Protein lysine methylation contributes to modulating the response of sensitive and tolerant Arabidopsis species to cadmium stress

Funding information Agence Nationale de la Recherche, Grant/ Award Number: ANR-10-LABX-49-01; Department of Plant Biology and Breeding of the Institut National de la Recherche Agronomique; Fondation de Coopération Scientifique Rovaltain; Fonds de la Recherche Scientifique-FNRS, Grant/Award Number: PDR T.0206.13; Région Auvergne RhôneAlpes; Toxicology Program of the Commissariat à l'Energie Atomique et aux Energies Alternatives Abstract The mechanisms underlying the response and adaptation of plants to excess of trace elements are not fully described. Here, we analysed the importance of protein lysine methylation for plants to cope with cadmium. We analysed the effect of cadmium on lysine-methylated proteins and protein lysine methyltransferases (KMTs) in two cadmium-sensitive species, Arabidopsis thaliana and A. lyrata, and in three populations of A. halleri with contrasting cadmium accumulation and tolerance traits. We showed that some proteins are differentially methylated at lysine residues in response to Cd and that a few genes coding KMTs are regulated by cadmium. Also, we showed that 9 out of 23 A. thaliana mutants disrupted in KMT genes have a tolerance to cadmium that is significantly different from that of wild-type seedlings. We further characterized two of these mutants, one was knocked out in the calmodulin lysine methyltransferase gene and displayed increased tolerance to cadmium, and the other was interrupted in a KMT gene of unknown function and showed a decreased capacity to cope with cadmium. Together, our results showed that lysine methylation of nonhistone proteins is impacted by cadmium and that several methylation events are important for modulating the response of Arabidopsis plants to cadmium stress.


| INTRODUCTION
As sessile organisms, land plants must deal with fluctuating levels of essential and non-essential trace elements in soils. Some plant species have the ability to colonize soils contaminated by toxic levels of metals and display remarkable leaf metal accumulation without visible toxicity symptoms. Understanding tolerance and accumulation of metals in these species, referred to as hyperaccumulators, offers the unique opportunity to uncover key mechanisms governing metal homeostasis and adaptation to challenging environments (for reviews, see Verbruggen, Hermans, & Schat, 2009;Kramer, 2010). Arabidopsis halleri, a close relative of A. thaliana and A. lyrata, is a model species for studying tolerance and accumulation of cadmium (Cd), one of the most toxic metal for living organisms (for reviews, see Kramer, 2010;DalCorso, Fasani, & Furini, 2013;Verbruggen, Juraniec, Baliardini, & Meyer, 2013;Moulis, Bourguignon, & Catty, 2014). While Cd and zinc tolerance seems to be constitutive in A. halleri, populations originating from different genetic units and from metallicolous or nonmetallicolous soils display important variability in terms of Cd accumulation Meyer et al., 2015;Schvartzman et al., 2018;Stein et al., 2017). This intraspecific variability suggests adaptation at the local scale and, possibly, the involvement of different molecular mechanisms to account for metal accumulation and tolerance traits.
In the last 15 years, the combination of genetic, 'omics' and functional approaches in both tolerant and non-tolerant species have contributed considerably to the understanding of Cd toxicity, tolerance and accumulation. Key mechanisms involved in metal uptake, translocation, chelation with ligands, vacuolar sequestration and cell signalling have been characterized (for reviews, see Villiers et al., 2011;DalCorso et al., 2013;Clemens, Aarts, Thomine, & Verbruggen, 2013;Clemens & Ma, 2016). The coordination of these processes is accomplished through multilevel regulatory mechanisms, including the epigenetic, transcriptional and post-translational levels Haak et al., 2017). Despite recent progress, the role of posttranslational modifications (PTMs) in the response and adaptation of plants to Cd and other trace elements is still poorly documented.
Given their pivotal role in the regulation of many cellular processes, it is anticipated that PTMs are important for plants to cope with biotic and abiotic stresses (Dahan, Koen, Dutartre, Lamotte, & Bourque, 2011;Kosova, Vitamvas, Prasil, & Renaut, 2011). To date, the most studied modification in the field of metal stress is protein phosphorylation. It has been shown that Cd poisoning in A. thaliana or Medicago sativa is accompanied by the activation of several mitogen-activated protein kinases (Jonak, Nakagami, & Hirt, 2004;Liu et al., 2010) and SNF1-related protein kinases (Kulik et al., 2012). Also, it has been reported that Cd stress induces the hyper-phosphorylation of the eukaryotic translation initiation factor eIF2a (Sormani et al., 2011) and the plasma membrane H + -ATPase (Janicka-Russak, Kabala, & Burzynski, 2012). Metal stress is also responsible for changes in the pattern of PTMs on tubulin (Gzyl, Chmielowska-Bak, Przymusinski, & Gwozdz, 2015). Finally, it is worth noting that the production of reactive oxygen species triggered by Cd is responsible for the carbonylation of proteins, an irreversible oxidative modification leading to the degradation of damaged proteins by the proteasome. Several studies indicated that plants treated with Cd display an increased oxidation of proteins (e.g., Pena, Barcia, Azpilicueta, Mendez, & Gallego, 2012;Romero-Puertas, Palma, Gómez, Del Río, & Sandalio, 2002) and an upregulation of the 20S proteasome proteolytic pathway (e.g., Pena, Pasquini, Tomaro, & Gallego, 2006;Polge et al., 2009;Sarry et al., 2006).
Protein methylation is a very diverse PTM acting on different protein residues. It is widespread in eukaryotic proteomes, modifying both histones and non-histone proteins, and contributes to the finetuned regulation of protein function. The modification of the lysine (Lys) side chain of proteins is among the most frequent methylation event (Falnes, Jakobsson, Davydova, Ho, & Malecki, 2016;Lanouette, Mongeon, Figeys, & Couture, 2014). It is catalysed by two structurally different classes of protein Lys methyltransferases (KMTs), the SET domain-containing group (SDG) and the seven-beta-strand (SBS) superfamily, which are able to add one to three methyl groups to specific Lys residues in proteins (Serre, Alban, Bourguignon, & Ravanel, 2018). In plants, these enzymes have been shown to methylate histones and non-histone proteins involved in all aspects of cell biology (e.g., transcription, protein synthesis, metabolism). Protein Lys methylation can be reversible by the action of demethylases. Demethylases acting on non-histone Lys-methylated proteins have not been yet reported in plants (Serre et al., 2018).
The identification of the enzyme/substrate (KMT/protein) relationship is a critical step towards understanding the role of protein methylation (Falnes et al., 2016;Lanouette et al., 2014;Serre et al., 2018). As such, inactivating the genes coding KMTs is the best way to analyse the functional outcomes of methylation in vivo. Despite recent progress, the role of non-histone protein Lys methylation in regulating plant cellular functions is still limited (Serre et al., 2018). In particular, no information is available about the role of protein methylation in the response and adaptation of plants to metal stress. The present work is based on the assumption that this PTM could be important for plants to efficiently address stress situations induced by metals. This hypothesis is supported by the abundance and diversity of Lys-methylated proteins, possibly targeting components of metal transport, signalling pathways or detoxification machineries, and the recognized role of this modification in the regulation of protein function (Falnes et al., 2016;Lanouette et al., 2014). To test this hypothesis, we analysed the expression of the two main players involved in protein Lys methylation, that is, Lys-methylated proteins and genes coding KMTs, in three Arabidopsis species challenged with Cd. We used A. thaliana and A. lyrata non-tolerant plants and three populations of A. halleri from different genetic units and showing contrasting tolerance and accumulation of Cd Meyer et al., 2015;Schvartzman et al., 2018). First, we showed that some non-histone proteins are differentially methylated at Lys residues in response to Cd and we identified one of these proteins by MS/MS. Second, we showed that Cd stress has limited impact on the transcriptional regulation of KMT genes. Third, using a root growth inhibition assay with A. thaliana mutants disrupted in genes coding KMTs, we showed that nine out of 23 mutants have a tolerance to Cd that is different from that of wild-type seedlings. Finally, we characterized two of these mutants that are either more tolerant or more sensitive to Cd. Together, our results showed that Cd triggers changes in the expression of a few Lys-methylated proteins and KMT  . Plants were grown at 21 C, 70% air humidity, under short day conditions (8 hr of light per day) with a light intensity of 80 μmol of photons m −2 s −1 . After 4 weeks of growth, plants were maintained in CM or challenged with 0.2-5 μM CdSO 4 in CM for 7-10 days. At the end of the treatment, roots and leaves were harvested separately from each individual, washed twice in distilled water, dried with absorbent paper, and frozen in liquid nitrogen.

| Protein extraction and immunoblotting
Proteins from Arabidopsis tissues were extracted by grinding frozenpowdered tissues in 50 mM Tris-HCl, pH 8.0, 10% (vol/vol) glycerol and protease inhibitors (Roche Applied Science). Samples were centrifuged at 16,000×g for 20 min and the supernatant used as a source of soluble proteins. Pellets were suspended in the extraction buffer supplemented with 2% (wt/vol) SDS, incubated for 15 min at room temperature, and centrifuged as before to recover solubilized membrane proteins. Proteins were resolved by SDS-polyacrylamide gel (SDS-PAGE), transferred to nitrocellulose membrane, and probed with an antibody against trimethyl-Lys (abcam 76118). Membranes were also probed with antibodies against fructose bisphosphate aldolase (Mininno et al., 2012) and the beta subunit of ATPase (Agrisera, AS03 030) for the normalization of protein loading. Protein detection was achieved using the ECL Plus Western Blotting detection reagents.

| Identification of Lys-methylated proteins by mass spectrometry
Sample preparationsoluble proteins from root and leaf samples were resolved by SDS-PAGE and bands were cut in the range 25-30 kDa.
After washing with water and then 25 mM NH 4 HCO 3 gel bands were destained twice with 1 ml of CH 3 CN and dried at room temperature.
Disulphide bridges were reduced using 10 mM dithiothreitol at 56 C for 45 min and cysteine were alkylated using 55 mM iodoacetamide for 30 min in darkness. Gel bands were washed twice with 50% (vol/vol) CH 3 CN in 25 mM NH 4 HCO 3 , then dehydrated with CH 3 CN, and finally dried at room temperature. In-gel protein digestion was Mass tolerance was set at 10 ppm on full scans and 0.02 Da for fragment ions. Proteins were validated once they contained at least two peptides with a p-value <.05. Two additional filters were used to improve the identification of trimethylated Lys peptides: 1/selection of peptides with Mascot score ≥30, 2/discrimination of Lys trimethylation (mass shift of 42.04695) and Lys acetylation (mass shift of 42.01056) using a mass tolerance at 2 ppm. Ambiguous peptides were eliminated and spectra of interest were checked manually to confirm their sequence and the nature of modifications. Similar parameters were used to identify trimethylated Lys peptides in samples from A. halleri and A. lyrata but MS data were searched using a local database built using the A. lyrata genome resources (Alyrata_384_v2.1 from the Joint Genome Institute) (Hu et al., 2011;Rawat et al., 2015).

| Screening of A. thaliana mutants in protein Lys methyltransferase genes
Seeds of the T-DNA insertion lines in 23 KMT genes were obtained from the European Nottingham Arabidopsis Stock Centre. Mutants were genotyped by PCR using gene-and T-DNA-specific primers (Table S1). Amplicons were sequenced to map the insertion sites.
Seeds of Col-0 and homozygous KMT mutants were surface sterilized and sown onto Petri dishes containing half-strength Murashige and Skoog (MS/2) medium with 0.8% (wt/vol) agar. After 2 days of stratification at 4 C, plates were transferred to a growth chamber for 4 days (21 C, 70% air humidity, 18 hr of light per day, 80 μmol of photons m −2 s −1 ). Twenty seedlings per genotype were then transferred to square Petri dishes containing either MS/2 medium or MS/2 medium with 20 μM CdSO 4 . Plates were oriented vertically in the growth chamber and scanned (GS-800 scanner, BioRad) after 0, 3, 6, 8 and 10 days of treatment. The root length of each seedling was measured with the ImageJ software. Root length at Day 8 and root elongation rate (in cm day −1 ) between Days 3 and 8 were used as primary criteria to monitor the inhibitory effect of Cd. Also, the tolerance index (TI) to Cd was calculated for each line by dividing the primary parameter (length or elongation rate) measured in the presence of Cd by the one measured in the control condition. One hundred TI were calculated for each line by random sampling of one value in the Cd and control conditions, respectively, with replacement at each draw.

| Gene expression data mining
The expression of genes coding KMTs in A. thaliana seedlings challenged with Cd was analysed using data from the literature. First, we used our genome-wide CATMA microarray analysis of roots and shoots from 4-week-old A. thaliana plants exposed to 5 or 50 μM Cd for 2, 6 or 30 hr (GSE10675) (Herbette et al., 2006). Next, we collected four datasets from the Gene Expression Omnibus database.
They correspond to (a) 2-week-old seedlings grown in MS/2 agar plates and treated with 70 μM Cd for 2 hours (GSE90701) (Khare et al., 2017); (b) roots from 5-week-old plants grown in hydroponics and treated for 7 days with 1 μM Cd (GSE94314) (Fischer, Spielau, & Clemens, 2017); (c) roots from 3-week-old seedlings grown in hydroponics and challenged with 200 μM Cd for 6 hr (GSE22114) (Li et al., 2010) and (4) 7-day-old seedlings grown in MS/2 agar plates and treated for 6 hr with 200 μM Cd (GSE35869) (Jobe et al., 2012). Also, we analysed curated data from the comparative transcriptomic analysis of A. thaliana and A. halleri plants challenged with 10 μM or 25 μM Cd, respectively, in hydroponics for 2 hr (Weber, Trampczynska, & Clemens, 2006). Relative expression levels of KMT genes were retrieved from the different experiments, the ratio between Cd and control conditions were calculated from pairwise comparisons, and a non-parametric Student's t test was performed on log2 ratio to determine differentially expressed genes (DEGs, with p-value <.05) and the threshold was set at twofold (−1 ≤ log2 fold change ≥1). Finally, we used the RNA-seq data we obtained to analyse the tolerance strategies to Cd of two metallicolous populations of A. halleri (BioProject PRJNA388549) . In this experiment, gene expression in the I16 and PL22 populations was analysed after 10 days of treatment with 5 μM Cd in hydroponics. For the identification of DEGs, we selected genes with more than 10 read counts in any of the triplicate, applied a non-parametric t test (p-value <0.05) in pairwise comparisons, and used a 1.4-fold change threshold value (−0.5 ≤ log2 fold change ≥0.5).

| Determination of Cd by inductively coupled plasma mass spectrometry
Plant samples were dehydrated at 90 C, weighed for data normalization, and digested at 90 C for 4 hr in 65% (wt/wt) ultrapure HNO 3 .
Mineralized samples were diluted in 0.5% (vol/vol) HNO 3 and analysed using an iCAP RQ quadrupole mass instrument (Thermo Fisher Scientific GmbH, Germany). 111 Cadmium concentration was determined using a standard curve and corrected using an internal standard solution of 103 Rhodium added online.

| Statistical analysis
Non-parametric statistical analysis was performed on our datasets, which typically contain small sample sizes (n ≤ 20) and do not meet the assumptions of parametric tests (normal distribution and homogeneity of variance, as determined using the Shapiro-Wilk and Fisher tests, respectively). Multiple non-parametric comparisons were performed with the Dunnett's many-to-one test using the nparcomp package (Konietschke, Placzek, Schaarschmidt, & Hothorn, 2015) and the R computing environment. The Fischer's approximation method was used and the confidence level was set at 95%.

| Analysis of the patterns of Lys-methylated proteins in sensitive and tolerant Arabidopsis species exposed to Cd
We analysed the effect of Cd stress on the pattern of Lys-methylated proteins in roots and leaves of three Arabidopsis species showing contrasting Cd tolerance and accumulation. The analysis was done by immunoblotting and focused on Lys trimethylation on proteins other than histones because antibodies against mono-and dimethyl-Lys are less sensitive and specific than anti trimethyl-Lys antibodies (Alban et al., 2014). Moreover, the procedure used for protein isolation was not appropriate for the extraction of histones, which requires acidic or high salt conditions (Shechter, Dormann, Allis, & Hake, 2007). We used the Cd-sensitive species A. thaliana (ecotype Columbia, Col-0) and A. lyrata ssp. petraea, and the A. halleri species (Auby [AU] population) that displays Cd hypertolerance and hyperaccumulation traits . Plants were grown hydroponically for 5 weeks in a standard culture medium and then challenged with 5 μM CdSO 4 for 9 days.
In these conditions, the symptoms of Cd toxicity (growth inhibition, chlorosis and inhibition of photosynthesis) were visible for A. thaliana and A. lyrata plants, but not for A. halleri ( Figure S1). The patterns of Lys trimethylated proteins were complex with many polypeptides detected in root extracts and in leaf soluble extracts (Figure 1  . The patterns of Lys-trimethylated proteins in I16 and PL22 were similar to those observed for AU, and notably the doublet of proteins at 26-28 kDa in leaf soluble extracts was detected in control and stress conditions ( Figure S2). Together, these results indicate that Cd triggers changes in the steady-state level of some Lys-methylated proteins, with contrasting patterns depending on the Arabidopsis species and possibly their ability to tolerate and accumulate Cd.

| Identification of Lys-methylated proteins related to Cd stress in Arabidopsis
We used protein tandem mass spectrometry (MS/MS) to identify Lysmethylated proteins whose expression is modulated by Cd. The identification of Lys-methylated peptides by MS/MS is still challenging for many reasons (Wang, Wang, & Ye, 2017), including the low abundance and/or low methylation level of targets and the high false discovery rates for methylpeptides identification due to amino acid substitutions that are isobaric with methylation events (Hart-Smith, Yagoub, Tay, Pickford, & Wilkins, 2016;Ong, Mittler, & Mann, 2004).
To address this challenge, we focused on the identification of the abundant doublet of trimethylated proteins at 26-28 kDa for which the expression pattern was potentially interesting regarding Cd stress ( Figure 1). We used a filtering procedure adapted from Alban et al. (2014) to identify Lys-trimethylated peptides with high confidence.
Also, MS/MS data from A. lyrata and A. halleri were searched against a database built from the A. lyrata genome, and not against the A. thaliana genome, to improve the identification of Lys-methylated peptides. Using this procedure, we were able to identify Lystrimethylated peptides belonging to nine proteins in the gel bands of interest in the range 25-30 kDa (Table 1, Table S2 and Figure S3).  In leaf samples, where the immunodetection of the doublet of Lys-methylated proteins is species and condition dependent ( Figure 1a), we identified trimethylated Lys residues in EEF1A, RPL10A and six additional proteins ( Table 1). Methylation of chloroplastic fructose 1,6-bisphosphate aldolases at a specific Lys residue (Lys395) was reported earlier (Alban et al., 2014;Ma et al., 2016;Mininno et al., 2012), while the other proteins were not previously known to be methylated.
We compared western blot and MS/MS analyses to try to assign the major trimethylated proteins at 26-28 kDa. The detection pattern of peptides from RPL10A bearing a trimethylated Lys90 in root and leaf samples (Table 1)

| Expression of A. thaliana genes coding protein Lys methyltransferases in response to Cd
In  (Li et al., 2010). In these conditions, Cd triggered the upregulation of 12 genes and the downregulation of three genes in roots (Table S3). Among these genes, only SBS7 was differentially regulated at a lower Cd concentration. Also, the expression of SDG29 was upregulated following a short-term exposure to Cd. Together, these data indicate that the expression of a limited number of KMT genes is influenced by Cd in A. thaliana.

| Expression of genes coding protein Lys methyltransferases in A. halleri populations with different properties of Cd accumulation
To analyse whether Cd could modify the expression of genes coding KMTs in the Cd-tolerant species A. halleri we first used the comparative transcriptomic analysis from Weber et al. (2006). In this study, in  (Figure 2c). In PL22, three genes were induced by Cd in roots (SBS2 and SBS9) or in shoots (SDG52), and the SBS5 gene was downregulated by Cd in roots. The four genes regulated by Cd are predicted to code for KMTs modifying non-histone substrates (Serre et al., 2018), suggesting that these methylation events could be related to the tolerance and/or accumulation properties of the metallicolous PL22 population.

| Identification of protein Lys methyltransferase mutants from A. thaliana with altered tolerance to Cd
We used a screening procedure with knockout mutants to determine whether some KMT genes could play a role in the response of A. thaliana to Cd. We included only genes coding for KMTs modifying, or predicted to modify, non-histone substrates (Serre et al., 2018). Our selection comprised all genes (11) coding SBS enzymes and 15 genes coding SDG enzymes from Classes VI and VII. SDG enzymes from Classes I to V are known to methylate histones and some of them also accept non-histone substrates (Serre et al., 2018). Genes coding these enzymes were not included in our analysis since mutations in KMTs acting on histones, or on histones plus non-histone substrates, can lead to pleiotropic effects (e.g., Ndamukong et al., 2011), thus complicating the interpretation of the screening results. We obtained homozygous T-DNA insertion lines disrupting 23 of the selected genes (Table S1). Three genes could not be retained for the screening, of which PAP7 for which the mutation is lethal in photoautotrophic conditions (Grubler et al., 2017).
We analysed mutant seedlings for root growth inhibition by Cd, which is a simple and efficient method to assess tolerance to a toxic element (Remy & Duque, 2016). The procedure was set up using Col-0 seedlings and the cad2.1 null mutant that is hypersensitive to Cd (Howden, Andersen, Goldsbrough, & Cobbett, 1995). In brief, 4-dayold seedlings were transferred to MS/2 medium supplemented or not with 20 μM CdSO 4 and grown vertically for another 10 days in photoautotrophic conditions (no source of reduced carbon added to the medium) (Figure 3). Root length at Day 8 and root elongation rate between Days 3 and 8 were used as primary criteria to assess tolerance of the mutant lines to Cd (Figure 3). To address line-dependent differences in root growth that could interfere with the interpretation of the screening we also calculated the TI for the two primary parameters, which corresponds to the ratio between the values in Cdcontaining over CM (Metwally, Safronova, Belimov, & Dietz, 2005) ( Figure 3). The concentration of Cd in the medium (20 μM) was selected to produce a significant root growth inhibition (TI about 0.5) and to allow the identification of insertion lines that are either more tolerant or more sensitive to Cd than Col-0 in our experimental conditions.
The results of the screening procedure have been summarized in a heatmap displayed in Figure 4. Mutants were clustered in three mains categories. First, the calculated TIs for 14 insertion lines were comparable with the Col-0 ecotype. Second, five mutants (sdg51, F I G U R E 2 Expression of genes coding KMTs in the PL22 and I16 metallicolous populations of A. halleri challenged with Cd. RNAseq analysis was performed using root and shoot samples from plants exposed to 5 μM CdSO 4 for 10 days . (a) Principal component analysis of KMT genes expression in roots and shoots. (b) Differentially expressed KMT genes according to the genetic unit (PL22 vs. I16). The ratio between the steady-state expression level of each KMT gene in PL22 over I16 was calculated in all conditions (root ± Cd, shoot ± Cd) and differentially expressed genes (DEGs) were selected using a log2 fold-change ≥0.5 or ≤−0.5. (c) Differentially expressed KMT genes according to the Cd treatment (Cd vs. control). The ratio between the steady-state expression level of each KMT gene in Cd-treated over untreated plants was calculated in all conditions (root and shoot from I16 and PL22) and DEGs were selected using a log2 fold-change ≥0.5 or ≤−0.5. CM, control medium; Cd, medium containing 5 μM CdSO 4 sdg52, camkmt1, sbs7 and sbs9) displayed a higher tolerance to Cd than the wild type. Third, four mutants (sdg50, sbs2, sbs6 and sbs8) were found more sensitive to Cd than the wild type. Together, the screening procedure allowed for the identification of 9 out of 23 insertion mutants with a tolerance to Cd that is significantly different from that of the wild-type ecotype, suggesting that protein Lys methylation is part of the responses used by A. thaliana to cope with Cd stress.

| Characterization of a Cd-tolerant mutant deficient in calmodulin Lys methyltransferase
Two mutants identified in the screening were selected for further investigations. The first insertion line, camkmt1, was found more tolerant to Cd than the wild type ( Figure 4)  and seedling tolerance to salt, heat and cold stress (Banerjee et al., 2013).
The tolerance to Cd of the camkmt1 knockout line was verified using root growth assays and seedling biomass measurements using variable concentrations of the toxic metal (from 5 to 20 μM). For root elongation inhibition, the improved tolerance of camkmt1 was significant only at the highest Cd concentration (Figure 5a). For seedling growth inhibition, the inhibitory effect of Cd on biomass was significantly less important for camkmt1 than for the wild type at 10 and 20 μM Cd (Figure 5b). CaMKMT is involved in the methylation of the major calcium (Ca) sensor CaM (Banerjee et al., 2013) and Ca is known to alleviate Cd toxicity (Baliardini, Meyer, Salis, Saumitou-Laprade, & Verbruggen, 2015;Suzuki, 2005). Consequently, the tolerance of cam-kmt1 was analysed using a fixed concentration of Cd (20 μM) and fluctuating concentrations of Ca (0.5, 1 and 1.5 mM). Changes in Ca availability did not modify the growth of seedlings in the absence of Cd ( Figure 6). The inhibition of root elongation and seedling biomass by Cd was inversely correlated to Ca concentration in the medium.
Also, the camkmt1 line was found significantly more tolerant to Cd than the wild type at each Ca concentration tested ( Figure 6).
Together, these data validated our screening approach and confirmed the identification of a Cd-tolerant A. thaliana mutant affected in the methylation of CaM.
Then, we used inductively coupled plasma mass spectrometry (ICP-MS) to determine whether the difference in Cd-tolerance of camkmt1 could be due to changes in its capacity to take up the element from the environment. Cadmium was measured in roots and shoots of plants grown in hydroponics and challenged with various Cd concentrations (0.2, 1 and 5 μM) for 7 days. There was no significant difference in the absorption and translocation of Cd in camkmt1 as compared to Col-0 ( Figure S6). Thus, the tolerance to Cd of camkmt1 was not due to changes in Cd accumulation but rather to an improved capacity to cope with the toxic element.

| Characterization of a Cd-sensitive mutant affected in the protein Lys methyltransferase SBS2
The sbs2 line was selected for further investigations because it is more sensitive to Cd (Figure 4) and the SBS2 gene is upregulated in the roots of the A. halleri PL22 population challenged with Cd ( Figure 2c). Yet, the function of the SBS2 gene is unknown.
F I G U R E 5 Tolerance to Cd of the protein Lys methyltransferases camkmt1 and sbs2 mutants. Four-day-old seedlings were transferred onto MS/2 medium containing various amount of CdSO 4 and grown vertically for 10 days in photoautotrophic conditions. (a,c) Dosedependent inhibition of root growth by Cd. Root length was measured at Day 8. (b,d) Dose-dependent inhibition of seedling biomass by Cd. Seedling fresh weight was measured at Day 10. Each distribution represents n = 20 seedlings. Statistical significance determined using a non-parametric Dunnett's test is shown, with p < .05 (*), p < .01 (**) and p < .001 (***) F I G U R E 6 Tolerance to Cd of the calmodulin Lys methyltransferase camkmt1 mutant. Four-day-old seedlings were transferred onto MS/2 medium containing various amount of CdSO 4 and CaCl 2 and grown vertically for 10 days in photoautotrophic conditions. (a) Calcium-dependent inhibition of root growth by Cd. Root length was measured at Day 8. (b) Calcium-dependent inhibition of seedling biomass by Cd. Seedling fresh weight was measured at Day 10. Each distribution represents n = 20 seedlings. Statistical significance determined using a non-parametric Dunnett's test is shown, with p-value <.01 (**), and p-value <.001 (***) Similar to camkmt1, we first confirmed the phenotype of sbs2 by measuring the inhibition of root elongation and seedling growth with different concentrations of Cd. Root growth of sbs2 was significantly more inhibited by Cd than the wild type at all concentrations tested (5-20 μM; Figure 5c). Also, the biomass of sbs2 seedlings was lower than Col-0 seedlings for the three concentrations tested (Figure 5d), confirming the Cd-sensitive phenotype of sbs2.
To gain insight into the role of the SBS2 gene in the response to Cd, we selected a second independent insertion line, referred to as sbs2b. The T-DNA insertions were located in the fourth exon of SBS2 for sbs2b and downstream the fourth exon for sbs2, in a region that is either an intron or the 3 0 untranslated region of SBS2 transcript variants (Figure 7). Reverse transcription-PCR analysis indicated that the two lines are loss-of-function alleles with no detectable SBS2 transcripts. Also, root growth assays showed that sbs2b behaved as sbs2 and was less tolerant to Cd than wild-type seedlings (Figure 7). Together, these data indicated that the invalidation of the SBS2 gene is responsible for an increased sensitivity to Cd.
We analysed whether the uptake and distribution of Cd was affected in sbs2. The Cd content in roots and shoots of sbs2 was similar to that of Col-0 at any Cd concentration tested ( Figure S6). Thus, the increased sensitivity to Cd of sbs2 was not associated with an increased absorption of the toxic element from the medium but rather to a reduced capacity to deal with its deleterious effects.
Finally, as a preliminary approach to identify the substrate of the SBS2 methyltransferase, we used western blot analysis to compare the patterns of Lys-trimethylated proteins in sbs2 and Col-0. We could not observe any significant decrease in band intensity (hypomethylation) in sbs2 relative to Col-0 in soluble or membrane proteins from roots and shoots of seedlings grown in standard conditions ( Figure S7). Thus, the substrate of SBS2 is probably a low abundant protein that was not detectable by the current immunolabeling approach.

| DISCUSSION
The methylation status of Lys residues in proteins is controlled by KMTs and contributes to the regulation of protein properties in diverse biological processes. To address whether Lys methylation of non-histone proteins is important for metal tolerance in Arabidopsis species, we analysed the effect of Cd on the two partners participating in this PTM, that is, methylated proteins on the one hand and KMTs on the other hand. Using an immunoblotting approach, we showed that the Lys-methylation status of some proteins is influenced by a Cd stress in the roots and leaves of Arabidopsis (Figure 1).
Changes in methylation patterns were observed between Cd-tolerant and Cd-sensitive species and between treated and untreated plants.
This analysis provided the first evidence that the steady-state level of F I G U R E 7 Molecular characterization of the A. thaliana sbs2 and sbs2b mutants. (a) Structure of the SBS2 locus. Splicing variants predicted in the Araport database are shown (https://apps.araport.org/thalemine/portal.do?externalids = AT1G63855). Thin lines represent introns, grey boxes represent exons, thick dark lines represent untranslated regions and triangles indicate T-DNA insertion sites in the sbs2 (GK-911F08) and sbs2b (SALK_037552) mutants. Primers A and B are indicated by arrows. (b) Amplification of SBS2 transcripts by PCR. PCR was performed using reverse transcribed total RNA from 10-day-old seedlings with Primers A (ATGATGACTACTACGACGACGAC) and B (CTCAATACGATCTCAACCAACTGA) for SBS2, and ACT7-F (ACATCGTTCTCAGTGGTGGTCC) and ACT7-R (ACCTGACTCATC-GTACTCACTC) for ACTIN7. PCR products were resolved by agarose gel electrophoresis. Two major amplicons of 700-800 bp were amplified in Col-0, cloned and sequenced. They correspond to splicing variants 1 and 4. (c) Tolerance to Cd of the sbs2 and sbs2b mutants. Four-day-old seedlings were transferred onto MS/2 medium supplemented or not with 20 μM CdSO 4 and grown vertically for 10 days in photoautotrophic conditions. Root length was measured at Day 8. Each distribution represents n = 20 seedlings. Statistical significance determined using a non-parametric Dunnett's test is shown, with p-value <.001 (***) some methylproteins, or the stoichiometry of Lys methylation of these proteins, could be linked with metal stress and with the genetic diversity of the Arabidopsis species. Then, we used MS/MS to identify Lys-trimethylated proteins of low molecular weight (25-30 kDa) that displayed different expression profiles in the leaves of A. thaliana, A. lyrata or A. halleri in response to Cd. Using a specific pipeline for the identification of Lys trimethylation events we identified 12 methyl sites in nine proteins (Table 1). Six of these proteins and eight of the Lys-methylated sites were not previously known, illustrating the depth of the analysis. In addition, by using genomic resources of A. lyrata for the assignment of MS/MS spectra from A. lyrata and A. halleri samples, we were able to identify, for the first time, post-translationally modified proteins in these model species.  (Carvalho et al., 2008). We identified two Lys trimethylation sites in Arabidopsis RPL10A proteins. The first one (Lys46) has been previously identified as monomethylated by the RKM5 methyltransferase in the homolog of RPL10A from yeast (Webb et al., 2011). Trimethylation of Lys46 was detected only in the roots of A. thaliana grown in control conditions and, so, has probably no link with the response to metal stress. This assumption is supported by the observation that a mutation in the SBS1 gene, the ortholog of RKM5 ( Figure S5), did not change the tolerance to Cd of A. thaliana seedlings (Figure 4). The pattern of trimethylation of the second residue (Lys90) in RPL10A in leaves was influenced by Cd stress in a species-dependent manner ( Table 1). The functional outcome of Lys90 trimethylation in RPL10A is not known; the modification may contribute to the optimization of ribosomal function or may affect its non-canonical functions, as previously showed for RPL10A phosphorylation (Carvalho et al., 2008).
We also analysed the expression of genes coding protein Lys methyltransferases in response to Cd in wild-type A. thaliana and in populations of A. halleri with different capacities to tolerate and accumulate the toxic metal. In A. thaliana, the steady-state level of only two KMT genes is regulated by moderate concentrations of Cd (Table S3). In A. halleri, we showed that Cd induces a significant change in the expression of four KMT genes in the PL22 population, but none in the I16 population (Figure 2c). The transcriptomic, ionomic and metabolomic analyses of these two metallicolous populations from different European genetic units indicated that distinct strategies driven by different sets of genes have evolved for the adaptation to high Cd  or high zinc in soils . Since PL22 accumulates Cd in roots and shoots whereas I16 behaves as a Cd excluder, both in situ and in hydroponic conditions, these results suggest that the regulation of KMT genes expression in PL22 could be correlated with the level of Cd that is taken up from the environment and translocated to shoots.
The substrates of the KMTs encoded by these four genes (SDG52, SBS2, SBS5 and SBS9) are likely not histones ( Figure S5), suggesting that Lys methylation of non-histone proteins could contribute to the regulation of cellular mechanisms involved in Cd accumulation or detoxification in the PL22 population. The analysis of DEGs between I16 and PL22, regardless of the presence of Cd in the culture medium, identified 22 KMT genes (Figure 2b). This suggests that Lys methylation of histones and non-histone substrates could be part of the diverging adaptation strategies of metallicolous populations. The expression of KMT genes coding enzymes of the SDG family has been previously analysed in cotton plants stressed with high temperature (Huang et al., 2016) and in foxtail millet under different abiotic stresses (Yadav, Muthamilarasan, Dangi, Shweta, & Prasad, 2016). In these studies, the expression pattern of some KMT genes was significantly changed in stress conditions. These data, together with our results, suggest that protein Lys methylation could play a role in the responses of plants to a variety of abiotic stresses.
Finally, we used a screening procedure based on root growth inhibition assays to determine whether some KMT genes could be important for A. thaliana to cope with Cd. We showed that 9 out of 23 insertion mutants displayed a tolerance to Cd that was significantly different from that of wild-type seedlings (Figure 4). These KMTs belong to the SDG class VII (SDG50, SDG51 SDG52) and to the SBS family (SBS2, SBS6, SBS7, SBS8, SBS9 CaMKMT) and are known, or predicted, to modify non-histone targets (Serre et al., 2018), suggesting that Lys methylation of non-histone proteins is one of the regulatory mechanisms modulating the response of A. thaliana to Cd stress.
Two of the identified mutants were further investigated. The camkmt1 line is unable to methylate CaM (Banerjee et al., 2013) and is more tolerant to Cd than the wild type at each Ca concentration tested ( Figure 5). Cadmium is known to interfere with Ca homeostasis and the Ca/CaM system has been hypothesized to participate in heavy metal signalling (Baliardini et al., 2015;Gallego et al., 2012).
More generally, CaM has been implicated in the response and recovery to different stresses and CaM methylation has been proposed to play a regulatory role in these processes. Indeed, a camkmt1 null mutant displayed increased tolerance to salt, heat and cold stress whereas lines overexpressing CAMKMT were hypersensitive to these stresses (Banerjee et al., 2013). Together, these data suggest that Lys methylation of CaM also plays a role in the signalling cascade triggered by Cd, probably at a level that is common between different abiotic stresses. The precise role of Lys methylation in the modulation of CaM activity is still unclear.
Our data also indicated that the invalidation of the SBS2 gene in A. thaliana is associated with a decreased capacity to cope with Cd ( Figure 7). Also, the expression of SBS2 was increased in the roots of A. halleri PL22 plants challenged with Cd (Figure 2c), suggesting that the methylation reaction catalysed by SBS2 is useful to limit the deleterious effects of Cd. The function of SBS2 is still not known in plants.
Its ortholog in animal cells is METTL23 ( Figure S5). METTL23 is located in the cytoplasm and the nucleus, interacts with a subunit of the GA-binding protein transcription factor, but its target(s) has not been yet identified (Bernkopf et al., 2014;Reiff et al., 2014). The profiling of methylproteins in sbs2 and Col-0 by western blot was not sensitive enough to detect any change between the two lines ( Figure S7), providing no clues to the nature of the substrate(s) of the SBS2 enzyme. The identification of the target(s) of SBS2 is the next step to gain insight into the role of this methylation event under favourable growth conditions and in the response to Cd stress.
Together, the data presented in this study provide the first evidence for a link between the methylation status of Lys in non-histone proteins and the response of plants to a stress induced by Cd. They pave the way for the identification of cellular mechanisms that are regulated by protein Lys methylation and are important for plants to cope with toxic elements. To reach this goal one has to identify the KMT/substrate relationships to be able to modulate the methylation status of protein targets in vivo. To summarize, this work suggests that the characterization of the KMT involved in the methylation of Lys90 in RPL10A and the identification of the substrates of the KMTs encoded by genes that are modulated by Cd and play a role in Cd tolerance will provide significant insights into the role of protein Lys methylation during metal stress.