Human milk oligosaccharides promote synaptogenesis and neurite outgrowth in human cortical organoids

TL;DR

Human milk oligosaccharides boost brain development in lab-grown human brain models by enhancing nerve connections and growth.

Contribution

This study reveals a direct role of HMOs in promoting synaptogenesis and neurite outgrowth in human cortical organoids.

Findings

HMO treatment significantly enhanced neurite outgrowth and synaptogenesis in a dose-dependent manner.

Proteomic profiling showed upregulation of proteins linked to neuronal differentiation and synaptic maturation.

HMOs upregulate RNA splicing pathways in cortical organoids.

Abstract

The first 1000 days of a child's life represent a critical window for brain development, during which nutrition exerts profound effects on the trajectories of neurodevelopment. Human Milk Oligosaccharides (HMOs), a major component of human milk, are largely indigestible by infants and are known to influence immunity, microbiome composition, and gut-brain signaling, but their direct role in neurodevelopment remains poorly understood. Here, we investigated the impact of HMOs on human cortical organoids, a physiologically relevant in vitro model of early brain development. We found that HMO treatment significantly enhanced neurite outgrowth and synaptogenesis in a dose-dependent manner. Global proteomic profiling further demonstrated the upregulation of proteins associated with neuronal differentiation, synaptic maturation, and cytoskeletal remodeling. Our findings suggest that HMOs can influence neurodevelopmental processes and highlight a potential role for maternal milk components in early brain development.

Image 1

•Human milk oligosaccharides (HMOs) modulate synaptic density in human cortical organoids.•HMOs influence neural connectivity during early development.•Genetic background shapes HMO-mediated effects.•HMOs upregulate RNA splicing pathways in cortical organoids.

Human milk oligosaccharides (HMOs) modulate synaptic density in human cortical organoids.

HMOs influence neural connectivity during early development.

Genetic background shapes HMO-mediated effects.

HMOs upregulate RNA splicing pathways in cortical organoids.

Figures

Full text

Introduction

1

Human milk oligosaccharides (HMOs) are a diverse group of structurally complex carbohydrates representing the third most abundant solid component in human milk, following lactose and lipids [[1], [2], [3], [4]]. HMOs begin to be produced during pregnancy and are present in colostrum at high concentrations, with their composition dynamically changing throughout lactation [[5], [6], [7], [8]]. While HMOs are largely indigestible by the infant, a small fraction, particularly fucosylated and sialylated HMOs, is absorbed intact into the bloodstream, enabling systemic actions beyond the gut microbiome [9]. These molecules exert their functions through multiple mechanisms, including shaping the gut microbiome, directly interacting with epithelial and immune cells, and potentially influencing processes in distant tissues [[9], [10], [11], [12]]. HMOs have garnered significant attention for their critical roles in several developmental stages, influencing systemic processes that contribute to overall development and health during early life [13]. Recent studies have begun illuminating a potential role for HMOs in neurodevelopment, although the mechanisms underlying their effects on the developing brain remain largely unexplored.

Several studies comparing breastfed and formula-fed infants have identified positive associations between breastfeeding and brain development [[14], [15], [16], [17], [18], [19], [20]]. Breastfed individuals have consistently demonstrated higher IQ scores, larger brain sizes, and increased white matter volumes compared to their formula-fed counterparts [[14], [15], [16], [17], [18]]. Notably, disparities in brain structure and cognitive development between breastfed and formula-fed individuals have been observed as early as the first year of life and can persist through early childhood and into adolescence [19]. These differences are thought to be influenced, at least in part, by the significantly lower levels of bioactive components, such as sialylated oligosaccharides, present in bovine milk and infant formula when compared to human milk [20].

Sialylated HMOs, in particular, are believed to play a critical role in brain development, serving as a primary dietary source of sialic acid (Neu5Ac) for the developing infant brain [9]. During early infancy, the endogenous synthesis of sialic acid is limited due to low enzymatic activity, making the intake of sialylated oligosaccharides from human milk essential for proper central nervous system maturation [21]. One study specifically identified a strong association between the sialylated HMO 6′-sialyllactose (6′SL) and increased brain myelination in infants, suggesting that HMOs may contribute directly to neuronal maturation [22]. However, due to the limited accessibility of the developing human brain for direct study, the precise mechanistic role of HMOs in human neurodevelopment remains poorly understood.

Animal model studies have provided valuable insights into the potential neurodevelopmental effects of HMOs [[23], [24], [25], [26], [27]]. Research has shown that sialylated and fucosylated HMOs may influence cognitive function by promoting synaptic plasticity, enhancing myelination, and modulating neurotransmitter activity [23]. In vivo magnetic resonance spectroscopy (MRS) studies in piglets have further demonstrated that sialylated milk oligosaccharides alter neurotransmitter levels and brain metabolites, supporting the hypothesis that HMOs contribute to neural connectivity and cognitive processing [24]. Advanced imaging techniques, such as MRI, have revealed that dietary sialyllactose influences sialic acid concentrations in the prefrontal cortex and induces structural brain changes, particularly in the corpus callosum [25]. Recent studies using a knock-out mouse model lacking 6′SL or 3′SL showed that lactational deprivation of sialylated HMO led to impaired cognitive functions in adulthood, including deficits in attention, memory, and increased perseveration [23,26,27]. These changes were accompanied by altered hippocampal electrophysiology and site- and time-specific reductions in gene expression in the prefrontal cortex, including myelination-associated genes [23]. Together, these findings strengthen the argument that sialylated HMOs play a crucial role in early cognitive development by transiently modulating neuronal patterning in the developing brain.

The first 1000 days of life are widely recognized as a period of rapid neuronal growth and plasticity, during which foundational processes such as synaptogenesis, myelination, and circuit refinement are particularly active [28]. Nutrition during this window has long-lasting effects on cognitive, motor, and behavioral outcomes, underscoring the importance of early dietary components such as HMOs [29]. Despite growing evidence of their neurodevelopmental benefits, no studies to date have directly investigated the impact of HMOs on human neural cultures, representing a critical gap in understanding their role in human brain development at the cellular level. Here, we address this question using human cortical organoids, derived from induced pluripotent stem cells (iPSCs), as a model to study the effect of HMOs in the development of the central nervous system.

Methods

2

HMOs acquisition and extraction

2.1

Human milk was collected as part of UC San Diego's Human Milk Institute (HMI) human milk donation program. Twenty-eight (28) healthy donors provided milk in excess of their own baby's needs, and the milk was pooled to obtain an average mixture of HMOs. HMOs were analyzed as previously described [30,31]. After centrifugation, the lipid layer was removed, and proteins were precipitated from the aqueous phase by addition of ice-cold ethanol and subsequent centrifugation. Ethanol was removed from the HMO-containing supernatant by roto-evaporation. Lactose and salts were removed by gel filtration chromatography over a BioRad P2 column (100 cm × 316 mm, Bio-Rad) using a semi-automated fast protein liquid chromatography (FPLC) system. HMO composition was measured by HPLC-FL. Only pooled HMOs with less than 1% lactose were used for experiments. Pooled HMOs (pHMOs) contained 33.4% 2′FL, 13.0% 3FL, 3.8% DFLac, 2.8% 3′SL, 2.9% 6′SL, 6.6% LNT, 1.3% LNnT, 8.1% LNFP1, 5.0% LNFP2, 6.0% DFLNT, 1.1% DSLNT, and less than 1% each of other, more complex HMOs.

Endotoxin (LPS) was removed from all pHMO and individual HMOs used for in vitro experiments by Detoxi-Gel Endotoxin Removing columns (Pierce Thermo Scientific, Rockford, IL, USA) according to the manufacturer instructions. The whole procedure was performed under sterile conditions with the use of pyrogen-free UltraPure distilled water, ethanol 200 proof and sterile filter tips to avoid LPS contamination. After collecting the flow-through, containing HMO, the purified samples were frozen immediately and lyophilized until completely dry. To minimize the amount of LPS, the HMO samples were processed twice on the endotoxin removing columns.

Generation of cortical organoids

2.2

Cortical organoids were generated following a semi-guided differentiation protocol from induced pluripotent stem cells (iPSCs) [32,33]. Organoids from 2 control lines (WT83 and KOLF) were used in this study to assess the reproducibility of the results. Healthy iPSC lines were maintained on Matrigel-coated dishes and fed daily with mTeSR+ (StemCell Technologies). On day 0 of the differentiation protocol, cells are dissociated using StemPro™ Accutase™ Cell Dissociation Reagent (#A1110501, Thermo Fisher Scientific) and centrifuged for 3 min at 0.6 rpm. The resulting pellet is resuspended in mTeSR Plus medium supplemented with 10 μM ROCK inhibitor, 10 μM SB431542 (#04-0010, Stemgent), and 1 μM Dorsomorphin (#3093, R&D Systems). Approximately 2.5 million cells are seeded into each well of an AggreWell™400 plate (#34411, STEMCELL Technologies) to standardize aggregate formation. After 24 h, embryoid bodies are transferred to 6-well plates and maintained under continuous orbital rotation at 95 rpm.

Neural induction

2.2.1

On day 3, mTeSR Plus is replaced by Media 1 (Neurobasal™ Medium (#21103049, Thermo Fisher Scientific) supplemented with 1x GlutaMAX™ (#35050079), 1% N2 supplement (#17502048), 2% Gem21 NeuroPlex™ (#400-160-010, GeminiBio), 1% MEM Non-Essential Amino Acids (#11140050), 1% Penicillin-Streptomycin (#15140122), 10 μM SB431542, and 1 μM Dorsomorphin). Organoids are cultured in Media 1 for six days, with media changes every other day.

NPC proliferation phase

2.2.2

On day 9, organoids are transferred to Media 2 (Neurobasal™ supplemented with 1x GlutaMAX™, 2% Gem21 NeuroPlex™, 1% NEAA, and 1% Penicillin-Streptomycin) containing 20 ng/mL bFGF (PeproTech). Daily media changes are performed for seven days. From day 16 to day 22, organoids are cultured in Media 2 supplemented with both 20 ng/mL bFGF and 20 ng/mL EGF (PeproTech), with media changes every other day.

Neuronal differentiation phase

2.2.3

On day 22, organoids are transitioned to Media 3 [Media 2 supplemented with 10 μg/mL Brain-Derived Neurotrophic Factor (BDNF), 10 μg/mL Glial-Derived Neurotrophic Factor (GDNF), 10 μg/mL Neurotrophin-3 (NT-3) (all from PeproTech), 200 μM l-ascorbic acid (Sigma Aldrich), and 1 mM dibutyryl-cAMP (StemCell Technologies)]. Media is changed every other day for six days.

From day 28 onward, organoids are maintained in Media 2 with media changes every 3–4 days.

Immunofluorescence

2.3

To confirm cortical specification following completion of the differentiation protocol, 30-day-old brain organoids were analyzed by immunofluorescence. Organoids were fixed overnight at 4 °C in 4% paraformaldehyde (PFA; #PI28906, Thermo Fisher Scientific), cryoprotected in 30% sucrose, embedded in optimal cutting temperature (OCT) compound (#AGR1180, Agar Scientific), and sectioned using a cryostat.

Cryosections were washed three times with DPBS (#14200075, Thermo Fisher Scientific), then blocked for 1 h at room temperature in a solution of 3% bovine serum albumin (BSA; #700-110-100, GeminiBio) and 0.1% Triton X-100 (#1610407, Bio-Rad). Primary antibodies were diluted in blocking solution and incubated overnight at 4 °C in the dark. The following markers were used to confirm neuronal and cortical lineage identity: FOXG1 (1:500, ab196868, Abcam), MAP2 (1:2000, ab5392, Abcam), SOX2 (1:300, AF2018, R&D Systems), Nestin (1:300, ab22035, Abcam), and PAX6 (1:300, 901301, BioLegend). After three PBS washes, fluorescent secondary antibodies were applied for 1 h at room temperature in the dark. Sections were washed again and imaged using a Dragonfly confocal microscope with Z-stack acquisition at 40 × magnification.

HMO treatment

2.4

At 3 months of age, organoids were removed from orbital shaker and transferred individually to a 96 well plate. Organoids were treated with HMO solution at different concentrations (10 μg/mL, 50 μg/mL and 100 μg/mL) for 30 days. HMOs were added to organoid media and exposed to organoids twice a week for 4 weeks. For WT83, 10 organoids were used for each condition (*n=*10); for KOLF, 24 organoids were used for each condition (n = 24). Diameter measurements were taken at the end of treatment using GelCount (Oxford Optronix).

Synaptic puncta quantification

2.5

After 1 month of treatment, organoids followed the immunofluorescence protocol and imaged on a Dragonfly Microscope using Z stack compiled images at an objective resolution of 63X. Immunofluorescence was performed using the HOMER1 (1:500, 160,003, Synaptic Systems) VGLUT1 (1:500, 135,311, Synaptic Systems) and MAP2 (1:2000, ab5392, Abcam) antibodies. Synaptic Puncta was quantified using Imaris 10.1 software (Oxford Instruments) by calculating co-localization of HOMER1 and VGLUT1 when in contact with MAP2. 10 images were analyzed for all 4 treatment conditions in both cell lines (WT83 and KOLF).

RNA extraction and qPCR

2.6

RNA was extracted from cortical organoids after the 30-day treatment with HMOs using RNeasy Plus Mini Kit Extraction Kit (Qiagen, Germany) according to the manufacturer's instructions. RNA yield was assessed using a NanoDrop One Spectrophotometer (Thermo Fisher, USA). One microgram of RNA was treated with DNase and reverse transcribed using the QuantiTect Reverse Transcription Kit (Qiagen) following the manufacturer's protocol. Real-time PCR was carried out using TaqMan Custom Gene Expression Array plates. With a total volume of 20 μL per reaction, 10 μL of cDNA + RNase-free water was combined with 10uL of TaqMan Master Mix (#4369016, Thermo Fisher, USA). qPCR reaction was carried out in 96-well plates sealed with MicroAmp® Optical Adhesive Film in a CFX Opus 96 (Bio-Rad, USA) using the cycling conditions recommended by the manufacturer's protocol. The relative quantification of each target was obtained according to the 2^−ΔΔCt^ method after normalization with HPRT1 gene.

Bulk proteomics analytical method

2.7

Proteins were extracted using 8 M urea prepared in 100 mM triethylammonium bicarbonate (TEAB) buffer. Cell lysis was achieved by two cycles of sonication on ice, each lasting 20 s at 60% amplitude. Protein concentration was determined using the bicinchoninic acid (BCA) assay. For each sample, 50 μg of total protein was reduced by 5 mM Tris (2-carboxy-ethyl) phosphine hydrochloride (TCEP) for 30 min at room temperature, followed by alkylation with 15 mM chloroacetamide for 30 min in the dark at room temperature. Proteins were then digested overnight (18 h) at 37 °C with sequencing-grade modified trypsin (Promega, V511A) at an enzyme-to-substrate (E:S) ratio of 1:50 (w/w). Digestion was quenched by acidification with formic acid (FA) to a final concentration of 1%. Samples were centrifuged at 18,000×g for 10 min to remove insoluble debris. Peptides were subsequently desalted using C18 Stage tips.

All samples for bulk proteomics were analyzed using a Thermo Scientific Vanquish Neo UHPLC system coupled to a Thermo Scientific Orbitrap Astral mass spectrometer equipped with an EASY-spray source. Each of the individual tryptic digest sample concentration was normalized for a total on column load of 250 ng and injected directly from a low-binding Thermo Fisher 200ul vial (PN: 60180-1655) in triplicate. The Vanquish Neo was configured to the trap-and-elute workflow. Sample loading utilized combined pressure and flow rate control where the maximum pressure was set to 800 bar and the flow rate capped at 200 μL/min. Peptides were separated on a Thermo Scientific EASY-Spray 75um x 15 cm, 2um(PN: ES75150PN) column heated to 50 °C. Separation was performed over a 21 min active gradient from 4% to 35% B with flow rates starting at 900 nL/min for the first 2 min then decreased to 300 nL/min for 19 min, this is then followed by 3 min of column washing at 500 nL/min at 95%B. Mobile phase A was 0.1% formic acid in water and mobile phase B was 80% acetonitrile and 0.1% formic acid in water. For MS1 analyses, MS1 spectra were acquired in the Orbitrap every 0.6 s at a resolution of 240,000 with a precursor mass range from 380 to 980 m/z in positive mode. Normalized AGC targets were set to 500% with a maximum injection time of 5 ms. For MS2 analyses, data was acquired on the Astral analyzer in DIA mode with a normalized HCD collision energy set to 29% and a default charge state of 2. Non-overlapping DIA precursor isolation windows were set to 2 m/z spanning 380-980 m/z (same as MS1 range) with a maximum injection time of 3 ms and a normalized AGC target value of 500%. Astral analyzer scan range was set to 150-2000 m/z.

Peptide identification protein inference and quantitation

2.8

Mass spectrometry data were processed using Spectronaut software (v19, Biognosys AG, Schlieren, Switzerland), employing default parameters unless otherwise specified. Peptide identification and protein inference were performed against the Homo sapiens sequences (83,347 sequences) downloaded from UniProt (SwissProt + TrEMBL) on May 2, 2025. The analysis was performed using Spectronaut's directDIA (Data-Independent Acquisition) workflow. Search parameters were configured assuming fully specific Trypsin digestion with up to two missed cleavages. Peptides considered for identification ranged from a minimum of 7 to a maximum of 52 amino acids in length. Carbamidomethylation of cysteine was set as a fixed modification, while acetylation (protein N-terminus) and oxidation of methionine were considered as variable modifications. False Discovery Rates (FDRs) were estimated using the target-decoy strategy, with a cutoff of 1% (q-value ≤0.01) applied at both the precursor and protein group levels. Protein inference was conducted based on the principle of parsimony, grouping identified peptides to the minimum set of proteins that sufficiently explain the observed peptide evidence. Quantitative information was extracted at the MS2 level, according to Spectronaut's established DIA quantitation algorithms. To account for variations in sample loading and instrument response, cross-run normalization was applied.

Downstream data analysis

2.9

Spectronaut's Results were exported and organized using PatternLab for Proteomics [34] import module to enabled PatternLab's TFold Analysis [35] and in house Python scripts for further Gene Ontology Analysis. The proteins quantified in at least 70% of the samples within each group were subjected to statistical and graphical analyses using the R programming environment. Abundance values were log10-transformed prior to analysis. Heatmaps were constructed using the R package pheatmap, based on ANOVA p-values to highlight differential expression patterns and using Euclidean distance and the complete linkage method for hierarchical clustering of the proteins. Additionally, row-wise normalization (z-score) was applied to highlight relative abundance patterns across conditions for each selected protein. Gene Ontology (GO) enrichment analyses were performed using Cytoscape [36] and STRING [37].

TFold analysis for differential abundance assessment

2.10

TFold analysis was used to identify differentially abundant proteins between pairs of experimental conditions by integrating both statistical significance and the magnitude of fold change in a data-driven manner. For each protein, TFold calculates a p-value from a two-sample t-test alongside its associated fold change. Unlike traditional approaches that use a fixed fold-change cutoff, TFold employs a variable threshold that adapts according to the statistical significance of each result. The stringency of the fold-change requirement is governed by a parameter that is automatically optimized by the algorithm in conjunction with the Benjamini-Hochberg [38] procedure for false discovery rate (FDR) control. This optimization seeks to maximize the number of proteins identified as differentially abundant while maintaining the desired FDR. Consequently, proteins with stronger statistical support are subject to more permissive fold-change criteria, whereas those with weaker support must exhibit larger differences in abundance. Only proteins that meet both the adaptive fold-change requirement and the significance threshold (q $[eqn]$ 0.05), as determined by the FDR correction, are ultimately reported as significantly differentially abundant.

Results

3

Characterization of iPSC-derived cortical organoids exposed to HMOs

3.1

To date, cortical organoids represent one of the most suitable models to study early human brain development, due to their remarkable similarities to the human fetal brain [39]. To study the effect of HMOs on neurodevelopment, cortical organoids were generated following a 28-day differentiation semi-guided protocol including 3 phases: neural induction, neural proliferation, and neural differentiation [32,33] (Fig. 1A). Given that previous studies have reported individual variability in HMO-mediated responses among neonates, we employed human cortical organoids derived from two genetically distinct neurotypical donor iPSC lines (WT83 and KOLF), which were previously characterized [9,40]. The cortical identity of organoids was confirmed at 1 month of age, following completion of the differentiation protocol, through immunofluorescence of the forkhead box protein G1 (FOXG1) cortical marker, microtubule associated protein 2 (MAP2) neuronal marker, SRY-box transcription factor 2 (SOX2) and Nestin neural stem cell markers, and paired box protein (PAX6) cortical progenitor marker (Fig. 1B).Fig. 1Generation and Characterization of Cortical Organoids. A) Cortical organoids were generated from induced pluripotent stem cells (iPSCs) following a semi-guided differentiation protocol and cultured until maturation, scale bar = 1000 μm. B) Representative images of organoids revealing cortical identity through immunofluorescence of FOXG1, MAP2, SOX2, Nestin and PAX6. Scale bar = 70 μm.Fig. 1

Due to the molecular variety of HMOs and the inaccessibility of newborn tissue, the extent to which HMOs reach and act upon the neonatal brain is currently unknown [40]. Circulating HMO concentrations average 1-10 μg/mL, with reported interindividual variability driven by maternal HMO composition, lactation stage, and infant metabolism [41,42]. Mothers start producing HMOs during the first trimester of pregnancy, raising the possibility that fetal brain development is exposed to HMOs in utero, with continued exposure postnatally through breastfeeding [[43], [44], [45]]. We selected cortical organoids at 3 months of differentiation, a developmental stage characterized by a predominance of differentiated glutamatergic neurons and glia, emerging network oscillations, and a nascent balance between excitation and inhibition [33,39]. At this stage, cortical organoids display sufficient cellular diversity and electrophysiological activity, providing a physiologically relevant model for quantitative analyses of synaptic density and neuronal arborization. Organoids were exposed to three concentrations of HMO solution: 10 μg/mL, 50 μg/mL, and 100 μg/mL, which span and exceed reported serum levels in infants to account for potential differences in HMO bioavailability and local brain exposure. HMOs were continuously added to the culture media twice a week for 30 days.

After one month of treatment, 4-month-old organoids were collected for synaptic puncta quantification and a bulk proteomic analysis (Fig. 2A). The diameter of post-treatment organoids was measured using GelCount, Oxford Optronix (Fig. 2B). The mean diameter of WT83 control organoids was 1363.2 μm, and no significant difference was observed among the treated groups (10 μg/mL, 1404.1 μm; 50 μg/mL, 1452 μm; 100 μg/mL, 1411.8 μm) (Fig. 2C).Fig. 2HMO Treatment on Cortical Organoids. A) At 3 months of age, organoids were treated with a combination of more than 200 human milk oligosaccharides during every media change for 1 month. Following HMO exposure, organoids were collected for synaptic puncta quantification and a bulk proteomics analysis. Created in BioRender. Martins, A (2025) https://BioRender.com/h1ey5n1B). Representative images of post treatment organoids exposed to 10 μg/mL, 50 μg/mL, 100 μg/mL and not exposed to HMOs. Green rings represent the area of focus of the imaging equipment (GelCount, Oxford Optronix, UK), while the red rings indicate the measured diameter of the organoid. C) Organoid diameter measurements following 1 month of HMO treatment. WT83 organoids (*n=*10) did not show significant differences between groups. However, KOLF organoids (*n=*24) reveal a decrease in size. This reduction could reflect increased neuronal differentiation, leading to decreased proliferation, or it may result from stress associated with transfer to 96 well plate. Data were analyzed by One-Way ANOVA test with Tukey's posttest. Data are expressed as mean ± SEM. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001 indicates statistically significant difference compared to controls WT83 or KOLF and between groups.Fig. 2

However, KOLF organoids exhibited a decrease in size among the groups. While control organoids showed a mean diameter of 1425.3 μm, all treated groups revealed a significant size reduction: 10 μg/mL, 1383 μm (p < 0.05); 50 μg/mL, 1300.8 μm (p < 0.001); 100 μg/mL, 1333.3 μm (p < 0.01). The reduction in diameter may indicate an HMO-induced shift from cellular proliferation towards neuronal differentiation, a hypothesis explored in subsequent analyses.

Exposure to human milk oligosaccharides induces synaptogenesis and neurite outgrowth in cortical organoids

3.2

Synaptogenesis and neurite outgrowth are key neurodevelopmental processes that are particularly active during the critical window of cognitive and motor development in the first 1000 days of life [28]. To assess the impact of HMO exposure, the number of mature excitatory synapses in organoids was quantified. Synapses were defined as the co-localization of pre- and post-synaptic proteins (vesicular glutamate transporter VGLUT1 and HOMER1, respectively) when in contact with neurites (MAP2) (Sup Fig. 1A). Using Imaris software, a surface was created to 3D map neurite volume and co-localized synaptic puncta in 20 μm Z-stacked images (Fig. 3A).Fig. 3Synaptic Puncta Quantification. A). A surface mask was applied to quantify neurite volume and synaptic clefts, defined as regions of overlap between HOMER1 and VGLUT1 puncta colocalized with MAP2. Quantification was performed using n = 4 technical replicates per 3 biological replicates. Scale bar = 3 μm. See also Figure S1B) Results indicate a positive correlation between increasing HMO concentration, synaptic puncta number and neurite volume. A positive correlation between HMO exposure and cell number is also noted at 50 μg/mL or higher, indicating a potential impact on cell proliferation as well. Data were analyzed by the One-Way ANOVA test with Tukey's post test. Data are expressed as mean ± SEM. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 indicates statistically significant difference compared to control WT83 and between groups. ^$^ p < 0.05; ^$$^ p < 0.01; p < 0.0001 indicates statistically significant difference compared to control KOLF and between groups.Fig. 3

Our analysis reveals increased synaptic puncta in treated groups compared to untreated groups in both cell lines (Fig. 3A and B). Treatment with 50 μg/mL resulted in a twofold increase in synapse number in both WT83 and KOLF organoids compared to untreated controls (p < 0.05), with further enhancement observed at 100 μg/mL, consistent with a dose-dependent response (p < 0.0001). Additionally, we observed a consistent increase in neurite volume: WT83 organoids showed a twofold increase at 100 μg/mL (p < 0.001), whereas KOLF organoids exhibited a twofold increase at both 50 μg/mL and 100 μg/mL (p < 0.0001). Our findings indicate that HMOs promote synaptogenesis and neurite outgrowth in human cortical organoids, with responses varying by genetic background. In our model, 50 μg/mL appears to be a critical threshold: at this concentration, both WT83 and KOLF organoids exhibited a significant increase in synaptic puncta, suggesting that 50 μg/mL may represent the minimum effective dose needed to elicit robust neural responses in cortical organoids.

To evaluate whether HMOs influence cellular proliferation, the number of nuclei per area was quantified. No consistent trend was observed across cell lines. In WT83 organoids, nuclear count increased with HMO exposure, showing a significant elevation at 50 μg/mL (p < 0.05) and 100 μg/mL (p < 0.001), indicating a positive association (Fig. 3B). In contrast, KOLF organoids exhibited no correlation between HMO concentration and nuclear density. Furthermore, as organoid diameter did not correspond to the changes in nuclear count, these findings should be interpreted with caution [46].

The proteome of HMO-exposed cortical organoids

3.3

To further examine the effects of HMOs on brain organoid development, we conducted proteomic profiling of cortical organoids derived from both KOLF and WT83 iPSC lines. Organoids from each condition (exposed to 0, 10, 50, or 100 μg/mL HMOs, n = 8 per group in triplicate) were subjected to mass spectrometry-based analysis. Across all samples, we identified 7940 proteins, of which 6072 exhibited statistically significant changes in abundance (FDR <0.05), allowing for the identification of dose- and lineage-dependent proteomic signatures.

Preliminary comparative analysis revealed that KOLF and WT83 organoids exhibit different sensitivities to HMO exposure. For the KOLF line, principal component analysis (PCA) revealed a moderate separation between treatment groups along PC1 and PC2, with HMO-treated samples clustering away from control in a dose-dependent manner (Fig. 4A). In contrast, the WT83 PCA displays a clearer segregation, with PC1 capturing over 30% of the variance and distinctly separating all three HMO concentrations from the control group. Consistent with this trend, a combined dendrogram shows dose-dependent separation of treated KOLF samples from controls, while WT83 displays greater sensitivity to 10 μg/mL and 50 μg/mL HMO concentrations (Sup Fig. 2A). In KOLF, the number of differentially abundant proteins increased proportionally with HMO concentration (at 10 μg/mL: 275, at 50 μg/mL: 368, at 100 μg/mL: 600), whereas WT83 displayed a more pronounced response at intermediate doses of 10 and 50 μg/mL (at 10 μg/mL: 1780; at 50 μg/mL: 2646; at 100 μg/mL: 422) (Sup Fig. 2B). To further interpret these changes, a Gene Ontology (GO) enrichment analysis was performed on proteins with differential abundance between untreated and treated groups (Fig. 4B). This analysis revealed the upregulation of terms associated with synaptic structures, including synapse, glutamatergic synapse, focal adhesion, and vesicle, confirming our previous observations by immunostaining that HMOs impact synaptic density and neuronal connectivity. However, we observed upregulation of several terms related to the spliceosomal complex, suggesting that HMO exposure can modulate pathways beyond neuronal maturation.Fig. 4The proteome of HMO-exposed cortical organoids. A) Principal component analysis (PCA) of variance-stabilized Proteomic sequencing data showing sample separation according to HMO treatment condition (10 μg/mL, 50 μg/mL and 100 μg/mL). For the WT83 cell line, the first two components explain 31.34% (PC1) and 18.76% (PC2) of the variance; for KOLF, 20.07% (PC1) and 14.62% (PC2). B) Volcano plot generated using the TFold module within PatternLab V. The analysis uses a t-test combined with an optimized, variable fold-change cutoff to maximize the number of identified proteins while maintaining a Benjamini-Hochberg false discovery rate (FDR) of 0.05. Each point represents a single protein, plotted by its log_2_(fold change) on the x-axis and its -log_10_(t-test p-value) on the y-axis. Proteins in blue are considered statistically differentially abundant (FDR ≤0.05 and pass the fold-change cutoff). Proteins in green satisfy the fold-change cutoff but not the FDR threshold; their classification is therefore defined by the Benjamini–Hochberg adjustment rather than the plotted p-value. Red dots represent proteins that do not meet any criteria. C) The most enriched “Cellular Component” GO terms for differentially abundant proteins in WT83 and KOLF organoids out of 7940 identified proteins, analyzed using STRING and Cytoscape. The two conditions shown are: WT83 at 50 μg/mL and KOLF at 100 μg/mL. For each term, the x-axis indicates enrichment signal, circle size reflects the gene count, and color corresponds to the statistical significance (FDR).Fig. 4

Proteomic profiling indicates lineage and dose-specific responses to HMOs across neurodevelopmental pathways

3.4

To delineate neuronal-specific responses to HMO exposure, we conducted enrichment analysis for neuron-associated proteins across all treatment conditions, resulting in the identification of 1094 unique gene IDs. From this curated subset, the top 25 most differentially expressed proteins were selected for each cell line and visualized using hierarchical clustering heat maps (Fig. 5A and B). This analysis revealed cell line–specific expression trajectories and dose-dependent modulation of neurodevelopmental protein networks, providing a refined view of the molecular programs engaged during cortical organoid maturation.Fig. 5Proteomic Signature of HMO-Treated Organoids. Heatmaps were generated in R using the pheatmap package to show the relative abundance (z-score) of hierarchically clustered proteins. Red indicates high abundance; yellow indicates low. A) Top 25 differentially abundant neuron-associated proteins. B) Relative expression of neuron-associated mRNA EFNB1, HOMER1and ENO2 was analyzed by qPCR. The values were normalized with the endogenous HPRT1 gene. Data were analyzed using One-Way ANOVA test with Tukey's post-test. Values were obtained from three biological replicates and expressed as mean ± SEM. ∗p < 0.05; ∗∗p < 0.01 indicates statistically significant differences when compared to the control. C) Top 50 differentially abundant splicing-associated proteins. D) Relative expression of splicing-associated mRNA LSM3, HSPA8 and RBM39 was analyzed by qPCR. The values were normalized with the endogenous HPRT1 gene. Data were analyzed using One-Way ANOVA test with Tukey's post-test. Values were obtained from three biological replicates and expressed as mean ± SEM. ∗p < 0.05; ∗∗p < 0.01 indicates statistically significant differences when compared to the control.Fig. 5

In KOLF-derived organoids, treatment with 50 μg/mL HMOs resulted in the upregulation of proteins associated with synaptogenesis and structural plasticity, including Vimentin, SLIT2, and Ephrin-B3, as well as markers linked to neuronal migration such as Vinculin and Filamin A [[47], [48], [49], [50], [51]]. Exposure to 100 μg/mL further enhanced expression of proteins indicative of neuronal maturation (Enolase 2, Microtubule-associated threonine 1), cytoskeletal remodeling (beta-tubulin TUBB2B), and synaptic activity (nerve growth factor, VGF, and sodium/potassium transporting ATPase subunit alpha-3, ATP1A3) [[52], [53], [54], [55], [56]]. In contrast, both concentrations led to significant downregulation of growth associated protein GAP43, a marker of immature neurons and regulator of axonal remodeling [57], Teneurin-2 (TENM2), a synaptic adhesion molecule critical for inhibitory synapse formation [58], and complement C1q binding protein (C1QBP), a mitochondrial regulator implicated in synaptic function and neurodevelopmental signaling [59]. Together, these findings suggest that HMO exposure promotes a shift from immature to more mature neuronal states in KOLF cortical organoids, marked by enhanced synaptic and structural complexity.

Although distinct protein signatures were identified, WT83-derived organoids appear to engage in similar neurodevelopmental pathways. At 10 and 50 μg/mL, we observed an upregulation of proteins associated with synaptic maturation and plasticity (Ephrin-B1), neurite outgrowth (IQ motif containing GTPase activating protein 1, IQGAP1), neural stem cell self-renewal (nuclear receptor subfamily 2 group E member 1, NR2E1), and cortical identity (FOXG1) [[60], [61], [62], [63]]. At 100 μg/mL, the proteomic profile shifted to include elevated expression of glial fibrillary acidic protein (GFAP) and apolipoprotein E (APOE), two canonical astrocyte markers, suggesting an HMO-induced acceleration of astrocyte differentiation [64,65]. Notably, HMO exposure led to a reduced expression of MAP6, a microtubule-associated protein previously found to be downregulated during synaptic activation [66], and calcium dependent secretion activator (CADPS), a calcium-dependent regulator of synaptic vesicle exocytosis [67]. Additionally, we observed decreased levels of Neurofascin (NFASC), a neural adhesion molecule previously shown to be downregulated by maternally transferred autoantibodies without overt developmental consequences [68]. These findings raise the possibility that certain HMOs may mimic maternal autoantibody effects, modulating adhesion-related pathways during early brain development.

qPCR of selected targets TUBB2B, ENO2 and EFNB1 showed expression changes in both KOLF and WT83 cell lines consistent in direction with our proteomic results. Although the magnitude of change did not reach statistical significance for all genes, EFNB1 exhibited a significant change, supporting the proteomic trend observed (Sup Fig. 3A).

Compared to KOLF, WT83-derived organoids exhibited greater sensitivity to HMO exposure, showing proteomic responses as early as 10 μg/mL, whereas KOLF organoids responded only at higher concentrations (≥50 μg/mL). Nevertheless, both lines seem to converge on shared neurodevelopmental programs, including pathways involved in synaptic maturation, neuronal differentiation, and neurite outgrowth.

Proteomic analysis of cortical organoids exposed to HMOs reveals upregulation of mRNA splicing pathways

3.5

Interestingly, a comparative analysis of proteins differentially abundant in KOLF and WT83 organoids exposed at 100 μg/mL HMO revealed substantial overlap in both upregulated and downregulated protein sets. 278 upregulated and 176 downregulated proteins common to both genotypes were identified (Sup Fig. 4A). Despite genotype-specific responses, GO enrichment of uniquely upregulated proteins in each line converged on RNA splicing pathways, including mRNA splicing via the spliceosome and RNA splicing via transesterification reactions. The integrated protein-protein interaction (PPI) network highlights key nodes within these pathways, including proteins in the HNRNP and SRSF families, underscoring a shared upregulation of splicing machinery following HMO exposure (Sup Fig. 4B). To further explore the splicing mechanisms in action, we performed a targeted enrichment analysis of splicing-associated proteins across all treatment conditions. This analysis identified 306 differentially abundant proteins, with the top 50 most significantly altered proteins visualized in a heatmap (Fig. 5C and D).

We observed increased levels of multiple splicing regulators that influence synaptic function and neuronal connectivity, including NOVA alternative splicing regulator 1 (NOVA1), RNA binding motif protein (RBM39), serine and arginine rich splicing factor (SRSF1), and synaptotagmin binding cytoplasmic RNA interacting protein (SYNCRIP) [[69], [70], [71], [72]]. Additionally, several members of the heterogeneous nuclear ribonucleoprotein family (HNRNPK, HNRNPL, and HNRNPM) displayed differences in protein abundance, suggesting a broader modulation of alternative splicing programs relevant to neurodevelopment through neuronal differentiation and maturation pathways [73]. Notably, at 100 μg/mL HMO treatment, the only two proteins consistently upregulated in both lines were the heat shock protein 8 (HSPA8), a protein involved in SNARE complex assembly and important for neurotransmitter release [74], and U6 snRNA and mRNA degradation associated protein (LSM3), a component of the spliceosomal machinery [75]. Influenza virus NS1A binding protein (IVNS1ABP) and polypyrimidine tract binding protein 1 (PTBP1) were specifically upregulated in WT83 organoids, two factors previously implicated in regulating the differentiation of neuronal progenitor cells [76,77]. The upregulation of HSPA8 and LSM3 proteins was also observed through qPCR, although changes in transcriptional levels were not observed for RBM39, HNRNPM and SRSF1 (Sup Fig. 3A). Indeed, prior studies have shown that transcript and protein abundances often diverge, particularly in the brain, where extensive post-transcriptional regulation contributes to this decoupling [78]. The significant enrichment of RNA splicing pathways suggests a potential mechanistic link between HMO exposure and the post-transcriptional regulation of genes essential for neural differentiation and synaptic maturation.

The coordinated upregulation of key splicing regulators, including factors known to modulate neuron-specific isoform expression, raises the possibility that HMOs promote synaptogenesis and neurite outgrowth, at least in part, through alternative splicing programs, shaping the neuronal proteome during early development.

Discussion

4

Human milk, the optimal source of early-life nutrition, is enriched with bioactive components that extend far beyond basic sustenance [[1], [2], [3], [4]]. Among these, HMOs have emerged as key modulators of infant health, yet their potential roles in neurodevelopment remain poorly defined [14,20]. In this study, we hypothesized that HMOs promote neuronal maturation during early neurodevelopment using human cortical organoids, a physiologically relevant model of the developing human brain [79]. The results described here provide the first evidence of HMO activity in human in vitro brain cells, advancing previous findings from animal models and bridging a significant gap in the literature. Given the limited presence of HMOs in infant formula [20], our study highlights the importance of further exploring the possible contribution of HMOs to neurodevelopment.

The elaboration of dendrites, axons, and synapses in the cortex peaks during the first year of life [80]. Glutamatergic synapses expressed through VGLUT1, which are known to appear around the third trimester, provide a reliable reference for assessing postnatal neurodevelopment [81]. Our findings support a potential role for HMOs in early neurodevelopment, as reflected by dose-dependent increases in synaptogenesis and neurite outgrowth. Our data aligns with the hypotheses of several previous studies evaluating other markers of brain maturation. Positive associations have been observed between HMO availability and increases in grey and white matter volume, as well as in myelination among neonates [14,[17], [18], [19],22]. Furthermore, animal studies have consistently demonstrated that specific HMO species enhance cognitive performance, as evidenced by improvements in memory and learning tasks compared to HMO-deprived counterparts [[23], [24], [25], [26], [27]]. Additionally, modeling this process using cortical organoids provides evidence that HMOs can interact with brain cells. However, further studies are needed to establish direct effects on functional brain connectivity.

While both cell lines demonstrated enhanced neuronal maturation, the specific dynamics revealed nuanced, lineage-specific responses. In KOLF organoids, for example, the consistent expansion of neurite volume across HMO concentrations may reflect progressive changes in neuronal network structure, although further studies are needed to confirm this. In contrast, WT83 organoids exhibit a significant increase in neurite volume only at 100 μg/mL, aligning with proteomic data showing downregulation of microtubule-associated proteins and upregulation of factors involved in neural stem cell maintenance at 10 and 50 μg/mL. These findings suggest that, in WT83, lower HMO concentrations may preferentially activate pathways involved in progenitor maintenance and early neurogenesis, or may be insufficient to trigger structural maturation, thereby delaying neuronal network expansion until higher exposure levels are reached. At 10 μg/mL, organoids did not display significant difference in any of the functional assays performed here. However, it remains possible that lower-level perturbations could exert indirect effects later in development, a timeframe not captured in the present study.

Proteomic profiling revealed that the two cell lines respond differently to the same concentration of HMOs. Variations in HMO response have been reported in many studies before, but without a reproducible and accessible model, they were never deeply investigated [[16], [17], [18], [19]]. In this study, we identified that KOLF-derived brain organoids responded in a dose-dependent way, increasing the number of DEGs as concentration increases and mainly activating pathways related to neuronal differentiation, synaptogenesis, and connectivity, while downregulating markers linked to immature neurons and synaptic inhibition. In contrast, WT83-derived brain organoids displayed more sensitivity to 10ug/mL and 50 μg/mL, exhibiting bigger fluctuations in activated pathways. The proteome of WT83 organoids was marked by early induction of neural stem cell maintenance and neurite extension programs and followed by a transition toward astrocytic protein expression at higher doses, suggesting that HMOs may also modulate glial-neuronal interactions. Additionally, HMOs seem to regulate metabolic pathways in both lines, with KOLF showing an increase in ATP1A3 and IGF2BP1, which influence energy metabolism, while WT83 upregulates APOE, indicating a role in lipid metabolism [56,65]. These data suggest that HMOs elicit complex, lineage and dose-specific effects on cortical development, integrating synaptic maturation, cell-type specification, and metabolic regulation. Nevertheless, synaptic puncta increased consistently across both lines. And at 100ug/mL, both lines had significantly higher synaptic density and neurite volume, consistent with normal development of the brain circuits [80].

The upregulation of spliceosome-related proteins raises the possibility that HMOs impact RNA processing and gene regulation machinery, suggesting broader systemic effects on neural cell biology. We observed a consistent upregulation of the heat shock protein HSPA8 in both KOLF and WT83 organoids exposed to HMOs in a dose-dependent manner. HSPA8 is known to support neurodevelopment by facilitating SNARE complex assembly and synaptic vesicle recycling, thereby promoting efficient neurotransmitter release and synaptic function [74]. In a non-dose-dependent manner, we also identified increased expression of splicing modulators SRSF1 and RBM39 in both lines, which are involved in neuronal differentiation and the regulation of alternative splicing events essential for synaptogenesis [70,71]. Interestingly, PTBP1, a key regulator that represses premature neuronal differentiation and maintains neural progenitor identity, was specifically upregulated in WT83 organoids, consistent with previous data showing high levels of NSC maintenance protein NR2E1 and suggesting, once again, that individual genotypes respond uniquely to the presence of HMOs [77]. These findings align with previous work in mouse models showing that short-term changes in gene expression, particularly due to 6′-sialyllactose (6′SL), can alter neuronal patterning [23]. Our results extend this idea by suggesting that HMOs, through the upregulation of splicing regulators, actively shape the neuronal proteome.

In this study, we observed that HMOs modulate synaptic density and neural connectivity during a critical window in brain development using a relevant human model system. In addition, our findings demonstrate that brain organoids derived from different individuals respond uniquely to the same concentrations of HMOs, suggesting that the natural diversity of HMOs in human milk may be evolutionarily tailored to match the infant's genetic background. Nevertheless, infant formulas remain an essential alternative for mothers unable to breastfeed, underscoring the importance of deepening our understanding of HMOs and their biological effects. The current lack of clinical and neuroimaging studies focused on the role of HMOs on a cellular level highlights a significant gap in the field. Given the critical influence of early-life nutrition on long-term neurodevelopmental outcomes and its profound implications for public health, further research using human-based models is essential to unravel the precise mechanisms by which HMOs contribute to brain development.

Limitations of study

5

This study was designed as an exploratory investigation to assess whether developing human cortical organoids exhibit molecular and structural responsiveness following HMO exposure. Several limitations should therefore be acknowledged. First, untreated organoids served as the sole control condition. The absence of an unrelated carbohydrate comparator precludes conclusions regarding HMO specificity. The observed molecular changes may therefore reflect broader carbohydrate-responsive signaling mechanisms, including activation of surface G protein-coupled receptors. Future studies incorporating defined non-HMO sugars will be required to resolve structural specificity. Second, a pooled HMO preparation was used, preventing attribution of effects to individual oligosaccharide species. Dissecting the contribution of defined fucosylated or sialylated HMOs will be essential to establish mechanistic resolution. Third, the extent to which HMOs reach the developing brain and cross the blood-brain barrier in vivo remains unclear [40]. While we have empirically defined 50 μg/mL as a working concentration for this study, our proteomic analysis reveals that at lower concentrations (10 μg/mL), WT83-derived organoids already display a significant proteomic signature. Accordingly, the organoid system should be interpreted as a reductionist platform to test potential direct neural responsiveness under controlled exposure conditions, rather than as a physiological model of in vivo HMO distribution.

Author contributions

A.R.M. and L.B. conceptualized and supervised the study. L.C. performed data acquisition, analysis, and drafting of the manuscript. N.M. and C.S. contributed to data interpretation and analytical method development. K.S. isolated and purified the HMOs. A.M., P.C.C., J.C., B.T., and I.L. performed proteomic analysis. All authors critically reviewed, edited, and approved the final manuscript.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:Alysson Muotri reports financial support was provided by National Institutes of Health. Paulo Carvalho reports financial support was provided by Brazil's National Council for Scientific and Technological Development. Lars Bode reports financial support was provided by Family Larsson-Rosenquist Foundation. Alysson Muotri reports a relationship with Tismoo that includes: board membership and equity or stocks. Alysson Muotri has patent issued to US11821895B2. Alysson Muotri has patent issued to US9340775B2. Alysson Muotri has patent issued to US9725695B2. Alysson Muotri has patent issued to US20150119327A1. Alysson Muotri has patent issued to WO2011017404A3. Alysson Muotri has patent issued to US20180306780A1. Alysson Muotri has patent issued to US9696297B2. Alysson Muotri has patent issued to EP4222268A1. Alysson Muotri has patent issued to US20240277712A1. Alysson Muotri has patent issued to WO2024031058A2. Lars Bode has patent issued to US9675649B2. Lars Bode has patent issued to US10160986B2. Lars Bode has patent issued to WO2019169188A1. Lars Bode has patent issued to US20200230161A1. Dr. Muotri is a co-founder and has an equity interest in TISMOO, a company dedicated to genetic analysis and human brain organogenesis, focusing on therapeutic applications customized for the disorder autism spectrum and other neurological disorders. Dr. Muotri is also an inventor on several patents using brain organoids. The terms of this arrangement have been reviewed and approved by the University of California, San Diego, in accordance with its conflict-of-interest policies. Dr. Bode LB is a co-inventor on patent applications related to the use of HMOs in preventing NEC and other inflammatory diseases. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.