Computational protocol: Normal milk microbiome is reestablished following experimental infection with Escherichia coli independent of intramammary antibiotic treatment with a third-generation cephalosporin in bovines

Similar protocols

Protocol publication

[…] Isolated genomic DNA was used as a template for amplification of the V4 hypervariable region of the bacterial 16S rRNA gene using the primers 515F and 806R, which had been optimized for the Illumina MiSeq platform (Illumina Inc., San Diego, CA) [] as described previously [].Six runs of the Illumina MiSeq sequencer were needed for sequencing of all samples. In each run, 279 samples and a sequencing control were pooled in an equimolar library and sequenced using the MiSeq reagent kit V2 for 300 cycles. Bioinformatics was performed using quality-filtered indexed reads, which were concatenated into a single FASTA file and uploaded in the open-source pipeline Quantitative Insights into Microbial Ecology (QIIME) version 1.9.1 [], using computer resources of the Cornell Boyce Thompson Institute (Ithaca, NY). Sequences were handled following the default settings of the pipeline. Quality filtering was performed as described previously []. Open-reference taxonomic assignment into operational taxonomic units (OTUs) with 97% identity was achieved using UCLUST [], RDP classifier [], PyNAST [], and the Greengenes [] reference database. Rare OTUs with less than five sequences in each run, and samples with less than 1000 reads were excluded from the database. Within-sample diversity (α-diversity) was assessed through Shannon diversity index calculated in a randomly selected subset of the OTU database obtained through the script single_rarefaction.py in QIIME at a rarefication level of 1500 reads per sample. Between-sample microbial diversity (β-diversity) was assessed through phylogenetic-based weighted UniFrac [] distances, calculated in QIIME through the script beta_diversity.py and the distance matrix obtained was used for comparison between groups. [...] The UNIVARIATE procedure of SAS version 9.4 (SAS Institute Inc., Cary, NC) was used to perform descriptive analyses. Non-normally distributed variables (i.e., SCC and CFU) were normalized through log transformation. Longitudinal changes in the microbial profile was assessed through description of the relative abundances of the 25 most abundant bacterial families using the tabulate function of JMP Pro 12 (SAS Institute Inc., Cary, NC), and relative abundances of all the remaining families were combined into a single cluster, defined as “Other.” Variables of interest were evaluated between challenged, unchallenged, treated, and untreated quarters with repeated measures ANOVA using the GLIMMIX procedure of SAS. Significance of pairwise comparisons were adjusted through the Tukey-Kramer multiple comparison correction. Outcomes were the relative abundance of Enterobacteriaceae, Shannon diversity index, LogSCC, LS, and LogCFUs; explanatory variables were challenge (challenged versus control quarter), treatment (treated versus untreated quarter), time relative to experimental challenge, and their two- and three-way interactions.To assess the effect of treatment, stratified analysis of covariance was performed in challenged and unchallenged quarters separately. To account for possible differences that occurred between intramammary infection and first treatment at 48 h, the average of values observed between challenge and treatment (i.e., 0, 6, 12, 24, 36, 42, and 48 hours relative to challenge) was included as a covariate in these models. Variables of interest were the relative abundance of Enterobacteriaceae, Shannon diversity index, LogSCC, LS, and LogCFUs; explanatory variables were treatment (treated versus untreated quarter), time relative to experimental challenge, and their two-way interactions. Teat nested within a cow was considered a random effect in all statistical models. The first-order ante-dependence covariance structure was selected because it resulted in the smallest Schwarz’s Bayesian information criterion value. Differences with P ≤ 0.05 were considered significant. Descriptive analyses of sequencing results are presented as average and standard deviation. All other results are presented as the least-square means followed by the respective standard error of the mean.Three animals (animals E, G, and J—Additional file ) experienced elevated SCC on 3 days preceding intramammary infusion of bacteria and for that reason did not develop an infection following the challenge with E. coli. A fourth animal (animal L—Additional file ) acquired a natural intramammary infection in one of the unchallenged quarters caused by an E. coli strain different from the experimental challenge strain. Data collected from these four animals was only used to compare challenged, infected versus challenged, uninfected quarters. All other analyses assessing the effect of experimental infection and treatment with ceftiofur did not include data from animals E, G, J, and L.Data were handled as follows: samples without a SCC value due to clotted milk (i.e., clinical mastitis) received a value of 30,000,000 SCC; samples with a CFU value that indicated too numerous to count received a value of 60,000 CFU. The rationale for choosing these arbitrary values was to assign a number that was larger than the largest observation for that variable in the dataset (i.e., the largest SCC observed was 27,255,000 and the largest CFU observed was 58,000).Multivariate analysis of milk microbiome was implemented in QIIME and R (R Core Team, Vienna, Austria) []. Analysis of similarities (ANOSIM) was performed in non-rarefied data using the vegan [] package in R. Groups shown to be significantly different in ANOSIM underwent Analysis of Composition of Microbiomes (ANCOM) [] carried out in QIIME version 2.0.6 [], in an attempt to identify which OTUs were driving the main differences between groups. Microbiome changes over time in challenged and control quarters were visualized through principal coordinates analysis (PCoA) of weighted UniFrac distances calculated in QIIME and plotted using EMPeror []. […]

Pipeline specifications

Software tools QIIME, UCLUST, RDP Classifier, PyNAST, UniFrac, JMP Pro, EMPeror
Databases Greengenes
Applications Miscellaneous, Phylogenetics, Metagenomic sequencing analysis, 16S rRNA-seq analysis
Organisms Escherichia coli, Bos taurus
Chemicals Cephalosporins