The metalloid arsenic (As) and the heavy metal cadmium (Cd) are highly toxic elements present in the environment, and they pose a considerable risk to living organisms owing to their accumulation in the soil and water (Godt et al., 2006; Zhao et al., 2009). Typical toxic effects of Cd on plants are the inhibition of primary root growth and increased lateral root density (Hu et al., 2013). Although the influence of Cd on root hairs has not been researched in detail, studies have shown that Cd at specific concentrations (that inhibit root growth by about 50%) increases root hair density and length in various plant species including radish (Vitoria et al., 2003), barley (Durcekova et al., 2007), sorghum (Kuriakose and Prasad, 2008), and Rhode grass (Kopittke et al., 2010). However, the molecular mechanism of root hair enhancement in response to Cd and As has not been reported to date.

Root hairs are tubular outgrowths from specific epidermal cells that function in nutrient and water absorption (Peterson and Farquhar, 1996). In Arabidopsis (Arabidopsis thaliana), root anatomy is well determined, and the different steps of root hair development including fate determination, initiation, bulge formation, and tip growth have been defined (Schiefelbein and Somerville, 1990; Dolan et al., 1993; Galway et al., 1994). By definition, epidermal cells located over the anticlinal junction between two underlying cortical cells differentiate into trichoblast cells that develop root hairs, while epidermal cells that lie above the outer periclinal cortical cells become non-hair cell atrichoblasts (Masucci et al., 1996; Dolan, 2006).

In Arabidopsis, several genes and transcription factors are involved in epidermal cell fate determination. Previous studies showed that root epidermal cells differentiate into non-hair cells through the action of the transcriptional complex WER-GL3/EGL3-TTG1, which consists of the R2-R3 MYB transcription factor WER (WEREWOLF), the basic Helix-Loop-Helix (bHLH) transcription factors GL3 (GLABRA3) and EGL3 (ENHANCER OF GLABRA3), and TTG1 encoding a protein of 341 amino acid residues containing four WD-40 repeats. By contrast, a root hair cell is produced if the transcriptional complex CPC/TRY/ETC1 (CAPRICE/TRIPTYCHON/ENHANCER OF TRY AND CPC1, one-repeat Myb protein)-GL3/EGL3-TTG1 is present in the epidermal cell (Wada et al., 1997; Lee and Schiefelbein, 2002; Schellmann et al., 2002; Kirik et al., 2004; Simon et al., 2007; Benitez and Alvarez-Buylla, 2010). The generation of each complex depends on the relative abundance of two competing transcription factors, WER/MYB23 and CPC/TRY/ETC1 (Wada et al., 1997; Lee and Schiefelbein, 1999). In non-hair cells, high expression levels of WER make the complex to be mainly an active form, which promotes the expression of GL2 (GLABRA2, homeodomain transcription factor) and the repression of RHD6 (ROOT HAIR DEFECTIVE 6, bHLH transcription factor) transcription, suppressing the expression of atrichoblast-specific genes and preventing root hair formation (Masucci and Schiefelbein, 1994, 1996; Rerie et al., 1994; Lee and Schiefelbein, 1999; Menand et al., 2007; Bruex et al., 2012). On the other hand, in root hair cells, the perception of a JACKDAW (JKD, zinc finger transcription factor)-induced positional signal in the cortex by SCM (SCRAMBLED, leucine-rich repeat receptor-like kinase) inhibits WER expression (Kwak et al., 2005; Hassan et al., 2010). Eventually, reduced WER expression as well as competitive inhibition of WER binding to GL3/EGL3-TTG1 by CPC/TRY/ETC1 (Lee and Schiefelbein, 2002) lead to the formation of an inactive complex, CPC-GL3/EGL3-TTG1, and the repression of GL2 and root hair cell differentiation. The WER-GL3/EGL3-TTG1 complex positively regulates the transcription of CPC/TRY/ETC1 and negatively regulates that of GL3/EGL3 in atrichoblast cells. Therefore, it was proposed that CPC/TRY/ETC1 translocates from non-hair cells to hair cells and GL3/EGL3 in the opposite direction via plasmodesmata (Kurata et al., 2005; Kang et al., 2013).

Genetic analysis of root hair mutants demonstrated that the phytohormones ethylene and auxin regulate root hair formation independently or downstream of the WER-GL3/EGL3-TTG1 or CPC-GL3/EGL3-TTG1 pathway (Schiefelbein, 2000; Muller and Schmidt, 2004). For instance, ROOT HAIR DEFECTIVE6 (RHD6), a downstream gene of GL2, encodes a bHLH transcription factor (Qing and Aoyama, 2012) and its mutation results in a decrease of root hair number as well as a shift in the position of root hair emergence (Masucci and Schiefelbein, 1994; Menand et al., 2007). However, this defect can be reversed by the application of auxin (IAA) or ACC (ethylene precursor), implying that RHD6 plays a role in root hair formation (probably initiation) via auxin and/or ethylene related pathways. Mutations in the AXR2 (AUXIN RESISTANT2) gene, which encodes the Aux/IAA protein, result in a decline in root hair number (Pitts et al., 1998; Nagpal et al., 2000; Rigas et al., 2013). AXR2 represents a converging point of the auxin and ethylene pathways, since treatment with IAA or ACC does not recover root hair numbers in axr2 mutants (Masucci and Schiefelbein, 1996). Mutants of the AUX1 gene, which encodes an auxin influx carrier, display shorter root hairs (Pitts et al., 1998; Schmidt and Schikora, 2001; Yu et al., 2015). Mutants of AXR1 (AUXIN RESISTANT1), which encodes a ubiquitin-activating enzyme E1, have shorter root hairs than the WT (Cernac et al., 1997; Pitts et al., 1998). Regarding ethylene, root hair formation is suppressed by the ethylene biosynthesis inhibitor AVG (Tanimoto et al., 1995; Masucci and Schiefelbein, 1996) and enhanced by ACC (ethylene precursor) treatment (Tanimoto et al., 1995). Nevertheless, the role of ethylene in root hair formation remains unclear, since the ethylene signaling mutants etr1 (ETHYLENE RESPONSE1) and ein2 (ETHYLENE INSENSITIVE2) show normal root hair density (Masucci and Schiefelbein, 1996). However, ein2 mutants have shorter root hairs (Rahman et al., 2002). In addition, the ethylene response mutant ctr1 (CONSTITUTIVE TRIPLE RESPONSE1), which is constitutively active in ethylene signal transduction, possesses a greater root hair density (Masucci and Schiefelbein, 1996), and the ethylene overproducing mutant eto1 exhibits longer hairs (Strader et al., 2010) than the WT. Furthermore, ethylene promotes auxin biosynthesis and transport, implying that ethylene may be involved in root hair formation through the auxin pathway (Ruzicka et al., 2007; Swarup et al., 2007).

In the present work, we examined the enhancement of root hair density and length by Cd and As(III) based on the hypothesis that root hair fate determination and morphogenesis are affected by cadmium and arsenic. Our results suggested that cadmium and arsenic increase root hair density and length by modulating the expression of genes involved in fate determination and the initiation and elongation of root hairs. Based on our results, we proposed a model in which cadmium and arsenic enhance root hair density and length by promoting the convergence of developmental (TTG1/GL2/GEM) and hormonal (auxin and ethylene) signaling pathways.

In the present study, we showed that Cd and As(III) increased root hair density probably by converting non-hairy cells to hairy cells through the modulation of the expression of genes involved in root hair fate determination (Figures 3 and 9). First, Cd and As(III) downregulated TTG1, which is essential for generating the WER-GL3/EGL3-TTG1 complex; this impaired its ability to upregulate GL2, which resulted in cells entering the root hair cell fate pathway. Second, Cd and As(III) may inhibit GL2 expression directly and render it unable to inhibit RHD6 transcription, which is involved in hair initiation, as well as impairing the expression of atrichoblast-specific genes. Third, Cd and As(III) promoted GEM expression, positively regulating RHD6 expression, which was downregulated in gem Arabidopsis (Figure 8); this led to the upregulation of root hair cell differentiation genes and the determination of hair cell fate.

The next step of root hair development may be root hair morphogenesis, including initiation and elongation. RHD6 and AXR2 are downstream components of fate determination that initiate root hair morphogenesis, since rhd6 and axr2 mutants have reduced root hair density (Masucci and Schiefelbein, 1994; Rigas et al., 2013). Our data showed that RHD6 expression was increased by Cd and As(III), resulting in increased root hair density and length. RHD6 expression was higher in the ttg1 mutant, in which root hair density is increased, and was lower in the gem mutant, in which root hair number is reduced, than in the control. AXR2 expression was enhanced by Cd and As(III), remained unchanged in ttg1, and was reduced in gem plants. Auxin and ethylene promote root hair initiation via AXR2 independently of RHD6, since they rescue low root hair numbers in the rhd6 mutant but not in the axr2 mutant (Masucci and Schiefelbein, 1996). Although the ctr1 mutant has a higher number of root hairs (Masucci and Schiefelbein, 1996), it is not clear whether it functions in hair initiation, since supporting data was not reported thereafter.

The auxin signaling components AUX1 and AXR1 are involved in root hair elongation since aux1 and axr1 mutants have shorter hairs (Pitts et al., 1998). Consistent with this result, AUX1 and AXR1 were upregulated at concentrations of Cd and As(III) that promoted root hair elongation (Figure 3). In addition, their expression was decreased in gem plants, which have shorter root hairs (Figure 8), and not affected in ttg1 Arabidopsis, which has a higher root hair number (Figure 6). These results confirmed that GEM is also involved in root hair elongation. In accordance with our results regarding AUX1, a previous report indicated that Cd stress induces AUX1 expression in rice root hairs, and the root hair length in osaux1 mutant rice is shorter than that of the WT (Yu et al., 2015). The involvement of auxin in root hair elongation is supported by our results showing increased levels of auxin in the root tip in parallel with the increased root hair length caused by Cd and As(III) (Figure 4C).

Ethylene signaling components also seem to play a role in root hair elongation, since the ethylene insensitive mutant ein2 has shorter hairs (Rahman et al., 2002), while the ethylene overproducing mutant eto1 has longer root hairs (Strader et al., 2010). In support of this, the ethylene signaling components ETR1 and EIN2 were upregulated in response to Cd and As(III), stimulating root hair elongation (Figure 3). Consistent with this result, the negative regulator of ethylene signaling CTR1 was downregulated by Cd and As(III). In the case of ttg1 plants showing an increase in root hair density, the expression of ethylene signaling genes was not altered (Figure 6). By contrast, ethylene signaling genes were repressed in gem plants, which display a reduction in root hair density and length (Figure 8). Regarding the relationship between ethylene and auxin, ethylene is able to enhance auxin synthesis and transport (Ruzicka et al., 2007; Swarup et al., 2007), and ethylene promotes root hair elongation via auxin-induced microtubule randomization (Takahashi et al., 2003). In addition, ethylene may be involved in root hair initiation, since ACC (ethylene precursor) enhances root hair number (Tanimoto et al., 1995) and AVG (ethylene biosynthesis inhibitor) reduces root hair number (Masucci and Schiefelbein, 1996).

It is also possible that Cd promotes root hair elongation by increasing ROS production, which activates root hair cell elongation (Takeda et al., 2008). This hypothesis is supported by reports that root hair length was induced by Cu (Pasternak et al., 2005) and Cr (Castro et al., 2007), which are associated with changes in ROS metabolism and auxin homeostasis.

It is also suggested that Cd enhances root hair formation by inducing Fe deficiency through the inhibition of Fe uptake, since Fe deficiency promotes root hair production and elongation by enhancing ethylene production (Schmidt and Schikora, 2001). Cd reduces Fe uptake by competing for the iron transporter IRT1 (Vert et al., 2002).

During the last decades, dynamic gene regulatory network (GRN) models have been used extensively to investigate cell fate determination (von Dassow et al., 2000; Albert and Othmer, 2003; Espinosa-soto et al., 2004; Almaas, 2007). These models integrate experimental data and produce a dynamic model to explore developmental patterns. Such models are potentially useful and can improve our knowledge regarding biological developmental systems, since they provide powerful tools to interpret contrasting data. In addition, new hypotheses can arise from these models that can be experimentally validated. For instance, Benitez et al. (2007) used a dynamic model to show that positional SCM signals in the cortex can function either positively or negatively on WER and CPC expression within the active or inactive complex, respectively. Although WER expression is proposed to be repressed by the SCM signal (Kwak et al., 2005), the possibility of a direct effect on CPC could not be excluded (Wada et al., 1997; Lee and Schiefelbein, 2002; Costa and Dolan, 2003; Benitez et al., 2007). In the model presented here, we attempted to show how Cd and As (III) affect the GRN in root epidermal cells based on the transcriptional levels of the genes examined. Our model (Figure 9) is only a schematic integration of the active molecules in each cell type. Moreover, this does not include the complete GRN model components within each cell type, in contrast to dynamic models presented in recent studies (Benitez et al., 2008; Kwak and Schiefelbein, 2008; Savage et al., 2008). Information derived from dynamic GRN studies (Benitez et al., 2007, 2008; Kwak and Schiefelbein, 2008; Benitez and Alvarez-Buylla, 2010) can improve our understanding of cell fate determination and patterning, as well as possible interactions within the GRN and cell to cell communication in Arabidopsis root epidermal cells. In this context, the GRN model can not only predict and provide novel insights into GRN component interactions involved in root hair patterning in response to Cd and As, but also help link heavy metal signaling pathways and genetic regulatory network components under different experimental condition in future studies.

For further experiments, the conversion of non-hair cell to hair cells (change of fate) induced by Cd and As(III) should be examined using morphological analyses. In addition, the expression of various fate-determining genes in response to metals, such as WER, GL3, EGL3, CPC1, TRY, and ETC1, should be examined. Analysis of the effects of metals on recently identified genes expressed in hairy and non-hairy cells is also meaningful (Bruex et al., 2012).

The density and length of root hairs are increased in response to Cd and As(III) in Arabidopsis by the following mechanisms: First, root hair fate determination (differentiation) is stimulated through the upregulation of GEM (positive regulator) and the downregulation of TTG1/GL2 (negative regulators). Second, the initiation of root hair is promoted by the upregulation of RHD6. Third, the elongation of root hair is promoted by increasing auxin content and the expression of the auxin signaling genes AXR2, AUX1, and AXR1. Finally, root hair elongation is stimulated by the upregulation of the ethylene signaling genes ETR1 and EIN2, and the downregulation of CTR1.

SH designed and supervised whole research, and wrote the article with contributions of all the authors; RB and DK equally performed most of the experiments; JK performed an initial experiment.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.