Limosilactobacillus reuteri and Lacticaseibacillus rhamnosus GG differentially affect gut microbes and metabolites in mice with Treg deficiency

Abstract

Treg deficiency causes a lethal, CD4⁺ T cell-driven autoimmune disease called IPEX syndrome (immunodysregulation, polyendocrinopathy, and enteropathy, with X-linked inheritance) in humans and in the scurfy (SF) mouse, a mouse model of the disease. Feeding Limosilactobacillus reuteri DSM 17938 (LR 17938, LR) to SF mice reprograms the gut microbiota, reduces disease progression, and prolongs lifespan. However, the efficacy and mechanism of LR, compared with other probiotics, in producing these effects is unknown. We compared LR with Lacticaseibacillus rhamnosus GG (LGG), an extensively investigated probiotic. LR was more effective than LGG in prolonging survival. Both probiotics restored the fecal microbial alpha diversity, but they produced distinct fecal bacterial clusters and differentially modulated microbial relative abundance (RA). LR increased the RA of phylum_Firmicutes, genus_Oscillospira whereas LR reduced phylum_Bacteroidetes, genus_Bacteroides and genus_Parabacteroides, reversing changes attributed to the SF phenotype. LGG primarily reduced the RA of genus_Bacteroides. Both LR and LGG reduced the potentially pathogenic taxon class_γ-proteobacteria. Plasma metabolomics revealed substantial differences among 696 metabolites. We observed similar changes of many clusters of metabolites in SF mice associated with treatment with either LR or LGG. However, a unique effect of LR was to increase the abundance of plasma adenosine metabolites such as inosine, which we previously showed had immune modulatory effects. In conclusion: 1) different probiotics produce distinct signatures in the fecal microbial community in mice with Treg deficiency; and 2) when comparing different probiotics, there are strain-specific microbial products with different anti-inflammatory properties, reinforcing the concept that “one size does not fit all” in the treatment of autoimmune disease.

NEW & NOTEWORTHY In the treatment of Treg-deficiency-induced autoimmunity, Limosilactobacillus reuteri DSM 17938 (LR) showed greater efficacy than Lacticaseibacillus rhamnosus GG (LGG). The study demonstrated that two different probiotics produce distinct signatures in the fecal microbial community in mice with Treg deficiency, but with many similarities in global plasma metabolites in general. However, there are strain-specific microbial products with different anti-inflammatory properties, reinforcing the concept that “one size does not fit all” in the treatment of autoimmune disease.

INTRODUCTION

Tregs maintain immune homeostasis and play a pivotal role in immune tolerance (1). Forkhead box protein 3 (Foxp3) is a major transcription factor that is associated with Treg cell development and function (2). Genetic mutations or deletions of the Foxp3 gene result in a primary immunodeficiency (PID) disease known as IPEX syndrome (immunodysregulation, polyendocrinopathy, and enteropathy, with X-linked inheritance) in humans (3, 4). The most common clinical presentation of IPEX syndrome is the classical triad of severe enteropathy, type I diabetes, and eczema of early onset (at 1.5–2 mo) (3–7). The scurfy (SF) mouse, bearing a mutation in the Foxp3 gene, displays a similar clinical phenotype, with early onset dermatitis, progressive multiorgan inflammation, and death within the first month of life caused by a lymphoproliferative syndrome (8, 9). It is currently unclear as to the trigger that initiates these immune responses.

Probiotics have the capacity not only to induce large-scale changes in the host microbiota but also to modulate the global metabolic functions of intestinal microbiomes (10–14). Probiotic Limosilactobacillus reuteri DSM 17938 (LR 17938 or LR) has been shown to have a mutualistic relationship with the human host (15). Commercially available worldwide, LR was derived from Lactobacillus reuteri ATCC 55730, a strain isolated from a Peruvian mother’s breast milk that was cured from two plasmids encoding antibiotic resistance (15, 16). LR inhibits pathogen growth and modulates the immune system (13). In humans, LR reduces the severity of acute infant diarrhea (17–19), may reduce the incidence of necrotizing enterocolitis (NEC) in premature infants (20–22), and decreases crying time in babies with colic (23, 24).

We have tested the effects of LR in several mouse models of human inflammatory and autoimmune diseases. We found this strain prevents experimental NEC in newborn animals by inhibiting the Toll-like receptor (TLR) 4-mediated NF-κB pathway (25), facilitating the induction of Tregs and lowering the number of inflammatory effector T cells (Teffs) in the intestinal mucosa (26, 27), a process requiring TLR2 (28). In addition, LR reduced the severity of experimental autoimmune encephalomyelitis (EAE), a mouse model of multiple sclerosis mainly driven by T_H1- and T_H17-induced inflammation (29). Both of these conditions were associated with altered gut microbiota (29–34).

In testing the effect of LR on SF mice, we found that LR prolonged the survival of SF mice and reduced organ inflammation as well as T_H1/T_H2-associated cytokines. Our studies identified a novel mechanism wherein the A_2A expressed predominantly on T cells was required for probiotic protection, inasmuch as LR had no therapeutic effect on SF mice with genetic knockout of receptor A_2A (35, 36). We discovered that the adenosine metabolite inosine (formed from adenosine by the action of adenosine deaminase, ADA) was reduced in plasma of SF mice and restored by LR treatment. Oral administration of inosine to SF mice also prolonged the survival of SF mice and reduced inflammation. Inosine similarly controlled inflammation by acting as an A_2A agonist (36). We found that the SF mouse demonstrates dynamic changes of gut microbial dysbiosis over the 22 days of lifespan, and early treatment of SF mice with LR modulates gut microbiota and its associated metabolites (36). However, we did not determine if the effect is specific to LR.

In the current study, we compared the effects of probiotic Lacticaseibacillus rhamnosus GG (LGG) (37) and LR with respect to mouse survival, microbiota, and microbial-associated metabolites in Treg deficiency. LGG is one of the most widely used and well-documented probiotic strains, often used to prevent or treat gastrointestinal infections, antibiotic-associated diarrhea, respiratory infections, and allergies (38, 39). Our studies aimed to further enhance our understanding of the specificity of probiotics in modulating autoimmunity, using a well-characterized model of Treg deficiency.

MATERIALS AND METHODS

Animals

Wild-type (WT) C57BL/6J and heterozygous B6.Cg-Foxp3^sf/J mice were purchased from Jackson Laboratories and allowed to acclimatize for 2 wk before experimentation. SF mice with hemizygous B6.Cg-Foxp3^sf/Y were generated by breeding heterozygous B6.Cg-Foxp3^sf/J female (Jackson Lab 004088) to C57BL/6J male (Jackson Lab 000664) mice. Only male mice were used in this study due to the Foxp3 gene existing on the X chromosome. In each litter of breeding pairs, all males were either SF used as the experimental groups or WT littermates used as the control groups. All mice were housed in the animal facility at the UT Health Science Center in Houston. This study was carried out in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals (NIH) and the Institutional Animal Care and Use Committee (IACUC). The study was approved by the IACUC (protocol numbers: AWC-14-056 and AWC-17-0045).

Preparation of Probiotics and Treatment of SF Mice

Probiotic LR and LGG were provided by BioGaia AB (Sweden) and were prepared as described previously (36). Briefly, probiotics were anaerobically cultured in deMan-Rogos-Sharpe (MRS) medium at 37°C for 24 h followed by plating in MRS agar at specific serial dilution and culturing anaerobically at 37°C for 48–72 h. Quantitative analysis of bacteria in culture media was performed by comparing absorbance at 600 nm of cultures at known concentrations, using a standard curve of bacterial colony forming unit (CFU)/mL grown on MRS agar.

Mice were fed probiotics or media (control) by oral gavage starting from the date of postnatal day 15 (d15) once we noted the SF clinical phenotype (crusted skin on tail and ears and/or deformed ears), and probiotics were given according to the protocols indicated in Supplemental Fig. S1A (all Supplemental material is available at https://doi.org/10.6084/m9.figshare.14062001) and Fig. 1A to observe the survivals. We fed 10⁷ CFU daily, in a volume of 100 µL, using a polypropylene feeding tube (22 gauge × 25 mm, Instech Laboratories, Inc., Plymouth Meeting, PA). The experiments include four groups of SF mice: SF controls (SF, n = 17) or SF mice fed with LR for 1 wk (SF + LR 1 wk, n = 10), for 2 wk (SF + LR 2 wk, n = 9), or for 3 wk (SF + LR 3 wk, n = 10). We also studied SF mice fed with control media (MRS) (SF + MRS 1 wk, n = 6; SF + MRS 2 wk, n = 6; and SF + MRS 3 wk, n = 7). To compare LR with LGG, SF mice were fed 10⁷ CFU/day of each probiotic, given by gavage daily in 100 µL, starting from postnatal d15 for 2 wk, with comparison groups denoted as LR (SF + LR, n = 9) and LGG (SF + LGG, n = 10). We observed the mice daily, recording survival after 2 wk of probiotic treatment. In another sets of experiments, groups WT (n = 7), SF (n = 5), SF + LR (n = 5), and SF + LGG (n = 7) were fed with media from postnatal d15 to d21 to collect plasma and cecal content for analysis of microbiota and metabolomics. Cecum and rectum were opened, and the contents were removed into sterilized microcentrifuge tubes and immediately frozen in liquid nitrogen and stored at −80°C freezer. We euthanized mice at d22 of age (before weaning) to investigate gut microbiota and metabolomics while avoiding microbial variations due to weaning (Fig. 1A).

Figure 1.

The effect of Limosilactobacillus reuteri DSM 17938 (LR) or Lacticaseibacillus rhamnosus GG (LGG) on the survival rate of scurfy (SF) mice. A: scheme for LR or LGG treatment on SF mice for sample analysis and survival observation. B: survival curves of SF, wild-type (WT), SF + LR, and SF + LGG mice (Kaplan–Meier survival curves and log-rank test; n = 9–17 mice per group). CFU, colony-forming unit; MRS, deMan-Rogos-Sharpe.

Determining the Population Structure of the Microbiota

Feces from cecum to rectum of mice were collected. The Louisiana State University School of Medicine Microbial Genomics Resource Group (http://metagenomics.lsuhsc.edu) performed sequencing and bioinformatics. Genomic DNA extraction was performed using QIAamp Fast DNA Stool Mini Kit (Qiagen, Germantown, MD), and bacterial profiling by amplification and sequencing of the V4 region of the 16S rRNA gene was performed as previously described (40). The relative abundance (RA) of each operational taxonomic unit (OTU) was examined at phylum, class, order, family, genus, and species levels. Bacterial α- and β-diversity metrics, as well as taxonomic community assessments, were produced using QIIME 1.8 (open source, http://qiime.org/) (41).

Plasma Global Metabolomics Analysis

We stored plasma samples at −80°C, and when individual studies were completed, samples were sent to Metabolon (https://www.metabolon.com/) for processing and measuring metabolomics profile in plasma from mice of WT, SF, SF + LR, and SF + LGG (36). Sixty hundred and ninety-six compounds of known identity in plasma were detected by a nontargeted metabolomics analysis platform including ultra-performance liquid chromatography-tandem mass spectrometer and gas chromatography-mass spectrometry. Global metabolomics profile data included fold changes of biomolecules from SF/WT, SF + LR/SF, and SF + LGG/SF mice, as well as information derived from the Interactive Pathways Explorer tool from Metabolon.

Statistical Analysis

For data with a Gaussian distribution, results were presented as means ± SE. Significance was determined using one-way ANOVA corrected for multiple comparisons with Tukey and Dunnett’s post-tests, or two-way ANOVA for multiple comparisons with a Bonferroni test. For data with Gaussian-approximate distribution, Kruskal–Wallis analysis was used for comparisons. Kaplan–Meier survival curves were graphed, and the comparison was analyzed using Logrank with χ² test. The statistical analysis was performed using Prism version 4.0 (GraphPad Software). Differences were reported as significant when P < 0.05.

RESULTS

Feeding Probiotic Prolongs the Survival of Treg-Deficient Scurfy Mice

We previously showed that oral feeding of LR 17938 either as an early treatment (starting at d8 of age) or later (starting on d15 of age) once daily prolongs the survival of SF mice from < 1 mo to > 4 mo of age (36). Initially, we aimed to optimize the treatment duration for different probiotic comparisons. We orally fed SF mice by gavage, 10⁷ CFU/day, daily starting on postnatal d15 once we observed the SF clinical phenotype, for 1, 2, or 3 wk. Subsequently, we stopped feeding LR to observe survival of the SF mice (Supplemental Fig. S1A). We found that oral feeding of LR to SF mice, with any studied duration of LR feeding, significantly increased the survival rate. The survival rate of SF mice with 2- or 3-wk treatments was significantly higher than that with 1-wk treatment; however, there were no significant differences between 2-wk and 3-wk treatment durations (Supplemental Fig. S1B). This finding indicated that 2-wk treatment could be used to compare the effects of LR and LGG on survival of SF mice (Fig. 1A). Results showed that SF mice died before 30 days of age without probiotic treatment, whereas both LGG and LR significantly prolonged the survival of SF mice (SF + LR vs. SF, P < 0.0001; SF + LGG vs. SF, P = 0.0149). However, LR exerted a greater effect on lifespan than LGG: the median lifespan of mice with LR was d73 versus LGG d35.5, and the longest lifespan of mice with LR was d151 versus LGG d66 (SF + LR vs. SF + LGG, P = 0.0053; Fig. 1B).

Probiotic Treatment Reprograms Gut Microbiota in SF Mice

We have previously reported that LR 17938 significantly changes the gut microbial dysbiosis identified in SF mice (36). To compare LR with LGG, we found that 1-wk feeding of probiotic (LR or LGG) led to important changes in the gut microbiota. We observed an increase in α diversity determined by the Shannon diversity index comparing both LR and LGG with SF mice (SF vs. WT, P = 0.008; SF + LR vs. SF, P = 0.014; SF + LGG vs. SF, P = 0.023; and SF + LGG vs. SF + LR, P = 0.088, Fig. 2A). Only LR increased microbial evenness (SF vs. WT, P = 0.0008; SF + LR vs. SF, P = 0.027; SF + LGG vs. SF, P = 0.088; and SF + LGG vs. SF + LR, P = 0.061, Fig. 2B). Either LR or LGG feeding shifted the microbial community composition, as shown by Jaccard_emperor β diversity, when comparing either SF + LR with SF (P < 0.001); SF + LGG with SF (P < 0.001); or SF with WT (P < 0.001). However, the two probiotics yielded populations of fecal microbiota which were comparable with that of WT mice, without a significant difference between them (Fig. 2C).

Figure 2.

Gut microbial alpha- and beta diversity affected by Limosilactobacillus reuteri DSM 17938 (LR) or Lacticaseibacillus rhamnosus GG (LGG). A: gut microbial Shannon diversity analysis and B: gut microbial evenness comparing scurfy (SF) + LR vs. SF; SF + LGG vs. SF; SF + LGG vs. SF + LR, and SF vs. wild-type (WT) (one-way ANOVA; n = 4–7 mice per group). C: Jaccard_emperor β diversity comparing SF + LR vs. SF; SF + LGG vs. SF; or SF vs. WT (n = 4–7 mice per group).

Taxa belonging to the phylum Firmicutes (∼60%) and Bacteroidetes (∼25%) dominated the core microbiota of WT mice (Fig. 3, A–C). The relative abundance of specific bacteria was calculated based on the OTUs of the bacteria normalized to total identified OTU counts by 16S rRNA gene sequencing in the sample. Even though we found comparable changes attributable to different probiotics in gut bacterial clusters, LR and LGG differentially modified some bacteria in the feces of SF mice. We found that at the phylum level, the relative abundance of Firmicutes was reduced in SF mice; however, Firmicutes were increased by LR (Fig. 3B). The relative abundance of Bacteroidetes was increased in SF mice but was reduced by LR (Fig. 3C). In contrast, LGG feeding did not reverse these bacterial changes in SF mice (Fig. 3, B and C).

Figure 3.

Relative abundance of predominant bacteria (>1% in any sample) at the phylum level. The relative abundance of specific bacteria was calculated based on the operational taxonomic units (OTUs) of the bacteria normalized to total identified OTU counts by 16S rRNA gene sequencing of the sample. A: overall relative abundance of bacteria compared among groups. B: the core microbiota of wild-type (WT) mice: the phylum Firmicutes (∼60%); and C: the core microbiota of WT: phylum Bacteroidetes (∼25%). (one-way ANOVA; n = 4–7 mice per group). LR, Limosilactobacillus reuteri DSM 17938; LGG, Lacticaseibacillus rhamnosus GG; SF, scurfy; OD1, Parcubacteria; TM7, Saccharibacteria.

LGG increased the relative abundance of proteobacteria at the phylum level (Fig. 4A). LGG was shown to increase the relative abundance of class level of α-Proteobacteria (Fig. 4B) and β-Proteobacteria (Fig. 4C), but LGG reduced the γ-Proteobacteria (Fig. 4D). Of importance, LGG significantly reduced the relative abundance of Enterobacteriaceae (Fig. 4E) classified as γ-Proteobacteria, as did LR (Fig. 4, D and E). Finally, LR increased the relative abundance of family_Ruminococcaceae, genus_Oscilospira [phylum_Firmicutes] (Fig. 5A) and reduced the relative abundance of family_Bacteroidaceae, genus_Bacteroides (Fig. 5B) and family_Porphyromonadaceae, genus_Parabacteroides (Fig. 5C) [phylum_Bacteroidetes]. LGG but not LR reduced the relative abundance of genus_Bacteroides (Fig. 5, B and C).

Figure 4.

Relative abundance of Proteobacteria affected mainly by Lacticaseibacillus rhamnosus GG (LGG). Phylum Proteobacteria (A); class_α-proteobacteria (B); class_β-proteobacteria (C); class_γ-proteobacteria (D); class_γ-proteobacteria, order_Enterobacteriales; family_Enterobacteriaceae (E) comparing groups wild-type (WT), scurfy (SF), SF + Limosilactobacillus reuteri DSM 17938 (LR), and SF + LGG (one-way ANOVA; n = 4–7 mice per group, data are presented as means ± SE).

Figure 5.

Relative abundance of bacteria at the genus level changed by Limosilactobacillus reuteri DSM 17938 (LR) or Lacticaseibacillus rhamnosus GG (LGG) in scurfy (SF) mice. Family_Ruminococcoceae, genus_Oscillospira (A); family_Bacteroideceae, genus_Bacteroides (B); family_Oirogtrinibadaceael, genus_Parabacteroides (C) comparing groups wild-type (WT), SF, SF + LR, and SF + LGG (one-way ANOVA; n = 4–7 mice per group, data are presented as means ± SE).

Plasma Metabolomic Profiles Are Modulated by Treg Deficiency and Probiotics

Metabolites of commensal bacteria in the lumen and mucosa play a key role in microbe-host interactions (42, 43). We analyzed metabolomic profiles of plasma obtained from WT, SF, and SF with either LR (SF LR) or LGG (SF LGG) after overt symptoms.

Changes in Circulating Metabolites in Response to Probiotic Treatment

There were 696-named biochemicals in plasma that were identified and analyzed. When comparing SF and WT mice, 202 recognized metabolites showed significant changes, with 79 metabolites upregulated and 123 metabolites downregulated. When we compared SF + LR with SF without LR, 419 metabolites were significantly changed—with 336 metabolites upregulated and 83 metabolites downregulated by LR. Comparing SF + LGG with SF without LGG, 316 metabolites were significant changed—with 248 metabolites upregulated and 68 metabolites downregulated.

Similar Changes in Metabolites Associated with LR and LGG

Principal component analysis (PCA) was successful in segregating samples by genotype (WT and SF) and by treatment (control vs. SF + LR, or control vs. SF + LGG) on Components 2 and 1, respectively (Fig. 6A). These results are reflective of profound metabolic differences being present among plasma samples driven primarily by mouse genotype. Of importance to the aims of the study, samples from SF mice treated with different probiotics (LR or LGG) tended to segregate close to each other, suggesting a comparable metabolic profile among these groups. Hierarchical cluster analysis (HCA), a stepwise clustering method that groups metabolically similar samples close to one another, illustrated similar results with successful clustering of samples by probiotic treatment with WT and SF groups clustering to the right (Fig. 6B). N-Acetyl-kynurenine, a tryptophan metabolite strongly associated with systemic inflammation when indoleamine 2,3-dioxygenase (IDO) is activated by proinflammatory cytokines, catalyzes the conversion of tryptophan to kynurenine (Fig. 6Ca). N-Acetyl-kynurenine level was increased in the plasma in SF mice compared with that in WT control, an effect which was reduced by both probiotics, LR or LGG (Fig. 6Cb).

Figure 6.

Plasma metabolomics profiles are modified by Limosilactobacillus reuteri DSM 17938 (LR) or Lacticaseibacillus rhamnosus GG (LGG) treatment of scurfy (SF) mice. A: principal component analysis (PCA) clustering from plasma metabolites of wild-type (WT), SF, SF + LR, and SF + LGG (n = 6–8 mice per group). B: heat map of 696 metabolites in plasma of WT, SF, SF + LR, and SF + LGG mice. Each lane represents a different mouse. C: metabolite N-acetyl-kynurenine associated with systemic inflammation (a) and changed by LR and LGG in SF mice (b) comparing groups WT, SF, SF + LR, and SF + LGG (one-way ANOVA; n = 4–7 mice per group, data are presented as means ± SE); **P < 0.01, SF + LR or SF + LGG compared with SF.

Unique Changes in Metabolites by LR or LGG

Comparative analysis of significantly altered biochemicals in groups SF + LR or SF + LGG compared with SF without probiotic treatment revealed that 47 (of 419 with significant changes) and 30 (of 316 with significant changes) metabolites were uniquely altered by LR and LGG treatment, respectively. Among these unique changes by LR or LGG treatment, we identified several differentially significant changes linked to LR but not to LGG.

Altered Inosine Metabolites Linked to LR

We found a significant decrease in inosine and its associated metabolites, xanthine and xanthosine, in SF-control in relation to WT-control. Notably, treatment of SF mice with LR resulted in significantly increased levels of inosine, xanthine, and xanthosine (Fig. 7, A and B), confirming the results from our previous study showing that the interaction of microbiota and inosine-A_2A receptor inhibits T_H1 and T_H2 cell-associated autoimmune disease (35, 36). In contrast, any changes of inosine metabolites by LGG treatment were not statistically significant (Fig. 7, A and B).

Figure 7.

Altered inosine metabolites by Limosilactobacillus reuteri DSM 17938 (LR) compared with Lacticaseibacillus rhamnosus GG (LGG). A: plasma intensity scale of inosine levels among the groups of wild-type (WT), scurfy (SF), SF + LR, and SF + LGG (n = 6–8 mice per group). B: the Interactive Pathway Analysis of nucleotides. The up- (red circle) or down- (blue circle), or no- (white circle) regulation of metabolites represent the group comparison of SF vs. WT, SF + LR vs. SF, or SF + LGG vs. SF. Each (red or blue or white) circle represents one metabolite. Each numbered circle represents the each nucleotide metabolism indicated in the figure.

Changes in Central Energy Metabolism—the Glycolysis-Tricarboxylic Acid Cycle

Glycolysis represents one of several pathways by which bacteria can catabolize glucose. SF mice showed a decline in glucose and the glycolytic intermediate 3-phosphoglycerate (3-PG) compared with WT mice. Interestingly, treatment of SF mice with LR but not LGG significantly increased relative levels of pyruvate and lactate, indicating increased glucose availability to be used in tricarboxylic acid (TCA) cycle. Significant decreases in several TCA cycle metabolites including α-ketoglutarate, succinate, fumarate, and malate were noted in SF control compared with WT control mice. LR treatment ameliorated the reduced levels of succinate, fumarate, and malate along with aconitate, whereas LGG treatment reversed succinate only (Fig. 8A).

Figure 8.

Differentially changed tricarboxylic acid (TCA) cycle and gamma-glutamyl (GG)-peptides by Limosilactobacillus reuteri DSM 17938 (LR) compared with Lacticaseibacillus rhamnosus GG (LGG). The Interactive Pathway Analysis: metabolites in energy (TCA cycle) (A); metabolites in peptide (GG-peptides) (B). The up- (red circle) or down- (blue circle) or no- (white circle) regulation of metabolites represent the group comparison of scurfy (SF) vs. wild-type (WT), SF + LR vs. SF, or SF + LGG vs. SF. Each (red or blue or white) circle represents one metabolite. Each numbered circle represents the each nucleotide metabolism indicated in the figure.

Evidence of Alterations in Redox Homeostasis with Probiotic LR Treatment in SF Mice

The tripeptide glutathione (GSH) is a small molecular weight thiol with antioxidant properties. Cysteine and methionine metabolism supports glutathione production, but there appeared to be minimal changes in cystathionine to cysteine conversion in probiotic-treated SF mice compared with SF controls. However, LR treatment upregulated the levels of 12 different gamma-glutamyl (GG) peptides (GGglutamate, GGhistidine, GGisoleucine, GGleucine, GG-alpha-lysine, GG-epsilon-lysine, GGmethionine, GGphenyalanine, GGthreonine, GGtryptophan, GGtyrosine GGvaline), whereas LGG treatment upregulated only two different GG peptides (GGglutamate and GGtryptophan) in SF mice (Fig. 8B). The addition of GG residues to amino acids plays a key role in the GG cycle, which maintains glutathione homeostasis and redox status (44).

LGG Effects on Bile Acid Metabolism

LGG uniquely upregulated both primary and secondary bile acid metabolism in SF mice compared with SF control mice. The upregulated primary bile acids in the plasma of SF mice which were fed LGG included cholate, with a fold change (FC) of 15, glycocholate (FC: 7.4), and β-muricholate (FC: 4.4). The upregulated secondary bile acids included 3-dehydrocholate (FC: 16.2), deoxycholate (FC: 4.4), and ursodeozycholate (FC: 3.6).

DISCUSSION

The probiotic L. reuteri DSM 17938 (LR) has immune modulatory functions and has been shown to have beneficial effects in infants with acute diarrhea (17, 18), infantile colic (23), and necrotizing enterocolitis (NEC) (21, 22), and in rodent models of NEC (25, 27, 45) and autoimmunity (35,36, 46). In the current study, we compared LR with another well-characterized probiotic, LGG, in the Treg-deficient SF mouse. We demonstrate that LR has more potency to prevent autoimmunity and prolong survival in this model than LGG, although both probiotics were beneficial. We observed that both probiotics restored fecal microbial alpha diversity, but they produced distinct fecal bacterial clusters and differentially modulated relative abundance of specific taxa. However, plasma metabolomics profiles from two different probiotic treatments were comparable, reflecting similar metabolic effects in vivo, even though some unique changes in metabolomics by LR were observed.

A Shift in Microbial Community Composition

LR treatment clearly altered the microbiota in this model, with markedly increased Firmicutes and reduced Bacteroidetes. After LR gavage feeding to SF mice, there were no observable Porphyromonadaceae, Bacteroides, or Parabacteroides, whereas Ruminococcaceae, important short-chain fatty acid (SCFAs) producers (47), were preserved. SCFAs, such as pentanoate (48), have anti-inflammatory mechanisms—such as inducing tissue levels of the anti-inflammatory cytokine IL-10 and suppressing the generation of proinflammatory cytokine IL-17a. We suggest that the preservation of certain genera by LR supplementation may have a role in immunomodulation. On the other hand, LGG’s major impact was to increase α- and β-Proteobacteria, which are mostly classified as commensal bacteria, even though the exact roles or functions of these microbes are unclear. LGG reduced γ-Proteobacteria, which represent a large class of Gram-negative bacteria that includes a number of potential pathogens such as Klebsiella, Enterobacter, Citrobacter, Salmonella, Escherichia coli, Shigella, Proteus, and Serratia (49). These pathogens occasionally spread to the bloodstream, resulting in life-threatening complications. Interestingly, even though LR did not affect the relative abundance of α- and β-Proteobacteria, it clearly reduced Enterobacteriaceae in SF mice. Both LR and LGG reduced Enterobacteriaceae, suggesting their potential applications for controlling or preventing the colonization or infection by pathogenic Enterobacteriaceae. A possible mechanism for Lactobacilli to reduce Enterobacteriaceae would be by reducing intestinal pH by producing lactic acid, acetic acid, formic acid, and other acids. Importantly, Lactobacilli can also secrete certain antimicrobial molecules, such as reuterin, ethanol, fatty acid, hydrogen peroxide, and bacteriocins to exert the antimicrobial activity (50–52). Lactobacilli have demonstrated the ability to inhibit several bacterial pathogens that belong to Enterobacteriaceae family, including Klebsiella oxytoca (53), E. coli (54), Shigella spp. (55), and carbapenem-resistant Enterobacteriaceae (CRE) (56). Recent studies showed a correlation between antimicrobial activity and metabolic profile produced by LR in cell-free supernatants, with the antimicrobial effect likely related to pH-dependent compounds (57).

Adenosine Receptor-Mediated Inflammatory Control

One factor that may result in the additional benefit of LR in this model was its property to act via leukocyte adenosine receptors. It is known that the lethal lymphoproliferative syndrome in SF mice is predominately mediated by T_H1 and T_H2 cell-induced pathology (58, 59). Activation of the adenosine/inosine/adenosine receptor 2A (A_2A) is one of the main mechanisms by which Treg cells control T_H1 and T_H2-induced autoinflammation (60, 61). In our previous studies, we demonstrated that LR treatment starting at day of life 8 increased plasma inosine levels. Inosine, a purine capable of activating adenosine receptors, was found to suppress differentiation of T_H1/T_H2 cells in vitro (36). Moreover, other studies reported that CGS21680, an agonist of A_2A receptors, inhibited the proliferation of CD4⁺ T cells that were isolated from WT mice but failed to block the proliferation of cells obtained from A_2A^−/− mice (60). These findings demonstrated that even in the absence of Tregs, the activated A_2A receptor may play a critical role in the suppression of T_H1 and T_H2 cells. However, LGG may not share this mechanism of action, because in LGG-treated mice inosine, hypoxanthine, and xanthosine were not upregulated.

Markedly Increased Gamma-Glutamyl Peptides

An unmistakable finding of this study was that LR and not LGG increases the levels of 12 different gamma-glutamyl (GG) peptides. GG peptides are reported to exhibit increased transport properties, because the GG moiety promotes translocation across lipid barriers (44). For example, the addition of GG residues to amino acid increases the translocation of amino acid across placental membranes and thereby helps to maintain glutathione homeostasis and redox status. Furthermore, gamma-glutamyl dipeptides are indicative of improved oxidative stress coping mechanisms (62). For example, microbial communities of patients suffering from ulcerative colitis were associated with a lower disease activity if they exhibited a capacity for quenching reactive oxidative species (ROS). Reduced ROS were associated with enhanced gamma-glutamyl transferase (GGT) activity, shown by a correlation between enrichment of GG peptides and preserved glutathione levels (63).

Changes in the gut microbiota in response to LR may play a role in the regulation of GG amino acids. It has been shown that the enrichment of Akkermansia muciniphila and Parabacteroides correlates with reduced fecal GGT activity and levels of colonic and serum GG peptides (64). The mechanisms by which microbes are able to change the levels of these GG peptides warrant further investigation. It will be important to determine how the microbiota-associated gamma glutamylation of AA plays a role in immunomodulation in other inflammatory and immune processes.

Bile Acid Metabolism Regulated by LGG

Bile acids function as hormones or nutrient signaling molecules that help to regulate glucose, lipid, lipoprotein, energy metabolism, and inflammatory responses (65, 66). Bile acids constitute the products of co-metabolism by both host and microbiome (67). Chronic low-grade inflammation associated with insulin resistance and inflammatory bowel disease (IBD) has shown to inhibit bile acid signaling and disrupt lipid metabolism (65). Currently, there are no specific reports on dysregulated bile acid pools in SF mice or IPEX syndrome. However, in our current studies, it showed that LGG upregulated several primary and secondary bile acids in SF mice. Secondary bile acids can act in multiple ways as signaling molecules, through activation of the nuclear Farnesoid X receptor (FXR), vitamin D receptor, and through surface expressed G-protein-coupled receptors (GPCRs) (66). Engagement of these compounds beneficially affect many physiological aspects (66, 68), especially, recently shown bile acids in regulating Treg-T_H17 axis by promoting Treg generation and reducing inflammatory T-cell function (69). However, accumulated higher levels of circulating bile acids have cytotoxic effects. The cytotoxicity of bile acids is dependent on their structural formation associated with their hydrophobicity. Increased serum levels of hydrophobic bile acids, such as chenodeoxycholic acid (CDCA) and deoxycholic acid (DCA), have been associated with colon cancer, gallstones, and other gastrointestinal (GI) diseases (70, 71). Recent studies in the mouse model of liver fibrosis showed that LGG prevents liver fibrosis through inhibiting hepatic bile acid synthesis and enhancing bile acid excretion (72).

The aim of this study was to determine the specificity of LR in Treg-deficiency-induced autoimmunity. From a clinical standpoint, comparative efficacy research for different probiotics in various diseases is very unlikely to take place. We believe that there is clear evidence that specific probiotics may be reliable in benefitting specific inflammatory conditions (12), but there are too many probiotics and too vast a number of diseases to allow comparison of the numerous probiotics; and the appropriate doses have also not been determined. It will also be impossible to conduct in vitro high-throughput studies of the different microbial species using intestinal epithelia or bioreactors, because these models may not be applicable to human diseases. However, in our opinion, a sound approach is to investigate well-established animal models to test a limited number of probiotics with proven physiological effects. Our study is one-step in this direction, and although IPEX syndrome is rare, LR may be beneficial in other Treg-deficiency-related monogenic immune disorders and possibly in other autoimmune conditions (73, 74), early-onset inflammatory bowel disease (IBD) (75), or graft-versus-host disease (GVHD) (76, 77).

SUPPLEMENTAL MATERIAL

Supplemental material is available at https://doi.org/10.6084/m9.figshare.14062001.

GRANTS

This work was supported by National Institutes of Health/National Center for Complementary and Integrative Health (NIH/NCCIH) Grant R01AT007083 (to J. M. Rhoads and Y. Liu) and by National Institute of Allergy and Infectious Diseases (NIH/NIAID) Grant R03AI117442 (Y. Liu). The authors acknowledge support of BioGaia AB in other research projects, but BioGaia did not fund this project.

DISCLOSURES

The authors acknowledge support of BioGaia AB in other research projects, but BioGaia did not fund this project. Stefan Roos has part-time employment by the company BioGaia AB, Stockholm, Sweden. Other authors declare no conflict of interest, financial or otherwise.

AUTHOR CONTRIBUTIONS

Y.L. and J.M.R. conceived and designed research; Y.L., T.K.H., E.S.P., J.F., and M.L. performed experiments; Y.L., T.K.H., C.M.T., E.S.P., J.F., and M.L. analyzed data; Y.L., T.K.H., C.M.T., S.R., and J.M.R. interpreted results of experiments; Y.L. and T.K.H. prepared figures; Y.L. drafted manuscript; Y.L., T.K.H., C.M.T., E.S.P., J.F., M.L., S.R., and J.M.R. edited and revised manuscript; Y.L., T.K.H., C.M.T., E.S.P., J.F., M.L., S.R., and J.M.R. approved final version of manuscript.