Increased root hair density by loss of WRKY6 in Arabidopsis thaliana

Plant material

Arabidopsis thaliana L. genotypes were chosen on basis of previous results. The ecotypes Ga-0 and Kondara were chosen because of their strong root hair length and density adaptation to low P in a pervious study (Stetter, Schmid & Ludewig, 2015). wrky6 mutants were obtained from T. Fujiwara (Kasajima et al., 2010) and tetraploid Col-0 (4×) seeds were obtained from D. Chao (Chao et al., 2013) to study the influence of WRKY6 and polyploidy on root hair traits. Col-0 was included as reference and background of wkry6 mutants.

Split plates and growth conditions

Vertically placed split squared Petri dishes (120 mm × 120 mm × 17 mm, Greiner bio-one, Kremmünster, Austria) with a top compartment containing all nutrients, and a lower test compartment with or without P_i (KH₂PO₄ substituted by KCl) were used. The growth medium was composed of 5 mM KNO₃, 2 mM MgSO₄, 2 mM Ca(NO₃)₂, 70 µM HBO₃, 14 µM MnCl₂, 1 µM ZnSO₄, 0.5 µM CuSO₄, 10 µM NaCl, 0.2 µM Na₂MoO₄, 40 µM FeEDTA, 4.7 mM MES and 1 mM KH₂PO₄, 43 mM sucrose, pH was adjusted to 5.7 with 1 M KOH, solidified with 1.2% phytoagar.

Seeds were derived from plants grown in nutrient-rich garden soil (Einheitserde EET Einheitserde- und Humuswerke, Sinntal-Jossa, Germany), and surface sterilized with 70% ethanol and 0.05% TritonX 100 for 15 min under continuous shaking and were washed twice with 70% ethanol for 5 min. After washing, the seeds were dried on filter paper under a clean bench. Eight to fifteen seeds were placed 1 cm above the split. Plates were stored for two days at 4°C in darkness, before being kept under 21°C and continuous light for 12 h. After light treatment, plates were kept in cold and darkness for another day before being transferred into the growth chamber. Plants were grown under constant light (180 µE) and 21°C.

Image acquisition and analyses

For each accession replicated images of six to ten roots per treatment were taken with 20× magnification with a Zeiss AxioCam MRm (Carl Zeiss, Jena, Germany) monochrome camera under a Zeiss Stemi 2000-c binocular microscope (Carl Zeiss, Jena, Germany) to assess root hair length and density. Root hairs were measured and counted using ImageJ (http://imagej.nih.gov). To investigate the root cell length and diameter, roots were treated with a 1:250 diluted propidium iodide (2.5 mM) solution in water. Roots were stained for 210 s, before being washed. Images were acquired 1 cm above the root tip, where root hairs were fully developed, once the root was 3 cm long. Images were taken at 20× magnification with a Zeiss LSM 700 microscope (Carl Zeiss, Jena, Germany). 3D image stacks were performed to observe the cross sections. The interval between the stacks was adjusted between 1.5 and 2 µm. Epidermal cell length, root diameter, cortex cell diameter, epidermal cell diameter, and the number of epidermal cells adjacent to each cortex cell were analyzed using ImageJ (http://imagej.nih.gov). For measuring the epidermal cell length, a representative epidermal cell being in contact with two cortical cells was measured for each sample. Ectopic root hairs (in non-hair positions) were encountered very rarely, not sufficient to be included in statistical analysis. For measuring the cortex cell diameter, the means of three neighboring cortex cells were measured. The epidermal cell diameter was quantified from the epidermal cells adjacent to these three cortical cells.

Statistical data analysis

A mixed linear model was used for data analysis, reflecting the experimental design. The following model was employed for the diversity study: ${\displaystyle Y=\mathrm{GEN}\times \mathrm{TRT}:\text{(REP/PLATE)/PLANT}}$ where, Y = Response variable (root hair density or root hair length), GEN = Fixed effect of the genotype, TRT = Fixed effect of the treatment, REP = Random effect of the replicate, PLATE = Random effect of the Petri dish, PLANT = Residual effect of the plant, x = main effects + interaction, / = nested relationship.

Data analysis was performed with the MIXED procedure in SAS/STAT of SAS^® 9.3. Variance homogenity was reviewed and outliers were identified and removed based on studentized residuals. Values with studentized residuals larger than 4 were removed from the dataset.

GWA mapping

Associated SNPs with Bonferroni corrected p-values smaller than 0.0001 were analyzed for gene identification based on the TAIR10 gene annotation (www.arabidopsis.org). We used sequence data of 71 accessions to analyze the candidate region on chromosome 1 with higher resolution. A window of 20,000 bp up- and downstream of the associated SNP was extracted and the genomic data were imputed using MaCH before SNPs with allele frequencies <0.05 were dismissed. Association mapping was performed with MLMM (Segura et al., 2012), to apply association mapping on the genomic region around a SNP, which previously has been found in high association with root hair density adaptation to local P_i scarcity (Stetter, Schmid & Ludewig, 2015).

The experiment to evaluate the root hair length and density of wrky6 was laid out as split- plot design with three replicates. For the root cell measurement, a split-plot design with two complete replicates was chosen. For each replicate one petri dish per treatment and accession was prepared and four plants per petri dish were sampled.

Genome-wide association mapping of root hair density and response to absence of local P_i

Root hairs of more than 160 Arabidopsis accessions were recently phenotyped (Fig. 1, inset) on vertical split agar plates (Stetter, Schmid & Ludewig, 2015). Genome wide association (GWA) mapping had identified candidate genes potentially involved in root hair density and surface acclimation to the absence of P_i. However, mutant analysis with few candidate genes did not identify direct relation to P_i sensing. Rather, mutations in candidate genes typically increased root hair densities and imposed P_i insensitivity, suggesting that these genes repress root hair density under P_i availability, rather than promote hair density in the absence of P_i (Stetter, Schmid & Ludewig, 2015).

While re-evaluating previous data, a non-significantly associated SNP in close proximity to WRKY6 captured our attention. WRKY6 encodes a transcription factor that physically interacts with the PHO1 promoter and is involved in Pi homeostasis (Chen et al., 2009). While the SNPs associated with WRKY6 did not have significant p-values in the original GWA mapping (Stetter, Schmid & Ludewig, 2015), fine mapping with 71 accessions for which whole genome sequences were available identified five highly significant SNPs 6 to 13 kb upstream to WRKY6 (Fig. 1). Furthermore, the coding sequence of WRKY6 was polymorphic at 11 positions among these 71 sequenced genotypes, of which three nucleotide exchanges lead to amino acid changes: 250V/G, 443S/G and 450V/A. These alleles were encountered at frequencies between 6–23% When calculated for individual alleles, the mean density and length of these alleles was similar, indicating that these polymorphisms are unlikely affecting WRKY6 function, but the mean hair density was significantly different for a few low frequency alleles (4–6%) in the promoter, in introns or in exons. This was most obvious under control conditions (Fig. 2D) and resulted in different responsiveness of these alleles.

Root hair length and density differences in wrky6-3 and other Arabidopsis accessions

The root hair density, length, root diameter and the response to scarce local P_i were subsequently analyzed with a wrky6-3 mutant, to address a potential further function of this gene in P_i-related root hair traits. Root hair density of this loss-of-function mutant was significantly increased in the presence of P_i by 35% and by 29% in scarce P_i, compared to the wild type Col-0 (Fig. 2A). In addition, root hair length increased significantly in the mutant, but was not different between control conditions and local absence of P_i (Fig. 2B).

The root hair density and length were also measured in an accession that is strongly stimulated by −P_i (Ga-0) and in an accession with reduced root hair length and density under −P_i conditions (Kondara). Their hair patterns were significantly different to Col-0 (Figs.2A and 2B). Interestingly, the root diameter was similar in most genotypes, but increased in tetraploid Col-0 (4×), especially in the absence of P_i (Fig. 2C) and Ga-0.

Primary root morphological differences in accessions and wrky6-3

Because different root hair density can have various mechanistic reasons, we microscopically compared root cell length, diameter and radial arrangements in diploid and tetraploid Col-0, Ga-0, Kondara and wrky6-3. Longitudinal sections of propidium iodide-stained roots revealed the length of rhizodermal cells, diameter of epidermal and cortical cells, as well as the number of neighboring cells in radial sections (Fig. 3). The data revealed that Kondara, an accession that almost completely lacked root hairs when assessed with low light microscopic resolution under −P_i, clearly had small bulbs of root hair initiations (Fig. 3). Thus, root hairs were initiated similar as in control conditions, but failed to elongate under −P_i (Fig. 3).

Consistent with a thicker root and greater radial root diameter in polyploids, the epidermal cell diameter in Col-0 (4×) was larger than in diploid Col-0 (Fig. 4A). Similarly, the epidermal cells were stretched in Col-0 (4×) control conditions compared to diploids, but were strongly reduced in −P_i, to similar cell lengths as diploid genotypes. The epidermal length was similar in all diploids, including wrky6-3 and was not affected by different P_i (Fig. 4B). By contrast, the two accessions with the thickest root diameter (Ga-0 and Col-0 4×) also had the largest cortex cell diameters, although these were only significantly different in −P_i to Kondara. wrky6-3 appeared to have significantly smaller cortex cells under −P_i than its reference diploid wild type, Col-0. The epidermal cell number per cortex cell did not differ between different P_i conditions and was identical between Col-0 2× and Col-0 4×. However, due to the large cortical cells in Ga-0, the ratio between epidermal and cortical cells was significantly increased, while this number was reduced in wrky6-3 under −P_I (Fig. 4D).

Diversity of root hairs and response to local lack of phosphate

In this study split agar plates were used to determine root hair responses to lack of P_i. The Arabidopsis plants were not systemically P_i-deficient, as the top part of the root had access to sufficient P_i. The results are consistent with the idea that different genes control the systemic and local responses to insufficient P_i supply (Chiou & Lin, 2011). −P_i leads to a correlated increase of root hair density in Col-0 (Bates & Lynch, 2001; Müller & Schmidt, 2004; Narang, Bruene & Altmann, 2000), but not in Kondara. The latter genotype was morphologically not different from Col-0 in cortical or epidermal cells, indicating that other genetic factors, not simply the radial cellular root architecture and morphology, were responsible for genotypic differences. This is corroborated by the fact that root hair bulbs were detected under −P_i, but hairs did not elongate, leading to macroscopically hairless roots under −P_i. Ga-0, which had very low hair density under control conditions, but strongly increased the root hair density in −P_i, had larger cortex cell diameter than Col-0, leading to higher numbers of epidermal cells per cortex cells, both under +P_i and −P_i. As root hairs are normally only found in the epidermal cells with direct contact to two cortical cells, this can explain the relative low hair cell density in this genotype under control conditions.

Polyploid plants generally have larger cells than diploids and this was confirmed for diploid Col-0 (2×) and the isogenic, colchicine-doubled tetraploid Col-0 (4×). Tetraploid Col-0 had thicker root diameter, which resulted from larger individual cell diameters. The radial ratio of cortical and epidermal cells, however, was unchanged, and the longitudinal epidermal cell length was only increased under +P_i. The increased root hair length in tetraploid Col-0 (4×) may thus be related to generally larger root cells, while the unaltered cortex/epidermal cell ratios explain little difference in hair densities.

When the primary root tip comes in contact with a P_i-deficient medium, primary root growth is reduced, leading to shorter longitudinal elongation of cells (Svistoonoff et al., 2007). A longitudinal reduction in epidermal cells was, however, only found in tetraploid Col-0, but not in the other genotypes, which suggests that the generally small effect of less dense root hairs in the −P_i compartment is due to other factors. That polyploidy also caused higher shoot K⁺ concentrations and salt resistance in various A. thaliana accessions (Chao et al., 2013), may in part be related to root hair-mediated uptake of K⁺, another nutrient primarily acquired via root hairs.

WRKY6 is associated with root hair traits

Fine association mapping of the root hair density response to low local P revealed associated highly significant SNPs in close proximity to the WRKY6 transcription factor gene. As SNPs in such a small region are linked, WRKY6 might be the causal gene associated; causal SNPs are not necessarily closer linked to the trait in GWA or more significant than nearby non-causal SNPs (Atwell et al., 2010). WRKY6 was initially identified as defense and senescence regulator in the leaves, but is also strongly expressed in the roots (Robatzek & Somssich, 2001; Robatzek & Somssich, 2002). However, WRKY6 also negatively regulates PHO1, a key phosphate transporter gene involved in xylem loading and P_i homeostasis, by directly binding to its promoter and inhibiting PHO1 expression (Chen et al., 2009). This link to P_i nutrition is supported by the fact that WRKY6 is also crucial for arsenate resistance, since arsenate is a structural analog of phosphate and taken up via the same route (Castrillo et al., 2013). P_i deficiency releases the WRKY6 binding to the PHO1 promoter (Chen et al., 2009) and also directly represses the expression of the phosphate transporter gene PHT1;1, a systemically controlled P_i-starvation induced gene and many other genes (Castrillo et al., 2013). Even further, unrelated functions of WRKY6 in abscisic acid signaling were recently identified . (Huang et al., 2016), but it is important to note that these phenotypes were identified in ectopic overexpressor mutants, while minor or no phenotypes were identified in loss-of-function alleles (Castrillo et al., 2013; Chen et al., 2009; Huang et al., 2016). Although ectopic overexpression, especially of multigene-family transcription factors, is often used to identify novel functions and phenotypes in transgenics, results from such overexpressors may need to be taken with caution, if previously non-expressed genes are activated in cell types in which the relevant gene was not or low expressed in the wild type (Zhang, 2003). Furthermore, the ectopic overexpression WRKY6, WRKY18, WRKY53 and WRKY70 resulted in similar stunted transgenics, with altered leaf morphology and altered flowering time (Ulker & Somssich, 2004), questioning the specificity of the overexpression approach.

A loss-of-function mutant of WRKY6 had longer and denser root hairs than Col-0, which were also less responsive to the lack of P_i. WRKY6 thus appears not only to repress phosphate transporters, but also root hairs. It is strongly up-regulated by a toxic arsenate pulse (Castrillo et al., 2013) and under boron deficiency (Kasajima et al., 2010) further establishing the link to nutrition. Several nutrients thus affect WRKY6. Root hairs indiscriminately increase the uptake of several nutrients, thus, WRKY6 may be generally involved in linking low nutrient availabilities with root hair traits. Small morphological differences to Col-0, especially in the cortex diameter and the reduced epidermal/cortical cell ratio, may partially explain the root hair promotion in wrky6-3, as this is expected to lead to more epidermal cells in the H-position. However, WRKY6 overexpressor mutants clearly had ample root hairs at different P_i supply (Chen et al., 2009; Huang et al., 2016). In addition to many other functions, WRKY6 may act as repressor of root hair density and length, rather than regulator of the P_i deficiency response (Huang et al., 2016; Robatzek & Somssich, 2001; Robatzek & Somssich, 2002; Skibbe et al., 2008).

It is noteworthy that the association mapping did not yet identify already known, well-established, crucial genes for root hair traits. Although it is possible that these are not causal for the natural variation seen in Arabidopsis, it is likely that larger populations of Arabidopsis accessions are mandatory to unravel the genetic architecture of natural variation of root hair traits and the acclimation to low P.