Maternal Antibiotic Treatment Impacts Development of the Neonatal Intestinal Microbiome and Antiviral Immunity

Epidemiologic evidence suggests that frequent courses of antibiotics alter the gastrointestinal tract (GIT) microbiota during infancy and childhood and may be a significant risk factor in the development of future allergic and autoimmune diseases (1). Children under 10 years of age receive >40 million courses of antibiotics per year and several courses under the age of 2 years (2, 3). Nearly 40% of pregnant mothers and infants are treated with antibiotics during the perinatal, immediate postnatal, and neonatal periods (4, 5).

Infants emerge from a relatively sterile in utero environment (6) and are colonized by maternally derived vaginal and GIT microbiota at birth (7, 8). In the first year of life, the GIT microbiota is characterized by low complexity and is marked by instability likely related to developmental status and dietary influences (9, 10). Antibiotic exposure, even short courses, alters the normal colonization patterns and assembly of GIT microbiota during infancy in both density and complexity (11, 12). Antibiotic treatment of pregnant mice confers an altered microbiota to their offspring, similar to what is observed in humans (13–15). Progressive colonization of the intestinal tract during infancy stimulates the development of mucosal and systemic immune tissues and immune cell populations (16, 17). Simultaneously, infants become exposed to a vast array of environmental Ags, pathogens, and most of the vaccines that they will receive in their lifetime. It is unknown how altering the GIT microbiota early in life affects immune function at a stage when immune responses against pathogens are known to be suboptimal.

In adult mice, transient or permanent alterations in the density and diversity of the GIT microbiota affect the maintenance and function of immune cell populations required for mucosal and systemic immune homeostasis (18–21). T cell immunity is critical for control and clearance of a viral infection, and antibiotic treatment of adult mice alters antiviral innate and adaptive immunity (22, 23). The differentiation and functional capacity of infant T cells differs from adult T cells (24). Yet few studies have specifically examined how alterations in the density and complexity of the GIT microbiota (dysbiosis) shapes systemic adaptive immune function or antiviral immunity during infancy.

We showed previously that maternal antibiotic treatment/treated (MAT) was associated with reduced Ab production following immunization of infant mice <15 days of age (25). Our goals in this study were to characterize GIT dysbiosis in a model of maternal and perinatal antibiotic treatment and examine its impact on infant antiviral immunity during systemic vaccinia virus infection. We further characterize the infant microbiota and show that MAT results in profound alterations in the GIT microbiota of mothers and their infants. Surprisingly, MAT infants did not inherit an intestinal microbiota reminiscent of their mothers but were almost entirely colonized by the commensal Enterococcus faecalis. MAT infant mice could not survive infection with vaccinia, in contrast to control (CTRL) infant mice. We noted alterations in innate immune cell populations early during infection and at its peak. However, the most notable impact was on CD8⁺ T cell function. MAT infants demonstrated diminished IFN-γ CD8⁺ T cell effector function following infection. CD8⁺ T cells isolated from uninfected MAT infant mice also could not sustain IFN-γ and TNF-α cytokine production upon TCR engagement in vitro. We conclude that MAT has a deleterious impact on GIT bacterial colonization that results in reduced antiviral immunity during a vulnerable period of early life in infant mice.

Pairs of age-matched adult conventional C57BL/6J mice were obtained from The Jackson Laboratory (Bar Harbor, ME), bred, infected, or maintained under specific pathogen–free (SPF), BSL-2, and specific pathogen Helicobacter–free (SPHF) conditions in the Columbia University Medical Center and the Columbia Center for Infection and Immunity mouse facilities. Mice transferred between facilities were allowed to acclimate for ≥1 wk prior to initiation of experiments or collection of stool samples. All animal procedures were conducted according to the National Institutes of Health guidelines for the care and use of laboratory animals and were approved by the Columbia University Medical Center Institutional Animal Care and Use Committee.

Ampicillin (AAP Pharmaceuticals), streptomycin (X-gen Pharmaceuticals), and clindamycin (Cleocin Pediatric; Pharmacia and Upjohn) were mixed into sterile drinking water at a final concentration of 1 mg/ml. Mice were allowed to drink the water ad libitum for the duration of the experiments. Antibiotic-containing water was given to pregnant mothers ≥3–5 d prior to the expected birth of a litter. Water was replaced and/or refreshed every 3 d for the duration of the experiment. At sacrifice, stool or cecal contents were collected from adult or infant mice, respectively, weighed, and snap-frozen for storage at −80°C prior to DNA extraction.

DNA was extracted from snap-frozen stool (adult mice) or cecal stool (infant mice) using the QIAamp Stool DNA mini-kit (QIAGEN, Valencia CA), according to the manufacturer’s instructions. Ceca were homogenized in Buffer ASL (QIAGEN), and DNA was isolated as above. The quantity (16S rRNA gene copies/mg stool or cecal content) of total and specific intestinal bacteria (Bacteroidetes, Firmicutes, Lactobacillus) was measured by quantitative real-time PCR (qPCR) using universal and group-specific 16S rRNA gene primers (25) and the QuantiTect SYBR Green PCR kit (QIAGEN). qPCR was performed on an ABI ViiA 7 real-time PCR system (Applied Biosystems, Foster City, CA). Bacterial DNA was quantified using serial dilutions of plasmid standards curves constructed with reference bacteria specific for each bacterial group analyzed.

For 16S rRNA gene pyrosequencing, DNA was extracted using a modified protocol of the QIAmp DNA Stool Mini Kit (QIAGEN). For each set of DNA extractions, two empty 2-ml tubes were taken through every step of the extraction to serve as negative CTRLs for downstream 16S rDNA PCR. Prior to extraction, all tubes, columns, and 0.1- and 0.5-mm glass beads (MO BIO Laboratories) were UV irradiated twice. All kit extraction reagents were aliquoted and UV irradiated as above. For each sample, stool pellets or cecal contents were resuspended in Buffer ASL (QIAGEN) and transferred to UV-irradiated 2-ml Safe-Lock tubes (Eppendorf) containing UV-irradiated glass beads. To optimally lyse Gram-positive bacteria, samples were further disrupted by bead beating in a TissueLyser (QIAGEN) for 5 min at 30 Hz and incubated for 5 min at 95°C. The remaining steps for extraction followed the manufacturer’s protocol and were performed in a UV hood. DNA concentration and purity were determined using a NanoDrop ND-100 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and stored at −80°C.

Amplification of the V1–V3 region of bacterial 16S rDNA for pyrosequencing was performed on DNA from stool and/or cecal content from CTRL and MAT mothers and infant mice using 16S rDNA composite primers consisting of FLX Titanium adapters, a sample barcode, and bacterial 16S rDNA-specific V1 (27F) and V3 (534R) primer sequences, as previously described (26). Plate caps, 96-well PCR plates, microcentrifuge tubes, and UltraClean water (MO BIO Laboratories) were UV irradiated in a Spectrolinker XL-1500 UV cross-linker (3000 × 100 μj/cm2) prior to PCR setup in a UV hood. Each 20-μl PCR reaction consisted of 1× AccuPrime Buffer II, 0.75 U AccuPrime Taq DNA Polymerase High Fidelity (Life Technologies), 2.5 U Sau3AI restriction enzyme, 200 nM each primer, and 100 ng sample DNA. Before addition of primers, the PCR master mix was UV irradiated to reduce downstream amplification of any potential contaminant DNA in reagents. The master mix was aliquoted into 96-well plates, and barcoded primers were added to each reaction. The 96-well plate was incubated at 37°C for 30 min to facilitate Sau3AI digestion of any remaining dsDNA contaminants. After digestion, the plate was incubated on ice for 5 min, and sample DNA was added to each reaction and immediately placed in the thermal cycler. Extraction reagent and PCR reagent CTRLs were included to control for any bacterial DNA contamination. The PCR cycling conditions were 95°C for 5 min, 35 cycles of 95°C for 20 s, 56°C for 30 s, and 72°C for 5 min. All PCR products were run on 1% agarose gels stained with ethidium bromide, and products were purified using the QIAquick Gel Extraction Kit. PCR products were further purified using AMPure magnetic purification beads (Beckman Coulter Genomics). AMPure-purified products were quantified with the Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen). Equimolar ratios of each sample were combined to create DNA pools of barcoded libraries for sequencing on the Roche 454 GS-FLX Titanium platform.

16S rRNA gene sequences were analyzed using the open source software package QIIME (Quantitative Insights Into Microbial Ecology; http://qiime.org/) v1.7 and v1.8. Sequences from each run were demultiplexed and quality filtered based on the following criteria: length outside bounds of 200 and 1000 nt, number of ambiguous bases exceeds limit of six, missing quality score, mean quality score below a minimum of 25, maximum homopolymer run exceeds a limit of six, number of mismatches in the primer exceeds the limit of zero, and include sequences without a discernible reverse primer. Sequencing data were further denoised using flowgram clustering with QIIME’s built in denoiser. Reverse primers were removed using fine-tuned BLASTing, as well as positional criteria. Chimeric sequences were identified with usearch61, which uses a combination of de novo and reference-based chimera-detection algorithms. The 13.5 release of the Greengenes dataset was used as the reference dataset. Identified chimeric sequences were removed. Demultiplexed, trimmed, and quality-filtered sequences were deposited in the Metagenomics Rapid Annotation using Subsystem Technology database under project ID 16329 (http://metagenomics.anl.gov/).

Open-reference operational taxonomic unit (OTU) picking was carried out with usearch61_ref, using Greengenes as the reference dataset at a similarity threshold of 97% (roughly corresponding to species-level OTUs). Representative sequences from the OTUs were aligned to a prealigned database of sequences (the Greengenes core set) using PyNAST, with quality thresholds set with a minimum sequence length of 150 nt and a minimum percent identity of 75%. PyNAST alignment failures were investigated by blasting all sequences that failed to align. Taxonomies were assigned to OTUs using the RDP Classifier trained on the Greengenes 13.5 dataset. α Diversity metrics (including Observed species, PD whole tree, and Chao1) were calculated, and rarefaction was plotted to investigate differences between groups for diversity within samples based on the abundance of various taxa within a community. To compare bacterial communities based on their composition between individuals and groups, β diversity (unweighted UniFrac and Bray-Curtis metrics) was assessed and visualized with principal coordinate analysis (PCoA) and bar charts. For each comparison (within and between groups), Bray-Curtis dissimilarities were used to calculate the mean and SEM, and differences between group dissimilarities were evaluated statistically using a two-tailed Student t test.

Recombinant vaccinia-OVA (vac-OVA) (27) was used to infect mice by i.p. injection at an inoculum of 5 × 10³–1 × 10⁵ PFU. DNA was purified from kidneys using the DNeasy Blood and Tissue Kit (QIAGEN, Valencia, CA), according to manufacturer instructions. Viral genomes were quantified using primers specific for vaccinia ribonucleotide reductase (Vv14L), as described by Freyschmidt et al. (28). qPCR was performed using the ABI ViiA 7 real-time PCR system (Applied Biosystems). To determine the vaccinia virus genome copies in samples, a standard curve was generated using DNA from purified vac-OVA stock and converted to copy number. Viral copies were normalized to the amount of tissue.

Naive mice or mice 3–8 d postinfection were euthanized by CO₂ inhalation. Peritoneal exudate cells (PECs) were aspirated following lavage of the peritoneum with 1–2 ml sterile PBS. Spleens, mesenteric lymph nodes, and thymus were mechanically disrupted to obtain single-cell suspensions. Splenocytes were treated with ACK buffer to lyse RBCs and washed with FACS Medium (HBSS containing 1% FBS and 0.1% sodium azide) prior to cell surface staining. Lymphocytes were stained with optimal concentrations of the following Abs: CD3ε (clone 145-2C11), CD4 (clone GK1.5), CD8α (clone 53-6.7), CD11b (clone M1/70), CD11c (clone N418), CD19 (clone 6D5), CD25 (clone PC61), CD27 (clone LG.3A10), CD44 (clone IM7), CD49b (clone DX5), CD62L (clone MEL-14), CD69 (clone H1.2F3), CD71 (clone RI7217), CD103 (clone 2E7), CD317 (PDCA-1; clone 927), B220 (clone RA3-6B2), F4/80 (clone BM8), GR1 (clone RB6-8C5), I-A/I-E (clone M5/114.15.2), KLRG1 (clone 2F1/KLRG1), NK1.1 (clone PK136), NKG2D (clone CX5), PD-1 (clone RMP1-30), and Siglec-H (clone 551) (all from BioLegend). Cells were analyzed on an LSR II flow cytometer (Becton Dickinson) using CellQuest software. Data were analyzed using FlowJo v10 analysis software (TreeStar).

Splenocytes harvested from vac-OVA–infected mice were incubated at a concentration of 2 × 10⁶ cells/ml in RPMI 10 (10% FBS, HEPES, l-glutamine, 2-ME, gentamicin sulfate, penicillin, streptomycin), with or without SIINFEKL peptide (5 μM), PMA (10 ng/ml), and ionomycin (1 μg/ml), and with 2 μl/ml Brefeldin A (BioLegend) for 5 h at 37°C. Splenic CD8⁺ T cells from 15–17-d-old uninfected infant mice were isolated by MACS selection, following the manufacturer’s instructions (Miltenyi Biotec), with the addition of biotinylated anti-CD71 Ab (clone RI7217; BioLegend). Cells were cultured at a density of 7.5 × 10⁵ cells/200 μl/well of a 96-well tissue culture and stimulated with plate bound anti-CD3 (1 or 10 μg/ml, clone 145-2C11; BioLegend) and soluble anti-CD28 (2 μg/ml; clone 37.51; BioLegend) in RPMI 10. Five hours prior to harvest, at 24 or 72 h, Brefeldin A was added to the cultures. After stimulation, cells were washed and incubated with Abs for CD8α, CD25, CD69, CD44, and CD62L for 30 min on ice. The cells were fixed in Fixation Buffer (eBioscience) for 15 min at 4°C and permeabilized using 1× permeabilization Wash Buffer (BioLegend) for 30 min. Abs for IFN-γ (clone XMG1.2) and TNF-α (clone MP6-XT22; both from BioLegend) were added, and cells were stained for 30 min at 4°C. Following staining, cells were washed in permeabilization buffer and in FACS medium prior to analysis. Splenocytes harvested at day of infection (doi) 3 and doi 7 from vac-OVA–infected infant mice were surface stained, as above, prior to fixation and permeabilization using the True-Nuclear Transcription Factor Buffer Set (BioLegend), according to the manufacturer’s instructions. Bcl-2 (clone BCL/10C4; BioLegend) was added, and cells were stained overnight at 4°C and then washed in permeabilization buffer and in FACS medium prior to analysis.

Ninety-six–well MultiScreen-HA Filter Plates (Millipore) were coated overnight at 4°C with murine IFN-γ–specific mAb (10 μg/ml, clone R4-6A2; BioLegend). Splenocytes isolated from mice at doi 8 were added at 2-fold dilutions of 10⁶ cells/well in the presence of OVA peptide SIINFEKL (1 μM). After a 24-h incubation at 37°C and 5% CO₂, the plates were washed and incubated with biotinylated murine IFN-γ–specific mAbs (5 μg/ml, clone XMG1.2; BioLegend) for 24 h at 4°C. This was followed by incubation with streptavidin-alkaline phosphatase (Sigma-Aldrich) at a 1:1000 dilution for 2 h at room temperature. The spots were visualized and counted after development with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium substrate (Sigma-Aldrich).

Statistical analysis was performed using GraphPad Prism 4.0. Data were analyzed by one-way or two-way ANOVA with the Holm–Sidak posttest or by an unpaired two-tailed Student t test with Welch’s correction, as indicated. Differences in the abundance of bacteria were determined with linear discriminant analysis effect size (LEfSe) analysis, which couples tests of statistical significance with measures of effect size to rank the relevance of differentially abundant taxa (29). For LEfSe analysis, an α value of 0.05 for the Kruskal–Wallis test and a log-transformed linear discriminant analysis score of 2.0 were used as thresholds for significance. In all analyses, p < 0.05 was considered statistically significant.

To maximally deplete the GIT microbiota of adult pregnant mice in density and composition, we used a combination of antibiotics that have broad-spectrum activity against luminal bacteria of mice and humans (1). Ampicillin and clindamycin are used commonly in the clinical setting. Although streptomycin is used less often, its activity mirrors gentamicin and is not absorbed from the intestine. This combination of antibiotics appreciably, but not completely, reduces aerobic and anaerobic Gram-positive and Gram-negative microbes in the GIT microbiota of adult mice within 3 d (25) and limits the possible expansion of vancomycin-resistant microbes. Pregnant mice were treated with this antibiotic mixture in their drinking water for 3–5 d prior to birth of a litter and throughout the duration of the suckling/preweaning period when vaccinia infection experiments were conducted (Fig. 1A). We specifically chose this short period of treatment before the birth of a litter to prevent any possible deleterious effects of the antibiotics on prenatal development. Thus, the infant mice were born to a mother with an altered intestinal microbiota. Although the infants may experience limited exposure to antibiotics during nursing, the levels transferred into milk are quite low and subtherapeutic (30–32). There were no appreciable differences in the size of litters or weights of infant mice between MAT and CTRL groups, nor did any adult mice treated with antibiotics exhibit signs of dehydration or weight loss.

To assess compositional changes in the GIT microbiota mediated by antibiotic treatment, we performed qPCR with universal bacterial and group-specific primers targeting the 16S rRNA genes of the phyla Bacteroidetes and Firmicutes and the genus Lactobacillus. These represent the dominant bacterial taxa in the murine GIT microbiota (33) and can be used to assess major changes in total bacteria and the major bacterial subgroups of the mouse GIT consortium (34). Because infant mice do not excrete significant stool pellets, we isolated their cecal contents and used fecal pellets from their mothers to assess differences in the microbiota. We found that the microbiota of MAT infant mice was affected in a manner that seemed to parallel their mothers’ (Fig. 1B). MAT infant mice exhibited a significant reduction in total bacteria (measured by the universal 16S rRNA primer set) and the bacterial taxa Bacteroidetes, Firmicutes, and Lactobacillus compared with CTRL infant mice. The antibiotic treatment consistently reduced the total bacterial load in the stool by 10,000–100,000-fold in MAT compared with CTRL adult and infant mice.

Although highly quantitative, qPCR can only target a subset of the diverse bacterial taxa inhabiting the GIT microbiota. Therefore, we sought to characterize the GIT microbiota with greater resolution. 16S rRNA gene pyrosequencing was used to assess the overall diversity and composition of the microbiota in stool of mothers and cecal contents from infants. The mothers were housed separately in different cages, and the infants were born to one of the mothers. Open-reference OTU clustering (97% clustering) with QIIME revealed a total of 1077 species-level OTUs from 16S rRNA gene pyrosequencing. α Diversity measures (observed species, PD whole tree, and Chao1) indicated significant differences in bacterial diversity and richness between all groups. Rarefaction analysis based on observed species is shown in Fig. 1C (data not shown for the other α diversity metrics). Antibiotic treatment significantly reduced all α diversity measures in MAT mothers compared with CTRL mothers (p < 0.001 for all three α diversity metrics, Student t test). Consistent with the development of the infant GIT microbiota (8), CTRL infant mice had lower diversity than their mothers, as might be expected in preweaning infant mice consuming only breast milk. MAT mothers’ microbiota diversity was even lower than CTRL infants (p < 0.01, for all three α diversity metrics, Student t test), suggesting a profound effect of antibiotic treatment on the diversity of the GIT microbiota in mothers in this model. Remarkably, although MAT infants were not directly treated with antibiotics, they exhibited the lowest overall GIT microbiota diversity, which was significantly lower than that of CTRL infants (p < 0.01, for all three α diversity metrics, Student t test). Diversity was also significantly reduced in MAT infants compared with MAT mothers (p < 0.01, Student t test), indicating that MAT has a profound impact on GIT microbiota diversity of their offspring.

To investigate relationships among the four groups of mice based on differences in phylogenetic diversity, PCoA of unweighted UniFrac distances was evaluated (Fig. 1D). The resulting PCoA plot revealed clustering of all four groups based on treatment and maternal versus infant groups. Antibiotic treatment explained the majority of the variance between groups, and both MAT mothers and infants separated from CTRL mothers and infants along PC1. Along PC2, MAT mothers separated dramatically from MAT infants, suggesting distinct differences between the microbiota of these two groups.

Bacterial distribution was dominated by 10 bacterial genera in mothers and their infants (Fig. 1E). Compositional analysis revealed that GIT microbiota of MAT infant mice was almost entirely dominated by Enterococcus spp., most closely related to E. faecalis strains (99.40–99.61% sequence similarity based on Greengenes BLAST of representative sequences). In contrast, MAT mothers were almost entirely dominated by Streptococcus spp. (most closely related to S. thermophilus strains; 99.59–99.80% sequence similarity) and Ralstonia spp. CTRL mothers and their infant mice demonstrated a diverse GIT microbiota composition. CTRL mothers were characteristically dominated by Bacteroidetes (Bacteroidales) and Firmicutes (Clostridiales and Oscillospira) members. CTRL infant mice had a microbiota rich in Lactobacillus, as would be expected based on their breast milk diet and from previous studies on the development of infant mouse GIT microbiota (8, 9). Thus, our taxa-specific qPCR results were consistent with 16S rRNA gene pyrosequencing, showing reduced quantity and relative abundance of the dominant bacterial taxa in MAT mothers and their infants, respectively. 16S rRNA gene pyrosequencing further showed near-complete dominance, in terms of relative abundance, of E. faecalis in MAT infant mice and S. thermophilus in their mothers.

There were no major differences in systemic lymphoid organ cellularity and distribution of major lymphocyte populations noted in CTRL and MAT infant mice (Supplemental Fig. 1). The CD4/CD8 ratio in spleens and thymus of MAT infant mice was higher than in CTRLs at 10–21 d of life. In addition, the percentage of splenic CD71⁺ cells, recently determined to have immune-suppressive capacity in infants (35), was notably lower in MAT mice after 12 d of age.

We hypothesized that GIT microbiota dysbiosis in infants would disrupt the immune response to viral challenge. We chose to focus on the response against vaccinia virus, because a systemic infection can be achieved by i.p. infection, and CD8⁺ T cells are required to clear the infection and to maintain durable immunity (36). We infected infant mice with a recombinant vaccinia strain expressing the whole-protein OVA (vac-OVA) so that we could track OVA Ag-specific responses. We observed that 15-d-old CTRL mice were able to survive infection when they were infected with up to 1 × 10⁵ PFU vac-OVA virus, whereas MAT mice succumbed to infection (Fig. 2A). MAT mice survived following infection with 5 × 10³ PFU vac-OVA, indicating that dysbiosis shifts susceptibility and reduces resistance to higher infectious inoculum. However, even at the lower infectious inoculum, viral burden was higher in MAT mice compared with CTRL mice (Fig. 2B).

Strikingly, MAT infant mice appeared well until 7–10 d following infection with vac-OVA (Fig. 2A), which corresponds with the peak of the antiviral CD8⁺ T cell response (37). We determined that i.p. infection of adult mice resulted in robust CD8⁺ T cell responses in the spleen and within the peritoneal cavity, the primary site of infection, where activated immune effector T (Teff) cells were detected at 7–10 d following infection (data not shown). We assessed CD8⁺ T cell responses of MAT and CTRL infant mice by harvesting lymphocytes from individual spleens and pooling PECs obtained by lavage of the peritoneum. All of the subsequent experiments were performed with a sublethal inoculum of virus in 15–20-d-old infant mice to compare bulk and Ag-specific CD8⁺ Teff cell phenotype and function in CTRL and MAT mice.

At doi 8, there was a high frequency of CD8⁺ T cells in the spleens and PECs of CTRL and MAT mice (Fig. 3A). However, the ratio of CD8⁺ T cells between uninfected and infected infant mice was significantly reduced in the PEC of MAT mice relative to CTRLs (Fig. 3A). Prior to infection, the frequency and absolute number of naive splenic CD8⁺ T cells was similar between CTRL and MAT infant mice (Fig. 3B). Sufficient numbers of lymphocytes could not be obtained from the peritoneum of uninfected mice, making similar analysis impossible. At doi 3, MAT and CTRL splenic CD8⁺ T cells still maintained a high proportion of naive T cells (Fig. 3C). By doi 8, the Teff subset (defined as CD44⁺CD62L⁻) of CD8⁺ T cells had also expanded in the spleen and PECs of MAT and CTRL infant mice.

Activated CD8⁺ T cells differentiate into distinct subpopulations during a primary viral infection and are categorized by functional activity and the coordinate expression of specific cell surface markers, including, KLRG1 and TNFR (CD27) (38, 39). The expression of CD27 is also associated with generation and maintenance of long-lived memory CD8⁺ T cells following vaccinia infection (40, 41). To further delineate the phenotype of splenic and PEC effector/memory CD8⁺ T cells in infant mice following vac-OVA infection, we examined the expression of KLRG1 and CD27 on CD8⁺ Teff cells at the peak of the antivaccinia viral response (doi 8). We found that the proportion of CD8⁺ T cells accumulating in the peritoneum with central memory T cells (Tcms) and Teff phenotypes was higher in CTRL mice than in MAT mice (Fig. 3C, 3D). This distinction was not apparent in the splenic CD8⁺ T cells (Fig. 3C, 3D). In addition, we noted that CD44 expression was higher on CTRL versus MAT PEC CD8⁺ T cells (mean fluorescence intensity: 12,491 versus 3,258; Supplemental Fig. 2A). Analysis of KLRG1 and CD27 revealed that the percentage of PEC CD27⁺CD8⁺ T cells from MAT infant mice was also lower relative to that of CTRL infant mice (Fig. 3D).

One explanation for the enhanced morbidity observed in MAT infant mice following infection with vaccinia could be insufficient functional responsiveness of Ag-specific effector CD8⁺ T cells. To assess this, we examined the OVA-specific IFN-γ response of T cells isolated from the spleens of infected mice (Fig. 4) using peptide-specific stimulation assays. Although the expansion and phenotype of CD8⁺ T cells in the spleens of infected MAT and CTRL infants were almost identical (Fig. 3), they were functionally different. At doi 3, the percentage of Ag-specific IFN-γ–producing CD44⁺CD8⁺ T cells was similar in CTRL and MAT infant mice (Supplemental Fig. 2B). By doi 8, despite similar absolute numbers of CD8⁺ T cells (Fig. 4B), the Ag-specific IFN-γ response of CTRL infant splenic CD8⁺ Teff cells remained significantly higher than in MAT infants (Fig. 4A, 4C). This was confirmed in a separate experiment by ELISPOT (Fig. 4D, independent experiment shown in Supplemental Fig. 2B), in which the MHC OVA-specific peptide SIINFEKL was used to assess the response of Ag-specific CD8⁺ T cells.

The antiapoptotic molecule Bcl-2 plays a critical role in the survival of CD8⁺ Teff cells during acute viral infections and their transition to memory cells (42). We hypothesized that the reduced frequency and absolute number of IFN-γ–producing CD8⁺ Teff cells seen in MAT infected infant mice at doi 8 could be due to a defective expression of Bcl-2 early during infection. Analysis of Bcl-2 expression in the CD8⁺ Teff subset from CTRL and MAT infant mice at doi 3 revealed that MAT CD8⁺ Teff cells expressed significantly reduced levels of Bcl-2 (Fig. 4E). However, at doi 7, consistent with the peak of the viral response, CTRL and MAT CD8⁺ Teff cells expressed low levels of Bcl-2 (Fig. 4E). Interestingly, compared with CTRL CD8⁺ Teff cells, MAT Teff cells exhibited significantly increased expression of the inhibitory receptor PD-1 at doi 3, and this remained elevated at doi 7 (Fig. 4F). Taken together, these results suggest that CD8⁺ Teff cells from MAT infected infant mice express markers that are associated with cell death or exhaustion.

The differential in vivo IFN-γ response of activated MAT CD8⁺ T cells following systemic infection with vac-OVA could suggest an intrinsic defect in the T cells that developed in the context of a dysbiotic GIT microbiota in these mice. To address this possibility, we isolated splenic CD8⁺ T cells from 15–17-d-old CTRL and MAT uninfected infant mice to assess their activation, differentiation, cytokine production, and proliferative capacity in response to TCR and CD28 stimulation. As shown in Fig. 5A, at 24 and 72 h poststimulation, the percentage of CD8⁺ T cells that differentiated into Teff cells was equivalent in CTRL and MAT infant mice. Moreover, the frequencies of CD69⁺ (Fig. 5B) and CD25⁺ CD8⁺ Teff cells from CTRL and MAT cells were similar (Supplemental Fig. 3A), as was their proliferative capacity (Supplemental Fig. 3A). At 24 h poststimulation, the percentage of IFN-γ–producing CTRL and MAT CD8⁺ Teff cells was similar (Fig. 5C). However, by 72 h poststimulation, the percentages of IFN-γ–producing (Fig. 5C) and TNF-α–producing (Supplemental Fig. 3B) CD8⁺ Teff cells were significantly reduced in MAT Teff cells compared with CTRL Teff cells. The in vitro stimulation results suggest that, despite similar phenotype and proliferation, MAT infant CD8⁺ Teff cells are unable to sustain production of cytokines following initial TCR engagement, and this defect could not be overcome by stronger TCR/CD28 stimulation. These results seem to parallel the in vivo observations of early and equivalent frequency of IFN-γ–producing CTRL and MAT CD8⁺ Teff cells, followed by a substantial reduction later during infection. Collectively, the in vivo and in vitro results hint that intestinal dysbiosis may result in the development of CD8⁺ T cells with intrinsic defects in effector function.

Reductions in subsets, inflammasome activation, and antiviral gene expression profiles of monocytes, dendritic cells (DCs), and NK cells in antibiotic-treated adult mice were described (22, 23, 43, 44). Thus, we sought to further characterize the MAT infant model and conducted an analysis of innate subsets also implicated in antivaccinia immunity (45, 46). Analysis of splenic DCs at doi 3 revealed that MAT infant mice had a reduced frequency and absolute number of MHC-II^hiCD11c^hi DCs (Fig. 6A). Subset analysis indicated similar frequencies of CD8α⁺, CD8α⁻, and plasmacytoid DCs in CTRL and MAT infant mice but a significantly higher frequency CD11b⁻CD103⁺ DCs in MAT infant mice. We also observed differences in the percentage and phenotype of accumulating NK cells in the spleens and peritoneum of MAT and CTRL mice at doi 8 (Fig. 6B, Supplemental Fig. 4). CTRL infants consistently exhibit a higher percentage of NK1.1 cells in the peritoneum following infection (Supplemental Fig. 4), and a higher proportion of these NK cells express CD49b⁺ in both the spleen and PECs, which is associated with a more mature subset (47). In addition, the expression of markers associated with activation, such as CD11b, CD62L, and NKG2D (48, 49), were also expressed more highly in CTRL splenic and PEC NK cells. Collectively, these observations indicate that a deficiency in certain innate immune populations is also affected by dysbiosis in infant mice and could contribute to poor antiviral immunity.

Previous studies demonstrated that mice of identical strains housed in different facilities are distinguished by their GIT microbiota and that alterations can occur following transfer from one facility to another (50, 51). During the course of our experiments, we initially bred mice in an SPHF facility (standard SPF and Helicobacter pylori–free conditions) where additional strict attention to maintenance of sterile conditions is emphasized. We suspected that the environmental impact of such hygienic conditions could also impact immunity of CTRL infant mice born in the SPHF facility. Indeed, we observed that CTRL infant mice were more susceptible to vaccinia infection if they were born in the SPHF environment (Fig. 7A). This susceptibility disappeared when mice were transferred to or bred in a conventional SPF BL2 facility prior to infection (Fig. 2). We reasoned that the susceptibility could be driven by changes in GIT microbiota in these mice. α Diversity measures did not reveal significant differences in overall bacterial diversity between mothers and their infants born in SPF versus SPHF facilities (Fig. 7B). PCoA analysis revealed that the microbiota of CTRL infants born in the SPHF facility was distinct from that of their counterparts born in an SPF facility, who clustered closely with their mothers (Fig. 7C). SPHF-born CTRL infants were notably separated along PC1 and PC2, along with some of their mothers, indicating that the GIT microbiota of these infants is affected by the environment and perhaps is unstable in its development. Intragroup Bray-Curtis dissimilarity revealed a significantly higher degree of dissimilarity within mothers and their infants housed in the SPHF facility compared with mothers and infants housed in the SPF facility. The greatest dissimilarity was found in CTRL infants housed in the SPHF environment (Fig. 7D). These results suggest a high degree of interindividual variation in the microbiota of mothers and infants in the SPHF facility. Intergroup Bray-Curtis dissimilarity revealed that infants born in the SPHF environment were significantly more dissimilar from their SPHF mothers than were infants born in the SPF environment compared with their SPF mothers (Fig. 7D). Thus, infants born in the SPF environment seem to acquire a microbiota more similar to their mothers than do infants born in the more hygienic SPHF environment. Furthermore, the dissimilarity between infant mice born under SPHF and SPF conditions was significantly greater than the dissimilarity between mothers from the two facilities (Fig. 7D), suggesting that the hygienic conditions of the SPHF environment have the greatest impact on the infants. LEfSe analysis identified overrepresentation of Clostridiaceae, SMB53, and Turicibacter spp. in CTRL infant mice born in SPHF facilities compared with SPF facilities, whereas CTRL infant mice born in SPF facilities had more unclassified Clostridiales, Anaerotruncus, and unclassified Lactobacillus (Fig. 7E). These results suggest that differences in specific bacterial taxa, rather than loss of overall diversity, may contribute to the increased susceptibility to vaccinia infection in the SPHF facility. Collectively, these results highlight the environmental influence of mouse facility breeding and husbandry practices that can subtly affect the developing infant GIT microbiota, with resulting deleterious immunologic outcomes.

We demonstrate that GIT dysbiosis occurring in parallel with a developing immune system during infancy appears to alter adaptive immune responses to systemic viral infection against vaccinia. GIT dysbiosis reduces innate and adaptive immune responses against influenza and lymphocytic choriomeningitis virus in infected adult mice (22, 23). Our study is one of the few reports to specifically address the impact of GIT dysbiosis on infant antiviral immunity. As we show in our model, GIT dysbiosis that developed in MAT infant mice was a consequence of antibiotic treatment of their mothers during the perinatal and postnatal period. GIT dysbiosis also developed in CTRL infant mice born under hygienic conditions. The disrupted colonization of the infant GIT microbiome caused by antibiotic treatment of the mothers resulted in enhanced susceptibility to vaccinia infection. This was characterized by decreased Ag-specific IFN-γ–producing systemic CD8⁺ Teff cells observed during the course of infection in vivo.

Unexpectedly, the impact of the dysbiotic host intestinal environment also appears to exert an influence on intrinsic CD8⁺ T cell function. Although MAT infant CD8⁺ T cells elicit a burst in IFN-γ following in vitro TCR stimulation and infection, this appears to be short-lived. We observed corresponding decreased Bcl-2 expression and increased PD-1 expression in MAT infant CD8⁺ T cells early during infection. This differential activation and expression of regulatory molecules in MAT CD8⁺ T cells hint at altered intrinsic function in the dysbiotic state. Huang et al. (52) also observed altered TCR responsiveness, including increased activation-induced cell death, in splenic CD4⁺ T cells of adult mice that were colonized with a limited flora during infancy. How a dysbiotic intestinal microbiota can influence this is presently unknown. However, short-chain fatty acids produced by the microbiota are shown to modulate regulatory T cell generation (53). An imbalance in these or other microbial-derived metabolites could, through direct or indirect means, also modify the responsiveness of T cell subsets and, thereby, viral immunity.

The development and function of innate subsets in infants are also generally less well characterized. Recently, Deshmukh et al. (54) reported that antibiotic treatment during the perinatal period led to decreased granulocytosis and neutrophil homeostasis in infant mice. This resulted in their enhanced susceptibility to E. coli and Klebsiella pneumoniae infection regulated by IL-17–producing innate lymphoid cells. NK cells and monocyte/DC subsets could also be modulated by the microbiota (19, 44). In germ-free mice, NK cells develop, albeit with a greater fraction expressing an immature phenotype, and they are poorly primed and activated by microbial ligands or during infection (43). MAT infants also appear to harbor less mature (CD49b⁺) and activated (NKG2D⁺, CD11b⁺, CD62L⁺) NK cells at the peak of infection. Diminished NK cell activity early during infection could also restrain the MAT CD8⁺ Teff cell response by allowing uncontrolled viral replication (as we observed) that overwhelms the capacity of the infant T cell response (48). The dynamic contribution of NK cells to antiviral immunity in MAT infants requires a more detailed analysis of their development, phenotype, and function within the infant dysbiotic environment.

Infant CD8⁺ T cells may already be biased toward a short-lived effector phenotype (24), and a dysbiotic GIT microbiota may further skew this. In an adult mouse model of respiratory CMV infection, the microbiota was required to promote viral-specific CD8⁺ memory T cells (21). Analysis of memory T cell development, expansion, and maintenance, as regulated by the GIT microbiota in other adult infection or infant mouse models, has not been explored. Human infants treated with antibiotics in the first week of life experience profound deficits in the density and diversity of their GIT microbiota (11, 12). Coincidentally, and in accordance with clinical practice guidelines, many infants receive their first vaccine just prior to hospital discharge. An impact of GIT dysbiosis on the intrinsic function, differentiation, and maintenance of Teff and memory T cells could have important implications. For example, infants may not be maximally protected if they receive their first vaccines or are infected during a period of GIT dysbiosis.

The profoundly defective response to a systemic viral infection in our MAT infant mice is associated with a dramatic reduction in the quantity and diversity, as well as an altered composition, of the intestinal microbiota. This alteration in infant microbiota by MAT was demonstrated by other investigators using similar protocols and length of treatment to assure stable alteration of the intestinal flora, even though such heavy usage is less likely to occur in the human clinical situation. Although we cannot exclude the possibility that antibiotics are transferred to the infant via mother’s milk or that translocation of maternal microbes into the bloodstream and milk of lactating mothers leads to colonization of their infants, we think that this is not the case. The MAT infant mice in our model do not simply acquire their own mother’s altered GIT microbiota. Instead, they are uniquely dominated by E. faecalis, a GIT commensal that also has pathogenic potential. E. faecalis is often among the first GIT colonizers, and it promotes the successive colonization of beneficial microbes in the infant GIT (55). Human newborn stool isolates of E. faecalis can modulate IL-8 (attenuated) and IL-10 (promoted) cytokine expression in colonic epithelial cell lines (55, 56). This may support a relatively suppressed and Th2-shifted immune status that is coordinate with mucosal commensal colonization and exposure to environmental Ags following birth. The MAT infant mice in our model are unique in maintaining colonization with this commensal. GIT microbiota and immune cell–reconstitution experiments will allow us to confirm the influence of a dysbiotic GIT microbiota and determine the parameters that promote recovery of the GIT microbiota and antiviral immunity in MAT infant mice.

Finally, we observed that hygienic environmental conditions affect the GIT microbiota of CTRL infant mice not treated with antibiotics, also resulting in enhanced susceptibility to vaccinia infection. Analysis of the GIT microbiota of infant mice born in the SPHF facility revealed that compositional differences distinguished these infants from those born in a typical SPF environment. Turicibacter was previously identified as a commensal regulated by CD8⁺ T cells in an adult restricted-flora mouse model; when CD8⁺ T cells are deficient, Turicibacter blooms (57). Our analysis of the GIT microbiota was performed on age-matched uninfected infant mice; thus, we cannot firmly conclude that the CTRL infant mice destined to succumb to vaccinia infection were those that harbored a GIT microbiota distinct from resistant CTRL infants. This will require prospective analysis, along with closer examination of cellular immune responses. Nevertheless, these results underscore the sensitive relationship that exists in the regulation of antiviral immunity and the effect of hygiene on the acquisition and progressive colonization of the GIT microbiota during infancy.

Infants are at greater risk for infection by, and dissemination of, lower viral inoculum; therefore, the ability to generate and maintain an adequate immune response is critical for survival. The global improvements in hygiene and liberal antibiotic-prescribing practices are associated with increasing rates of autoimmune and allergic disorders. These illnesses may be imprinted during infancy, only to manifest later in life. Identifying whether and how the GIT microbiota promotes immunity against viruses is also important for our general understanding of immune system development during infancy and has important implications for vaccine design, immunization strategies, medical practice, and disease prevention. The maternal–infant model presented in this article also has relevance to the human circumstance, particularly as it relates to full-term or preterm infants who were exposed to antibiotics prior to, during, and after birth.

We thank Dr. Donna Farber for review, editorial comments, and suggestions during the preparation of this manuscript and Dr. Michael Starnbach for support of E.S.N.L.-S. during the early development of the infant MAT model.

The authors have no financial conflicts of interest.