The histone deacetylase HDA19 controls root cell elongation and modulates a subset of phosphate starvation responses in Arabidopsis

The length of root epidermal cells and their patterning into files of hair-bearing and non-hair cells are genetically determined but respond with high plasticity to environmental cues. Limited phyto-availability of the essential mineral nutrient phosphate (Pi) increases the number of root hairs by longitudinal shortening of epidermal cells and by reprogramming the fate of cells in positions normally occupied by non-hair cells. Through analysis of the root morphology and transcriptional profiles from transgenic Arabidopsis lines with altered expression of the histone deacetylase HDA19, we show that in an intricate interplay of Pi availability and intrinsic factors, HDA19 controls the epidermal cell length, probably by altering the positional bias that dictates epidermal patterning. In addition, HDA19 regulates several Pi-responsive genes that encode proteins with important regulatory or metabolic roles in the acclimation to Pi deficiency. In particular, HDA19 affects genes encoding SPX (SYG1/Pho81/XPR) domain-containing proteins and genes involved in membrane lipid remodeling, a key response to Pi starvation that increases the free Pi in plants. Our data add a novel, non-transcriptionally regulated component of the Pi signaling network and emphasize the importance of reversible post-translational histone modification for the integration of external signals into intrinsic developmental and metabolic programs.

The formation of root hairs, tip-growing unicellular extensions of root epidermal cells emerging from specialized cells referred to as trichoblasts, is a well-explored model for the specification and differentiation of plant cells. In Arabidopsis and other members of the Brassicales, root hairs are arranged in a striped pattern of longitudinal files of hair and non-hair cells that is controlled by a sophisticated mechanism in which cell-to-cell-movement of transcriptional regulators and extracellular signaling pathways regulate the allocation of the alternating binary cell fate [1][2][3] . A positional signal, probably induced by JKD in the root cortex 4 , is perceived by the leucine-rich repeat receptor kinase SCM localized on the plasma membrane of epidermal cells 5,6 . The signal is supposedly more abundant in epidermal cells that contact an anticlinal wall of underlying cortical cells (H position) owing to the larger surface contact area relative to cells that are located over periclinal cortical cells (N position). This bias derived from cortical cells is amplified by a feedback mechanism that dictates higher SCM abundance in H positions relative to cells in N positions 7 . SCM suppresses the non-hair transcription factor WER, which leads to higher WER abundance in non-hair cells relative to hair cells. Together with the bHLH MYC transcription factor

Results
Mutations in Class I HDAs compromise the Pi deficiency-induced increase in root hair density. Pi deficiency affects the development of epidermal cells in a nutrient-specific manner, resulting in an increase in both length and density of root hairs 23,24 . The molecular mechanism governing these changes remains elusive. Root hair patterning under control (Pi-replete) conditions is dependent on correct acetylation of the N-terminal tails of core histones 20,21 . To investigate whether reversible histone acetylation is also important for the Pi deficiency-induced root hair phenotype, we tested homozygous mutants or knock-down lines defective in the expression of HDA18 (class II of the RPD3/HDA1 superfamily), HDA17, an unclassified member of the RPD3-like superfamily, and the class I members HDA6, HDA9 and HDA19 for their root hair phenotypes under control and low Pi conditions. All lines showed a significant increase in root hair density when grown on low Pi media. This increase was less pronounced in all three class I HDAs, which was in particular evident in the hda19 mutant (Fig. 1a). HDA19-RNAi plants showed a response similar to that of the hda19 mutant. Transgenic plants over-expressing HDA19 driven by the 35S promoter (35S:HDA19) produced more root hairs under low Pi conditions when compared to the wild type (Fig. 1). Notably, HDA19-RNAi formed fewer root hairs than the wild type also under control conditions and mutants defective in the expression of HDA18 and HDA17 produced significantly more root hairs than the wild type under Pi-replete conditions. Root hair length was increased by more than two-fold upon Pi starvation; a Pi deficiency-induced increase was observed for all lines except for hda18 (Fig. 1b). Under Pi-replete conditions, only for HDA9-RNAi plants a deviation in root hair length from that of the wild type was observed.
HDA19 activity controls epidermal cell length. In Arabidopsis, the driving force for the increase in root hair density upon Pi deficiency is a diminished longitudinal elongation of trichoblasts 11 . To investigate whether the change in the frequency of root hairs caused by altered HDA19 expression is a consequence of altered trichoblast length, we analyzed the length of epidermal cells of the genotypes under investigation. To this end, we compiled micrographs of individual roots comprising the first six to eight millimeter of the root tip and measured the length of the cells in relation to their distance from the quiescent center. This procedure was repeated with three different roots for each genotype (wild type, HDA19-RNAi and 35S:HDA19) and growth type (Pi-replete and low Pi). The respective root hair phenotypes are shown in Fig. 2. Cell length measurements revealed that for both growth types HDA19 abundance is critical in controlling the length of both trichoblasts and atrichoblasts (Fig. 3). While knock-down lines produced markedly longer cells than the wild type, short epidermal cells were observed in 35S:HDA19 plants. HDA19 Pi-replete conditions, which were on average comparable to those of the wild type. The distribution of the cell lengths along the roots is shown in Fig. 3c-j.
Overexpression of HDA19 induces additional hair cell fate assignment in Pi-deficient plants. In addition to a decrease in cell length and, associated with this response, an increase in root hair density in the H position, growth on Pi-depleted media increases the probability of epidermal cells in the N position to adopt the H cell fate 24 . To investigate whether the increase in root hair density in 35S:HDA19 plants in response to Pi starvation is solely due to shortening of epidermal cells or whether additional H cell fate acquisition contributes to the phenotype, we conducted a detailed analysis of the frequency of root hairs in ectopic positions based on cross-sections. This analysis revealed a dramatic (33-fold) increase in the number of ectopic roots in 35S:HDA19 plants when grown on low Pi media. Although this massive increase was due to the very low number of ectopic roots in this genotype under Seedlings were grown on control and low Pi media for 14 days. Data represent means of ten plants for each genotype and growth type ± SE. The asterisks denote statistically significant differences to Pi-replete wildtype plants based on Student's t-test (*P < 0.05, **P < 0.01, ***P < 0.001). HDA19-Ri, HDA19-RNAi.  HDA19 abundance affects leaf responses to Pi starvation. A further hallmark of Pi-deficient plants is an accumulation of anthocyanin in leaves to protect nucleic acids and chloroplasts from photo-oxidative damage due to limited photosynthesis. In wild-type plants, a circa 4-fold increase in anthocyanin concentration was observed after transfer of the plants to low Pi media (Fig. 5a,b). Pi-deficient hda19 mutants showed a very small increase in anthocyanin levels over Pi-replete plants; in Pi-deficient HDA19-RNAi plants anthocyanin levels were reduced relative to the wild type when grown on Pi-replete or on low Pi media. Anthocyanin concentrations in 35S:HDA19 plants did not differ from that of the wild type ( Fig. 5a,b). Since anthocyanin synthesis is controlled by R3 and R2R3 MYB transcription factors, which also regulate epidermal cell fate (i.e. CPC and GL3), it was tempting to assume that HDA19 abundance altered root hair density by altering their expression. However, transcript abundance of CPC and GL3 did not differ significantly across the genotypes under investigation.
Determination of leaf Pi concentrations by ICP-OES did not reveal significant differences between the wild type and 35S:HDA19 plants under both growth conditions. In HDA19-RNAi plants, the Pi concentrations in leaves was significantly decreased under control conditions, but no changes were observed when plants were grown on low Pi media (Fig. 5c). Pi concentrations in roots did not differ between the genotypes. Iron accumulates in Pi-deficient plants 13 and may interfere with cell elongation. As anticipated from previous studies, the concentration of iron was dramatically increased under Pi-deficient conditions but did not vary much among the genotypes, except a significant increase in roots of 35S:HDA19 plants grown on low Pi media (Fig. 5d). No differences in leaf iron concentration were observed between the genotypes under investigation.

HDA19 modulate the expression of Pi-responsive genes.
To understand the molecular events that determine the phenotypes of lines with altered HDA19 expression, we conducted transcriptional profiling experiments of HDA19-RNAi and 35S:HDA19 plants grown under control and low Pi conditions. A subset of 86 Pi-responsive genes was expressed two-fold lower in HDA19-RNAi plants when compared to 35S:HDA19 lines (Supplemental Table 1). Among the genes with the highest difference was SPX3 that showed a 17-fold difference between the two genotypes. SPX1, which encodes a key regulator of the Pi starvation response, and another SPX (SYG1/Pho81/XPR) domain protein, At2g38920, were less Pi-responsive in HDA19-RNAi lines. Furthermore, MGD3, a key enzyme in galactolipid biosynthesis, and the purple acid phosphatase PAP23 were down-regulated in HDA19-RNAi plants. Galactolipid biosynthesis is induced by Pi deficiency as part of metabolic pathway during which phospholipids (PLs) in membranes are replaced with the galactolipid digalactosyldiacylglycerol and the sulfolipid sulfoquinovosyldiacylglycerol ('membrane lipid remodeling' 25 ). Excretion of PAPs aimed at re-mobilizing Pi represents a further key PSR response. PAP23 is co-expressed with SPX1/SPX3 and genes involved in galactolipid biosynthesis that were strongly induced upon Pi starvation such as MGD3. The phospholipase D PLDζ2 and the glycerophosphodiester phosphodiesterase GDPD6 showed reduced expression in the HDA19 knockdown line. Also PLDζ 2 and GDPD6 have been associated with the hydrolysis of PLs as part of membrane lipid remodeling 26,27 . Notably, induction of PLDζ2 upon Pi starvation was dependent Seedlings were grown on control and low Pi media for 14 days. Data represent means of ca. 40 cross-section from 10 plants for each genotype and growth type ± SE. Asterisks denote statistically significant differences to Pi-replete wild-type plants based on Student's t-test (*P < 0.05, **P < 0.01, ***P < 0.001).
on functional CPC, a positive regulator of the root hair cell fate 28 . GDPDs convert lysophosphatidylcholine derived from PLs to glycerol-3-phosphate, a precursor of galactolipids. CAR8, another repressed gene in the HDA19-RNAi line, is tightly co-expressed with genes involved in galactolipid biosynthesis. Five genes with reduced expression in HDA19-RNAi plants encode protein kinases (CRK24, CRK6, CRK45, HT1, At2g17170), indicating a possible post-transcriptional regulation of the cell length. Also of note, two genes involved in cell-to-cell signaling, RALFL9 and RALFL27, members of the peptide family with similarity to tobacco Rapid Alkalinization Factor (RALF), were lower expressed in HDA19-RNAi relative to 35S:HDA19 plants.
Consistent with a role of HDA19 as a gene repressor, slightly more genes (106) showed higher induction upon Pi starvation in HDA19-RNAi relative to 35S:HDA19 plants. The most pronounced differences were observed for several genes encoding unknown proteins (Supplemental Table 2). Interestingly, 11 genes in this suite (IGMT2, GSTL1, At3g01260, NMT3, SKOR, GSTF3, At5g22555, At2g43590, At4g01430, At4g19370 and FRD3) were previously defined as Fe-responsive 29 , while only one such gene (NAS2) was among the genes that showed lower expression in the HDA19-RNAi line. About half of the genes in this group were up-regulated upon iron deficiency and the other half showed reduced expression in iron-deficient plants, suggesting that a lack of functional HDA19 compromises cellular iron homeostasis. Consistent with a role in cell differentiation, several genes (20 out of 106) encode proteins that are predicted to be located in the extracellular region. Three genes (At5g48540, At1g07560 and ATPI4Kγ3) encode protein kinases.
To validate the observed changes in gene expression, we determined the transcript levels of selected genes by qRT-PCR. In roots of Ws-2, the wild type of the transgenic lines, SPX3 transcripts were dramatically increased by Pi deficiency. To allow comparison with the widely used Col-0 accession, we also determined the expression of SPX3 in Col-0 roots, which showed a similar increase in transcript abundance. In both growth types of the HDA19 knock-down lines, SPX3 transcript levels were markedly Data represent means of ten plants for each genotype and growth type ± SE. Asterisks denote statistically significant differences to Pi-replete wild-type plants based on Student's t-test (*P < 0.05, **P < 0.01, ***P < 0.001).
Scientific RepoRts | 5:15708 | DOi: 10.1038/srep15708 reduced relative to the wild types; no changes were observed for the 35S:HDA19 line (Fig. 6). A similar picture was observed for SPX1, albeit with less pronounced changes. miRNA399 is a systemic key regulator of the PSR. To determine whether systemic Pi signaling is affected in lines with altered HDA19 expression, we measured the abundance of mature miR399d and its primary transcript (pri-miR399d) by Northern blotting and qRT-PCR, respectively. Northern blot analysis of miR399d revealed a massive increase in RNA abundance upon Pi deficiency with a slight reduction in the HDA19-RNAi line and a more pronounced band for roots of the 35S:HDA19 line (Fig. 6a). Expression analysis of pri-miR399d supported this pattern. In the wild type, the abundance of pri-miR399d was circa 7-fold increased (Fig. 6b). Pri-miR399d expression was not significantly altered in the knock-down line; although, a trend of reduced expression was noted. In 35S:HDA19 plants, a marked increase in pri-miR399d transcript levels was observed in roots from plants grown under both Pi-replete and low Pi conditions (Fig. 6b), indicative of higher miR399d activity in 35S:HDA19 plants. It thus appears that HDA19 is critical in systemic Pi signaling.

Discussion
Reversible modification of core histones is an important regulator of gene activity, modulating DNA-templated processes such as transcription and repair. Acetylation of the ε -amino group of specific lysine residues neutralizes the positive charge of the histone tails, thereby weakening the interaction with the DNA and increasing the accessibility of the DNA for transcriptional regulators 30 . Histone acetylation is generally associated with gene activation. However, histone PTMs are not ultimately dictating gene expression and the interpretation of histone marks has been described as translation of a 'histone language' rather than deciphering a rigid code 31 . Thus, although HDAs are generally defined as transcriptional repressors, decreased acetylation is not a mandatory partner of gene repression. This is supported by the fact that in the present investigation the number of Pi-responsive genes that were greater than two-fold increased in HDA19-RNAi relative to 35S:HDA plants were only slightly higher than genes with two-fold lower expression in the knock-down line. The phenotypic changes observed in the transgenic lines may thus be caused by both activation and repression of genes by altered HDA19 abundance. by a yet unidentified signal that presumably originates in the root cortex. Mutants that cannot sense or transduce this signal such as scm or wer produce short cells characteristic of root hair-forming cells 11 . Mathematical modeling suggested that lack of perception or reduced strength of the positional signal delays the time point at which non-hair cells exit the default trichoblast pathway, resulting in short, trichoblast-like epidermal cells 11 . It was further suggested that the cortical bias is reduced in Pi-deficient plants, leading to a reduction in epidermal cell length and reduced rigidity of epidermal patterning. The current data imply that HDA19 is modulating this signal, causing opposite cell length in knock-down and 35S:HDA19 lines. HDA19 might either affect the generation of the positional signal, its perception or regulate signaling nodes that relay the cortical bias to down-stream target genes. A weaker signal defines a less strict pattern with a higher probability of epidermal cells entering the (default) hair fate, leading to an appreciable increase in ectopic root hairs in 35S:HDA19 lines when grown on low Pi media. This scenario is supported by the observation that compromising the positional signal by separating the epidermal cell layer from the cortex in Brassicaceae caused all epidermal cells to enter the hair cell fate 32 .
Ethylene is an important modulator of both root hair morphogenesis and cell elongation 33,34 . Ethylene affects root hair initiation downstream of RHD6 and does not interfere with the early transcriptional regulators that dictate the fate of the cells [34][35][36] . HDA19 has been previously implicated in ethylene signaling 37 , and it may be assumed that HDA19 affects ethylene production or signaling, which in turn alters the root hair phenotype in lines with altered HDA19 expression. While we cannot exclude that HDA19 exerts its effect via ethylene signaling, our results are more in line with an altered strength of a cortical signal. Ethylene induces long root hairs, which was not observed in the 35S:HDA19 line. Also, ethylene induces the formation of ectopic root hairs, which was only observed when plants were grown on low Pi medium. Moreover, ethylene-related genes that were previously found to be up-regulated in leaves of 35S:HDA19 lines such as ERF1 37 did not differ in expression between 35S:HDA19 and HDA19-RNAi roots, further arguing against the assumption that HDA19 chiefly acts through ethylene signaling on the root phenotype. Together, the data can be more readily explained by modified strength or perception of a positional signal in transgenic lines with altered expression of HDA19.

Histone modifications may interact to regulate responses to Pi starvation. Previously, we
identified the PHD finger protein ALF6 in a forward genetic screen as a critical player for root hair elongation under Pi-deficient conditions 18 . ALF6 is a bona fide histone reader that binds to H3K4me3 38 . Beside compromised root hair elongation, alf6 mutants showed a pleiotropic phenotype comprising reduced anthocyanin accumulation and altered root architecture in response to low Pi, suggesting that ALF6 is an upstream regulator of several Pi-related processes. Intriguingly, similar to what has been observed on plants with altered HDA19 abundance, various genes involved in membrane lipid remodeling were among the putative targets of ALF6, suggesting a close connection between membrane lipid remodeling and root hair differentiation. Transcription of both HDA19 and ALF6 is not responsive to Pi starvation, indicating that some of the key players escape detection in transcriptional or proteomic surveys. The two studies also indicate that substantial regulation of cellular Pi homeostasis occurs at the histone level. It is tempting to speculate that histone acetylation and methylation interacts to control the transcription and/or post-transcriptional processing of genes that impact the phenotypic readout. Notably, it was reported that H3K4 trimethylation was absent in the HDA6 mutant axe1-5, indicating that HDA activity is required for H3K4 methylation 39 , supporting such a scenario.

Iron-phosphate interactions may impact intercellular communication. Several lines of evi-
dence have provided support for a role of iron in the attenuation of root hair elongation in response to Pi starvation. Generally, Pi-deficient plants accumulate a surplus of iron despite strongly down-regulated iron acquisition genes 13,15 . Notably, restricted elongation of primary roots upon Pi deficiency does not occur in the absence of iron, which led to the assumption that this response does not primarily reflect an acclimation to Pi deficiency but is rather caused by iron toxicity 40 . Recently, this view has been challenged by the observation that a putative ferroxidase, LPR1, and the P5-type ATPase PDR2 are required to allow for meristem-specific iron and callose deposition which in turn regulates symplasmic communication in the RAM and elongation zone during Pi starvation 41 . Several iron-responsive genes are differentially regulated between HDA19 knock-down and overexpression lines, with higher expression in the HDA19-RNAi line. Assuming that a sophisticated control of iron uptake and transport is required to regulate cell-to-cell signaling in Pi-deficient plants, it appears plausible that the strength of the positional signal that controls epidermal cell fate and longitudinal cell length is modulated in plants with altered HDA19 abundance due to subtle alterations in iron distribution. Growth on low Pi media may further decrease the signal, thereby increasing the probability of hair formation in the N position.
In conclusion, our data highlight a novel non-Pi-responsive modulator of the PSR that controls cell length and, as a consequence, root hair density. In addition, the Pi deficiency-induced expression of key regulatory factors such as the SPX domain-containing transcription factors SPX3 and SPX1 is decreased in HDA19 knock-down lines. It is plausible that such changes are mediated by interactions of histone PTMs and Pi-responsive factors. These bind to histone writers and histone readers that control transcription and fine-tune cell length and gene expression to the availability of Pi.
Scientific RepoRts | 5:15708 | DOi: 10.1038/srep15708 Methods Plant growth conditions. Arabidopsis thaliana (L.) Heynh. plants were grown in a growth chamber on a solidified medium as described by Estelle and Somerville 42 . Seeds of the accession Wassilewskija (Ws-2) and Columbia (Col-0), HDA6-RNAi (CS24038), HDA9-RNAi (CS30879), had17 (SALK_090088) and hda18 (SALK_006938) plants were obtained from the Arabidopsis Biological Resource Center (Ohio State University). Plants over-expressing HDA19 under the control of the 35S promoter (35S:HDA19), HDA19 RNAi plants, hda19 mutants (hda19-1) and HDA6-RNAi plants have been described previously 37,43,44 . Seeds were surface-sterilized by immersing them in 5% (v/v) NaOCl for 5 min and 70% ethanol for 7 min, followed by four rinses in sterile water. Seeds were placed onto Petri dishes and kept for 1 d at 4 °C in the dark, before the plates were transferred to a growth chamber and grown at 21 °C under continuous illumination (50 μ mol m −2 s −1 ; Phillips TL lamps). The medium was solidified with 0.4% Gelrite pure (Kelco), the pH was adjusted to 5.5. Low Pi conditions were obtained by growing plants on media containing 2.5 μ M KH 2 PO 4 , while the Pi concentration of control medium was 2.5 mM KH 2 PO 4 . The lower concentration of potassium due to the reduced KH 2 PO 4 concentration was compensated for by the addition of KCl.
Microarray analysis. The Affymetrix gene chip Arabidopsis ATH1 Genome Array was used for microarray analysis. Total RNA sample were prepared as described above. All RNA samples were quality assessed by using the Agilent Bioanalyzer 2100 (Agilent, Santa Clara, USA). Complementary RNA synthesis was performed by use of the GeneChip One-Cycle Target Labeling Kit (Affymetrix, Santa Clara, USA). Hybridization, washing, staining, and scanning procedures were performed as described in the Affymetrix technical manual.
Gene expression data were imported directly into GeneSpring (version 11.5, Agilent, Santa Clara, USA). The software was used to normalize the data per chip to the 50 th percentile and per gene to the control samples. Genes that were flagged as absent in two replicates were not considered in the analysis. P values for the Benjamini and Hochberg method (false discovery rates; FDRs) were calculated by GeneSpring. Transcripts were defined as differentially expressed that showed delta signal changes larger than the mean expression value of the whole data set or 2-fold changes with P < 0.05.

Real-time RT-PCR.
For qRT-PCR, total RNA was extracted from the roots using the RNeasy Plant Mini Kit (Qiagen) and DNase treated with the Turbo DNA-free Kit (Ambion) following the manufacturer's instructions. cDNA was synthesized using SuperScript III reverse transcription kits (Invitrogen) following the manufacturer's instructions. Real-time PCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems) on an Applied Biosystems 7500 Fast Real-Time PCR System with programs recommended by the manufacturer. Samples were normalized first to an endogenous reference (AtTUA) and then the relative target gene was determined by performing a comparative Δ Δ Ct. The following primers were used: TUA (At5g19770) fwd: GTGCTGAAGGTGGAGACGAT, rev: AACACGAAGACCGAACGAAT; SPX1 fwd: GAAGAGCACAATCGCTGCCTT, rev: TGGCTTCTTGCTCCAACAATG; SPX3 fwd: GCG CCGGTGGAATCTATTTT, rev: GCAACTCCTTGTGGTGGATGA; pri-miR399 fwd: TTACTGGGCG AATACTCCTATGG, rev: ATTTTACTTGCATATCTAGCCAATGC.

Measurement of root hair length and density.
Confocal images with a scale bar of 100 μ m at 10 X resolution were used for measuring the root hair length. A ZEISS DISCOVERY V.12 microscope equipped with an ocular scale bar was used for measuring root hair density at 2 to 6 mm from the tip of the primary root. Statistical significant deviations from the wild type were determined by Student's t-test. Micrographs were taken between 0 to 8 mm from the tips of primary roots.
For cross-sections, root samples were fixed, dehydrated and then embedded in Technovit 7100 (Heraeus Kulzer, Wehrheim) resin in gelatin capsules. Transverse sections (30 μ m) were cut using a RM 2255 Leica microtome (Leica, Nussloch, Germany). Sections were dried and stained with toluidine blue (0.05%) on glass slides and examined using bright-field on an Imager Z1 microscope (Zeiss, Jena, Germany).
Confocal microscopy and cell length measurements. Plants were placed in 10 mg/ml propidium iodide solution (PI) for one minute and gently rinsed with water for two minutes. The root was removed and mounted in fresh water. The roots where then observed using a confocal laser scanning microscope (Zeiss LSM510 Meta). The peak excitation λ and emission λ for PI was 536 nm and 620 nm, respectively.
The cell length of trichoblasts and atrichoblasts was measured using ImageJ (http://rsb.info.nih.gov/ ij/). The position of each cell was calculated from the cumulative length of all cells between the cell and the quiescent center. The data sets were then smoothed and interpolated into 25-mm-spaced data points using a kernel-smoothing routine 45 , which was performed using a Microsoft Excel macro that enable the average calculation between replicate roots. Northern blot analysis. Total RNA was extracted from roots with TRIzol ® Reagent (Invitrogen) and Scientific RepoRts | 5:15708 | DOi: 10.1038/srep15708 ICP-OES analysis. Mineral nutrient analysis was determined by inductively coupled plasma optical emission spectrometry (ICP-OES) as described in Rodríguez-Celma et al. 29 . Five plants were harvested per treatment and genotype.