PA treatment induces extensive metabolic reprogramming
Untargeted UPLC-MS/MS profiling of shoots and roots of tomato seedlings 24h after the second PA treatment was used to further investigate the molecular basis of PA-IR.
After pre-processing, 4692 features were retained in shoots, of which 1672 were differentially abundant (DA). 985 features were more abundant in PA-treated than in mock-treated shoots, whereas 687 were less abundant. A substantial number of DA features were unique to either mock (17)- or PA-treated (396) samples. In roots, 2549 features were retained, of which 186 were significantly DA between PA- and mock-treated samples - a much smaller proportion than in shoots. There was a clear bias towards overaccumulation after PA treatment in roots: 177 features were significantly more abundant, while 9 were less abundant. The smaller effect of PA on the root metabolome is reflected by principal component analysis and hierarchical clustering, both of which show complete separation between PA- and mock-treated shoot samples, but only partial separation for root samples (Figure 6 ).
Pathway analysis identified five metabolic pathways enriched upon PA treatment in tomato shoots: phenylpropanoid biosynthesis (P < 0.001), stilbenoid and diarylheptanoid biosynthesis(P < 0.001), flavonoid biosynthesis (P = 0.020),riboflavin metabolism (P < 0.001) and thiamine metabolism (P = 0.030). In roots, four pathways were enriched:phenylpropanoid biosynthesis (P < 0.001),stilbenoid and diarylheptanoid biosynthesis (P < 0.001), flavonoid biosynthesis (P = 0.040) and galactose metabolism (P = 0.030). A full list of pathways is shown inSupplementary Table S4 .
PA itself was found in both shoot and root samples from PA-treated plants, both as free PA and as a hexose conjugate. This indicates that PA is taken up, conjugated and possibly systemically transportedin planta . The hexose conjugate was eleven times more abundant than free PA in shoots, and thirteen times more abundant in roots. Both free and conjugated PA were approximately ten times more abundant in shoots than in roots.
A second feature of interest is the defense hormone SA, which was 6.2 times more abundant in PA-treated than in mock-treated shoots (P < 0.001). By contrast, SA abundance in roots was unaffected by foliar PA treatment. To confirm the effect of PA treatment on SA seen after untargeted metabolome profiling, and to verify the absence of responses to other major plant hormones suggested by our transcriptome data, we also performed targeted UPLC-MS/MS measurement of SA, JA, abscisic acid and indole-3-acetic acid. We observed a similar SA accumulation in tomato shoots after PA treatment (4.8-fold increase) in this experiment, and found no major changes in the level of the other measured phytohormones (Supplementary Table S6 ).
Since foliar PA treatment induces SA accumulation in shoots, we investigated whether SA accumulation is necessary for PA-IR by testing the efficacy of PA in tomato plants expressing the NahGtransgenic construct, which converts SA to catechol (Brading, Hammond-Kosack, Parr & Jones 2000). Plants expressing NahG were more susceptible to M. incognita (+14%, P = 0.037),confirming the role of SA in defense against M. incognita(Martínez-Medina et al. 2017). However, PA was almost as effective in NahG plants (-36%, P < 0.001) as in wild type plants (-43%, P < 0.001; Figure 6 ), which suggests that PA-IR against M. incognita is at least partially SA-independent.
Besides PA and SA, 59 other DA features were annotated with a high degree of confidence by matching against an in-house database using mass and retention time. A further 226 DA features were putatively annotated to a metabolite class by matching their m/z against the METLIN database (Guijas et al. 2018). Both features identified against the in-house database and annotated using METLIN were notably enriched in PPP derivatives and flavonoids. PPP derivates identified via the in-house database include para- coumaroyl quinic acid (4-fold increase after PA treatment, P < 0.001), syringin (12-fold increase, P < 0.001), feruloyl hexose (3-fold reduction, P < 0.001), benzoic acid (66-fold increase, P < 0.001) and benzoyl hexose (16-fold increase, P < 0.001). The PPP precursor phenylalanine was two times more abundant in PA-treated than in mock-treated shoots (P < 0.001). Several PPP derivatives were also DA in roots, but generally with smaller fold changes. Examples include para- coumaroyl quinic acid (2.3-fold increase, P < 0.001), syringin (2.2-fold increase, P = 0.031) and benzoyl hexose (12.6-fold increase, P = 0.002).
Among DA features, 153 had m/z values whose only METLIN matches were flavonoids or flavonoid glycosides and whose retention times were compatible with these metabolite classes. Of these 153 features, 67 were less and 86 were more abundant in PA-treated than in mock-treated shoots. Remarkably, 50 of the 86 upregulated flavonoids were unique to PA-treated samples; these flavonoids were either absent in mock-treated plants, or present below the limit of detection. No putative flavonoids were unique to mock-treated samples. In root samples, twelve putative flavonoids were DA, all of which were upregulated and six of which were unique to PA-treated samples.
There is a strong overlap between DA features in root and shoot: 156 out of 186 DA features (84%) in roots were also DA in shoots. Putatively annotated features that were more abundant in both PA-treated roots and shoots include two benzoic acid hexosides, a cinnamoyl hexoside,para- coumaroyl quinic acid, a hexose conjugate of 5-hydroxyferulic acid, syringin, seven unidentified flavonoid glycosides and three unidentified flavonoids. A full list of DA features, alongside their empirical formulas and putative annotations, is provided inSupplementary Table S5 .