Exploring the Immunomodulatory Potential of Human Milk: Aryl Hydrocarbon Receptor Activation and Its Impact on Neonatal Gut Health

Human milk samples were obtained from a surplus of milk donated to the Dutch Human Milk Bank (n = 22). The formula used in this study, Nutrilon neonatal start, Nutricia (Hoofddorp, The Netherlands), was a commercially available product, prepared according to the manufacturers’ recommendations.

Both pasteurized and unpasteurized samples were stored at −20 °C. To separate the milk into skimmed components, milk was centrifuged at 4 °C for 20 min, which resulted in the dissociation of the skim and fat layers. Subsequently, the fat layer of the milk was carefully removed and stored separately at −20 °C.

Neonatal fecal samples were collected in the Amsterdam UMC and MMC Veldhoven (The Netherlands) from healthy full-term (n = 9) and preterm neonates (n = 19) at birth and after 4 weeks, as part of the eNose/NEC study. All the neonates, except for one full-term baby, were breastfed. The fecal samples were extracted from the diapers and immediately stored at −80 °C (Table 1). Study approval was provided by the local institutional review boards of all the participating centers (Amsterdam UMC and Amsterdam VUMC). Written informed consent was obtained from the parents or legal caretakers.

Neonatal fecal samples (preterm n = 19; full-term n = 9) and all the layers of the divided milk (human n = 22; formula, n = 6) were processed for Trp metabolite measurement using liquid chromatography–tandem mass spectrometry (LC-MS). The Trp metabolites’ concentrations and composition from the fecal and milk samples were quantified as previously described [27]. The samples were pretreated as follows: (1) for the fecal samples, 3 mg of lyophilized fecal material was mixed with 900 μL of MeOH/H₂O mixture (1:1), added according to internal standards; and, (2) for the milk samples, 100 μL of milk was mixed in 300 μL of methanol and 100 μL of a solution of internal standards [28].

LC-MS was performed as previously described. A calibration curve was created for each metabolite by calculating the intensity ratio obtained between the metabolite and its internal standard. These calibration curves were then utilized to determine the concentrations of each metabolite in the fecal samples. The specific calibration ranges and internal standard concentrations associated with the fragmentation parameters have been previously described [29].

The Trp metabolism was assessed by measuring the total metabolites involved in each pathway. The sum of the kynurenine pathway consisted of picolinic acid, 3-OH-kyurenin, quinolinic acid, kynurenine, 3-OH-anthanilic acid, kynurenic acid, and xanthurenic acid. The sum of the serotonin pathway consisted of 5-OH-tryptophan, N-acetyl-serotonin, 5-OH indole acetic acid, and serotonin. Finally, the sum of the indole pathway comprised tryptamine, indole-3-acetamide, indole-3-lactic acid, indole-3-aldehyde, indole-3-acetic acid, tryptophol, and indole-3-sulfate. These metabolites and their order are illustrated in Supplementary Figure S1.

The kynurenine pathway was quantified through the activity of the enzyme indoleamine 2,3-dioxygenase (IDO1), which was assessed by measurement of the kynurenine/Trp ratio [15]. The Trp metabolism was assessed by measuring the total metabolites involved in each pathway.

A luciferase-based assay was used to assess AHR activity in response to different samples of human milk, formula, and specific indoles, as described previously [16]. Briefly, mouse kidney H1L1.1c2 cells (kindly shared by Dr. Michael Denison, University of California, Davis, CA, USA), which contained a dioxin response element-driven firefly luciferase reporter plasmid pGudLuc1.1 after the Cyp1a1 gene (the AHR target gene), were seeded in a 96-well plate and stimulated with 10x diluted milk or the indoles for 4 h. Cell lysate was then used to measure luciferase activity using a luminometer. A synthetic AHR-activating positive-control 6-Formylindolo[3,2-b]carbazole (FICZ) as well as a negative control (culture medium) were used to confirm effectiveness and normalize the activity.

Caco2 cells, an immortalized human colon colorectal adenocarcinoma cell line (Caco2 line, passages 5–19; HTB-37; American Type Culture Collection, Manassas, VA, USA), were used as a model for the human intestine [30]. The cells were routinely maintained in Dulbecco’s Modified Eagle Medium (DMEM, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% heat-inactivated Fetal Calf Serum (FCS) (Life Technologies, Inc., Invitrogen, CA, USA), 100 μmg/mL penicillin and streptomycin antibiotic, and L-glutamine (Biocambrex, Verviers, Belgium). The cells were seeded in a 96-well plate and incubated for 24 h at 37 °C in a 5% carbon dioxide humidified atmosphere. The cells were then supplemented for 24 h with FICZ (3.5 nmol/mL), indole-3-acetic acid (IAA; 60 pmol/mL), indole-3-sulfate (I3S; 150 pmol/mL), or indole-3-lactic acid (ILA; 1 nM/mL) (Thermo Fisher Scientific). Finally, the cells were stimulated with tumor necrosis factor (TNFα, 50 ng/mL) (Thermo Fisher Scientific) for 4 h or kept unstimulated. The supernatant was collected and stored at −80 °C, and the cells were collected into TRIzol and stored at −80 °C (Thermo Fisher Scientific) for a subsequent protein and gene expression analysis.

Human fetal intestinal tissues (gestational age 18–20 weeks) were obtained by the HIS Mouse Facility of the Amsterdam UMC from a clinic (The Netherlands). Fetal human organoids (HFO) were generated from isolated crypts issued from fetal small intestine as previously described in [31]. Briefly, after being cut open longitudinally, the intestine was cut into small pieces of ~5 mm and thoroughly washed with ice-cold phosphate-buffered saline (PBS). The tissue was then incubated in 2 mM Ethylenediaminetetraacetic acid (EDTA) diluted in PBS for 30 min at 4 °C. Next, the tissue was washed with PBS + 10% FCS (Bondinco, Alkmaar, The Netherlands), separating the dissociated cells into the supernatant, which was filtered through a cell strainer (70 μm). The isolated crypts were resuspended in Matrigel (Corning, NY, USA) and derived into organoids using a specific growth medium containing advanced DMEM supplemented with Glutamax (Invitrogen, Carlsbad, CA, USA), 1 M of Hepes, 100 U-mg/mL penicillin-streptomycin, a B27 supplement, a neurobasal supplement, EGF (Invitrogen, Carlsbad, CA, USA), n-Acetylcysteine (Sigma-Aldrich, Burlington, NJ, USA), Noggin, and Rspondin (home made by stably transfected HEK cells) and incubated at 37 °C, 5% CO₂. The medium was refreshed every 2–3 days, and the organoids were passaged by mechanical disruption intervals of 6–10 days, as described previously [31].

RNA from Caco2 cells and HFOs was isolated using the bioline ISOLATE II mini kit (BIO-52073, Bioline, Waddinxveen, The Netherlands) according to the manufacturer’s protocol, and RNA quality was checked by Tapestation 2200 (Agilent Technologies Netherlands B.V., Amstelveen, The Netherlands). cDNA was synthesized with Random Hexamer primers (Promega, Leiden, The Netherlands), Oligo dT primers (Thermo Fisher Scientific), deoxyribonucleotide triphosphate (dNTPs) (Thermo Fisher Scientific), 1× RT-buffer (Thermo Fisher Scientific), Revertaid Transcriptase (Thermo Fisher Scientific), and Ribolock RNAse Inhibitor (Thermo Fisher Scientific).

Quantitative RT-qPCR was performed on a BioRad iCycler using the sensifast SYBR No-ROX Kit (GC-biotech Bio-98020, Meridian, Cincinnati, OH, USA) according to the manufacturer’s instructions. The most stable reference genes were selected from a panel of nine reference genes using GeNorm [32]. The relative expression of the genes with N0 values was obtained using LinRegPCR [33], and the values were normalized with the reference genes. Gene expression of the target genes was determined using the LinRegPCR software (version 2021.1) [33]. The primer sequences (Sigma Aldrich) for the AHR gene were F 5′-TTACAGGCTCTGAATGGCTT-3′ and R 5′-TTTTCTGGAGGAATCTGGTCT-3′; those for the IL8 gene were F 5′-AAATTTGGGGTGGAAAGGTT-3′ and R 5′-TCCTGATTTCTGCAGCTCTGT-3′; those for reference genes GAPDH were F 5′-CCATGTTCGTCATGGGTGTG-3′ and R 5′-GGTGCTAAGCAGTTGGTGGTG-3′; and those for CYCLOPHILLIN were F 5′-ACGGCGAGCCCTTGG-3′ and R 5′-TTTCTGCTGTCTTTGGGACCT-3′. The primer sequence for CYP1A1 has been previously described [15].

The IL8 concentration was determined by the sandwich enzyme-linked immunosorbent assay (ELISA; R&D Systems, Abingdon, UK) according to the manufacturer’s protocol. Optical density was measured using a Synergy HT plate reader (BioTEK, Beun de Ronde, Abcoude, The Netherlands).

Immunofluorescence staining of AHR was performed on Caco2 cells. The cells were fixed with 4% paraformaldehyde and then permeabilized with 0.1% Triton™ X (Sigma-Aldrich). The cells were blocked with 2% Bovine Serum Albumin (BSA). The cells were labelled with Mouse Anti-AHR Monoclonal Antibody (dilution: 1:50, FF3399, eBioscience™ (Product # 14-9854-82) overnight and labelled with Goat anti-Mouse IgG (H+L), Superclonal™ Recombinant Secondary Antibody, and Alexa Fluor 488. The nuclei (dilution: 1:2000; panel b: blue) were stained with SlowFade^® Gold Antifade Mountant with DAPI (Thermo Fisher Scientific). The images were captured at a 60× magnification and processed using ImageJ (Rasband, WS., ImageJ, U. S. National Institutes of Health, Bethesda, MD, USA, version 1.54f, 29 June 2023).

The HFOs were split into single cells and seeded at 10⁵ cells/mL onto cell culture inserts for 24-well plates with pores 0.4 µm in diameter (cellQART^®, Product #932040). The HFOs were kept in a culture for 3 weeks as described in the Section 2.6. The HFOs were then treated with the indoles and FICZ prior to TNFα challenge for 72 h. A measurement of the trans-epithelial electrical resistance (TEER) was used to characterize the formation of a tight intestinal cell monolayer. TEER electrodes were recorded everyday with an EVOM-resistant meter (WPI, Sarasota, FL, USA), with one electrode placed on the luminal side and the other electrode on the basolateral side, with the two different electrodes only being separated by the intestinal monolayer. The TEER measurements of the filters with the cells were subtracted by blank filters, and the result was multiplied by the membrane area to obtain the TEER measurement in Ω·cm².

All analyses and graphical representations were carried out using GraphPad Prism version 9.1.0 (San Diego, CA, USA). The data were analyzed for normal distribution using the Anderson–Darling test, the d’Agostino and Pearson test, the Shapiro–Wilk test, and the Kolmogorov–Smirnov test. For the data which failed the normality test, nonparametric tests were used to analyze significant differences. Two-group analyses were performed using the two-tailed Student’s t test or the nonparametric Mann–Whitney test. The median (interquartile range) and means ± SEM were used to present the data from one of two independent experiments. Multiple comparisons were performed using a one-way analysis of variance (ANOVA) and the post hoc Tukey test or nonparametric Kruskal–Wallis test followed by a post hoc Dunn test. Significance was considered with p < 0.05. The different tests are specified in each figure legend.

Our initial objective was to investigate whether there is a differential abundance of Trp metabolites in the feces of preterm neonates compared to full-term neonates. To this end, LC-MS was employed to determine the composition and concentration of Trp metabolites in the feces of a Dutch cohort of preterm (n = 19) and full-term (n = 9) neonates, at birth and 4 weeks postnatal (Figure 1A). The Trp metabolomic quantification focused on the three metabolic pathways of Trp, namely, the kynurenine, serotonin, and indole pathways (Figure S1A). No difference was observed in the fecal concentration of Trp between preterm and full-term neonates at birth and 4 weeks postnatal (Figure 1B). In addition, at birth, no significant differences in Trp metabolites were observed between preterm and full-term neonates if the sum of the metabolites involved in each Trp metabolic pathways was calculated (Figure 1C). The full-term neonates showed higher levels of Trp metabolites in all three Trp metabolic pathways (Figure 1C). Furthermore, at a single-metabolite level, the preterm fecal samples displayed significantly lower levels of xanthurenic acid and serotonin and higher levels of 3-OH-kynurenine (Figure 1D). In the kynurenine pathway, kynurenine was elevated in the preterm neonates, whereas kynurenic acid was elevated in the full-term neonates (Figure S2B). At 4 weeks postnatal, the preterm neonates showed no significant difference in the sum of metabolites involved in the kynurenine pathway (Figure 1C). However, at a single-metabolite level, picolinic acid, kynurenic acid, and xanthurenic acid were significantly higher in the feces of the full-term neonates (Figure 1E). Furthermore, at 4 weeks postnatal, the preterm neonates displayed significantly lower fecal levels when regarding the total Trp metabolites involved in the serotonin pathway compared to the full-term neonates (Figure 1C). This difference was mainly explained by a significant reduction in serotonin, N-acetyl-serotonin, and 5′HT-OH-indole acetic acid in the feces (Figure 1E). Moreover, the preterm neonates showed significantly lower levels of total Trp metabolites involved in the indole pathway compared to the full-term neonates (Figure 1C). Particularly, the preterm neonates showed significantly lower levels of indole-3-lactic acid (ILA), the main contributor in the sum of indoles in neonates (Figure 1E). Furthermore, the preterm neonates displayed a significantly lower conversion of ILA from Trp in comparison to the full-term neonates, as shown through the ILA-to-tryptophan ratio (Figure 1F). In addition, we confirmed the ability of ILA to activate AHR in a dose-dependent manner, with the concentration of ILA measured in the feces of preterm neonates being too low to activate AHR (Figure 1G). Moreover, when looking at associations between the preterm and full-term metabolites, the ILA-to-indole-3-aldehyde (I3Ald) ratio displayed the highest association signature power linked to full-term neonates (AUC, 0.905, p = 4.0644 × 10⁻⁴, Figure S2D). The rest of the measured Trp metabolites were not significantly different (Figures S2A–C and S3A–C).

We next profiled the Trp metabolites in human milk (HM, n = 22) and commercially available infant milk formula (F, n = 3) (Figure 2A). When comparing human milk and formula, no significant difference in the Trp levels was observed (Figure 2B). We then assessed the sum concentrations of all the metabolites contained in each distinct Trp metabolic pathway. No difference was observed between human milk and formula in the sum concentrations of the metabolites from the kynurenine pathway (Figure 2C). However, at a single-metabolite level, picolinic acid was significantly higher while 3OH-anthranilic acid, xanthurenic acid, kynurenic acid, and kynurenine were significantly lower in the formula compared to human milk (Figure 2D). Interestingly, human milk contained significantly higher levels of metabolites from the serotonin pathway compared to formula (Figure 2C). This reflected the relatively high concentration of 5-OH-indole acetic acid in human milk (Figure 2D). All the other measured Trp metabolites were not significantly different (Figure S3A–C). With these data, our next step was to investigate whether the bacterial-derived metabolites were different in the two types of milk. We then assessed the composition and levels of the metabolites representing the indole pathway in human milk and commercial infant formula. The total indole metabolites’ level was significantly higher in the human milk samples compared to the formula (Figure 2C). Specifically, indole-3-acetic acid (IAA) and indole-3-sulfate (I3S), two Trp metabolites known for being AHR agonists, were detected at significantly higher levels in human milk (Figure 2D).

Here, we investigated the functional relevance of neonatal ILA and human milk IAA and I3S, focusing on their ability to trans-activate the downstream signaling of AHR in epithelial cells. First, we investigated the capability of high (full-term) and low (preterm) ILA levels to activate AHR in a dose-dependent manner using a mouse luciferase reporter cell system. At the concentration detected in the feces of preterm neonates, ILA was not able to activate AHR, while the full-term ILA concentration exhibited a high activation of AHR (Figure 1F and Figure 3A). We then investigated whether the levels of IAA and I3S detected in human milk could similarly trigger AHR activation. Both IAA and I3S significantly activated AHR at the physiological concentrations measured in human milk (Figure 3B).

AHR metabolites exhibit varying affinities for the human AHR receptor compared to their interaction with the mice AHR receptor [31]. Thus, we investigated whether ILA, IAA, and I3S would similarly activate AHRs in human epithelial cells. To visually track the translocation of AHR to the nucleus of human epithelial cells, we challenged Caco2 cells with ILA, IAA, and I3S, at the levels detected in the feces and the milk, and with the synthetic AHR activator FICZ. Notably, in the untreated Caco2 cells, AHR localization was confined exclusively to the cytoplasm (Figure 3C,D), whereas challenge with IAA, I3C, ILA, and FICZ induced the translocation of AHR to the nucleus (Figure 3C,D). Moreover, the nuclear translocation of AHR upon challenge with ILA, IAA, and I3S was fully mitigated with an AHR inhibitor (CH223191↗) (Figure 3C), confirming the AHR-specific route of action of these indoles.

Having confirmed the ability of IAA, I3S, and ILA to trigger AHR activation at physiological concentrations, along with the subsequent downstream target gene activation (CYP1A1), we then studied the possible role of these AHR-activating metabolites in reducing cytokine-induced epithelial IL8 production in Caco2 and human fetal intestinal organoids (HFOs). In both Caco2 cells and HFOs, TNFα stimulation induced the production of IL8. Guided by findings from our Trp metabolite analysis, we then exposed the Caco2 cells and HFOs to the highest concentrations of AHR activation-inducing indoles observed in the human milk and full-term fecal samples, as described in Figure 4A. In the Caco2 cells, TNFα significantly increased the expression of IL8, which was significantly reduced by IAA, ILA, and FICZ pretreatment (Figure 4B). However, treatment with I3S did not yield such a significant affect (Figure 4B). Upon treatment with CH223191↗, ILA, IAA, and FICZ lost their ability to reduce the expression of IL8 (Figure 4C). The same results were obtained when measuring the production of IL8 in the supernatant of the cells (Figure 4D). In addition, gene expression of the AHR downstream target gene CYP1A1 was induced in all the conditions, confirming specific AHR activation upon ILA, IAA, I3S, and FICZ challenge (Figure 4E). Upon treatment with CH223191↗, ILA, IAA, and FICZ lost their ability to activate AHR, shown by a decreased expression of the AHR target gene CYP1A1 (Figure 4E).

To corroborate these findings, we employed human HFOs, a complex in vitro model which more closely resembles the immature intestinal environment of neonates. The HFOs were derived from the intestines of fetuses from 18 to 22 weeks of gestation and then pretreated with ILA, IAA, I3S, and FICZ prior to TNFα challenge (Figure 5A). Upon ILA and IAA treatment, the HFOs produced significantly lower levels of TNFα-induced IL8 protein in the supernatant of the culture (Figure 5B). No difference was evident in TNFα-induced IL8 production after FICZ and I3S pretreatment. Finally, barrier integrity was measured in the HFOs grown two-dimensionally in transwell systems. The HFOs were exposed apically to the IAA and I3S measured in human milk and challenged with TNFα for 72 h. Upon TNFα exposure, the HFOs showed a significant loss in barrier integrity, which was characterized by a reduction in TEER (Figure 5C). The HFOs exposed to ILA and IAA were able to maintain a stable TEER from TNFα-induced barrier loss, whereas I3S showed a significant reduction in barrier integrity (Figure 5C,D and Figure S5).

These data demonstrate that physiological concentrations of IAA and ILA can reduce the production of the neutrophil chemotactic factor IL8 in fetal intestinal epithelial cells and protect barrier integrity.