Maternal diet shapes infant microbiota and defensive capacity against infections in early life via differential human milk composition

Summary

Background

Maternal nutritional status and dietary profile during pregnancy and lactation have short- and long-term impacts on offspring health. However, there is an incomplete understanding of the mechanisms behind these health effects. This study aims to assess the effect of maternal diet on the health of offspring by examining to unravel the impact of maternal diet on offspring health outcomes and evaluate the link between maternal nutrition, human milk immune components and neonatal colonisation as potential mechanisms that mediate the influence of maternal diet in the incidence of infant infections.

Methods

To assess this objective, we used two complementary approaches by which a clinical observational study based on the MAMI birth cohort guided a preclinical interventional analysis using a neonatal rat model of rotavirus-induced gastroenteritis.

Findings

The findings in both approaches demonstrated that a maternal diet rich in plant-based protein, fibre and polyunsaturated fatty acids, was linked to reduced incidence and severity of infections in offspring that would be mediated by beneficial modulation of the gut microbiota and immune system. Specifically, in the suckling rats, a predominant Th1 immune response and an enhanced virus-specific response were observed. Moreover, human milk IgA and rat milk IgG2c played a key protective role that complemented the effects of maternal diet.

Interpretation

These results strengthen the importance of maternal diet during pregnancy and lactation supporting infant health.

Funding

The study was supported by LaMarató-TV3 (DIM-2-ELI, ref. 2018-27/30-31).

Research in context.

Evidence before this study

• -A diet rich in vegetables, polyunsaturated fatty acids, and fibre is associated with health benefits for children, adults and pregnant women.
• -Human milk IgA plays several protective and immunomodulatory roles in the offspring.
• -The influence of maternal diet on the immune composition of human milk is not clear.

Added value of this study

• -A maternal diet rich in plant-based protein and dietary fibre and low in saturated fats reduces the severity and incidence of offspring infections.
• -Following a maternal diet low in fibre and rich in animal protein, cholesterol and MUFA together with having low levels of human milk IgA seemed to predispose infants to infections in early life. Moreover, in such circumstances, microbial diversity at one month of age could serve as a crucial indicator of the probability of infections during the early stages of life.
• -The IgA in human milk may have a protective role against infant infections, while in rat breast milk IgG2c is the mediator.
• -A maternal diet enriched in plant protein and dietary fibre and low in saturated fats induced gastrointestinal and immune system maturation in lactating rats, with a predominant Th1 response and an increased specific response against the virus causing the infection.

Implications of all the available evidence

• -The findings of this research can help in offering improved evidence-based dietary guidance for pregnant and lactating mothers during this pivotal stage of infant development, to counteract early life infections.
• -This research can also aid in better understanding the relationship between maternal diet, bioactive components of breast milk, infant microbiota, and infant immunity. However, further research is needed to draw more robust conclusions and to better understand the mechanisms involved.

Introduction

Diet is considered key in preventing health risks in both adult and infant populations.¹ Adequate nutrition is needed to establish optimal host-microbe interactions² and proper host functioning, including immune response.³ During pregnancy, a determinant period for the human development, an adequate maternal diet is essential for an optimal infant health.4, 5, 6, 7 Nevertheless, the mechanisms underlying these health effects are still poorly understood.

Maternal dietary macronutrients are well known to affect foetus development and the infant during lactation. During gestation, long chain-polyunsaturated fatty acids (LC-PUFAs) are trans-placentally transferred and involved in the development of the foetus.⁸ After birth, the newborn’s energy intake is mainly supplied by human milk lipids, which are also dependent on the maternal diet.⁹ Despite the importance of the maternal diet for foetal and infant development, little is known about the influence of the maternal diet on human milk (HM) immune components, such as immunoglobulins (Igs), cytokines (CKs) and adipokines, which are known to play a role on newborn immune development.¹⁰

This study hypothesised that the maternal diet during gestation and lactation has an impact on the health of offspring by influencing HM immune components and neonatal microbiota colonisation. While previous studies within the same cohort demonstrated associations between maternal diet and both HM oligosaccharide profile¹¹ and microbiota profile,¹² the present study extends these findings by exploring the potential link between maternal dietary patterns and immune components in HM, addressing also the impact of this diet- HM immune components link on the infant defensive capacity. The results from the observational cohort showed that key maternal dietary components within a usual diet are linked to infant infection prevalence. Thus, to validate such observation, we performed a preclinical maternal dietary intervention using an acute offspring gastroenteritis model triggered by an oral administration of rotavirus (RV).13, 14, 15 Murine maternal diets were adapted from the specific dietary compounds observed in the clinical observational study: one related to lower infant infections (rich in fibre and vegetal protein) and the other related to higher infant infections (low in fibre, vegetal protein and rich in fats). The preclinical dietary intervention allowed the confirmation of the clinical observations and provide stronger evidence supporting the relationship between the maternal diet with infant infection resolution, immunological response dynamics and intestinal microbiota composition.

Methods

Experimental model and subject details

Human clinical study

Participant families and clinical records

From a previous study¹⁶ where 120 mother-infant pairs were included, a subset of eighty-five healthy mothers-infant pairs from the MAMI birth cohort¹⁷ with available milk samples and infant faeces as well as clinical and dietary data were included in this study. Infant data including mode of birth, sex, breastfeeding practices, antibiotic exposure, growth parameters, and also, number of infections: gastrointestinal infections (gastroenteritis, diarrhoea), respiratory infections (bronchitis, pharyngitis, nasopharyngitis, and pneumonia), and others (otitis, conjunctivitis, etc) from birth to 12 months, were available from clinical records. Maternal clinical data was also available.

Maternal dietary records and clusters

Maternal dietary records were collected using a 140-item Food Frequency Questionnaire (FFQ) about their regular diet during the pregnancy¹⁸ allowing us to obtain macronutrient and micronutrient intakes per day. The FFQ covers usual foods and recipes, as well as the frequency of consumption (daily, weekly, monthly, or during the end of pregnancy and birth time), the number of times the participant consumed a particular food item, the median portion and the size of each participant’s portion. All the food surveys were conducted by trained nutritionists. FFQ information was analysed using data from the Food Composition Tables developed by the Centro de Enseñanza Superior de Nutrición Humana y Dietética.¹⁹ The intake of soluble and insoluble fibre types was determined by using the Marlett food composition tables.²⁰ Information about the use of probiotic, prebiotic and/or supplements were also collected during the interview. A dietary intake-based clustering using Jensen-Shannon distance and partitioning around medoid (PAM) clustering to find different dietary profiles global dietary information including mono-, poly-unsaturated and saturated fatty acids, animal and vegetal proteins, carbohydrates, polysaccharides and soluble and insoluble fibre. The contribution of these nutrients to the total energy was calculated in percentage and used for the clusterization. The optimal number of clusters was calculated by the Calinski-Harabasz index (CHI) and it revealed two distinct clusters as described.¹⁶ This information permitted the identification of two dietary patterns in this cohort, named Diet 1 and Diet 2, reported previously by our team²¹ (Supplementary Fig. S1).

Biosample type, collection and storage

Human milk (HM) samples were obtained at day 7 after delivery. Women received oral information and written instructions of protocols for self-collection of milk samples. To normalise the collection, milk samples were collected from one breast, cleaned with soap or 0.5% of chlorhexidine solution in order to reduce the skin bacteria, using a sterile pumper (discarding the first drops to normalise the collection) in the morning before lactation, as previously reported¹² and then frozen at −80 °C after the collection. Infant faeces were also collected longitudinally at 7 days, 1 month, 6 months and 12 months of life, as reported elsewhere.^21,22 Infant faecal samples were collected in sterile containers, by the parents, at home and in the morning, following detailed instructions, stored at −20 °C and transported within 24 h after collection, for later storage at − 80 °C until analysis.

Preclinical study

Animals and conditions

Eight-week-old female (n = 27) and male Lewis rats (n = 15) were obtained from Janvier (Le-Genest-Saint-Isle, France) and maintained in polycarbonate cages (2 rats per cage) in the animal facilities of the Diagonal Campus at the University of Barcelona (UB) one week before mating. After mating, females were individually housed in cages. Animals had access to water and food ad libitum. Conditions of humidity and temperature were controlled in a 12 h light-12 h dark cycle. To control the influence of diets and virus inoculation, dams and pups were weighed and monitored daily.

Rat diet composition

The diets for the preclinical intervention were adapted considering the human dietary profiles and provided by Inotiv (Inotiv Tecklad™ Diets, West Lafayette, IN, USA). All food was appropriate for gestation and lactation, but the purpose was to have a reference diet (RD (AIN-93G)) and two other diets: a diet (D1) enriched in plant protein, fibre, and fish oil and a diet (D2) with a higher quantity of animal protein and lard compared with the other dietary profiles, mimicking the observation in the human study (Supplementary Table S1).

Virus and infection challenge

Simian SA-11 (produced by the Enteric Virus Group, UB) was the RV strain selected for the project. SA-11 was intragastrically inoculated (2 × 10⁸ tissue culture infectious dose, TCID₅₀/rat in 100 μL of PBS) in pups at day 5 of life, as in previous articles,^13,23, 24, 25 1 h after separation from their dams. Low-capacity syringes (Hamilton Bonaduz) were used in the oral inoculation.

Preclinical study experimental design

Three different diets (RD, D1, and D2) were provided to the female rats during the gestation (21 days) and the lactation (21 days) periods. 9 female rats out of a total of 27 were randomly assigned to each dietary group. The rats that get pregnant were included in the study (n = 5 for RD, n = 7 for D1, n = 8 for D2) and were allowed to deliver naturally. A pup’s birth was recorded as day 1 of life. The litters were adjusted to 9 pups/litter; when they were formed by 8 pups, the following litter within the same group was adjusted to 10 pups. Pups had free access to the nipples; 70% of the pups were intragastrically inoculated with RV (infected pups) while the remaining 30% received the same volume of phosphate-buffered solution (PBS) at day 5 of life (non-infected pups). These uninfected pups from the RD group were included in the present study as a reference group. Ideally, in a 9 litter-size group of animals, three pups from each litter were euthanised at day 8, three on day 21 (end of the suckling period) and three on day 28 (experimental endpoint) (Supplementary Fig. S1). In each point, one non-infected pup per litter was sacrificed. To avoid sex-related bias, animals were allocated in each group and time point to achieve a balanced female-to-male ratio (∼50:50). Euthanasia was performed by intramuscular anaesthesia with ketamine (90 mg/kg) (Merial Laboratories S.A., Barcelona, Spain) and xylazine (10 mg/kg) (Bayer A.G.) and, subsequently, exsanguination by cardiac puncture. Pups’ sample collection at day 8, 21, and 28 consisted of blood, stomach content, caecal content and small intestine collection. Additionally, Breast milk (BM) was obtained at day 21.

Rat plasma collection

Maternal blood samples at day G0, which is one day before mating, were obtained from the saphenous vein in heparin tubes. Blood samples from dams and pups were collected by cardiac puncture in heparin tubes at day 8, 21 and 28. The haematologic composition was analysed immediately using a Spincell haematology analyser (Spincell 3, Spinreact, Girona, Spain). The blood was centrifuged (10,000×g for 5 min at room temperature) and the plasma was collected and stored at −80 °C until Ig quantification.

Rat milk sample collection

Milk of each dam was collected at day 21 just before euthanasia. Mothers were separated from pups and were immediately intramuscularly anesthetised with ketamine (10 mg/100 g) (Merial Laboratories S.A.). After 30 min, 2 U.I. of oxytocin (Syntocinon 10 U.I./mL, Alfasigma, Bologna, Italy) were intraperitoneally administered as in previous articles.²⁶ Five to 10 min later, gentle manual stimulation from the base to the top of the teats was performed to express milk into sterile microtubes taking the milk drops by a pipette with sterile pipette tips. Centrifugation (12,000×g for 5 min at 4 °C) was performed to collect the intermediate aqueous phase (lactic serum), which was then stored at −80 °C until Ig quantification analysis.

Method details

Immunoglobulin, cytokine and adipokine measurement

Human samples

The quantification of immune factors (GM-CSF, IFN-γ, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-9, IL-10, IL-12p70, IL-13, IL-17A, IL-18, IL-21, IL-22, IL-23, IL-27, TNF-α, IgE, IgG1, IgG2, IgG3, IgG4 and adiponectin) was performed by ProcartaPlex™ multiplex immunoassay (Thermo Fisher Scientific, Vienna, Austria), following the manufacturer’s instructions as in previous studies.27, 28, 29 The 96-well plates were run on a Luminex Instrument and analysed in ProcartaPlex Analyst Software (MAGPIX® analyser, Luminex Corporation) at the Flow Cytometry Unit of the Scientific and Technological Centres of the University of Barcelona (CCiT-UB). The quantification of leptin was performed by a Quantikine® Colourimetric Sandwich ELISA Kit (R&D Systems, Minneapolis, MN, USA) following the manufacturer’s instructions. Data were analysed by Multiskan Ascent v2.6 software (Thermo Fisher Scientific, Vienna, Austria).

The quantification of IgA and IgM was performed by an ELISA kit from Bethyl Laboratories (Montgomery, TX, USA) and an ELISA kit from Cloud-Clone Corp. (Houston, TX, USA), respectively following the manufacturer’s instructions. These kits included the quantification of the secretory form of both types of Igs. Assay sensitivities are detailed in Supplementary Table S2.

Rat samples

The quantification of Ig (IgA, IgM, IgG1, IgG2a, 2b, 2c) in milk, mother’s plasma and pup’s plasma was performed by ProcartaPlex™ Multiplex immunoassay (Thermo Fisher Scientific, Vienna, Austria), as described above for human milk samples. Assay sensitivities are detailed in Supplementary Table S2.

Plasma (d8, d21, d28), milk (d21) and stomach content (d8) concentrations of anti-RV total Ig were determined by ELISA analysis as previously described.¹³ The highest dilution of the standard used corresponded to 1 arbitrary unit (AU).

Rat faecal collection and clinical evaluation after viral challenge

Faecal collection was performed daily (d4 – d13) from all groups. Stools were weighed and evaluated in terms of diarrhoea severity, scoring from 1 to 4 (diarrhoea index, DI) based on colour, texture and amount as follows: normal (DI = 1), loose yellow-green (DI = 2), totally loose yellow-green (DI = 3), and large amounts of watery faeces (DI = 4).³⁰ A score higher than 2 indicates diarrhoeic stools. For further profiling, such as faecal SA-11 shedding, the faecal samples were frozen at −20 °C. The percentage of diarrhoeic animals (%DA) indicates the incidence of diarrhoea, taking into consideration the number of animals with diarrhoeic faeces with respect to the number of animals in each group.

Faecal SA-11 shedding

Faecal samples from day 6 (first day post-inoculation) were homogenised using a Pellet Pestles Cordless Motor (Merck KGaA, Darmstadt, Alemania) with PBS (10 mg/mL) before being centrifuged at 170×g for 5 min at 4 °C. Supernatants were frozen at −20 °C until the quantification of SA-11 virus particles by ELISA, as previously described.^23,30 Titrated dilutions of UV-inactivated SA-11 particles (from 10⁶ to 10⁴/mL) were used for a standard curve.

Small intestine histology

Portions (1 cm) of the central section of the small intestine corresponding to day 8 (within the diarrhoea period) were dehydrated, paraffin-embedded and cut into 5 μm cross sections using a microtome (RM2135, Leica, Wetzlar, Germany). The sections were stained with haematoxylin-eosin as previously reported¹⁴ and with periodic acid-Schiff.³¹ Photographs of the small intestine morphometry were taken using a light microscope (Olympus BX41, Olympus Corporation, Shinjuku Tokyo, Japan) with a 100× objective. The number of goblet cells and the width, length, and area of the villi and the crypts depth were quantified by ImageJ software as in previous studies.¹⁴

Quantification of gene expression by real-time PCR

A portion (1 cm) of the central section of the small intestine was collected in a sterile microtube with RNAlater solution (Invitrogen, Carlsbad, MA, USA) during sample collection at day 8. As previously described,²⁴ sections were homogenised for 30 s in lysing matrix tubes (MP Biomedicals, Illkirch, France) using a FastPrep-24 instrument (MP Biomedicals). The RNeasy®Mini Kit (Qiagen, Madrid, Spain) was used to isolate the RNA and RNA purity and concentration were immediately obtained with a NanoPhotometer (BioNova Scientific S.L., Fremont, CA, USA). The thermal cycler PTC-100 Programmable Thermal Controller and the TaqMan® reverse transcription reagents (Applied Biosystems, AB, Weiterstadt, Germany) were used to obtain cDNA from the isolated RNA. Later, gene expression with real-time PCR (ABI Prism 7900 HT, AB) was performed using the specific PCR TaqMan® primers (AB): Afp (Rn00560661_m1, Inventoried [I]), Fcgrt (Rn00583712_m1, I), Prdm1 (Rn03416161_m1, I), Muc2 (Rn01498206_m1, I), Ocln (Rn00580064_m1, I) and Cldn2 (Rn02063575_s1, I). The housekeeping gene Gusb (Rn00566655_m1, I) was used to normalise relative gene expression using the 2-ΔΔCt method.³² Data was represented as the expression percentage in each group normalised to the RD group mean, which was established as 100%, as in previous studies.^29,33

DNA extraction from human and rat faecal samples and microbiota profiling

Total DNA was isolated from the human stool samples and from the rat caecal material (50–100 mg) using the Mater-Pure DNA extraction kit (Epicentre, Madison, WI, USA), with some modifications, as described in Selma-Royo et al.³⁴ After the DNA extraction, DNA was purified using a DNA purification kit (Macherey–Nagel, Duren, Germany) following the recommended protocol and the final DNA concentration was measured using a Qubit® 2.0 fluorometer (Life Technology, Carlsbad, CA, USA).

The composition of human faecal and rat caecal microbiota was assessed by sequencing of the V3–V4 variable region of the 16S rRNA gene with primers described by Klindworth et al.³⁵ Libraries were prepared following the Illumina 16S rRNA gene metagenomic sequencing library preparation protocol (Cod. 15044223 Rev. A). The libraries were then sequenced using a 2 × 300 bp paired-end run on a MiSeq-Illumina platform (FISABIO sequencing service, Valencia, Spain). After sequencing, the residual adaptors were removed using Trimmomatic v. 039³⁶ by the sequencing server.

Statistical analysis and computational analysis of microbiota profiling

All statistical analyses were performed using IBM SPSS Statistics 30 (IBM, USA) and R version 4.1.2 (R foundation, Austria). Data are expressed as the mean ± standard error of the mean (SEM) unless otherwise specified. Shapiro–Wilk and Levene’s tests were used to determine normality and homogeneity of data variance, respectively. Student’s t-tests and Mann–Whitney U tests were used to assess significant differences between the two groups (D1 vs. D2) in MAMI cohort, while X² tests were used to compare frequencies. One-way ANOVA followed by post-hoc testing (Dunnet T3 or Bonferroni depending on the homogeneity of data variance) and Kruskal Wallis tests followed by multiple comparison testing with Dunn’s test with Bonferroni correction were performed to assess significant differences between groups. To assess time-dependent variables, a repeated measures ANOVA test was performed. Additionally, to assess dependent variables (diet and infection in the clinical part), non-parametric Aligned Rank Transform for non-parametric factorial ANOVA (ART-ANOVA) was applied. The Spearman correlation coefficient was used to identify correlations between variables. Differences were considered statistically significant when p value < 0.05.

Bacterial diversity analysis was done using raw reads, which were quality controlled and filtered (Quality; 20 and length; 150 bp) using fastqc (v0.11.8) and trimGalore (v0.6.4_dev; https://github.com/FelixKrueger/TrimGalore). The paired-end reads with a minimum overlap of 30 bp were joined using Fastq-join.³⁷ The sequences were trimmed of primers and distal bases and singletons were removed with USEARCH v11.³⁸ zOTUs mapping to the human genome (GRCh38) using the Burrow–Wheeler Aligner in Deconseq v0.4.3 were filtered out. The resulting reads were denoised and chimeras were filtered with UNOISE3.³⁹ Taxonomic assignment of zOTUs was performed using the Ribosomal Database Project.⁴⁰ The zOTUs were aligned with MAFFT,⁴¹ and then we made a phylogenetic tree with FASTTREE⁴² that was then midpoint-rooted. Those reads classified as mitochondria and chloroplast were removed before the statistical analysis. The calculations of richness of microbiota composition (Observed, Chao1 and ACE) and evenness indices (Shannon and Simpson) for each sample were done using the “phyloseq” R package,⁴³ and differences by group were assessed using Kruskal–Wallis non-parametric tests. The same package was used to calculate the β-diversity, with the normalisation of the data to log-ratio, and representations based on Unifrac Unweighted and Bray distances that were plotted using ggplot2. Non-metric multi-dimensional scaling (NMDS) plots based on Bray–Curtis dissimilarity measures were used to visualise the AUC factor according to study diet group while projecting this variable and dominant genera vectors using the R package vegan.⁴⁴ Permutational analysis of variance (PERMANOVA) was performed with the Adonis2 function from the vegan package.⁴⁴ Furthermore, a canonical correspondence analysis (CCA) was performed using the vegan package at the genus level after percentage normalisation. To explore the potential microbial markers from each of the analysed groups, we assessed the LEfSe to identify microbial genera enriched in the two clusters.⁴⁵ An LDA score (log10) > 3 was considered significant. FDR Benjamini–Hochberg correction was added for pairwise comparison. We tried to identify two groups of variables whose balance was more associated with the response variable, using the R package Selbal,⁴⁶ as well as MaAsLin2,⁴⁷ which is a complete R package to efficiently determine the multivariable association between microbial communities.

To identify microbial features associated with different groups and their relationship with the microbiota, the selbal package was employed in the R environment. This package is designed to perform microbial feature balance analysis using penalised logistic regression models, allowing for the selection of relevant features in high-dimensional datasets. The selbal.cv function was used to perform cross-validation. The accuracy of the model was evaluated using the area under the ROC curve (AUC) as the optimisation criterion for logistic regression. This approach allows for the identification of microbial balances that are significantly associated with the variables, providing a deeper understanding of the interactions.

Ethics

The clinical human study was approved by the Ethical Committee of the Hospital Clínico Universitario, Valencia and Ethical Committee of Bioethics CSIC (ClinicalTrials.gov Identifier: NCT03552939; protocol described elsewhere¹⁷).

The preclinical procedures were approved by the Ethical Committee of the University of Barcelona (CEEA-UB, 255/18) and by the Catalan Government (Ref. 10933). The required sample size was estimated using the GRANMO program. IgG plasma concentration was selected as the primary outcome variable to ensure the detection of statistically significant differences between groups. The calculation assumed a two-sided alpha level of 0.05, a dropout rate of 5%, and included a minimum of three dams per group, consistent with previous studies, due to the considerable variability observed between litters.

Role of funders

The funding sources had no involvement in the study design, data collection, data analysis or interpretation, nor in the writing of the manuscript.

Results

Maternal diet influences newborn microbiota and infant infections during the first year or life

We first investigated the dietary patterns of the mothers in order to evaluate the effect of maternal diet on the infant's health outcomes, mainly on infections during first year of life. Based on dietary food frequency questionaries, two distinct maternal dietary profiles were identified in 85 mothers from the MAMI cohort (Supplementary Fig. S1A). The first cluster (D1, n = 33) was characterised by women with higher vegetal protein, fibre and carbohydrate intakes. The other maternal dietary pattern (D2, n = 52) was composed of mothers with a diet enriched in animal protein and lipids, especially cholesterol (Table 1). The values were under usual diet consumption and balanced diets. While no significant differences were observed in the clinical characteristics of the two groups, the prevalence of infections from birth to one year significantly differed between the two maternal dietary groups. D1 presented both a lower percentage of infants with infections (p = 0.018) and a lower total number of infections throughout the period (p = 0.040) (Fig. 1A and B, respectively) suggesting an impact of the maternal diet on infant defensive capacity against pathogens.

Table 1.

Maternal dietary intakes and mother-infant birth cohort characteristics of the two dietary groups.

Maternal dietary intakes	D1, n = 33	D2, n = 52	p value
Total protein (g/day)	106.21 ± 2.51	123.00 ± 2.54	0.23
Animal protein (g/day)	52.59 ± 2.66	76.94 ± 2.27	<0.001
Vegetal protein (g/day)	51.25 ± 1.45	42.66 ± 1.15	<0.001
Lipids (g/day)	101.99 ± 2.42	120.24 ± 1.99	0.19
Cholesterol (g/day)	254.09 ± 12.60	353.37 ± 10.16	0.003
SFA	21.26 ± 0.84	25.89 ± 1.20	0.13
TRANS	0.01 ± 0.001	0.01 ± 0.001	0.32
MUFA	41.22 ± 1.96	49.14 ± 1.61	0.05
PUFA	15.64 ± 0.53	17.67 ± 0.67	0.10
n-6 CLA	0.01 ± 0.002	0.003 ± 0.001	0.47
n-3 ALA	0.15 ± 0.02	0.13 ± 0.01	0.84
n-3 EPA	0.15 ± 0.01	0.19 ± 0.01	0.43
n-3 DPA	0.08 ± 0.01	0.09 ± 0.01	0.10
n-3 DHA	0.39 ± 0.04	0.44 ± 0.03	0.37
Carbohydrates (g/day)	285.38 ± 5.17	227.04 ± 3.96	<0.001
Polysaccharides (g/day)	146.95 ± 7.41	118.20 ± 3.39	0.16
Starch (g/day)	29.59 ± 2.05	28.13 ± 1.53	0.72
Lactose (g/day)	9.41 ± 1.40	14.77 ± 1.66	0.01
Total dietary fibre (g/day)	41.42 ± 1.39	31.42 ± 1.09	<0.0001
Cellulose (g/day)	9.80 ± 0.39	7.08 ± 0.30	0.70
Insoluble dietary fibre (g/day)	26.74 ± 0.89	18.38 ± 0.81	<0.0001
Soluble dietary fibre (g/day)	5.24 ± 0.19	3.70 ± 0.18	<0.0001
Insoluble hemicellulose (g/day)	9.45 ± 0.42	5.91 ± 0.31	<0.0001
Soluble hemicellulose (g/day)	3.32 ± 0.15	2.37 ± 0.13	<0.001
Insoluble pectin (g/day)	3.15 ± 0.18	2.32 ± 0.12	0.004
Soluble pectin (g/day)	1.69 ± 0.09	1.17 ± 0.07	0.001
Klason lignin (g/day)	4.04 ± 0.12	2.75 ± 0.12	<0.0001
Phytosterols (mg/day)	122.60 ± 5.85	131.67 ± 6.15	0.83
Polyphenols (mg/day)	2052.95 ± 122.18	1585.36 ± 68.25	0.90
Total retinoids (mg/day)	304.29 ± 16.93	415.16 ± 38.06	0.01
Mother-Infant birth cohort characteristics
Maternal characteristics
Pre-gestational BMI (kg/m2), mean ± SEM	22.07 ± 0.46	23.04 ± 0.51	0.19
Weight gain over the pregnancy (kg), mean ± SEM	12.05 ± 0.72	12.08 ± 0.66	0.98
Antibiotic during pregnancy, yes (%)	8 (24.24)	18 (34.62)	0.31a
Antibiotic at birth, yes (%)	15 (45.45)	17 (32.69)	0.24a
Perinatal antibiotic, yes (%)	18 (54.55)	28 (53.85)	0.95a
Gestational age (weeks), mean ± SEM	39.71 ± 0.20	39.51 ± 0.17	0.437
Mode of delivery: Vaginal birth, yes (%)	21 (63.64)	35 (67.31)	0.73a
First-time mothers, yes (%)	19 (57.58)	28 (53.85)	0.35a
Infant characteristics
Gender: Female, yes (%)	18 (54.55)	28 (53.85)	0.95a
Exclusive breastfeeding until 7 days, yes (%)	29 (87.87)	46 (88.46)	0.94a
Exclusive breastfeeding until 6 months, yes (%)	29 (87.87)	39 (75)	0.15a
Birth weight (kg), mean ± SEM	3.30 ± 0.10	3.23 ± 0.05	0.50
BMI z-score, mean ± SEM
Birth	−0.35 ± 0.18	−0.30 ± 0.12	0.81
7 days	−0.43 ± 0.18	−0.57 ± 0.11	0.50
15 days	−0.31 ± 0.17	−0.59 ± 0.13	0.20
1 month	−0.56 ± 0.15	−0.56 ± 0.15	0.98
6 months	−0.20 ± 0.15	−0.33 ± 0.12	0.51
12 months	0.51 ± 0.16	−0.01 ± 0.13	0.02

Dietary components were standardised by total energy intake to 2500 kcal/d. Student t tests^a and X² test were used to assessed differences between the groups on quantitative and categorical variables, respectively. p < 0.05 (shown in bold letters) was considered statistically significant. BMI, body mass index; SEM, standard error of the mean. The dietary groups (D1 and D2) were identified using a partitioning clustering algorithm based on the k-medoids method, with the optimal number of clusters determined by the silhouette method.¹⁶

Fig. 1.

Impact of maternal dietary components on infant infections, and also on HM immune composition. (A) Percentage of children with infections from birth to 12 months of age. Infections include both gastrointestinal and respiratory maladies. (B) Number of infants’ accumulated infections from birth to one year. The area under the curve (AUC) is indicated in figure. (C) Spearman correlations between the number of offspring infections (from birth to one year) and maternal dietary intake. Spearman correlation coefficients are denoted by the colour in the colour-bar. Bold frames represent statistically significant correlations (p < 0.05). M, month. (D) HM Th1/Th2 ratio according to diet profile and presence of infant infection (from birth to one year). (D1-I, n = 8; D1-NI, n = 24; D2-I, n = 24; D2-NI, n = 25). I, Infection; NI, no infection. Th1/Th2 ratio refers to the relationship between Th1 (IgG1 + IgG2 + IgG3) and Th2 (IgG4) Ig. (E) HM IgA concentration according to diet profile and presence of infant infection (from birth to one year). (F) HM IL-10 concentration according to diet profile and presence of infant infection (from birth to one year) variables. The values are represented on a logarithmic scale. The cytokine was not detected in all samples because the concentration fell below the detection limit in the multiplex assay. The detectability for each subgroup was: 41.7% in D1-NI, 37.5% in D1-I, 40.0% in D2-NI, and 54.2% in D2-I. (G and H) Spearman’s correlation test between HM IgA and the number of infections (from birth to one year) in (G) D1 mothers (n = 33) and (H) D2 mothers (n = 52). Results are expressed as percentages (A) and as mean ± SEM (B, D–F). X² test (A), Mann–Whitney U test (B) and ART-ANOVA followed by Kruskal Wallis test with multiple comparison with Bonferroni correction (D–F). p < 0.05 was considered statistically significant. SFA, Saturated Fatty Acids; MUFA, Monounsaturated Fatty Acids; PUFA, Polyunsaturated Fatty Acids; CLA, Conjugated Linoleic Acid; ALA, α-Linoleic Acid; EPA, Eicosapentaenoic Acid; DPA, Docosapentaenoic Acid; DHA, Docosahexaenoic Acid; M, month.

We tried to unravel which specific components from the maternal diet could be related to the observed protective effect. Thus, specific associations between dietary components and infant number of infections during the first year of life were identified by performing Spearman’s rank bivariate correlations. Maternal consumption of fibre (both soluble and insoluble) (ρ = −0.299, p = 0.007), polyphenols (ρ = −0.315, p = 0.004) and carotenoids (ρ = −0.256, p = 0.022) were negatively correlated, with the number of infections in the first 12 months of life. Contrary, the intake of cholesterol, n-3 α-linoleic acid (ALA), total protein and animal protein showed a positive association with the infant’s infection during this period (ρ = 0.285, p = 0.010; ρ = 0.232, p = 0.038; ρ = 0.247, p = 0.027; and ρ = 0.228, p = 0.042, respectively) (Fig. 1C).

Following, we hypothesised that the effect of diet on maternal human milk could be one of the mechanisms behind this effect. Therefore, we assessed the relationship between diet and HM immunological profile. Significant differences in the HM Th1/Th2 ratio and IgA concentration were found when considering the interaction between maternal diet and infant infection later in life (Fig. 1D and E). Specifically, in the D2 group, mothers with infants with infections (D2-I) had a higher HM Th1/Th2 ratio compared to the mothers with infants without infections (D2-NI). These findings suggest that high concentration of these HM factors, together with a diet rich in animal protein and saturated fatty acids (SFA) are associated with more infections in the offspring. This was not observed in the HM from mothers with a dietary pattern D1. Similarly, HM IgA concentration also differed between dietary profiles when considering infant infections. HM IgA was lower in D2 mothers whose children had infections compared to those belonging to this group but that did not suffer infections (Fig. 1E). Indeed, only among mothers from D2 groups and not from D1, a negative correlation was found between the HM IgA and the number of infections on the infant (Fig. 1G and H). In addition, IL-10 tended to be increased in the non-infected groups (p = 0.073) (Fig. 1F). The concentration of the other measured immune components did not show differences between the two dietary profiles (Supplementary Tables S3–S5). Taking together, our results suggest an effect of the maternal dietary pattern on the incidence of infant infection during the first year of life that would be mediated by the transference of IgA through human milk. Thus, the interaction between the maternal diet and the milk immune composition seems to influence the development of early life infections.

We performed non-metric multidimensional scaling (NMDS) on the HM immune components (7 Igs, 18 CKs, and 2 adipokines) and we observed how some dietary components that differed between the two dietary clusters strongly overlaid with the NMDS plot (Supplementary Fig. S2B), such as total retinoids. Complementary to the strength of the dietary vectors with the NMDS plot (Supplementary Fig. S2B), the Spearman correlations showed associations between these dietary factors and some immune components (Supplementary Fig. S2C). The intake of insoluble hemicellulose, which is important in the characterisation of the dietary profiles (Table 1), was significantly negatively correlated with interferon (IFN)-γ (ρ = −0.217, p = 0.047), interleukin (IL)-5 (ρ = −0.22, p = 0.044), tumour necrosis factor (TNF)-α (ρ = −0.273, p = 0.012) and leptin (ρ = −0.221, p = 0.044) (Supplementary Fig. S2C). Moreover, the intake of total retinoids was positively correlated with HM IgM concentration (Supplementary Fig. S2C).

In a previous study we demonstrated that both maternal and infant microbiota was linked to the maternal diet during gestation.²² Considering this, we investigated how the intestinal microbiota of breastfed infants was affected by maternal diet during gestation and its relation to the predisposition for suffering infections.

With that aim, we analysed the gut microbiota of infants that suffered any kind of infection during the first year of life (Fig. 2, Supplementary Figs. S3 and S4). Regarding β-diversity, the NMDS analysis using the Bray–Curtis distance showed a distinct overall structure of the infant gut one-month-old microbiota of those children who suffered infections during early life compared to those who remained healthy (ρ = 0.13, p = 0.001) (Fig. 2A). We further explored the effect of infection, considering the number of infections and classifying the children accordingly into three groups (0, 1 or >2 infections). At 1 month of age, significant differences were observed between those children who underwent more than 2 infection processes compared to both those who remained healthy (R2 = 0.077, p = 0.024) and those who suffered only one infection (R2 = 0.512, p = 0.039) (Fig. 2A). These changes were no longer observed in samples collected at six or twelve months of life (Supplementary Fig. S3). Similarly, infection processes during early life also impacted α-diversity. The richness of the gut microbial community from one-month-old infants who had no infection was significantly lower than those who experienced one or more infections in their first year of life (Fig. 2B). Altogether, our results suggest that children that underwent several infections during early life had a distinct colonisation process with potential influences in infant development.

Fig. 2.

Infant gut microbiota at one-month is associated with maternal diet and the predisposition of infants to infections during the first year of life. (A) Non-metric multi-dimensional scaling (NMDS) plots based on Bray–Curtis dissimilarity measures to identify microbial groups associated to the infections and also, associated to AUC and key microbial genera. (B) Richness, Simpson, and Shannon measures of α-diversity at 1 month considering the number of infections during the first year of life (0 infections, 1 infection, ≥2 infections). (C) Richness, Simpson, and Shannon measures of α-diversity at 1 month considering maternal diet and infant infections during the first year of life. (D) The β-diversity of gut microbiota assessed using PCoA of Unifrac Unweighted distances, in relation to maternal diet and the occurrence of infant infections during the first year of life. (E) Key microbial genera linked to maternal diet group using mixed models. A, 0 infections; B, 1 infection; C, ≥2 infections. PERMANOVA test (A) and Kruskal Wallis non-parametric test (B, C) were used to assess significant differences between groups and are indicated in the figures. ∗p < 0.05, ∗∗p < 0.01. p < 0.05 was considered statistically significant.

As mentioned, offspring from mothers with a D1 dietary profile seemed to have less incidence of infections during early life (Fig. 1A); thus, we included the analysis of how the maternal diet profile could modulate the observed effect that the infection has on the neonatal gut microbiome. Our results revealed that children with infections with mothers from D2 groups (D2-I) exhibited a higher α-diversity in comparison to those with non-infection (D2-NI) at one month of age (Fig. 2C). Contrary, this effect was lost in infants from D1 mothers suggesting that maternal diet could exert its protective effect through a modulation of the α-diversity of neonatal gut microbiota. Despite the effect of the maternal diet on the β-diversity of the infants assessed by Unifrac distances was not significant (R2 = 0.048, p = 0.380; Fig. 2D) and thus, the overall structure of the microbial community was not altered; some members of the gut microbiota were impacted. At compositional level, a decrease in the Streptococcus genus and an increase in pro-inflammatory bacteria such as E. coli, Klebsiella, Enterococcus and Haemophilus were observed in the D2 group (AU-ROC = 0.832, p < 0.05) (Fig. 2E). Indeed, considering the number of infections, children with more than 2 infections during the first year of life were characterised by an increased relative abundance of Enterococcus and Klebsiella genera in faecal microbiota of one-month-old infants and negatively associated with Bifidobacterium and Streptococcus (Fig. 2A). In addition, specific associations (Supplementary Fig. S3) when controlling for infant age and other demographic variables were found with both maternal diet group and infant infections. Due to the dramatic effect of the time in the establishment of infant microbiota, it was complex to have markers associated with infections. Nevertheless, some gut microbiota members were related to infants with 1 or more infections such as Ruminococcus, Anaerobutyricum and Clostridium IV.

Neonatal rat acute gastroenteritis amelioration is influenced by the maternal diet composition

The observational clinical analysis in the cohort MAMI revealed a potential effect of maternal diet on the incidence of infant infections during the first year of life. In order to explore the potential mechanisms behind this observation, a preclinical approach was developed. Thus, an acute gastroenteritis infection model was used, based on a rotavirus (RV) infection and diarrhoea assessment in neonatal rats. The maternal diet profiles D1 and D2 from the clinical study were adapted for rodents specifically considering those dietary components which were positively or negatively correlated with infants’ infections (i.e., fibre and cholesterol, respectively) (Supplementary Table S1). The clinical D1 pattern was adapted for rodent diet 1 (D1) by enriching it in vegetal protein, fish oil and fermentable fibre, while the clinical D2 pattern was adapted (D2) by adding higher amounts of animal protein and SFA. A reference diet (RD) was also included in the study as a control (Supplementary Table S1). The diets were administered to the rats throughout gestation (21 days) and lactation (21 days). On the 5th day of life, pups were orally inoculated with the virus and displayed diarrhoea until day 10–11 of life (Supplementary Fig. S1B).

Consistent with previous studies,13, 14, 15 the low-grade infection in the offspring from RD mothers had similar effects on growth compared to the non-infected pups from the same dietary group. While the RV inoculation in this model did not impact the neonatal growth, maternal diet did. The offspring of the mothers with a dietary profile slightly enriched in animal protein and SFA (D2) had a higher body weight than those from the D1 group during the study (Fig. 3A), and a higher body weight increase in the diarrhoea period (Supplementary Fig. S5A) compared to D1 (p < 0.001) infected offspring.

Fig. 3.

Gastroenteritis is ameliorated by maternal diet D1 enriched in vegetal protein, fibre and fish oil. (A) Mean animal weights from days 2–13 of life. (B–C) Clinical indices of diarrhoea: severity (B) and incidence (C). The severity is expressed with the diarrhoea index (DI) on a scale of 1–4. Scores of DI ≥ 2 indicate presence of diarrhoea. The incidence of diarrhoea is represented as the percentage of diarrhoeic animals (%DA), which corresponds to the percentage of animals displaying DI scores ≥2 in each group. The mean area under the curve (AUC) of severity and incidence, S-AUC and I-AUC, respectively, is displayed on the right side of the graph. (D) The faecal weight, as an objective indicator of the severity of diarrhoea, was calculated as the mean value in the diarrhoea period. (E) Faecal SA-11 rotavirus shedding on day 6, which is the first day post inoculation. RD animals were under the cut off value in the technique. (F) Spearman’s correlation test between faecal SA-11 particles and DI at day 6. (G) Blood counts of leucocyte types: lymphocytes, granulocytes and monocytes. (H) Relative intestinal gene expression of Afp calculated with respect to RD, which corresponded to 100% of transcription (dashed line). Results are expressed as mean ± SEM (n = 15–43 depending on the sample used for each subpanel result). ^αp < 0.05 RV + RD vs. RV + D1, ^βRV + RD vs. RV + D2, ^#RV + D1 vs. RV + D2, ∗RD vs. RV + RD (A,B,C,H). Two-way repeated measures ANOVA and one-way Kruskal Wallis test followed by Dunn’s multiple comparison test with Bonferroni correction (A), one-way ANOVA followed by post hoc with Bonferroni correction (B, C) and one-way Kruskal Wallis test followed by Dunn’s multiple comparison test with Bonferroni correction (D, E, G, H) were used to assess significant differences between groups and are indicated in the figures. p < 0.05 was considered statistically significant. RD, Reference diet; D1, dietary pattern 1; D2, dietary pattern 2; RV, Rotavirus. RD group, non-infected pups in the maternal reference diet group; RV + RD group, infected pups in the maternal reference diet group; RV + D1, infected pups in the maternal D1 diet group; RV + D2, infected pups in the maternal D2 diet group.

We evaluated the gastroenteritis caused by RV inoculation in terms of severity (diarrhoeic index or DI and faecal weight), and incidence (percentage of diarrhoeic animals, DA%), as in previous studies.13, 14, 15^,23 The DI scores (from 1 to 4) were around 2–3 in all infected animals, indicating that mild diarrhoea in infected suckling rats was achieved. In contrast, uninfected suckling rats in the RD group showed DI scores around 1–1.5 (Fig. 3B). Likewise, the DA% and the faecal weight were also higher in the RV + RD group with respect to the RD group (Fig. 3C and D, respectively). When examining the effect of maternal dietary profile, infected D1 pups showed a lower diarrhoea severity throughout the diarrhoea period (d5-d11), evidenced by a reduction in the area under the severity curve with respect infected RD and D2 groups (Fig. 3B); indicating the positive effect of the maternal dietary intervention with (D1 in the amelioration of offspring’s infection). This positive effect was also observed when analysing the incidence of diarrhoea (Fig. 3C); again, D1 group displayed a curve below the other two groups. Although D1 ameliorated the severity of diarrhoea, this observation was not related to the viral shedding that was assessed by quantifying the RV particles in the faeces at day 6 (the first day post inoculation²³), and was similar to the infected RD and D2 groups (Fig. 3E). Interestingly, viral shedding was positively correlated with the DI at day 6 (Fig. 3F), suggesting that when DI is higher, faecal RV load is also higher.

Maternal dietary intervention also influenced the blood leucocyte count, since suckling rats from the D2 group had higher lymphocyte blood counts than those from D1 (Fig. 3G), indicating that D2 animals required higher mobilisation of blood lymphocytes to counteract infection than D1 pups, presumably leading to the more severe infection process observed in the diarrhoea period as reported above. We then examined the jejunum intestinal architecture and gene expression (Supplementary Fig. S6). The induction of diarrhoea by RV in this model was reported previously to result in the upregulation of antiviral genes in the host and downregulation of various genes that play a role in intestinal maturation and absorptive processes, including alpha-fetoprotein (Afp), One cut Homeobox 2 (Onecut2), and C–C motif chemokine ligand 19 (Ccl19).⁴⁸ In agreement with this, infected animals from the RD group showed Afp downregulation compared to non-infected RD pups (p = 0.004). Moreover, this impairment was even clearer in the RV-infected D2 suckling rats as they showed significantly lower gene expression than the other RV-infected suckling rat groups (RD, D1) (Fig. 3H). On the contrary, maternal D1 improved this RV-associated disruption of the intestinal maturation. Analysis of jejunum intestinal architecture and gene expression of tigh junction proteins and mucin 2 were similar between groups (Supplementary Fig. S6). Thus, the protective effect of D1 on the neonatal anti-RV response is not likely mediated by its influence on the intestinal barrier.

The diet-dependent amelioration of diarrhoea is mediated by the pups’ antibody response

We hypothesised that the modulation of the antibody response against the infection by the maternal diet during pregnancy and/or lactation could be one of the mechanisms involved in the reduction of the infection severity by the D1 dietary profile. Thus, determination of a global antibody (Ab) profile was performed in plasma during the peak of diarrhoea, on day 8. RV infection increased plasma IgA, IgM and IgG levels, as previously observed¹⁴ (Table 2). Interestingly, D1 pups presented higher levels of total Igs in plasma compared to the others infected groups. Specifically, this increase was due to an increase in IgG, and particularly in IgG2c (involved in Th1 response). As IgG2c was also clearly increased in reference infected pups (RV + RD) compared to reference non-infected pups (RD), the rise in the D1 group suggests a higher Th1 immune response against the virus in this group which could explain the lower infection.

Table 2.

Concentration of Igs in plasma in the diarrhoea period (day 8).

	RD	RV + RD	RV + D1	RV + D2	p values
					RD vs. RV + RD	RV + D1 vs. RV + RD	RV + D2 vs. RV + RD	RV + D2 vs. RV + D1
Total	1026.16 ± 83.09	1477.82 ± 89.39a	1812.88 ± 67.82b	1361.82 ± 79.04c	0.014	0.039	0.807	0.001
IgA (%)	0.78 ± 0.00 (0.08 ± 0.01)	2.43 ± 0.46a (0.16 ± 0.03)	2.25 ± 0.50 (0.12 ± 0.02)	1.57 ± 0.31 (0.11 ± 0.02)	0.036 0.154	1.000 0.373	0.528 0.890	1.000 1.000
IgM (%)	2.51 ± 0.48 (0.24 ± 0.03)	4.96 ± 0.54a (0.33 ± 0.02)a	6.18 ± 0.83 (0.33 ± 0.04)	4.44 ± 0.55 (0.32 ± 0.03)	0.006 0.045	0.738 1.000	1.000 1.000	0.180 1.000
IgG (%)	1022.87 ± 82.68 (99.68 ± 0.02)	1470.43 ± 88.76a (99.51 ± 0.04)a	1804.45 ± 67.12b (99.55 ± 0.05)	1355.80 ± 78.51c (99.57 ± 0.04)	0.014 0.008	0.040 1.000	0.823 0.787	0.001 1.000
IgG1 (%)	33.97 ± 10.82 (3.07 ± 0.72)	85.10 ± 13.69 (5.36 ± 0.80)	101.16 ± 12.14 (5.34 ± 0.57)	66.26 ± 12.73 (4.27 ± 0.72)	0.095 0.235	0.687 1.000	1.000 1.000	0.110 1.000
IgG2a (%)	272.08 ± 58.54 (25.62 ± 3.71)	368.28 ± 38.65 (24.83 ± 2.08)	312.30 ± 12.50 (17.35 ± 0.41)b	290.25 ± 21.19 (21.60 ± 1.47)	0.132 0.733	1.000 0.004	0.707 0.818	1.000 0.073
IgG2b (%)	254.14 ± 32.94 (24.42 ± 1.13)	349.44 ± 23.49a (23.68 ± 0.44)	411.92 ± 24.00 (22.87 ± 1.11)	354.72 ± 21.33 (26.62 ± 1.51)c	0.050 0.622	0.290 0.266	1.000 0.752	0.430 0.012
IgG2c (%)	462.68 ± 41.18 (46.89 ± 5.25)	667.60 ± 53.91a (46.13 ± 2.94)	979.07 ± 40.68b (54.44 ± 1.46)	644.58 ± 50.27c (47.51 ± 2.76)	0.029 1.000	0.001 0.064	1.000 1.000	0.001 0.072
Th1	716.82 ± 43.00	1017.05 ± 68.14a	1390.99 ± 45.90b	999.30 ± 55.59c	0.003	0.001	1.000	<0.001
Th2	306.05 ± 68.64	453.38 ± 51.28	413.46 ± 23.08	356.50 ± 30.29	0.095	1.000	0.313	0.500
Th1/Th2	2.84 ± 0.48	2.71 ± 0.32	3.45 ± 0.13	3.11 ± 0.26	0.910	0.109	1.000	0.409

Data are the mean ± SEM (n = 6–15/group).

p < 0.05 (shown in bold letters) was considered statistically significant.

^ap < 0.05 compared to the reference diet (RD).

^bp < 0.05 compared to the RV + RD group.

^cp < 0.05 compared to the RV + D1 group by Kruskal Wallis test followed by Dunn’s multiple comparison test with Bonferroni correction. p < 0.05 was considered statistically significant. Th1/Th2 ratio refers to the relationship between Th1 (IgG2b + IgG2c) and Th2 (IgG1 + IgG2a) Ig.

To further study the IgG2c increase found in D1 pups’ plasma, the levels of the different types of Ig in rat BM were also quantified. While similar levels of BM Ig types were found between dietary groups (Fig. 4A and B), the IgG2a and IgG2c proportions tended to be different depending on maternal diet (Fig. 4C) with the IgG2c proportion tending to increase in D1 BM compared to D2 BM (without being statistically significant). In contrast, IgG2a (involved in Th2 response) tended to decrease in D1 BM and was significantly decreased in D1 pups’ plasma (Fig. 4C and Table 2, respectively), supporting a maternal influence on IgG subtype levels in offspring plasma, via lactation, which is impacted by maternal diet. In addition, an association between BM IgG2c and the load of faecal RV SA-11 particles was found (Supplementary Fig. S7).

Fig. 4.

Analysis of the Ig profile in milk at the end of lactation, the anti-RV Ab levels in rat breast milk, digested milk and maternal plasma. (A) NMDS for the concentration of BM Igs (d21). Categorical variable (Dietary Group) is represented by colour. Stress: 0.1167257. (B) Ig levels in BM at the end of lactation (d21). (C) IgG subtype proportions (%) in BM at the end of lactation (d21). (D) Concentration of anti-RV Abs (Arbitrary Units/mL) by ELISA in pup plasma at d8, d21 and d28. (E) Fold change of anti-RV Abs (Arbitrary Units/mL) concentration from the beginning of the dietary intervention to the end of the study (G0 → d21). (F) Concentration of anti-RV Abs (Arbitrary Units/mL) in stomach content of pups at day 8. (G) Concentration of anti-RV Abs (Arbitrary Units/mL) in milk at the end of lactation. Results are expressed as mean ± SEM (n = 5–8/group, except for (A) which has n = 5–15/group). Kruskal Wallis test followed by Dunn’s multiple comparison tests with Bonferroni correction was used to assess the significance of the differences between groups (A). p value expressed in the figure was derived from the Kruskal Wallis test (G). p < 0.05 was considered statistically significant.

We investigated also the total anti-RV specific Ab by the determination of total anti-RV specific Ab on days 8, 21 (the end of lactation period) and 28 (one week after weaning) to observe the specific response during and after infection resolution (Fig. 4D). As expected, levels during the infection (d8), tended to be higher in the RV + RD group with respect to the RD group (p = 0.087). Similar to the global response, RV + D1 also exhibited increased anti-RV Ab concentration compared to the RV + RD group. This increase did not persist afterwards (d21 and d28).

To evaluate if the observed increase in anti-RV specific Abs in D1 plasma on day 8 is due to maternal transmission and/or maternal influence, these Abs were also quantified in maternal plasma and breast milk (BM) at day 21 and in offspring stomach content at day 8, as a representation of the ingested BM on that day (Fig. 4). Anti-RV Ab levels in maternal plasma from the beginning of the intervention (G0) to the end (d21) and in the milk did not differ according to diet (Fig. 4E–G), suggesting that the specific antiviral immunity in D1 neonatal rats is not dependent on the maternal transmission. The expression of the receptor of IgGs (FcRn) at intestinal level was analysed, and it was observed a lower FcRn gene expression in D1 compared to D2 neonatal rats (Supplementary Fig. S6D). Since FcRn levels are described to decrease physiologically throughout suckling,^14,15 D1 diet seems to accelerate the intestinal maturation of the offspring.

Maternal diet impacts the microbiota of infected offspring and mitigates the impact of infection

The infection and the maternal diet impacted the overall structure of the neonatal caecal microbiome assessed by permanova analysis based on weighted UniFrac distances (R2 = 0.232, p = 0.001) (Fig. 5A). This was confirmed by the canonical correspondence analysis (CCA) that showed significant differences between groups, highlighting the differential compositional profile of the D2 group (Fig. 5B).

Fig. 5.

The neonatal caecal microbiota of infected 8-day-old rat offspring is influenced by maternal diet. (A) β-diversity of the microbiota, assessed using weighted UniFrac distances and plotted through PCoA. (B) Canonical correspondence analysis (CCA) at the genus level. (C) α-diversity indexes measured as observed ASVs, Chao1, and Shannon indices. (D–E) Taxonomic biomarkers identified in both dietary groups by LEfSe analysis. Plots showed the significant Linear Discriminant (LDA) scores. Results are derived from n = 5–14 animals/group. PERMANOVA test (A, B) and Kruskal Wallis non-parametric test (C) were used to determine the significant effect of the diet on the overall structure of the gut microbiota. p < 0.05 was considered statistically significant.

While the dietary intervention resulted in an effect on the caecal microbiota at day 8, it did not influence the α-diversity of the microbial populations (Fig. 5C). At the compositional level, using linear discriminant analysis (LEfSe), we found that the offspring’s microbiota from the D1 group was characterised by Rothia and Streptococcus genera while the D2 group was enriched in Clostridium_sensu stricto and Romboutsia genera when compared to RD + RV group (Fig. 5D and E). Furthermore, we assessed the relationship between members from the caecal microbiome and the defensive capacity against retroviral infection. Interestingly, Rothia spp., which was a biomarker for the D1 group microbial community showed the strongest correlation with the physiological variables related to the defensive capacity, including plasma anti-RV Ig in pups’ plasma (Supplementary Fig. S8C) suggesting the potential role of microbiome modulation on the observed protection against the infection related to the maternal diet.

Discussion

Here, we integrated a clinical and a preclinical study to unravel the link between maternal diet, HM immune composition, neonatal microbiota that may contribute to infant disease prevention. An appropriate maternal diet seems to be pivotal to healthy pregnancies and to reducing viral infections’ impact both on mothers and offspring.^5,49,50 In the present study, a maternal diet rich in fermentable fibre and vegetable protein during gestation and lactation reduced infant infections and its severity in both clinical and preclinical approaches. Furthermore, we previously observed that maternal diet is important in the infant’s immune status at birth.²⁷ Therefore, we wanted to assess whether a maternal diet-HM immune composition interaction exists to counteract early life infections and to understand its influence on neonatal microbiota composition.

Our data showed that the dietary pattern enriched in fibre and vegetal protein and poor in lipids such as cholesterol and MUFA (D1) decreased both the presence and the number of offspring infections within the first year of life without affecting infant growth. Dietary fibre is well known to have beneficial actions during pregnancy, and accordingly, the recommended intake during pregnancy is increased.⁵¹ Dietary fibre is degraded by commensal gut microbiota into short chain fatty acids (SCFAs) which have several associated beneficial outcomes both at gastrointestinal and immune levels.^52,53 In fact, Needell et al. observed that a SCFA maternal treatment decreased virus-induced inflammation in the offspring.⁵⁴ Therefore, it could be suggested that the beneficial impact observed in our study is mediated by these microbial metabolites.

The gold food standard for infants’ growth, development, protection, and immune development is HM due to its nutrients and bioactive compounds.^10,55, 56, 57, 58 These components change throughout lactation and are influenced by other factors, such as diet. However, the literature about maternal diet’s influence on HM immune components, such as IgA, is controversial.⁵⁹ Regarding the immune components in HM from mothers following both type of diets, a wide range of bioactive components such as Igs, Cks, leptin and adiponectin were established, and overall, these bioactive factor levels were similar independently of the maternal dietary pattern. Surprisingly, we found that the concentration of some HM immune components was different in mothers whose infants developed infection during the first year of life, but only in infants from the D2 group. Thus, following a maternal diet rich in animal protein, cholesterol and MUFA together with having low levels of HM IgA and IL-10 and a higher ratio of HM Th1/Th2 associated Igs seemed to predispose infants to infections in early life. Taken together, it can be suggested that low levels of IgA and IL-10 in HM predispose to infant infections in mothers whose dietary habits align with the D2 profile but do not when the maternal diet is rich in healthier components, such as fibre or vegetable protein.

The neonatal microbiota undergoes a rapid colonisation and transformation and is characterised by a low diversity,60, 61, 62 with Bifidobacterium being the predominant genus (up to 90%). Bifidobacterium spp. have developed various strategies for utilising human milk oligosaccharides (HMO). However, their relative abundances decline rapidly after the introduction of solid foods during the weaning process.⁶³ Maternal diet determines both the maternal intestinal^21,64 and milk¹² microbiota, which, in turn influences the neonatal microbiota.^22,52,54 In a previous study conducted on the same cohort, it was demonstrated that the maternal diet during gestation was associated with the microbiota composition of both the mother and newborn at birth.²² Specifically, higher neonatal microbiota richness was positively linked to SFA and animal protein intake and negatively associated with vegetable protein intake. Subsequently, in this study we investigated how the microbiota of nursing infants was influenced by the maternal diet during gestation and the infants' susceptibility to infections. It was found that infants with higher microbiota diversity at one month of age were more prone to infections within the first year of life. This association was particularly notable among children exposed to a maternal diet enriched in animal protein and fats and poor in fibre (D2). This could potentially be related to the association between maternal consumption of SFA and animal protein, and the subsequent augmentation in neonatal microbiota diversity, as previously mentioned.²² Consequently, these findings imply that if a mother adheres to a diet low in dietary fibre and rich in animal protein and fats, the microbial diversity at one month of age can serve as a critical indicator of the likelihood of infections during the early stages of life.

RV inoculation induced moderate diarrhoea in the preclinical intervention study, with peak viral shedding the day after infection and consequent lymphocytosis with high plasma anti-RV Ab and Th1 associated Ig levels. Interestingly, the infected suckling pups from mothers fed D1 chow, but not D2 chow, had less diarrhoea severity and incidence. This lower infection in D1 animals was accompanied with lower FcRn gene expression and lower blood lymphocyte counts than D2 pups indicating higher immune maturation⁶⁵ and a lower systemic inflammation,^66,67 respectively. The positive effect of D1 on infection was also related to a change in the infant humoural response, as infected D1 pups displayed an enhanced Th1 IgG response (IgG2b and especially IgG2c), thus supporting gastroenteritis resolution. This immune maturation effect was not observed in those pups whose mothers followed the reference or D2 diets. IgG2c levels increased in the plasma of pups during infection (d8) in the D1 group, and there was also a tendency for an increase in BM itself. This suggests that the positive effect in pups could be directly derived from passive transfer through breastfeeding. However, the role of maternal IgG2c on pups’ intestine is not well established yet. Rat IgG2c is the analogue of IgG3 in mice⁶⁸ and it was reported that this Ig, together with IgG2b, regulates responses in neonatal intestines to translocating microbes and influence T cell differentiation.⁶⁹ In fact, we observed that mothers with higher levels of the Th1 isotype IgG2c in their BM had pups with lower faecal RV particle levels and diarrhoea scores, regardless of diet, indicating a protective role. This finding is important in this context as general literature about Igs and BM is typically more focused on the role of IgA.^59,70, 71, 72, 73 However, this preclinical result contrasts with the findings obtained in the clinical intervention, where HM IgA had a protective role while higher levels of HM Th1 IgG associated isotypes appeared to be detrimental. This can be explained because the maternal-infant transmission of Igs in humans is different from that of rats. In humans IgA is the predominant Ig in HM,⁵⁹ while in rats it is IgG.²⁶ Despite these differences between humans and rats, it is worth noting that the predominant IgG subtypes present in BM are Th1 associated,^28,59 in contrast to the intrauterine Th2 predominance,^28,74,75 which prevents intrauterine rejection⁷⁶ during pregnancy.

On the other hand, the anti-RV response is also boosted in infected D1 pups. The increase in neonatal plasma anti-RV Ab was dependent on the maternal diet group but independent on passive BM immune transmission, as BM anti-RV Ab levels were similar in digested milk during the infection and in BM at the end of the suckling period. Therefore, maternal D1 plays a role in the development and maturation of the infant immune system to actively counteract infections, beyond the passive immunisation conferred by BM.

An impact on cecal microbiota composition was found, even though both maternal dietary patterns (D1 and D2) did not change intestinal histomorphometry or expression of TJ proteins. Our results revelead that individuals whose mothers followed a maternal diet characterised by high fibre intake, fish oil and plant-based protein had a distinct microbiota profile compared to those who did not. Notably, this shift characterised by an increased abundance of Rothia and Streptococcus spp., among other changes, seems to be sufficient to exert an immune impact at both the intestinal and systemic levels. Previous preclinical studies showed the key role of neonatal gut microbiota in immune system development.²⁴ Thus, in our case study, it could be helping to attenuate RV infection. However, other analysis should address the specific molecular pathways behind the association of Rohia spp. with the observed systemic immune effects.

In conclusion, maternal dietary composition plays an important role in counteracting early life infections both at clinical and preclinical levels, by impacting BM immune composition, intestinal development, microbiota and immune status of the offspring. Further studies are needed to clarify the maternal diet influence on immune system development and on maternal transmission of BM immune components through lactation.

Limitations

The preclinical study was designed to match the clinical study regarding dietary patterns. However, the components of the human dietary patterns were not exactly the same in the human and the rodent studies as they were adapted for another species (rat) and corrected for the pregnancy and neonatal periods. Additionally, our findings indicate that D1 dietary profile enhanced intestinal and immune maturation in infant rats during infection, but due to ethical considerations, we had no infant plasma or intestinal samples from the clinical cohort to confirm these results in humans. Moreover, the current study involves maternal intervention during both gestation and lactation, but it would be valuable to establish distinct cohorts to investigate the effects of maternal dietary patterns solely during gestation and solely during lactation in order to clarify the importance of each period. Overall, further research is required to better understand the specific dietary components and the critical windows during which maternal interventions may positively influence the development of the infant's immune system. Although the present study did not address the underlying molecular mechanisms, the findings underscore the importance of future research aimed at investigating these pathways in greater depth to better understand the biological processes involved.

Contributors

Conceptualisation: F.J.P.-C. and M.C.C.

Methodology: K.R.-A., M.S.-R. and R.C.-R.

Investigation: K.R.-A., M.S.-R., R.C.-R., S.G., C.M-C., and M.C.

Visualisation: K.R.-A., M.S.-R. and R.C.-R.

Funding acquisition: F.J.P.-C. and M.C.C.

Project administration: M.J.R.-L., F.J.P.-C. and M.C.C.

Supervision: M.J.R.-L., F.J.P.-C. and M.C.C.

Writing – original draft: K.R.-A., M.S.-R. and R.C.-R.

Writing – review & editing: S.G., C.M-C., M.C., M.J.R.-L., F.J.P.-C. and M.C.C.

Data sharing statement

Data is available upon request. The 16S rRNA gene sequence data generated is available through NCBI Sequence Read Archive Database under project accession numbers: BioProject ID PRJNA614975 for the clinical part and BioProject ID PRJNA1119440 for the preclinical part.

Declaration of interests

None declared.