CUL4 associates with DDB1 and DET1 and its downregulation affects diverse aspects of development in Arabidopsis thaliana

Authors


*(fax +49 30 838 54345; e-mail hellmann@zedat.fu-berlin.de).

Summary

Cullins are central scaffolding subunits in eukaryotic E3 ligases that facilitate the ubiquitination of target proteins. Arabidopsis contains at least 11 cullin proteins but only a few of them have been assigned biological roles. In this work Arabidopsis cullin 4 is shown to assemble with DDB1, RBX1, DET1 and DDB2 in vitro and in planta. In addition, by using T-DNA insertion and CUL4 antisense lines we demonstrate that corresponding mutants are severely affected in different aspects of development. Reduced CUL4 expression leads to a reduced number of lateral roots, and to abnormal vascular tissue and stomatal development. Furthermore, cul4 mutants display a weak constitutive photomorphogenic phenotype. These results therefore assign an important function to CUL4 during plant development and provide strong evidence that CUL4 assembles together with RBX1 and DDB1 proteins to form a functional E3 ligase in Arabidopsis.

Introduction

E3 ubiquitin ligases are key regulatory components of the ubiquitin proteasome pathway. They specifically target substrate proteins for degradation via the 26S proteasome, and it is now generally accepted that they control most aspects of plant development (for reviews see Schwechheimer and Villalobos, 2004; Smalle and Vierstra, 2004). Many E3 ligases are multimeric protein complexes that contain cullin proteins as scaffolding subunits. The Arabidopsis thaliana genome encodes six cullin proteins which share general sequence similarity within a region of around 200 amino acids, denoted the cullin homology (CH) domain (Shen et al., 2002; Wirbelauer et al., 2000).

Of these six cullins, CUL1 (At4g02570) and CUL2 (At1g43140) are part of an SCF (SKP1-Cullin-F-box) type E3 ligase (Risseeuw et al., 2003). This class of E3 ligases consists of the cullin, an ASK (Arabidopsis Skp1-like) protein, a RING-finger protein RBX1 and a substrate adaptor with an F-box domain (Gray et al., 2001; Risseeuw et al., 2003; Shen et al., 2002). Although the function of CUL2 remains unclear in plants, CUL1 clearly participates in the regulation of various pathways, such as phytohormone signal transduction, flowering, senescence and embryogenesis (Durfee et al., 2003; Gray et al., 2002; Guo and Ecker, 2003; Hellmann et al., 2003; Potuschak et al., 2003; Shen et al., 2002; Woo et al., 2001). Like CUL1, the cullins 3a (At1g26830) and 3b (At1g69670) also associate with RBX1 but use BTB/POZ domain proteins as substrate adaptors (Dieterle et al., 2005; Figueroa et al., 2005; Gingerich et al., 2005; Thomann et al., 2005; Weber et al., 2005). CUL3 proteins are involved in ethylene biosynthesis, embryogenesis and phytochrome A signal transduction (Dieterle et al., 2005; Figueroa et al., 2005; Gingerich et al., 2005; Wang et al., 2004). The fifth cullin, APC2 (At3g46910), is part of an anaphase promoting complex/cyclosome (APC/C) which is most probably involved in mitotic processes (Capron et al., 2003).

By contrast, almost nothing is known about the sixth and last CH-domain-containing cullin from Arabidopsis, which is called CUL4 (At5g46210). The closest family members in Arabidopsis are CUL3a and -3b, and like these cullins it interacts with RBX1 in yeast two-hybrid (Y2H) assays and apparently in planta with the COP9 signalosome (Dohmann et al., 2005; Lechner et al., 2002; Shen et al., 2002). Its biological role and association with other proteins, however, remains elusive. Currently, knowledge of how Arabidopsis CUL4 might function in plants has been inferred mainly from work on the human ortholog hCUL4a, which forms functional E3 ligase complexes with hRBX1 and hDDB1 (damaged DNA-binding 1; Chen et al., 2001; Hu et al., 2004; Matsuda et al., 2005; Wertz et al., 2004). hDDB1 represents the substrate adaptor unit, like BTB/POZ or F-box proteins in other cullin-based E3 ligases. Interaction between hCUL4a and hDDB1 requires the second and fifth α-helices of hCUL4a and the central region of hDDB1 (Hu et al., 2004). hCUL4a also assembles with hCAND1, which is a negative regulator of cullin-dependent E3 ligases (Chuang et al., 2004; Hu et al., 2004). The stability of a variety of proteins such as hDDB2, hSTAT1, hSTAT3, hHOXA9, hCDT1 and hc-Jun is linked to hCUL4a (Andrejeva et al., 2002; Higa et al., 2003; Matsuda et al., 2005; Ulane and Horvath, 2002; Ulane et al., 2003; Wertz et al., 2004; Zhang et al., 2003; Zhong et al., 2003). Since these proteins are involved in transcriptional processes, cell cycle control and the integrity of DNA, it is not surprising that loss of this cullin broadly affects development and chromosomal architecture: for example, it leads to embryonic defects in mice and Caenorhabditis elegans, and to decondensed chromosomes in the fission yeast Schizosaccharomyces pombe (Li et al., 2002; Osaka et al., 2000; Zhong et al., 2003).

hDDB1 was first identified in studies of patients with xeroderma pigmentosum (XP), who are hypersensitive to UV irradiation because of defects in DNA repair (Chu and Chang, 1988; Feldberg, 1980). The protein assembles with hDDB2 to form a heterodimer known as the DDB complex that binds to damaged DNA and is involved in excision repair of damaged nucleotides (for an overview see Wittschieben and Wood, 2003). In response to UV light, hDDB1 acts together with hCUL4a and hRBX1 as a substrate adaptor to target hCDT1 and hDDB2 proteins for ubiquitination (Hu et al., 2004; Matsuda et al., 2005).

In addition to binding hDDB2, hDDB1 also interacts with hDET1 (de-etiolated 1) and hCOP1 (constitutive photomorphogenesis 1) to established a DCXDET1-COP1 (DDB1/CUL4a/X-box) E3 ligase that controls stability of the transcription factor hc-Jun (Wertz et al., 2004). The X-box denotes an as yet unidentified motif that allows interaction between DDB1 and the heterodimer hDET1/hCOP1 (Wertz et al., 2004). Interestingly, all subunits of DCXDET1-COP1 are highly conserved and corresponding orthologs are present in plants (Ishibashi et al., 2003; Schroeder et al., 2002; Schwechheimer et al., 2001; Shen et al., 2002). For example, Arabidopsis has two highly related DDB1 proteins, named DDB1a and DDB1b (Schroeder et al., 2002). Analysis of T-DNA insertion mutants showed that DDB1a null mutants do not have an apparent mutant phenotype, whereas loss of DDB1b is embryo lethal (Schroeder et al., 2002). In plants, DDB1a associates with the light-signaling intermediate DET1 and the ubiquitin-conjugating variant COP10 (Schroeder et al., 2002; Yanagawa et al., 2004). The precise functions of COP10 and DET1 are still unclear. COP10 acts a repressor of photomorphogenesis and has been proposed to destabilize positive regulators of light signaling such as the bZIP transcription factor HY5, in concert with the COP9 signalosome and COP1 (Osterlund et al., 2000; Yanagawa et al., 2004). Phenotypical similarities between cop1 and det1 mutants in turn suggest that both proteins are active in the same or overlapping pathways (Ma et al., 2003). There is evidence that DET1 is involved in protein degradation of the LHY (late elongated hypocotyl) protein, an important component of the circadian oscillator (Song and Carre, 2005). In addition, work on the tomato ortholog of DET1 (tDET1) showed that the protein interacts with the non-acetylated tail of the core histone H2B, and it has thereby been proposed to modulate chromatin structure and to affect light-dependent transcriptional activities (Benvenuto et al., 2002). In this context it is noteworthy that SpCul4 from the fission yeast S. pombe and human hDDB1 are also involved in heterochromatin assembly: hDDB1 interacts with histone acetyl transferase (HAT) complexes (Martinez et al., 2001), and SpCul4 associates with SpRik1, a DDB1-related protein, and the histone methyl transferase SpClr4 (Jia et al., 2005).

Taken together, these data imply a situation in which Arabidopsis CUL4 might undergo complex assemblies similar to hCUL4a and suggest that it may participate in important processes such as the cell cycle, light-dependent growth control, modulation of chromosomal structure and DNA repair.

Here we show that Arabidopsis cullin 4 indeed assembles with DDB1 proteins, RBX1 DET1, and DDB2, and we provide evidence that DDB1 acts as the bridging protein between the cullin and DET1 in planta. By the use of cul4 mutants with strongly reduced levels of CUL4 protein we demonstrate that the cullin influences a broad variety of developmental processes. Most obviously cul4 mutants display developmental defects in leaf and lateral root development. The mutants are also affected in the growth of the vascular tissue, and dark-grown seedlings display weak de-etiolated morphologies. The results presented here provide novel information on the assembly of Arabidopsis DDB1 proteins with CUL4 and reveal the importance of this cullin during plant development.

Results

Arabidopsis cullin 4 is a protein of 793 amino acids with a predicted molecular weight of 91.5 kDa. CUL4 contains a conserved cullin or CH-domain from amino acids 466–615 (http://elm.eu.org/; Puntervoll et al., 2003; Figure 1a). The protein has an extended N-terminal region of around 65 amino acids that is unique to this particular cullin in Arabidopsis (Figure 1b). The most related sequences in non-photosynthetic organisms are the human cullins 4a and 4b (Shen et al., 2002). AtCUL4 in fact appears to be more homologous to human hCUL4a than to its closest Arabidopsis family members AtCUL3a and AtCUL3b (Figure 1c).

Figure 1.

 Structure of Arabidopsis CUL4.
(a) Schematic drawing of full-length CUL4 with the cullin homology (CH) domain.
(b) Partial alignment of N-terminal regions of Arabidopsis cullins CUL1 (164 residues), CUL3a (111 residues), CUL3b (111 residues) and CUL4 (240 residues) using the clustalw program (Thompson et al., 1994); black-shaded residues are identical, gray-shaded residues are conserved.
(c) Sequence identities (dark gray-shaded fields) and similarities (light gray-shaded fields) between Arabidopsis (At) cullins and the human (h) CUL4a using the matgat program (Campanella et al., 2003). Blosum50; First Gap 12; Extending Gap 2.

Arabidopsis CUL4 associates with DDB1a and RBX1

Based on the work of Wertz et al. (2004) and Hu et al. (2004) we investigated whether Arabidopsis CUL4 assembles with DDB1. The Arabidopsis genome encodes two highly related DDB1 proteins, DDB1a (At4g05420) and DDB1b (At4g21100), which are >90% identical to each other (Schroeder et al., 2002). Yeast two-hybrid studies showed that the full-length CUL4 protein could interact with DDB1a and that this interaction is specific, in as much as no assembly was seen with CUL3a (Figure 2a,b).

Figure 2.

 CUL4 interacts with DDB1a, DDB1b and RBX1.
(a) Left panel: Yeast two-hybrid analysis shows interaction of full-length and partial CUL4 with full-length DDB1a and RBX1 proteins. Right panel: schematic drawing of the different constructs tested.
(b) Yeast two-hybrid analysis of full-length and partial DDB1a with CUL4 (left panel) and the different constructs tested (right panel). SDII, selection medium for transformation with bait (pBTM116-D9) and prey (pACT2) plasmids supplemented with leucine and histidine; SDIV, selection medium for interaction studies without leucine and histidine supplements. Photos were taken from single spots.
(c–h) Interaction between CUL4 and DDB1a and a DDB1b300−694 protein was confirmed by pulldown assays.
(c–f) DDB1a and CUL41−453 interaction assays with bacterially expressed and purified GST, GST:CUL41−453 and GST:DDB1a300−666 proteins. Input represents 1 μl of [35S]-methionine-labeled protein used for pulldown assays.
(f) In vitro translated CUL4 co-precipitates with GST:DDB1b300−694 but not with GST.
(g) CUL4 co-precipitates from plant extracts with GST:DDB1a300−666 but not with GST.
(h) Overexpressed myc-tagged DDB1a protein is specifically detectable in transgenic plants but not in Col0 wild type.
(i) myc-tagged DDB1a co-precipitates with GST:CUL41−453 from plant extracts.

To map the domains required for CUL4–DDB1a interaction, different Y2H expression constructs for partial CUL4 and DDB1a proteins were generated (Figure 2a,b). As shown in Figure 2(a) the first 453 amino acids of CUL4 were sufficient to allow CUL4–DDB1a assembly. By contrast, deletion of the first 304 amino acids of the cullin (CUL4304−793) completely abolished interaction with DDB1a. This loss of interaction was not based on protein misexpression since RBX1 was still able to assemble with the C-terminal of the partial CUL4 protein (Figure 2a). These findings demonstrate that the N-terminal portion of CUL4 is required for association with DDB1a and that the conserved cullin domain is not required for this interaction. To determine the domains of DDB1a required for interaction with CUL4, several truncated versions of DDB1a were generated. Deletion of the first 298 amino acids (DDB1a299−1089) did not affect interaction with CUL4 and a DDB1a protein fragment containing only amino acids 300–666 (DDB1a300−666) was sufficient for interaction with full-length CUL4 (Figure 2b).

Interaction between CUL4 and DDB1a was further confirmed by in vitro pulldown assays. In these experiments, the fusion proteins glutathionine-S-transferase (GST):CUL41−453 and GST:DDB1a300−666 were incubated with in vitro translated and l-[35S]-methionine-labeled CUL41−453, DDB1a300−666, or full-length DDB1a. As shown in Figure 2(c) and (d), full-length and partial DDB1a proteins co-precipitated with GST:CUL41−453. Interestingly, the partial proteins of GST:DDB1a300−666 and in vitro translated CUL41−453 were sufficient to permit interaction between the proteins (Figure 2e). To investigate whether DDB1b can also interact with CUL4, a DDB1b fragment stretching from amino acids 300–694 (DDB1b300−694) was cloned and used for a pulldown assay since this region was sufficient for DDB1a–CUL4 association. As shown in Figure 2(f), GST:DDB1b300−694 did interact with a in vitro translated full-length CUL4. This result provides evidence that both DDB1 proteins can be in a complex with the cullin in planta. The GST:DDB1a300−666 fusion protein was also used to successfully pull down CUL4 from plant extracts (Figure 2g). Detection of precipitated CUL4 was performed using a polyclonal antibody raised against a CUL4-specific N-terminal peptide. To investigate the interaction of DDB1a–CUL4 with in planta expressed DDB1a we generated transgenic Arabidopsis plants that expressed DDB1a under the control of a 35S promoter (here referred to as 35S::myc-DDB1a; Figure 2h). The 35S::myc-DDB1a plants grew like wild-type plants, indicating that ectopic expression of DDB1a has no major impact on Arabidopsis development (data not shown). Using GST:CUL41−453 as bait, myc-tagged DDB1a was successfully pulled down from plant extracts (Figure 2i). Collectively, these results are in agreement with the findings from Hu et al. (2004) showing interaction between CUL4 and DDB1 in mammalian cells, and demonstrate that in Arabidopsis the N-terminal region of CUL4 interacts with the central domain of DDB1a.

DET1 and DDB2 interact with DDB1a but not with CUL4

It has been shown that Arabidopsis DDB1a directly interacts with DET1 (Schroeder et al., 2002). It was therefore of interest to test whether DET1 can also associate with CUL4. Using Arabidopsis DET1 as prey and either DDB1a or CUL4 as bait in Y2H assays we could detect interaction of DET1 with DDB1a but not with CUL4 (Figure 3a). Additionally, we tested DDB1a300−666 for interaction with DET1 because this partial DDB1a protein was sufficient to interact with CUL4. However, DET1 did not interact with DDB1a300−666, indicating that DET1 and CUL4 interact with different regions of DDB1a (Figure 3a). The Arabidopsis genome encodes for one protein, At5g58760/DDB2, which is 22.8% identical and 40.4% similar to human hDDB2. Since hDDB2 is one of the first described hDDB1-interacting proteins, it was of interest whether the Arabidopsis ortholog can also interact with DDB1a. As shown in Figure 3(b), DDB2 indeed interacted with DDB1a but not with CUL4, indicating that DDB1a undergoes comparable complex associations as described for its human counterpart.

Figure 3.

 DDB1a associates with DET1 and DDB2.
(a) Yeast two-hybrid analysis shows assembly of Arabidopsis DET1 with full-length DDB1a but not with CUL4 or with partial DDB1a300−666 proteins.
(b) Likewise, DDB2 interacts in Y2H assays with DDB1a.
(c) Specific detection of overexpressed myc-tagged tDET1 in transgenic 35S::myc-tDET1 plants.
(d) myc-tDET1 co-precipitates with GST:CUL41−453 and GST:RBX1.
(e) Immunoprecipitation assays using a CUL4-specific antibody demonstrate co-precipitation of myc-tDET1 and CUL4. Plant extracts represent 40 μg of total protein extract from 35S::myc-tDET1 plants.

CUL4 associates with DET1 in planta

To investigate whether CUL4 and DET1 are present in the same complex in planta we first tried to express myc-tagged Arabidopsis DET1 under the control of a 35S promoter. However, we were not able to generate transgenic plants reproducibly expressing myc-DET1, most likely because of silencing problems (data not shown). To overcome this problem we used the DET1 gene from tomato (tDET1), which is 81.3% identical at the amino acid level to Arabidopsis DET1 (Mustilli et al., 1999). In this case stable Arabidopsis lines were established that expressed tDET1 fused to a myc epitope under the control of a 35S promoter (35S::myc-tDET1; Figure 3b). These plants did not display any obvious phenotypes (data not shown). Because interaction of DET1 and DDB1a has already been demonstrated in planta (Schroeder et al., 2002), we focused on whether CUL4 and RBX1 can associate with DET1 in plant extracts. For this approach a GST:CUL41−453 fusion protein was used that contains the DDB1a-interacting part of CUL4 as well as a full-length GST:RBX1 fusion protein. As shown in Figure 3(c) both GST fusion proteins could co-precipitate myc-tDET1 from Arabidopsis extracts, whereas the GST control alone could not. In addition, in planta association of CUL4 with myc-tDET1 was demonstrated by immunoprecipitation (IP) of CUL4 and the co-IP of myc-tDET1 (Figure 3d). In summary, these results demonstrate that CUL4 associates with DET1 in planta and that assembly of the two proteins is likely to be bridged by DDB1a.

CUL4 is expressed constitutively

To determine CUL4 expression patterns in Arabidopsis, total RNA was isolated from flowers, stems, rosette leaves and roots for Northern blot analysis. Hybridization with a CUL4-specific probe showed that the mRNA was present in all tissues examined (Figure 4a). In addition, CUL4 mRNA showed a slight light–dark regulation of expression. For this approach, RNA was isolated from leaves of 3-week-old plants cultured in a 12-h day:night cycle. Samples were taken every 4 h over one 24-h cycle. Hybridization with a CUL4-specific probe revealed a slight upregulation of CUL4 expression within the first 4 h of the night cycle (Figure 4c). For more detailed information on tissue-specific expression patterns, a 1.64 kb promoter was cloned in front of a GUS reporter gene (pCUL4::GUS) and used for plant transformation. Analysis of transgenic pCUL4::GUS plants showed GUS expression mainly in young leaves and in the central and major veins of the vascular tissue (Figure 4c,d). In roots, expression was only detectable in the older parts of the primary and lateral roots (Figure 4e,f). We could never detect any staining in emerging lateral root buds. Finally, in flowers a strong expression was detectable mainly in pollen which was confirmed by in situ hybridization (Figure 4g–i).

Figure 4.

 RNA and promoter-GUS expression analysis of CUL4 in different tissues.
(a) CUL4 is expressed in roots, leaves, shoot and flowers.
(b) Analysis of CUL4 expression in a 12-h day–night cycle.
(c–g) A CUL4 promoter drives GUS expression in shoot and central leaf veins (c, d), primary roots (e), root tip (f) and pollen (g).
Pollen-specific expression was confirmed by in situ hybridization with (h) showing results of the antisense and (i) with the sense probe on flower tissue.

Identification and generation of cul4 mutant plants

To gain information about the function of CUL4 during plant development, two T-DNA insertion lines from the SALK collection (SALK_077684 and SALK_084869) and one insertion line from GABI-Kat (600H03) were obtained and analyzed for effects on CUL4 expression levels (Alonso et al., 2003; Rosso et al., 2003). Analyses of the SALK lines showed that CUL4 expression was either not affected by the T-DNA insertion (SALK_077684; T-DNA insertion is located around 150 bp upstream of the ATG) or that we could not detect the presence of the T-DNA within the CUL4 gene (SALK_084869; data not shown).

However, the T-DNA insertion in the GABI-Kat line was confirmed by PCR (data not shown), and selfing of plants heterozygous for the T-DNA insertion showed a 3:1 segregation pattern on sulfadiazine-containing medium with 823 resistant plants and 292 sensitive ones. This segregation pattern indicated the presence of just one T-DNA insertion in this line. The insertion was found to be located within the 12th intron, and analysis of homozygous lines (denoted cul4-1) by Northern blotting and RT-PCR revealed reduced levels of CUL4 mRNA (Figure 5a,c).

Figure 5.

 Reduction of CUL4 expression by using T-DNA insertion and antisense lines.
Schematic drawings of (a) T-DNA insertion within the 12th intron of CUL4 genomic DNA and of (b) CUL4 cDNA with the first 623 bases used for antisense RNA expression.
(c) T-DNA insertion in cul4-1 strongly reduces expression of CUL4 but expression is still detectable, as demonstrated by Northern blot (upper panel) and RT-PCR (lower panel) on two cul4-1 plants.
(d, e) Overexpression of the first 623 bp of CUL4 strongly reduces endogenous expression of CUL4. The antisense effect is specific to CUL4 since (d) CUL1 and CUL3b are not affected, and neither are (e) DDB1a, DET1 and COP1. Asterisks mark full-length CUL4 (upper asterisk) and antisense (lower asterisk).

Additional cul4 knockdown mutants were generated using a CUL4 antisense construct (denoted ascul4). Specifically, the first 623 bp of CUL4 after the ATG were used because this region is highly specific to CUL4 and so should allow its targeted reduction without affecting expression of the other cullins (Figure 5b). Two hundred transgenic plants were generated, of which around 20% showed a strongly reduced growth. Of these, two representative lines, named ascul4-100 and ascul4-102, were chosen for further detailed analysis. As confirmed in Figure 5d, CUL4 expression was clearly downregulated in these plants whereas CUL1 and CUL3b expression were not affected. Analysis of DDB1a, COP1 and DET1 expression also did not reveal any major changes in the ascul4-100 and -102 backgrounds (Figure 5e). Furthermore, the strength of the ascul4 phenotypes strictly correlated with the level of CUL4 reduction, demonstrating the specificity and efficiency of the antisense approach (Figure 6a).

Figure 6.

 Reduced CUL4 expression leads to reduced growth and aberrant leaf development.
(a) Changes in growth in ascul4 lines are strictly correlated with reduced CUL4 expression levels.
(b) Both ascul4 and cul4-1 mutants show aberrant leaf development which is manifested by split leaves and irregular leaf shapes.
(c) Aberrant vascular tissue with unconnected strands present in the leaf of ascul4 plants (indicated by arrows). Size bars represent 1 mm.
(d) Leaves of ascul4 lines developed higher numbers of double stomata that were attached to each other (circled area in upper right picture and lower right enlargement). Code to shading in graph: black bar, Col0; dark gray bar, ascul4-100; light gray bar; ascul4-102. Size bars represent 10 μm.

CUL4 is involved in leaf and root development

The results from expression analyses suggested that CUL4 might participate in different events during development. Indeed, phenotypic analysis of ascul4 and cul4-1 plants showed that they regularly developed misshapen cotyledons (data not shown) and leaves (Figure 6b). With respect to wild-type leaves, two main differences were observed: first, the central vein was often split into single independent vascular strands running next to each other, and second, the leaves often developed isolated vascular strands (Figure 6c). In addition, differentiated leaves of ascul4 lines also showed abnormal stomatal development with a high frequency of stomata touching each other (Figure 6d). It is noteworthy that in the cul4-1 mutants the aberrant leaf development phenotypes were only observed in cotyledons and first leaves, and with increasing age they grew comparably with wild-type Col0 plants (Figure 7a). This strongly contrasted with ascul4-100 and ascul4-102 plants, which remained dwarf-like throughout the life of the plant (Figure 7a). In addition, at the onset of bolting the two ascul4 lines also had fewer rosette leaves in comparison with wild-type plants (Figure 7b) and this was not observed in cul4-1 mutant plants (data not shown).

Figure 7.

 CUL4 is required for at several stages of development and affects photomorphogenesis.
(a) ascul4 lines remain small throughout the life cycle whereas cul4-1 mutants develop like wild-type plants with the onset of flowering.
(b) ascul4 lines developed fewer leaves at the time of bolting.
(c) Western blot analysis of Col0, ascul4-102 and cul4-1 revealed that CUL4 protein levels are equally affected in ascul4-102 and cul4-1 seedlings, whereas at the rosette stage cul4-1 clearly contains more CUL4 protein than ascul4-102. CDC2 was used as a loading control.
(d) cul4 mutant plants develop fewer lateral roots in comparison with wild-type Col0 plants. Images and quantifications of lateral roots were taken 10 days post-germination. Scale bar represents 1 cm.
(e) Six-day-old, dark grown cul4 mutant seedlings show a weak constitutive photomorphogenic phenotype, most clearly evident in the opening of cotyledons. Bar coloring in (d) is consistent for (b) and (e).

To understand better these phenotypic differences between CUL4 antisense and T-DNA insertion mutants, CUL4 protein levels were examined at seedling and rosette stages. At the seedling stage both mutant lines showed strongly reduced CUL4 protein levels (Figure 7c), but at the rosette stage cul4-1 plants clearly expressed the protein whereas it was barely detectable in the ascul4 plants (Figure 7c). Consequently, the observed phenotypical dissimilarities between cul4-1 and ascul4 are most probably due to different the CUL4 content during different stages of development in the mutants.

Besides the aberrant leaf growth, all cul4 mutants showed subtle changes in root development. Although primary root growth was unaffected, the number of lateral roots was clearly reduced (Figure 7d). Here, ascul4 lines and cul4-1 developed fewer lateral roots and these were shorter in comparison with the roots of wild-type plants (Figure 7d). In addition, the first lateral roots appeared at a greater distance from the hypocotyl in comparison with what is found in wild-type plants. One cause of aberrant root development could be related to changed sensitivities towards the phytohormone auxin. However, the roots of ascul4 mutants showed normal responses towards the synthetic auxin 2,4-dichlorophenoxy acetic acid (2,4-D) in root elongation assays (data not shown).

CUL4 participates in photomorphogenesis

Because CUL4 associates with DET1, we speculated that cul4 mutants might display defects in photomorphogenesis, as has been described for cop1 and det1 mutants (Ang and Deng, 1994; Pepper and Chory, 1997). To identify defects in the photomorphogenic response, cul4-1 and ascul4-102 seedlings were cultured for 6 days in complete darkness before examining changes in the apical hook, hypocotyl length and cotyledon development. We could not detect any major changes in the formation of the apical hook. However, both ascul4-102 and cul4-1 seedlings had a tendency to develop slightly shorter hypocotyls in comparison with Col0 plants (Figure 7e). Furthermore, both cul4 mutant lines showed a high frequency of seedlings with open cotyledons, which occurred more often in the antisense lines (around 78%) in comparison with the cul4-1 mutant background (around 55%; Figure 7e). These findings suggest that CUL4 participates in some aspects of photomorphogenesis, such as cotyledon opening, but given the strong reduction in CUL4 protein content it appears probable that the cullin is not a major regulator of this developmental process.

Discussion

The data presented in this work provide a variety of new information about the function of Arabidopsis CUL4. This includes the description of Arabidopsis RBX1–CUL4–DDB1a assembly and an in planta demonstration of an association between CUL4 and DET1 via DDB1a. In addition, the described cul4 mutant phenotypes provide an overview of CUL4 function during Arabidopsis development and reveal that this cullin is important for several processes.

It is apparent that a protein–protein interaction between DDB1 and CUL4 in animals and plants is highly conserved. In both human and Arabidopsis the N-terminal region of CUL4 and a central section of DDB1 are required for assembly of the two proteins (Hu et al., 2004; Wertz et al., 2004; this work). Additionally, the N-terminal interaction of CUL4 with DDB1a in Arabidopsis is typical of previously described interactions between a cullin and its substrate adaptor protein (Figueroa et al., 2005; Pintard et al., 2003; Schulman et al., 2000). One can also expect DDB1b to interact with CUL4, since both proteins are more than 90% identical at the amino acid level (Schroeder et al., 2002). Given this high level of conservation it is reasonable to predict that both Arabidopsis DDB1 proteins will act as substrate adaptor proteins for CUL4-based E3 ligases.

It is noteworthy that at present the only known substrate adaptor protein that directly interacts with CUL4 is human DDB1. Provided that Arabidopsis DDB1a serves as a substrate adaptor protein, it is striking that cul4 mutants are strongly affected in development whereas ddb1a null mutants do not display an apparent phenotype (Schroeder et al., 2002; this work). This difference is probably linked to the fact that CUL4 is a single copy gene in humans whereas in Arabidopsis two highly related and possibly functionally redundant DDB1 proteins are present (Schroeder et al., 2002; Shen et al., 2002). Nonetheless, given the strong cul4 mutant phenotypes it will be of interest to test whether only RBX1 and DDB1 proteins interact with CUL4 in Arabidopsis. In comparison, other Arabidopsis E3 ligases that contain cullins 1 or 3 are likely to assemble with multiple substrate adaptor proteins because the Arabidopsis genome encodes around 700 F-box proteins and >80 BTB/POZ-domain-containing proteins (Dieterle et al., 2005; Gagne et al., 2002; Gingerich et al., 2005). On the other hand, DDB1a and DDB1b alone may be sufficient to bring a high degree of diversity to a CUL4-based E3 ligase, as suggested by the work of Hu et al. (2004) and Wertz et al. (2004). Here, hDDB1 can act directly as the substrate adaptor to target hCDT1 or hDDB2 for degradation but also represents a mechanism involving the higher-order E3 ubiquitin ligase DCXDET1-COP1.

Based on our findings one can hypothesize how CUL4 acts at the molecular level in plants. For example, the appearance of isolated strands of vascular tissue in ascul4 lines might indicate abnormal auxin distribution in the leaf. Although roots of the cullin mutants displayed normal auxin sensitivity, defects in vascular tissue are generally associated with changes in auxin distribution (Feugier et al., 2005). Conversely, the observed aberrant stomatal development is often connected to defects in cell-cycle control (Larkin et al., 2003), and in this context a variety of mutants with a similar stomata phenotype such as sdd1, yoda, flp or tmm have been described (Berger and Altmann, 2000; Bergmann et al., 2004; Lai et al., 2005; Nadeau and Sack, 2002). Cell cycle defects might also hold true for lateral root development (de Jager et al., 2005), which is affected in all cul4 mutants. In particular, participation of CUL4 in cell cycle control appears plausible because in humans the replication licensing factor hCDT1 is a target for a hCUL4a-hDDB1 E3 ligase (Hu et al., 2004), and related proteins are encoded in Arabidopsis (Castellano et al., 2004). Other proteins for which stability has been shown to depend on human hCUL4a are hSTAT1, hSTAT3, hHOXA9 and hDDB2 (Andrejeva et al., 2002; Matsuda et al., 2005; Ulane and Horvath, 2002; Ulane et al., 2003; Zhang et al., 2003; Zhong et al., 2003). By performing Blast searches on the Arabidopsis proteome database (http://www.Arabidopsis.org/Blast/), we could not find any hSTAT1- or hSTAT3-related proteins (unpublished data). In contrast, hHOXA9 contains a WOX-like (WUSCHEL-like homeobox; Haecker et al., 2004) homeodomain which can be found in Arabidopsis and appears to represent the ubiquitination site in hHOXA9 proteins (Zhang et al., 2003). However, Blast searches using other parts of hHOXA9 did not lead to any significant alignments with Arabidopsis proteins and it remains open whether WOX-domain-containing proteins are targets for CUL4-dependent degradation. Finally, hDDB2 is involved in the recognition of UV-damaged DNA, a critical step in nucleotide excision repair (Wittschieben and Wood, 2003). Since Arabidopsis DDB1a protein assembles with DDB2, one could speculate that CUL4 functions in DNA repair mechanisms in plants.

The demonstration of an association of CUL4 with DET1 in planta, along with the findings from Wertz et al. (2004) in humans, suggests the possibility of a DCXDET1-COP1 E3 ligase in Arabidopsis. At this point it will be interesting to test whether Arabidopsis DET1 interacts with COP1, as was demonstrated for the human orthologs. In Arabidopsis, both proteins are well known to be key players in repressing photomorphogenesis (for an overview see Hardtke and Deng, 2000). Although CUL4 assembles with DET1 in planta, it is surprising that dark-grown cul4 mutants display only subtle photomorphogenic phenotypes. Similarly, ddb1a null mutants do not show an obviously abnormal de-etiolated phenotype (Schroeder et al., 2002). However, genetic analysis demonstrated that dark-grown det1/ddb1a double mutants developed an enhanced de-etiolated phenotype in comparison with the single mutants, confirming that both proteins are active in the same or overlapping pathways (Schroeder et al., 2002). The weak phenotype of cul4 mutants could be due to the fact that the mutants were only knocked-down and still contained functional CUL4 protein. However, given the strong reduction in expression of CUL4 at the seedling stage it is possible that the putative Arabidopsis CUL4-based DCXDET-COP1 E3 ligase is not a crucial player in the repression of photomorphogenesis. Rather, it suggests a situation in which such an E3 ligase is involved only in specific aspects, such as cotyledon opening, whereas DET1 and COP1 might independently act on other developmental stages.

Finally, DET1 may play a role in heterochromatin assembly by histone acetylation (Benvenuto et al., 2002; Schroeder et al., 2002). DDB1 interacts with HAT complexes that acetylate histones, and the fission yeast SpCul4 was recently described to associate with the histone methylase Clr4 (Jia et al., 2005; Martinez et al., 2001). Taken together it will be interesting to investigate whether CUL4 acts in a DCXDET1 E3 ligase to establish chromatin structures in plants.

Experimental procedures

Plant material

Arabidopsis thaliana plants Columbia 0 ecotype (Col-0) were grown on soil in a greenhouse and growth chamber at 20°C under long day conditions (16 h light:8 h dark). For investigation of CUL4 light–dark-dependent expression, plants were cultured in growth chambers for 21 days in a 12 h light:12 h dark photoperiod. Photomorphogenic phenotype analyses were done in sterile culture with 6-day-old seedlings grown at 20°C on minimal medium in complete darkness after 2 days of vernalization at 4°C and 6 h of light exposure.

Cloning, plant vectors and transformation

The CUL4 promoter (1614 bp) was amplified from genomic Col-0 DNA with specific primers (PCUL4FW tattagtaagtttaagcgagg; PCUL4RW SpeI actagtgaaaatgggtgaaaattgtgt) and subcloned into pCR2.1 using a TOPO-TA cloning kit (Invitrogen, Karlsruhe, Germany) before subcloning into the SpeI site of the mini binary vector pCB308 (Xiang et al., 1999). Correct orientation of the promoter in front of the GUS cDNA was confirmed by sequencing. CUL4, DDB1a and DET1 cDNAs were cloned with specific primers (CUL4attB1 aaaaagcaggctatatgtctct tcctaccaaacgctctactttc; CUL4attB2 agaaagctgggtctaagcaagataattgtatatctgagggt; DDB1attb1 aaaaagcaggctatatgagctcatggaac; DDB1attb2 agaaagctgggttcagtgaagcctagtg; DET1attb1 aaaaagcaggctatatgttcacaagcggtaacgtc; DET1attb2 agaaagctgggttcatcgcctaaaatggatattg; DDB1b900attb1 aaaaagcaggctactctattgcatcttcc; DDB1b2082attb2 agaaagctgggttgtaagttcgccctccc; DDB2attb1 aaaaagcaggctatatggcgacggagtacgagcg; DDB2attb2 agaaagctgggtctactttgttgtccaaacatagac) from a cDNA library (Minet et al., 1992) or mRNA using Pfu-polymerase or an RT-PCR kit (Qiagen, Hilden, Germany), respectively. Full-length and partial products were directly shuffled in one-tube Gateway BP/LR-reactions (Invitrogen) into Gateway-compatible binary, yeast and Escherichia coli expression vectors pGWB18, pBTM116-D9, pACT2 and pDEST15 (Invitrogen), respectively. The CUL4 antisense construct was generated by amplifying a partial 623 bp CUL4 cDNA (CUL4FW atgtctcttcctaccaaacgctctactttc; CUL4RW623 gatgcttccggaaaagctgc) which was first subcloned into pCR2.1 TOPO-TA vector (Invitrogen) and subsequently into BamHI/XbaI sites of the mini binary vector pCB302-3 (Xiang et al., 1999). All fragments were fully sequenced. Binary vectors were introduced by electroporation into Agrobacterium tumefaciens strain GV3101, and were subsequently used for plant transformation according to Clough and Bent (1998).

Yeast two-hybrid assays

A lexA-based Y2H system was used as described in Weber et al. (2005) with pBTM116-D9 (kindly provided by Dr Erich Wanker) as the bait plasmid and pACT2 (Clontech, Mountain View, CA, USA; GenBank accession no. U29899) as the prey plasmid. Both vectors were modified to introduce a Gateway cassette (Invitrogen). Complementary DNAs of the different genes (CUL3a, CUL4, RBX1 and DDB1a) were cloned into pDONR221 (Invitrogen) and subsequently introduced into pBTM116-D9 and pACT2. For Y2H assays, yeast cells were transformed with bait and prey plasmid constructs as described previously (Weber et al., 2005). Cells were grown on synthetic dextrose (SD) minimal medium (Sambrook and Russell, 2001) supplemented with leucine and histidine (SDII). Selected colonies were diluted 1:2000 in autoclaved distilled water before transfer to SD minimal medium without supplements (SDIV) as a control for interaction studies. All constructs were controlled for autoactivation before being used for interaction assays. Photographs of single drops from diluted colonies were taken 3 days after transfer.

Immunoprecipitation, pulldown and Western blot assays

In vitro translated CUL4, DDB1a and DDB1a300−666 were synthesized with the TNT-reticulocyte lysate system (Promega, Mannheim, Germany) using l-[35S] trans-labeled methionine (Amersham, Freiburg, Germany). For pulldown assays from reticulocyte lysate CUL4, DDB1a and DDB1a300−666 were incubated at 4°C for 2 h with GST, GST:DDB1a300−666 and GST:CUL41−453, respectively. Binding assays and washings were done in standard buffer (50 mm Tris/HCl pH 7.5, 150 mm NaCl, 5 mm MgCl2, 0.2% NP-40). Six-day-old seedlings germinated on minimal American Trudeau Society (ATS) medium were used for pulldown assays from plant extracts. Pulldown assays with GST and GST:DDB1a300−666 and washings were done as described before (Hellmann et al., 2003) using a modified standard buffer [100 mm Tris/HCl pH 7.5; 300 mm NaCl, 0.1 mm mercaptoethanol, 1 mm phenylmethylsulfonyl fluoride (PMSF) and 10 μm MG132 (Sigma, Taufkirchen, Germany)]. Proteins were resolved on SDS-PAGE. Products were detected by autoradiography. Western blot and immunodetection were done as described before (Hellmann et al., 2003). Antibodies used were anti-myc (Upstate, Dundee, UK; no. 05-724), polyclonal raised in rabbit against the N-terminal peptide of CUL4 (SFDLESLYQAVDNLC) and secondary horseradish peroxidase conjugated antibodies (Santa Cruz Biotechnology, Inc, Santa Cruz, CA, USA).

Northern blot and RT-PCR analysis

For RT-PCR, total RNA extraction and DNase I digestion were done using a NucleoSpin RNA Plant kit (Macherey-Nagel, Freiburg, Germany). Reverse transcriptase-PCR was done using a one-step RT-PCR Kit (Qiagen). Ninety nanograms of total RNA was used for each RT-PCR reaction (25 cycles, 61°C annealing temperature, 1 min elongation time) with gene-specific primers (CUL4ATG atgtctcttcctaccaaacgctctactttc, CUL4RW650 aacttcaggggccagagaaag, actin2FW tacaacgagcttcgtgttgc, actin2RW gattgatcctccgatccaga). Reactions were done according to the Qiagen manual with 25 cycles and 60°C annealing temperatures. For expression analysis in seedlings, flower, shoot, leaf and root total RNA was harvested from 4-week-old soil-grown plants. Northern blot analysis and radioactive hybridization were done using standard techniques as described before (Hellmann et al., 2003).

In situ hybridization, GUS staining and microscopic work

Flower tissue from 4-week-old soil grown plants was fixed in freshly prepared 4% paraformaldehyde and used for embedding in paraffin. Sectioning and in situ hybridization were done according to Jackson (1991). A 650 base pair fragment from the 5′-end of CUL4 was used as a template for RNA synthesis. Labeling of sense and antisense RNA was done with a DIG-RNA labeling kit (Roche, Mannheim, Germany). For detection of GUS expression, different tissues were incubated for up to 24 h in GUS-staining solution according to Jefferson (1989). Tissue was destained overnight in 100% ethanol. Pictures were taken with an Olympus Digital camera C-4040Zoom mounted on a stereo microscope SZX12 (Olympus, Hamburg, Germany) or an Axioskop 2 plus microscope (Zeiss, Jena, Germany).

Acknowledgements

We would like to thank Sutton Mooney for critical reading, Dr E. Wanker for donating the vector pBTM116-D9, Dr T. Nakagawa for providing binary pGWB vectors, and Alexandra Rieck for technical work. This work was supported by the Deutsche Forschungsgemeinschaft (grants HE3224/1-5 to P.H. and HE3224/5-1 to A.B.), a FIRB grant from the Italian Ministry of Research (MIUR) and support from the Centre National de la Recherche Scientifique (CNRS) to E.L. and M.D.

Ancillary