Feeding Type Shapes Infant Gut Microbiota and Metabolite Profiles in a Simulated Colonic Model

Abstract

The establishment of the infant gut microbiota is strongly influenced by feeding type, with human milk (HM) and infant formula (IF) driving distinct microbial profiles. This study compared the effects of donated raw human milk (HM) and Holder-pasteurised HM (Past-HM) on microbial composition using a 4-stage in vitro gut model inoculated with pooled faecal samples from five healthy 3–6-month-old infants. Colonic digesta were sampled over a 48-h period following introduction of raw HM, Past-HM and IFs for microbial diversity by quantitative polymerase chain reaction (qPCR) and 16S rRNA gene sequencing, and short-chain fatty acid (SCFA) analysis by gas chromatography. Microbial profiling revealed 2026 operational taxonomic units (OTUs) across eight phyla and 165 genera. Past-HM and goat milk-based IF (GIF) promoted Bifidobacterium dominance and produced higher levels of total SCFA, especially acetic acid, compared to HM. Spearman correlation linked Bifidobacterium with acetic acid and Escherichia-Shigella with propionic acid. PCA showed OTU composition (Dim1, 72.6%) and SCFA profiles (Dim2, 19.8%) and distinguished control and milk-inoculated groups. Feeding type was the dominant factor shaping microbiota and metabolite profiles, exerting a stronger influence than incubation time (PERMANOVA p < 0.05). These findings underscore the pivotal role of early feeding choices in directing both microbial colonisation and microbial metabolic activity in the infant gut.

1. Introduction

A normal infant gut microbiota is essential for digestion, protection against pathogens, and development of the immune system, and microbial metabolites, including short-chain fatty acids (SCFA), play an essential role in this regard [1,2,3]. Colonisation of the newborn gut by bacteria begins immediately after birth and is influenced by the birth delivery mode and initial feeding practices. It results from exposure to facultative anaerobes (mostly Escherichia coli, Staphylococcus spp., and Streptococcus spp.) from the mother’s vaginal canal, for a normal delivery. Soon after, the infant’s gut turns anaerobic, allowing strict anaerobes (Bifidobacterium spp., Clostridioides spp. and Bacteroides spp.) to develop along with lactobacilli. This process includes transfer of bacteria from the mother’s breast milk to the infant’s gut, although the underlying mechanisms for transfer of microbes from the mother’s gut to breastmilk are unclear [4,5,6]. After colonisation, bacterial diversity rapidly increases in the first year of life and stabilises by five years of age. Disturbances affecting the gut microbiota and their metabolites during this period are associated with various childhood diseases, including allergies, infections, metabolic disorder, and immune dysregulation [5,7,8,9].

Breastfeeding is recommended for the first 6 months of life by the World Health Organisation (WHO) and by the American Academy of Paediatrics (AAP) [10]. The benefits of breastfeeding during this early stage of life have been confirmed by numerous studies [11,12,13,14]. Breastfeeding promotes a healthy gut microbiota, leading to optimum immune system development [5,15,16,17]. Breastfed babies develop a high abundance of Bifidobacterium spp. in the gut, since these bacteria can utilise milk oligosaccharides present at uniquely high concentrations in human milk (HM) [15,18,19,20,21]. Breastfed infants have a higher relative abundance of Bifidobacterium and Lactobacillus species [22], and lower relative abundance of taxa such as Enterobacteriaceae, Proteobacteria, Veillonella, Clostridioides, and Bacteroides [20]. HM is also rich in antimicrobial proteins that help protect against pathogens, such as lactoferrin, lysozyme, and SIgAs [15,17,23].

In cases where breastfeeding is not possible, mothers may rely on infant formulas (IFs), or in the case of low-birth-weight or preterm infants, pasteurised donor human milk (DHM) [24,25]. Feeding with IF alters the gut microbiota, resulting in a more “mature and adult-like” microbiota profile, with an increase in the prevalence of Clostridioides difficile, Bacteroides fragilis, and Escherichia coli and lower numbers of Bifidobacterium spp., which may increase the long-term risk of metabolic and immune diseases [4]. Commercial IFs are usually derived from cow’s milk, often with added synthetic human milk oligosaccharides (HMOs); however, there is also demand for goat’s-milk-based IFs since these have a higher concentration and diversity of HMO-like oligosaccharides [26,27,28]. Both IF products and DHM undergo processing that alters their properties. DHM undergoes Holder pasteurisation as part of essential microbiological safety procedures at milk banks, impacting the HM’s minor components [25,29].

The present study aimed to investigate the impact of different infant feeding practices on the infant gut microbiota using an in vitro gut model system designed to mimic the environment of an infant’s gastrointestinal tract. Using this model, results showed differences in the effects of IFs and DHM relative to natural HM on the development of the infant gut microbiota.

2. Materials and Methods

2.1. Materials

Ethics approval for collecting HM and infant faecal inoculum samples was obtained from the RMIT University Human Research Ethics Committee (Project numbers 22831 and 23249, respectively). Human milk (mature lactational stage) was collected from three breastfeeding women, and each sample was divided into two batches. One batch was kept unpasteurized (raw group), and the other was pasteurised (pasteurised group) following Holder pasteurisation conditions (62.5 °C for 30 min in a water bath). IFs consisted of a goat’s-milk-based infant formula (GIF) (Oli 6 stage 1, Nuchev Pty, Ltd., Melbourne, Australia) and cow’s-milk-based infant formula (CIF) (S-26 Original Stage 1, Wyeth Nutrition, Baulkham Hills, NSW, Australia) obtained from a local supermarket and prepared as per the manufacturer’s instructions. Neither had added HMOs or probiotic bacteria. A pooled faecal inoculum was prepared from the stool samples of five different babies aged between 3 and 6 months with consent from their mothers. Purified water (Type 1 Ultrapure, Merck Millipore Milli-Q Direct 16 Water Purification System, Bayswater, Victoria, Australia) was used for all solutions and buffer preparations. Pancreatin (USP grade) was purchased from MP Biomedicals (Seven Hills, Victoria, Australia). Pepsin, lipase from Rhizopus oryzae, and bile salts powder were purchased from Southern Biological (Alphington, Victoria, Australia). Mucin, α-amylase, α-chymotrypsin, sodium dithionite, and sodium thioglycolate were purchased from Merck (Bayswater, Victoria, Australia). SnakeSkin^TM Dialysis Tubing (3.5 kDa MWCO) was purchased from Thermo Fisher (Scoresby, Victoria, Australia).

2.2. Development and Optimisation of an In Vitro Simulator of the Infant Intestinal Microbial Ecosystem

To evaluate the effect of feeding on the gut microbiota composition and SCFAs, we constructed an in vitro gastrointestinal model as described by Gallier et al. with minor modifications [30]. This was a four-stage static model, simulating the gastrointestinal tract of a 3–6-month-old baby. This model incorporated stomach, duodenum, jejunum/ileum, and colon experimental conditions, but not an oral phase, since oral digestion is nominal at this age. Enzymes and bile salt concentration for gastrointestinal digestion were calculated as described by Gallier et al. [30] and Shani-Levi et al. [31]. Enzyme levels in adults were determined according to the Infogest in vitro digestion tool (http://www.proteomics.ch/IVD/ accessed on 5 November 2024) and adjusted to match the expected levels in term babies (3–6 months old) [30,31,32,33]. The experimental workflow is shown in Figure 1.

In the stomach phase, 15 mL of each sample, either control (water), human milk (from three different donors), its pasteurised version (Past-HM), goat-milk-based infant formula (GIF), or cow-milk-based infant formula (CIF), was added. Infant formulas were prepared according to the manufacturer’s instructions. Each feeding type was mixed with 0.575 mL of gastric fluid (0.65 g/L KCl and 3.65 g/L NaCl), 10 µL of α-amylase (30 mg/mL), 0.335 mL of lipase (50 mg/mL), and 0.1275 mL of pepsin (20 mg/mL), and the mixture was then adjusted to pH 3.2 using 1 M HCl. The mixture was incubated for 1 h at 37 °C with shaking at 30 rpm.

In the duodenum phase, 1 mL of pancreatin (with trypsin activity ≥ 25 USP U/mg) and 0.13 mL of bile salt solution (0.1 g/mL) were added to 16.25 mL of gastric digesta [34]. Purified water was added to adjust the volume to 27.5 mL, and the pH was raised to 6.7 using 1 M NaOH. The mixture was then incubated for 30 min. Following this, 25 mL of the digesta was transferred into a dialysis tube (3.5 kDa molecular weight cut-off) for the jejunum/ileum phase. The tube was submerged in a sodium bicarbonate solution (3.75 g/L in purified water), which was replaced hourly during the 3 h incubation period.

For the colon phase, 20 mL of digesta from the jejunum/ileum phase was transferred to a 100 mL Schott bottle containing 49.5 mL of colonic medium previously inoculated with 0.5 mL of faecal inoculum (30% w/v). The colonic medium, prepared in purified water, consisted of dipotassium phosphate (4.8 g/L), monopotassium phosphate (14.9 g/L), sodium bicarbonate (2 g/L), yeast extract (2 g/L), peptone (2 g/L), mucin (1 g/L), Tween 80 (2 g/L), and cysteine hydrochloride (0.5 g/L). A pooled faecal inoculum was prepared using stool samples collected from five infants aged 3 to 6 months, with RMIT ethics approval for collection as described. The inoculum was a 30% w/v mixture, combining equal amounts of the five stool samples in a solution containing dipotassium phosphate (8.8 g/L), monopotassium phosphate (6.8 g/L), sodium dithionite (0.015 g/L), and sodium thioglycolate (0.1 g/L), dissolved in purified water. The prepared inoculum was aliquoted and stored at −80 °C for use in all subsequent experiments.

Schott bottles were maintained under anaerobic conditions at 37 °C and incubated for 48 h. Samples were collected at 0, 24, and 48 h during colonic fermentation. For each time point, 1 mL of colonic fermentation liquid was analysed to assess gut microbiota composition using qPCR (using specific primers targeting Lactobacillus and Bifidobacterium) and Illumina 16S rRNA gene sequencing. Short-chain fatty acid (SCFA) production was also quantified (see later). Additionally, pH measurements were recorded at the above time points using a standard pH meter (Hanna Model 211, Hanna Instruments Pty Ltd., Keysborough, Victoria, Australia).

2.3. Evaluation of Microbial Community Changes

To track changes in the gut microbial community during colon fermentation, bacterial DNA in colon digesta collected at different time points was extracted for analysis. Bacterial genomic DNA was extracted from 250 µL of colonic digesta using the QIAamp PowerFecal Pro DNA kit (MO BIO Laboratories, Inc., Carlsbad, CA, USA) following the manufacturer’s instructions and used for qPCR quantification of lactobacilli and bifidobacteria and for Illumina sequencing.

For quantification of bacteria by qPCR, amplification using SensiFAST^TM SYBR^® No-ROX (Bioline, London, UK) was used. Each reaction consisted of SensiFAST^TM SYBR^® No-ROX (Bioline) master mix, primers at a final concentration of 400 nM each, 1 µL of template DNA, and nuclease-free water in a final volume of 20 µL. Universal eubacteria primers and primers specific for Lactobacillus spp. and Bifidobacterium spp. were purchased from Integrated DNA Technologies, Coralville, IA, USA. Universal eubacteria and Lactobacillus spp. primer sequences were as described by Bergstrom et al. [35]. Bifidobacterium spp. primers were as described by Tannock et al. [36].

A standard qPCR curve was prepared with pure cultures of a Lactobacillus rhamnosus strain and Bifidobacterium longum subsp. infantis BB536 using serial dilutions and plating to determine the relationship between gene copies (qPCR) and viable counts (CFU/mL). Counts were performed by plating using lactobacilli MRS agar [37] (Difco) (Thermo Fisher, Scoresby, Victoria, Australia) supplemented with 0.05% (w/v) cysteine for bifidobacteria, and incubation of plates anaerobically at 37 °C. qPCR was performed using a CFX Connect Real-Time System (BioRad, Hercules, CA, USA) with the following temperature cycling conditions: 95 °C for 3 min, followed by 40 cycles of 30 s at 95 °C and 30 s at 56 °C or 60 °C (for Lactobacillus/Bifidobacterium or eubacterial detection, respectively). A melt curve at the end of 40 cycles was conducted to detect primer-dimer formation. It should be noted that using this method, there was no distinction between live vs. dead bacterial cells.

Total bacterial DNA extracted from the colonic fermentation suspension was sent to the Australian Genome Research Facility (AGRF) in Melbourne, Australia, for Illumina bacterial 16S RNA gene amplification and sequencing. The V3–V4 region of 16S rDNA was amplified using 16S rRNA primers 341F and 806R. Cycle conditions are as follows: initial step of 98 °C for 30 s, followed by 30 cycles of dissociation at 98 °C for 10 s, annealing at 60 °C for 10 s, and final extension at 72 °C for 5 min. Thermocycling was completed with an Applied Biosystem 384 Veriti and using Platinum SuperFi II Mastermix (Life Technologies, Mulgrave, Victoria, Australia) for the primary PCR. The first-stage PCR was cleaned using magnetic beads, and samples were visualised on 2% SYBRTM E-GelTM (Thermo-Fisher). A secondary PCR to index the amplicons was performed with PlatinumTM SuperFi II PCR Master Mix (Thermo Fisher). The resulting amplicons were cleaned again using magnetic beads, quantified by fluorometry (Quantifluor^® System, Promega, Alexandria NSW, Australia) and normalised. The equimolar pool was cleaned a final time using magnetic beads to concentrate the pool and then measured using a High-Sensitivity D1000 Tape on an Agilent 2200 TapeStation (Agilent Technologies, Mulgrave, Victoria, Australia). The pool was diluted to 5 nM, and molarity was confirmed using a Qubit High Sensitivity dsDNA assay (Thermo Fisher). This was followed by sequencing on an Illumina MiSeq (San Diego, CA, USA) with a V3, 600-cycle kit (2 × 300 base pairs paired end).

Bacterial 16S rRNA gene sequences were analysed using the Illumina DRAGEN BCL Convert 07.021.645.4.0.3 pipeline. In brief, the demultiplexed raw reads were primer-trimmed and quality-filtered using the cutadapt plugin, followed by denoising with DADA2 (via q2-dada2) [38] in R (version 4.4.1). Taxa were assigned to amplicon sequence variants (ASVs) using the q2-feature-classifier and sklearn naïve Bayes training and classification algorithms [39].

2.4. SCFA Measurement with LC-MS

Short-chain fatty acids, the products of bacterial metabolism, were extracted from the colonic digesta and derivatised for analysis using liquid chromatography coupled with mass spectrometry (LC-MS) [40]. To the colon digesta, acetonitrile (HPLC grade, Thermo Fisher) was mixed with samples at a 2:1 ratio, followed by vortexing for 2 min and centrifugation for 15 min at 13,000× g. The supernatant was collected and derivatised as described by Li et al. [40]. The derivatisation was conducted in a 2 mL centrifuge tube by sequentially adding 50 µL each of the following solutions: 50mM 3-Nitrophenylhydrazine hydrochloride (3-NPH·HCl), 50 mM 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC), and 7% pyridine to 100 µL of supernatant. Samples were then placed in a water bath (30 °C) for 30 min with shaking. After completing the reaction, 250 µL of purified water was added to each sample.

Derivatised SCFAs were separated by an Atlantis PREMIER BEH C18 AX column (2.1 × 150 mm, 1.7 µm, Waters, Rydalmere, NSW, Australia) on a Vanquish UPLC system (Thermo Scientific, Waltham, MA, USA) with two compartments: the sample tray was maintained at 15 °C and the column at 55 °C. Mobile phase A consisted of 0.1% formic acid in water, and mobile phase B consisted of 0.1% formic acid in acetonitrile. Gradient elution was performed by increasing mobile phase B from 5 to 50% over 20 min and to 100% over the next 3 min. The flow rate was 0.2 mL/min, and the injection volume was 5 µL. The mass spectrometer was operated in positive (4.2 kV) ionisation mode over a mass range of 120–1200 m/z with a resolution of 60,000 (FTMS mode). SCFAs were identified using Xcalibur 4.0 (Thermo Fisher) based on retention time and accurate mass matching. The quantification of SCFA was performed using external calibration curves.

2.5. Statistical Analyses

All experiments were conducted in biological triplicate and three technical replicates. Colonic pH, bacterial-species-specific qPCR, and SCFA concentration data analyses were conducted via one-way ANOVA with Tukey’s post-hoc test (GraphPad Prism v.9 for Windows, GraphPad Software, San Diego, CA, USA). Data are expressed as mean ± S.E.M., and a p-value of ≤0.05 was considered statistically significant. Microbial community analyses (relative abundance, Shannon index, Simpson index, Bray–Curtis dissimilarity matrix) were performed in R 4.4.1 using MicrobiomeAnalyst R and vegan packages (https://github.com/xia-lab/MicrobiomeAnalystR and https://cran.r-project.org/web/packages/vegan/index.html, respectively, accessed on 1 August 2024) for RStudio [41]. Significant differences were tested with permutational multivariate analysis of variance (PERMANOVA) using 999 permutations. Spearman correlation coefficients and principal component analysis (PCA) were conducted with Tidyverse and FactoMineR [42] (https://cran.r-project.org/web/packages/tidyverse/index.html and https://cran.r-project.org/web/packages/FactoMineR/index.html, respectively, accessed on 1 October 2024) for R Studio to assess the correlations between treatment, microbiota, and metabolite profiles. Heatmaps were generated in ggplot2 (https://cran.r-project.org/web/packages/ggplot2/index.html, accessed on 3 October 2024) in Tidyverse.

3. Results

3.1. Colonic pH

The in vitro gut model used in this study was a four-stage model with gastric, small intestinal (duodenum, jejunum/ileum), and colonic stages. The latter was inoculated with the standardised pooled infant faecal inoculum before inoculation with milk or control samples, comprising either sterile water (control), raw or Past-HM, or cow or goat milk infant formulas (CIF and GIF, respectively). At 0 h, aliquots taken from the colonic stage all had a similar pH of 6.4–6.5 (Figure 2). As incubation progressed, except for the water control, the pH decreased to pH 4.5–5 (all HM- and IF-inoculated samples). At 24 h, the lowest pH was recorded for GIF (4.49), followed by that for pasteurised HM (4.80). The pH decrease was consistently greater for pasteurised HM samples than for unpasteurized HM samples, although this difference was not statistically significant. However, the pH decrease for GIF was significantly greater (p < 0.05) than that for raw HM at 24 h, and for CIF and pasteurised HM at 48 h (Figure 2).

3.2. Bifidobacterium and Lactobacillus spp. Changes Measured by qPCR

Over the 48 h sampling period, there was an increase in bacterial growth (determined by qPCR and non-specific 16S rDNA primers) across all taxa in the colonic medium. When bacterial quantification by qPCR was repeated using the genus-specific primers, results showed that Lactobacillus and bifidobacteria levels (CFU/mL) increased throughout the incubation period (Figure 3 and Figure 4). At 24 h, Lactobacillus counts were highest for the control, presumably due to their presence in high numbers in the faecal inoculum, and numbers were also significantly higher for GIF compared with CIF (Figure 3). At 48 h, no statistically significant differences were seen except for CIF, which showed lower Lactobacillus numbers (p < 0.05) compared to the remaining samples (Figure 3). Bifidobacterium numbers were initially similar at time zero (6 to 7 log CFU/mL, which was higher than the numbers seen for lactobacilli), with no statistically significant differences between groups (Figure 4). At 24 h, however, the HM and GIF samples (but not CIF) showed a significant increase in levels compared to the control. At 48 h, the same pattern was seen; however, the difference for GIF was no longer statistically significant. At both 24 and 48 h, levels of bifidobacteria for CIF were consistently less than those seen for GIF or HM (Figure 4).

3.3. Microbial Community Changes

To evaluate microbial changes (by microbial community profiling, or MCP), 16S rRNA gene sequencing was conducted on DNA extracted from the colonic digesta. A total of 2060 operational taxonomic units (OTUs) that belonged to eight phyla and 165 genera were identified across all samples. Since our model was designed with stool samples from 3-to-6-month-old babies, both breastfed (n = 3) and formula-fed (n = 2), we expected that a diverse microbial community would be present and further develop after bacterial metabolism of the various substrates present in raw and pasteurised HM and each IF.

At the start of the incubation (time point 0 h), all samples had a similar microbiota composition, dominated by Bifidobacterium (>50%), Collinsella, and to a lesser extent Escherichia-Shigella, presumably reflecting the microbial composition of the pooled faecal inoculum (Figure 5). During fermentation, at 24 h and 48 h, the control samples showed a decline in relative levels of Bifidobacterium and an increase in Escherichia/Shigella and Bacteroides. Relative levels of other Enterobacterales (non-Escherichia/Shigella), which might include pathogens, were also greatest for the control at 48 h. In contrast, all milk-inoculated (HM or IF) samples showed higher relative abundance of Bifidobacterium (medium blue bars), with GIF showing the highest relative abundance at 48 h. Interestingly, at 24 h, the Past-HM sample had a higher relative abundance of bifidobacteria than raw HM. In addition, at 24 h, bifidobacteria levels for CIF were noticeably lower than those for the other samples, with correspondingly greater levels of Escherichia/Shigella. Lactobacilli were present in low proportions compared to the other bacterial taxa, especially compared with levels of bifidobacteria. Enterococcus was also present in relatively low amounts. Both qPCR (Figure 3) and MCP (Figure 5) results consistently showed that HM and GIF were better at stimulating the growth of bifidobacteria than CIF.

Alpha diversity, a measure of the OTU diversity within each group, was assessed by two different methods. First, the Shannon entropy index (H′) was calculated to assess diversity in microbiota at the different sampling times (Figure 6). At 0 h, inoculation with HM increased the microbial diversity compared to the control, as expected. At 24 h and 48 h, control samples showed a significant increase in diversity relative to 0 h, while inoculation with HM and IF caused a significant decrease in diversity (p < 0.05 compared to control). This was most likely due to the growth of Bifidobacterium spp. outcompeting the other bacteria. At 48 h, the control group’s diversity index remained higher than all four milk-inoculated groups, indicating that feeding the model with HM or IF leads to a relatively lower species diversity. Supporting this, PERMANOVA analysis using Bray–Curtis showed that this difference was statistically significant (Pr > F) (p = 0.004).

The Simpson’s dominance index (D) (Figure 7) shows the richness and evenness of species within a microbial community. At the initial sampling (0 h), the diversity across all samples was relatively uniform, with D values of around 0.4. By 24 h, a significant increase in diversity was observed for the control sample, with D dropping to nearly 0. Conversely, all HM and IF samples showed an increase in D indices, indicating reduced species diversity and dominance by a limited number of species compared to the control. The results (Figure 7) further showed that feeding the model with Past-HM reduced gut microbial diversity to a greater extent than feeding with raw HM. Inoculation with IF samples also led to reduced diversity, with only a few genera dominating (Escherichia-Shigella for CIF at 24 h, and Bifidobacterium for GIF at 48 h).

The Bray–Curtis dissimilarity metric was used to compare the different inocula (groups) based on relative abundance at the genus level (β-diversity), followed by hierarchical clustering based on the distance of OTUs. Results indicated three distinct clusters (Figure 8). Samples at 0 h formed one cluster, suggesting similarity of samples at baseline regardless of the inoculum type. A second cluster included just the control samples at 24 h and 48 h, which remained essentially unchanged over time. The third cluster contained all remaining HM and IF samples at 24 and 48 h. Within this, raw HM samples formed a distinct sub-cluster, regardless of time point. Overall, the tree structure indicated that HM treatment and IF types play a greater role than incubation time in shaping the microbial community. The deep branching for control samples at 24 h and 48 h reflected their higher dissimilarity relative to all four milk inoculations, supporting an effect due to feeding treatments.

3.4. Short-Chain Fatty Acid Production

Concentrations of SCFAs in the gut model colonic stage were determined over time (

Table 1. Average concentration of short-chain fatty acids in the colonic stage of the in vitro infant gut model, over 0, 24, and 48 h for control, raw, and pasteurised HM, GIF, and CIF samples. Results are the mean ± S.E.M (µg/mL) for n = 3 replicates. Statistically significant differences between groups at each time point are indicated by asterisks (p ≤ 0.05).

Fatty Acid Concentration (µg/mL)
Time (h)	Control			Raw HM			Pasteurized HM			GIF			CIF
	0	24	48	0	24	48	0	24	48	0	24	48	0	24	48
Acetic acid	4.44 ± 1.07	94.2 ± 12.74	102.2 ± 22.8	5.05 ± 2.28	125.6 ± 15.7	153.7 ± 23.4	4.58 ± 1.21	144.0 ± 18.9	163.7 ± 12.0 *	2.52 ± 0.07	135.5 ± 13.1	181.8 ± 4.85 *	4.21 ± 1.20	133.1 ± 7.00	152.9 ± 8.82
Propionic	0.19 ± 0.03	46.1 ± 6.82	66.1 ± 13.1	0.16 ± 0.04	2.37 ± 1.36	4.99 ± 2.03	0.16 ± 0.03	0.58 ± 0.06	0.49 ± 0.13	0.11 ± 0.004	0.49 ± 0.07	0.40 ± 0.08	0.12 ± 0.01	1.06 ± 0.24	2.09 ± 1.46
Butyric	0.16 ± 0.03	0.56 ± 0.15	12.7 ± 2.34	0.39 ± 0.04	0.18 ± 0.02	0.21 ± 0.04	0.09 ± 0.00	0.04 ± 0.00	0.04 ± 0.00	4.35 ± 0.18	1.56 ± 0.23	2.04 ± 0.05	0.31 ± 0.012	0.17 ± 0.002	0.19 ± 0.02
Valeric	0.03 ± 0.00	0.06 ± 0.01	1.04 ± 0.19	0.03 ± 0.01	0.01 ± 0.00	0.01 ± 0.00	0.03 ± 0.00	0.01 ± 0.00	0.01 ± 0.00	0.07 ± 0.002	0.019 ± 0.002	0.02 ± 0.001	0.019 ± 0.002	0.006 ± 0.0005	0.008 ± 0.001
Total	4.06 ± 0.75	132 ± 24.6	152 ± 40.4	5.83 ± 2.41	128 ± 15.4	159 ± 24.0	5.49 ± 1.28	145 ± 19.0	165 ± 12.1	7.22 ± 0.12	138 ± 13.5	184 ± 4.89	4.85 ± 1.19	134 ± 6.99	155 ± 10.31

). Levels of SFCAs fluctuated based on their continuous production and utilisation by bacteria in a dynamic microbial environment. At 0 h, all groups (colonic samples) had relatively low levels of SCFAs (<1 μg/g). As fermentation progressed, acetic acid became the dominant SCFA present. The concentrations of acetic, propionic, and butyric acids within groups increased at 24 h relative to time point zero, except for butyric acid in the GIF sample, which was highest at 0 h (4.35 μg/g; p = 0.0076 when compared with control). Levels of propionic and butyric acids at 24 h and 48 h relative to the controls were lower. The control had the highest levels of butyric acid at 48 h, followed by GIF (p = 0.0003 vs. control). CIF had lower butyric acid levels than GIF at 48 h (p = 0.0005), while HM and CIF showed similar levels to the control at 48 h. Propionic acid levels significantly increased in the control over time. Although HM, GIF, and CIF showed an increase in propionic acid concentration, the levels were significantly lower than for the control at 24 h (p < 0.01 for all samples vs. control). Both SCFAs were likely utilised by clostridia (or other anaerobic bacteria) present in the medium. Pasteurisation of HM resulted in elevated levels of total SCFAs compared to levels in raw HM at 24 h and 48 h. At 48 h, acetic acid levels in Past-HM (163.7 ± 12.0 μg/g) were higher than in the controls (102.2 ± 22.8 μg/g) (p < 0.05), and also higher than those in raw HM. Overall, SCFA levels were greatest for the GIF-inoculated samples. PERMANOVA analysis showed that changes in SCFA production were significantly affected by both the sample factor (p < 0.05) and time factor (p < 0.001).

3.5. Multivariate Analysis of Gut Microbiota Composition and SCFA Production

Spearman correlation coefficient analysis was conducted to assess the correlation between the relative abundance of bacterial genera and types of SCFA produced (Figure 9). In the plot, the colours and sizes of circles indicate the strength and magnitude of different correlations. Bifidobacterium spp. showed a strong positive correlation with levels of acetic acid; conversely, Bacteroides and Phascolarctobacterium showed a strong negative correlation with acetic acid levels, presumably because they would use it as a growth substrate. Bacteroides and Phascolarctobacterium genera showed a positive correlation with other SCFAs, except for propionic acid. Escherichia-Shigella spp. also showed a strong positive correlation with propionic acid and a weaker positive correlation with acetic and butyric acids.

Principal component analysis (PCA) was performed on the results (Figure 10). Dim1 (OTU variable) proved the most critical dimension for distinguishing between groups, capturing 72.6% of the total variance, while Dim2 (SCFAs) accounted for 19.8% of the variance. The controls formed a distinct group, separate from both the pasteurised and raw milks, primarily along Dim1. At 0 h, all samples clustered together on the negative side of Dim1. After fermentation, the milk samples and controls formed two distinct clusters. The control groups at 24 h and 48 h formed a separate cluster on the positive side of Dim1, showing substantial differences from the other groups. The pasteurised and raw HM groups at 24 h and 48 h had differences that were more pronounced along Dim2. These results suggest that the difference between controls (no milk inoculation) versus tests (milk inoculations) was greatest, while differences within each group (i.e., within each test inoculation), though significant, were relatively less significant.

4. Discussion

This study demonstrates that the feeding type, whether HM, Past-HM, or IF, plays a decisive role in shaping the gut microbiota structure and function. In a simulated colonic model inoculated with pooled faecal samples from 3–6-month-old healthy infants, we observed distinct microbial succession patterns and SCFA profiles across feeding groups. Both HM and IF selectively enriched Bifidobacterium and reduced alpha diversity relative to controls. Holder pasteurisation had no significant effect on the functional capacity of HM. Interestingly, GIF more closely mirrored HM in microbial outcomes than CIF. Together, these findings highlight the importance of early nutritional inputs in guiding the trajectory of gut microbial colonisation during infancy.

Bifidobacteria are critical components of the neonatal and infant HM gut microbiome. They colonise the gut soon after birth, and this is crucial for proper digestion, gut health, pathogen resistance, and immune system development [2,3,19,43]. The type of infant feeding affects their levels and (more broadly) the gut microbiome. In this study, using an in vitro infant gut model, we showed that in the absence of HM or IF, bifidobacterial numbers in the colonic stage of the model declined over time (Figure 4). Conversely, there were greater bifidobacteria numbers in milk- or IF-inoculated digesta. Further, our results show that pasteurisation of HM does not adversely impact its properties for bifidobacterial growth, presumably because HMOs are not heat-labile. They might even increase in concentration after heat treatment [44].

In addition, the results of our study showed that the Escherichia-Shigella group became more prominent with CIF, reaching a higher relative abundance compared to the other milks or GIF at the 48 h mark. Levels of lactobacilli were significantly lower (p < 0.05) in the colonic stage when CIF was used, compared to levels for the controls or HM or GIF (Figure 3). This might be due to the higher concentrations of oligosaccharides in GIF, which are more abundant and structurally more similar to those in HM than in CIF [26]. Supplementation of IFs with HMOs has also been shown to stimulate the growth of bifidobacteria and lactobacilli, both in vitro and in clinical trials [21].

The increases in bifidobacteria and lactobacilli in the gut model system were reflected by changes in pH, itself a useful indicator of gut health [45,46]. The colonic pH at baseline was similar across all samples, at pH 6.5. (Figure 2) This is similar to the faecal pH of newborns in modern environments where breastfeeding is not generally maintained by mothers long after birth [47]. Our results also suggest that pasteurised HM leads to a greater level of fermentation in the colon compared to raw HM; however, this requires further investigation. In vitro studies have shown that lower pH conditions suppress Bacteroides spp. and favour the growth of Bifidobacterium spp. Such acidic environments selectively favour the growth of beneficial taxa and inhibit others, including potential pathogens [27,48]. Clinical data also align with these findings, showing that higher stool pH correlates with increased relative abundance of Clostridioides spp. and Enterobacterales and a reduction in Bifidobacteriaceae [49]. Nonetheless, a lower pH might also create favourable conditions for certain harmful microbes, too. In summary, pH levels affect the composition and abundance of specific microbial taxa [49,50].

An increase in the relative abundance of Bifidobacterium spp. following co-culture with pasteurised HM may be attributed to the formation of prebiotic compounds such as lactulose and HMOs [51,52]. Studies by Daniels et al. [44] and de Segura et al. [53] have documented the generation of such compounds following pasteurisation, which might explain the observed increase in Bifidobacterium spp. Additionally, pasteurisation could influence or induce the formation of bioactives that inhibit pathogen growth. This dual action of pasteurised HM, simultaneously promoting beneficial bacteria and suppressing harmful bacteria, highlights the nuanced role of the infant diet in shaping gut microbiota.

At 48 h, in our model system, the highest Bifidobacterium relative abundance (Figure 5) was observed for GIF. This is presumably due to the presence of greater levels of oligosaccharides with higher similarity to HMOs in goat’s milk and GIF [26,30]. Oligosaccharides in CIF are structurally different and lower in concentration and diversity than those in GIF [54]. The ability of GIF to promote the growth of Bifidobacterium has also been shown to be comparable to that of HM in other studies [36,55]. The roles of Bifidobacterium spp. and Lactobacillus spp. in forming a healthy gut microbiome, preventing various health issues later in life, have been described [56,57,58]. For example, some Bifidobacterium species, especially B. longum subsp. infantis, produce bacteriocins that inhibit harmful bacteria, thereby enhancing gut health and stability [59]. Furthermore, Bifidobacterium supplementation has been shown to improve gut microbiome diversity and intestinal barrier function in preterm infants [60]. Lactobacillus species also contribute to a healthy gut environment by producing lactic acid, which lowers pH, helping to mitigate dysbiosis and inhibit pathogens [60,61]. While Bifidobacterium and Lactobacillus are vital for a healthy gut microbiome, the complexity of microbial interactions underscores the importance of a diverse microbiota and its metabolites for optimal health outcomes in infants [62].

Microbial community diversity results provide valuable insights into the impact of feeding on the gut microbial diversity of babies. In this study, microbial diversity was initially similar across samples, but by 24 h, the control group had a higher diversity, while milk samples, Past-HM and GIF, exhibited reduced diversity, with greater relative proportions of Bifidobacterium spp. Further in vivo studies are necessary to explore the long-term implications of these findings for infant gut health and to better understand the underlying mechanisms driving these changes in microbial diversity.

SCFAs play a crucial role in gut health, acting as a chemical interface between bacteria and the host. SCFA production is influenced by microbial fermentation of dietary substrates, with different pathways depending on the bacterial strains and available substrates [63]. In this study, acetate was the predominant SCFA across all samples, with consistent increases over time, particularly in milk-treated groups, where levels were significantly higher compared to the control. Although the overall SCFA profiles were similar between raw and pasteurised milk groups, the pasteurised samples exhibited a significant increase in acetate levels at the 48 h mark. This suggests that components of pasteurised human milk may be more readily fermentable by certain gut bacteria, leading to enhanced SCFA production. Moreover, the observed increase in Bifidobacterium spp. at 24 h may have contributed to the higher acetate production at 48 h [64,65].

Bifidobacterium species utilise fucosylated and sialylated HMOs to produce acetate, lactate, and formate [66,67]. The higher abundance of Bifidobacterium observed in this study aligns with the elevated acetate levels in milk digesta compared to the control. Clinical studies support this correlation, showing that HMO-fed infants have a higher relative abundance of Bifidobacterium spp. compared to formula-fed infants, who have a gut microbial community with a higher abundance of Firmicutes and butyrate-producing bacteria [28]. Additionally, studies have shown that when Enterobacterales (e.g., E. coli, Shigella, and Klebsiella) dominate the microbiota, acetate levels decrease while succinate levels increase [68]. The literature also reports a positive correlation between Clostridiales abundance and butyrate levels [67]. In this study, the observed decrease in butyrate concentrations after 24 h likely reflects the low levels of Clostridiales in the starting faecal inoculum.

Multifactorial analysis revealed additional information on the relationship between microbiota composition and SCFA production. Spearman correlation showed that Bifidobacterium spp. was positively associated with acetate, while Bacteroides and Phascolarctobacterium were negatively associated with acetate but positively correlated with other SCFAs. These contrasting patterns highlight the diverse metabolic capacities of bacterial taxa in shaping the gut SCFA profile. Such insights are critical for guiding targeted strategies to modulate the microbiota and optimise SCFA production for health benefits.

Although the in vitro gut model cannot fully replicate the physiological complexity of the human body, it remains a valuable approach for investigating microbiota–host interactions under controlled conditions, while avoiding the ethical challenges of infant studies. Such models are particularly useful for assessing the nutritional and health-promoting properties of HM and IF, and whether HM’s beneficial properties are lost due to heat treatment. HM plays a critical role in reducing the risk of infections and inflammatory diseases, especially in vulnerable premature and low-birth-weight infants. While heat treatment may reduce some of these benefits, the overall nutritional and physiological advantages of HM appear to outweigh these losses.