Mutual control of intracellular localisation of the patterning proteins AtMYC1, GL1 and TRY/CPC in Arabidopsis

The initiation of Arabidopsis thaliana trichomes serves as an excellent model system with which to study how cells are specified in a temporal and spatially defined manner by a gene regulatory network (Balkunde et al., 2010; Tominaga-Wada et al., 2011; Grebe, 2012). In wild type, trichomes are formed at the leaf basis of young leaves without a recognisable positional reference to other leaf structures (Hülskamp et al., 1994), and the analysis of genetic mosaics ruled out the possibility that the selection is based on cell lineage (Larkin et al., 1996; Schnittger et al., 1999). It is therefore postulated that trichome patterning is governed by cell-cell interactions between initially equivalent cells (Ishida et al., 2008; Pesch and Hülskamp, 2009; Balkunde et al., 2010).

In the current models of trichome patterning, a regulatory network of positively and negatively acting genes explains how trichomes are selected in a field of competent protodermal cells. The positive regulators are represented by a group of three proteins: the WD40 protein TRANSPARENT TESTA GLABRA 1 (TTG1) (Koornneef, 1981; Galway et al., 1994; Walker et al., 1999); the R2R3 MYB-related transcription factor GLABRA 1 (GL1) (Oppenheimer et al., 1991); and the basic helix-loop-helix (bHLH)-like transcription factor GLABRA 3 (GL3) (Koornneef et al., 1982; Hülskamp et al., 1994; Payne et al., 2000). GL1 and GL3 act partially redundantly with their homologues MYB23 and EGL3, respectively (Kirik et al., 2001; Zhang et al., 2003; Kirik et al., 2005). The negative regulators are a class of seven homologous R3 single-repeat MYB genes that act in a partially redundant manner (Schellmann et al., 2002; Kirik et al., 2004a; Kirik et al., 2004b; Wang et al., 2007; Tominaga et al., 2008; Wang et al., 2008; Wester et al., 2009; Wang et al., 2010; Gan et al., 2011). Among these, TRIPTYCHON (TRY) and CAPRICE (CPC) show the strongest single-mutant phenotypes (Hulskamp et al., 1994; Wada et al., 1997; Schellmann et al., 2002). TTG1, an R2R3MYB and a bHLH protein form a trimeric activator complex, in which TTG1 and the MYB both bind to the bHLH protein (Payne et al., 2000; Zhang et al., 2003; Zimmermann et al., 2004; Kirik et al., 2005; Digiuni et al., 2008; Gao et al., 2008; Wang and Chen, 2008; Zhao et al., 2008). The inhibition of the activator complex is governed by the binding of an R3MYB to the bHLH protein, thereby replacing the R2R3MYB (Payne et al., 2000; Bernhardt et al., 2003; Esch et al., 2003). The immediate downstream gene of this machinery promoting the differentiation of trichome cells is GLABRA 2 (GL2) (Rerie et al., 1994).

The activators and inhibitors of trichome patterning are engaged in a gene regulatory network (Benítez et al., 2007; Benítez et al., 2008), from which two principles acting in parallel can be extracted (Pesch and Hülskamp, 2009). First, the activator-inhibitor model explains pattern formation by a feedback loop between the activators and the inhibitors, in which the activators activate the inhibitors, which in turn repress the activators. The inhibitors mediate the cell-cell communication by intercellular movement. Second, the activator depletion model explains patterning by the removal of the mobile activator TTG1 in neighbouring cells through trapping by GL3 in trichome initials (Bouyer et al., 2008; Balkunde et al., 2011).

In addition to GL3 and EGL3, the bHLH gene AtMYC1 has been shown to participate in the regulation of trichome patterning (Symonds et al., 2011; Zhao et al., 2012). Formally, AtMYC1 acts as a positive regulator of trichome initiation because trichome number is reduced in mutants (Symonds et al., 2011; Zhao et al., 2012). Promoter swap experiments showed that GL3/EGL3 can replace AtMYC1, whereas AtMYC1 cannot rescue gl3 egl3 mutants, suggesting a redundant but also a divergent function (Zhao et al., 2012). The AtMYC1 gene was initially isolated in a screen for Antirrhinum DELILA homologues in Arabidopsis (Urao et al., 1996). Expression analyses suggest that AtMYC1 functions downstream of the patterning genes but upstream of GL2. Chromatin immunoprecipitation experiments have shown that GL3/GL1 bind to the AtMYC1 promoter chromatin (Morohashi and Grotewold, 2009), suggesting transcriptional regulation of AtMYC1 by GL1 and GL3. In the root hair system, AtMYC1 is controlled by WER and CPC in a similar manner to that described for GL3/EGL3 (Bernhardt et al., 2005; Bruex et al., 2012). Its function upstream of GL2 was inferred from the finding that GL2 expression is reduced in atmyc1 mutants (Zhao et al., 2012). The AtMYC1 protein has been grouped in a phylogenetic tree together with GL3, EGL3 and TT8 in subgroup IIIf of Arabidopsis bHLH proteins (Heim et al., 2003). Yeast two-hybrid assays revealed that, in addition to GL3, AtMYC1 protein also interacts with most of the other patterning proteins including the above-mentioned members CPC, TRY, TTG1, GL1 and MYB23 (Zimmermann et al., 2004; Tominaga et al., 2008; Zhao et al., 2012). However, in contrast to GL3 and EGL3, the AtMYC1 protein appears to be unable to form homo- or heterodimers with GL3/EGL3 (Zhao et al., 2012).

Here, we explore the molecular basis of AtMYC1 function. We show that GL2, TRY and CPC expression patterns are unchanged in atmyc1 mutants and overexpression lines. In contrast to GL3 and EGL3, AtMYC1 protein is located in the cytoplasm. When AtMYC1 is co-expressed with GL1, GL1 is relocated into the cytoplasm. Co-expression of AtMYC1 with TRY or CPC leads to the recruitment of AtMYC1 into the nucleus, as well as to transport of TRY/CPC from the cytoplasm into the nucleus. Our genetic analysis indicates that AtMYC1 inhibits the function of TRY/CPC.

Plants were grown on soil at 24°C with 16 hours light/day. Plant transformations were performed by the floral dip method (Clough and Bent, 1998). We used the Col-0 wild type, gl3-3 egl3-3 (GK 545D05; SALK_077439) and atmyc1-1 (Salk 057388) to transform 35S:GL3, 35S:EGL3 and 35S:AtMYC1 in the respective backgrounds. 5′-AtMYC1:AtMYC1, 35S:YFP-AtMYC1, 35S:YFP-NLS-AtMYC1 and 35S:YFP-NES-AtMYC1 were transformed in atmyc1-1 plants. 5′-TRY:GUS, 5′-CPC:GUS and 5′-GL2:GUS were transformed in Col-0 plants. The respective T2 lines of 5′-TRY:GUS, 5′-CPC:GUS and 5′-GL2:GUS were crossed with atmyc1-1. The 35S:GL1, gl1 (SalK_039478), try-JC and cpc-1 alleles (Wada et al., 1997; Larkin et al., 1999) were crossed with one selected T2 line of 35S:AtMYC1 Col-0 and with atmyc1-1. The insertion of all atmyc1 alleles was confirmed: Salk_006354, Salk_057388, Salk_056899, Gabi_621-F09 and Sail_277_H01. The different combinations of the atmyc1 mutants were created by crossing of atmyc1-1, gl3-3 (GK 545D05) (Rosso et al., 2003; Jakoby et al., 2008) and egl3 (Salk_019114) (Wester et al., 2009). The rescue efficiency is presented by box-whisker plots (Tukey, 1977).

Particle bombardment (Pesch and Hülskamp, 2011), protoplast isolation (Col ecotype) and transfection (Spitzer et al., 2006) were performed as described previously. Tobacco leaves were transiently transformed by infiltration with Agrobacterium tumefaciens (Agrobacterium strain GV3101 pMP90RK) and co-infiltrated with the antisilencing strain RK19 (Voinnet et al., 1999).

CDSs were cloned in pENTR1A/pENTR4 (Invitrogen) or introduced through BP recombination in pDONR201 (Invitrogen). Mutations in TRY, CPC and GL1 were created by site-directed mutagenesis (supplementary material Table S3). The vector pENTR1A-w/o-ccdB was created by EcoRI digestion and religation to remove almost the entire Gateway recombination cassette between the attB sequences of pENTR1A. This construct was used in LR recombinations to produce expression vectors without CDS. AtMYC1-w/o-Stop-pDONR201 was created via PCR and BP recombination in pDONR201 using AtMYC1-pDONR201 as template (supplementary material Table S3).

NLS-AtMYC1-pDONR201 and NES-AtMYC1-pDONR201 are derivatives of AtMYC1-pDONR201. To generate these constructs, tandem copies of BamHI and XhoI restriction sites were inserted directly after the start codon of AtMYC1 in AtMYC1-pDONR201 to obtain BamHI-XhoI-AtMYC1-pDONR201 (supplementary material Table S3). In the next step, the SV40 NLS was generated by annealing the oligonucleotides BamHI-NLS-XhoI-for and BamHI-NLS-XhoI-rev encoding amino acids LQPKKKRKVGG. The NES from protein kinase inhibitor (PKI) was produced by annealing the BamHI-NES-XhoI-for and BamHI-NES-XhoI-rev oligonucleotides encoding the amino acids LQNELALKLAGLDINKTGG. The annealed linkers were inserted into BamHI-XhoI-AtMYC1-pDONR201 using the BamHI and XhoI restriction sites (supplementary material Table S3).

The 5′ AtMYC1 region (-1 to -1963) was cloned in pAM-PAT-GWPro_35S (GenBank AY436765) (supplementary material Table S3). The complementation construct 5′-AtMYC1:AtMYC1 was created by recombination of 5′-AtMYC1-pAMPAT with AtMYC1-pDONR201. Reporter constructs were created by recombining pENTR-GUS-201 (Invitrogen) with 5′-TRY-pAM-PAT, 5′-CPC-pAM-PAT and 5′-GL2-pAM-PAT (Weinl et al., 2005; Wester et al., 2009). 35S:CDS constructs were generated by LR recombination of the respective CDS entry clones (GL3, EGL3, AtMYC1) and pAM-PAT-GWPro_35S (GenBank AY436765). 35S:AtMYC1-YFP was created by LR recombination of AtMYC1-w/o-Stop-pDONR201 and pEX-G-YFP (Feys et al., 2005). The N-terminal YFP, RFP and CFP fusion constructs were created through recombination of the respective entry clones or pENTR1A-w/o-ccdB with pENS-G-YFP, pENS-G-CFP and pNmR (Feys et al., 2005) (gift of A. Schrader, Botanical Institute, Cologne, Germany), respectively. The 35S:GFP-YFP construct was described previously (Digiuni et al., 2008). For BiFC analysis, the pBATL vectors pCL112 YFP N-term and pCL113 YFP C-term (kindly provided by Sean Chapman, SCRI, Dundee, UK) were used after LR recombination of the entry clones containing the different CDSs.

For the nuclear trapping system, the Gateway recombination cassette was removed by HindIII and XbaI digestion from pBluescript-GW-RekA (gift of S. Bieri, MPI for Plant Breeding, Cologne, Germany) and introduced into pVT-U (Vernet et al., 1987) to create pVT-U-attR. Patterning genes were introduced by LR recombination with the respective entry clones. The plasmid vector pNS and pNS-GL3 were described previously (Ueki et al., 1998; Balkunde et al., 2011). AtMYC1 CDS was cloned as a SalI-HindIII fragment from SalI-AtMYC1-HindIII-pGEM-T-easy (supplementary material Table S3) into the SalI site of pNS to create pNS-AtMYC1 (supplementary material Table S3). GL1 CDS was cloned as a SalI-XhoI fragment into the respective sites of pNS to produce pNS-GL1. pNS-TRY was created by cloning the TRY CDS as a PvuII-NcoI fragment into SalI (blunted) and NcoI cut pNS.

To produce control vectors without CDS fusions, the destination vectors (pC-ACT2-attR, pAS-attR and pVT-U-attR) were recombined with pENTR1A-w/o-ccdB.

To fuse luciferase or protein A to patterning proteins the two destination vectors pcDNA3-Rluc-GW and pTREX-dest30-ntPrA (Blasche et al., 2013) were recombined by LR reaction with the entry clones of interest or with pENTR1A-w/o-ccdB.

Yeast two-hybrid assays were performed as described previously (Gietz et al., 1995). The nuclear trapping assays used pNS and pVT-U vectors that were co-transformed in the yeast strain EGY48 (Clontech). The nuclear import of the NESLexAD fusion proteins was assayed on medium lacking leucine and histidine or lacking leucine, histidine and uracil.

For LUMIER assays, proteins were transiently co-expressed in HEK293TN cells (BioCat/SBI LV900A-1) as hybrid proteins with Staphylococcus aureus protein A or with the Renilla reniformis luciferase tag fused to their N-terminus. The LUMIER assay is based on the technique used by Barrios-Rodiles et al. (Barrios-Rodiles et al., 2005) and was optimised for our proteins as follows. Constructs (1.5 μg plasmid DNA) were co-transfected into 1×10⁶ HEK293TN cells in 6-well plates using 10 μl Lipofectamine 2000 (Invitrogen). After 48 hours, the medium was removed and cells were washed with PBS. The cells were collected by centrifugation (600 g, 15 minutes) and then lysed on ice in 50 μl ice-cold lysis buffer [20 mM Tris pH 7.5, 250 mM NaCl, 1.1% Triton X-100, 10 mM EDTA, 10 mM DTT, Protease Inhibitor Cocktail (Roche, 1 836 170) and benzonase (Novagen, 70746; 0.0125 units per μl final concentration)] for 1 hour. Subsequently, lysates were centrifuged at 15,000 rpm (21,380 g) for 15 minutes at 4°C, and the supernatant was mixed with 10 μl PBS-prewashed sheep anti-rabbit IgG-coated magnetic beads (Invitrogen, Dynabeads M280; 2 mg/ml final concentration) and incubated for 1 hour on ice. The beads were collected using a magnetic holder (Neodym-Magnets, 1.3 Teslar) and luminescence measured in a microtitre plate reader (BMG Fluostar Optima). Luciferase activity was determined for the input and the pulldown. Lysis buffer, PBS and untransformed cells served as a negative control. Pulldown efficiency was calculated by: [RLU pulldown/(RLU input×5)]×100.

Real-time PCR reactions were performed using the SYBR Green Real-Time PCR Master Mix (Applied Biosystems). To quantify RNA expression levels, published primers pairs were used: CPC, TRY primers (Morohashi et al., 2007), TRY primers (Zhao et al., 2008). Actin primers were used as an endogenous control as previously described (Zhao et al., 2008). Three biological replicates, each including two technical replicas, were measured. Relative RNA levels were calculated according to the comparative Ct (2^-ΔΔCt) method.

GUS activity was assayed as described (Sessions et al., 1999). Light and fluorescence microscopy were performed using a Leica DMRE microscope or a Leica MZ16F binocular microscope. Images were processed using Adobe Photoshop Elements 7.0. Co-bombardments in Arabidopsis cells and BiFC analyses in leek cells were studied with a Leica TCS SP2 confocal laser-scanning microscope using a UV HCX APO L 40.0×/0.80 W or UV/HCX APO L U-V-I 63.0×/0.90 W water-immersion objective or a PL FLUOTAR L 20.0×/0.40 BD dry objective. YFP was excited at 514 nm and emitted light was detected at 520-560 nm. RFP was excited at 561 nm and detected in the 570-650 nm range. Colocalisation studies with YFP and RFP fusion proteins in leek cells were imaged with a Leica TCS SPE confocal laser-scanning microscope using an ACS APO 10.0×/0.30 dry objective. YFP was excited at 488 nm and emitted light captured at 511-550 nm. RFP was excited at 561 nm and detected in the 570-602 nm range. Simultaneous detection of RFP and YFP was performed using the sequential scanning facility of the microscopes.

Quantification of fluorescence signal intensities was performed using Leica LAS AF and Leica LCS software. Images were created by overlying the single stack images of the cell. Each pixel in the confocal image was assigned a greyscale intensity value that ranges from black (0) to white (255). The pixel sum (sum of the measured greyscale intensity values) of the overlying image was measured for two regions of interest (whole cell and nucleus). The pixel sum for the whole cell was defined as 100% and the nucleus pixel sum expressed as a percentage of this. Images of Arabidopsis leaves were created using the Leica MZ16F binocular microscope and the z-stack function of Leica Application Suite Version 3.7.0. Scanning electron microscopy (SEM) images were taken with an FEI Quanta 250 FEG scanning electron microscope.

Transformed leaf material was homogenised in liquid nitrogen using a mortar and pestle. Leaf powder was incubated with lysis buffer [150 mM NaCl, 1% Triton X-100, 50 mM Tris HCl pH 8, 0.1% SDS, 10 mM DTT, Roche Complete Protease Inhibitor Cocktail (1 tablet/40 ml)] for 30 minutes on ice, centrifuged at 10,000 g? for 15 minutes at 4°C, the supernatant denatured by addition of 6× concentrated protein sample buffer (0.375 M Tris HCl pH 6.8, 12% SDS, 60% glycerol, 0.6 M DTT, 0.06% Bromophenol Blue) and heated for 15 minutes at 95°C. Samples were analysed by western blot using mouse anti-GFP IgG (Roche, 1:1000) and anti-mouse IgG (Sigma, 1:50,000), both in 5% milk powder in PBST (PBS with 0.1% Tween 20). For detection of protein bands the SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific) and the ImageQuant LAS 4000 (GE Healthcare) detection system were used.

As recently described for three alleles (Symonds et al., 2011; Zhao et al., 2012), we found a strong reduction in trichome number in all five atmyc1 alleles tested in the Col wild-type background. When comparing the trichome number for each on the first two leaf pairs, we found a similar degree of reduction in all alleles independent of the T-DNA insertion position (supplementary material Table S1, Fig. S1A). We focused our analysis on the atmyc1-1 allele (SALK-057388), which does not produce a full-length mRNA and is therefore considered a null allele (supplementary material Fig. S1D).

Analysis of the trichome number in atmyc1-1 egl3 (Salk-019114) and atmyc-1 gl3-3 double mutants compared with wild type showed subtle but statistically significant differences (Table 1). Since egl3 and gl3-3 single mutants also exhibit a reduced trichome number, this phenotype is considered additive. When we examined trichome branch number in the single and double mutants, we found no significant genetic effect on each other (Table 1).

Overexpression of AtMYC1 under the control of the CaMV 35S promoter resulted in an ∼2-fold increase in trichome number in the Col background (supplementary material Fig. S2; Tables 2, 3). This raised the question of whether 35S:AtMYC1 can rescue the gl3 egl3 phenotype. We compared 35S:GL3, 35S:EGL3 and 35S:AtMYC1 with respect to their ability to rescue the gl3 egl3 trichome phenotype (Fig. 1A). We found full rescue with 35S:GL3, partial rescue with 35S:EGL3, and no rescue with 35S:AtMYC1. This indicates that, despite its function as an activator of trichome development, AtMYC1 cannot replace GL3/EGL3 function. This is consistent with the finding by Zhao and co-workers that AtMYC1 expression under control of the GL3 and EGL3 promoters cannot rescue the gl3 egl3 double mutant (Zhao et al., 2012). Conversely, 35S:GL3 or 35S:EGL3 caused substantial overproduction of trichomes in the atmyc1-1 mutant background (Fig. 1B), whereas 35S:AtMYC1 showed a moderate rescue. This indicates that GL3 and EGL3 do not require AtMYC1 to induce extra trichome formation.

The finding that AtMYC1 cannot replace GL3 and EGL3 function prompted us to assess the ability of AtMYC1 to regulate downstream genes. We analysed the temporal and spatial expression of pTRY:GUS, pCPC:GUS and pGL2:GUS in atmyc1-1 mutants and the corresponding Col background. At all leaf developmental stages we found indistinguishable expression patterns for pTRY:GUS, pCPC:GUS and pGL2:GUS in atmyc1-1 mutants and wild type (Fig. 2A-F; supplementary material Fig. S3A-F). Also, overexpression of AtMYC1 did not alter the expression pattern of pTRY:GUS, pCPC:GUS or pGL2:GUS (Fig. 2G-I; supplementary material Fig. S3G-I). To test whether the expression levels of TRY, CPC or GL2 are changed in atmyc1-1 mutants or AtMYC1 overexpression lines, we employed qPCR (Table 4). Although overexpression of AtMYC1 had no substantial influence on the expression levels, a slight but significant reduction was seen for CPC and GL2 in the atmyc1 mutant. The latter is consistent with the findings of a previous study (Zhao et al., 2012).

Using the yeast two-hybrid method, AtMYC1 has been reported to interact with various proteins involved in trichome patterning, including ETC3 (Tominaga et al., 2008), GL1 and TTG1 (Symonds et al., 2011), WER and TRY (Zhao et al., 2012). The interactions with GL1, WER and TRY were confirmed by bimolecular fluorescence complementation (BiFC) (Zhao et al., 2012). We aimed to confirm and extend the characterisation of these interactions. Our yeast two-hybrid screens confirmed these interactions and also revealed, as expected, that AtMYC1 interacts with CPC (supplementary material Fig. S4). We further show that single amino acid exchanges (R/D) in the MYB-R/B-like bHLH transcription factor interface of the MYB factors (Zimmermann et al., 2004) strongly reduce the interaction of GL3 with GL1, TRY and CPC. These amino acid exchanges in the respective MYB factors also eliminate the interaction with AtMYC1. These data suggest that binding of AtMYC1 with the MYB factors occurs by a similar mechanism to that employed with GL3.

To provide independent support of the interaction between AtMYC1 and the other patterning proteins we used the luminescence-based mammalian interactome (LUMIER) assay (Barrios-Rodiles et al., 2005). Protein fusions to protein A (ProtA) or to luciferase were expressed in human HEC293TN cells. We immunoprecipitated the ProtA-tagged protein with IgG beads and quantified the co-immunoprecipitated proteins by luciferase assay. This confirmed the interaction of AtMYC1 with GL1, TRY, CPC and TTG1 (Table 5).

We used transient expression assays to determine the intracellular localisation of AtMYC1. YFP-AtMYC1 localisation was studied in Agrobacterium-transfected tobacco cells. In contrast to YFP-GL3 (Fig. 3C), we found YFP-AtMYC1 fluorescence predominantly in the cytoplasm (Fig. 3B,E). This was unexpected for a bona fide transcription factor. We therefore studied the localisation behaviour in more detail. First, we confirmed the integrity of the YFP-AtMYC1 fusion protein in a western blot using a GFP antibody (Fig. 3L). We then analysed the localisation of AtMYC1 in four additional cell types: Arabidopsis protoplasts (Fig. 3H,I), Arabidopsis cotyledons (Fig. 3F), onion cells (Fig. 3K) and leek epidermal cells (supplementary material Fig. S5). In all tested cell types, AtMYC1 was localised predominantly in the cytoplasm. By contrast, GL3 was localised, as expected, in the nucleus (Fig. 3G,J). This visual impression was confirmed by quantifying the percentage of nuclear and cytoplasmic fluorescence (Tables 6, 7). To test whether the localisation of AtMYC1 depends on the construction of the fusion proteins, we also studied the C-terminal fusion of YFP to AtMYC1 (AtMYC1-YFP). Again, we found a predominantly cytoplasmic localisation (Fig. 3I). Finally, we excluded the possibility that fusion to the YFP protein affects localisation by showing that an N-terminal fusion with RFP (RFP-AtMYC1) is also found predominantly in the cytoplasm (Fig. 4, Table 6).

Our finding that AtMYC1 localisation differs from that of its MYB interaction partners raised the question of whether AtMYC1 can recruit them from the nucleus into the cytoplasm or vice versa. We tested this by co-expressing RFP-AtMYC1 and YFP-GL1 or YFP-TRY in Arabidopsis cotyledon cells (Fig. 4). YFP-GL1 is localised mainly in the nucleus. In cells co-expressing RFP-AtMYC1 and YFP-GL1 we found YFP-GL1 in the cytoplasm (Fig. 4; supplementary material Fig. S5). Surprisingly, co-expression of RFP-AtMYC1 and YFP-TRY led to the localisation of RFP-AtMYC1 in the nucleus (Fig. 4). This was confirmed by a quantitative analysis using the YFP and RFP tags reciprocally to exclude biases caused by the tags (Tables 6, 7). The quantification revealed that the co-expression of TRY and AtMYC1 also caused an accumulation of TRY in the nucleus. When expressed alone, between 40% and 67.5% of the fluorescent protein fused to TRY was found in the cytoplasm. When co-expressed with AtMYC1, TRY protein was almost exclusively found in the nucleus (95.1%, 92.9%, see Tables 6 and 7). When combining RFP-AtMYC1 and YFP-CPC, we found similar, although slightly less drastic, localisation behaviours as with TRY (Table 7). The relocalisation was also seen in leek cells, demonstrating that this recruitment is not dependent on cell type (supplementary material Fig. S5).

Upon co-expression of AtMYC1, GL1 and TRY, we found all three in the nucleus (supplementary material Fig. S5), suggesting that TRY can counteract the AtMYC1-dependent recruitment of GL1 into the cytoplasm.

To further corroborate these findings, we studied the localisation of the GL1-AtMYC1, TRY-AtMYC1 and CPC-AtMYC1 interactions using BiFC assays. The TRY-AtMYC1 and CPC-AtMYC1 interactions occurred predominantly in the nucleus, whereas the GL1-AtMYC1 interaction took place in the cytoplasm (supplementary material Fig. S6).

We used the nuclear transportation trap (NTT) assay (Ueki et al., 1998) to independently test the nuclear transport behaviour of AtMYC1 in a more stringent manner. In the NTT assay the protein under consideration is expressed as a translational fusion to the artificial transactivator LexAD (LexA DNA-binding domain combined with GAL4AD transactivation domain) fused to a nuclear export signal (NES) from the HIV Rev protein. The NES can trigger the nuclear export of proteins lacking a nuclear localisation signal (NLS). A functional NLS is sufficient to overcome the NES-mediated nuclear export resulting in activation of the LEUCINE2 (LEU2) reporter gene. Using the NTT assay we found that AtMYC1, in contrast to GL3 (Balkunde et al., 2011), could not activate the LEU2 reporter, indicating that it does not contain a functional NLS. Significantly, additional expression of TRY or CPC led to the activation of LEU2, whereas co-expression of GL1 did not activate the reporter. These findings confirm the results of our transient assays and support the conclusion that TRY and CPC can trigger nuclear transport whereas GL1 cannot direct AtMYC1 into the nucleus (Fig. 5).

The reduction in trichome number in atmyc1-1 mutants and the supernumerary trichomes in 35S:AtMYC1 lines indicate that AtMYC1 is a positive regulator of trichome development. At the molecular level, AtMYC1 modulates the behaviour of GL1 by recruiting it from the nucleus to the cytoplasm, and AtMYC1 is recruited by TRY or CPC to the nucleus. This raises the question of whether the interaction with GL1 and/or TRY and CPC is biologically relevant. To test this, we analysed the genetic interactions between AtMYC1, GL1, CPC and TRY using combinations of overexpression lines and mutants. The try atmyc1 double mutant showed an increased cluster frequency, indicating that the atmyc1 mutant enhances the try cluster phenotype (Table 2). Overexpression of AtMYC1 in try mutants caused an increase in trichome number, but no change in cluster frequency as compared with try mutants (Table 2). Thus, extra trichome initiation by overexpressed AtMYC1 is independent of the presence of TRY. The atmyc1 cpc double mutants showed essentially additive phenotypes. This experiment was performed with cpc1-1 in the Ws background (Table 3) and repeated with the cpc1-2 allele in the Col background in order to exclude the possibility that a quantitative trait locus (QTL) modifies the trichome patterning phenotype in the double mutant (supplementary material Table S2). Overexpression of AtMYC1 in cpc mutants caused a slight increase in cluster frequency. Lines overexpressing both GL1 and AtMYC1 showed a substantial increase in trichome number (Fig. 6C). Frequently, we found large clusters in which the outermost trichomes appeared smaller than at the centre, suggesting that they had been initiated later in leaf development (Fig. 6E). Overexpression of AtMYC1 has a similar genetic effect as the absence of TRY in cpc try and 35S:GL1 try mutant lines (Schnittger et al., 1998; Schellmann et al., 2002), suggesting that AtMYC1 is negative regulator of TRY.

In order to test whether the intracellular localisation of AtMYC1 is biologically relevant, we created N-terminal fusions with an NLS (Kalderon et al., 1984) and with an NES (Wen et al., 1995; Matsushita et al., 2003). We first tested the intracellular localisation of the respective YFP-tagged fusion proteins by particle bombardment in leek cells. As expected, YFP-NES-AtMYC1 was found in the cytoplasm and YFP-NLS-AtMYC1 in the nucleus (Fig. 7). YFP-NLS-AtMYC1 and RFP-TRY co-expression resulted in an accumulation of both fusion proteins in the nucleus (Fig. 7). In cells co-expressing YFP-NES-AtMYC1 and RFP-TRY we found the RFP signal in the nucleus and the cytoplasm, rather than predominantly in the nucleus (Fig. 7). The co-expression of YFP-NLS-AtMYC1 or YFP-NES-AtMYC1 with CFP-GL1 led to an accumulation of CFP in the nucleus or cytoplasm, respectively (Fig. 7). Together, these data indicate that the mislocalisation of AtMYC1 triggers a change in the intracellular localisation of TRY and GL1.

In a second set of experiments, we constitutively expressed YFP-AtMYC1, YFP-NES-AtMYC1 and YFP-NLS-AtMYC1 in atmyc1 mutants and assessed the rescue of trichome number. As individual transformants may show a wide range of phenotypes, we studied at least 50 T1 plants (Pesch and Hülskamp, 2011) in order to catch the full phenotypic spectrum for our comparisons. In 35S:YFP-AtMYC1 atmyc1-1 lines we found an average of 151.3±92 trichomes on the first four leaves (n=50 T1 lines). Compared with 35S:YFP-AtMYC1, the 35S:YFP-NES-AtMYC1 construct exhibited significantly less efficient rescue of atmyc1-1 (112.1±44.5; n=51 T1 lines; Wilcoxon signed-rank test, P=0.02). Conversely, the 35S:YFP-NLS-AtMYC1 construct rescued the atmyc1-1 mutant significantly better (176.5±81.5; n=50 T1 lines; Wilcoxon signed-rank test, P=0.005). Together, these data show that changing the intracellular localisation of AtMYC1 changes its biological activity.

AtMYC1 is one of the genes regulating trichome and root hair patterning in Arabidopsis thaliana. It belongs to the large family of bHLH factors in Arabidopsis, which contains many key regulators of plant development. A phylogenetic analysis of the bHLH domains of 133 bHLH proteins placed AtMYC1 in subgroup IIIf (Heim et al., 2003). In addition to AtMYC1, this subgroup contains three further members, GL3, EGL3 and TT8, which are all involved in TTG1-dependent pathways related to trichome, root hair and seed coat development and anthocyanin production (Broun, 2005; Koes et al., 2005; Ramsay and Glover, 2005; Lepiniec et al., 2006; Serna and Martin, 2006; Balkunde et al., 2010; Feller et al., 2011; Tominaga-Wada et al., 2011). In support of this phylogenetic classification, several observations have suggested that AtMYC1 acts redundantly with GL3 and EGL3 in the root hair patterning system (Bruex et al., 2012) and in the context of trichome patterning (Zhao et al., 2012). In addition, AtMYC1 shows interactions with other proteins, similar to those reported for GL3, EGL3 and TT8, in the yeast two-hybrid assay (Zimmermann et al., 2004; Tominaga et al., 2008; Zhao et al., 2012) and in pulldown experiments (this study). However, two lines of genetic evidence indicate that the molecular function of AtMYC1 differs from that of GL3, EGL3 and TT8. First, in contrast to the gl3 egl3 double mutant the atmyc1 gl3 or atmyc1 egl3 double mutants show an additive phenotype but not an enhancement of the single-mutant phenotypes of gl3 or egl3. Second, the expression of AtMYC1 under the control of the GL3 or EGL3 promoter (Zhao et al., 2012) or the 35S promoter (this study) cannot rescue the gl3 egl3 mutant phenotype. A divergent function of AtMYC1 is consistent with two phylogenetic studies that analysed 147 bHLH factors in Arabidopsis or bHLH factors in rice and Arabidopsis (Toledo-Ortiz et al., 2003; Li et al., 2006). These studies placed AtMYC1 in a clade distinct from GL3, EGL3 and TT8.

A possible molecular function can be deduced from the intracellular localisation of AtMYC1 protein. In contrast to a previous study (Zhao et al., 2012), we found AtMYC1 predominantly in the cytoplasm. As the AtMYC1 CDS was successfully used for the rescue experiments and shows the expected protein-protein interactions in yeast and in pulldown assays, the AtMYC1 sequence used in this study is fully functional. AtMYC1 was consistently found in the cytoplasm of cells from four different plant species (tobacco, onion, leek and Arabidopsis) with N- and C-terminal YFP fusions and with an RFP fusion. We confirmed our results with the yeast NTT system, which is specifically designed to test the nuclear transport of proteins. Thus, our results consistently show that AtMYC1 predominantly localises in the cytoplasm. Mislocalisation of AtMYC1 protein, either to the cytoplasm or to the nucleus, showed that the correct localisation of the protein is important for its regulatory function. It is interesting to note that AtMYC1 contains a monopartite NLS sequence at its N-terminus that is predicted to target the protein predominantly to the nucleus with a similar efficiency as that predicted for GL3 (supplementary material Table S4). This suggests that the AtMYC1 NLS sequence is masked under normal conditions either by protein folding or binding to additional factors, which are likely to be general factors as the cytoplasmic localisation was observed in various species and cell types.

Our observations of the quantitative relocalisation of AtMYC1, TRY/CPC and GL1 are based on the overexpression of the respective proteins. Therefore, we do not know the relative distribution of these proteins under native conditions. Our attempts to study this in planta have thus far failed because AtMYC1, TRY or GL1 fusions with YFP or RFP were below the detection limit when expressed under the endogenous promoters. Nevertheless, it is evident that the four proteins have the property to balance each other’s localisation. Given that GL1 and TRY/CPC are involved in the transcriptional regulation of other genes, it is likely that their relocalisation changes their activity: the recruitment of AtMYC1 by TRY/CPC and concomitant accumulation of TRY/CPC in the nucleus reflects the fact that AtMYC1 binding to TRY/CPC can modulate the balance between nuclear and cytoplasmic TRY/CPC protein. The in vivo net effect is likely to be a suppression of both TRY and CPC activity. This is suggested by several genetic observations. If AtMYC1 represses TRY and/or CPC one would expect lines overexpressing AtMYC1 to share phenotypic aspects of try and cpc mutants. A repression of CPC activity is consistent with an increased trichome number in 35S:AtMYC1 lines. A repression of TRY activity by AtMYC1 could explain the substantial increase in trichome number and the slightly increased production of trichome clusters in 35S:AtMYC1 cpc and the large trichome clusters in 35S:AtMYC1 35S:GL1 lines because this phenotype is reminiscent of the 35S:GL1 try phenotype (Schnittger et al., 1998; Szymanski and Marks, 1998). Also, the decrease in trichome number by 50% compared with the wild type in the atmyc1 mutant can be interpreted as being caused by enhanced activity of CPC or TRY. The recruitment of GL1 by AtMYC1 from the nucleus into the cytoplasm is likely to impair the activity of GL1. However, such a repressive role of AtMYC1 on GL1 is not consistent with our genetic finding that the reduced trichome number phenotype of 35S:GL1 plants is rescued in a 35S:AtMYC1 background. This suggests that the impact of AtMYC1 on the localisation of GL1 has little biological relevance or that the impact on TRY/CPC overwrites this effect. The latter is consistent with the observation that the relocalisation of GL1 by AtMYC1 is counteracted in the presence of TRY protein.

How is AtMYC1 integrated in the gene regulatory network? A putative repressive function of AtMYC1 on TRY/CPC is difficult to functionally integrate into existing models. One attractive possibility for its function can be inferred from the finding that the AtMYC1 promoter is a direct target of the GL1-GL3/EGL-TTG1 activator complex (Morohashi and Grotewold, 2009). This suggests that the activator complex, and therefore also GL1, might activate the expression of AtMYC1 as well as of TRY/CPC. On top of this regulation at the transcriptional level we propose that the protein activities of TRY/CPC and GL1 are coupled through AtMYC1. As AtMYC1 can recruit GL1 from the nucleus into the cytoplasm and TRY/CPC from the cytoplasm into the nucleus the regulatory effect should be antagonistic (Fig. 8). In our model, the concentration of GL1 in the nucleus is reduced in the presence of AtMYC1, whereas the concentration of TRY in the nucleus is increased. It is possible that, in addition to this cell-autonomous effect, a reduced level of cytoplasmic TRY/CPC would lead to reduced lateral movement and therefore to less repression in the neighbouring cells. Consistent with this scenario, the distance between trichomes is markedly increased in atmyc1 mutants as compared with wild type. This type of negative-feedback loop could serve not only to fine-tune the network but would also provide an additional mechanism to modulate lateral repression.

We thank Birgit Kernebeck for excellent technical assistance; Irene Klinkhammer for cell suspension culture work; Hans-Peter Bollhagen (AG Büschges, Zoological Institute, University of Cologne) for help with SEM; Sigfried Werth for photographic documentation; and L. Leson for the installation of the magnetic holder. Vectors were kindly provided as follows: pAM-PAT-GW by B. Ülker (MPIZ, Cologne); pET3A-Ubi-PrA by M. Koegl (German Cancer Research Center, Heidelberg); pBluescript-GW-RekA by S. Bierei (MPIZ, Cologne); TRY, CPC, GL3, TTG1 entry clones by U. Herrmann (Department of Biosciences, Oslo, Norway); EGL3 and AtMYC1 entry clone by I. Zimmermann; pENTR1A-w/o-ccdB by Arp Schnittger (Institut de Biologie Moléculaire des Plantes du CNRS, Strasbourg, France); and TTG1Δ26 entry clone by Rachappa Balkunde (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA).