Antibiotic use in early life impairs MAIT cell–mediated immunity in adulthood

In a multicenter pediatric study, 73% of children received at least one course of antibiotics during their first year of life (Wu et al., 2016). While antibiotics can be beneficial, 26% of children across 32 hospitals were prescribed an antibiotic that was suboptimal due to ineffectiveness for the target bacteria, extended prophylaxis, or overly broad treatment (Tribble et al., 2020). The widespread and often inappropriate use of antibiotics during infancy is associated with deleterious health outcomes later in life, including asthma and recurrent respiratory tract infections (Aversa et al., 2021; Donovan et al., 2020; Mitre et al., 2018; Ni et al., 2019; Patrick et al., 2020; Zhou et al., 2021). Although antibiotics have been suggested to deplete commensal microbes that promote respiratory immunity (Donald and Finlay, 2023; Tamburini et al., 2016), the mechanisms by which this occurs have not been established, preventing the development of effective therapeutics.

While conventional T cells recognize peptides presented by polymorphic major histocompatibility complex (MHC) proteins, many innate-like T cells are specific for non-peptidic microbial antigens, including lipids and metabolites, presented by monomorphic MHC class Ib molecules (Godfrey et al., 2015). Mucosal-associated invariant T (MAIT) cells comprise the predominant subset of innate-like T cells in most human tissues (Provine and Klenerman, 2020), reaching up to 9% of pulmonary T cells in healthy individuals (Hinks et al., 2016). MAIT cells express semi-invariant T cell receptors (Vα19-Jα33 in mice and Vα7.2-Jα33/20/12 in humans) that recognize microbial derivatives of riboflavin (vitamin B2) synthesis presented by the MHC class-I–related (MR1) molecule (Kjer-Nielsen et al., 2012; Provine and Klenerman, 2020). Because MR1 is highly conserved between humans and mice (Huang et al., 2009), MAIT cells from both species respond to the same microbial metabolites. Riboflavin synthesis is broadly conserved among bacteria and fungi, so MAIT cells produce IFN-γ and/or IL-17A in response to a wide array of respiratory pathogens, with protection in murine models demonstrated for Francisella tularensis, Klebsiella pneumoniae, and Legionella longbeachae (Georgel et al., 2011; Gold et al., 2010; Hartmann et al., 2018; Hinks et al., 2016; Jahreis et al., 2018; Le Bourhis et al., 2010; Meierovics et al., 2013; Wang et al., 2018; Zhao et al., 2021). While viruses do not synthesize metabolites, MAIT cells express receptors for the cytokines IL-12, IL-15, and IL-18, enabling them to release IFN-γ and granzyme B in response to viral infections (Ussher et al., 2014; van Wilgenburg et al., 2016), with protection against influenza shown in mice (van Wilgenburg et al., 2018). The importance of these MR1-restricted T cells has been demonstrated in humans, where a single-nucleotide polymorphism in the Mr1 gene is associated with increased susceptibility to tuberculosis (Seshadri et al., 2017), while a patient with recurrent bacterial and viral infections was found to have a nonfunctional MR1 variant (Howson et al., 2020). In children with bacterial or viral pneumonia, 30% of pulmonary MAIT cells produced IL-17A compared to only 2% of conventional CD4⁺ T cells (Lu et al., 2020), demonstrating their outsized contribution to respiratory immunity.

Commensal microbes promote the development of MAIT cells through their synthesis of riboflavin intermediates (Constantinides et al., 2019; Legoux et al., 2019). Although MAIT cells are nearly absent from the peripheral tissues of germ-free (GF) mice (Constantinides et al., 2019; Koay et al., 2016; Legoux et al., 2019; Treiner et al., 2003), microbial colonization of 1-wk-old GF neonates, but not 7-wk-old GF adults, is sufficient to restore MAIT cell abundance (Constantinides et al., 2019). Since MAIT cells require early-life exposure to riboflavin-synthesizing commensals (Constantinides and Belkaid, 2021), we hypothesized that antibiotic use during this period may impede their development and subsequently impair respiratory immunity. Here we demonstrate that riboflavin-synthesizing bacteria are transiently enriched in early life and antibiotic treatment during this period is sufficient to deplete these commensals and hinder MAIT cell development. The resulting decrease in MAIT cell abundance renders adult animals more susceptible to bacterial pneumonia. Administering a riboflavin-synthesizing commensal during antibiotic treatment restores MAIT cell development and respiratory immunity, suggesting that probiotics may counter the detrimental effects of antibiotic use in early life.

While antibiotics alter the composition of the intestinal microbiome (Bokulich et al., 2016), the extent that they deplete riboflavin-synthesizing commensals has not been established. To identify antibiotics that may inhibit MAIT cell development, we cultured murine intestinal commensals with each of the 32 most prescribed antibiotics in neonatal intensive care units (Hsieh et al., 2014). The introduction of solid food during infancy causes diversification of the intestinal microbiome, including the appearance of obligate anaerobes (Bokulich et al., 2016; Sanidad and Zeng, 2020), so we isolated the cecal and colonic contents of wild-type (WT) C57BL/6J mice during weaning. Microbial samples were cultured anaerobically for 24 h in chopped meat media supplemented with carbohydrates, which supports growth of a broad range of anaerobic bacteria, and the maximum mg/kg infant dose of each antibiotic as recommended by the U.S. Food and Drug Administration. Optical density measurements revealed that microbial growth was most strongly inhibited by β-lactam antibiotics, including ampicillin, piperacillin, cefazolin, amoxicillin, cefotaxime, and dicloxacillin (Fig. 1 A).

The ribD gene is necessary to generate the riboflavin derivative recognized by MAIT cells and is indicative of the other riboflavin synthesis enzymes due to their organization within the riboflavin operon (Constantinides et al., 2019; Corbett et al., 2014; Garcia-Angulo, 2017). To quantitate how antibiotics affect the riboflavin synthesis capacity of the intestinal microbiota, we designed degenerate primers that were complementary to highly conserved sequences within the ribD gene (Fig. S1 A). After verifying that the assay was specific to ribD (Fig. S1 B), we assessed the relative abundance of riboflavin synthesizers by quantitative polymerase chain reaction (qPCR). Although metronidazole and vancomycin had minimal effects on culture growth (Fig. 1 A), they reduced the relative abundance of ribD by 493- and 17-fold relative to the untreated control, respectively (Fig. 1 B), suggesting outgrowth of non-riboflavin synthesizers. 16S rRNA gene sequencing revealed that both antibiotics allowed the growth of bacteria within the Enterococcaceae and Lactobacillaceae families (Fig. 1 C), which exhibit low conservation of riboflavin synthesis among their species (Constantinides et al., 2019). PICRUSt2 prediction of microbial function confirmed that metronidazole and vancomycin decreased abundance of the riboflavin synthesis pathway (Fig. S1 D). Conversely, the relative abundance of ribD was unaffected or increased in cultures that were dominated by Bacteroidaceae (Fig. 1 C), indicating growth of riboflavin-synthesizing species from this family. We concluded that antibiotics could reduce the relative abundance of riboflavin synthesizers, either by selective inhibition (e.g., vancomycin and metronidazole) or broad suppression of intestinal commensals (e.g., ampicillin).

Since mice begin weaning by 2 wk of age, we hypothesized that the transition from nursing to solid food and the accompanying reduction of maternal immunoglobulins increases the abundance of riboflavin synthesizers within the intestinal microbiota. Longitudinal analysis of untreated mice revealed a predominance of Lactobacillaceae in the feces at 1 wk, which resulted in a low capacity to synthesize riboflavin (Fig. 2, A and B). The abundance of riboflavin synthesizers markedly increased by 2 wk due to the emergence of Bacteroidaceae and Tannerellaceae, which have a high prevalence of this biosynthetic pathway (Constantinides et al., 2019). However, the appearance of other bacterial families that have a low prevalence of the riboflavin biosynthesis pathway decreased the relative proportion of riboflavin synthesizers, resulting in a transient enrichment during weeks 2–4 (Fig. 2, A and B).

Oral β-lactam antibiotics are the first-line pediatric therapy for otitis media, pharyngitis, sinusitis, and urinary tract infections, while oral vancomycin and metronidazole are recommended for gastrointestinal infections (Centers for Disease Control and Prevention, 2017, The Royal Children’s Hospital Melbourne, 2020). Consequently, ampicillin, vancomycin, and metronidazole are the 1st, 5th, and 44th most administered drugs in neonatal intensive care units, respectively (Hsieh et al., 2014). Treatment courses are typically 10–14 days (Centers for Disease Control and Prevention, 2017, The Royal Children’s Hospital Melbourne, 2020), so we determined whether this duration was sufficient to impair MAIT cell development in vivo. Mice received ampicillin for 2-wk intervals followed by a fecal microbiota transplant (FMT) from untreated age-matched animals to limit the resulting microbial dysbiosis to the treatment period (Fig. 2 C). Because the frequency of MAIT cells is highest in murine skin and correlates with their abundance in other tissues (Constantinides et al., 2019), we quantified cutaneous MAIT cells in adult 8-wk-old animals by flow cytometry (Fig. 2 D and Fig. S2). Ampicillin treatment during weeks 2–4 decreased MAIT cells by two- to threefold, while treatment during weeks 1–3 and 3–5 had less severe effects (Fig. 2, E and F). This period of susceptibility corresponded to the reported predominance of immature CD44⁻ MAIT cells within the thymi of 2-wk-old mice, where they remain enriched through 4 wk of age (Koay et al., 2016). Analysis of thymi from 4-wk-old mice at the conclusion of 2 wk of antibiotic treatment revealed a significant reduction in the abundance of MAIT cells (Fig. 2, G and H). A decrease in the abundance of mature CD24⁻CD44⁺ MAIT cells and an increase in the frequency of immature CD24⁺CD44⁻ MAIT cells in ampicillin-treated mice relative to controls was observed (Fig. 2, I and J). This is similar to what has been reported in GF mice (Koay et al., 2016), indicating that antibiotic treatment at this age disrupts the developmental progression of MAIT cells.

Since a single course of ampicillin was sufficient to impair MAIT cell development, we assessed whether other T cell subsets were impacted by antibiotic use in early life. Mice received ampicillin from weeks 2–4 and were subsequently co-housed with untreated age-matched animals for a week to model cessation of antibiotics (Fig. 3 A). While early-life ampicillin treatment had modest effects on conventional CD4⁺ and CD8⁺ T cells and invariant natural killer T (iNKT) cells in the lungs, these cell types were not impacted in the skin. We did observe ∼1/3 reduction in γδ T cells and regulatory T (T_reg) cells within the skin and lungs (Fig. 3, B and C). Prior work has demonstrated that Vγ6⁺ Vδ1⁺ T cells are reduced immediately following antibiotic treatment (Paget et al., 2015), but the persistence of this decrease weeks after microbial reconstitution has not been described. T_reg cells accumulate in murine skin during infancy in response to microbially induced CCL20 and are reduced in GF animals (Scharschmidt et al., 2017). Since the accumulation of cutaneous T_reg cells peaks at day 13 (Scharschmidt et al., 2015), the reduction caused by 2–4-wk ampicillin treatment suggests that continuous microbial exposure is necessary to retain these cells within tissues. While early-life ampicillin treatment caused slight reductions of γδ T cells and T_reg cells, MAIT cells were reduced by threefold across both the lungs and the skin (Fig. 3, B and C).

Because our in vitro assay demonstrated that ampicillin broadly suppresses growth of intestinal commensals (Fig. 1 A), the inhibition of MAIT cell development may have been due to the combined loss of microbial ligands for pattern recognition receptors and derivatives of riboflavin synthesis. To establish whether antibiotics that selectively deplete riboflavin synthesizers impair MAIT cell development, we administered vancomycin or metronidazole during the 2–4-wk period. Feces collected at the conclusion of the antibiotic treatment revealed that vancomycin and metronidazole depleted the riboflavin synthesis capacity of the microbiota in vivo without inhibiting microbial growth (Fig. 3, D and E), as observed in vitro. Prior work demonstrated that intestinal commensals differentially transcribe riboflavin synthesis genes in vivo and the riboflavin operon is post-transcriptionally regulated by binding of the riboflavin product flavin mononucleotide (El Morr et al., 2024; Winkler et al., 2002). Therefore, we assessed the abundance of MAIT cell agonists during antibiotic treatment using 8D12 MAIT hybridoma cells that release IL-2 in response to 5-OP-RU activation in a dose-dependent manner (Fig. 3 F). While fecal homogenates from untreated controls elicited IL-2 production, administration of ampicillin, vancomycin, or metronidazole decreased activation of the MAIT hybridoma as much as the addition of an anti-MR1 blocking antibody (Fig. 3 G), indicating a loss of stimulatory metabolites. Vancomycin and metronidazole permitted growth of bacterial families with a low conservation of riboflavin synthesis, including Bifidobacteriaceae, Enterococcaceae, and Lactobacillaceae (Fig. 3 H). 2–4-wk treatment with either vancomycin or metronidazole was sufficient to inhibit MAIT cell accumulation within the lungs and, to a lesser extent, the skin (Fig. 3, I–L). Together, these results indicate that administration of antibiotics in early life can impair MAIT cell development by depleting stimulatory metabolites.

Antibiotic use during the first 6 mo of life increases the risk of recurrent respiratory tract infections in children (Zhou et al., 2021). Since MAIT cells produce IFN-γ and/or IL-17A in response to a wide array of respiratory pathogens (Georgel et al., 2011; Gold et al., 2010; Hartmann et al., 2018; Hinks et al., 2016; Jahreis et al., 2018; Le Bourhis et al., 2010; Meierovics et al., 2013; Wang et al., 2018; Zhao et al., 2021), we sought to determine if early-life antibiotic treatment could subsequently impair immunity to bacterial pneumonia.

The gram-negative, facultative, intracellular bacterium F. tularensis causes hundreds of infections each year, resulting in pneumonic tularemia when inhaled (U.S. Department of Health and Human Services, 2022). Analysis of TCRβ-deficient mice indicated that αβ T cells are essential for survival in F. tularensis live vaccine strain (LVS), while γδ T cells are dispensable since antibody depletion or genetic ablation did not alter immunity (Yee et al., 1996). Transfer of MAIT cells into immunodeficient Rag2^−/−Il2rg^−/− mice improved survival in F. tularensis LVS and was dependent on their production of TNF, IFN-γ, and GM-CSF (Zhao et al., 2021). Although some F. tularensis strains avoid triggering a MAIT cell response by not synthesizing riboflavin (Shibata et al., 2022), F. tularensis LVS encodes the riboflavin synthesis pathway. Following retropharyngeal inoculation with the F. tularensis LVS, the abundance of MAIT cells within murine lungs increased more than 75-fold, while other T cell subsets did not proliferate extensively (Fig. 4 A), which is consistent with prior studies (Meierovics et al., 2013; Zhao et al., 2021). To establish the susceptibility of our mice, we inoculated WT animals with varying doses of F. tularensis LVS and monitored disease severity by weight loss and survival (Fig. S4). Mice that received 1 × 10³ or more colony-forming units (CFUs) succumbed to the infection, while animals that received 6 × 10² CFUs or less survived indicating that the median lethal dose (LD₅₀) is 6–10 × 10² CFUs for our inoculation route (Fig. S4 B).

To establish whether early-life antibiotic use impairs pulmonary immunity, we inoculated ampicillin and vehicle control-treated mice with a sublethal 4 × 10² CFU dose of F. tularensis LVS (Fig. 4 B). Mice that received antibiotics in early life lost twice as much body weight and failed to recover 14 days after infection (Fig. 4 C). Ampicillin-treated mice had half the number and frequency of MAIT cells compared to control animals (Fig. 4, D and E), indicating that the antibiotic-mediated impairment of MAIT cells persisted throughout the infection. At 8 wk old, mice that received early-life ampicillin had proportions of T-bet⁺ MAIT1 and RORγt⁺ MAIT17 cells similar to untreated controls and their production of granzyme B, IFN-γ, and IL-17A were comparable (Fig. S3), indicating that the decreased immunity was due to fewer MAIT cells rather than impaired effector function. Following an LD₅₀ dose of 8 × 10² CFUs of F. tularensis LVS (Fig. 4 F), WT mice that received early-life ampicillin were significantly more susceptible than vehicle control-treated animals, with a survival rate that mirrored MAIT cell–deficient Mr1^−/− mice (Fig. 4 G). Ampicillin-treated Mr1^−/− mice succumbed to the infection similarly to the control Mr1^−/− animals, demonstrating that antibiotic-mediated impairment of immunity is mediated by the loss of MAIT cells. Transfer of MAIT cells following ampicillin treatment was sufficient to minimize weight loss during a sublethal infection with F. tularensis LVS (Fig. 4, H–I), indicating that the persistent effect of early-life antibiotic use was reduced MAIT cell abundance.

Due to the abundance of Bacteroidaceae species during the 2–4-wk period (Fig. 2 A) and their susceptibility to vancomycin and metronidazole (Fig. 1 C and Fig. S1 D), we hypothesized that a riboflavin-synthesizing member of this family may be sufficient to restore MAIT cell development. We identified the human fecal isolate Bacteroides thetaiotaomicron strain VPI-5482, which retains the riboflavin synthesis pathway and the PER-1 β-lactamase that hydrolyzes penicillins and cephalosporins (Paterson and Bonomo, 2005), enabling coadministration during ampicillin treatment. As a non-riboflavin synthesizing control, we selected an ampicillin-resistant Enterococcus faecium isolated from human feces. In vitro testing confirmed that supernatants from B. thetaiotaomicron but not E. faecium–stimulated MAIT hybridoma cells in an MR1-dependent manner (Fig. 5 A). Mice received daily oral gavages of either B. thetaiotaomicron or E. faecium during the ampicillin treatment from 2 to 4 wk of age (Fig. 5 B). Dosing with B. thetaiotaomicron, but not E. faecium, increased the abundance of MAIT cell ligands in fecal and cecal samples during the treatment window (Fig. 5, C and D). B. thetaiotaomicron and E. faecium were detected in their respective samples at 4 wk of age, but cohousing for 1 wk with untreated, age-matched mice normalized the microbiota across treatment groups (Fig. 5 E). Concomitant administration of B. thetaiotaomicron, but not E. faecium, was sufficient to increase the number and frequency of pulmonary MAIT cells in adult mice (Fig. 5, F and G). This probiotic treatment had a systemic effect as a similar increase in the number and frequency of MAIT cells was also observed in the skin of adult mice (Fig. 5, H and I). To determine whether administration of a riboflavin-synthesizing probiotic is sufficient to restore immunity, mice were inoculated with an LD₅₀ dose of 8 × 10² CFUs of F. tularensis LVS at 8 wk of age (Fig. 5 J). Animals that received B. thetaiotaomicron were less susceptible than mice that received E. faecium or no probiotic during ampicillin treatment, with survival rates comparable to control mice that were not administered antibiotics (Fig. 5 K). Thus, early-life antibiotic use subsequently impairs MAIT cell–mediated immunity by reducing riboflavin synthesis capacity and availability of MAIT cell ligands, which can be ameliorated by administering a riboflavin-synthesizing probiotic.

Our findings reveal a transient enrichment of riboflavin-synthesizing intestinal commensals in early life that is necessary for the development of peripheral MAIT cells. Predictive functional analysis of human stool determined that the riboflavin biosynthesis pathway is more abundant in children than adults (Radjabzadeh et al., 2020), so there may be a corresponding period for microbial induction of MAIT cells in humans. Within the first year of life, the frequency of MAIT cells in human blood increases ∼10-fold and these cells acquire their capacity to produce cytokines (Swarbrick et al., 2020). During this period, the infant intestinal microbiome contains an abundance of bacteria from the families Enterobacteriaceae, Bacteroidaceae, and Clostridiaceae (Bokulich et al., 2016), which have a high prevalence of riboflavin synthesis (Constantinides et al., 2019). However, their collective abundance decreases following the first year (Bokulich et al., 2016), suggesting a transient enrichment in riboflavin synthesizers during infancy. Although it remains unclear whether human MAIT cell development depends on this period, children between the ages of 1.5–13 years old, with a mean age of 6.5 years, that receive hematopoietic stem cell transplants regain only 10% of their pre-transplant MAIT cell frequencies a year later, even though conventional T cells fully recover (Ben Youssef et al., 2018).

Our in vitro screen identified multiple antibiotics that depleted riboflavin synthesizers within the murine intestinal microbiota, including ampicillin, vancomycin, and metronidazole. Both ampicillin and vancomycin decreased microbial diversity within the human infant microbiota and early-life antibiotic treatment had persistent effects on the microbial composition (Gasparrini et al., 2019). We found that mice treated with ampicillin, vancomycin, or metronidazole from 2 to 4 wk of age developed fewer MAIT cells in their tissues and this impairment persisted into adulthood. The prevalence of immature CD161⁻ MAIT cells within the blood of human neonates and infants indicates that extrathymic signals are necessary for their development in early life (Koay et al., 2016; Swarbrick et al., 2020). Given that the microbiota promotes maturation of murine MAIT cells (Constantinides et al., 2019; Koay et al., 2016; Legoux et al., 2019), the development of human MAIT cells is likely vulnerable to antibiotic use during infancy. Disparate antibiotic use in early life may explain why the range of MAIT cell frequencies in pediatric and adult peripheral blood is broader than the abundance in neonatal cord blood (Swarbrick et al., 2020).

Following early-life antibiotic treatment, mice were more susceptible to bacterial pneumonia later in life. In addition to increasing the risk of recurrent respiratory tract infections in children (Zhou et al., 2021), antibiotic use during infancy more than doubles the odds of developing pediatric asthma (Donovan et al., 2020; Mitre et al., 2018; Ni et al., 2019; Patrick et al., 2020), with the risk increasing ∼20% with each course of antibiotics (Aversa et al., 2021; Donovan et al., 2020). Since MAIT cells have been shown to restrict allergic airway inflammation in response to Alternaria alternata and house dust mite extracts (Cait et al., 2024; Ye et al., 2020), additional research is warranted to assess whether the increased prevalence of asthma is due to the depletion of MAIT cells. Additionally, the decrease in cutaneous MAIT cells following early-life antibiotics suggest that immunity within the skin may also be impaired. Antibiotic use during infancy has been shown to increase the risk of atopic dermatitis in children (Aversa et al., 2021), although it remains unclear whether MAIT cells are beneficial or detrimental during this inflammatory skin disease (Imahashi et al., 2023; Naidoo et al., 2021). Since administration of ampicillin hindered thymic development of MAIT cells, antibiotic use in early life may also affect MAIT cell responses in other tissues. Recent work has demonstrated that MAIT cells undergo maturation in the intestines (Bugaut et al., 2024), so fewer immature cells from the thymus may impair intestinal immunity. Additionally, MAIT cell maturation within the intestines likely depends on microbial antigens since the ratio of type-1/17 MAIT cells is altered in the absence of the MyD88 protein that facilitates Toll-like receptor signaling.

To maintain the riboflavin synthesis capacity of the microbiome during antibiotic treatment, we administered a human fecal isolate of B. thetaiotaomicron, which was sufficient to increase MAIT cell abundance and restore respiratory immunity. Due to the abundance of Bacteroidaceae within the intestinal microbiota of infants (Bokulich et al., 2016), members of this bacterial family may be essential for MAIT cell development. However, we demonstrated that multiple antibiotics depleted Bacteroidaceae from the murine microbiota. Human newborns that received the β-lactams amoxicillin and cefotaxime had less Bacteroides and the abundance of multiple taxa within this genus was decreased over a year following antibiotic treatment (Reyman et al., 2022). Identifying which riboflavin synthesizers are the primary determinants of MAIT cell abundance and developing therapeutics to restore them is of interest for human health.

Our work identifies weaning as a critical period for the development of MAIT cells, which depend on the transient enrichment of riboflavin-synthesizing commensals in early life. Antibiotics that are readily prescribed to infants can deplete riboflavin synthesizers and administration during this period abrogates MAIT cell development, rendering adult mice more susceptible to infection. Concomitant administration of a riboflavin-synthesizing commensal during antibiotic treatment was sufficient to restore murine MAIT cell development and immunity, indicating that probiotic supplementation may ameliorate the immunologic consequences of early-life antibiotic use.

The antibiotic screen was adapted from previous work to analyze early life microbiota (Chen et al., 2020). Whole digestive tract from proximal small intestine to rectum were excised from 2 each of two-and-a-half and three-week-old mice and quickly transferred into an anaerobic chamber (Coy Lab Products). Cecal and colon contents were harvested, passed through a 70-μm filter using 5 ml of pre-reduced phosphate-buffered saline (PBS) with 0.1% cysteine (PBSc), diluted 12-fold with Chopped Meat Carbohydrate Broth (Anaerobe Systems), and adjusted to 0.1 OD after subtracting a media blank. 0.8 ml aliquots were distributed into deep-well 96-well plates in duplicate with each antibiotic. Antibiotics were acquired from MedChemExpress in 10 mM DMSO or water. Antibiotic concentrations were derived from maximum neonatal or pediatric mg/kg dose. This dose was adapted to a molar assay concentration by assuming homogenous diffusion and 70% water by mass. Assay concentration was limited to 150 μM to limit DMSO concentration to 1.5%. DMSO concentration was normalized to all samples and controls in mother dilutions which were prepared with 25× concentration for transfer to the 96-well deep-well assay plate. Samples were incubated for 24 h at 37°C. 100 μl of each sample was transferred to an optically clear 96-well plate for optical density measurement. DNA was purified from the remainder using DNeasy Ultraclean Microbial Kit (Qiagen) for downstream analysis.

qPCR analysis of riboflavin synthesis capacity was performed using PowerTrack SYBR Green (A46109; Applied Biosystems) with a Bio-Rad C1000 Touch Thermal Cycler with CFX96 Real-Time System. RibD primers were designed using a multiple alignment of reference genomes from 4 riboflavin-synthesizing bacterial species previously isolated from our facility: Enterocloster bolteae, B. thetaiotaomicron, Muribaculum intestinale, and Duncaniella spp; RibD-F (5′-CCMAAYCCBATGGYVGGDG-3′; RibD-R: 5′-AGCADGGYTCSAGRSTBACRTA-3′; B=C,G,T, D = A,G,T, M = A,C, R = A,G, S=C,G, V = A,C,G, Y=C,T). V3–V4 hypervariable region of the bacterial 16S rRNA gene was amplified with 338F and 806R primers (338F: 5′-ACT​CCT​ACG​GGA​GGC​AGC​AG-3′; 806R: 5′-GGACTACHVGGGTWTCTAAT-3′). ribD relative abundance was quantified using the Pfaffl method relative to 16Sv3v4, relative to no-antibiotic control samples (Pfaffl, 2001).

Amplification of the V3–V4 hypervariable region of the bacterial 16S rRNA gene was performed using 338F and 806R primers similar to those previously described (Chen et al., 2020), which contained 12-base sample barcodes (N), unique molecular identifiers (X) to correct for PCR duplication artifacts, and Illumina sequencing components (5′-AAT​GAT​ACG​GCG​ACC​ACC​GAG​ATC​TAC​ACT​CTT​TCC​CTA​CAC​GAC​GCT​CTT​CCG​ATC​T NNN​NNN​NNN​NNN​ACT​CCT​ACG​GGA​GGC​AGC​AG-3′) and reverse primer (5′-CAAGCAGAAG ACGGCATACGAGATNNN​NNN​NNNNNNGTGACTGGAGTTCAGACGTGTGCTCTTCCGATCTXXXXXXXXXXGGACTACHVGGGTWTCTAAT-3′). Amplification was performed using 50 ng DNA, 10 μl 2× KAPA HiFi HotStart ReadyMix (Kapa Biosystems), and 0.35 μM of each primer. Samples with low DNA concentration were amplified using a maximum volume. PCR cycling was performed in two stages: 95°C for 3 min, 5 cycles of 98°C for 20 s, 57°C for 15 s to optimize primer binding to genomic amplicon region, 72°C for 45 s, and a final 1 min elongation at 72°C. This was followed by 95°C for 3 min, 20 cycles of 98°C for 20 s, 67°C for 15 s to optimize primer binding to PCR amplicon, 72°C for 45 s, and a final 1 min elongation at 72°C. Samples with abundant template concentration were pooled and purified using Agencourt AMPure XP beads (Beckman Coulter). Samples with low template concentration were purified individually and quantified using Quant-iT dsDNA Assay Kit (Thermo Fisher Scientific) prior to pooling. Sequencing was performed using Illumina MiSeq. Sequencing reads were demultiplexed using second barcode indices and processed to remove PCR duplication artifacts using unique molecular identifier sequences (Chen et al., 2020). Sequences were analyzed for taxonomic identification using the nf-core ampliseq workflow with FastQC, Cutadapt, MultiQC, QIIME2, and DADA2.

Fecal DNA was purified using the QIAmp PowerFecal Pro DNA Kit (Qiagen), amplified and analyzed as previously described. Sequencing was performed using Illumina Nextseq.

16S rRNA gene amplicon sequence variants (ASVs) and relative abundances per sample were used for functional prediction of microbial communities with Phylogenetic Investigation of Communities by Reconstruction of Unobserved States, v2.6.2 (PICRUSt2) (Douglas et al., 2020; Wright and Langille, 2025) and its dependencies (Barbera et al., 2019; Czech et al., 2020; Louca and Doebeli, 2018; Ye and Doak, 2009). The analysis followed the standard PICRUSt2 pipeline with default parameters, except for the modification of the –min_align parameter from the default 0.8 to 0.7 to prevent exclusion of a large fraction of sequences. Setting the parameter to 0.7 resulted in <0.03% of ASVs being excluded from the analysis.

WT C57BL/6J mice were acquired from the rodent breeding colony at Scripps Research, which is supplied by The Jackson Laboratory. Mr1^tm1Gfn (Mr1^−/−) mice were generated by Dr. Susan Gilfillan (Washington University School of Medicine, St. Louis, MO, USA) and was backcrossed to C57BL/6 for >8 generations (Treiner et al., 2003). Mice were bred and cared for in a facility accredited by the American Association for the Accreditation of Laboratory Animal Care at the Scripps Research Institute. All experiments were conducted at Scripps Research in accordance with an Animal Safety Protocol (21-0003) approved by the Scripps Research Institutional Animal Care and Use Committee. All experiments were age and sex matched. Mice used as controls included untreated non-littermates bred from the same colony as the experimental groups.

In vivo antibiotic doses were adapted from human neonatal and pediatric doses at or below maximum daily mg/kg dose: ampicillin 100 mg/kg/day, vancomycin 40 mg/kg/day, and metronidazole 30 mg/kg/day. Mice younger than 3 wk old were weighed daily and treated by oral gavage with antibiotic dissolved in PBS at concentrations necessary to deliver a volume of 10 μl/g. After weaning, mice were administered antibiotics in drinking water supplemented with 9.6 mg/L equal sweetener (3.2 mg/kg/day). All control mice used for survival experiments were treated with vehicle controls.

To rapidly reconstitute microbiota following antibiotic treatment for evaluation of the MAIT cell developmental window, each group received a FMT gavage from age-matched donor mice. Cecum and colon microbiota were collected anaerobically as with preparation for the in vitro assay and 50 μl/mouse of the filtered PBSc slurry was used for FMT. Mice treated in all other experiments were co-housed with age-matched mice for 1 wk following antibiotic treatment for microbiome reconstitution.

For the isolation of skin cells, ears were excised, split into ventral and dorsal halves, and placed dermal side down for 1 h 45 min at 37°C with 5% CO₂ in 500 μl RPMI 1640 with 20 mM HEPES (Corning), 50 mM β-mercaptoethanol (Gibco), 1 mM sodium pyruvate (Corning), 2 mM glutaGRO (Corning), 1 mM nonessential amino acids (Corning), 100 U/ml penicillin (Corning), and 100 mg/ml streptomycin (Corning) (Supplemented RPMI), 0.5 mg/ml DNase I (Sigma-Aldrich), and 0.25 mg/ml of Liberase TL (Roche) (Digest Media). After incubation, 500 μl Supplemented RPMI with 0.5 mg/ml DNase I, 3% fetal bovine serum (FBS) and EDTA was added to stop digestion. Digested tissue was then passed through a 70 μm filter in Supplemented RPMI with 0.5 mg/ml DNase I and 3% FBS (3% DNase), centrifuged, and resuspended in 500 μl Supplemented RPMI with 10% FBS (10% RPMI).

For the isolation of lung cells, lungs were excised, diced, and incubated in Digest Media for 45 min in a 37°C water bath with vortexing every 15 min. The digested lungs were passed through a 70-μm filter with 3% DNase, centrifuged, resuspended in 5 ml of 37% Percoll (Sigma-Aldrich), centrifuged, resuspended in 2 ml RBC Lysis Buffer (420302; BioLegend), incubated for 3 min at room temperature, diluted with 3 ml 3% DNase, centrifuged, and resuspended in 500 μl 10% RPMI.

For the isolation of thymocytes, thymi were excised and passed through a 70-μm filter with 3% DNase. MAIT cells were magnetically enriched using PE-conjugated MR1-5-OP-RU tetramer and MojoSort Mouse anti-PE nanobeads, according to the manufacturer’s protocol (BioLegend).

Fluorophore-conjugated antibodies were purchased from BD Biosciences, BioLegend, or Invitrogen. 5-(2-oxopropylideneamino)-6-D-ribitylaminouracil (5-OP-RU)–loaded mMR1 and PBS57-loaded mCD1d tetramers were acquired from the National Institutes of Health Tetramer Core Facility (Corbett et al., 2014). For extracellular staining, cells were incubated in complete RPMI at room temperature for 1 h. For in vitro restimulation, cells were incubated with phorbol 12-myristate 13-acetate (PMA) and ionomycin for 2.5 h at 37°C with 5% CO₂. Cells were fixed for 1 h and then stained intracellularly for transcription factors for 1 h at 4°C using eBiosciences FOXP3/Transcription Factor Staining Buffer Set (00-5523-00; Invitrogen). All staining mixes included TruStain FcX rat anti-mouse CD16/32 (101320 BioLegend). Dead cells were removed by staining with LIVE/DEAD Fixable Blue Dead Cell Stain Kit (Invitrogen Life Technologies) or Zombie NIR Fixable Viability Kit (Biolegend). Cytometric analysis was performed using a Cytek Aurora Spectral Flow Cytometer (Aurora Biosciences) with SpectroFlo software (Cytek). Data were processed using FlowJo (BD). MAIT cells were gated mMR1 tetramer⁺ TCRβ⁺ TCRγδ⁻ CD90.2⁺ CD45⁺ LIVE/DEAD⁻; iNKT cells were gated as mCD1d tetramer⁺ mMR1 tetramer⁻ TCRβ⁺ TCRγδ⁻ CD90.2⁺ CD45⁺ LIVE/DEAD⁻; CD4 T cells were gated as FoxP3⁻ CD4⁺ CD8β⁻ mCD1d tetramer⁻ mMR1 tetramer⁻ TCRβ⁺ TCRγδ⁻ CD90.2⁺ CD45⁺ LIVE/DEAD⁻; regulatory T cells (Treg) were gated as FoxP3⁺ CD4⁺ CD8β⁻ mCD1d tetramer⁻ mMR1 tetramer⁻ TCRβ⁺ TCRγδ⁻ CD90.2⁺ CD45⁺ LIVE/DEAD⁻; CD8⁺ T cells were gated as CD8β⁺ CD4⁻ mCD1d tetramer⁻ mMR1 tetramer⁻ TCRβ⁺ TCRγδ⁻ CD90.2⁺ CD45⁺ LIVE/DEAD⁻; and γδ T cells as TCRγδ⁺ TCRβ⁻ CD90.2⁺ CD45⁺ LIVE/DEAD⁻.

2 × 10⁴ cells of WT3-MR1 fibroblasts cultured in RPMI without folic acid and supplemented with 10% (vol/vol) fetal bovine serum and 2 mM glutaGRO (Corning) (10% RPMI without folate) were added to each well of a treated plate and incubated for 4 h at 37°C with 5% CO₂. 5-OP-RU was generated by incubating 100 mM 5-amino-6-D-ribitylaminouracil (5-A-RU) with methylglyoxal for 45 min at room temperature in the dark prior to dilution in 10% RPMI without folate for use in this assay. Fecal samples were prepared by diluting weighed pellets in 10% RPMI without folate to generate a 10 mg/ml solution. Sterilized glass beads were vortexed for 1 min with the diluted fecal samples and supernatants were filtered through a 0.22-μM syringe filter. Bacterial supernatants were prepared by pelleting the bacteria at 4,000 × g for 5 min and filtering the supernatant through a 0.22-μM syringe filter. Bacterial supernatants were further diluted 10-fold in 10% RPMI without folate. 20 μl of the prepared fecal or bacterial supernatants were added to the cultured WT3-MR1 cells for a final 100-fold dilution in the assay. Supernatants were incubated with the WT3-MR1 cells for 2.5 h at 37°C with 5% CO₂ before being rinsed off with 10% RPMI without folate. 10 μg/ml of anti-MR1 (clone 26.5) antibody diluted in 10% RPMI without folate was added to any wells that required blocking of the MR1 receptor and incubated for 5 min at room temperature. 100 μl of 10% RPMI without folate was added to any wells that did not require blocking. 1 × 10⁵ 8D12 MAIT hybridoma cells resuspended in 10% RPMI without folate were added to each well of the assay and incubated overnight at 37°C with 5% CO₂. Supernatants from cocultures were collected after centrifugation at 526 × g for 5 min to remove cells. IL-2 was detected by enzyme-linked immunosorbent assay per the manufacturer protocol (Thermo Fisher Scientific).

F. tularensis subsp. holarctica, Centers for Disease Control and Prevention live vaccine strain (NR-646; BEI Resources) was streaked on cysteine heart agar. Single colonies were selected for growth in tryptic soy broth with 0.1 g/L cysteine at 37°C. This was diluted to OD₆₀₀ 0.01 and grown to generate a CFU-OD curve. For in vivo experiments, the same inoculation and growth protocol was used, and the culture was grown until OD reached 0.6–0.8. CFU/ml was calculated and used for infection inoculates, which were diluted in PBS and confirmed by plating of inoculate prior to infection. Age- and sex-matched mice were weighed, anaesthetized with isoflurane, suspended, and inoculated retropharyngeally via reflexive aspiration. Body weight was assessed daily and mice were euthanized if weight loss exceeded 20–25%.

B. thetaiotaomicron, VPI-5482 (29148; ATCC) and E. faecium (BAA-2320; ATCC) were grown anaerobically at 37°C in brain heart infusion (BHI) media supplemented with hemin and vitamin K (BHI-S) overnight to stationary phase and plated to determine stationary CFU/ml. Broth culture was concentrated by centrifugation and resuspended in PBS for daily administration via oral gavage of 1 × 10⁸ CFU during antibiotic treatment.

MAIT cells were isolated and adoptively transferred as previously described (Chen et al., 2019). Briefly, MAIT cells were expanded in donor mice by infection with a sublethal dose of F. tularensis LVS. After 14 days, single cell suspensions were prepared from the lungs of donor mice, as described above, but excluding lysis of red blood cells. MAIT cells were magnetically enriched using PE-conjugated MR1-5-OP-RU tetramer and MojoSort Mouse anti-PE nanobeads, according to the manufacturer’s protocol (BioLegend). Approximately 2.5 × 10⁴ MAIT cells in 100 μl of sterile PBS were transferred into each mouse at 7 wk of age by retroorbital injection. Recipient mice were rested for 1 wk before subsequent infection.

Statistical analyses were performed using Prism 10 software (GraphPad). Pairwise comparisons with 2 groups were performed using two-tailed unpaired t tests; analyses with >2 groups were performed using ordinary one-way ANOVA with Šidák’s multiple comparison correction or multiple unpaired t tests with Holm–Šidák’s multiple comparison correction where stated; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. “ns” = not significant. Survival statistics were performed using Mantel–Cox test.

Fig. S1 shows the design and use of degenerate ribD primers to quantify the abundance of this gene, as well as supplemental analyses of the predicted abundance of the riboflavin synthesis pathway in vitro samples using PiCRUST2. Fig. S2 shows the gating strategy used to identify various immune cell populations. Fig. S3 shows supplemental data regarding the functional analysis of MAIT cells, including frequency of T-bet and RORγt expression and cytokine production. Fig. S4 shows supplemental data regarding infections with F. tularensis.

All sequencing data are available under BioProject accession PRJNA1271113. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the corresponding author upon request.

The authors acknowledge the Scripps Research Department of Animal Resources, Flow Cytometry Core, and Genomics Core and thank Dr. Olivier Lantz (Institut Curie, Paris, France) for the WT3-MR1 cells and 8D12 MAIT hybridoma cells, Dr. Jeffrey Aubé (University of North Carolina, Chapel Hill, NC, USA) for providing 5-A-RU, and the NIH Tetramer Core Facility for mCD1d and mMR1 tetramers.

This work was supported by the National Institutes of Health (K22AI146217, R21AI171697, and R35GM151347 to Michael G. Constantinides), the National Science Foundation (Graduate Research Fellowship to Gabrielle R. LeBlanc), the Natural Sciences and Engineering Research Council of Canada (Doctoral Postgraduate Scholarship to Dominic Haas), and Scripps Research.

Author contributions: Gabrielle R. LeBlanc: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, validation, visualization, and writing—original draft, review, and editing. Adam L. Sobel: conceptualization, data curation, formal analysis, investigation, methodology, project administration, software, validation, visualization, and writing—original draft. Jonathan Melamed: funding acquisition, investigation, and methodology. Dominic Haas: funding acquisition and investigation. Eduard Ansaldo: formal analysis. Aiko M. Cirone: investigation. Elizabeth Murguia: investigation. Michael G. Constantinides: conceptualization, formal analysis, funding acquisition, methodology, project administration, supervision, validation, visualization, and writing—original draft, review, and editing.