Ubiquitin/26S proteasome (UPS)-dependent proteolysis of a variety of cellular proteins plays an essential role in many basic cellular processes. UPS impacts transcriptional regulation by controlling the stability, and thus the activity, of numerous transcription factors (TFs). In Arabidopsis, trichome development and flavonoid metabolism are intimately connected, and several TFs have been identified that simultaneously control both processes. Here we show that UPS-dependent proteolysis of two of these TFs, GLABROUS 3 (GL3) and ENHANCER OF GL3 (EGL3), is mediated by ubiquitin protein ligase 3 (UPL3). Cell-free degradation and in planta stabilization assays in the presence of MG132, an inhibitor of proteasome activity, demonstrated that the degradation of GL3 and EGL3 proteins is 26S UPS-dependent. Yeast- or protoplast-based two-hybrid and bimolecular fluorescent complementation assays showed that GL3 and EGL3 interact via their C-terminal domains with the N-terminal portion of UPL3. Moreover, both TFs are stabilized and show increased activities in a upl3 mutant background. Gene expression analyses revealed that UPL3 expression is negatively affected by mutation in the gl3 locus, but is moderately upregulated by the overexpression of GL3, suggesting the presence of a regulatory loop involving GL3 and UPL3. Our findings underscore the importance of post-translational controls in epidermal cell differentiation and flavonoid metabolism.
In Arabidopsis, trichome development and flavonoid metabolism are intimately connected and influenced by hormonal and developmental signals, as well as by various biotic and abiotic stresses (Deikman and Hammer, 1995; Teng et al., 2005; Solfanelli et al., 2006; Cominelli et al., 2008; Gonzalez et al., 2008; Loreti et al., 2008; Gonzalez, 2009; Shan et al., 2009). The spatial and temporal expression of genes required for trichome development and anthocyanin biosynthesis is orchestrated by a ternary activator complex, comprising an R2R3 family MYB transcription factor (TF), a basic helix-loop-helix (bHLH) TF and a WD40 protein, and is considered a hallmark of combinatorial gene regulation in plants (Broun, 2005; Ramsay and Glover, 2005; Hichri et al., 2011). Arabidopsis trichome development is positively regulated by the action of GLABROUS 1 (GL1, an R2R3-MYB TF), GLABROUS 3 and ENHANCER OF GLABROUS 3 (GL3 and EGL3, respectively, both of which are bHLH factors), and TRANSPARENT TESTA GLABRA 1 (TTG1, a WD40 protein). Together, these proteins positively regulate the expression of GLABROUS 2 (GL2) that encodes a homeodomain TF required for trichome morphogenesis (Szymanski et al., 1998). The R2R3-MYB/bHLH/WD40 complex that positively regulates the late structural genes in the Arabidopsis flavonoid biosynthetic pathway involves MYB75/90/113/114, GL3/EGL3/TT8 and TTG1 (Walker et al., 1999; Borevitz et al., 2000; Nesi et al., 2000; Payne et al., 2000; Zhang et al., 2003; Baudry et al., 2004; Zimmermann et al., 2004; Stracke et al., 2007; Allan et al., 2008; Gonzalez et al., 2008). In addition, small single-repeat R3 MYB proteins, such as TRY, CPC and MYBL2, compete with R2R3-MYBs for binding with the bHLH factors and disrupt the activator complex, thereby suppressing trichome development and anthocyanin synthesis (Wada et al., 1997; Matsui et al., 2008). The regulatory network is further complicated by the addition of small regulatory RNAs. It has been shown that the R2R3-MYBs, MYB75, MYB90 and MYB113, are the targets of TAS4-siR81(–) and MIR828 (Rajagopalan et al., 2006; Hsieh et al., 2009).
The bHLH TF family is one of the largest known groups of TFs in Arabidopsis, with more than 160 members divided into 12 groups and subgroups based on structural similarities (Heim et al., 2003). GL3 and EGL3, which encode two homologous members of subgroup IIIf, are involved in trichome differentiation and anthocyanin biosynthesis (Payne et al., 2000; Zhang et al., 2003). GL3 and EGL3 function in a partially redundant manner, and have distinct expression patterns. Both are highly expressed in leaf primordia, but in the mature leaf, EGL3 is expressed at low levels both in pavement cells and trichomes, whereas GL3 expression is limited to trichomes (Zhao et al., 2008). Despite the GLABROUS designation, mutations in the GL3 locus do not cause a glabrous phenotype sensu strictu, but instead lead to the development of fewer, less branched trichomes (Hulskamp et al., 1994). Whereas the anthocyanin content in gl3 mutants is comparable with that in the wild type, egl3 mutants are characterized by a reduction in anthocyanin accumulation, with no apparent trichome defects, suggesting functional differentiation. However, both trichome and anthocyanin biosynthesis defects were enhanced in gl3 egl3 double mutants, revealing overlapping and complementary functions of GL3 and EGL3 (Zhang et al., 2003).
Although significant progress has been made in understanding the role of the R2R3-MYB/(E)GL3/WD40 complex in developmental and biochemical pathways, it is unclear whether this transcriptional activator complex is regulated post-translationally. Among the currently known post-translational control mechanisms, regulated proteolysis by the ubiquitin/26S proteasome system (UPS) stands out as a major regulator of TF activity (Thomas and Tyers, 2000; Dhananjayan et al., 2005). The UPS comprises E1, E2 and E3 enzymes, the concerted actions of which are responsible for the conjugation of polyubiquitin chains that target proteins for proteolysis by the multisubunit 26S proteasome (Smalle and Vierstra, 2004; Vierstra, 2009; Lyzenga and Stone, 2012). E3 enzymes typically determine substrate specificity by recognizing one or a small subset of target proteins, and, in plants, are classified into two subgroups, RING/U-boxes and HECT ubiquitin ligases (Smalle and Vierstra, 2004; Mazzucotelli et al., 2006; Hotton and Callis, 2008). RING/U-box E3s act as adapters between cognate E2s and target proteins. HECT E3s accept and covalently link ubiquitin from an E2 to a conserved Cys in their HECT domain before transfer to the protein substrate (Rotin and Kumar, 2009). The Arabidopsis genome encodes seven HECT E3s that are grouped into four subfamilies (UPL1/2, UPL3/4, UPL5 and UPL6/7; Downes et al., 2003). To date, a specific proteolysis target has only been identified for UPL5 (Miao and Zentgraf, 2010).
Genetic studies with UPL3 family members revealed a role in trichome development (Downes et al., 2003). Whereas the majority of leaf trichomes in wild-type plants are tri-radiate, upl3 and the allelic kak mutants have increased frequencies of trichomes with five or more branches (Downes et al., 2003; El Refy et al., 2003). The positive role of GL3 in trichome branching suggests that it is a candidate target for UPL3 (Downes et al., 2003). Subsequent double mutant analyses have further strengthened this hypothesis (Sako et al., 2010). The supernumerary trichome branching in kak seedlings was suppressed in a kak gl3 double mutant background, proving that GL3 plays a key role in the UPL3 loss-of-function phenotype. GL3 and EGL3 are homologous proteins (and thus potentially recognized by a single E3), and their combined actions control both trichome development and anthocyanin accumulation. Therefore, we hypothesized that, in addition to GL3, EGL3 activity is also under proteasome-dependent proteolysis control.
Here, we show that both GL3 and EGL3 are unstable TFs and are targeted for UPS-dependent proteolysis. We have provided multiple lines of evidence to demonstrate that the proteasomal degradation of GL3 and EGL3 is mediated by UPL3. Moreover, mutation in the gl3 locus negatively affects UPL3 expression, whereas overexpression of GL3 moderately upregulates it, suggesting the presence of a regulatory loop involving GL3 and UPL3.
GL3 and EGL3 are 26S proteasome targets
We first investigated whether GL3 and EGL3 are 26S proteasome targets using a cell-free degradation assay (Figure 1). Tagged GL3 (GST-GL3-FLAG) and EGL3 (GST-EGL3-FLAG) recombinant proteins were expressed and purified from Escherichia coli (Figure S1a,b). Equal quantities of recombinant proteins (100 ng) were then added to total protein extract from 1-week-old Col-0 seedlings supplemented either with DMSO (solvent control) or MG132, a proteasome activity inhibitor. Recombinant GL3 and EGL3 proteins were rapidly degraded under these conditions (GST-GL3-FLAG half-life, T1/2 = 24.23 min; GST-EGL3-FLAG T1/2 = 26.90 min; Figure 1); however, the addition of MG132 caused marked protein stabilization (T1/2 > 90 min; Figure 1). We concluded that both TFs are unstable and targeted for UPS-dependent degradation.
To independently test whether GL3 and EGL3 are degraded via UPS, we ectopically expressed both TFs in stable transgenic lines in their respective gl3-4 and egl3-2 backgrounds. To facilitate immunodetection, GL3 and EGL3 were C-terminally fused to a 3xFLAG tag. Both tagged versions complemented their respective loss-of-function mutations, indicating that the 3xFLAG tag did not interfere with either GL3 or EGL3 function (Figure S2a–d). To monitor protein stability, the resulting GL3-FLAGox (gl3-4) and EGL3-FLAGox (egl3-2) lines were treated with 200 μm cycloheximide (CHX) and protein depletion rates were analyzed by immunoblotting (Figure 2a,c). Consistent with the results from the cell-free degradation assay (Figure 1), both GL3 and EGL3 were unstable, and GL3 had a shorter half-life compared with EGL3. The transcript levels for both transgenes were not affected by CHX treatment, indicating that the depletion of GL3 and EGL3 was strictly based on the regulation of protein stability (Figure S3a).
Next, we tested the effect of chemical inhibition of proteasome activity on GL3 and EGL3 stability (Figure 2b,d). Three-week-old seedlings were treated for 120 min with CHX alone or in combination with MG132. As expected from the results of the cell-free degradation assays (Figure 1), both GL3 and EGL3 degradation were inhibited by MG132 (Figure 2b,d). Similar to CHX treatment, combined CHX and MG132 treatment did not affect transgene transcript levels (Figure S3b).
Identification of lysines required for GL3 and EGL3 degradation
Most proteins degraded by the proteasome are marked with a polyubiquitin chain covalently linked to a Lys residue by the sequential action of E1, E2 and E3 enzymes (Kurepa and Smalle, 2008; Kurepa et al., 2009). The identification of ubiquitination site(s) in any given target protein is often a challenge because of the large number of Lys residues in many proteins and the lack of clearly defined ubiquitin modification motifs (Catic et al., 2004). GL3 and EGL3 contain 34 (5.3% of the total amino acids) and 35 (5.9% of the total amino acids) Lys residues, respectively. UbPred, a web-based program trained on verified ubiquitination sites identified by large-scale proteomics studies, was used to identify ubiquitination sites in GL3 and EGL3 (Figure S4a; http://www.ubpred.org; Radivojac et al., 2010). For further analyses, we focused on the high-confidence putative ubiquitination sites K535 and K536 in GL3, and K391, K392, K493 and K495 in EGL3. Medium-confidence putative ubiquitination sites K66 and K346 in GL3, and K65 in EGL3, were also analyzed (Figure S4a).
Potential ubiquitin-bearing Lys residues were mutated, and mutant proteins were expressed and purified from E. coli (Figure S1a,b). The stability of the recombinant proteins was compared in a cell-free degradation assay. Interestingly, all high-confidence putative ubiquitination sites were represented by vicinal or adjacent lysines (Figure S4a). For some target proteins, mutation of the preferential ubiquitination site Lys is counteracted by the use of a neighboring Lys as the destabilizing residue (Ju and Xie, 2006). We reasoned that adjacent residues pairs, K535 and K536 (GL3), K391 and K392 (EGL3), and K493 and K495 (EGL3), are potential alternative acceptor sites because of their positions. Therefore, double mutants of these adjacent lysines were generated to increase the chance of protein stabilization. Lysine → arginine substitutions at medium-confidence putative ubiquitination sites K66 (GL3), K346 (GL3) and K65 (EGL3), as well as the high-confidence putative ubiquitination sites K391 and K392 (EGL3), did not lead to stabilization of the recombinant proteins (Figure 3a,b). On the other hand, the paired lysine → arginine substitutions at K535 and K536 of GL3, and at K493 and K495 of EGL3, led to significant stabilization for both TFs (Figure 3a,b), suggesting that these residues are needed for ubiquitin-dependent degradation. Notably, these Lys residues are located between the C-terminal bHLH and ACT-like dimerization domains of GL3 and EGL3 (Figure 3a,b). In addition, amino acid sequence alignment of the C-terminal domains of selected bHLH proteins from subgroup IIIf showed that K536 in GL3, and K495 in EGL3, are evolutionarily conserved (Figure S4b).
The increased stability of GL3 and EGL3 proteins containing lysin → arginine substitutions prompted us to further investigate the functional effects of this increased stability by comparing the transcriptional activity of the mutant with the corresponding wild-type protein. Plasmids expressing GL3, GL3K535R,K536R, EGL3 or EGL3K493R,K495R, fused to the GAL4 DNA binding domain, were co-electroporated into Arabidopsis protoplasts with a reporter containing the firefly luciferase gene under the minimal CaMV 35S promoter and 5X GAL response elements. Protoplasts expressing wild-type GL3 or EGL3 produced higher luciferase activity compared with the basal level activity of the reporter plasmid (Figure 3c); however, compared with wild-type GL3 (11.8 ± 0.8-fold) and EGL3 (6.0 ± 0.2-fold), considerably higher luciferase activity was detected in protoplasts expressing GL3K535R,K536R (22 ± 2.2-fold) or EGL3K493R,K495R (12 ± 1.3-fold), suggesting that an increased accumulation of the mutant proteins is most likely responsible for increased activity (Figure 3c).
UPL3 mediates the proteasome-dependent degradation of GL3 and EGL3
Mutations of the UPL3 gene encoding HECT ubiquitin protein ligase 3 lead to an increase in both trichome number and branching (Downes et al., 2003; El Refy et al., 2003). Because the function of UPL3 is to earmark proteins for degradation, it follows that one or more activators of trichome development (such as GL1, GL3, ZWI, FRC2, FRC4 and STI) are targeted by this E3 (Downes et al., 2003). Using a combination of biochemical and genetic assays, we tested whether GL3 and EGL3 are targeted for degradation by UPL3 (Figure 4).
For an E3 to interact with its target, it needs to be co-localized with its target proteins. Previous studies have shown that GL3 and EGL3 are expressed in developing leaves and root cells (Bernhardt et al., 2005; Zhao et al., 2008). Both GL3 and EGL3 expression are higher in developing and developed trichomes. In mature leaves, GL3 expression is mostly restricted to trichomes, whereas EGL3 is expressed in both trichome and surrounding cells. Compared with GL3, EGL3 expression is lower in trichomes but higher in surrounding cells (Zhao et al., 2008). Thus, we first examined the tissue-specific expression of UPL3 by using a stable transgenic line that expresses a UPL3pro:uidA transgene in the Col-0 background (Figure 4a). As expected, higher GUS activity was detected in the trichomes of mature leaves. UPL3 promoter-driven GUS expression was also detected in the vascular tissue and pavement cells of the mature leaf, as well as in cotyledonary leaves and roots of 4-day-old transgenic seedlings (Figure 4a). UPL3 showed an overlapping expression pattern with the expression of its candidate targets GL3 and EGL3.
Next, we compared the stability of recombinant GST-GL3-FLAG and GST-EGL3-FLAG proteins in a cell-free degradation assay using extracts from wild-type and upl3-2 seedlings (Figure 4b,c). The T-DNA insertion in upl3-2 is thought to express a truncated UPL3 protein that contains the N-terminal target interaction domain, but not the C-terminal catalytic HECT domain (Downes et al., 2003). Thus, this mutant UPL3 protein is predicted to interact with its target proteins, but cannot catalyze polyubiquitination and targeting to the 26S proteasome. Indeed, both recombinant GL3 and EGL3 were more stable in upl3-2 extracts compared with the wild type (Figures 4b,c), suggesting that UPL3 is involved in the regulation of GL3 and EGL3 stability.
To further investigate the role of UPL3 in GL3 and EGL3 proteolysis, we generated upl3-2 gl3-4 and upl3-2 egl3-2 double mutant lines expressing GL3-FLAGox and EGL3-FLAGox, respectively, and the resulting triple homozygous lines were selected using a combination of genotyping and western blotting (Figure S5). As expected for an E3 loss-of-function mutant and its target proteins, steady-state levels of both FLAG-tagged GL3 and EGL3 proteins were higher in upl3-2 gl3-4 and upl3-2 egl3-2 backgrounds compared with gl3-4 and egl3-2 single mutant controls (Figure 4d,e). Treatment with 200 μm CHX confirmed that both GL3 and EGL3 proteins are stabilized in the absence of UPL3 (Figure 4d,e). No difference in transcript levels of the transgenes was observed in any of the samples tested, thus confirming that the changes in protein abundance are caused by a post-translational mechanism (Figure S6).
Finally, we characterized trichome phenotype and anthocyanin accumulation in GL3-FLAGox upl3-2 gl3-4 and EGL3-FLAGox upl3-2 egl3-2 seedlings (Figures 4f,g). GL3-FLAGox upl3-2 gl3-4 seedlings, but not EGL3-FLAGox upl3-2 egl3-2 seedlings, developed large, aberrant trichomes with increased branch number that formed a reticulate texture on the leaf surface (Figure 4f). GL3-FLAGox upl3-2 gl3-4 and EGL3-FLAGox upl3-2 egl3-2 seedlings also accumulated higher anthocyanin levels compared with GL3-FLAGox gl3-4 and EGL3-FLAGox egl3-2 seedlings (Figures 4g and S7). Collectively, these results further accentuated the potential role of UPL3 as the E3 that mediates the ubiquitin- and proteasome-dependent degradation of both GL3 and EGL3.
UPL3 physically interacts with GL3 and EGL3
To determine whether UPL3 indeed targets GL3 and EGL3, we next tested the physical interaction with both TFs in yeast two-hybrid, protoplast-based two-hybrid and bimolecular fluorescent complementation (BiFC) assays (Figures 5 and 6). The UPL3 protein contains two recognizable functional domains: a C-terminal catalytic HECT domain and an N-terminal cluster of four Armadillo repeats, hypothesized to form the target interaction site (Downes et al., 2003). Because of the large protein size of UPL3 (1888 amino acids; Downes et al., 2003) and the well-documented function of Armadillo repeats in other proteins (Tewari et al., 2010), we generated a bait construct comprised only of the N-terminal domain (UPL31–470) fused to the GAL4 activation domain (Figure 5a). To facilitate domain mapping and avoid the auto-activation of bHLH constructs (Zimmermann et al., 2004), we prepared prey constructs containing the N- or C-terminal domains of either GL3 (GL31–209 and GL3400–637) or EGL3 (EGL31–205 and EGL3402–596), fused to the GAL4 DNA binding domain (Figure 5a). Following co-transformation of bait and prey constructs, only yeast lines containing UPL31–470, in combination with GL3400–637 or EGL3402–596, grew on selective media (Figure 5b). This suggested that UPL3 and (E)GL3 do, in fact, interact, and that the N-terminal domain of UPL3, with the Armadillo repeats, recognizes the C-terminal domains of GL3 and EGL3. Because GL3 and EGL3 are highly homologous and act similarly in the two-hybrid assay, we selected GL3 to further delineate the domain interacting with UPL3. UPL31–470 interacted only with GL3400–637, the C-terminal domain that contains the bHLH and ACT-like domains (Figure 5c). The C-terminal region of GL3 has been demonstrated to form homodimers and heterodimers with the C-terminal region of EGL3 (Payne et al., 2000; Zhang et al., 2003). We confirmed that the dimer formation is mediated by the C terminus of GL3 in a yeast two-hybrid assay, although, individually, the bHLH domain (GL3400–495) or the ACT-like domain (GL3552–637) did not homodimerize, as indicated by the lack of growth on quadruple selection medium (Figure S8). Taken together, these results indicate that the dimeric conformation mediated by both bHLH and ACT domains is required for interaction with UPL3.
To independently test these interactions, we performed a protoplast-based two-hybrid assay (Figure 6a). The N-terminal domain of UPL3 (UPL31–470) was expressed under the control of the MMV promoter as an N-terminal fusion with the GAL4 DNA binding domain, and full-length GL3 and EGL3 cDNAs were expressed under the control of the CaMV 35S promoter (Figure S9). Combinations of these effector plasmids were co-transformed into Arabidopsis protoplasts, along with a luciferase reporter containing five tandem repeats of the GAL4 response element upstream of a minimal CaMV 35S promoter. The MMVpro:BD-UPL31–470 construct alone increased luciferase activity (4.0 ± 0.2-fold), indicating self-activation in plant cells; however, the co-expression of MMVpro:BD-UPL3 with 35Spro:GL3 or 35Spro:EGL3 led to 8.02 ± 0.32-fold and 5.8 ± 0.30-fold increases of luciferase activity, compared with the respective controls, again suggesting that UPL31–470 interacts with both GL3 and EGL3 (Figure 6a).
Finally, the interactions of UPL3 with GL3 or EGL3 were confirmed by BiFC assay (Figure 6b). Plasmids containing full-length GL3 or EGL3 cDNA, fused to the N-terminal half of YFP (GL3-NYFP and EGL3-NYFP, respectively), and UPL31–470 fused to the C-terminal half of YFP (UPL3-CYFP), were co-expressed in Arabidopsis protoplasts. A strong YFP fluorescence signal was detected in the nucleus (marked by DAPI staining), indicating the reconstitution of the active YFP protein as a result of the interaction between UPL3 and GL3 or EGL3 (Figure 6b). From this set of results, we conclude that UPL3 physically interacts with both GL3 and EGL3.
UPL3 expression is altered in gl3 mutant and GL3-ox lines
To test whether GL3 and/or EGL3 are involved in a feedback mechanism, we compared UPL3 transcript levels in Col-0, single-mutant lines gl3-4 and egl3-2, double-mutant line gl3 egl3, and the complementation lines GL3-FLAGox (gl3-4) and EGL3-FLAGox, using real-time PCR (Figure 7). UPL3 expression was roughly twofold higher in GL3-FLAGox (gl3-4) plants compared with the Col-0 wild type, and Approximately 50% of the wild-type level in the gl3-4 and gl3 egl3 lines. In contrast, the UPL3 transcript levels in egl3-2 and EGL3-FLAGox (egl3-2) plants did not significantly differ from the wild-type level (Figure 7). Thus, it appears that UPL3 gene expression is moderately activated by GL3 but not by EGL3.
In recent years, a wealth of information has accumulated on the transcriptional regulation of trichome development and anthocyanin biosynthesis in Arabidopsis. However, little is known about the post-translational modification of TFs that control these pathways. The post-translational regulation of TFs by UPS is a highly dynamic system that recognizes and responds to changes in the cellular microenvironment. The importance of UPS in plants is signified by numerous studies that implicate regulated and UPS-dependent proteolysis in many signaling and metabolic pathways (Xie et al., 2002; Smalle and Vierstra, 2004; Jang et al., 2005; Zhang et al., 2005; Qi et al., 2011; Wang et al., 2011; Li et al., 2012). Recently, it has been demonstrated that the light-inducible R2R3MYB, MdMYB1, controlling anthocyanin accumulation in apple skin, is targeted by the E3 ligase, COP1, for proteasomal degradation (Li et al., 2012). The jasmonate ZIM-domain (JAZ) repressor proteins interact with bHLH proteins (GL3, EGL3 and TT8) and R2R3-MYBs (PAP1 and GL1) to repress JA-induced anthocyanin accumulation and trichome formation. Following exogenous JA treatment or triggered endogenous JA biosynthesis, the JA signal induces degradation of JAZ proteins through the 26S proteasome pathway in a COI1-dependent manner. bHLH-MYB-WD40 proteins are consequently released to form an active complex that induces the expression of structural genes (Qi et al., 2011). Our work extends the understanding of the roles of UPS in bHLH-MYB-WD40-regulated transcription. GL3 and EGL3 proteins are short-lived, and can be significantly stabilized by chemical inhibition of the proteasome system (Figures 1 and 2). The proteolytic activity of UPS appears to control the levels of GL3, EGL3 and possibly their co-regulators. The emerging picture unveils a sophisticated relationship between UPS and gene regulation of trichome development and anthocyanin biosynthesis. The formation and activation of the bHLH-MYB-WD40 transcriptional complex is mediated by a number of biotic, and abiotic signals, including JA-induced degradation of JAZ proteins (Qi et al., 2011). As the cellular level of bHLH-MYB-WD40 complexes increases, however, the bHLH proteins, i.e. GL3 and EGL3, induce UPS-mediated degradation (this work), thereby maintaining a delicate balance of transcriptional activity. This balance may be necessary to prevent the hyperactivation of bHLH-MYB-WD40-mediated transcription that would result in severe trichome and anthocyanin phenotypes.
The lysine-dependent degradation of TFs via UPS has been demonstrated in great detail in animals (Batonnet et al., 2004; Vosper et al., 2009). Whereas a lysine residue (K133) located in the bHLH domain of the myogenic regulator, MyoD, is critical for its degradation (Batonnet et al., 2004), the proteolysis of the proneural bHLH factor, neurogenin (NGN), depends on both canonical (lysine) and non-canonical (cysteine) residues in the protein (Vosper et al., 2009). Lysine-dependent UPS-mediated degradation also appears to govern the post-translational modification of GL3 and EGL3. In GL3 and EGL3, the mutation of a pair of lysine residues (K535K536 in GL3 and K493K495 in EGL3), located between the bHLH domain and the C-terminal ACT domain, significantly delayed the degradation of these proteins (Figure 3a,b). Moreover, the mutant proteins are more active than wild-type proteins (Figure 3c), most likely as a result of the increased stability.
Wild-type Arabidopsis have triradiate trichomes and nuclear DNA content is restricted to 32C, whereas trichome cells of the upl3-2 mutant, often with four or five branches, undergo an extra round of endoreplication (64C) (Downes et al., 2003). These observations indicate that, as a negative regulator of trichome branching and endoreplication, UPL3 targets factors that promote these processes for proteasomal degradation. Evidence presented here strongly supports the view that UPL3 mediates the proteasomal degradation of GL3 and EGL3: first, the spatial expression pattern of UPL3 is similar to that of its putative targets, GL3 and EGL3 (Zhao et al., 2008) (Figure 4a); second, the recombinant GL3 and EGL3 proteins are significantly more stable in upl3-2 plant extracts compared with the wild type (Figure 4b,c); and finally, ectopic expression of GL3 or EGL3 in upl3-2 mutant plants results in aberrant trichomes and anthocyanin accumulation, most likely as a consequence of the higher accumulation of the two proteins (Figure 4d–g).
The physical interaction of an E3 ligase with its target protein is a prerequisite for efficient degradation by UPS. The N-terminal region of UPL3 contains a cluster of four ARM repeats that share sequence homology with the ARM region of Arabidopsis importin, a member of the nuclear pore complex responsible for importing proteins into the nucleus (Downes et al., 2003). An alpha solenoid structure formed by multiple ARM repeats is known to interact with various protein partners. We have demonstrated here that the N-terminal ARM repeat domain of UPL3 interacts with the C-terminal region of GL3 or EGL3 that contains the bHLH and ACT domains (Figure 5 and 6). The GL3/EGL3 C-terminal regions are known to dimerize (Payne et al., 2000; Zhang et al., 2003). We have further determined that dimeric conformation of the bHLH-ACT domain is required for UPL3 interaction. The bHLH- and ACT-mediated dimerization of subgroup-IIIf bHLH TFs promotes the activation of different subsets of genes in the anthocyanin biosynthetic pathway (Kong et al., 2012). Our findings suggest that, in this particular conformation, the bHLH TFs are subjected to UPL3-mediated regulation.
E3 ligases play a critical role in the post-translational regulation of many regulatory proteins and enzymes; however, little is known about their own transcriptional regulation. GL3 binds to the promoters of a number of trichome and anthocyanin pathway-related genes, such as MYBL2, CPC, ETC1 and GL2, as well as its own promoter, consequently influencing the expression of these genes (Morohashi et al., 2007). Recently, chromatin immunoprecipitation (ChIP) has shown that GL3 binds the UPL3 promoter, although the significance of this binding has not been determined (Morohashi and Grotewold, 2009). Our work attests to the importance of GL3 binding to the UPL3 promoter. Mutation in the gl3 locus negatively effects UPL3 expression, whereas overexpression of GL3 upregulates it (Figure 7). These findings suggest the presence of a regulatory loop involving GL3 and UPL3; however, GL3 does not appear to be the sole transcriptional regulator of UPL3. The detection of UPL3 transcripts in gl3-4, egl3-2 and gl3 egl3 mutants suggests a redundancy of factors involved in UPL3 regulation. Considering the wide range of functions for UPL3 in proteolysis, the involvement of other factors in its regulation is not surprising.
Based on our results we propose a working model for the post-translational control of TFs regulating trichome development and anthocyanin biosynthesis (Figure 8). In Arabidopsis, the R2R3MYB-bHLH-WD40 activator complex triggers the downstream transcriptional cascade that culminates in trichome development and anthocyanin accumulation. In a dynamic cellular environment, GL3, EGL3 and other regulators in the complex are under multiple levels of control. Of these multilayered and interconnected control mechanisms, post-translational modification is a particularly dynamic and reversible system for fine-tuning the amplitude and timing of gene expression. As the cellular GL3 concentration increases, it binds to the UPL3 promoter, possibly with other co-factors, and upregulates UPL3 expression. An, as yet, unidentified signal initiates the binding of the E2 ligase to the HECT domain of UPL3, and transfers the ubiquitin. UPL3, through its N-terminal armadillo repeat domain, interacts with the C termini of GL3/EGL3. Ubiquitin is transferred to internal lysine residues, and the ubiquitinated GL3 and EGL3 proteins are targeted by UPS for degradation. Whether other TFs in the complex are also targets of UPL3 or another E3 ligase remains undetermined. Proteasomal degradation of the light-regulated apple MYB, MdMYB1, by COP1 suggests that additional factors are very likely to be involved in the post-translational regulation of these TFs. Experiments are under way to explore the post-translational regulation of other regulators controlling Arabidopsis trichome development and anthocyanin biosynthesis.
Plant materials and growth conditions
All mutant and transgenic lines were in the Arabidopsis thaliana accession Col-0. Seeds were sterilized and plated on half-strength MS medium (Caisson Laboratories, http://www.caissonlabs.com). Plants were grown in a controlled environment chamber (16 h of light at 22–24°C and 8 h of dark at 17–19°C). Arabidopsis T-DNA insertion mutants and their identification numbers are described in Appendix S1.
To generate GL3 and EGL3 overexpression lines, full-length cDNAs were cloned into a modified pCAMBIA1300 vector containing the CaMV 35S promoter and rbcS terminator. Three tandem repeats encoding the FLAG epitope (5′-GACTACAAAGACGATGACGACAAG-3′ and 5′- GACTACAAAGACGATGACGACAAG-3) were fused in-frame to the C-terminal end of GL3 and EGL3 cDNA (Figure S9). The resulting constructs, p35Spro:GL3-FLAG and p35Spro:EGL3-FLAG, were introduced into Agrobacterium tumefaciens strain GV3850. The gl3-4 and egl3-2 plants were transformed using the floral-dip method (Clough and Bent, 1998).
Homozygous lines overexpressing GL3-FLAGox in gl3-4, and EGL3-FLAGox in egl3-2 backgrounds, respectively, were generated by crossing corresponding parental homozygous lines. Putative double homozygous lines were selected based on trichome phenotype, immunoblotting analysis and genotyping with the primers listed in Table S1. Homozygous GL3-FLAGox and EGL3-FLAGox lines in double homozygous upl3-2 gl3-4 or upl3-2 egl3-2 backgrounds were selected by genotyping F2-segregating progeny.
To generate UPL3pro:uidA transgenic plants, a 2.1-kb fragment upstream of the UPL3 coding region was cloned into a modified pKYLX71 vector containing the rbcS terminator and the β-glucuronidase coding sequence (uidA) interrupted by an intron. Col-0 plants were transformed using the floral-dip method. GUS histochemical staining was performed as previously described (Jefferson et al., 1987).
Protein purification and cell-free degradation assays
To purify recombinant GL3 and EGL3 proteins, GL3-FLAG and EGL3-FLAG were PCR amplified from the plant expression vectors and cloned into the pGEX4T1 vector (GE Healthcare Biosciences, http://www.gelifesciences.com). The resulting pGEX4T1-GL3 and pGEX4T1-EGL3 plasmids were used as templates for site-directed mutagenesis to identify ubiquitination sites. Mutagenesis was carried out as described by Zheng et al. (2004), and mutations were verified by sequencing using a Beckman Coulter Sequencer CEQ-8000 (Beckman Coulter, https://www.beckmancoulter.com). Wild-type and mutant constructs were transformed into BL21 cells containing pRIL (Stratagene, now Agilent, http://www.genomics.agilent.com) and the recombinant proteins were purified as described in Appendix S1. Cell-free extracts were essentially prepared as previously described by Wang et al. (2009); (Appendix S1).
Cycloheximide and MG132 treatments
For protein stability and proteasome inhibition assays, 3-week-old seedlings were incubated with either 200 μm CHX alone, or 200 μm CHX and 200 μm MG132. CHX and MG132 stocks were prepared in water and DMSO, respectively, and all control experiments contained equal quantities of solvent.
Total protein extracts were prepared as described by Kurepa and Smalle (2011), separated on SDS-PAGE (12% acrylamide for GL3 and 10% for EGL3) and transferred to nitrocellulose membranes (Bio-Rad, http://www.bio-rad.com). Membranes were stained with Ponceau S to ensure equal loading and quality of transfer, blocked with 3% BSA, probed with anti-FLAG antibody (clone M2; Sigma-Aldrich, http://www.sigmaaldrich.com), followed by HRP-conjugated secondary antibody (Thermo Scientific, http://www.thermoscientific.com). Signal was detected using Super Signal West Pico Chemiluminescent Substrate (Thermo Scientific). Relative protein quantities were determined densitometrically using ImageJ (http://rsb.info.nih.gov/ij). Anti α-tubulin antibody (SantaCruz Biotechnology Inc., http://www.scbt.com) was used for the detection of α-tubulin.
Yeast two-hybrid assay
The N-terminal region of UPL3 (UPL31–470) was cloned into pAD-GAL4-2.1 (Stratagene). Different N- and C-terminal fragments of GL3 (GL31–209, GL3400–495, GL3400–554, GL3552–637 and GL3400–637) and EGL3 (EGL31–205 and EGL3402–596) were cloned in pBD-GAL4 Cam (Stratagene). C-terminal fragments of GL3 (GL3400–495, GL3400–554, GL3552–637 and GL3400–637) were cloned in pAD-GAL4-2.1 (Stratagene). To determine the interaction, individual pBD clones were cotransformed with pAD-UPL31-470 into Saccharomyces cerevisiae Y2H Gold strain (Clontech, http://www.clontech.com). To determine homodimerization, different domains of GL3, fused to AD domains (GL3400–495, GL3400–551, GL3552–637 and GL3400–637) and BD (GL3400–495, GL3400–551, GL3552–637 and GL3400–637) were co-transformed into Y2H Gold strain. Transformed colonies were selected on synthetic dropout (SD) medium lacking leucine and tryptophan (–Leu, –Trp). Colonies were then screened for growth on SD medium lacking adenine, histidine, leucine and tryptophan (–Ade, –His, –Leu, –Trp).
Transient protoplast assay
The construction of reporter and effector plasmids for protoplast assay is described in Appendix S1. Isolation of protoplasts from Arabidopsis T87 cell suspension culture (ABRC stock number CCL84839), electroporation of reporter and effector plasmids into protoplasts, and measurement of luciferase and GUS activities were performed using the protocols developed for tobacco protoplasts (Pattanaik et al., 2010b). All experiments were repeated three times with four replicates for each sample.
For the BiFC assays, the expression vectors pA7-NYP and pA7-CYP that contain either the N- or C-terminal halves of the yellow fluorescent protein (YFP) under the CaMV 35S promoter and nopaline synthase (nos) terminator were used. Full-length GL3 and EGL3 cDNAs were cloned into pA7-NYFP to create 35Spro:GL3-NYFP and 35Spro:EGL3-NYFP, respectively. UPL31–470 was cloned into the pA7-CYP vector to create 35Spro:UPL31-470-CYFP (Figure S9).
Different combinations of constructs were co-electroporated into Arabidopsis protoplasts and incubated in the dark at 22–24°C. After 14–16 h, cells were stained with 4′,6-diamidino-2-phenylindole (DAPI), according to the manufacturer's protocol (Invitrogen, http://www.invitrogen.com) and visualized under an inverted microscope (Eclipse TE200; Nikon, http://www.nikon.com), as described previously (Ohad et al., 2007; Pattanaik et al., 2011).
Total RNA was isolated from 3-week-old seedlings, and used for synthesis of first-strand cDNA and quantitative RT-PCR (qPCR) as described by Pattanaik et al. (2010a). All PCR reactions were performed in triplicate and repeated two times. The comparative cycle threshold (Ct) method (bulletin no. 2; Applied Biosystems, http://www.appliedbiosystems.com) was used to measure transcript levels. ACT2 (At3g 18780) and TUB3 (At5g 62700) were used as reference genes (Huang et al., 2007).
Morphometric analyses, anthocyanin quantification and statistical analyses
Leaves and seedlings were photographed using an Olympus SZX12 microscope (Olympus, http://www.olympus-global.com). Anthocyanins were extracted using the acid–methanol method, and quantified using the equation A530 – A650 × 0.25 (Rabino and Mancinelli, 1986). At least five seedlings per line were used to determine the average anthocyanin levels. To determine the significance of difference between means, data were analyzed by Student's t-test.
We are grateful for the intellectual and technical guidance from Drs Jan Smalle and Jasmina Kurepa of the University of Kentucky. We thank Dr Erich Grotewold of The Ohio State University, Columbus, OH, for providing us with gl3 egl3 double mutant seeds. We appreciate Kathy Shen for her helpful suggestions and critical reading of the manuscript. This work is supported by a grant from the Kentucky Tobacco Research and Development Center, University of Kentucky.