Impact of a GABA-producing <em>Lactococcus lactis</em> on microbiota and mycobiota during CNS inflammatory: Part 3

The gut microbial populations are key regulators of immune homeostasis; simultaneously, they are constantly exposed to changes, including those promoted by disease. In experimental autoimmune encephalomyelitis (EAE), the most used animal model to study multiple sclerosis (MS), we and others have documented changes in the abundance of bacterial taxa following the induction of the disease. Some of those changes affect bacteria capable of producing gamma-aminobutyric acid (GABA). Because of the importance of GABA as an inhibitory neurotransmitter and as an immunomodulatory factor that regulates immune cell function and decreases inflammation, we previously genetically engineered a Lactococcus lactis strain to overproduce GABA (P8s-GAD L. lactis). In this work, C57BL/6 Envigo (Env) and Jackson Laboratories (Jax) mice were administered 5 x 108 colony-forming units (CFU) of P8s-GAD L. lactis, unmodified L. lactis (P-L. lactis), or Medium and EAE induction. The administration of P8s-GAD L. lactis was protective in Env mice, while no protective effects were observed in Jax mice. Using Jax mice, we compared the effects of the treatments on EAE, microbiota by 16S rRNA sequencing, and the mycobiota by IST sequencing, in mice treated with the bacterial strains, either with or without glutamic acid, using samples collected at the beginning of the treatment, EAE induction, 14 days after, and at the end of the experiment (day 28). To test whether the bacterial treatments affected the CNS, a whole-brain proteomics analysis was performed in mice at day 28. Despite the lack of protective effects, the treatment with GABA-producing L. lactis (P8s-GAD L. lactis) in EAE resulted in changes to the gut microbiome. Additionally, the proteomics analysis revealed a change in protein profile when mice were treated with P8s-GAD L. lactis. Our study highlights the importance of controlling the source of mice for probiotic and microbiota research in the context of experimental models of immune-mediated diseases. Methods Animals and housing conditions Ten-week-old female C57BL6 NHsd mice (Envigo RMS, Inc., Indianapolis, IN, USA) (RRID:MGI:2161078; Env) and C57BL6 Jax mice (Jackson Laboratories, Bar Harbor, ME, USA) (RRID: IMSR_JAX:000664; Jax) were randomly housed in groups of 5 in wire-top cages with a 12-h light/dark cycle (22 ± 1 ºC; 23-33% humidity). All animals weighed approximately 20 grams. Mice had free access to food and water, with all care and procedures following Eastern Washington University (EWU) and Boise State University (BSU) institutional policies for animal well-being and health. Teklad 22/5 rodent diet (Envigo RMS) was used in Env mice, while Laboratory Rodent Diet 5001 (LabDiet, Inc., Richmond, IN) was used in experiments with Jax mice at BSU. Mice had free access to food and water and were housed and handled ethically, following all care and procedures according to an approved Institutional Animal Care and Use Committee (IACUC) for animal well-being and health under IACUC protocols 2018-11-01 (EWU) and AC22-018 (BSU). Whether Env, Jax, or both were used is depicted in the figure legends. In all EAE experiments, each experimental group consisted of 10 mice (n=10). The sample size for the EAE experiments was selected based on a power analysis performed using previous observed findings by ANOVA on day 25 vs. untreated (p = 0.045 for treatment vs. control) on a desired power of 90% and a significance level of 0.05 (1). EAE induction and clinical score evaluation Animals were given one week to acclimate to the housing environments. EAE was induced using an EAE induction kit (Hooke KitTM EK-2110, Hooke Laboratories, Lawrence, MA). The kits use MOG35-55 in emulsion with complete Freund’s adjuvant (CFA) and Pertussis toxin (PTX) in a glycerol buffer. On day 0 of EAE induction, the MOG35-55-CFA emulsion was diluted with PTX in phosphate-buffered saline and injected subcutaneously. PTX toxin injection was repeated the following day (day +1). Mice were scored under blinded observations as described by us and others: 0 – no detectable signs of EAE, 0.5 – distal limp tail, 1.0 – complete limp tail, 1.5 – limp tail and hind limb weakness, 2.0 – unilateral partial hind limb paralysis, 2.5 – bilateral partial hind limb paralysis, 3.0 – complete bilateral hind limb paralysis, 3.5 - complete bilateral hind limb paralysis and partial front limb paralysis, 4.0 – quadriplegia, 5 – dead animal. Attrition: mice that were found dead received a clinical score of 5. Losses not caused by EAE induction were not observed. Bacterial growth, dosing preparations, and treatments Lactococcus lactis-P8s-GAD and Lactococcus lactis-P (unmodified strain, carrying plasmid without engineered construct) strains were cultured in M17 broth (BD Difco, USA) or agar (Thermo Scientific Chemicals, USA) with 0.5% glucose (Sigma-Aldrich, USA) and 5 mg mL-1 erythromycin (GM17 erm media) at 30 °C without aeration overnight (20). After culturing for 18 hr, the strains were diluted to an OD600 of 0.2 in GM17 or GM17 erm and incubated at 30 °C for three hours. At the time of sample collection, the bacterial cell number was determined using a spectrophotometer (OD600). Viable plate count experiments were used previously to correlate cell number with spectrophotometer absorbance at OD600 (20). Mice were treated five times (consecutive)/week with oral gavages with 5 x 108 CFU/mouse of L. lactis strain P-L. lLactis(P) or P8s-GAD L. lactis (P8s) suspended in sterile saline. Depending on the experiments, the treatments were done in Env or Jax mice. Dosing timing varied within studies, and treatments started on the day of EAE induction (Day 0) or seven days before EAE induction (Day minus 7 (Day -7). The treatments continued throughout the experiments until the end (Day 28). The dosing strategies used are depicted in figures and corresponding legends. For some experiments, the GM17 medium used for the bacterial administration was supplemented with 200 mM glutamic acid hydrochloride (Sigma, reference number G2128). One experiment was performed with drinking water supplemented with 200 mM glutamic acid hydrochloride, which was provided to the animals ad libitum. Stool sample collections and 16S ribosomal RNA (rRNA) and ITS sequencing Stool samples were collected in sterile tubes on days -7, 0, 14, and 28 post-EAE induction (dpi) and stored at -80°C before use. Seven animals per group (n=7) were used for microbiome and mycobiome stool analysis. Samples were sent to the University of Wyoming for DNA extraction and sequencing. Qiagen DNA stool extraction kits were used for DNA isolation. DNA aliquots (1 ng/ml DNA) were analyzed by PCR using primers specific to the variable regions 3 and 4 (V3-V4) of the prokaryotic 16S rRNA gene and the fungal internal transcribed spacer (ITS) gene. Library preparation and V3-V4 amplicon sequencing were performed on the Illumina MiSeq (RRID: SCR_016379) platform. A modified protocol with the Nextera XT kit was used for library preparation, and sequencing was performed using MiSeq V2 (2x250bp) chemistry. The sequencing protocol involved a combined amplification with forward and indexed-reverse primers. After sequencing, the microbiome data were analyzed using Nephele (RRID: SCR_016595), a cloud-based web application from the Office of Cyber Infrastructure and Computational Biology (OCICB), National Institute of Allergy and Infectious Diseases (http://nephele.niaid.nih.gov/; 2016). QIIME (RRID: SCR_008249) was used for analysis (26), and RStudio (RRID: SCR_000432) for statistical analysis (27). Reads that were demultiplexed were clustered into OTUs using an open reference approach by comparison with the SILVA_99 database, allowing sequences to cluster at 99% similarity. Analyses included the identification of chimeras and removal using uchime. The abundance of each taxon was analyzed using the phyloseq package in RStudio (18). The compositional heterogeneity of each sample's microbial community at every timepoint and each taxonomic level was visualized using PCoA scaling and the ordinate function in the phyloseq package and using the Bray-Curtis dissimilarity index calculated using the metaMDS function in the vegan package. Whole-brain proteomics Whole brains were collected at the end of the experiment (day 28) and store at -80°C before use. The samples were processed at the IDeA National Resource for Quantitative Proteomics, Little Rock, AR. Total protein from each sample was reduced, alkylated, and purified by chloroform/methanol extraction prior to digestion with sequencing-grade modified porcine trypsin (Promega, Madison, WI). Tryptic peptides were then separated by a reverse phase Ion-Opticks-TS analytical column (25 cm x 75 mm with 1.7 mm C18 resin) supported by an EASY-Spray nano-source and stabilized with a Heater THOR Controller (Ion-Opticks, Fitzroy, Australia) at 60°C. Peptides were trapped and eluted from a (PepMap Neo, 300um x 5mm Trap) using a Vanquish Neo UHPLC nano system (Thermo Scientific, Waltham, MA), which kept the samples at 11°C before injection. Peptides were eluted at a flow rate of 0.350uL/min using a 35 min gradient from 98% Buffer A (0.1% formic acid, 0.5% acetonitrile in water):2% Buffer B (80% acetonitrile, 20% water, 0.1% formic acid) to 94.5:5.5 at 0.1 minutes to 56:44 at 27.1 minutes followed by a column wash of 45:55 at 29.7 minutes to 1:99 at 35 minutes followed by equilibration back to 98:2. Eluted peptides were ionized by electrospray (2.5 kV) followed by mass spectrometric analysis on an Orbitrap Astral mass spectrometer (Thermo Scientific). Precursor spectra were acquired from 380-980 Th, 240,000 resolution, normalized AGC target 200%, maximum injection time 3 ms. DIA acquisition on the Orbitrap Astral was configured to acquire 199, 3 Th window from 380-980 Th, 25% HCD Collision Energy, normalized AGC target 100%, maximum injection time 3 ms. Fragment (MS2) scan range from 150-2000 Th with an RF Lens (%) set to 40. Data Analysis: Following data acquisition, data were searched using Spectronaut (Biognosys, version 20.1; Biognosys, Zurich, Switzerland) against the UniProt Mus musculus database (3rd version of 2025) using the directDIA method with an identification precursor and protein q-value cutoff of 1%, generate decoys set to true, the protein inference workflow set to maxLFQ, inference algorithm set to IDPicker, quantity level set to MS2, cross-run normalization set to false, and the protein grouping quantification set to median peptide and precursor quantity. Fixed Modifications were set to Carbamidomethyl (C), and variable modifications were set to Acetyl (Protein N-term), Oxidation (M). Protein MS2 intensity values were assessed for quality using ProteiNorm (28). Data normalization and statistical analysis are shown below. Statistical analysis For EAE clinical scores and body weight changes, group differences were estimated using repeated measures and two factors (independent variables time and treatment) mixed-effect ANOVA followed by Tukey’s multiple comparison posthoc test, with p < 0.05 considered statistically different. Group differences in disease onset and severity scores were evaluated using non-parametric Kruskal-Wallis followed by Dunn’s multiple comparisons tests, with p < 0.05 considered statistically different. The R software package was used for the statistical analysis of the microbiota and mycobiota. ADONIS was used to determine group differences in gut microbiota and mycobiota compositions estimated by PCoA analysis (adjusted p values provided), and differences in microbiota alpha diversity and Firmicutes:Bacteroidetes (F/B) ratios were evaluated using non-parametric Kruskal-Wallis followed by Dunn’s multiple comparisons testing, with p < 0.05 considered statistically different. The statistical analysis of the microbiota and mycobiota by ADONIS was complemented with a two-independent variables two-way ANOVA followed by Tukey’s test to assess group differences at the phylum, family, and genus level, obtaining adjusted p-values when comparing all mice and timepoints combined for each taxonomical level. The statistical analysis was performed with GraphPad Prism (RRID: SCR_002798), version 10. In the proteomics analysis, the data were normalized using Variance Stabilizing Normalization (VSN) (29) and analyzed using proteoDA to perform statistical analysis using Linear Models for Microarray Data (limma) with empirical Bayes (eBayes) smoothing to the standard errors (30, 31). Proteins with an FDR-adjusted p-value < 0.05 and a fold change > 2 were considered significant.