The microbiota regulates murine inflammatory responses to toxin-induced CNS demyelination but has minimal impact on remyelination

Significance People with multiple sclerosis have a microbiota distinct from healthy controls, and there is growing interest in how these differences might contribute to the onset and progression of CNS autoimmunity. However, the impact that the microbiota may also have on the endogenous regeneration of myelin—remyelination—has not yet been explored. Here we show that inflammatory responses during remyelination depend upon the microbiota, being modulated by antibiotics or probiotics or in germ-free mice. In contrast, these interventions had minimal impact on the activity of oligodendrocyte progenitor cells, with only supratherapeutic doses of antibiotics having an inhibitory effect. Our results suggest that endogenous CNS remyelination is largely resilient to interventions that modify the microbiota.

Fluorescently immunolabelled sections were imaged using a Leica-SP5 confocal microscope with LAS software (Leica Microsystems, Wetzlar, Germany). Depending on the staining, either Z-stacks spanning the section widths were obtained and combined, or a single planar image was taken, and either a 20x or 40x lens was used.

Histological analysis of remyelination
To produce semi-thin resin sections for toluidine blue staining, glutaraldehyde-fixed tissue was dissected into pieces of maximum 1 mm thickness and stained with 2% osmium tetroxide overnight at 4°C. Samples were then processed into resin (TAAB Laboratories Equipment Ltd., Aldermaston, UK), after which 0.75μm sections were cut using a microtome (Leica RM 2065) and stained with 1% toluidine blue.
These samples were further analysed by electron microscopy. 0.75μm sections were stained with aqueous 4% uranylacetate and lead citrate and visualised on a Tecnai G2 80-200keV transmission electron microscope. A minimum of five micrographs per animal were captured at 5000x from the splenium of the corpus callosum, close to the midline.

Microglia isolation and culture
Microglia were isolated from 3-month old adult C57BL/6 mice using a Magnetic-Activated Cell Sorting (MACS) protocol (Fig. 3A). Mice were euthanised by CO2 overdose and posterior cervical dislocation, and each brain provided cells for one replicate. Tissue was diced into small pieces and incubated at 37°C for 30 minutes in a dissociation solution, consisting of 34 U/ml papain (Worthington, Lakewood, NJ) and 20 μg/ml DNAse (Gibco, Thermo Fisher Scientific) in HALF (Hibernate-A equivalent, made in house). After this, tissue was triturated with a fire-polished glass pipette, passed through a 70μm cell strainer (Millipore) and centrifuged for 20 minutes at 800g in 22.5% Percoll (GE Healthcare, Little Chalfont, UK). The pellet, containing single cells, was labelled with magnetic bead-conjugated antibodies for CD11b (Miltenyi Biotech, Woking, UK), and CD11b + microglia were eluted by MACS according the manufacturer's instructions. Microglia were cultured at 2x10 4 cells per well of a poly-d-lysine-coated 96-well microplate (Corning, NY), in DMEM/F12 (Gibco) supplemented with 10% foetal bovine serum (FBS, Biosera, Heathfield, UK), 2% B27, 500μM N-acetylcysteine and 1% penicillin-streptomycin. After 48 hours, media was changed to macrophage serum-free medium (Thermo Fisher Scientific), containing the antibiotic treatments. Following a further 48 hours, 10μg/ml myelin debris was added to each well for 4 hours. Myelin debris had been isolated from 2 to 3-month old C57BL/6 mice by discontinuous sucrose gradient centrifugation, as previously described (3,4). After this incubation, un-internalised debris was removed by washing with cold PBS and cells were fixed with 4% PFA.

Antibiotic treatments in vitro
To apply antibiotic treatments (all Sigma-Aldrich) at doses approximating those of in vivo exposure, steady state plasma concentrations (CSS(P)) were estimated for each antibiotic administered to mice in their drinking water (Fig. S2A). These estimations were based on literature values of oral bioavailability (F), clearance (CL), and area-under-the-curve ratio of cerebrospinal fluid (CSF) to plasma (AUCCSF/AUCP), for ampicillin and sulbactam (5-7), ciprofloxacin (8-10) and metronidazole (10)(11)(12). Daily water consumption was assumed to be 4 ml/day/mouse.

Immunocytochemistry
Fixed cells in 96-well plates were blocked for 1 hour with 5% NDS and 0.1% triton in PBS. For the cell proliferation assay, EdU was labelled at this point using the Click-iT EdU Alexa Fluor 647 Imaging Kit (Thermo Fisher Scientific). Cells were then incubated with primary antibodies diluted in blocking solution overnight at 4°C. Primary antibodies: mouse anti-CNPase 1:500 (C5922, Sigma-Aldrich), rabbit anti-Iba1 1:1000 (019-10741, Wako), rat anti-MBP 1:500 (MAB384, Millipore), rabbit anti-Olig2 1:500 (AB9610, Millipore). Cells were then washed three times with PBS and incubated with relevant fluorophoreconjugated secondary antibodies (see "Immunohistochemistry") diluted in blocking solution for 1 hour at RT and protected from light. Nucleic acids were stained by 10 minutes incubation with 1μg/ml Hoechst, followed by three more PBS washes. Images were acquired automatically using a Cell Insight CX5 (Thermo Fisher Scientific).

Faecal polymerase chain reaction (PCR)
To determine the microbial load of antibiotics-treated mice, DNA from faecal pellets was extracted as described previously (1), quantified using QuantiT PicoGreen reagent (Invitrogen) and adjusted to 1 ng/μl. Copy numbers of the 16S rRNA gene were quantified by quantitative RT-PCR using generic eubacterial primers (Tib MolBiol, Berlin, Germany), and expressed per ng of total DNA.
To confirm microbial status of GF and ex-GF mice, DNA from faecal pellets was isolated using a QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. 1μl eluted DNA was added to 24μl PCR SuperMix (Thermo Fisher Scientific) with 200nM UniF/R primer pairs (13), which recognise a 147bp conserved region of bacterial 16S rRNA. The conditions for the PCR were 95°C for 5 minutes, then 25 cycles of 95°C for 30s, 52°C for 30s and 72°C for 45s, and finally 72°C for 7 minutes. The PCR products were then separated on a 1% agarose gel and visualised under UV light.

Detection of faecal / serum metabolites using GC-MS
Fresh faecal pellets were collected and stored in sterile Eppendorf tubes at -80°C. Serum samples were obtained from 100-200μl left ventricular blood collected prior to perfusion. The blood samples were incubated in sterile Eppendorf tubes for 1 hour at RT to allow coagulation, then separated by centrifugation for 15 minutes at 1500g. Serum was collected and stored at -80°C.
Short chain fatty acids (SCFAs) were extracted using a modified Bligh and Dyer method (14). In short, 15-25mg of faeces or 20µl of serum was transferred to a pre-chilled plastic tube and extracted with ice-cold 2:2:1 methanol:chloroform:water containing internal standard, following vigorous mixing, sonication and centrifugation (16,000g, 20 minutes). For serum samples, 100µl of the aqueous phase was transferred to pre-chilled glass tubes. Faecal samples were extracted twice, and a combined total of 300µl aqueous phase was transferred to pre-chilled glass tubes. Aqueous phases were dried down under nitrogen at 4°C, and derivatised as described previously (15). SCFAs were measured by gas chromatography-mass spectrometry (GC-MS), on a Trace GC Ultra coupled to a Trace DSQ II mass spectrometer (Thermo Fisher Scientific). Derivatised samples were diluted 1:1 with hexane, and 2µl was injected onto a 50m x 0.25mm (5% phenyl-arylene, 95% dimethylsiloxane) column with a 0.25µm ZB-5MS stationary phase (Phenomenex, Macclesfield, UK). Full-scan spectra were collected at three scans per second over a range of 50 to 650 m/z. Data processing was carried out using Xcalibur (version 2.2, Thermo Fisher Scientific), and peaks were assigned based on the National Institute of Standards and Technology (USA) library. All solvents were of HPLC-grade or higher.

Image analysis
Cell counts were semi-automated, using a combination of Fiji, CellProfiler and CellProfiler Analyst software (16). Fiji was used to create maximum projections of Z-stacks acquired by confocal microscopy. The region of interest (ROI i.e. the lesion area) was manually defined by a blinded observer using a composite image, based on Hoechst + hypercellularity and non-specific background staining. Individual channels were then extracted and imported into CellProfiler, where the nuclear channel (Hoechst) was cropped to the predefined ROI, and nuclei were identified as primary objects. Other channels were normalised to the median background intensity to correct for variability in staining. A number of intensity features were measured for the nuclear and peri-nuclear regions of each cell in every channel, and this data was exported as a master database file. In CellProfiler Analyst, a training set of >50 cells was specified per group and used to train a classifier based on the feature set. For CD68 and P2ry12, which are expressed to different degrees by cells within the lesion, these were grouped into "high" and "low" classes of expression, and examples of cells from these training sets are provided (Fig. S1M, N). The training data was increased until the classifier gave consistently comparable results to manual counting in sample images. Finally, "per lesion" cell counts were extracted for each image using this classifier.
To quantify the area of a lesion occupied by myelin debris, images of tissue stained for dMBP were imported into a CellProfiler pipeline, which applied a threshold to each image determined by the background (median) staining. The area of the image above this threshold was considered positive for myelin debris and was expressed as a fraction of the total lesion area. For morphological analysis of microglia, Iba1 + microglia were identified in CellProfiler Analyst, and this population were then further analysed within CellProfiler to quantify morphological features of each cell's Iba1 + mask and skeleton.
To quantify remyelination from toluidine blue-stained resin sections in the probiotic study, slides of the 10 lesions (5 per group) were independently ranked by two experienced, blinded investigators (GG and CZ) according to the extent of remyelination. Ranks were based upon the proportion of each lesion with thin myelin sheaths characteristic of remyelination, compared to areas with persistently demyelinated axons. The assigned numerical order (1-10) was used for subsequent non-parametric statistical tests. To quantify remyelination from electron microscopy images, only axons contained entirely within each field were counted. For these axons, the internal and external diameter of the myelin sheath were then traced using a freehand selection tool in Fiji. The g-ratio was calculated as the ratio between the diameters of two circles with areas equal to the internal and external selections respectively. Axons with a circularity <0.7 or aspect ratio >2.5 were excluded from further analysis.

6
All statistical analysis was carried out using a Jupyter Notebook with Python 2. In vivo experiments contained the following numbers of biological replicates per group: antibiotics lysolecithin study: n = 4-6 mice, germ-free cuprizone study: n = 4-5 mice, probiotic lysolecithin study: n = 3-5 mice. These group sizes were chosen based on previous work and were thought to be sufficiently powered to detect meaningful differences in the OPC / inflammatory response to demyelination. For in vivo cell counts, generally 3-4 technical replicate sections were counted and averaged per biological replicate. For in vitro cell assays, 3-5 technical replicate wells were averaged for each of 4-5 biological replicate studies.
Data was tested for normality of residuals (Kolmogorov-Smirnov test) and homogeneity of variance (Levene's test). Data sets passing both of these criteria were compared by either unpaired Student's t-test (if 2 groups), or one-way ANOVA with Tukey HSD post hoc tests (if >2 groups). Non-parametric data was compared by Mann-Whitney U test (2 groups) or Kruskal-Wallis test with Dunn's post hoc test (>2 groups). For in vitro assays, treated conditions were compared to control conditions using a paired-samples t-test with the Holm-Bonferroni correction for multiple comparisons. For ranking analysis of remyelination, groups were compared using the Mann-Whitney U test. For all statistical tests, differences were considered significant if p<0.05, and the respective test is described in each figure legend.
In all bar plots, the height of the bar represents the group mean, with an error bar representing the standard error of the mean (SEM). In vivo data are overlaid with strip plots, in which a grey point represents the value for each individual animal.