The experiment was conducted at the INRAE dairy research farm (La boire, Marcenat, https://doi.org/10.15454/1.5572318050509348E12). Procedures were evaluated and approved by the French Ministry of Education and Research (APAFIS #4062-2015043014541577 v5), and carried out in accordance with French and European guidelines and regulations for animal experimentation. Information provided in the manuscript complies with the essential recommendations for reporting of the ARRIVE guidelines.
Animals, diets and experimental design
Eighteen female Holstein (n = 12), Montbéliarde (n = 4) and Holstein x Montbéliarde (n = 2) calves (42.07 ± 3.85 kg birth weight) were enrolled in the study at birth. Calves were kept with their dam for a few hours but systematically received 2 L of warm colostrum of good quality (≥ 50 g IgG per L) that is conserved at − 20 °C until use. Calves were individually housed for the first week of life and bottle fed 3 L of milk twice daily (0700 h and 1800 h). After the first week, calves were group housed according to treatment with ad libitum access to water and hay. Calves were fed up to 8 L of milk per day through the use of an automated milking system (De Laval, Sursee, Switzerland). Similarly, calves had access to calf starter (STARTIVO, Centraliment, Aurillac, France) from four weeks of age with a maximum daily allowance of 2 kg in the pre-weaning period. In the immediate post-weaning period, calves had access to 2 kg of GENIE ELEVAGE (Centraliment, Aurillac, France). Chemical composition of dietary ingredients is presented in Table S1.
Calves were randomly assigned at birth to either a treatment (3-NOP, 3 mg 3-NOP/kg BW, n = 10 up to week 23, a heifer was removed from the herd due to infertility as a result of being born a twin, i.e., a free-martin, n = 9 at week 60) or control (CONT, placebo premix containing SiO2 and propylene glycol only, n = 8, nine calves were recruited but one calf died early in the study) group, such that breed distribution and birthweight were balanced across groups. The 3-NOP supplement contained 10% 3-NOP diluted in propylene glycol and adsorbed on SiO2, such that 30 mg of the supplement was fed per kilogram of body weight to achieve the above target dose of 3-NOP. The 3-NOP and placebo were mixed with water (300 mg/mL water) and administered daily via an oral gavage approximately 2 h after feeding. Calves were treated daily from the day of birth, following consumption of colostrum, until 14 weeks of age. All calves were weaned at week 11 using the step-down method over two weeks. After weaning, all calves were group housed in a single pen to replicate normal production practices.
Calves were weighed weekly. Daily individual milk and concentrate intakes prior to weaning were recorded using automated feeders. Total group intake of hay and concentrate, and all refusals in the post-weaning period were recorded twice weekly.
Sampling
All calves were sampled for rumen fluid and faecal content at 1, 4, 11, 14, 23 and 60 weeks of life. Sampling at week 11 was conducted immediately prior to weaning and sampling at week 14 was conducted just prior to cessation of the treatment. Samples of rumen liquid were obtained via oesophageal tubing at least 2 h after feeding. Aliquots (1 mL) of rumen liquid were immediately frozen in liquid nitrogen and stored at − 80 °C until DNA extraction. Additional rumen liquid aliquots were taken for analysis of volatile fatty acids and ammonia as previously described25,26. At week 60, 3 mL of ruminal fluid was added to 3 mL of methyl green formalin saline (MFS) solution (35 mL/L formaldehyde, 0.14 mM NaCl, and 0.92 mM methyl green) and stored in the dark at room temperature until protozoa were counted. At each sampling period, calves were rectally finger-stimulated with sterile-gloved hand to facilitate the collection of a faecal sample, which was immediately frozen in liquid nitrogen and stored at − 80 °C until DNA extraction. Blood samples were taken via jugular venepuncture into a heparin tube at week 11, 14 and 23 for metabolic analysis. Blood was immediately centrifuged at 1500×g for 10 min at 4 °C. Plasma was stored at − 80 °C until analysis.
Methane measurements
Methane emissions were recorded using the GreenFeed system (C-Lock Inc., Rapid City, South Dakota, USA) during two time periods. Firstly, from weaning (week 11) until week 23, one GreenFeed system was programmed using C-Lock Inc. software to deliver a maximum of six rotations of a feed dispensing cup, delivering ~ 6 g of pellet concentrate GENIE ELEVAGE (as fed) per rotation, with intervals of 30 s between each rotation, so that 36 g of pellet was delivered during each visit. During the second phase of measurement when heifers were 57 to 60 weeks of age, two GreenFeed systems were utilised with software programmed to deliver a maximum of six rotations of a feed dispensing cup, delivering approximately 45 g of pellet (as fed) per rotation, with intervals of 30 s between each rotation, so that 270 g of pellet was delivered during each visit. During the second measurement period, calves were separated according to treatment group and allocated to one of two GreenFeed systems. An adaptation period of one week preceded a 4-week experimental recording period. After two weeks, calves were rotated into the alternate pen to eliminate any possible biases between the two GreenFeed systems. During all measurement periods, a minimum of 3 h was required between visits. The calf starter pellets described above were used as an enticement. Recorded methane measurements were included if the total time spent in the feeder was > 3 min with calves visiting the feeder a minimum of three times per day to ensure repeatability of the recorded measurements27.
DNA extraction and amplicon sequencing
Genomic DNA (gDNA) was extracted from each rumen and faeces sample using a bead-beating and on-column purification28. DNA extracts were quantified on a Nanodrop 1000 Spectrophotometer (Thermo Fisher Scientific, France) and run on a FlashGel System (Lonza, Rockland, Inc.) to check integrity. Approximately 15 µg of rumen or faecal gDNA were sent to Roy J. Carver Biotechnology Center (Urbana, IL61801, USA) for microfluidics PCR amplification (Biomark HD, Fluidigm, South San Francisco, USA) and HiSeq Illumina paired end sequencing. Selected primers for amplification targeting the V3–V5 region of 16S rRNA gene of bacteria, 16S rRNA gene of archaea, fungal ITS2 and protozoal 18S rRNA gene are presented in Table S2. After amplification all samples were pooled. The library was sequenced on one lane of a HiSeq V2 Rapid flowcell for 251 cycles from each end of the fragments using a HiSeq 500-cycle SBS sequencing kit version 2 (Illumina, San Diego, USA).
Bioinformatic analysis
All pipelines have a quality control step, removing sequences with Phred scores of < 25 and trimming based on expected amplicon lengths, merging paired reads, chimera identification and removal, and OTU picking. Rare OTUs (tripletons or less) were removed from the total OTU dataset. All sequences were submitted to the Sequence Read Archive under BioProject ID PRJNA641997.
Archaeal output fastq files were analysed by downstream computational pipelines (QIIME software v.1.33.229; length ≥ 250, maximum five primer mismatches, < 8 homopolymers). Chimeric sequences were detected using the UCHIME algorithm (USEARCH 6.1) to run both de novo and reference based chimera detection. Sequences were clustered at the 99% sequence similarity level using GreenGenes v13.8 database using an open reference-based OTU picking approach with the QIIME algorithm and usearch61 method with default parameters30,31. For rumen, 2,554,377 reads were obtained, average 20,272 per sample, 87.2% passed quality control and 7.8% chimeric sequences were eliminated. The filtered OTU table had 7987 picked OTUs. For faeces, 1,497,269 reads were obtained, average 13,368 per sample, 97.7% passed quality control and 5.0% chimeric sequences were eliminated. The filtered table had 1380 picked OTUs.
For bacteria and protozoa and for fungi we followed default settings of IM Tornado and PIPITs pipelines, respectively. Raw bacterial (1,929,957 total and 15,317 average per sample; and 2,760,706 reads, average 24,649 per sample, for rumen and faeces, respectively) and protozoal (1,567,407 total and 12,440 average per rumen sample) reads were analysed using the IM-Tornado pipeline32, a tool designed to analyse sequencing data producing two separate reads that do not overlap. For rumen contents 90.2% and 87.3% of the reads passed the quality trimming for bacteria and protozoa, respectively; for faecal contents, the good quality reads represented 88.9% and 91.6%. As reads are not overlapping, IM Tornado is performing analysis simultaneously on Read1, Read2 and merged R1R2. At the dereplication step, all quality trimmed reads were collapsed into a set of unique reads which accounted for (average of Read1, Read2 and R1R2) 28 ± 8% and 25 ± 8% in the bacterial rumen and faecal datasets and 30 ± 9% and 16 ± 7% for the protozoa. In all datasets, chimeric sequences represented less than 1%. Silva database release 123_1 was used in both cases to assign taxonomy. OTU picking resulted in 3551 OTUs for rumen bacteria, 2303 for faecal bacteria, 2303 for rumen protozoa and 465 for faecal protozoa. Faecal protozoa OTUs were not associated to known symbiotic ciliates associated to feed degradation. As a preliminary analysis revealed no differences between groups, they are not further presented in the Results section.
Fungi ITS2 fastq files (4,402,028 total reads and an average 34,937 per sample; and 2,211,525 reads, average 19,746 per sample, for rumen and faeces, respectively) were analysed downstream using the automated PIPITS pipeline (https://github.com/hsgweon/pipits33). Quality filtered reads represented 96.1% and 95.4% in rumen and faecal contents, respectively. In both datasets 82% of the reads were taxonomically annotated. OTUs were picked using open-reference OTU picking method and the its_12_11_otus reference taxonomy, provided by the UNITE database (https://unite.ut.ee). For rumen 30,100 OTUs were picked and 16,390 OTUs were picked for faeces.
Absolute quantification of archaeal and bacterial gene copies in ruminal contents
Gene copies of 16S rDNA for archaea and bacteria in week 11, 14, 23 and 60 weeks of life were quantified using a qPCR approach. Primers used are summarized in Table S2; reaction assay and temperature cycles for archaeal and bacterial 16S rDNA were conducted as previously described26,34. Triplicate qPCR quantification was performed on 20 ng of extracted gDNA. Amplifications were carried out using SYBR Premix Ex Taq (TaKaRa Bio Inc., Otsu, Japan) on a StepOne system (Applied Biosystems, Courtabeuf, France). Absolute quantification of 16S rDNA copies involved the use of standard curves prepared with gDNA of Methanobrevibacter ruminantium DSM 1093 and Prevotella bryantii DSM 11,371 as described34. Results are expressed as log10 gene copies/mL of rumen liquid. PCR efficiencies were 104% and 95% for archaeal and bacterial 16S rDNA, respectively.
Plasma metabolite analysis
Plasma samples (100 µl) for week 14 and 23 were thawed at room temperature and deproteinised with 300 µl of cold methanol. After centrifugation (14,000×g, 10 min, + 4 °C), supernatants were evaporated using a Genevac EZ-2 evaporator (Genevac SP Scientific, Ipswich, UK), and dried residues were dissolved in 50/50 (v/v) acetonitrile/water containing 0.1% formic acid. The mixture was transferred to an autosampler vial, and 10 µl were injected into a LC–MS/ToF system. To evaluate the system’s reproducibility and stability, nine quality-control (QC) samples were prepared by mixing equal volumes of all plasma samples. These QCs were analysed every ten samples, i.e. three times throughout the LC–MS analytical run. Metabolic profiles were then analysed using fast liquid chromatography (1200 Series, Agilent, France) coupled to a time-of-flight mass spectrometer (microTOF, Bruker, Germany). Separation was performed on an Acquity HSS T3 column (Waters, France) using a water/acetonitrile (both containing 0.1% of formic acid) gradient at a flow rate of 0.4 mL/min. The linear gradient of 0% to 100% of acetonitrile was applied for 13 min before returning to the initial conditions and equilibrating for 7 min.
The MS system was operated in positive ionization mode with a scan range of 50–800 m/z. The capillary was set to − 4.5 kV, the nebulizer was operated at 2 bars; the dry gas was set to 8 L/min at a temperature of 200 °C. The capillary exit was set to 90 V with skimmer 1 set to 30 V. The time of flight (ToF) was calibrated by using lithium formate (ions at m/z 90 and 800). For accurate mass acquisition, a formate acetate solution was infused during the run at a flow rate of 100 μL/min monitoring for positive ion mode.
Dietary chemical composition
Dietary samples of hay and starter concentrate were ground to pass through a 1-mm sieve and stored at 4 °C for analysis according to the Association of Official Analytical Chemists35. Organic matter was determined by ashing at 550 °C for 6 h (Association of Official Analytical Chemists [AOAC] method number 923.03). Crude protein was determined by the Dumas method (CP; N × 6.25; AOAC method number 992.15). Cell wall components (NDF and ADF) were determined with residual ash (AOAC methods number 200.04 and 973.18) and starch was determined by a polarimetric method36.
Statistical analysis
Calf growth, intake, physiological parameters (VFA and methane production) were tested using PROC mixed in SAS 9.4 (SAS Institute Inc., Cary, NC, United States) with repeated measures analysis. Means were compared using the LSMEANS/DIFF statement with treatment, week, and the interaction of treatment × week as fixed terms; calves nested within treatment as a random effect and week as a repeated measure. The inclusion of breed in the model was not significant and thus, it was not included. Initial and final BW were analysed using a model similar to that described above, but excluding week as a repeated measure. Alpha diversity metrics: Shannon and Simpson diversity indices, richness, and evenness for all microbial communities and qPCR abundance were analysed independently for each time point using the PROC GLM procedure in SAS.
OTUs tables per week and microbial group were uploaded separately in R, and OTU tables were TSS (Total Sum Scaling) normalized after addition of pseudocounts to the whole datasets to deal with zero-values. Subsequent analysis on these normalized tables was performed using mixOmics37 on centred log-ratio (CLR) transformed compositional data. Initial PLS-DA analysis was done on 10 components for performance assessment. Distance and number of components were manually tuned by microbial group and by week, after examining the error rates (perf function). Tuning of the sparse-PLS-DA allowed to select the number of components and variables for the final model. Further, in an attempt to explain fermentation parameters and methane production with respect to OTU abundance, heatmaps were made (cim function on mixOmics) based on the sparse-PLS results on three components. The canonical mode was used with a threshold value of 0.7 and default Euclidian distance and Ward linkage.
Microbial community composition was compared across treatments in each week using a t-test and where appropriate a post hoc FDR test was used to eliminate false significant data38. For plasma metabolite analysis, MS data were converted to NetCDF format using DataAnalysis 3.4 software (Bruker Biospin, Germany), and processed using a Galaxy web instance workflow for metabolomics (W4M)39. The data were first filtered on retention time, and signals outside the range (< 0.4 min or > 22 min) were removed. Background noise was also removed by subtracting masses found in blank samples (solvents). Signal intensities were then normalized within-batch using a linear regression model40. After filtration and normalisation, the number of features was reduced from 2426 to 1175. A data matrix containing mass and retention time with associated signal intensities for all detected peaks was generated and analysed by multivariate analysis using SIMCA-P software (Umetrics, v. 13.01, Sweden). Principal component analysis (PCA) was performed to look for clusters between treatments or identify potential outliers. A tight clustering of QC samples was obtained in a PCA model indicating stable analytical conditions throughout all measurements.
Source: Ecology - nature.com
