Natural variation of root exudates in Arabidopsis thaliana-linking metabolomic and genomic data

Abstract

Many metabolomics studies focus on aboveground parts of the plant, while metabolism within roots and the chemical composition of the rhizosphere, as influenced by exudation, are not deeply investigated. In this study, we analysed exudate metabolic patterns of Arabidopsis thaliana and their variation in genetically diverse accessions. For this project, we used the 19 parental accessions of the Arabidopsis MAGIC collection. Plants were grown in a hydroponic system, their exudates were harvested before bolting and subjected to UPLC/ESI-QTOF-MS analysis. Metabolite profiles were analysed together with the genome sequence information. Our study uncovered distinct metabolite profiles for root exudates of the 19 accessions. Hierarchical clustering revealed similarities in the exudate metabolite profiles, which were partly reflected by the genetic distances. An association of metabolite absence with nonsense mutations was detected for the biosynthetic pathways of an indolic glucosinolate hydrolysis product, a hydroxycinnamic acid amine and a flavonoid triglycoside. Consequently, a direct link between metabolic phenotype and genotype was detected without using segregating populations. Moreover, genomics can help to identify biosynthetic enzymes in metabolomics experiments. Our study elucidates the chemical composition of the rhizosphere and its natural variation in A. thaliana, which is important for the attraction and shaping of microbial communities.

Figure 1

Hierarchical clustering of metabolic features from (a) exudates ESI(−), (b) ESI(+) and of (c) genetic distances. (a+b) Features were obtained by UPLC/ESI(−)-QTOF-MS (a) or UPLC/ESI(+)-QTOF-MS (b) from exudate samples and differed from the blank (Welch test, p < 0.05). Intensities were corrected for batch effects using SVA and subjected to average linkage clustering with correlation as a distance measure. (c) Variant tables of the 19 genomes project were reduced to coding regions, as annotated by TAIR. The sum of all mismatches was used as a distance matrix for average linkage clustering. Dendrograms were cut at a correlation threshold of 0.95 (dashed line). As cluster numbers were not comparable, consistent clusters were coloured across ion modes as a visual guidance.

Figure 2. Colour-coded intensity matrix of differential metabolites occurring in exudates.

Integrated peak areas were log-transformed and scaled to zero mean and standard variance. A Welch-test was used to find differentially abundant metabolites between the 19 accessions.

Figure 3. Workflow for matching metabolic patterns of absence with stop codons in genes annotated as AraCyc enzymes.

For the metabolic data, 384 out of 455 metabolic features from the ESI(−) data set were absent in at least one accession. 38 of them were annotated as monoisotopic peak [M] by CAMERA. Approximately 32,000 stop codons were detected. 1,588 of AraCyc enzyme-encoding genes displayed a prematurely ended amino acid sequence possibly representing non-functional enzymes that can be causative for metabolite absence.

Figure 4. Natural and T-DNA insertion knockouts of SCT.

(a) Relative transcript levels of SCT in root tissue as determined by qPCR, PP2A as reference, normalized to Rsch-4, mean ± s.e.m., n = 3. (b) Peak area counts of cyclic didehydro-di(coumaroyl)spermidine sulfate in exudates, mean ± s.e.m., n = 3.

Figure 5. Robinin absence is linked to a stop codon in the UGT91A1 encoding gene.

(a) Peak area counts, mean ± s.e.m. (n = 3) with absence in Wu-0 (highlighted in red) (b) MS/MS spectrum of robinin, 30 eV, (c) extracted ion chromatogram at m/z 739.21 with kaempferol 3-O-Rha(1→2)Glc 7-O-Rha eluting at 3.9 min and the galactose-conjugated robinin eluting at 4.3 min not detected in the natural knockout Wu-0 and T-DNA insertion line SALK_088702C, (d) relative transcript levels of UGT91A1 in roots as determined by qPCR, PP2A as reference, normalized to Col-0, mean ± s.e.m., n = 4, (e) schematic representation of the UGT91A1 gene (one exon) and the loss-of-function mutations in Wu-0 and SALK_088702C.

Figure 6. Biosynthetic pathway of cyclic didehydro-di(coumaroyl) spermidine sulfate.

Di(coumaroyl)spermidine is synthesized by SCT and subsequent oxidative ring closure and sulfonylation leads to cyclic didehydro-di(coumaroyl) spermidine sulfate, PAPS = 3′-phosphoadenosine-5′-phosphosulfate.