. *P < 0.05, **P < 0.001.

We determined differences in bacterial communities between groups and over time using β diversity and PERMANOVA analyses and found bacterial community compositions differ between feeding groups (P = 0.001; Figure 3A) and change over time (P = 0.001; Figure 3B). Main effects of both feeding type (P = 0.002) and time (P = 0.001) also explain variation in NMDS plots based on Bray–Curtis distances (Figure 3C,D). Ruminococcus was associated with IF‐fed piglets, whereas Collinsella was associated with HM‐fed piglets. However, the test of multivariate homogeneity of group dispersion was significant for time (P < 0.05) indicating that groups were not equally dispersed in the multivariate space and there could be additional underlying ecological or environmental variability within groups.

Figure 3.

Principal coordinates analysis (PCoA) ordinations using the Jaccard distance matrices of piglet fecal microbiome samples from: (A) all human milk (HM; shown in maroon) fed and infant formula (IF; shown in blue) fed piglet samples across all timepoints; and (B) all samples at all timepoints with increasing color darkness by study weeks (lightest blue = BL [baseline] and darkest blue = Nec. [necropsy]). P = 0.001 for feeding main effect; P = 0.001 for week main effect; β dispersion not significant. Ellipses represent 95% confidence areas. Values in parentheses reflect the percentage of variation explained by each axis. Nonmetric multidimensional scaling (NMDS) ordinations using the Bray–Curtis distance matrices of piglet fecal microbiome samples from: (C) HM (shown in maroon) fed and IF (shown in blue) fed piglet samples across all timepoints with arrows denoting the 10 most significant genera vectors; and (D) all samples at all timepoints with increasing color darkness by study weeks (lightest blue = BL [baseline] and darkest blue = Nec. [necropsy]) with arrows denoting the 10 most significant genera vectors. P = 0.002 for feeding main effect; P = 0.001 for week main effect; and P < 0.05 for β dispersion for week main effect. Ellipses represent 95% confidence areas. PCoA plots using the Bray–Curtis distance matrix were similar and can be found in Supporting Information: Figure S3.

Intestinal inflammation

No differences in IL‐1β, IL‐8, TNF‐α, and IL‐10 concentrations were found between groups within any intestinal region (Figure 4A). There were also no group differences in fecal calprotectin concentrations over time. Mean concentrations of inflammatory markers can be seen in Supporting Information: Table S5.

Figure 4.

(A) Boxplots of cytokine concentrations (interleukin 1 β [IL‐1β], interleukin 8 [IL‐8], and tumor necrosis factor α [TNF‐α]; all in pg/ml) in human milk (HM; maroon) fed and infant formula (IF; blue) fed piglets from jejunal, ileal, and colonic regions. No significant differences were detected between feeding groups; (B) Heatmap of significant associations between various intestinal inflammatory markers (jejunal, ileal, and colonic IL‐1β, IL‐8, and TNF‐α and fecal calprotectin) and the overall 20 most relatively abundant genera observed in both human milk and infant formula feeding groups. Only correlations that were significant (P < 0.05) are shown, with larger circles showing smaller P values and strength of those associations denoted by color (blue = positive correlation; red = negative correlation). FC, fecal calprotectin; Jej, jejunum; Ile, ileum; Col, colon.

Oligodendrocyte maturation

Within each sampled region of the brain, six subsections (three gray matter, three white matter) were counted for mature (APC) and immature (PDGFRα) oligodendrocytes from 252 images (Figure 5A–D). The APC cell counts were higher in white matter of HM‐fed piglets (P < 0.004) vs IF‐fed piglets (Figure 5E). No other differences were observed in cell counts between groups.

Figure 5.

(A) Representative images of fields of view (FOV) for brain oligodendrocyte quantification of human milk‐fed piglets (A) gray matter and (B) white matter, and infant formula‐fed piglets (C) gray matter and (D) white matter. 4′,6‐Diamidino‐2‐phenylindole (DAPI) = blue; adenomatous polyposis coli (APC; mature) = green; platelet‐derived growth factor receptor α (PDGFRα; immature) = red. (E) Quantification of oligodendrocyte cell type (APC): mature; PDGFRα (immature) per FOV (0.05 mm²) and respective brain matter region (gray matter vs white matter) compared between human milk (HM; maroon) and infant formula (IF; blue) groups. *P < 0.05.

DISCUSSION

Bacterial community composition differed significantly between feeding groups and over time (Figures 1, 2, 3), markers of inflammation did not differ between groups (Figure 4A) but certain bacterial genera were associated with specific inflammatory markers (Figure 4B), and HM‐fed piglets had more mature oligodendrocytes in white matter (Figure 5B). Our findings reinforce previous research suggesting early life nutrition impacts fecal bacterial community composition. ¹⁰ , ¹¹ , ¹² Our study examines the temporal dynamics of microbial colonization in response to different early feeding exposures. To understand how food source influenced bacterial composition over time, we used ordination methods. The binary Jaccard distance matrix compared presence‐absence of genera, whereas the Bray–Curtis dissimilarity matrix compared relative abundance of genera. In the PCoA plots using the Jaccard distance matrix, more distinct clustering by week was observed (Figure 3B) than for Bray–Curtis (Supporting Information: Figure S3B). This suggests different genera were present or absent in fecal samples collected near day 2 relative to day 28, which reflects the dynamics of succession in the microbiome during colonization. Relative abundances of Streptococcus and Escherichia‐Shigella decreased over time and suggest GI microbiome community alterations continue throughout early development (Figure 2B–C).

Although our study did not find differences between feeding groups among the 20 most relatively abundant genera, 10 of those genera contributed towards variation in the NMDS ordinations (Figure 3C–D). Notably, a feeding group by time interaction was observed for Escherichia‐Shigella (Figure 2C). This suggests complex interactions among bacteria during early GI bacterial community development.

Inflammatory cytokines did not differ regionally between groups (Figure 4A); fecal calprotectin concentrations did not change between groups or time (Supporting Information: l Table S5). This finding is not altogether novel; previous work also observed no differences in IL‐8 between HM‐fed and IF‐fed infants. ⁴¹ Fecal calprotectin studies exhibit great variability: some found lower fecal calprotectin in HM‐fed infants compared with IF‐fed infants at 1‐month of age, ⁴¹ whereas others found no differences between groups. ⁴² The overall relatively low concentrations of inflammatory markers in this study is interesting because previous research demonstrated that low‐grade inflammation occurs in the developing intestines. In fact, low‐grade inflammation may be important for proper intestinal tract, GI microbiome, and immune development, whereas the immune system is trained to distinguish between commensal and pathogenic bacteria. ⁴³ The lack of general intestinal inflammation could in part be due to the controlled laboratory environment and may not reflect environmental exposures serving as immune stimuli in newborn infants.

Multiple associations were observed between inflammatory markers and relative abundances of the 20 most abundant genera (Figure 4B). Although the physiological basis of these associations cannot be determined in this study, it provides groundwork for further exploration of these relationships. Notably, Ruminococcus and Collinsella were negatively correlated with certain proinflammatory markers, whereas Streptococcus and Escherichia‐Shigella were positively correlated with others. This is interesting because butyrate‐producing bacteria, including Ruminococcus, have recently been associated with reduced neuroinflammation. ⁴⁴ As both Streptococcus and Escherichia‐Shigella decreased in relative abundance throughout the course of the study (Figures 1B, 2B,C), this may explain the relatively low intestinal inflammatory profiles observed as both genera are associated with inflammatory conditions. ⁴¹ , ⁴⁵ Additionally, Streptococcus does not clearly associate with HM‐ or IF‐fed infant populations, ¹² , ⁴¹ nor does Escherichia‐Shigella clearly associate with necrotizing enterocolitis in infants. ⁴⁵ Further investigation is warranted to understand these relationships and their physiologic effects in infant/piglet populations.

Immature and mature oligodendrocytes were quantified (Figure 5A–B) from the right prefrontal cortex in both white and gray matter as a measurement of maturation. ⁴⁴ , ⁴⁶ The prefrontal cortex was used because of its rapid development and susceptibility to early life environmental exposures. ⁴⁷ White matter injury (and/or the loss of premyelinating oligodendrocytes like APC cells) is associated with risk for neurodevelopmental impairment. ¹⁹ Intriguingly, the only notable difference between groups was found in white matter APC+ cells. This may indicate greater maturation of oligodendrocytes in the white matter of HM‐fed piglets or IF‐fed piglets experience delayed oligodendrocyte maturation. A previous MRI study found increased white matter volume in HM‐fed infants supporting this finding and suggesting nutritive and/or nonnutritive components found in HM and IF influence white matter development. ¹⁹ Future research assessing oligodendrogenesis is needed to confirm this hypothesis.

We acknowledge several study weaknesses. There were inconsistent numbers of fecal and brain samples over time for analyses because of inadequate sample amounts and technical difficulties, respectively. Based on observed differences and anticipated variability, 14 pairs of feeding‐trial piglets (28 total) are needed for 80% statistical power based on our preliminary functional neurodevelopment data. However, only six pairs of piglets were included in this exploratory study. Our small sample size limits our ability to connect GI microbiome and neurodevelopmental markers. Further, using a control group of sow‐fed piglets would be useful for comparison. Nevertheless, this study is a step forward in an area that is under‐developed. We measured several biomarkers to explore potential gut–brain axis connections over time with repeated fecal collections. The neonatal piglet model controls for variables that often confound human studies (ie sociodemographic) and allows for tissue collections (intestinal and brain samples) that are unethical in human.

Future research is necessary to investigate mechanisms underlying these connections while considering the myriads of microbe‐to‐microbe and microbe‐to‐host interactions, which could mediate potential gut–brain connections and emphasize the importance of infant nutrition on multiple systems. Additional studies could incorporate metagenomic and metabolomic analyses to increase comprehensive understanding of microbiome dynamics during early life.

AUTHOR CONTRIBUTIONS

Heidi Sellmann: Methodology; software; data curation; investigation; formal analysis; writing—original draft; Writing—review and editing. Janet E. Williams: Methodology; software; data curation; validation; supervision; resources; writing—review and editing. Klas Udekwu: Methodology; validation; resources; writing—review and editing. Ashley McDonough: Methodology; software; data curation; investigation; validation; supervision; project administration; resources; writing—review and editing. Katie Heckathorn: Methodology; data curation; writing—review and editing. Laurel Nuñez: Methodology; data curation; investigation; project administration; writing—review and editing. Yimin Chen: Conceptualization; investigation; validation; formal analysis; supervision; visualization; writing—review and editing; project administration; methodology; funding acquisition; resources; writing—original draft.

CONFLICT OF INTEREST STATEMENT

None declared.

Supporting information

HS Manuscript Submission Supplemental Revised.

Sellmann H, Williams JE, Udekwu K, et al. The relationship between infant feeding types, gut microbiome, intestinal inflammation, and neurodevelopment in a neonatal piglet model. J Parenter Enteral Nutr. 2026;50:121‐130. 10.1002/jpen.70019