RESULTS
Biothiol concentrations in roots and shoots varied greatly depending onArabidopsis genotypes according to the HPLC-DAD analysis (Fig. 1; Supplementary Table 2). Glutathione was present in all genotypes, but it was at remarkably low concentrations in the γECS mutants cad2-1 ,pad2-1 and rax1-1 , when compared with the levels of the wild type Col-0 (all mutants were unequivocally identified by PCR amplification and sequence alignment; Supplementary Fig. 3). In the phytochelatin-defective mutant cad1-3 , the GSH concentration in shoots was almost 2-fold compared with the wild-type. In the presence of 3 µM Hg, PC2 ((γGlu-Cys)2-Gly), PC3 ((γGlu-Cys)3-Gly) and PC4 ((γGlu-Cys)4-Gly) were found in Col-0 roots, whereas only PC2 and PC3appeared in rax1-1 roots. However, none of those PCs were observed in cad2-1 and pad2-1 mutants. As expected, we could not detect PCs in cad1-3 under Hg stress, in spite of the fact that GSH concentrations were the highest observed both in shoots and roots, doubling the concentration found in Arabidopsis Col-0 (Fig. 1; Supplementary Table 2) .
The decrease of GSH levels in pad2-1 , cad 2-1 andrax1-1 and the inability to synthesize PCs in cad1-3 were accompanied by significant changes in ascorbic acid (ASA), reduced (GSH) and oxidized glutathione (GSSG) concentrations, as measured by HPLC-ESI-MS(TOF) (Table 1). Arabidopsis mutants treated with 3 µM Hg had ASA concentrations in roots well above of values found in Col-0, which almost doubled in shoots. On the other hand, GSH concentrations followed the same pattern found using HPLC-DAD, with pad2-1 ,cad2-1 and rax1-1 having the lowest values both in roots and shoots, whereas the GSH concentration in cad1-3 was 2-fold higher than Col-0. Exposure to 3 µM Hg led to general increases in GSH concentrations in shoots and roots of all mutant genotypes, effect particularly intense in cad1-3 . With respect to GSSG, concentrations were one order of magnitude lower than those of GSH, but they changed with a similar pattern. As a result, there were minimal changes in the relative content of GSSG irrespective of genotype and occurrence of Hg stress.
The marked changes in ASA and GSH/GSSG contents observed in response to Hg suggested possible alterations in the redox balance of mutant shoots and roots. We firstly analysed the concentrations of Hg in shoots and roots of Arabidopsis , which accumulated largely in roots (shoots Hg concentrations was less than 1% of that found in roots; Fig. 2a). All γECS mutants had similar Hg levels in roots, which were approximately 50% of the concentration found in Col-0 and cad1-3plants (Fig. 2a). However, Hg concentrations in shoots were not statistically different between genotypes. In parallel, we determined chlorophyll fluorescence parameters, and observed that non-photochemical quenching was severely impaired in γECS and PCS mutant genotypes both in control and 3 µM Hg-treated plants (Fig. 2b), confirming that limiting biothiols metabolism led to stress in leaves. GR activity, an enzyme specifically sensitive to this Hg (Sobrino-Plata et al. 2009), was not affected in roots (Fig. 3a), but was impaired by Hg in Col-0 andcad1-3 roots with an almost complete inhibition in cad2-1 ,pad2-1 , and rax1- 1 mutants (Fig. 3a). Despite such inhibition, the amount of GR protein did not change appreciably in shoots and roots even under Hg-stress independently of the genotype (Fig. 3b). With regard to γECS in shoots, protein accumulation under control conditions was remarkably lower (40-50%) in γECS andcad1-3 mutants than in Col-0. Mercury stress also led to a marked decrease in Col-0 leaves, reaching similar values in all genotypes (Fig. 3b). However, there were no differences among genotypes and Hg stress levels in root γECS, probably due to the low signal obtained by α-γECS immunodetection (high background; Fig. 3b).
Our previous study established that part of the ability of plants to withstand Hg toxicity depends on the formation of Hg-PCs complexes, such as HgPC2 (Hg(γGluCys)2Gly) and HgPC3 (Hg(γGluCys)3Gly) (Carrasco-Gil et al., 2011). Full HPLC-ESI-MS(TOF) analysis of biothiol ligands and Hg-biothiol complexes (Hg-PCs) in shoots and roots of showed clear differences between all studied Arabidopsis genotypes (Fig. 4). In shoots, we could only detect free PC2([PC2-H]; m/z 538.1) and PC3 ([PC3-H];m/z 770.2) ligands, whereas in roots there were oxidized variants of free PCs, such as ([PC3oxd-H];m/z 768.2), Hg-PC complexes like HgPC2([HgPC2-H]; m/z 738.1) and HgPC3 ([HgPC3-H];m/z 970.1). The graphical table included in Fig. 4b shows the groups of free ligands, oxidised PCs and Hg-biothiol complexes, found in shoots and roots of all Arabidopsis genotypes. The results inrax1-1 and cad2-1 roots closely resembled those found for Col-0, where we detected [HgPC2-H] and [HgPC3-H], albeit with a rather weak signal (data not shown). On the other hand, in pad2-1 we only found GSH ([GSH+H]+; m/z 308.1) and GSSG ([GSSG+H]+; m/z 613.3) in roots and shoots, which were better detected in positive mode, in addition to PC2, that was just over the background signal. As expected, cad1-3 did not accumulate free PCs or Hg-PCs complexes.
Recent studies indicated that toxic elements (Cd and As) are chelated with PCs in roots impeding translocation to shoots and potentially helping plants to attenuate stress, in a manner that metal(loid)-PCs complexes would mass in root vacuoles (Liu et al., 2010; Mendoza-Cózatl et al., 2008) However, to some extent metal(loid)s may travel to shoots bound to organic ligands such as PCs (Shi et al., 2019). To determine whether Hg had a similar behaviour, we studied the possible occurrence of biothiols and Hg-PCs complexes in xylem sap by HPLC-ESI-MS(TOF) using both positive and negative modes. Since Hg blocks water movement through plant vascular tissues, we used a Schölander pressure chamber to generate sufficient root pressure to impulse xylem water movement. All xylem sap samples were checked for phloem or broken cells fluids contamination by measuring MDH activity, which indicated that cross-contamination was negligible in all Arabidopsis genotypes under Hg stress (Supplementary Fig.3). The compounds GSH ([GSH+H]+; m/z 308.1) and GSSG ([GSSG+H]+; m/z 613.3) appeared in the xylem sap of all genotypes, albeit signals were lower in γECS mutants (data not shown). The characteristic PC2 peak ([PC2+H]+; m/z 540.1) appeared in xylem sap of Col-0 and, at very low intensity, also inrax1-1 (Fig. 5). This compound coeluted with another ofm/z 538.1, which was tentatively identified as oxidized PC2 (PC2oxd). However, PC2 or PC2oxd were not detected incad2-1 , pad2-1 and, as expected, PCS mutant cad1-3 .
To confirm the nature of PC2oxd we run in parallel a hydroponic experiment with Col-0 Arabidopsis treated with 10 µM Cd for 72 h. In this case, we got a better signal in MS(TOF) in negative mode with a m/z 536.1 ([PC2oxd-H]) (Supplementary Fig. 4a); molecular ion that was subjected to tandem MS (-MS2), and was compared with those obtained using PC2 (m/z 538.13) and PC2oxd (m/z 536.1) standards, which had characteristic daughter ions atm/z 254.1 and 128.0 (Supplementary Fig. 4c). Incidentally, we were unable to observe any Cd-PC complex, in spite of using ESI-MS(TOF) settings appropriate for detection of CdPC2, as we obtained the characteristic peaks associated with the natural Cd isotopic distribution (major [CdPC2-H] peak at m/z650.0) by direct injection of a Cd:PC2 standard (Supplementary Fig. 4b).
In Col-0 xylem sap, along with to PC2 and PC2oxd we found only a compound with the characteristic Hg-isotopic fingerprint that could correspond to Hg-PCs complexes, which was tentatively assigned to HgPC2, eluting separately from free biothiol ligands (Fig. 5a). The MS(TOF) spectrum (in positive mode) of the detected compound ([HgPC2+H]+; m/z 740.1) fitted well with theoretical data and also with a Hg:PC2standard mixture (1:1) (Fig. 5b). The identity of the m/z 740.1 ion peak of Col-0 xylem samples was confirmed using tandem MS/MS analysis. The same Hg:PC2 standard mixture was used to set up analytical conditions, and the m/z 740.1 mother ion was selected and sent to the collision cell for fragmentation (MS2). Several major daughter ions appeared withm/z 609.1, 536.1 and 508.1 both in the HgPC2standard and the Col-0 xylem sap (Fig. 5c). Some of these ions were tentatively identified by comparing with those detected in Hg-biothiol complexes analysis as follows: m/z 609.1 was assigned to [HgPC2-Glu]+; m/z 536.1 matched [PC2oxd-2H]+, andm/z 508.1 was assigned to [HgGSH+H]+. Further identification of the m/z 536.1 ion, with the highest intensity peak, was obtained after a second fragmentation (MS3) resulting in various ions. The MS3 spectra of both the xylem sap and the standard mixture were also very similar, with a major m/z 507.1 daughter ion (possibly [GSH-H]+), with a second m/z489.1 ion also present in both samples (Fig. 5d). Therefore, we can assert that HgPC2 complexes could be transferred from roots to shoots via xylem flux, process that did not occur inrax1-1 , cad2-1 and cad1-3 mutants. Nevertheless, we could not determine to what extent Hg flows to shoots via xylem, since our ICP-MS analysis failed to detect Hg above background levels, probably due to the small volume of sample collected (10-50 µL).
In view of the relevant role that biothiol metabolism has in tolerance to Hg and Hg speciation in plants, we analysed the expression pattern of 20 genes involved in sulphur uptake, assimilation and incorporation to biothiols under Hg-stress (Gigolashvili & Kopriva, 2014). The expression pattern was organ-dependent, with some genes being over-expressed in the shoots of certain mutants treated with Hg (Fig. 6), whereas in the roots we only detected gene down-regulation under Hg stress (Fig. 7). Regarding transcription factors in shoots, MYB28was induced only in the γECS-mutants cad2-1 and pad2-1under Hg exposure, whereas MYB51 was suppressed in rax1-1 . On the other hand, both MYB28 and MYB51 were down-regulated in roots under Hg-stress, especially in rax1-1 andcad1-3 mutants (Fig. 7). We also found significant down-regulation of SLIM1 in roots of all mutant Arabidopsisgenotypes (Fig. 7; Suppl. Tables 3 and 4).
Among the genes involved in sulphur incorporation and assimilation in shoots, the sulphur transporter SULTR1;2 had the highest over-expression in Hg-treated cad2-1 , rax1-1 andcad1-3 plants, whereas a strong repression was observed in Col-0 (Fig. 6; Suppl. Table 3). A similar repression appeared in Col-0 for ATP sulphurylase (ATPS3 ) and APS reductase (APR1 andAPR3 ) (Fig. 6). On the other hand, ATPS4 (only inpad2-1 ), APR1 , APR2 and APR3 (only inrax1-1 ), were over-expressed in pad2-1 , rax1-1 andcad1-3 plants treated with 3 µM Hg. With regard to GSH and PCs metabolism, we only observed a minor down-regulation under Hg stress, particularly significant for cad2-1 and pad2-1O-acetylserine (thiol) lyase (OASTLA and OASTLB ) genes. Interestingly, expression of the phytochelatin synthase genesPCS1 and PC2 decreased in leaves in Hg-treated plants, being particularly significant in pad2-1 , rax1-1 andcad 1-3 (Fig. 6, Supplementary Table 3). Finally, in roots under Hg stress we only observed significant gene down-regulation, mostly in the mutant genotypes. Especially relevant was the down-regulation of sulphate transporters, including a remarkable decrease forSULTR1;2 in cad2-1 , rax1-1 and cad1-3 (Fig. 7). The expression of other sulphur transporters decreased, including that of SULTR2;1 in rax1-1 and cad1-3 , andSULTR3;5 , which was very intense in all Hg-exposedArabidopsis genotypes. With regard to sulphur assimilation genes, the most consistent changes occurred in cad1-3 , whereATPS1 , ATPS3 , SiR , OASTLB , OASTLC ,γECS , GSH-S , PCS1 and PCS2 expression decreased in plants treated with 3 µM Hg (Fig. 7, Supplementary Table 4).