Considering that this project forms part of a larger study that consists of multiple (inter-disciplinary) aims, participant recruitment and selection information [21], additional clinical and physiological measurements [21], as well as complementary metabolic investigations [4, 14] based on sub-divisions of the current cohort, have already been published and may be referred to for additional information not pertinent to this investigation.
Participants
The participants included in this investigation were selected at random and participation was completely voluntary. Prior to the marathon, all participants were required to complete a health and dietary questionnaire (with an additional menstrual cycle questionnaire for female participants), of which individuals receiving or using any anti-inflammatory treatments, chronic medication, as well as those with any food allergies, cardiovascular complications or musculoskeletal disorders and injuries were excluded from the study. All athletes were instructed to refrain from exercising and/or using any alternative recovery modalities (heat, cryotherapy, inflammatory drugs, antioxidant vitamins, compression garments, etc.) during the recovery period of this investigation. Withal, the use of anti-bacterial mouth wash was prohibited as a means of conserving the proposed bacterial nitrate-nitrite conversion of beetroot juice in the oral cavity. All the participants gave written and informed consent before the commencement of any analysis. An overview of the participant characteristics/demographics is presented in Table 1.
Clinical samples and supplementation
Blood samples of 31 athletes (19 males; 12 females) were obtained (antecubital venesection) before (P0), after (P1), as well as 24 h (P2) and 48 h (P3) after completing the Druridge Bay Marathon (Northumberland, UK) [21]. During the two consecutive days following the race, athletes received either beetroot juice (n = 15 athletes; 9 males and 6 females) or isocaloric placebo supplements (n = 16 athletes; 10 males and 6 females). Placebo samples consisted of a maltodextrin, protein powder, and fruit squash mixture, with a similar macro-nutrient content to that of the beetroot juice supplement (containing approximately 400 mg of phenolic compounds and 194 mg of the pigment, betanin), as described by Clifford, Allerton [21] and indicated in Table S2. These supplements were placed in containers that were indistinguishable in appearance and were consumed as follows: 3 × 250 ml supplements on the day of the marathon (immediately after P1 sampling, ±3 h post-race, and at 20:00), 3 × 250 ml supplements on the first day after the marathon (upon waking-up, with lunch, and with supper), and 1 × 250 ml supplement upon waking on the second day post-marathon. Participant groups were matched according to predicted marathon finishing times and did not significantly differ in terms of recorded dietary intake (determined using Nutritics dietary analysis software), or the number of males/females per group [21]. P0 samples were collected at participant-convenient times preceding the race and patients were required to be in a hydrated yet fasted (for at least 4 h) state, whereas P1 samples were acquired within 30 min after completing the race, thus dictating the approximate time of P2 and P3 collection. All the blood samples were collected in 10 mL vacutainer vials and placed on ice before being transported to the Northumbria University (Newcastle upon Tyne, UK), Faculty of Health and Life Sciences (Department of Sport, Exercise and Rehabilitation) laboratory, for further processing. Initial sample processing included clotting at room temperature for 30 min, followed by a 10 min centrifugation step (3000 g). The serum (supernatant) was then extracted, immediately frozen (− 80 °C), and transported on dry ice to the North-West University, Human Metabolomics: Laboratory of Infectious and Acquired Diseases for metabolomics analyses. All samples were stored at − 80 °C until analysis commenced. A schematic representation of the larger metabolic study design is presented in Fig. S1 of this investigation and the Supplementary material of Stander, Luies [14].
Total metabolome extraction and derivatisation
As previously described [4, 14], all samples, including pooled quality control samples (containing 50 μl of each sample), were subjected to an in-house total metabolome extraction (SOP number: HM-MET-056) and traditional TMCS derivatisation before being analysed. To summarise; 50 μl of internal standard (3-phenylbutyrate; 0.45 μg/ml), dissolved in a chloroform:methanol:milliQ water (1:3:1) solution, was added to smaller aliquots (50 μl) of the samples. While on ice, 300 μl of ice-cold acetonitrile was added to the aliquots, whereupon it was mixed for 2 min (REAX D-91126 vortex; Heidolph Instruments GmbH & Co.KG, Schwabach, Germany), and centrifuged for 10 min at 4000 rpm. The supernatants of the samples were then extracted, transferred to glass GC-MS vials, placed in a heating block set to 40 °C, and dried under a stream of nitrogen gas for approximately 45 min. Using a Hamilton syringe, 25 μl of methoxamine hydrochloride (dissolved in 15 mg/ml pyridine) was added to each vial, which proceeded to incubate for 90 min at 50 °C. Finally, samples were derivatised with 40 μl BSTFA (enriched with 1% TMCS) for 60 min at 60 °C, before being transferred to a new GC vial containing a vial insert.
GCxGC-TOFMS analysis and processing
The randomised samples were injected (1 μl; 1:3 split ratio) into the Pegasus 4D GCxGC-TOFMS system (LECO Africa (Pty) Ltd., Johannesburg, South Africa), using the Gerstel auto-sampler (Gerstel GmbH and co. KG, Mülheim van der Ruhr, Germany). The carrier gas (purified helium) was set to flow at a constant rate of 1 ml/min, while the injector temperature was held at 270 °C. The primary oven, containing a Restek Rxi-5MS capillary column (30 m; 0.25 μm diameter and 0.25 μm film thickness), was programmed with an initial temperature of 70 °C, which incrementally (4 °C/min) increased until a final temperature of 300 °C was reached (maintained for 2 min). The secondary oven, containing a Restek Rxi-17 capillary column (1 m; 0.25 μm diameter and 0.25 μm film thickness), was set at 85 °C, which increased with 4.5 °C/min until a final temperature of 300 °C was reached (maintained for 2 min), while the thermal modulator pulsed cold and hot streams nitrogen gas every 3 s for a duration of 0.5 s. The mass spectra (ms) of the first 400 s of each run was discarded (regarded as solvent delay), whereafter ms of ions (50–800 m/z) were acquired at 200 ms/s. The transfer line and ion source were held at 270 °C and 220 °C, respectively, with a detector voltage of 1600 V and filament bias of − 70 eV. The data generated from the GCxGC-TOFMS was processed (deconvolution, peak alignment and identification) using the ChromaTOF Software (LECO Corporation), as described by Stander, Luies [4].
Data processing and statistical analyses
The dataset obtained was normalised relative to the internal standard, and plasticizers, analytical contaminants, and column-related compounds were removed. Hereafter, a 50% zero value filter, zero value replacement (with random values below the detection limit), 50% quality control coefficient of variation (QC-CV) filter (retaining metabolites with a CV ≤ 50%), log transformation, and auto-scaling were performed.
Following these clean-up steps, the data was subjected to a variety of multivariate and univariate statistical methods using MATLAB [32] (in conjunction with a PLS [33] toolbox), as a means of selecting those metabolites pertinent to the aim of this investigation. To comprehensively address the aim of this investigation, multiple statistical objectives, and therefore comparisons, were required. In summary, paired statistical analysis of the beetroot juice-ingesting cohort was performed to confirm whether this cohort indeed recovered to a pre-marathon-related state within 48 h (statistical objective A), as has already been confirmed for the placebo group [14]. Hence, the P2 and P3 serum metabolite profiles of the beetroot-ingesting cohort were respectively compared to that of the P0 profile to identify any differentiating metabolites which would oppose metabolic recovery. Multivariate analyses included multilevel (ML) principal component analyses (PCA) and ML-partial squares discriminant analysis (ML-PLS-DA), while univariate analyses consisted of a paired t-test, and effect size tests, to assess statistical and practical significance respectively. To control for false discovery rates (FDR) associated with large scale multiple testing, as in the case of untargeted metabolomics datasets, t-test p-values were adjusted using the Benjamini-Hochberg procedure (limiting FDRs to 5%). All metabolites (P0 vs P2 and P0 vs P3) with a BH adjusted p-value ≤0.05 and an effect size d-value ≥0.5 were deemed significant.
To determine whether or not beetroot juice ingestion expedites the metabolic recovery trend of athletes within 48 h post-marathon, unpaired statistical analyses of the beetroot cohort vs placebo cohort were performed (statistical objective B). Foremost, the metabolic progression for both treatment groups over time was compared by using statistical models that accounted for the entire experimental design and dependencies between measures, i.e. a two-way repeated-measures analysis of variance (RM ANOVA), and an unfolded PCA. The latter transforms a three-dimensional tensor into a two-dimensional matrix (Fig. S2), thus allowing for PCA [34]. To supplement these comprehensive statistical methods, recovery time-point-specific inter-cohort comparisons were performed to assess day-specific variation that may be apparent between the cohorts. For this, multivariate statistical methods included PCA, PLS-DA, whilst univariate methods comprised an independent samples t-test (FDR’s controlled to 5% according to the BH procedure), and independent effect size tests based on Cohen’s d-values. Here, variables were selected based on a BH adjusted p-value ≤0.05 or a Cohen’s d-value ≥1.0, thus considering both statistical and practical significance. As opposed to the equal statistical and practical relevance of marker selection in objective A, this objective’s selection was more stringent on practical relevance to capture slight differences that may be of practical importance.
Since the cohort sizes of this investigation are relatively small, multivariate models are less readily validated and could therefore only be used to visualise trends and variation, whilst univariate models, better equipped to avoid false discoveries, were employed for variable selection in both statistical objectives.
Source link