Search

Additional file 8 Suppl. Fig. 6. Proportion of linear model (LM, top row) and neural network (NN, bottom row) predictions with a R2 above (left column) or below (right column) the 95% confidence interval computed from the predictions with the random sets of genes for each gene set (there is no neural network model computed for the Complete BLUP full set).

Additional file 5 Suppl. Fig. 3. Relationship between Spearman’s correlations between module-trait (y-axis) and gene significance-kME (x-axis).

Additional file 10 Suppl. Fig. 8. Predictions scores on test sets for increasing numbers of the peripheral genes. Violin and boxplots of prediction R2 for the LM Ridge algorithm and for increasing sizes of the peripheral genes set (in brown) and the peripheral NG genes set (in orange), used for the predictions (in percent of the full set).

Additional file 4 Suppl. Fig. 2. Traits hierarchical ascendant clustering dendrogram. Clustering was performed from the correlations between traits with Ward method ("Ward.D2") by the R package pvclust. Approximately Unbiased (au, in red) and Bootstrap Probability (bp, in green) p-values indicated the degree of belief associated with clusters. Highly supported modules are framed by a red square, grouping (a) the mean sample diameter with the two circumference traits, (b) the S/G ratios with glucose composition, (c) the two C5/C6 together, and (d) the H/G ratios.

Additional file 3 Suppl. Fig. 1. PCA score plots on gene expression data. Each plot represents the distribution of the individuals on the 2 first axes of the PCA (representing 17,7% of the variation), colored by class of various experimental factors (Xylem and cambium scraper, extractor and extraction method, population, sequencing column, line and plate, the growth rate at harvest, sampling date, time, temperature, solar radiation, humidity and wind speed). Cofactors related to weather are presented in the 6 lower plots.

Additional file 7 Suppl. Fig. 5. Histograms of the centrality scores for the genes selected by Boruta at different p-values thresholds. Repartition of selected genes within the following gene sets is hilighted, with core in blue, peripheral NG in orange, peripheral in brown and other (NA) in black. Four p-value thhresolds for Boruta selections were considered: 0.01, 0.05, 0.1 and 0.2.

Additional file 6 Suppl. Fig. 4. Histograms of the centrality scores without (top) or with (bottom) the grey group. Core, peripheral and peripheral without grey sets are represented respectively by the blue, dark orange and orange bars. Random sets are distributed across the histogram and do not appear on this figure. Distribution of genes clustered in the grey module is represented by the grey bars, white bars are for other genes.

Additional file 9 Suppl. Fig. 7. Difference of prediction scores between algorithms (top) and sets (bottom). On the top panel, the difference between LM and NN prediction scores for the core (in blue), random (in grey), peripheral (in brown), peripheral (in orange) and Boruta gene sets (in green). On the bottom panel, the LM differences are in red and the NN differences in turquoise and the color filling the bar represents the difference between core and peripheral genes in brown, core and peripheral NG in orange and between the random sets in grey. For the random pairs, error bars represent the first and third quartiles of the differences between pairs of randomized sets and the bar corresponds to the median.

International audience; Understanding the mechanisms triggering variation of cell wall degradability is a prerequisite to improving the energy value of lignocellulosic biomass for animal feed or biorefinery. Here, we implemented a multiscale systems approach to shed light on the genetic basis of cell wall degradability in maize. We demonstrated that allele replacement in two pairs of near-isogenic lines at a region encompassing a major quantitative trait locus (QTL) for cell wall degradability led to phenotypic variation of a similar magnitude and sign to that expected from a QTL analysis of cell wall degradability in the F271 × F288 recombinant inbred line progeny. Using DNA sequences within the QTL interval of both F271 and F288 inbred lines and Illumina RNA sequencing datasets from internodes of the selected near-isogenic lines, we annotated the genes present in the QTL interval and provided evidence that allelic variation at the introgressed QTL region gives rise to coordinated changes in gene expression. The identification of a gene co-expression network associated with cell wall-related trait variation revealed that the favorable F288 alleles exploit biological processes related to oxidation-reduction, regulation of hydrogen peroxide metabolism, protein folding and hormone responses. Nested in modules of co-expressed genes, potential new cell-wall regulators were identified, including two transcription factors of the group VII ethylene response factor family, that could be exploited to fine-tune cell wall degradability. Overall, these findings provide new insights into the regulatory mechanisms by which a major locus influences cell wall degradability, paving the way for its map-based cloning in maize.

International audience; Maize has become an extensively cultivated crop in high latitudes like Northern Europe thanks to historical improvements of cold tolerance. However, earlier sowing increase the risk of exposure to longer chilling periods, affecting early growth and frequently plant performance and final yield. Understanding how maize responds to chilling periods is therefore a major task to both better understand maize local adaptation and improve agriculture. Here, we evaluated two sister double-haploid dent maize lines sharing 82% of their genome and displaying a contrasted tolerance to chilling. Using an Illumina stranded and paired-end mRNA-seq dataset from leaves of both sister lines grown under control and chilling conditions, we captured the allelic variation consequences at the transcript level. Clustering of differentially expressed gene profiles let us identify 574 genes differentially expressed, 513 and 61 being up- and down-expressed in the chilling-tolerant line compared to the chilling-sensitive line. We then explored how the variation in gene expression contributes to the variation in phenotypic traits. Genes associated with these traits were identified, paving the way for pinpointing candidate genes for chilling tolerance in future follow-up studies.

International audience; Botrytis cinerea is a nectrotrophic plant pathogen that leads to important economic losses in many crops. B. cinerea has a large arsenal of virulence, including the production of cell wall degrading enzymes, reactive oxygen species and toxins. B. cinerea has a wide host range and plant resistance to this pathogen is quantitative. Nitrogen (N) is a major limiting factor of plant growth and soils frequently lack sufficient N. Large quantities of N fertilizers are used worldwide and have an impact on crop disease. However, the variety of effects observed indicate that there is a complex relationship linking N availability to plant disease. We showed previously that high N supply increases the susceptibility of A. thaliana to B. cinerea (Fagard et al., 2014, J. Exp Bot 65: 5643-5656). A transcriptomic approach of B. cinerea-infected A. thaliana plants grown in low or high N has enabled us to identify both fungal and plant genes for which expression is affected by the plant’s N status. Characterization of corresponding fungal knock-out mutants led to the identification of novel B. cinerea virulence genes, including a GST. On the plant side, we identified the jasmonate signaling pathway as a key player in the modulation of A. thaliana susceptibility by N supply. Our current working model is that N availability for plants primarily affects the outcome of plant-pathogen interactions through its effect on plant defense and pathogen virulence gene expression.

Figure S6. Floral transition and hormone pathways. (a) Distribution of AHD genes in clusters. (b) Genes associated with gibberellin metabolism and responses. (PDF 56 kb)

Figure S3. RNA-Seq experiments and expressed gene distributions. (a) Library sizes of the biological replicates. (b) Read mapping for the different time points and biological replicates. (c) Distribution of the expressed genes in the genomes. (d) Differentially expressed genes (DEGs) and Venn diagram with the three main comparisons. (e) Distribution of the expressed genes in the main gene classes. (PDF 194 kb)

Recent literature on the differential role of genes within networks, including the omnigenic model, distinguishes core from peripheral genes in the layout underlying phenotypes. Cores are typically few, each of them highly contributes to phenotypic variation, but because they are so few, they altogether only explain a small part of trait heritability. In contrast, peripherals, each of small influence, are so numerous that they finally lead phenotypic variation. We collected and sequenced RNA from 459 European black poplars and built co-expression networks to define core and peripheral genes as the most and least connected ones. We computed the role of each of these gene sets in the prediction of phenotypes and showed that cores contribute additively to phenotypes, consistent with a downstream position in a biological cascade, while peripherals interact to influence phenotypes, consistent with an upstream position. Quantitative and population genetics analyses further revealed that cores are more expressed than peripherals but they tend to vary less and to be more differentiated between populations suggesting that they are more constrained by natural selection. Our work is the first attempt to integrate core and peripheral terminologies from co-expression networks and omnigenic theory. In the end, we showed, that there seems to be a strong overlap between them, with core genes from co-expression networks likely being a mixture of integrative hubs with a direct effect on phenotype in agreement with the omnigenic theory, and master regulators which control the overall metabolic flow towards the phenotype.

Figure S4. Analysis of lncRNAs. (a) Resources used to construct the lncTU dataset. (b) lncTUs differentially expressed for each comparison using the Benjamini-Hochberg (BH) test. (PDF 51 kb)

Figure S8. Expression of MAF5 and its antisense lncTU located in the 3′ end of MAF5 region at T0, T2 and T3. (a) Fold changes. (b) Browser snapshot showing the expression profiles. BR1: biological replicate number 1. (PDF 77 kb)

Figure S5. Differentially expressed endoreduplication-related genes extracted from the ThaleMine database. Log2 (Fold-Change) is reported. In black, non-statistically significant fold change values. (PDF 44 kb)