A role of Arabidopsis COP9 signalosome in multifaceted developmental processes revealed by the characterization of its subunit 3

The COP9 signalosome was originally identified as a repressor of light mediated development in Arabidopsis (Wei and Deng, 1992; Wei et al., 1994a; Wei et al., 1994b; Chamovitz et al., 1996). Plant development is highly plastic and optimized according to environmental cues, of which light plays a dominant role (Kendrick and Kronnberg, 1994). In response to light, Arabidopsis seedlings undergo photomorphogenic development characterized by a short hypocotyl, expanded cotyledons and straightened apices. By contrast, dark-grown seedlings follow a skotomorphogenic developmental program and display an elongated hypocotyl, closed cotyledons and apical hooks. Genetic screens for constitutive photomorphogenic seedling development in darkness identified 11 loci in Arabidopsis collectively known as the COP/DET/FUS genes (Chory et al., 1989; Deng et al., 1991; Miséra et al., 1994; Kwok et al., 1996). Dark-grown mutant seedlings of this group display a pleiotropic phenotype including short hypocotyl, open and expanded cotyledons, accumulation of anthocyanin, and constitutive expression of light regulated genes. The recessive nature of these mutations suggests that their encoded proteins are negative regulators of photomorphogenic development.

Of the 11 cop/det/fus loci, eight are required for the biogenesis of the COP9 signalosome. Mutations in any of these eight loci result in severely retarded seedling development and lethality after the seedling stage (Kwok et al., 1998). So far, at least four genes from this group, COP9, FUS6/COP11, FUS5 and COP8/FUS4, have been found to encode distinct subunits of the COP9 signalosome respectively (Chamovitz et al., 1996; Staub et al., 1996; Karniol et al., 1999; Serino et al., 1999). CSN5, in contrast, is encoded by two redundant genes in the genome (Kwok et al., 1998). Thus no mutation has been identified for CSN5.

Biochemical studies in both plant and animal systems have demonstrated that the COP9 signalosome is a highly conserved, nuclear enriched protein complex with 8 distinct subunits, designated CSN1 to CSN8 respectively (Deng et al., 2000). Both subunit composition and subunit amino acid sequences are conserved among diverged organisms implying a conserved cellular function of the COP9 signalosome. Interestingly, a remarkable subunit-to-subunit similarity between the COP9 signalosome and the lid subcomplex of the 26S proteasome has been observed (Glickman et al., 1998; Seeger et al., 1998; Wei et al., 1998). In mammals, the COP9 signalosome and/or its subunits have been implied in a variety of cellular pathways (Wei and Deng, 1999; Tsuge et al., 2000; Tomoda et al., 1999; Li et al., 2000; Bech-Otschir et al., 2001). In Arabidopsis, most of our understanding of its role in plant development is limited to the seedling stage because all mutants are lethal after the seedling stage (Castle and Meinke, 1994; Miséra et al., 1994; Kwok et al., 1996). Similarly, Drosophila mutations in the COP9 signalosome subunits result in lethality at the late larval or pupal stage (Freilich et al., 1999). The lethality is probably a result of the essential role of the COP9 signalosome in late developmental processes (Wei and Deng, 1999).

Recently, the viable transgenic lines exhibiting a reduced level of the COP9 signalosome using subunit 5 gene co-suppression and antisense approaches, provided an opportunity to examine the role of COP9 signalosome in other developmental processes (Schwechheimer et al., 2001). These transgenic lines had auxin-response defects similar to the loss-of-function mutants of the E3 ubiquitin ligase SCF^TIR1. Furthermore, the COP9 signalosome directly interacts with the SCF^TIR1 E3 ligase complex in vivo and was required for efficient degradation of PSIAA6, a candidate substrate of SCF^TIR1. A possible mechanism for the COP9 signalosome-mediated regulation of the SCF E3 ligases might be through the promotion of the RUB1 de-conjugation from AtCUL1 subunit of SCF complex (Lyapina et al., 2001). Indeed, the COP9 signalosome mutants over accumulate RUB1, a conjugated form of AtCUL1 in vivo (Schwechheimer et al., 2001). It is interesting to note that reduction of the COP9 signalosome resulted in phenotypes similar to those of mutants in AXR1, a component of an enzyme cascade that conjugates the ubiquitin-related protein RUB1 to the AtCUL1 subunit of SCF^TIR1 (del Pozo et al., 1999). Thus, RUB1 conjugation and de-conjugation cycles must work together to promote this SCF^TIR1 E3 ligase-mediated auxin response and degradation of SCF^TIR1 substrates.

In both mammals and plants, the interaction between the COP9 signalosome and SCF complexes is mediated through CUL1 and RBX1, the SCF core subunits (Lyapina et al., 2001; Schwechheimer et al., 2001). It is well accepted that distinct SCF-type E3 ubiquitin ligases share common core subunits such as CUL1 and RBX1 but differ in their F-box domain subunit, which is responsible for their substrate specificity (Deshaies, 1999). In Arabidopsis, there are over 300 F-box proteins in the sequenced genome. Several characterized F-box proteins, including UFO (Samach et al., 1999), COI1 (Xie et al., 1998), EID1 (Dieterle et al., 2001), ZTL (Somers et al., 2000), and FKF1 (Nelson et al., 2000) have been shown to play important roles in distinct developmental pathways. By interacting with the core subunits, it is therefore possible that the COP9 signalosome has the capacity to interact with many different SCF-type E3 ubiquitin ligases and therefore be involved in the respective developmental processes that these SCF E3 ligases regulate.

We report the molecular characterization of the COP9 signalosome subunit 3 (CSN3), which is encoded by the Arabidopsis FUS 11 locus. The fus11 mutant is defective in the COP9 signalosome and accumulates significant amount of multi-ubiquitinated proteins. Partial loss-of-function plants of the COP9 signalosome, generated using co-suppression approach, exhibited defects in multiple aspects of plant development, including pattern formation, phyllotaxy and organ identity. These results support the notion that the COP9 signalosome interacts with multiple SCF E3 ligases and is involved in multifaceted developmental processes.

The fus11-U203, fus6-T236, and cop10-1 mutants have been described previously (Miséra et al., 1994; Wei et al., 1994a; Wei et al., 1994b). The cop1-6 mutant was described by McNellis et al. (McNellis et al., 1994). The seeds were planted on agar plates containing GM medium plus 1% sucrose (Wei et al., 1994), and were cold treated at 4°C for 5-12 days before being transferred to the indicated growth chambers. The light intensity used was 156 and 111 μmole/m²/second for white and red light respectively.

The peptide sequences of cauliflower CSN3 (Serino et al., 1999) were used to search the NCBI gene bank and the Arabidopsis EST database. A partial cDNA clone of tomato (Lycopersicon chilense, U19099) was found to share high homology with two peptide sequences from cauliflower. This cDNA clone was obtained by RT-PCR from tomato. The tomato cDNA obtained was then used as a probe to screen a λ ZAP II cDNA library obtained from the Arabidopsis Biological Resource Center (Columbus, Ohio). Among the 60 positive plaques, 10 were chosen for further screening and in vivo excision. An in frame stop codon at the 5′ untranslated region indicated that two clones obtained are full-length cDNA clones.

To map the CSN3 gene, the full-length CSN3 cDNA was used as a probe to hybridize the CIC YAC filter obtained from the Arabidopsis Biological Resource Center (Columbus, Ohio) according to the instructions provided. Among the 5 positive YACs, two of them (CIC9B8 and CIC5H3) have been physically mapped and were found to be close to the FUS11 locus on the chromosome V.

The full-length cDNA of Arabidopsis CSN3 gene (ecotype Columbia) was cloned into the binary vector pPZPY122 at the BamHI and KpnI sites in sense orientation under the control of the 35S promoter (Yamamoto et al., 1998). The constructs were transformed into wild-type Arabidopsis in both Columbia and Landsberg ecotypes, and the fus11-U203 heterozygous plants via vacuum infiltration. 26 independent transgenic lines in the heterozygous fus11-U203 background were examined for possible complementation. For the transgenic plants in wild-type background, 30 independent lines were used for phenotype observation in both the Columbia and Landsberg background respectively. To facilitate gentamycin selection, the seeds were germinated on gentamycin plates (100 mg/l) for 3 days in darkness and then transferred to continuous white light. The greening process of the sensitive plants was completely blocked while the transgenic plants turned green normally.

Arabidopsis seedlings were homogenized in a Tris buffer (TB) containing 50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 10 mM MgCl_2, 5 mM EDTA, 5 mM DTT, 10% glycerol, with freshly added protease inhibitor, phenylmethylsulfonylfluoride (PMSF; 2 mM) and protease inhibitor-cocktail™ (Boehringer Mannheim) as instructed by the manufacturer. The homogenate was microcentrifuged for 15 minutes and the supernatant was filtered through a 0.2 μm filter (Gelman Sciences, Ann Arbor, MI) before loading onto a Superose 6 (HR 10/30) gel filtration column (Pharmacia). The column was equilibrated with a buffer containing 50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 10 mM MgCl₂, 5 mM EDTA, 2 mM DTT, and 10% of glycerol. After loading the sample, proteins were eluted in the same buffer at a flow rate of 0.2 ml/minute. Fractions of 0.5 ml each were collected, starting from the onset of the column void volume (7.5 ml) and were concentrated using StrataClean Resin (Stratagene) as described by Kwok et al. (Kwok et al., 1998). All manipulations were carried out at 4°C. Equal volumes of each fraction were used for SDS-PAGE and then subjected to immunoblot analysis.

A CSN3 recombinant protein was generated by fusing the full-length CSN3 protein to the His tag at the BamHI and EcoRI sites in the vector pRSET C (Invitrogen). The overexpressed proteins were purified by Ni-NTA superflow resin (QIAGEN) and used to immunize rabbits. Rabbit polyclonal antibodies to CSN1, CSN5, CSN7 and CSN8 were described previously (Kwok et al., 1999: Karniol et al., 1999; Staub et al., 1996; Chamovitze et al., 1996). The polyclonal antibodies against ubiquitin were purchased from Sigma.

All western blot and immunoprecipitation analyses were carried out as previous described (Staub et al., 1996) with minor modifications. The antibodies coupled with protein A beads were mixed with the protein extracts for 3 hours instead of 6 hours and the beads were washed with PBSM buffer supplemented with 0.4 M NaCl and 0.2% SDS. The immunoprecipitated proteins were released from the protein A beads and analyzed with SDS-PAGE. To detect the ubiquitinated proteins, the filter was autoclaved for 10 minutes to completely denature ubiquitin after the proteins blotted to the filter.

12-day-old transgenic seedlings without true leaves or leaf buds were collected. The roots and hypercotyls were removed. Total RNAs were extracted with Qiagen RNAeasy plant kit. Wild-type plants treated in the same way were used as control. Northern blots were carried out as previously described (Peng et al., 1996) and detected with a phosphoimage system.

To clone the gene encoding CSN3 of the Arabidopsis COP9 signalosome, we used the CSN3 peptide sequences obtained from the purified cauliflower COP9 signalosome (Serino et al., 1999) to search the NCBI and the Arabidopsis databases. At the time, no Arabidopsis EST clone or genomic clones corresponding to the peptide sequences were found; however, a highly homologous partial cDNA clone of tomato (U19099) was identified and used to screen an Arabidopsis cDNA library (Kwok et al., 1998). Ten Arabidopsis cDNA clones were recovered and further analyzed. A near full-length cDNA sequence was deduced that encoded for a 429 amino acid protein (Fig. 1A). The predicted protein sequence contains all three regions corresponding to the peptide sequences from cauliflower (Fig. 1B). The CSN3 protein contains an imperfect leucine zipper domain in the N-terminal region (amino acid 18 to 46) and a PCI domain in the C-terminal region (amino acids 194 to 362) (Fig. 1A). The Arabidopsis CSN3 shares 42% and 31% identities with its human and Drosophila counterparts respectively (Fig. 1A). Genomic Southern blot analysis and database searches of the completed Arabidopsis genome sequence revealed that the CSN3 is encoded by a single copy gene in Arabidopsis (data not shown).

While Arabidopsis CSN1 and CSN8 are exclusively present in the COP9 signalosome (Wei et al., 1994; Staub et al., 1996), CSN5 and CSN7 exists both in complex and monomeric forms (Kwok et al., 1998; Karniol et al., 1999). To examine whether CSN3 cofractionates with the COP9 signalosome and whether it also exists in any other form, we raised polyclonal antibodies against CSN3 and carried out gel filtration fractionation analysis of the total protein extracts of light-grown seedlings. The elution profiles of CSN3 and other representative subunits of the COP9 signalosome are shown in Fig. 2A. Consistent with previous observations, CSN1 is only present in the complex form and CSN5 existed in both monomeric and complex forms. The CSN3 protein had exactly the same elution profile as CSN1, indicating that CSN3 is only present in the COP9 signalosome.

To further substantiate that CSN3 is part of the COP9 signalosome, we carried out co-immunoprecipitation experiments using antibodies against CSN3 and CSN5 respectively. The immunoprecipitated proteins from crude cell extracts were then examined by protein blot analyses for the presence of CSN1, CSN3 and CSN5. As shown in Fig. 2B, CSN1, CSN5 and CSN3 were clearly precipitated by antibodies against either CSN5 or CSN3, but not by the corresponding pre-immunoserum. This result indicates that CSN3 is physically associated with the COP9 signalosome.

Among the 11 pleiotropic COP/DET/FUS loci, mutations in eight of them result in apparent absence of the COP9 signalosome (Kwok et al., 1998; Wei and Deng, 1999). Molecular characterization of the COP9 signalosome has demonstrated that COP9, FUS6/COP11, FUS5 and COP8/FUS4 encode subunits of the COP9 signalosome (Wei et al., 1994; Kaniol et al., 1999; Staub et al., 1996; Serino et al., 1999). To investigate whether CSN3 is also encoded by one of the eight loci, the full-length CSN3 cDNA was used as a probe to hybridize the CIC YAC filter, set at high stringency conditions. Two over-lapping YAC clones, CIC9B8 and CIC5H3, were found to contain the CNS3 gene. These two clones have been mapped to a region close to the FUS11 locus between markers nga 151 and nga 106, indicating that the FUS11 locus is a good candidate for the CSN3 gene.

To verify this prediction, the genomic CSN3 gene was obtained by PCR from both wild-type and fus11-U203 mutant seedlings. Sequence analysis revealed that the 2.9 kb transcribed genomic region of the wild-type CSN3 gene contains 11 exons and 10 introns (Fig. 3A). In the fus11-U203 mutant, there is a G to A mutation at the 3′ splicing junction site of the ninth intron (Fig. 3B). RT-PCR analysis of cDNAs from the fus11-U203 mutant indicated that this mutation resulted in an aberrant splicing to a cryptic receptor site within the ninth intron, introducing two novel amino acids followed by a stop codon. This stop codon results in early termination of the reading frame and a possible production of a mutant protein missing the C-terminal 44 amino acids (Fig. 3C). Protein gel filtration analyses revealed that CSN3, CSN1 and CSN5 were all absent in the fus11-U203 mutant, while CSN5 could be found exclusively in the monomeric form (Fig. 2A). This indicates that the last 44 amino acids of CSN3 play an essential role in either the assembly or stability of the COP9 signalosome.

To confirm that CSN3 is encoded by the FUS11 gene, the CSN3 cDNA under the control of the cauliflower mosaic virus 35S promoter was introduced into the fus11 mutant background to test functional complementation. Since the homozygous fus11 mutant is lethal, the transgene was stably transformed into heterozygous plants. Both the phenotype and the transgene segregation were analyzed in the T₂ generation to reveal the number of the T-DNA insertion loci and possible functional complementation. In the absence of phenotype rescue, the wild type to mutant ratio in the T₂ generation should be 3:1. With functional complementation by a single locus of the transgene unlinked to the FUS11 locus, the wild type to mutant ratio should be 15:1. If multiple transgene loci are involved, a higher wild type to mutant ratio is expected. The segregation ratios of five representative transgenic lines are summarized in Table 1. For the 4 lines with single insertion, the ratios of wild type to mutant were 12.3:1, 13.6:1, 14.2:1 and 17.0:1 respectively. For the line with two possible transgene loci, a ratio of 53 to 1 was observed. These results clearly indicated that the overexpressed CSN3 cDNA functionally complements the fus11 mutation.

As an initial step to extend our hypothesis that the COP9 signalosome may regulate multiple E3 ligase-mediated protein degradation, we examined the effect of the COP9 signalosome defect on the overall profile of cellular ubiquitinated proteins. To this end, total soluble proteins from wild-type and COP9 signalosome mutants were separated by gel filtration according to their native sizes. The collected fractions were then concentrated, further separated on SDS-PAGE, and subjected to protein blot analysis with anti-ubiquitin antibodies. Consistent results were observed regardless of which batch of ubiquitin antibodies were used and an example of this profiling of ubiquitinated proteins in wild type, fus11-U203, fus 6-T236 and cop1-6 mutant seedlings is shown in Fig. 4A. Two general changes were observed in the overall profiles of ubiquitinated proteins in the COP9 signalosome mutants. First, absence of the COP9 signalosome lead to an overall increase in multi-ubiquitinated proteins, especially in the high molecular mass range (the smearing high molecular mass bands in fractions 5-15). Second, dramatic changes in the levels of several relatively abundant ubiquitin conjugated proteins were evident (see arrows). Similar changes in the profile of cellular ubiquitinated proteins were observed in several other COP9 signalosome mutants, but neither the cop1 nor the cop10 mutants examined exhibited such a change compared to wild type (Fig. 4A and data not shown). Thus, the COP9 signalosome has a distinct function from other COP/DET/FUS protein in the ubiquitin/”proteasome pathway.

To test if the COP9 signalosome-promoted protein degradation is substrate specific, we examined the degradation of both HY5 and phytochrome A (phyA) proteins in the COP9 signalosome-deficient fus11-U203 mutant. HY5 and phyA are two light regulatory components whose light regulated degradation has previously been shown to be mediated by the 26S proteasome (Vierstra and Callis, 1999; Osterlund et al., 2000). While HY5 is degraded rapidly in darkness, phyA degradation depends on light stimulation. As shown in Fig. 4B, the COP9 signalosome deficiency completely inhibited degradation of HY5 in both dark-grown seedlings and in light grown seedlings subjected to dark adaptation. However, the COP9 signalosome deficiency had no detectable effect on light-dependent degradation of phytochrome A. Both light-grown seedlings and red light-treated dark-grown fus11 mutant seedlings had normal patterns of phyA degradation. Thus, the involvement of the COP9 signalosome in ubiquitin-proteasome mediated proteolysis appears to be substrate specific. Therefore alteration of the protein ubiquitination pattern in the COP9 signalosome mutants are likely limited to those proteins that are subjected to the COP9 signalosome regulation.

Since all the cop or fus mutants for genes encoding the COP9 signalosome are lethal after the seedling stage, the role of the COP9 signalosome in late developmental processes remains unclear. The recent examination of CSN5 partial loss-of-function transgenic lines revealed a role of the COP9 signalosome in auxin response (Schwechheimer et al., 2001). To further define the developmental role of the COP9 signalosome, we obtained over 30 transgenic lines in each of the Columbia and Landsberg ecotypes by stable transforming the 35S driven CSN3 cDNA transgene construct. The majority of the transgenic lines exhibited a variety of phenotypes in the multiple generations examined (Table 2). Western blot analyses with antibodies specifically against Arabidopsis CSN3 and other COP9 signalosome subunits revealed that all the plants with observed phenotypic defects from all lines examined had reduced amounts of the COP9 signalosome (Fig. 5A and data not shown). This indicates that a co-suppression event occurred in those transgenic lines, resulting in a reduction of CSN3 and the COP9 signalosome abundance. This reduction of the COP9 signalosome is likely responsible for the observed phenotypes.

A striking feature of these transgenic lines is that each independent line segregates a variety of different phenotypes regardless of the specific phenotype of the parental plants (with the exception of early seedling defects, see below). Although the phenotype of a given progeny varies, the overall spectra of phenotypes within a given line are quite consistent. In fact, the same phenotypic spectra have been inherited through at least four tested generations so far. It was also common to have an individual transgenic plant exhibiting multiple phenotypes. To better describe the observed phenotype in our partial loss-of-function COP9 signalosome transgenic plants, selected phenotypes are presented according to the organ and the developmental stage affected.

The vegetative growth of Arabidopsis consists of two consecutive phases, the juvenile stage and the adult stage. The first pair of true leaves at the juvenile phase are small, round and positioned in opposite phyllotaxy, whereas later adult leaves are spatulate, serrate and positioned in spiral phyllotaxy (Medford et al., 1992). In the CSN3 co-suppression plants, a variety of leaf shape defects have been observed at both the juvenile and adult phases (Fig. 6A). The selected leaf shapes include strip (panel a of Fig. 6A), needle (panel b of Fig. 6A), and horn (panel c of Fig. 6A). In addition, phyllotaxy changes have also been observed at the juvenile stage in 5 of the 30 transgenic lines in the Landsberg background. For example, panel d of Fig. 6A shows phyllotaxy pattern of three instead of two symmetric leaves in rosette pattern. Western blot analyses revealed that all these defective leaves have reduced levels of CSN3 and other COP9 signalosome subunits (Fig. 5A and data not shown). At the same time, over accumulation of ubiquitinated proteins were also detectable in these defective leaves (Fig. 5B). Together, these results suggest that proteasome-mediated protein degradation defects are probably responsible for the leaf developmental defects.

In the progeny of the CSN3 co-suppression lines, early seedling developmental defects have been constantly observed. As shown in Fig. 6B, the phenotypes included seedlings with single, triple and quadruple cotyledons (panels a, b and c of Fig. 6B), and seedlings without embryonic roots (panel f, Fig. 6B). These seedling phenotypes are usually below 3% of the progeny of the given transgenic lines. These early seedling phenotypes are very similar to the auxin related mutants such as PIN-FORMED and PINOID in cotyledon number change (Goto et al., 1987; Bennett et al., 1995), confirming our observation that the COP9 signalosome indeed regulates SCF^TIR1 and the auxin responses (Schwechheimer et al., 2001). In addition, adult phenotypes of the CSN3 co-suppressing lines also supported this conclusion. One frequent adult phenotype in our transgenic progeny (between 5% and 30%) during vegetative growth is the loss of apical dominance (panels d and e of Fig. 6B). This type of plants often has wrinkled leaves (panel e of Fig. 6B). The appearance resembles the phenotype of auxin resistant mutants (Estelle and Somerville, 1987) and those CSN5 co-suppression lines recently reported (Schwechheimer et al., 2001).

Another common seedling defect is that the seedlings fail to produce true leaves and leaf buds (Fig. 6C). Northern blot analysis of such seedlings demonstrated that the Shoot Meristemless (STM) gene was expressed at a lower level compared with the wild type (Fig. 6C, lower panel). This type of plants occurred at relatively higher frequency compared to other seedling phenotypes shown in Fig. 6B (Table 2). It is important to note that the parental plants producing defective seedlings tend to display adult developmental defects and have reduced COP9 signalosome abundance (data not shown) while other later developmental defects usually have no correlation with their parental phenotypes.

The reproductive growth of Arabidopsis begins with the transition from an adult vegetative meristem to an inflorescence meristem, which will give rise to the floral inflorescence. A variety of phenotypes were observed in the inflorescence of CSN3 transgenic plants as shown in Fig. 7A. The inflorescence in panel a has a phyllotaxy defect, so that four flowers were positioned on one side of the stem (see white arrow). In panel b, multiple poorly developed inflorescence branches radiated out from the same position, suggesting a possible defect in branching pattern. The inflorescence in panel c had a thick and expanded stem and the flowers originated without any particular phyllotaxy pattern – another malfunction in pattern formation. In addition to its big size, the inflorescence in panel d had no internode elongation at all, therefore the inflorescence somewhat resembles a sunflower head. While the plant shown in panel e had tightly packed siliques at the top end of inflorescence, and the plant shown in panel f has extremely short stalks. These developmental defects suggest that the COP9 signalosome is involved in multiple aspect of inflorescence development.

Although the percentage of plants with floral defects was not high (Table 2), the diversity of the flower phenotypes in the CSN3 transgenic plants was striking. Some selected examples are shown in Fig. 7B. In the flower in panel a, a gynoecium was developed within another gynoecium (see white arrow). While the stigmatic tissues of the later gynoecium were well developed (see green arrow), its two carpels failed to fuse at the top. Meanwhile, the adjacent flower only had two unfused carpels. The phenotype of this flower is similar to that of a weak AGAMOUS mutant (Sieburth et al., 1995). The gynoecium shown in panel b was developed within a partially fused sepal, while other floral organs were completely missing. Panel c presents an ‘empty’ flower, also described for the severe ufo mutants (Wilkinson and Haughn, 1995). In this flower, one of the two sepal-like structures at the first whorl organ position had ovule-like structures along the middle axis (see white arrow), indicating a partial homeotic transformation. The flowers shown in panel d either completely missed or had reduced numbers in petals, stamen and sepals, suggesting that probably both A and B type flower identity genes were affected. This hypothesis gains support from the phenotypes shown in panel e. Here the flowers had no sepals and stamens but still produced structurally defected petals and carpels. In contrast to the above examples, the flower in panel f had an extra stamen and an extra petal. These few examples suggest that the floral defects include pattern formation, homeotic organ transformation and organ number modification. Western blots using flowers from two representative flower types, as shown in panels c and e, are presented in Fig. 5A. The results clearly indicate that those defective flowers had a reduced COP9 signalosome level. In addition, multi-ubiquitin-conjugated proteins were also over accumulated in these transgenic flowers (Fig. 5B and data not shown).

We have presented definitive evidence that subunit 3 of the COP9 signalosome is encoded by the FUS11 gene in Arabidopsis. First, the Arabidopsis CSN3 protein cofractionated and co-immunoprecipitated with other known subunits of the COP9 signalosome. Thus CSN3 is a subunit of the COP9 signalosome. Second, the CSN3 gene mapped to the same region on chromosome V as the FUS11 locus. Further, a specific mutation in the CSN3 gene, which results in an alternative splicing and early termination of the reading frame, has been identified in the fus11-U203 mutant. Third, the fus11-U203 mutant failed to accumulate CSN3 and several other subunits of the COP9 signalosome. And finally, overexpression of CSN3 functionally complemented the fus11-U203 mutation. With this new identification of FUS11 encoding subunit 3, it will be of interesting to determine the relationship between the three remaining cop/det/fus loci required for the COP9 signalosome accumulation and the two uncharacterized subunits of the COP9 signalosome (Kwok et al., 1998; Serino et al., 1999).

The absence of the COP9 signalosome in the fus11-U203 mutant confirms our earlier observations that when any single subunit is defective, the complex becomes unstable and fails to accumulate. While some subunits, such as CSN5 and CSN7, are present also as monomers, other subunits, including CSN1, CSN3, CSN4 and CSN8, are degraded to undetectable levels in vivo. The fact that CSN3 is only present in the complex form indicates that CSN3 acts exclusively as a subunit of the COP9 signalosome. Thus, our manipulation of the CSN3 abundance should directly and specifically affect the COP9 signalosome level in vivo.

It is striking that reduction of the cellular COP9 signalosome abundance resulted in such a variety of developmental defects in Arabidopsis. The most common phenotypes include alterations in pattern formation, phyllotaxy, homeotic transformation and organ size or number modifications from embryogenesis to flower development (see Table 2). Although the phenotype among individual progenies of the transgenic lines varies greatly, several constant features are evident throughout our studies. First, the spectra of the phenotypes for most of the reduction-of-function lines are quite similar when the population is large enough. Second, for a given transgenic line and with the exception of embryogenesis defects, the spectrum of phenotypes is constant from generation to generation and independent of the phenotype of the parent plant. Third, the phenotypic defects apparent at late developmental stages were in general milder, while the early phenotypes are more severe and often detrimental to plant development.

The phenotypes observed in the transgenic lines provided us with the opportunity to systematically document the role of COP9 signalosome in plant development. It is likely that specific regulatory proteins are targets of the COP9 signalosome during plant development and the reduction of the COP9 signalosome resulted in altered regulation of the protein abundance and defects in their associated developmental processes. The diverse phenotypes we observed would be consistent with the notion that the COP9 signalosome interacts and regulates multiple E3 ligases and thus their substrate protein degradation. Besides the well-documented role of the COP9 signalosome in SCF^TIR1-mediated protein degradation and auxin response, some of our co-suppression line phenotypes also indicate that other SCF-type E3 ligases also require the COP9 signalosome for their proper function. For example, some aspects of the observed flower defects resemble those of the ufo mutants (Levin and Meyerowitz, 1995; Wilkinson and Haughn, 1995; Ingram et al., 1995). UFO is a F-box protein and has been suggested to be part of a SCF-type E3 ligase complex involved in flower development (Samach et al., 1999). Most likely, the ufo mutations cause a defect in the degradation of proteins critical for flower development, thus leading to an abnormal flower phenotype. It is thus plausible that some of the flower defects we observed in our CSN3 co-suppression lines are caused by an improper function of the SCF^UFO as a result of reduced COP9 signalosome levels. Beside TIR1 and UFO, several other F-box proteins have been reported to regulate plant defense (Xie et al., 1998), flower timing (Somers et al., 2000; Nelson et al., 2000) and light regulation (Dieterle et al., 2001). The different F-box proteins seem to be involved in distinct developmental processes but all share an ability to form a specific SCF E3 ligase complex with the same core subunits. Since Arabidopsis has more than 300 F-box proteins, it is quite possible that the COP9 signalosome could accomplish most, if not all, its function in regulating many developmental processes by modulating the activities of these distinct F-box-containing SCF E3 ligases.

How can individual progenies of the same transgenic line have different phenotypes and the different parts of the same transgenic plant have different phenotypes? It is well established that proteasome-mediated proteolysis plays a crucial rule in plant development, by ensuring that the activity level of its targets is controlled in a precise temporal and spatial manner according to developmental and environmental cues. Some of these protein degradation events likely require the function of the COP9 signalosome. The reduction of the COP9 signalosome would result in inappropriate activities of key developmental regulators above or below their critical thresholds, thus causing defects in developmental processes controlled by them. The stochastic nature of the phenotypes of CSN3 transgenic lines could be caused by variation of the levels of the COP9 signalosome among individual plants at different time points of the plant development, or in different cells and tissues within a plant, due to the temporal and spatial pattern of gene silencing. It is worth noting that the phenotype of the ufo mutant also appeared to be stochastic (Levin and Meyerowitz, 1995; Wilkinson and Haughn, 1995). UFO is a F-box protein involved in protein degradation as a component of the SCF E3 ligase complex (Samach et al., 1999). Thus it is not entirely impossible that stochastic phenotype defects might be a characteristic feature of protein degradation malfunction.

A direct role of the COP9 signalosome in the ubiquitin/”proteasome pathway is supported by several observations. In addition to its essential role in HY5 degradation in darkness (Osterlund et al., 2000) (Fig. 4B), the COP9 signalosome was reported to physically interact with a SCF-type ubiquitin E3 ligase and is essential for the E3 ligase mediated process (Schwechheimer et al., 2001). In this study, we provide evidence that there is an overall increase in cellular multi-ubiquitin-conjugated proteins in the COP9 signalosome-deficient mutants and in CSN3 transgenic plants with reduced levels of the COP9 signalosome (Fig. 4A, Fig. 5B). As the multi-ubiquitin-conjugated proteins are the substrates for the proteasome, their over accumulation in the COP9 signalosome-defective plants indicates a defect in the presentation of those substrates to the proteasome. Since the COP9 signalosome and SCF-type E3 ligases associate with each other physically in vivo (Schwechheimer et al., 2001), it is possible that the COP9 signalosome could function in releasing the multi-ubiquitin conjugated proteins from the E3 ligases and presenting them in a proper form to the 26S proteasome. This specific role of the COP9 signalosome could be related to its ability to promote RUB1 de-conjugation from the AtCUL1 subunit of the SCF ligase complex (Lyapina et al., 2001; Schwechheimer et al., 2001). The best case to be considered is the SCF^TIR1-mediated protein degradation and auxin response, where both AXR1-mediated RUB1 conjugation to AtCUL1 and the COP9 signalosome-promoted RUB1 de-conjugation are essential. It is feasible that this cycling of RUB1 conjugation and de-conjugation of the SCF ligase provide the dynamic conformation changes necessary for releasing the multi-ubiquitin conjugated substrate and presenting to the 26S proteasome. Clearly, this area should be a fertile ground for future study.

It is not clear at this point at which level the specificity of the COP9 signalosome to the substrate is conferred. We have shown that while the COP9 signalosome is essential for proteasome-mediated HY5 degradation, it does not play an observable role in proteasome-mediated phytochrome A degradation (Fig. 4) (Jabben et al., 1989; Osterlund et al., 2000). There are two possible ways by which the COP9 signalosome could act specificity toward selected substrates. First, the COP9 signalosome may only interact and regulate a selected group of E3 ligases in vivo. This is possible but not supported by the fact that the core subunits of the SCF complexes are responsible for contacting the COP9 signalosome (Lyapina et al., 2001; Schwechheimer et al., 2001). Alternatively, the COP9 signalosome could directly confer its specificity toward substrates. This could be achieved by directly influencing the E3 ligase conformation through its association, and/or enhance the interaction of the E3 ligase with selective substrates. Further studies will be required to sort out these alternative models.

We are grateful to Dr Magnus Holm for commenting this manuscript. This research is supported by a NSF grant (MCD-0077217) and a BARD grant (IS-3123-99) to X.-W. D. X.-W. D. is a NSF Presidential Faculty Fellow. ZP is a recipient of a National Institute of Health postdoctoral fellowship.