Auxin responses in mutants of the Arabidopsis CONSTITUTIVE PHOTOMORPHOGENIC9 signalosome.

The CONSTITUTIVE PHOTOMORPHOGENIC9 (COP9) signalosome (CSN) is an evolutionarily conserved multiprotein complex that interacts with cullin-RING type E3 ubiquitin ligases (CRLs). CSN subunit 5 (CSN5), which, when incorporated into CSN, can deconjugate the NEDD8 modification from the cullin subunit of CRLs, is essential for CSN's role in controlling CRL activity. Whether the CSN5 monomer, which is maintained in csn mutants such as csn3 or csn4, has a functional role, remains to be established. We performed a comparative gene expression-profiling experiment with Arabidopsis (Arabidopsis thaliana) csn3, csn4, and csn5 mutants, and we show here that these mutants cannot be distinguished at the transcriptional level. Furthermore, we show that csn3 csn5 mutants are morphologically indistinguishable from csn3 or csn5 mutants. Taken together, these data suggest that the CSN5 monomer does not have a function that leads to transcriptional or morphological changes in the csn mutants. We further examined auxin responses in csn mutants. Whereas CSN had previously been shown to be required for the auxin response-regulatory E3 complexes, specifically SCF(TIR1), the csn mutant phenotype suggests that CSN is not essential for auxin responses. We present physiological and genetic data that indicate that auxin responses are indeed only partially impaired in csn mutants and that this is not the result of maternally contributed CSN. Finally, we discuss these findings in the context of the current understanding of the role of neddylation and CSN-mediated deneddylation for CRL activity.

The CONSTITUTIVE PHOTOMORPHOGENIC9 (COP9) signalosome (CSN) is an evolutionarily conserved regulator of development in higher eukaryotes Schwechheimer, 2004). CSN was originally identified through the biochemical characterization of the COP9 protein from Arabidopsis (Arabidopsis thaliana; Wei et al., 1994;Chamovitz et al., 1996). Characteristic phenotypes of these csn mutants are photomorphogenic growth in the dark and postgermination growth arrest (Kwok et al., 1996).
The characterization of CSN in Arabidopsis, mammals, and yeasts revealed that CSN physically interacts with cullin-RING type E3 ubiquitin ligases (CRLs), such as the cullin1-containing SKP1-cullin1-F-box (SCF) protein complexes Schwechheimer et al., 2001). The CSN interaction is required for E3 ubiquitin ligase activity and E3 ligase degradation substrates are stabilized in csn mutants (Schwechheimer et al., 2001). CSN may impact on CRL function at least in part by its ability to deconjugate the ubiquitin-related protein NEDD8 (or RUB1) from the cullin subunit of CRLs . The deneddylation activity resides within CSN subunit CSN5, which to date is the only CSN subunit with a known biochemical activity. CSN5 has an interesting feature in that it is present in all eukaryotes not only as a subunit of CSN but also as a monomer (Freilich et al., 1999;Mundt et al., 1999Mundt et al., , 2002Maytal-Kivity et al., 2002;Oron et al., 2002;Dohmann et al., 2005). Interestingly, although csn subunit mutants typically lack the residual CSN complex and frequently also the residual CSN subunits, the CSN5 monomer is always maintained in these mutants (with the obvious exception of csn5 mutants). Because csn mutants contain almost exclusively neddylated cullins (where the nonneddylated cullins may represent de novo synthesized cullins) and because monomeric CSN5 does not have an activity toward neddylated cullins in vitro, it was concluded that the CSN5 monomer is inactive with regard to cullin deneddylation Schwechheimer et al., 2001;Cope et al., 2002;Gusmaroli et al., 2004;Dohmann et al., 2005). It can, however, formally not be ruled out that the CSN5 monomer has deneddylation activity toward as yet unknown NEDD8 conjugates that are distinct from the cullins.
Genetic data suggest that neddylation, as well as deneddylation, are required for efficient E3 ligase activity (Schwechheimer et al., 2001(Schwechheimer et al., , 2002. How neddylation and deneddylation regulate CRL activity is currently still a matter of debate. In the non-or deneddylated state, cullins interact with CULLIN-ASSOCIATED AND NEDDYLATION-DISSOCIATED1 (CAND1); cullins released from CAND1 can engage in CRL formation and are subsequently neddylated, resulting in increased E3 ligase activity and in part increased affinity for E2 conjugating enzymes (Kawakami et al., 2001;Liu et al., 2002;Zheng et al., 2002;Oshikawa et al., 2003;Goldenberg et al., 2004;Bornstein et al., 2006;Chew and Hagen, 2007). Recent studies on the assembly of CRLs and cullin neddylation propose that neddylation is dependent on the presence of the degradation substrate, its binding to the substrate recognition unit of the CRL (e.g. an F-box protein and the SKP1 adaptor protein), and the subsequent formation of a substrate-loaded holo-CRL complex (Bornstein et al., 2006;Chew and Hagen, 2007). According to the model that can be derived from these studies, a specific CRL is formed in the presence of its substrate and subsequently neddylated to prevent dissociation of the substrate recognition subunits. In the absence of the degradation substrate (e.g. after its complete degradation), deneddylation can occur and enables the disassembly of the respective CRL. In that way, deneddylation may provide the CRL core complex subunits, namely, cullins and RBX1, for the formation of other CRLs with distinct substrate specificities. It is noteworthy, however, that this model predicts that neither neddylation nor deneddylation are required for the formation or the activity of the CRL per se. Previous studies had implicated cullin deneddylation in controlling the stability of CRL subunits (e.g. the levels of specific F-box proteins are reduced in human cell lines, yeasts, and fungi with reduced CSN activity; Wee et al., 2005;Cope and Deshaies, 2006). However, in view of the more recent hypothesis that deneddylation may allow disassembly of CRL complexes, and in view of the fact that dened-dylation is impaired in csn mutants, this CRL subunit instability may also be a consequence of the impairment of CRL complex formation and the destabilization of CRL subunits that cannot be assembled into CRL holocomplexes.
In plants, the interaction of CSN with CRL-type E3 ubiquitin ligases was first identified in studies that focused on the auxin-regulatory CRL SCF TIR1 (Schwechheimer et al., 2001). SCF TIR1 promotes the degradation of the AUXIN/INDOLE ACETIC ACID (AUX/ IAA) transcriptional repressors in response to the perception of auxin by the TRANSPORT INHIBITOR RESPONSE1 (TIR1) F-box protein subunit Tan et al., 2007). In transgenic lines where CSN5 function is reduced via a CSN5 antisense construct, the AUX/IAA protein from pea (Pisum sativum), PsIAA6 (introduced into Arabidopsis as a transgene), is partially stabilized, auxin-dependent gene expression is inefficient, and physiological auxin responses are impaired (Schwechheimer et al., 2001). These observations strongly suggest that the CSN-SCF TIR1 interaction is required for efficient SCF TIR1 function. CSN was subsequently also shown to interact with other CRLs from plants such as the jasmonate-signaling SCF COI1 , the floral development regulatory SCF UFO , as well as with several E3 ligases from nonplant organisms Wang et al., 2003;Schwechheimer, 2004). It is thought that CSN is required for the proper function of all CRL-type E3 ubiquitin ligases.
The extent to which CSN is required for SCF TIR1mediated auxin responses remained to be determined. In the context of auxin response, this point is of particular interest because one phenotype of mutants deficient in the function of the F-box protein TIR1 and three of its functional paralogs, the AUXIN-SIGNALING F-BOX PROTEINS1 to -3 (AFB1-3), is a lack of root formation, which is most likely due to a failure to specify the hypophysis during embryogenesis (Dharmasiri et al., 2005). For the same reasons, mutants or transgenic plants that carry a stabilizing mutation in the genes encoding either IAA12/BODENLOS or its functional paralog IAA13, both of which are candidate SCF TIR1/AFB1-3 degradation substrates, lack a root . In contrast, all csn mutants examined to date form an embryonic root despite the fact that CSN is required for proper SCF TIR1/AFB1-3 function. Thus, although the csn mutant phenotype is generally very severe, it is weak with regard to the root development phenotype. This weak phenotype could suggest that CSN is not essential for SCF TIR1/AFB1-3 function or that CSN-and SCF TIR1/AFB1-3 -independent auxin response pathways exist that mediate auxin responses. An alternative explanation could be that the loss of CSN in csn mutants is compensated by functional CSN that is carried over from the parental plant to the progeny. Such a maternal contribution of CSN was observed in Drosophila csn mutants (Oron et al., 2002;Oren-Giladi et al., 2008).
In this study, we describe the results of a comparative transcriptome analysis that serves to examine whether the presence or absence of the deneddylating enzyme CSN5 leads to apparent gene expression changes in csn mutants. Furthermore, we examine the contribution of the csn5 mutation to the phenotypes of the csn3 mutant, which is deficient in CSN function. Finally, we address whether auxin responses are impaired or blocked in the csn mutants, whether there is evidence for a maternal contribution of CSN function, and we discuss our findings in the context of understanding the role of cullin neddylation and CSNmediated deneddylation.

RESULTS AND DISCUSSION
Comparative Expression Profiling of csn3, csn4, and csn5ab Mutants Arabidopsis mutants deficient in any one of the eight CSN subunits have a constitutive photomorphogenic phenotype and arrest growth at the seedling stage (Kwok et al., 1996;Gusmaroli et al., 2007). The indistinguishable phenotype of these mutants suggests that the loss of a single CSN subunit impairs CSN function to the same extent. However, all csn mutants other than mutants deficient in the two genes encoding CSN5 in Arabidopsis (csn5a csn5b double mutants; designated csn5ab throughout this article) and including csn3 and csn4 mutants maintain CSN5 as a monomer (Serino et al., 1999;Peng et al., 2001;Gusmaroli et al., 2004;Dohmann et al., 2005). Although the analysis of cullin deneddylation in csn mutants shows that this CSN5 monomer is not sufficient for cullin deneddylation, it remains formally possible that the CSN5 monomer has a deneddylation activity toward proteins other than cullins. We reasoned that differences between csn3, csn4, and csn5ab mutants may be resolved in a comparative transcript-profiling experiment, and that these differences may then allow us to identify gene expression changes that can be attributed to CSN5 monomer function. To obtain a homogeneous set of csn mutants for such a transcriptprofiling experiment, we examined mutants for CSN3, CSN4, and CSN5 in the Arabidopsis ecotype Columbia (Col-0) from the SALK T-DNA insertion mutant collection ( Fig. 1A; Gusmaroli et al., 2004;Dohmann et al., 2005;Rubio et al., 2005). Like all previously identified CSN subunit loss-of-function alleles, these csn3, csn4, and csn5ab (csn5a-2 csn5b-1) mutants have a constitutive photomorphogenic phenotype, arrest growth at the seedling stage, and are deficient in the deneddylation of Arabidopsis cullins as shown by western blots with antibodies directed Figure 1. Characterization of the csn mutant alleles used in this study. A, Genomic organization of the Arabidopsis genes encoding CSN subunits CSN3, CSN4, CSN5A, and CSN5B. Exons are depicted as black boxes, introns as lines. The positions of the relevant T-DNA insertion mutants from the SALK mutant collection are indicated. B, Phenotypes of 7-d-old wild-type and csn mutant seedlings. Scale bar 5 2 mm. C, Immunoblots of wild-type and csn mutant protein extracts reveal the accumulation of NEDD8-conjugated cullins CUL1, CUL3A, and CUL4 in the csn mutants. Neddylated cullins are marked by asterisks (*). Immunoblots with antibodies directed against CSN4 and CSN5 indicate the absence of the respective protein in the csn4 and csn5 (the csn5a-2 csn5b-1 double mutant) mutants, respectively. D, Gel filtration analyses of the csn mutant alleles confirm the presence of the CSN5 monomer in the csn3 and csn4 mutant allele. csn3 and csn5 mutants maintain a CSN4-containing protein complex that is similar in size to the CSN complex detected in the wild type as well as a smaller CSN4-containing protein complex or a CSN4 monomer. against CULLIN1 (CUL1), CUL3A, and CUL4 ( Fig. 1, B and C).
Also in line with previous studies, we found that the CSN5 monomer, but not the CSN complex form of CSN5, is maintained in the csn3 as well as in the csn4 mutants; both CSN5 forms are undetectable in the csn5ab mutant ( Fig. 1, C and D). Based on these observations, we claimed previously that CSN is not destabilized in the Arabidopsis csn5ab mutants (Dohmann et al., 2005). This notion finds support in the fact that csn5 loss-of-function mutants from other organisms also maintain the residual CSN complex (Mundt et al., 2002;Oron et al., 2002). We are, however, also aware of the observations made by others that only CSN1, CSN3, and CSN4 cofractionate in this protein complex in Arabidopsis csn5ab mutants (Gusmaroli et al., 2007). This finding suggests that the CSN4-containing protein complex identified in the csn5ab mutants is not the residual CSN complex that lacks only CSN5, but corresponds to a distinct protein complex of similar size that may also contain CSN1 and CSN3. Because we do not have access to a more complete set of CSN subunit antibodies, we are at this stage unable to resolve these issues. An apparent change in the abundance of cullins (e.g. a decrease of CUL3 as reported in this other study; Gusmaroli et al., 2007) is not apparent in our experiments.
When we examined CSN4 in the csn3 and csn5ab mutants, we found CSN4 to be maintained in both mutants as the subunit of a complex that is similar in size to the wild-type CSN complex and also as an apparent smaller protein complex or monomer, which to our knowledge has not been discussed in previous publications (Fig. 1, C and D). Both of these complexes are absent in the csn4 loss-of-function mutant. In summary, we conclude that the csn3, csn4, and csn5ab mutants have a common morphology and common cullin deneddylation defects, and that they differ from the wild type and from each other with respect to their CSN4 and CSN5 subunit maintenance, most importantly with respect to the presence and absence of the CSN5 monomer.
To examine whether the three csn mutants differ from each other at the gene expression level, we examined a set of gene expression data that had been obtained from 7-d-old dark-and light-grown wildtype, csn3-1, csn4-1, and csn5ab mutant seedlings using Affymetrix ATH1 arrays (Dohmann et al., 2008). These microarray data were analyzed using the R statistical programming environment and the Bioconductor modules Limma and Affy (Ihaka andGentleman, 1996, 2004;Irizarry et al., 2002;Smyth, 2004). The Limma module was used to fit a linear model to the data for all pairwise comparisons of csn mutants versus wild-type controls, and the empirical Bayes method was used to reduce the genewise sample variance (Smyth, 2004). The P values were then adjusted for multiple testing with the Benjamini and Hochberg method to control for false positives (Benjamini and Hochberg, 1995). In addition, the empirical Bayes approach adjusts raw P values for multiple testing and generates a B statistic that may be used for ranking differentially expressed genes (Smyth, 2004). Using this analysis, we identified a total of 5,780 genes that are misexpressed in at least one csn mutant (Supplemental Table S1). Using K-means clustering, these misexpressed genes were then grouped into five clusters. Within each cluster and in both physiological conditions tested, the expression profile for the three csn mutants is highly similar when compared to the expression of these genes in the wild type (Fig. 2). Based on this transcriptome-profiling experiment, we conclude that the three csn mutants have nearly identical gene expression defects that may be the molecular basis of their morphologically indistinguishable phenotype. Hence, we would also conclude that the CSN5 monomer does not have a specific function that leads to transcriptional changes in the csn mutants.

Characterization of csn Mutants Lacking CSN and CSN5 Monomer Function
Because the csn5ab mutant lacks CSN5 in the CSN complex as well as in the monomeric form, and because the CSN5 monomer is maintained in csn3 mutants, the similarity in the gene expression profile of the three csn mutants supports the notion that the CSN5 monomer does not have an important role in regulating growth and developmental responses in Arabidopsis. To test this hypothesis genetically, we wanted to examine csn3 csn5ab mutants that lack CSN as well as the CSN5 monomer. To this end, we crossed heterozygous csn3 and csn5ab mutants and then identified a line homozygous for csn5b-1 and segregating for the csn3-1 and csn5a-2 mutations. Provided that the loss of the CSN5 monomer does not alter the phenotype of the csn3 mutant, and based on the fact that the CSN3 and CSN5A loci are not linked, we expected one of seven progeny seedlings with a csn mutant phenotype to represent a csn3 csn5ab mutant (Table I). In the case of a genetic interaction, we would expect novel phenotypes in the csn3 csn5ab mutants or the csn3 csn5ab mutants to be underrepresented among seedlings with a csn mutant phenotype. In line with the notion that the absence of CSN5 monomer function does not affect the csn3 mutant phenotype, we identified the expected number of csn3 csn5ab mutants by genotyping in a population of 54 mutants with a csn phenotype and we found these double mutants to be morphologically indistinguishable from the single mutants (Table I; Fig. 1B).

Auxin-Dependent Growth Processes Can Occur in csn Mutants
The interaction between CSN and CRLs was initially demonstrated in Arabidopsis using the auxin response regulatory SCF TIR1 as a candidate E3 ligase (Schwechheimer et al., 2001). These studies showed that CSN and SCF TIR1 interact and that CSN is required for proper AUX/IAA degradation and proper auxin responses. Because these studies were conducted with weak csn5 mutants, namely, CSN5A antisense lines, it was not possible to judge to which extent SCF TIR1mediated auxin responses depend on CSN function. Because csn mutants develop a root (e.g. Fig. 1B), but mutants deficient in the degradation of the AUX/IAA proteins IAA12/BDL and IAA13 fail to establish a root meristem, it may be hypothesized that CSN function is not essential for the activity of SCF TIR1 and its functionally redundant SCF AFB1-3 E3 ligases (Dharmasiri et al., 2005;Weijers et al., 2005). Alternatively, it may be that functional CSN complexes are transmitted as a maternal contribution to the csn mutant progeny and that this contribution compensates at least in part for the loss of CSN in the csn mutants. Maternal contribution of CSN has been observed in Drosophila and appears to be required for successful embryogenesis of Drosophila csn mutants (Oron et al., 2002;Oren-Giladi et al., 2008). We wanted to test whether developmental processes such as root formation can occur in csn mutants in the absence of CSN (and a hypothetical maternal contribution of CSN). To this end, we generated calli from hypocotyls of wild-type seedlings and Figure 2. Transcript profiling of csn3, csn4, and csn5ab mutants grown in the dark and in the light. K-means clustering of 5,780 genes (see Supplemental  Table S1) that are differentially expressed in at least one csn mutant reveals that the three csn mutants have highly similar gene expression patterns: cluster 1, 1,720 genes that are up-regulated in dark-and lightgrown csn mutants in comparison to the wild type; cluster 2, 621 genes that are down-regulated in lightgrown csn mutants and whose expression is reduced in dark-grown wild-type and csn mutant seedlings; cluster 3, 1,138 genes that are down-regulated in dark-grown csn mutants and whose expression is reduced in light-grown wild-type and mutant seedlings; cluster 4, 943 genes that are up-regulated in dark-grown csn mutants and whose expression is upregulated in light-grown wild-type and csn mutant seedlings; and cluster 5, 1,354 genes that are repressed in dark-and light-grown csn mutants in comparison to the wild type. csn5 corresponds to the csn5ab double mutant. The CSN3, CSN5A, and CSN5B loci are unlinked.
c Expected ratio of the specific genotypes among plants with a csn mutant phenotype under the assumption that there is no genetic interaction between the csn3 and csn5 mutations. d x 2 5 3.75, which is smaller than x 2 0.05 5 9.49; degrees of freedom 5 4. of csn3, csn4, and csn5ab mutants, and subsequently regenerated roots from these calli on auxin-containing medium (Fig. 3). This experiment showed that csn mutant cells can undergo cell divisions in these specific growth conditions and that CSN is not required for the auxin-dependent root differentiation process.

Auxin-Induced Gene Expression Is Impaired But Not Blocked in csn Mutants
The interaction between CSN and SCF TIR1 was shown to be required for the efficient degradation of the AUX/IAA transcriptional repressors and consequently for the efficient activation of auxin-induced genes. In line with this mode of function, we found that the basal expression of at least 37 genes (20%) from a set of 188 auxin-induced genes is reduced in all three csn mutants in the dark as well as in the light ( Fig. 4A; Supplemental Tables S3 and S4). To be able to judge to what extent auxin-induced gene expression is impaired in csn mutants, we introduced the auxin response reporter DR5:GUS, whose activation requires the auxin-and SCF TIR1/AFB1-3 -dependent degradation of AUX/IAA repressors, into the csn mutant backgrounds (Ulmasov et al., 1997). Whereas we found that the basal activity of DR5:GUS is not significantly altered between the csn mutants and the wild type (Fig. 4, B-E), we found its expression to be severely compromised in the csn mutants following auxin induction (Fig. 4, F-J). Next, we examined the effect of auxin on cell cycle activity and introduced the cell division marker CYCB1.1:GUS, which is induced in the wild type (e.g. in response to auxin-stimulated lateral root formation) into the csn mutant backgrounds (Ferreira et al., 1994). In response to auxin treatment, we observed increased CYCB1.1:GUS expression in the wild type and in the mutants; however, only wild-type cells, but not the csn mutants, induced lateral root formation in response to auxin treatment (Fig. 4, N-Q). Taken together, these experiments suggest that auxin responses are partially, but not fully, impaired in csn mutants (Fig. 3).

Critical Role for AUX/IAA Dosage in csn Mutants
We next tested genetically whether csn mutants are sensitized toward a disturbance of the auxin-response pathway. To this end, we introduced a previously published transgene for the expression of the AUX/ IAA repressor IAA13 under the control of the IAA13 promoter (IAA13:MYC) as well as the tir1-1 mutant into the csn mutants (Ruegger et al., 1998;Weijers et al., 2005). Whereas the expression of IAA13:MYC or that of its functional paralog IAA12/BODENLOS (IAA12: MYC) in the wild type does not alter the morphology of roots from transgenic plants expressing these constructs, the expression of the stabilized AUX/IAA variants (here we used iaa13:MYC) results in the misspecification of the hypophysis during embryogenesis and, consequently, in a lack of primary root formation (Fig. 5, A-D; Weijers et al., 2005). Similarly, mutants deficient in the function of the auxin receptor F-box protein TIR1 do not have a root morphology phenotype, but mutants lacking TIR1 as well as other functional TIR1 paralogs (AFB1-3) sometimes have root formation defects, which can be explained by a hypophysis mis-specification. For our experiment, we reasoned that the introduction of a wild-type IAA13: MYC transgene or the tir1-1 mutation may cause a root specification defect in the sensitized csn mutant background by bringing IAA13 (and IAA12 dosage in the case of tir1-1) above a critical threshold. Indeed, in agreement with the hypothesis that csn mutants are sensitized to increased AUX/IAA levels, a fraction of Figure 3. Callus formation and root differentiation from callus in csn mutants. A to D, Calli induced from wildtype and csn mutant hypocotyls 3 weeks after transfer to callus induction medium (CIM). Scale bar 5 5 mm. E to H, Root differentiation from wild-type and csn mutant calli 3 weeks after transfer to root induction medium (RIM). Scale bar 5 5 mm. I to L, Magnification of the roots shown in E to H. Scale bar 5 0.5 mm. csn5 corresponds to the csn5ab double mutant.
the csn mutants harboring the IAA13:MYC transgene or the tir1-1 mutation developed a significantly shorter root (class II csn mutants) or failed to form a primary root (class III csn mutants; Fig. 5). Thus, these genetic interactions support the notion that csn mutants are sensitized toward AUX/IAA dosage and suggest that SCF TIR1/AFB1-3 -mediated protein degradation is only partially impaired in these mutants.

CONCLUSION
In our studies, we show by global transcript profiling that the loss of the deneddylating CSN subunit CSN5 leads to the same transcriptional changes as observed in csn3 and csn4 mutants. csn3 and csn4 mutants differ from the csn5ab mutant in that they still contain a CSN5 monomer (Fig. 1). The functional relevance of the CSN5 monomer, which is maintained in csn mutants from all organisms studied to date, remained to be established. Our transcript-profiling experiment (Fig. 2) as well as a genetic interaction study where we examined the loss of CSN5 function on the csn3 mutant phenotype (Table I) show that the CSN5 monomer does not have a function that leads to phenotypes at the transcriptional or morphological level.
We have further examined the role of CSN in the context of auxin responses mediated by SCF TIR1 and its functionally paralogous SCF AFB1-3 complexes (Figs. 3-5).
In summary, our cell biological and genetic studies show that auxin responses are not fully impaired in csn mutants, suggesting that CSN function is not essential for the activity of these E3 ligases. In previous studies, we had used an auxin-inducible luciferase reporter construct to examine auxin responses in csn5ab mutants and had concluded that auxin responses are blocked in these mutants (Dohmann et al., 2005). Although luciferase is a highly dynamic reporter protein, it is also a comparatively unstable reporter protein and subtle auxin responses may therefore have escaped detection in our previous experiments (Calderon-Villalobos et al., 2006). Through the use of the more stable GUS reporter in the context of the DR5:GUS and CYCB1.1:GUS and in physiological experiments, we have now been able to provide evidence that auxin responses are partially functional in csn mutants. This raises the question about the role of neddylation and deneddylation for CRL E3 ligase activity. Two elegant biochemical studies have recently demonstrated that the binding of a degradation substrate to the CRL degradation substrate receptor subunit induces CAND1 dissociation and cullin neddylation ( Fig. 6; Bornstein et al., 2006;Chew and Hagen, 2007). In this context, neddylation may serve to prevent CAND1 reassociation with the complex and thereby stabilize E3 ligase complexes and their activity. In turn, CSN-mediated deneddylation may serve to induce the disassembly of CRLs and the reassociation of CAND1 in the absence of the degradation substrate (e.g. fol-lowing its complete degradation or in response to a signaling event; Fig. 6). It is important to note, however, that according to this model neither neddylation nor deneddylation are required for E3 ligase assembly and activity. Rather, neddylation stabilizes the E3 ligases by preventing CAND1 reassociation and deneddylation destabilizes these complexes by permitting CAND1 association (Fig. 6). In line with this model, the auxin responses observed in the csn mutants may originate from the preformed and neddylated SCF TIR1/AFB1-3 complexes that are present in the mutants. In response Figure 5. Genetic interaction of csn mutants with IAA13:MYC and tir1-1. A to G, Phenotypes of 7-d-old light-grown seedlings with the specified genotypes. Scale bar 5 2 mm. E, Representative phenotype of a csn mutant such as csn3, csn4, or csn5ab. In the subsequent phenotype classification, seedlings with this phenotype were counted as class I csn mutants. F and G, Representative phenotypes of csn mutants with an apparent partial (class II) or full (class III) impairment of primary root formation. H to P, Classification of plants with a csn mutant phenotype as shown in E to G, with the genotypes as specified in the figure. The total number of csn mutants analyzed (n 5 number of csn mutants analyzed) was set to 100% in this analysis. There are no seedlings with a csn mutant phenotype among the progeny of IAA13:MYC lines K or tir1-1 L mutants (.500 seedlings were examined). csn5 corresponds to the csn5ab double mutant.
to auxin, the impairment of auxin responses may then arise from a limitation of free cullin and RBX1 subunits that are required to form additional CRLs specific for auxin response. This model would also predict that the auxin-dependent formation of holo-CRLs is impaired in the csn mutants. In turn, it can at present not be ruled out that CSN-and SCF TIR1/AFB1-3 -independent auxin response pathways exist that are responsible for the residual auxin responses observed in csn mutants, as reported here, and in mutants deficient in the TIR1and TIR1-related F-box protein functions (Dharmasiri et al., 2005). Further studies are required to address this issue and to substantiate these findings at the biochemical level.

Biological Material
All experiments were carried out with Arabidopsis (Arabidopsis thaliana) ecotype Col-0. The following T-DNA insertion mutant alleles were identified and obtained from the Nottingham Arabidopsis Stock Centre: csn3-1 (Salk_000593), csn3-2 (Salk_106465), csn4-1 (Salk_043720), and csn4-2 (Salk_053839; Rubio et al., 2005). The csn3 alleles were genotyped using the oligonucleotides CSN3-ATG-FW and CSN3-RV to detect the wild-type gene, and LBb1 and CSN3-E9-RV to test for the presence of the T-DNA insertion. The csn4-1 mutation was genotyped using CSN4-FW1 and CSN4-RV1 to test for the presence of the wild-type gene, and LBb1 and CSN4-RV1 to test for the presence of the T-DNA. The csn4-2 mutation was genotyped using CSN4-FW2 and CSN4-RV2 to test for the presence of the wild-type gene and LBb1 and CSN4-RV2 to test for the presence of the T-DNA. The csn5ab mutations were described previously (Dohmann et al., 2005). The DR5:GUS reporter was a gift from Tom Guilfoyle (University of Columbia; Ulmasov et al., 1997), the CYCB1.1:GUS construct (Ferreira et al., 1994) was obtained from Lieven de Veylder (VIB) and transgenic lines expressing iaa13:MYC and IAA13:MYC  were obtained from Dolf Weijers, Eva Benkova, and Gerd Jü rgens (Tü bingen University). All primer sequences are listed in Supplemental Table S2.

Microarray Analysis
For microarray analyses, three biological replicate samples were harvested from 7-d-old dark-or light-grown csn3-1, csn4-1, and csn5ab (csn5a-2 csn5b-1) mutant or wild-type ecotype Col-0 seedlings. The material was processed and hybridized to Affymetrix ATH1 gene chips as described (Dohmann et al., 2008). The microarray data can be retrieved as experiment GSE9728 at GeneExpressionOmnibus (GEO). Microarray analyses were performed in the R statistical programming environment using the Bioconductor modules Limma and Affy (Ihaka and Gentleman, 1996;Irizarry et al., 2002;Gentleman et al., 2004;Smyth, 2004). Raw expression values from each probe set were read from the original CEL files through the Affy module, and then the robust multichip average algorithm was applied to the dataset for background adjustment, quantile normalization, and summarization using median polish (Irizarry et al., 2003). The Limma module was then used to fit a linear model to the data for all pairwise comparisons of mutant versus control microarrays, and the empirical Bayes method was used to reduce the genewise sample variance (Smyth, 2004). The P values from the hypothesis tests were adjusted for multiple testing with the Benjamini and Hochberg method to control for false positives (Benjamini and Hochberg, 1995). In addition, the empirical Bayes approach automatically adjusts raw P values for multiple testing and generates a B statistic that may also be used for ranking differentially expressed genes (Smyth, 2004). All genes with an adjusted P value ,0.05 from the csn experiments were identified (Supplemental Table S1) and then subjected to K-means clustering using default settings of the Genespring GX package. The auxininduced genes (Supplemental Table S3) were identified based on a subset of data from a published experiment deposited in GEO as GSE630, namely, GSM9571 to GSM9576, GSM9595 to GSM9600, and GSM9620 to GSM9628 (Okushima et al., 2005). The same data analysis methods as described above were applied to this dataset, and a false discovery rate ,0.001 was used for Figure 6. Model of the role of neddylation and deneddylation in the assembly and disassembly of CRL E3 ubiquitin ligases based on the data recently published in Bornstein et al. (2006). In this model, degradation substrate availability leads to CRL holocomplex formation and dissociation of CAND1. Cullin neddylation stabilizes the CRL E3 ligase and prevents CAND1 reassociation. Depletion of the degradation substrate will allow for CRL disassembly, cullin deneddylation, and CAND1 reassociation.
To generate the csn3-1 tir1-1 double mutant, tir1-1 (Ruegger et al., 1998) was crossed with a CSN3/csn3-1 heterozygous mutant. CSN3/csn3-1 heterozygous seedlings were identified in the F 1 progeny by genotyping for the csn3-1 mutation, and tir1-1/tir1-1 CSN3/csn3-1 mutant seedlings were identified in the F 2 generation based on csn3-1 genotyping and tir1-1 mutant auxininsensitive growth on 0.075 mM 2,4D. csn3-1 tir1-1 double homozygous seedlings were then analyzed in the F 3 generation. To generate csn IAA13:MYC lines, the IAA13:MYC transgene  was crossed into heterozygous CSN/csn mutant seedlings. To test for the presence of the transgene, plants were genotyped using MYC-FW and IAA13-RV. Lines heterozygous for the csn mutations and carrying IAA13:MYC were identified in the F 1 progeny, IAA13:MYC homozygous csn/CSN heterozygous plants were identified in the F 2 generation and phenotypically analyzed in the F 3 generation. Primer sequences are listed in Supplemental Table S2.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S1. List of genes misexpressed in specified csn mutants grown in the light (L) or the dark (D).
Supplemental Table S2. List of primers for genotyping of mutant and transgenic lines.
Supplemental Table S3. List of auxin-regulated genes.
Supplemental Table S4. List of auxin-regulated genes that are misexpressed in all three csn mutants.