Clinical study protocol
Fecal samples for fecal microbiota transfer (FMT) were collected from one of the top 5 weight losers during a weight loss intervention study focusing on muscle mass regulation in postmenopausal women “Effects of negative energy balance on muscle mass regulation” (registered at https://clinicaltrials.gov, NCT01105143) at the Department of Endocrinology of the Charité- Universitätsmedizin, Berlin, Germany [33, 34].
This study was carried out in accordance with the recommendations of the International Conference on Harmonization Guidelines for Good Clinical Practice and the Declaration of Helsinki. The protocol of the study was approved by the local Ethics Committee of the Charité- Universitätsmedizin Berlin (EA2/050/10). All subjects gave written informed consent before participating in this study. Inclusion criteria comprised female gender, a BMI > 27 kg/m2, and postmenopausal status. Severe untreated medical, neurological, and psychiatric diseases within the last 5 years which may interfere with the planned interventions, such as unstable coronary heart disease, kidney and liver disease, systemic infections, endocrinological disorders, and hypertension (systolic blood pressure > 180 mm Hg, diastolic blood pressure > 110 mm Hg) were excluded by medical history assessment. Further exclusion criteria were changing dieting or smoking habits significantly in the last 2 months including a weight loss of 5 kg or more. Exclusion criteria for participants were also a history of medication, changes in smoking habits, or diets within the last 3 months, which may have significantly affected body weight. Participants with synthetic thyroid medications were not excluded if they were clinically euthyroid. A total of 80 overweight or obese female subjects were initially included in the study. Subjects were randomly assigned to the intervention and the control group, respectively.
The detailed study protocol has been reported elsewhere [11, 34, 35]. In brief, weight loss was induced by an established, standardized weight reduction program for 12 weeks in the intervention group. In the first 8 weeks of the 12-week calorie restriction period, weight loss was performed by a very-low-calorie diet (VLCD, 800 kcal/d) replacing all meals with a formula diet (Optifast 2®, Nestlé HealthCare Nutrition GmbH) [36]. Before and after these eight weeks of VLCD, stool samples were taken for downstream analyses and experiments from the one individual of the top 5 weight losers, who exhibited the strongest improvement in insulin sensitivity.
Mice and intervention protocol
Male GF C57BL/6J mice (15–24 weeks old, n = 33) were housed in gnotobiotic isolators with two to four mice per cage on a 12-h light-dark cycle. Mice were divided into three groups, each group in one individual isolator, respectively. Mice were colonized by FMT with a human gut microbiota, that was collected before (AdLib, n = 13) and after (CalRes, n = 9) a calorie-restricted dietary intervention and compared to an age-matched GF control group (n = 11). Animals were maintained on a chow diet (SSNIFF, V1534-300) which provides 9% kJ from fat, 33% kJ from proteins and 58% kJ from carbohydrates for 3 weeks after the FMT. Mice were sacrificed by cervical dislocation following anesthesia. This study was carried out in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and the Animal Welfare Act under the supervision of our institutional Animal Care and Use Committee. Animal protocols were conducted according to institutional ethical guidelines of the Charité-Universitätsmedizin, Berlin, Germany, and were approved by the Landesamt für Gesundheit und Soziales (approval number G 0085/17 LAGeSo Berlin, Germany) and comply with the ARRIVE guidelines.
Human FMT into germ-free mice
Human stool samples were collected before starting the intervention (AdLib) and after 8 weeks on the formula diet in the intervention group (CalRes). As previously described [34], volunteers were given a stool collection kit and were instructed to store collected stool samples in the freezer at − 20 °C until they were transported in a cooled container to our research unit. All samples were subsequently stored at − 80 °C. Each fecal sample was thawed and prepared in an anaerobic chamber. Stool samples were resuspended in phosphate-buffered saline (PBS) (1:10, Sigma-Aldrich), followed by centrifugation for 1 min at 500 rpm. The resulting preparation was externally sterilized, transferred into gnotobiotic isolators, and 200 μl were administered by oral gavage. Mice not receiving the FMT received oral gavages of autoclaved drinking water as a sham gavage. For the FMT, samples were chosen from an individual who was the top five weight losers of the program, who showed the strongest increase in insulin sensitivity during the weight loss phase and provided a complete set of stool samples (since not all individuals provided stool samples at each study visit).
Analysis of metabolic parameters
Mice were weighed every two days throughout the course of the experiment using a model EMB 200-2 scale (KERN & Sohn GmbH). Mice were fasted for 6 hours, and an oral gavage glucose tolerance test (OGTT) was performed. Fasted blood glucose levels were determined before a solution of glucose (10% Glucose, Braun) was administered (2 g glucose/kg body weight) by oral gavage. Subsequently, blood glucose levels were monitored at 15, 30, 60, and 120 min after glucose administration. Total fecal excretion and food consumption were obtained weekly from one representative cage per group. The energy density of the chow diet and fecal samples were determined using bomb calorimetry.
Bomb calorimetry
The energy content of the chow diet and mouse feces were analyzed using an Isoperibol Calorimeter 6200 instrument with a model 1108 oxygen bomb (Parr Instrument Co.), as described elsewhere. Briefly, the sample was pressed into a 1-g pellet and was placed into the bomb, which was filled with oxygen (3000 kPa), and placed in a bomb cylinder with 2000 ml distilled water. The increase in the temperature (∆T) of the surrounding water by the heat produced at combustion was measured. The energy content (E) of the pellet was calculated as follows:
The energy equivalent (W) specifies the energy required to raise the temperature of the surrounding water by 1 °C (W = [Calorie/°C]).
Fecal DNA extraction and sequencing
Fecal samples were collected from mice in the AdLib and the CalRes groups respectively at day 1, 3, 7, 14, and 21 after colonization with the human donor microbiota. DNA was extracted using the QIAamp Fast DNA Stool Mini Kit as detailed in the manufacturer’s protocol (Qiagen, USA). Library preparation for 16S rRNA gene sequencing was done according to the protocol of Illumina (USA), targeting the 16S V3 and V4 region, and sequenced on an Illumina MiSeq instrument with 2 × 300-bp v3 chemistry. Reads were demultiplexed using bcl2fastq (Illumina, San Diego, USA) and submitted to initial quality control by QCumber-2 (https://gitlab.com/RKIBioinformaticsPipelines/QCumber). In brief this pipeline trims sequencing adapters as well as low quality read ends and discards reads shorter than 50bp using Trimmomatic [37].
16S rRNA sequencing analyses
Demultiplexed reads were processed and denoised by DADA2 [38]. Taxonomy was assigned using the DADA2 implementation of the RDP classifier [39] using the DADA2-formatted training sets for SILVA138 (zenodo.org/record/3731176/files/silva_nr_v138_train_set.fa.gz). Species were assigned by exact matching against a reference (zenodo.org/record/3731176/files/silva_species_assignment_v138.fa.gz). A phylogenetic tree was constructed de novo via the DECIPHER and phangorn R packages. The optimized tree counts for a total of 1504 detected different amplicon sequence variants (ASVs), and taxonomy tables were converted into a phyloseq object [40] for further downstream analyses. No filtering except for unassigned taxa was performed prior to calculation of diversity metrics and ordination analysis of the complete dataset (1414 ASVs, 104 samples, day 1 to 21), while singletons were removed from differential abundance analysis in 20 samples at day 21 (284 ASVs).
Isolation of murine splenic immune cells
Spleen was homogenized, passed through 70 μm filters, washed, and subjected to red blood cell lysis (ACK lysing buffer, GIBCO). The red blood cell lysis was stopped by adding washing buffer (MaxPar Cell staining Buffer, Fluidigm), and the homogenate was then passed through 30 μm filters and washed again before final suspension in MaxPar Cell staining Buffer. The cell suspension was adjusted to 2x106 cells/500 μl.
Isolation of murine intrahepatic immune cells
Isolation of murine intrahepatic immune cells was performed as reported previously [41]. Briefly, whole liver was prepared by harvesting perfused liver lobes into 15 ml PBS. The liver was dissociated mechanically, followed by tissue digestion (0.5 mg/ml Collagenase Type IV (Worthington), 0.02 mg/ml DNAse I (Sigma-Aldrich), 2 % fetal calf serum, 0.6 % bovine serum albumin in HBSS) for 30 min at 37 °C in a rotation shaker (200 rpm). Hepatocytes were then pelleted (30 g, 1 min, room temperature (RT)), the supernatant was centrifuged (310 g, 4 min, 4 °C), and the pellet was resuspended in 30 % density gradient media (Percoll, Sigma-Aldrich) in HBSS followed by centrifugation (800 g, 30 min, RT) to enrich liver mononuclear cells. Following red blood cell lysis (ACK lysing buffer, GIBCO), the homogenate was washed again, and passed through 30-μm filters before final suspension in MaxPar Cell staining Buffer. The cell suspension was adjusted to 2 × 106 cells/500 μl.
Isolation of lamina propria mononuclear cells
Lamina propria mononuclear cells (LPMC) were isolated from colon sections (10 cm distal part) as described previously [42]. Briefly, the colon was opened longitudinally, cut into small pieces, and subsequently incubated twice with HBSS containing 1 mmol/l EDTA for 30 min, followed by enzymatic digestion (0.44 mg/ml Collagenase D and 20 μg/ml DNase I in RPMI, Sigma-Aldrich) for 60 min at 37 °C. After a filtration step through a 100μm filter, the LPMC were purified by 44/67% density gradient (Percoll™, GE Healthcare) centrifugation for 20 min at 600 × g.
Staining for mass cytometry
For barcoding, anti-CD45 antibodies were conjugated in house to 89Y, 147Sm, and 166Er. Up to six individual samples were stained with a combination of the different anti-CD45 antibodies for 30 min at 4 °C. Cells were then washed and pooled for surface staining. A total of 2 × 106 cells per sample were stained in a 96-deep-well plate with metal-conjugated antibodies (antibodies for mass cytometry as previously described [43]) for 30 min at RT. 0.5 mM cisplatin (Fluidigm) was added as live/dead cell marker for the last 10 min. Cells were then washed twice and the pellet was resuspended in 1 ml of nucleic acid Intercalator-Ir solution (12.5 nM Cell-ID Intercalator-Ir diluted in MaxPar Fix and Perm Buffer, Fluidigm), followed by incubation at RT for 30 min. Followed by two washing steps, the cells were resuspended in 100 μl formaldehyde-solution (1:10) and fixed overnight at 4 °C. The next day, cells were washed twice with ultrapure water and kept at 4 °C until mass cytometry measurement.
Mass cytometry measurement
Cells were analyzed using a CyTOF2 upgraded to Helios specifications, with software version 6.7.1014, using a narrow bore injector. The instrument was tuned according to the manufacturer’s instructions with tuning solution (Fluidigm) and measurement of EQ four element calibration beads (Fluidigm) containing 140/142Ce, 151/153Eu, 165Ho, and 175/176Lu served as a quality control for sensitivity and recovery. Directly prior to analysis, cells were resuspended in ddH2O, filtered through a 20-μm cell strainer (Celltrics, Sysmex), counted and adjusted to max. 8 × 105 cells/ml. EQ four element calibration beads were added at a final concentration of 1:10 v/v of the sample volume to be able to normalize the data to compensate for signal drift and day-to-day changes in instrument sensitivity. Samples were acquired with a flow rate of max. 300 events/s. The lower convolution threshold was set to 400, with noise reduction mode turned on and cell definition parameters set at event duration of 10-150 pushes (push = 13 μs). The resulting flow cytometry standard (FCS) files were normalized and randomized using the CyTOF software’s internal FCS-Processing module on the non-randomized (“original”) data. The default settings in the software were used with time interval normalization (100 s/minimum of 50 beads) and passport version 2. Intervals with less than 50 beads per 100 s were excluded from the resulting FCS file.
Mass cytometry data analysis
FCS files were compensated for signal spillover using CATALYST package and per channel intensity ranges were aligned between batches of measurements using the normalizeBatch function (cydar package). Cytobank [44] was used for manual debarcoding, gating of lymphocyte subsets and to perform viSNE on pre-gated subsets as described previously [43]. For each organ (colon, liver, spleen) semi-supervised population identification was conducted by graph-based clustering (R phenograph package) on total leukocytes as well as on events pre-gated for TCRab−CD19−NK1.1− and TCRab+ to better resolve the innate/myeloid compartment and T cells, respectively. Leaf nodes were merged into biologically relevant subsets by second-level hierarchical clustering while putting more weight on lineage-delineating markers.
Statistical analysis
The results are shown as the mean ± SD. A P-value of < 0.05 was considered significant. All analyses were performed using GraphPad Prism version 7 (GraphPad Software) and R version 3.4.0, available free online at https://www.r-project.org. Mathematical correction for multiple comparisons was made whenever indicated. Diversity metrics were calculated from microbial ASVs and principal coordinate analysis (PCoA) carried out using the phyloseq and vegan packages. Changes in alpha-diversity over time and between group were tested by ANOVA-type statistic using the nparLD package [45]. Bray-Curtis dissimilarity computed from variance-stabilized counts of the complete dataset was used to quantify and visualize compositional changes between microbial communities by PCoA. Global differences between groups and changes over time were tested on the dissimilarity matrix by permutational analysis of variance using adonis with 9999 replications. Differential abundance analysis of ASVs between CalRes and AdLib groups at day 21 was carried out using DESeq2. Abundances of significant ASVs (FDR < 0.01) were visualized by heatmap along with log2-fold change (LFC) values.
For statistical analysis of cell population abundances, we fitted a generalized linear mixed-effects model (GLMM) for each population using the lme4 package as previously described [46]. To take into account the day-to-day (batch) variability of the mass cytometry runs, we included batch as fixed effect in the models and all quantitative data presented are shown after batch-adjustment. P-values resulting from differential abundance testing were adjusted using the Benjamini-Hochberg procedure and an FDR-cutoff of 5% across all clusters/subsets and between-group comparisons. To estimate differences between manually gated immune cell subsets, one-way ANOVA followed by Tukey’s tests were applied.
To investigate associations between microbial composition and immune cell populations, we performed sparse canonical correlation analysis. For each organ, a semiparametric correlation matrix was estimated based on the latent Gaussian copula model using the mixedCCA package with selection of canonical correlation vectors using L1-penalization (lasso) and the Bayesian information criterion for unknown error variance [47]. Sparse canonical covariates were computed by matrix multiplication of the ranked variables of each dataset with its canonical vector. Each latent correlation matrix was ordered using the projection on its first principal component to visualize the cross-correlation structures and to additionally highlight the top ten taxa that either positively or negatively associate with the immunological datasets. All heatmaps and circular correlation plots were generated using the ComplexHeatmap and circlize.
The body weight and glucose time course data were analyzed by two-way repeated measures ANOVA followed by post-hoc (Bonferroni) test.