Computational protocol: Key role of lipid management in nitrogen and aroma metabolism in an evolved wine yeast strain

Similar protocols

Protocol publication

[…] Statistical analysis was performed with R software, version 3.1.1 [].We obtained three datasets, in which each variable of interest is a curve along the time (h) that we expressed in terms of consumed sugar (g/L). We chose to summarize these three datasets by modeling each curve with an adequate model and then extracting criteria of interest.First, for each condition, the biomass was modeled using a Weibull model with the drc package []. The one four-parameter Weibull model is written as follows:6fx=c+d-c1-exp-expblnx-lneThis four-parameter ascending function is asymmetric with an inflection point at time e. For each modeled function, we extracted several criteria of interest: µmax, defined as the maximal of the ratio f’(t)/f(t) for each t, expressed in h−1; the inflection point, expressed in terms of consumed sugar g/L; and the maximum biomass, expressed in 106 cells.Considering amino acid (AA) consumption, we modeled each AA under each condition with the drc package and a Weibull model. The four-parameter Weibull function is written as follows:7fx=c+d-cexp-expblnx-lneThis four-parameter decreasing function is asymmetric with an inflection point at time e. For each modeled function, we extracted the following criteria: the maximal rate, which is the maximum of the first derivative of the function expressed in mg/L.h, and the inflection point and the point at which the quantity of AA is null (called Point.AA0), both expressed in terms of consumed sugar (g/L).For these two parametric models, the normality of residual distributions and homogeneity of variance were studied with standard diagnostic graphs; no violation of the assumptions was detected.Each volatile compound under each condition was then modeled using a non-parametric model using the cellGrowth package []. The model used is a local regression and allows for the extraction of the inflection point expressed in consumed sugar (g/L), the maximal production in mg/L and the maximal rate (maximum of the first derivative in mg/L.h). To calculate the specific rate, we divided the first derivative of the model (the rate) by the population, as estimated above. Finally, we recorded the maximum specific rate (SRmax) and the time at which this maximum was reached, expressed in consumed sugar (g/L) (PointSRmax).To provide an overview of the dataset, principal component analysis (PCA) was carried out with the FactoMineR package [].Multivariate factorial analysis (MFA) was then performed for the two strains (Lalvin EC1118® and Affinity™ ECA5) at two levels of nitrogen (70 and 330 g/L) and two levels of phytosterols (2 and 8 mg/L). This analysis allowed for the study of links between the consumption of AA and volatile compound production []. [...] For each fermentation condition (SM330, 8 mg/L of phytosterols with the two strains), three independent fermentations were carried out in parallel and sampled when CO2 production reached 35 and 70 g/L, corresponding to two different phases of aroma metabolism. Cells (1x109 cells) were harvested by centrifugation at 1000g for 5 min at 4 °C, and the cell pellets were washed with DEPC-treated water and then frozen in methanol at -80 °C. Total RNA was extracted with Trizol reagent (Gibco BRL, Life Technologies) and was purified with the RNeasy kit (Qiagen). The quantity and quality of the extracted RNA were verified by spectrometry (NanoDrop 1000, Thermo Scientific). We used the Agilent 8 × 15 k gene expression microarrays (Design ID 038619 with 40 EC1118-specific genes, Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s instructions. Fluorescent cRNAs were synthesized from 100 ng of total RNA using the One color RNA Spike-In kit (Agilent Technologies). Labeled cRNA was purified with the RNeasy Kit (Qiagen). Microarrays were hybridized for 17 h at 65 °C in a rotating hybridization oven (Corning) with the Gene Expression Hybridization kit (Agilent). The hybridization signal was detected with a GenePix 4000B laser Scanner (Axon Instruments).The limma package [] was used to import and normalize the global microarray data (quantile method for normalization between arrays). The entire dataset is available in the “Gene Expression Omnibus Database” (No. GSE68354). Transcriptomic data were analyzed by two different methods.For each level of CO2 released (35 and 70 g/L) and based on this normalized dataset of 6200 expression data for the two strains, we used sparse partial least squares—discriminant analysis (sPLS-DA), which is an exploratory approach in a supervised context, to select the most important transcripts relative to the four samples []. We tuned the number of dimensions of the sPLS-DA to two and the number of variables to choose on these two dimensions to 500 (250 for each).Functional analysis was performed on the selected transcripts by time point to highlight significant functional groups according to the gene ontology (GO) process terms using the Genecodis program [] via the FDR method at a p value cutoff of 0.05 [].For each time point, MFA was then performed to obtain an overview of the dataset, which consisted of 513 variables measured for the two strains (Lalvin EC1118® and Affinity™ ECA5) and for the two sampling times. The dataset included a set of individuals described by two types of variables: the normalized expression of the 500 transcripts selected by the sPLA-DA according to the two strains and the 13 compounds (or ratios) produced during fermentation by the two strains. The MFA took the structure of the two groups of data into account and balanced the effect of each group of variables, enabling the study of links between expression data and volatile compounds production [].To determinate the differential gene expression between experimental conditions, a modified t-test was performed by filtering on confidence at p < 0.05, using the Benjamini and Hochberg false discovery rate as multiple testing corrections of the t-test p values []. The genes with different levels of expression were grouped according to gene ontology (GO) process terms using the Genecodis program []. […]

Pipeline specifications

Software tools FactoMineR, limma, GeneCodis
Databases GEO
Applications Miscellaneous, Gene expression microarray analysis
Organisms Saccharomyces cerevisiae
Chemicals Acetyl Coenzyme A, Alcohols, Amino Acids, Esters, Nitrogen