DEFECTIVE KERNEL 1 promotes and maintains plant epidermal differentiation

The production of specific function-adapted organ morphologies in plants is achieved by the exquisitely complex spatial and temporal control of cell proliferation, expansion and differentiation. The importance of this co-ordination is well illustrated in the cotyledon epidermis of Arabidopsis thaliana. To perform its basic functions this cell layer must produce both mature pavement cells (PCs), which can be very large and have a complex morphology, and appropriately spaced pairs of stomatal guard cells, which are tiny and adapted to their role in gas exchange (Glover, 2000). Producing this complex mosaic demands that neighbouring cells co-ordinate their growth.

Epidermal specification and differentiation are linked to the activity of subfamily IV of the plant-specific homeodomain-leucine zipper (HD-ZIP) transcription factors (TFs) (Abe et al., 2003; Depege-Fargeix et al., 2011; Di Cristina et al., 1996; Horstman et al., 2015; Javelle et al., 2011a,, 2010; Nakamura et al., 2006; Peterson et al., 2013; Roeder et al., 2012; San-Bento et al., 2014; Takada et al., 2013; Vernoud et al., 2009; Wu et al., 2011; Yang et al., 2011). Cell-cell communication influences both epidermal cell fate decisions and subsequent differentiation (reviewed by Javelle et al., 2011b; Schiefelbein et al., 2014; Takada and Iida, 2014). For example, several receptor-like kinases (RLKs) and ligand-like peptides control patterning of the stomatal lineage (reviewed by Richardson and Torii, 2013) and the CRINKLY4 (CR4) RLK, and its Arabidopsis homologue ACR4, are required for the control of epidermal identity in maize (Becraft et al., 1996) and the expression of HD-ZIP IV genes in Arabidopsis (San-Bento et al., 2014).

DEFECTIVE KERNEL 1 (DEK1) regulates growth coordination and epidermal characters in plants (Becraft et al., 2002; Johnson et al., 2005,, 2008; Lid et al., 2002). Loss of DEK1 function causes early embryo lethality in both Arabidopsis and maize (Becraft et al., 2002; Johnson et al., 2005). In Petunia and Arabidopsis, strong downregulation of DEK1 is accompanied by proliferation of disorganised callus-like cells (Ahn et al., 2004; Lid et al., 2005). DEK1 is also required for multicellular development in the moss Physcomitrella patens (Demko et al., 2014; Liang et al., 2013; Perroud et al., 2014). DEK1 has a modular structure, with several predicted transmembrane domains, a linker domain and a cytoplasmically localised C-terminal domain comprising a sequence similar to animal calpains and a C2-like domain (Lid et al., 2002). The complete C-terminal domain (which we refer to as the CALPAIN domain) is sufficient to complement dek1 loss-of-function alleles, suggesting that this is the active domain of the protein (Johnson et al., 2008; Liang et al., 2013). A weak DEK1 allele, dek1-4, which causes a single amino acid substitution in the C2-like domain, and mutant alleles of ACR4 and the HD-ZIP IV-encoding genes HDG11 and ATML1, were recently reported to lack sepal giant cells (GCs) (Roeder et al., 2012).

Here we aim to understand the role of DEK1 in epidermis differentiation, and particularly to clarify its relationship with the HD-ZIP IV TFs.

We found that the PCs in the adaxial epidermis of dek1-4 were less complex and more homogeneous than in wild-type (WT; Col-0) controls (Fig. 1). Three-day-old Col-0 cotyledon PCs included small cells with almost no lobes, intermediate cells with few lobes and larger, fully interdigitated cells (Fig. 1A). At the same stage, dek1-4 PCs ranged from showing no lobes (polygonal shape) to showing some interdigitation, although less than in Col-0 (Fig. 1B,C). We noted an abnormally high frequency of four-way, or near four-way, junctions (usually found in meristematic cells) in dek1-4 (Fig. 1B,C).

In the dek1-4 mutant, intermediate size and large cells were rounder than WT cells of similar size (Fig. 1D). Despite their shape differences, the average final cell size of PCs in dek1-4 and WT was similar (Fig. 1E). Thus, PCs in dek1-4 have, on average, less complex shapes. This cellular phenotype persisted in fully expanded cotyledons (Fig. 1H,I). Neighbouring PCs in dek1-4 frequently had shared straight walls (Fig. 1I) suggesting that they had undergone late divisions (observed in 100% of mutant seedlings, but only twice in 40 WT seedlings analysed). No defect in blade expansion was observed (Fig. 1J), but cotyledons were frequently epinastically curled (Fig. 1G).

Leaf blade curling was described in CYCLIN D3;1 (CYCD3;1)-overexpressing plants, where it was associated with dramatic changes in ploidy distribution (Scofield et al., 2013). However, no significant differences in ploidy distribution were detected between WT and dek1-4 cotyledons (Fig. 1K). Consistent with this, no differences in the expression of CYCD3;1, or other cell cycle-related genes such as LOSS OF GIANT CELLS FROM ORGANS (LGO), SIAMESE (SIM), KIP-RELATED PROTEIN6 (KRP6), HOBBIT (HBT) or DEL1, which encodes an endocycle inhibiting E2F-like protein, were observed between Col-0 and dek1-4 seedlings (supplementary material Fig. S1). Therefore, the phenotypes observed are not associated with ploidy defects.

dek1-4 leaves showed a significantly higher proportion of trichomes with more than three branches than WT controls (Wilcoxon rank sum test, P=1.7^–20) (Fig. 2A-C). A similar phenotype was described for hdg11 and hdg12 hdg11 mutants (Nakamura et al., 2006). We analysed the expression of HDG11, HDG12 and of additional HD-ZIP IV genes (ATML1, PDF2, HDG2) in the dek1-4 mutant and found that all genes tested were reproducibly downregulated (Fig. 2D). We then assessed the genetic relationship between DEK1, HDG11 and HDG12. hdg12-2 dek1-4 double mutants showed highly branched trichomes, as seen in the dek1-4 parental line, but did not exhibit additional defects. In dek1-4 hdg11-2 plants we observed no significant additivity of the trichome phenotypes of the two mutants (supplementary material Fig. S2). The sepals of hdg12-2 dek1-4 and of dek1-4 hdg11-2 mutants resembled those of dek1-4 mutants (supplementary material Fig. S3). Together, our results are consistent with multiple HD-ZIP IV proteins acting downstream of DEK1 to regulate epidermal differentiation.

To confirm whether the phenotype of the dek1-4 allele is caused by reduced DEK1 activity, we expressed an artificial microRNA (Ossowski et al., 2008) targeted to the DEK1 coding sequence (amiRDEK1) under the CaMV 35S promoter and under a dexamethasone (DEX)-inducible promoter (Craft et al., 2005; Samalova et al., 2005). T1 35S:amiRDEK1 plants showed similar phenotypes to those reported for AtDEK1RNAi lines (Johnson et al., 2005) and the dek1-4 mutant (Roeder et al., 2012). Some seedlings had fused cotyledons, discontinuous epidermis and died. Surviving seedlings had PCs with shallow lobes, true leaves with multi-branched trichomes, twisted rosette leaves with elongated petioles and sepals devoid of GCs (supplementary material Fig. S4). Three independent homozygous lines showing reduced expression levels of DEK1 (Fig. 3A) were analysed. The expression levels of all HD-ZIP IV genes previously analysed were not only reduced, but also correlated tightly with the levels of DEK1 transcript (Fig. 3A) and with the severity of PC phenotypes (Fig. 3B) and trichome branching defects (Wilcoxon rank sum test: P=4.24^–23 for line C, P=3.03^–23 for line L) (Fig. 3C). Aborted trichomes were observed (Fig. 3D) but not included in the statistical analysis.

To study the effects of overexpressing the active form of DEK1 (the CALPAIN domain) on the expression of HD-ZIP IV genes, we used transgenic seedlings harbouring a pRPS5A:CALPAIN-6XHIS construct (OEX [WT]). Endogenous levels of DEK1 were similar to that of WT plants, but CALPAIN transcript levels (transgene+endogenous) were several times higher than in Col-0 seedlings (supplementary material Fig. S5). No consistent changes in the expression of HD-ZIP IV genes were observed in these lines compared with WT seedlings (supplementary material Fig. S5), suggesting that HD-ZIP IV gene regulation by DEK1 is either indirect or requires additional factors.

Two independent homozygous lines in which amiRDEK1 expression was triggered only upon DEX application were selected. Induction of amiRDEK1 expression lowered DEK1 mRNA levels to about half those detected in control plants (supplementary material Fig. S4J). Upon flowering, control sepals showed a classic epidermal structure with GCs surrounded by smaller cells (Fig. 4A). In DEX-grown plants, we observed sepals with few or no GCs, as well as GC-like structures that had undergone late divisions after the onset of differentiation (Fig. 4B-D). Thus, reduced levels of DEK1 affect the maintenance of GC differentiation.

PC shape is affected in mutants with altered cortical microtubule (CMT) dynamics (Ivakov and Persson, 2013). The dek1-4 allele was introgressed into a line harbouring a GFP-TUA6 expression construct (Ueda et al., 1999) and the resulting plants analysed. In WT seedlings, small polygonal PCs have highly oriented CMTs arranged transversely with respect to their long axis, while in more lobed PCs CMTs usually orient more randomly and bundle densely in neck regions. In dek1-4, CMTs are mainly oriented perpendicularly to the long axis of the cell (supplementary material Fig. S6B,C). Quantification confirmed that CMTs in mutant PCs are more highly oriented than in WT PCs (supplementary material Fig. S6D).

To test whether their abnormal arrangement in the mutant was due to an inability of CMTs to rearrange their distribution, or was a consequence of cell shape defects, we studied CMT reorganisation upon ablation in cotyledons (Sampathkumar et al., 2014). CMTs in both mutant and WT cotyledons reoriented in cell files surrounding the ablation site (supplementary material Fig. S6E-H), suggesting that CMT dynamics are not disrupted in dek1-4. This observation supports the hypothesis that the CMT arrangement in the mutant is a consequence of altered cell shape.

A recent study has provided strong evidence that the expression of HD-ZIP IV-encoding genes is maintained by intercellular signalling mediated by ACR4 (San-Bento et al., 2014). We measured the height of epidermal cell contact zones and found that they are less uniform in height in dek1-4 than in WT seedlings (supplementary material Fig. S7). Consistent with a defect in epidermal cell adhesion (Singh et al., 2005), we also observed examples of cell separation in the cotyledons of 35S:amiRDEK1 lines and the abnormal accumulation of callose at epidermal cell boundaries in the expanding cotyledons of dek1-4 seedlings (supplementary material Fig. S7).

Unlike animal calpains, which exist in many isoforms, DEK1 is a single-copy gene unique to land plants (Liang et al., 2013; Lid et al., 2002). Here we show that defects in DEK1 perturb epidermal cell differentiation via a mechanism affecting the transcription of genes encoding HD-ZIP IV family TFs. The ancestral HD-ZIP IV protein, like DEK1, is thought to have arisen in the common algal ancestor of land plants, where it might have been crucial in the evolution of a specialised outer cell layer with the characteristics necessary for life on land, such as cuticle formation and stomatal development (Zalewski et al., 2013). Apoplastic signalling mediated by ACR4, like DEK1 function, is necessary for the maintenance of HD-ZIP IV gene expression (San-Bento et al., 2014). However, genetic analysis carried out in Arabidopsis and maize (Becraft et al., 2002; Roeder et al., 2012) suggests that DEK1 and ACR4 act in parallel.

A key requirement of the epidermis is that it should be continuous, implying the maintenance of a highly regulated zone of circumferential contact between neighbouring cells. A common feature of phenotypes associated with loss of DEK1 activity is epidermal cell separation, suggesting that DEK1 might regulate this zone of cell-cell contact (Ahn et al., 2004; Johnson et al., 2005; Lid et al., 2005). Here we show that although dek1-4 mutants produce a continuous epidermis, the zone of cell-cell contact between epidermal cells is perturbed, further supporting this hypothesis. We propose that DEK1 influences HD-ZIP IV gene expression, and thus epidermal differentiation, by regulating epidermal cell adhesion, a function that might have been crucial in the successful move of plants onto dry land.

Seeds were obtained from the Nottingham Arabidopsis Stock Centre unless otherwise stated. The dek1-4 allele was donated by Adrienne Roeder (Cornell University, USA) and introgressed into the Col-0 background. dek1-4 genotyping involved amplifying genomic DNA using the primers shown in supplementary material Table S1 and sequencing the resulting product with primer RG15 (supplementary material Table S1, Fig. S2). T-DNA insertion lines were genotyped using the primers described in supplementary material Table S1. A DEK1-specific artificial microRNA was designed using MicroRNA Designer (WMD, http://wmd3.weigelworld.org) and generated using the PCR primers in supplementary material Table S1 according to the protocol from WMD. The pre-amiRNA was recombined into pDONR221 (Invitrogen) and then into the Gateway-compatible binary vector pART27 (Lee and Gelvin, 2008) under control of a CaMV 35S promoter, or into the pOPON2.1 vector (a gift from Ian Moore, University of Oxford, UK). Plants expressing pRPS5A:CALPAIN-6XHIS were generated as described for pRPS5A:CALPAIN-GFP (Johnson et al., 2008), but with the insertion of an in-frame two times 6×His tag with a STOP codon in place of the GFP sequence.

Soil-grown plants were placed at 20°C and 55% relative humidity under a 16-h light/8-h dark cycle. For in vitro cultures, seeds were surface sterilised with chlorine gas, stratified on media for 2-3 days in the dark and germinated under a 16-h light/8-h dark cycle at 21°C. To induce DEK1 silencing, germinated seedlings were transferred after 5 days onto medium supplemented with 0.4% DMSO (as control) or 25 µm dexamethasone (DEX).

Scanning electron microscopy samples were mounted on graphite paste (Electron Microscopy Sciences) and visualised using an SH-3000 table-top scanning electron microscope (HIROX) at −30°C or −50°C with an accelerating voltage of 5 kV.

Cotyledon PC cell walls were stained with 1 mg/ml propidium iodide (PI) for 5 min, briefly rinsed, and imaged at a point midway between the midrib and margin and halfway along the blade. Cell areas and perimeters were measured using ImageJ (NIH). A ‘shape factor’ (SF=4π area/perimeter²) was calculated for each cell (Dewitte et al., 2007). CMT alignments were quantified using the ImageJ plug-in FibrilTool (Boudaoud et al., 2014). All confocal images were acquired using an LSM700 confocal microscope (Carl Zeiss).

For tissue imprints, we used an agarose-based method (Mathur and Koncz, 1997). Light micrographs were taken with an Axio Imager.M2 microscope (Carl Zeiss) using DIC optics.

Seedlings were grown vertically for 6 days on square plates and aerial parts harvested. Three independent biological replicates of 15-20 seedlings were used. RNA extraction, DNase treatment and qPCR analysis were carried out as described previously (Xing et al., 2013). Expression levels of each gene, relative to EIF4A, were determined as previously described (Ferrari et al., 2006). Primers used for qRT-PCR analysis are shown in supplementary material Table S1.

Cotyledons were harvested with forceps, chopped with a razor blade in nuclear lysis buffer (Dewitte et al., 2007), filtered and stained with DAPI. The nuclei-containing solution was analysed with a Partec ploidy analyser. Three biological replicates (20 cotyledons from ten independent plants for each replicate) were performed.

We thank Adrienne Roeder (Cornell University, USA) for providing the original dek1-4 mutant line and for helpful discussions; Ian Moore (University of Oxford, UK) for providing the binary vector pOPON2.1; and Jeremy Just (RDP, ENS de Lyon, France) for helping with statistical analysis.