Genetic analysis of CAND1–CUL1 interactions in Arabidopsis supports a role for CAND1-mediated cycling of the SCF^TIR1 complex

Abstract

SKP1-Cullin1-F-box protein (SCF) ubiquitin-ligases regulate numerous aspects of eukaryotic growth and development. Cullin-Associated and Neddylation-Dissociated (CAND1) modulates SCF function through its interactions with the CUL1 subunit. Although biochemical studies with human CAND1 suggested that CAND1 plays a negative regulatory role by sequestering CUL1 and preventing SCF complex assembly, genetic studies in Arabidopsis have shown that cand1 mutants exhibit reduced SCF activity, demonstrating that CAND1 is required for optimal SCF function in vivo. Together, these genetic and biochemical studies have suggested a model of CAND1-mediated cycles of SCF complex assembly and disassembly. Here, using the SCF^TIR1 complex of the Arabidopsis auxin response pathway, we test the SCF cycling model with Arabidopsis mutant derivatives of CAND1 and CUL1 that have opposing effects on the CAND1–CUL1 interaction. We find that the disruption of the CAND1–CUL1 interaction results in an increased abundance of assembled SCF^TIR1 complex. In contrast, stabilization of the CAND1–CUL1 interaction diminishes SCF^TIR1 complex abundance. The fact that both decreased and increased CAND1–CUL1 interactions result in reduced SCF^TIR1 activity in vivo strongly supports the hypothesis that CAND1-mediated cycling is required for optimal SCF function.

The ubiquitin/26S proteasome pathway regulates numerous aspects of plant growth and development. SKP1-Cullin1-F-box protein (SCF) complexes comprise the largest class of ubiquitin-ligases in Arabidopsis (1). These multisubunit enzymes consist of four subunits: the RING protein RBX1 binds to a C-terminal domain of the CUL1 scaffold protein, whereas the SKP1 adaptor protein binds to the N terminus of CUL1. The fourth subunit is any of several F-box proteins (FBPs) that bind SKP1 via their F-box domain. FBPs act as recognition factors to recruit specific substrates to the SCF complex for ubiquitination. With ≈700 FBP genes in the Arabidopsis genome, plants have the capacity to assemble many distinct SCF complexes, which likely play a prominent role in regulating many biological processes (1).

SCF^TIR1 contains the TIR1 FBP and regulates response to the plant hormone auxin by targeting repressors of auxin response known as Aux/IAA proteins for ubiquitin-mediated proteolysis (2). Several additional plant FBPs have been implicated in a diverse range of biological processes, including hormonal signaling, photomorphogenesis, organ development, self-incompatibility, defense response, and circadian clock control (3). How plants and other eukaryotes regulate this multitude of FBPs assembling with a common CUL1-RBX1 core complex to form distinct SCF complexes for controlling the stability of hundreds of substrates is unclear.

Studies in numerous eukaryotes have shown that the conjugation of the ubiquitin-like protein RUB/NEDD8 to the CUL1 subunit is required for optimal SCF ubiquitin-ligase activity (4–6). In Arabidopsis, mutations affecting the RUB-modification pathway reduce SCF^TIR1 activity, resulting in the stabilization of Aux/IAA proteins and diminished auxin response (7, 8). In vitro assays demonstrate that modification increases SCF activity (6, 9) possibly by facilitating the recruitment of E2-ubiquitin conjugates to the SCF complex (10). Hypermodification of CUL1 also disrupts SCF^TIR1 function in vivo (11), and cleavage of RUB off CUL1 by the COP9 signalosome (CSN) is required for optimal SCF activity. Although purified CSN inhibits SCF activity in vitro by removing RUB from CUL1, SCF substrates accumulate in CSN-deficient mutants, indicating reduced SCF activity in vivo (12). Together these findings suggest that RUB modification of CUL1 is highly dynamic, with cycles of RUB conjugation and removal being required for proper SCF activity.

Discovery of the human Cullin-Associated and Neddylation-Dissociated (CAND1) protein suggested how cycles of RUB conjugation and cleavage might regulate SCF activity. Biochemical studies found that CAND1 specifically binds unmodified CUL1 to form a ternary complex with the CUL1-RBX1 SCF catalytic core (13, 14). Because CAND1 binding precludes CUL1 from binding SKP1, CAND1 was proposed to act as a negative regulator of SCF complex assembly by sequestering CUL1. Because CAND1 only binds unmodified CUL1, a model invoking cycles of RUB modification and cleavage was proposed to modulate interactions between CUL1 and CAND1 to regulate SCF complex assembly (12). However, more recent biochemical and structural studies suggest that RUB conjugation occurs after SCF complex assembly (15, 16), raising new questions about the regulation of CAND1–CUL1 interactions.

Although these in vitro studies suggested that CAND1 acts as a negative regulator of SCF complexes, the identification of cand1 mutants in Arabidopsis indicated otherwise. Loss-of-function mutations in CAND1 confer multiple defects in growth and development, including reductions in auxin response and other SCF-regulated processes (17, 18). Furthermore, IAA7, an SCF^TIR1 substrate, and RGA, an SCF^SLY1 substrate, exhibit increased stability in cand1 mutants, clearly indicating that CAND1 positively regulates SCF function in vivo. These genetic findings can be reconciled with the in vitro data, however, if cycles of SCF assembly and disassembly are required for proper SCF activity in vivo (12). Such a requirement would explain how mutations in factors hypothesized to promote SCF disassembly, such as CAND1 and the CSN, could result in reduced SCF activity. Because many FBPs compete for access to a common CUL1-RBX1 catalytic core, SCF cycling may provide plasticity to a cell's complement of assembled SCF complexes. Cycling also has been proposed to protect FBP subunits from autoubiquitination in the absence of available substrates and could conceivably provide a mechanistic basis for promoting the preferential incorporation of FBP–substrate complexes into the SCF or the release of ubiquitinated substrates.

As a genetic test of the SCF cycling hypothesis, we characterized SCF^TIR1 assembly and function in Arabidopsis cand1 and cul1 mutants that differentially affect the CAND1–CUL1 interaction. Our results support the cycling model and demonstrate that CAND1–CUL1 interaction dynamics play a crucial role in the control of SCF activity in vivo.

Results

Opposing Effects of eta2-1 and axr6-2 on CAND1–CUL1 Interactions.

We have previously described a CAND1 mutant, eta2-1, that contains a missense mutation within a highly conserved region near the C terminus (17). eta2-1 plants exhibit diminished SCF^TIR1 activity and a corresponding reduction in auxin response. To examine how the mutation affects CAND1 function, we examined interactions between CAND1 and CUL1 by using coimmunoprecipitation assays. As previously reported, the eta2-1 mutation had no effect on CAND1 or CUL1 protein levels (Fig. 1A). However, although unmodified CUL1 efficiently coprecipitated with CAND1 from wild-type extracts, little if any CUL1 was observed in α-CAND1 precipitates from eta2-1 extracts (Fig. 1A). Thus, the eta2-1 point mutation greatly diminishes the ability of CAND1 to bind CUL1. Consistent with this finding, the eta2-1 mutation completely abolished the ability of CAND1 to interact with CUL1 in a yeast 2-hybrid assay (Fig. 1C).

Fig. 1.

Effects of the eta2-1 and axr6-2 mutations on CAND1–CUL1 interactions. (A and B) α-CAND1 antibody was used to precipitate CAND1 from Col, eta2-1, axr6-1, and axr6-2 seedling extracts. Precipitates were immunoblotted with CAND1 and CUL1 antisera. Densitometric quantification of CUL1 levels in the extracts and precipitates relative to Col are shown below each lane in B. (C) (Left) CAND1-CUL1 yeast 2-hybrid assays. β-gal activities are expressed relative to wild type. Error bars indicate SD (n = 3). (Right) Western blots of yeast extracts expressing the bait and prey constructs.

axr6-1 and axr6-2 are dominant-negative alleles of CUL1 that confer reduced auxin response when heterozygous and a seedling-lethal phenotype when homozygous (19). These two mutations are distinct missense alleles of the same amino acid in an N-terminal domain of CUL1 that is involved in both SKP1 and CAND1 binding (16). To investigate how axr6-1 and axr6-2 affect the CAND1–CUL1 interaction, we performed α-CAND1 immunoprecipitations with wild-type and homozygous mutant seedling extracts. In both axr6 mutants, we detected a dramatic increase in the amount of CUL1 coprecipitating with CAND1 (Fig. 1B). Like wild-type CUL1, however, the interaction was abolished by the eta2-1 mutation [supporting information (SI) Fig. S1A]. To investigate the possibility that axr6-2 increases affinity for CAND1, we examined the interaction in a yeast 2-hybrid system. In contrast to our results with plant extracts, axr6-2 did not obviously affect interactions with CAND1 in yeast (Fig. 1C). A possible explanation for this difference is that Saccharomyces cerevisiae does not contain a CAND1 ortholog. Thus, there may be factors regulating CAND1–CUL1 interactions in higher eukaryotes that are absent from the yeast 2-hybrid system.

eta2-1 and axr6-2 Differentially Affect SCF^TIR1 Complex Abundance.

If CAND1 regulates cycles of SCF assembly and disassembly by binding to CUL1, this cycle should be disrupted in eta2-1 and axr6-2 mutants. To test this possibility, we examined the SCF^TIR1 complex in eta2-1 mutants by using a gel-filtration assay. A construct containing the TIR1 promoter driving expression of a TIR1-HA-Strep-tagged transgene was introduced into tir1-1 plants. This transgene fully complemented the auxin response defects conferred by the tir1-1 mutation (L.D.N., unpublished data). The transgene was subsequently crossed into the tir1-1 eta2-1 background, and α-HA Western blots confirmed equivalent levels of TIR1-HA-Strep expression (Fig. 2A). Protein extracts of tir1-1[TIR1-HA-Strep] and tir1-1 eta2-1[TIR1-HA-Strep] (subsequently referred to as “wild-type control” and “eta2-1,” respectively) were fractionated and blotted with antibodies against CAND1, CUL1, HA, and ASK1 (Arabidopsis SKP1).

Fig. 2.

Effects of the eta2-1 and axr6-2 mutations on CCR and SCF complex abundance. (A) α-HA Western blot of tir1-1[P_TIR1::TIR1-HA-Strep] and tir1-1 eta2-1[P_TIR1::TIR1-HA-Strep] seedling extracts. RPN6 is shown as a loading control. (B) Extracts from 7-day-old seedlings were fractionated on a Superdex-200 gel-filtration column and fractions (1–20) were collected and blotted with α-CAND1, α-CUL1, α-HA, and α-ASK1 antibodies. Molecular mass standards are labeled at the bottom. (C) CUL1-ASK1 2-hybrid assays. β-gal activities are expressed relative to wild type. Error bars indicate SD (n = 3). axr6-3 is shown as an additional negative control. (D) A 6×His fusion protein containing IAA7 domain II (amino acids 71–100) was incubated with and subsequently purified from 1.5 mg of 7-day-old seedling extracts. A derivative containing the P⁸⁷S mutation, which abolishes TIR1 binding, was used as a negative control. Pulldowns were immunoblotted with antisera against CUL1.

In wild-type extracts, the ≈135-kDa CAND1 protein was detected predominantly in fractions 10–14, corresponding to molecular masses of ≈100–350 kDa (Fig. 2B). The CAND1–CUL1–RBX1 (CCR) complex has a predicted molecular mass of ≈235 kDa. The major CUL1 peak cofractionated with the upper distribution of CAND1 (Fig. 2B). Consistent with our findings that the eta2-1 mutation prevents CUL1 binding, the mutant protein was only detected in the low-molecular mass fractions (Fig. 2B). Thus, the CCR complex is present predominantly in fractions 10 and 11.

Surprisingly, the CUL1 elution profile was unaltered by the eta2-1 mutation (Fig. 2B). However, SCF complexes have a predicted molecular mass very similar to that of the CCR complex. Because CUL1 assembles into the CCR and SCF complexes in a mutually exclusive fashion, it seemed possible that the disruption of the CCR complex by the eta2-1 mutation results in more CUL1 assembling into SCF complexes, which would be consistent with the predictions of the SCF cycling model. Consistently, we found that the eta2-1 mutation results in an enrichment of TIR1 in fractions corresponding to the molecular mass of the SCF^TIR1 complex. The majority of TIR1 from wild-type extracts eluted in low-molecular-mass fractions the size of monomeric TIR1 or a TIR1–ASK complex (Fig. 2B). In contrast, TIR1 was nearly equally distributed between these low-molecular-mass fractions and fractions 10–11 in eta2-1 extracts. The eta2-1 mutation also caused a consistent increase in the relative amount of ASK1 that coeluted with CUL1 and TIR1 in fractions 10 and 11 (Fig. 2B). Together these results indicate that loss of the CCR complex in the eta2-1 mutant results in an increased abundance of the SCF^TIR1 complex.

Compared to eta2-1, the axr6-2 mutation had opposing effects on CCR and SCF complex abundance. A clear shift in the CAND1 elution profile was consistently observed, with more CAND1 coeluting with CUL1 in fractions 10 and 11 containing the CCR complex (Fig. 2B). Additionally, consistent with previous coimmunoprecipitation experiments indicating that the axr6-2 mutation dramatically reduced interactions with ASK1 (19), we find axr6-2 causes a reduction in the relative abundance of ASK1 that coelutes with CUL1 in SCF fractions 10 and 11 (Fig. 2B). In a yeast 2-hybrid assay, however, axr6-2 caused only a moderate reduction in the ability to interact with ASK1 (Fig. 2C). Together these findings suggest that the axr6-2 mutation may shift the CUL1-binding equilibrium in favor of CAND1, thus reducing SCF complex assembly. To further test this possibility, we performed pull-down assays with plant extracts by using a domain II peptide from IAA7, an SCF^TIR1 substrate (8). Consistent with there being less assembled SCF^TIR1, the axr6-2 mutation resulted in a reduction in the amount of CUL1 protein pulled down by the domain II bait peptide. However, CUL1 binding was partially restored in eta2-1 axr6-2 double mutants (Fig. 2D).

Loss of CAND1 Suppresses axr6-2.

To test the possibility that the axr6-2 mutation prevents CAND1–CUL1 dissociation, we generated double mutants with eta2-1. eta2-1 suppressed the seedling-lethal phenotype of axr6-2 homozygotes (Fig. 3A). Additionally, although eta2-1 is recessive, its suppression of axr6-2 is dominant, indicating increased sensitivity to CAND1 dosage. Examination of auxin inhibition of root growth revealed that eta2-1 axr6-2 double mutants display only a weak auxin response defect comparable to eta2-1 single mutants (Fig. S1B). Like eta2-1, the cand1-1-null allele (18) also suppressed the lethality of axr6-2 in a dominant fashion (Fig. 3B), indicating that suppression is not allele-specific, but is simply due to loss of CAND1 function. Thus, in the absence of CAND1, the axr6-2 mutant protein must be capable of fulfilling most, if not all, CUL1 functions. Curiously, although cand1-1 was largely epistatic to axr6-2, the axr6-2 mutation partially suppressed the dwarf phenotype of eta2-1 (Fig. 3 A and B).

Fig. 3.

Genetic interaction between eta2 and axr6-2. (A and B) Adult phenotypes of single and double mutants. (Inset) Terminal seedling-lethal phenotype of CAND1⁺ axr6-2 homozygotes. (Scale bars: 5 cm.) (C) α-CAND1 and α-CUL1 protein Western blots of 6-day-old seedling extracts. CUL1-RUB is indicated with an asterisk.

In addition to suppressing the lethality of axr6-2, mutations in cand1 also suppress the effects of axr6-2 on the cullin protein. As previously reported (19), axr6-1 and axr6-2 cause a modest increase in the CUL1 protein level and a decrease in the ratio of RUB modified to unmodified CUL1, implying a defect in RUB modification. However, the loss of CAND1 restores both wild-type cullin levels and normal RUB modification to axr6-1 and axr6-2 plants (Fig. 3C).

Loss of CAND1 Enhances axr6-3.

Unlike the dominant CUL1 alleles described above, axr6-3 is a recessive, temperature-sensitive mutation. We have previously shown that this mutation prevents SCF assembly by abolishing ASK1 binding, but does not affect interactions with CAND1 (20). In sharp contrast to the suppression seen with axr6-2, eta2-1 axr6-3 double mutants displayed very dramatic phenotypes. All double mutant plants died when grown at 28°C (Fig. 4 A and B). When grown at the more permissive temperature of 18°C, a few double mutants survived to adulthood, but exhibited extreme dwarfism, delayed senescence, and complete sterility (Fig. 4 C and D). Auxin response assays of eta2-1 axr6-3 double mutants revealed that, although eta2-1 enhances the auxin response defect of axr6-3 at semipermissive temperature (data not shown), axr6-3 is largely epistatic to eta2-1 at higher temperatures (Fig. 4E). This finding is consistent with our previous finding that axr6-3 prevents SCF assembly at 28°C (20). Thus, the severe growth defects of eta2-1 axr6-3 double mutants may be due to the effects of eta2-1 on CUL3- and CUL4-based ubiquitin ligases (21, 22), rather than further impairment of CUL1-based SCF complexes.

Fig. 4.

Genetic interactions between eta2-1 and axr6-3. (A and B) Forty-two-day-old axr6-3 and axr6-3 eta2-1 plants grown at 28°C. The double mutant fails to develop further after forming two to three pairs of true leaves. (C and D) Adult phenotypes of axr6-3 and axr6-3 eta2-1 double mutants grown at 18°C. (E) Inhibition of root elongation by 2,4-D. Seedlings grown at 20°C were transferred to 1 μM 2,4-D medium and grown an additional 4 days at 28°C. Error bars indicate SD (n ≥ 12).

eta2-1 Enhances a Weak Mutation in the CSN.

The eta6 mutant was isolated from the same tir1-1 enhancer screen used to isolate eta2-1 (17). Map-based cloning revealed that eta6 is a recessive allele of CSN1/FUS6, which encodes a subunit of the CSN (23). This finding was confirmed by allelism tests with fus6-1 and by complementation with a CSN1 transgene, prompting us to rename our mutant csn1-10 (Fig. 5A). csn1-10 is a transition mutation resulting in a Ser→Asn substitution at amino acid 305 (Fig. S2A) This residue is highly conserved among eukaryotic CSN1 orthologs. However, this mutation is in the last base of exon 4 and also affects mRNA splicing. Our analysis of CSN1 transcripts from csn1-10 plants detected two mRNA species: one that was correctly spliced and a second produced by using an errant splice site 32 bases upstream, resulting in a short deletion followed by a frameshift that truncates the protein. Western analysis revealed a reduction in the level of full-length CSN1 protein, but we did not detect the truncated derivative (Fig. S2B). At this time, we do not know whether the csn1-10 phenotypes are the result of the missense mutation, the reduction in CSN1 levels due to the splicing error, or a combination of both.

Fig. 5.

Genetic interactions between eta2-1 and csn1-10. (A) Adult Col, csn1-10, and csn1-10[FLAG-CSN1] plants. (B) Auxin-resistant root growth of csn1-10 and csn1-10 tir1-1 seedlings. Seedlings were transferred to 50 nM 2,4-D and grown an additional 4 days. Error bars indicate SD (n ≥ 10). (C) Seven-day-old etiolated seedlings. (D) Sixty-day-old eta2-1 csn1-10 double mutants. (E) Twenty-one-day-old fus6-1 seedling. Seedlings do not develop further. (F) Root growth assay on medium containing the synthetic auxin 2,4-D. Error bars indicate SD (n ≥ 12). (G) α-CUL1 Western blot of 11-day-old seedling extracts. CUL1-RUB is indicated with an asterisk. (Scale bars: 1 mm.)

Previous screens for constitutive photomorphogenesis/de-etiolated/fusca (cop/det/fus) mutants have identified CSN1 and several other CSN subunit mutations (24). All of these mutants exhibit a strong cop⁻ phenotype when grown in the dark and are seedling-lethal. Unlike these previously characterized csn mutants, csn1-10 does not display a cop⁻ phenotype and is viable throughout development (Fig. 5 A and C). csn1-10 seedlings do, however, exhibit auxin-resistant root growth (Fig. 5B) and reduced SCF^TIR1 activity (Fig. S2C). These findings suggest that csn1-10 is a weak allele that must retain sufficient CSN function for viability and repression of photomorphogenesis.

Because the CSN has been proposed to act in conjunction with CAND1 to regulate cullins, we generated csn1-10 eta2-1 double mutants to examine genetic interactions. While cand1 single mutants exhibit only a weak cop⁻ phenotype (17, 18), eta2-1 csn1-10 double mutants display a much stronger de-etiolation defect (Fig. 5C). When grown in the light, double mutants produced several pairs of very small leaves, but never transitioned into reproductive development (Fig. 5D). This phenotype is similar to severe csn mutants, although csn-null plants generally produce only a few pairs of leaves before arresting (Fig. 5E). The csn1-10 mutation causes a modest increase in the auxin response defect conferred by eta2-1 (Fig. 5F).

The CSN regulates SCF function via its metalloisopeptidase activity that cleaves RUB from CUL1 (25). CUL1 Western blots revealed a clear accumulation of RUB-modified CUL1 in csn1-10 seedling extracts. However, this defect was not enhanced by eta2-1 (Fig. 5G). Likewise, gel-filtration analysis of CSN4 and CSN5 (Fig. S2D) revealed that the eta2-1 mutation does not affect the elution profile of these subunits, suggesting that CSN assembly and interactions are not dependent on CAND1.

Discussion

Arabidopsis cand1 mutants exhibit several auxin response defects resulting from a reduction in SCF^TIR1 activity (17, 18). Although genetic studies indicate that CAND1 positively regulates SCF^TIR1 function, biochemical studies on human CAND1 suggested that it negatively regulated SCF assembly by sequestration of the CUL1 subunit. In vivo, SCF complexes are likely to be highly dynamic because many FBPs compete for access to a common CUL1-RBX1 catalytic core. Thus, if CAND1 facilitates cycles of SCF assembly and disassembly, the loss of CAND1 could result in reduced SCF activity. The SCF cycling model proposed that active SCF complexes containing RUB/NEDD8-modified CUL1 recruit the CSN, which removes RUB from CUL1 (12). Because CAND1 specifically binds unmodified CUL1, RUB cleavage could enable CAND1 to bind the CUL1-RBX1 core and dissociate the SCF. Although remodification of CUL1 was initially proposed to free CUL1 from CAND1 and promote SCF reassembly, recent findings, including this study, indicate that RUB modification occurs after CUL1 has been released from CAND1 and has assembled an SCF complex (15, 16). Using the genetic tools provided by Arabidopsis CUL1, CAND1, and CSN mutants, we have tested several aspects of this model.

The eta2-1 and axr6-2 mutations differentially affect CAND1–CUL1 interactions and SCF^TIR1 homeostasis. eta2-1 (17) and axr6-2 (19) are both single amino acid missense mutations located in close proximity within the CUL1–CAND1 interface (16). We demonstrate that these two mutations have opposing effects on CAND1–CUL1 interactions. Whereas the CAND1–CUL1 interaction is largely abolished by the eta2-1 mutation, a dramatic increase in CAND1–CUL1 complex abundance is seen in axr6-2 mutants.

Biochemical studies with human CAND1 found that its binding to CUL1 is antagonistic to SKP1 binding, suggesting that CAND1 negatively regulates SCFs by sequestering CUL1 to prevent SCF complex assembly (13, 14). Our studies of eta2-1 and axr6-2 mutants support this negative role for CAND1 in regulating SCF assembly. Our gel-filtration assays reveal that the reduced CAND1–CUL1 interaction in eta2-1 mutants results in the accumulation of the SCF^TIR1 complex (Fig. 2B). This conclusion is further supported by our previous results of IAA7 pull-down assays, which detected an increase in the relative amount of CUL1 pulled down from eta2-1 extracts compared with wild-type controls (17). In contrast, axr6-2 has a stabilizing effect on the CAND1–CUL1 complex, leading to a reduction in SCF assembly, as we observed diminished abundance of ASK1 in gel-filtration fractions corresponding to SCF complexes in axr6-2 extracts. Furthermore, noticeably less CUL1 copurified with an IAA7 bait protein in pull-down assays with axr6-2 extracts (Fig. 2D). This reduction was dependent on CAND1 because the eta2-1 mutation partially restored the amount of CUL1 copurifying with IAA7. Despite these opposing effects on SCF^TIR1 assembly, both eta2-1 and axr6-2 confer reduced SCF^TIR1 activity as indicated by the stabilization of Aux/IAA protein substrates and reduced sensitivity to auxin (17, 19), strongly supporting the hypothesis that CAND1-mediated cycling is required for optimal SCF^TIR1 activity in vivo (Fig. 6).

Fig. 6.

CAND1-mediated cycling. An active SCF complex containing RUB-modified CUL1 is shown at the top. After CSN-mediated RUB cleavage, CAND1 promotes SCF disassembly by dissociating CUL1-RBX1 from SKP1-FBP. SCF disassembly is blocked in the eta2-1 mutant, resulting in SCF complex accumulation. Dissociation of CUL1-RBX1 from CAND1 reassembles the SCF. However, this process is blocked in axr6-2 mutants. Upon release from CAND1 and SCF reassembly, the CUL1 subunit is RUB-modified, activating the SCF for substrate ubiquitination.

The effects of the dominant-negative axr6-1 and axr6-2 mutations on CUL1 are very interesting. These two missense mutations change Phe¹¹¹ to Trp and Ile residues, respectively. Previous CUL1 coimmunoprecipitation experiments with plant extracts suggested that these mutations reduce interactions with the ASK1 SCF subunit (19). Such a defect does not explain the dominant nature of these mutations however, because axr6-3 also is defective in ASK1 binding, but is recessive (20). Our findings that axr6-1 and axr6-2 cause a striking increase in the relative amount of CUL1 associated with CAND1 provides an explanation because the stabilized axr6-2–CAND1 complex could disrupt SCF cycling in AXR6/axr6-2 heterozygotes, preventing wild-type CUL1 from reassembling into SCF complexes efficiently. Thus, the previously reported strong reduction in ASK1 binding activity may largely be due to the inability of the mutant axr6-2 protein to dissociate from CAND1. Our double-mutant studies strongly support this possibility. Whereas axr6-2 homozygous plants exhibit a seedling-lethal phenotype, double mutants with cand1 are viable and only exhibit a weak auxin response defect comparable to cand1 single mutants. This clearly demonstrates that the lethality conferred by axr6-2 is dependent on CAND1 and that, in the absence of CAND1, axr6-2 has minimal effects on SCF activity.

The suppression of the RUB modification defect of axr6-1 and axr6-2 also suggests that these cul1 mutations block CAND1–CUL1 dissociation. The finding that axr6-1 and axr6-2 mutations confer reduced RUB modification despite these lesions being near the N terminus of the protein while RUB modification occurs near the C terminus (4) has previously been difficult to explain. Although loss of CAND1 does not alter the RUB modification profile of wild-type CUL1 (17, 18), cand1 mutations restore RUB modification to axr6-1 and axr6-2. This finding provides genetic support that CAND1 inhibits RUB modification of CUL1, as suggested by recent structural and biochemical studies of the human CCR complex (15, 16). This work highlights the question of what regulates CAND1–CUL1 dissociation. In vitro studies indicate that the addition of SKP1-FBP to preassembled CAND1–CUL1 complexes can promote dissociation (15). However, there also is evidence that an additional, unknown CUL1-associated factor may play an important role because the effect of SKP1-FBP addition is noticeably stronger with fractionated CUL1 versus recombinant CUL1 (15, 16). Our findings that axr6-2 “traps” CUL1 in the CAND1 complex without dramatically affecting CUL1 interactions with either CAND1 or ASK1 in yeast 2-hybrid assays are consistent with this possibility. The finding that axr6-3 completely abolishes ASK1 binding, but does not increase CAND1–CUL1 abundance (20), also suggests that axr6-2 may specifically inhibit dissociation.

Genetic Interactions with the CSN.

CAND1 is thought to act in concert with the CSN to regulate SCF disassembly (12). Previously described csn subunit mutants exhibit strong de-etiolation and seedling-lethal phenotypes. We isolated a weak csn1 allele that is viable and remains etiolated when grown in the dark, thus providing a genetic tool for future studies investigating CSN-mediated regulation of signaling processes beyond the seedling stage. The csn1-10 mutation moderately enhances the weak de-etiolation and auxin-resistant phenotypes of cand1 mutants. However, overall growth and development are dramatically affected as csn1-10 cand1 mutants exhibit extreme dwarfism and fail to transition into the reproductive phase of development. The fact that csn-null mutants are lethal and csn1-10 enhances cand1 suggests that these mutants are not simply impaired in CAND1-mediated cycling of cullin-based ubiquitin-ligases. Although both CAND1 and the CSN are thought to promote the disassembly of cullin-based ligases, only csn mutants are impaired in the RUB deconjugation from cullins. Previous studies in other systems have demonstrated that this defect of csn mutants can result in the autoubiquitination of FBPs, resulting in reduced SCF activity (26). Reduced FBP abundance also has been reported in animal cells in which CAND1 expression has been knocked down using siRNA (14). However, we do not detect a reduction in TIR1 abundance (Fig. 2A) or stability (data not shown) in our cand1 mutants. A separate study found that the endogenous levels of the Arabidopsis COI1 FBP were not affected by the loss of CAND1, but levels of an overexpressed UFO transgenic protein were diminished (18). Thus, although the loss of cand1 may not dramatically alter FBP stability, these mutants may be hypersensitive to perturbed CSN function. CAND1 may minimize the autoubiquitination of FBPs in csn1-10 plants by facilitating SCF disassembly. In csn1-10 cand1 double mutants, however, this protection may be lost, resulting in increased FBP autoubiquitination and severe defects in growth and development. Future studies need to address this possibility.

Materials and Methods

Plant Materials and Growth Conditions.

All Arabidopsis lines used are in the Col-0 ecotype. Seedlings were grown under sterile conditions on ATS nutrient medium (27) under long-day lighting at 20°C. For coimmunoprecipitation experiments, seedlings were grown in liquid ATS medium on a shaker at 20°C. Adult plants were grown in soil under long-day conditions at 18°C or 20°C. For root growth assays, 4-day-old seedlings were transferred to ATS medium supplemented with 2,4-dichlorophenoxyacetic acid (2,4-D), and root growth was measured after an additional 4 days. The 35S::Flag-CSN1 transgene was provided by X. W. Deng (Yale University, New Haven, CT). To construct the P_TIR1::TIR1-HA-Strep transgene, a fragment containing 1.7 kb of upstream promoter sequence, the 5′-UTR, and the TIR1 coding sequence without the stop codon was amplified by PCR from Col-0 genomic DNA and cloned into the pExtag-HAStrep vector (28). The construct was subsequently transformed into tir1-1 mutant plants by Agrobacterium-mediated transformation.

Yeast 2-Hybrid Assays.

The full-length CAND1 ORF was cloned into pJG4/5 (Origene) to make the prey construct. The pEG202-CUL1 bait construct was a gift from X. W. Deng (18). The K⁶⁸²R mutation in the RUB conjugation site and the axr6-2 mutation were introduced into the CUL1 bait construct by using the QuikChange XL site-directed mutagenesis kit (Stratagene). These plasmids were cotransformed with the P_LexA::lacZ reporter pSH18–34 (31) into yeast strain EGY48. Transformed colonies were inoculated into liquid media lacking histidine, uracil, and tryptophan supplemented with 2% raffinose. Cells were subcultured in 5 ml of the same media supplemented with 2% galactose to induce the expression of the AD-CAND1 fusion. After 6–8 h growth, cells were harvested for liquid β-gal assays as described previously (29). For the CUL1-ASK1 assays, the constructs have been previously described (20). β-gal activities from the integrated P_GAL1::lacZ reporter in strain YPB2 were determined from cultures at 1.5 × 10⁷ cells/ml grown at 24°C.

Antibodies, Immunoprecipitations, and Pull-Down Assays.

CAND1, CUL1, and ASK1 antibodies have been described previously (17). HA monoclonal antibody was obtained from Covance Research Products. CSN4, CSN5, and RPN6 antibodies were purchased from Biomol. CAND1-CUL1 coimmunoprecipitations were performed as per Quint et al. (20) using extracts prepared from 5- to 7-day-old seedlings. The 6×His-IAA7-dII pull-down assays were performed as described previously (17).

Gel-Filtration Chromatography.

Seven-day-old seedlings were homogenized in extraction buffer containing 50 mM Tris·HCl (pH 7.5), 150 mM NaCl, 10 mM MgCl₂, 1 mM EDTA, 10% glycerol, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 1× Halt protease inhibitors (Pierce). Homogenates were centrifuged for 15 min at 4°C to remove debris. Supernatants were spun again for 10 min and then filtered through a 0.2-μm filter (Pall). Then 600 μg of total protein was fractionated through a Superdex 200 10/300 GL column (Amersham). After loading the sample, proteins were eluted in filtered and degassed extraction buffer at a flow rate of 0.2 ml/min; 0.5-ml fractions were collected after the 6-ml void volume was reached and concentrated with StrataClean Resin (Stratagene). The column was calibrated by using gel-filtration calibration kits (Amersham) as per the manufacturer's instructions. All procedures were carried out at 4°C.