blood mononuclear cells
2.9.3 MS data processing and statistical analysis
The raw mass spectra data files were first converted to mzXML format using MSConvert by ProteoWizard [314, 315]. This was necessary to generate common text-based data that can be used for further processing and analysis. MSConvert was used with the following settings: Output format ‘mzXML’, Extension ‘32-bit’, including ‘Write index’, ‘Use zlib compression’ and a ‘PeakPicking filter’ set from one. The XCMS package in the Bioconductor project written in R 3.1.2 (“EPA time-course experiment”) and 3.0.1 (“salmon diet experiment”) and was used for data pre-processing, quality assessment of MS runs, normalisation and metabolomic fingerprinting [310, 316]. Peaks of overlapping m/z bins were grouped, followed by retention time correction using loess smoothing, regrouping of peaks, and filling missed peaks. The codes applied in each experiment for data pre-processing are detailed in Appendix II. Un-normalised data were plotted as
heatmaps to detect effects related to sample injection order. The quality of the MS run was visually assessed by plotting total ion count intensities of quality control samples. Normalisation and log-transformation were applied to raw data. The final number of biological replicates per group is shown in Table 2.4 (EPA time-course experiment”) and Table 2.5 (“salmon diet experiment”). For statistical evaluation of the metabolite data, each urine collection time-point was analysed separately.
Permutational multivariate analysis of variance using distance matrices (permutational MANOVA; from the Vegan package in R) was performed to evaluate differences in the urinary metabolite fingerprints between the treatment groups. Data were further visualised by Principal Component Analysis (PCA) and Partial Least Squares- Discriminant Analysis (PLS-DA). Analysis was first focussed on the ions with the ten highest PLS-DA loadings (“discriminant ions”; those that drive the separation between treatment groups) and statistical significance was evaluated using an ANOVA after log- transformation, followed by an LSD post-hoc test using a 5% significance level.
However, the identification of the “discriminant ions” based on database and literature search was largely not possible in the “EPA time-course experiment”, therefore data was further evaluated by calculating p-values for all ionisation products using a Kruskal-Wallis test followed by multiple testing correction using the false discovery rate (FDR) estimation [317]. For metabolite identification, masses and fragmentation patterns of ions were searched against literature and databases, including METLIN [311] and HMDB [312].
In the “salmon diet experiment”, “discriminant ions” with putative identification were analysed on a Thermo LC-MS/MS system (Thermo Fisher Scientific, Massachusetts, USA) to confirm their identity. The system was fitted with an Accela 1250 quaternary UHPLC pump, a 2 μl injection loop, and coupled to a Q Exactive Quadrupole-
Table 2.4 Numbers of urine samples obtained from C57BL/6J mice and Il10-/-mice fed AIN-76A diets, either unmodified, or enriched with oleic acid (OA) or eicosapentaenoic acid (EPA), and included in metabolomic data analysis. Numbers of biological replicates were the same for positive and negative ionisation mode. Urine was collected at 7.1 (T1), 9 (T2), 10.1 (T3) and 12 (T4) weeks of age as part of the “EPA time-course experiment”.
Genotype Diet Scheduled sampling
age (weeks) T1 T2 T3 T4 C57BL/6J AIN-76A 12 5 4 4 4 OA 9 10 8 12 9 8 7 9 EPA 9 7 9 12 9 9 7 9 Il10-/- AIN-76A 12 10 9 7 6 OA 9 10 9 12 9 9 7 9 EPA 9 10 9 12 5 5 5 5
Table 2.5 Numbers of urine samples obtained from C57BL/6J mice and Il10-/- mice fed 30% salmon or 30% control diets and included in metabolomic data analysis. Numbers of biological replicates differed between positive and negative ionisation mode. Urine was collected at 6.2 (T1), 9 (T2), and 11.5 (T3) weeks of age as part of the “salmon diet experiment”.
Ionisation mode Genotype Diet T1 T2 T3
Positive C57BL/6J 30% control 4 4 5 30% salmon 2 4 7 Il10-/- 30% control 5 4 4 30% salmon 6 4 7 Negative C57BL/6J 30% control 5 4 5 30% salmon 4 4 7 Il10-/- 30% control 5 4 4 30% salmon 8 4 7
was applied: (1) 95% eluent A and 5% eluent B over 1.5 min; (2) 100% eluent B over 8.5 min with a 1-min hold at 100% eluent B; (3) linear gradient to 95% eluent A and 5% eluent B over 1 min with a 3-min hold at 95% eluent A and 5% eluent B. Mass spectra were evaluated with XCalibur software and fragmentation patterns compared to literature and databases.
2.10
Microbiomic analysis of caecum digesta
Microbiomics aims to capture genomic information from the microbiota present in an environment. With the invention of next-generation sequencing, for example, 454 pyrosequencing (Roche), SOLiD sequencing (Life Technologies) or Illumina sequencing (Illumina, Inc.), the exploration of the role of the microbiota in GIT homeostasis was enabled [318]. Roche’s 454 FLX titanium sequencing targets the 16S gene region of the ribosomal RNA (rRNA) [319], a highly conserved gene in bacteria that consists of nine hyper-variable regions (V1-V9). During sequencing, the nucleotide sequences are read to a certain length and the reads are assigned to specific groups of organisms, the operational taxonomic units (OTUs). The advantage of Roche’s 454 pyrosequencing compared to other methods is the capability to generate a relatively large amount of sequences with a good average read length (approximately 400 base pairs) despite relatively low costs [320]. Thus a good coverage of the microbial genome structure (average bacterial gene length approximately 950 base pairs [321]) can be achieved enabling more robust identification of the resident bacteria [322].
2.10.1
Method overview
The workflow for the microbiomic analysis from caecum digesta is illustrated in Figure 2.7. Briefly, DNA was extracted from caecum digesta and amplified using two sets of primers targeting the hyper-variable regions 1-3 (V123) and 4-6 (V456) of the 16S rRNA gene. Amplicons were purified and pyrosequenced on an FLX titanium sequencer (Roche, Connecticut, USA). Data was denoised and chimera-checked in Quantitative Insights Into Microbial Ecology (QIIME) software.