The Arabidopsis COP9 Signalosome Subunit 7 Is a Model PCI Domain Protein with Subdomains Involved in COP9 Signalosome Assembly

Abstract

The COP9 Signalosome (CSN) is a multiprotein complex that was originally identified in Arabidopsis thaliana as a negative regulator of photomorphogenesis and subsequently shown to be a general eukaryotic regulator of developmental signaling. The CSN plays various roles, but it has been most often implicated in regulating protein degradation pathways. Six of eight CSN subunits bear a sequence motif called PCI. Here, we report studies of subunit 7 (CSN7) from Arabidopsis, which contains such a motif. Our in vitro and structural results, based on 1.5 Å crystallographic data, enable a definition of a PCI domain, built from helical bundle and winged helix subdomains. Using functional binding assays, we demonstrate that the PCI domain (residues 1 to 169) interacts with two other PCI proteins, CSN8 and CSN1. CSN7 interactions with CSN8 use both PCI subdomains. Furthermore, we show that a C-terminal tail outside of this PCI domain is responsible for association with the non-PCI subunit, CSN6. In vivo studies of transgenic plants revealed that the overexpressed CSN7 PCI domain does not assemble into the CSN, nor can it complement a null mutation of CSN7. However, a CSN7 clone that contains the PCI domain plus part of the CSN6 binding domain can complement the null mutation in terms of seedling viability and photomorphogenesis. These transgenic plants, though, are defective in adult growth, suggesting that the CSN7 C-terminal tail plays additional functional roles. Together, the findings have implications for CSN assembly and function, highlighting necessary interactions between subunits.

INTRODUCTION

The COP9 Signalosome (CSN) is a highly conserved protein complex found in all eukaryotes (reviewed in Wei and Deng, 2003). The CSN from most organisms purifies as a 450- to 550-kD complex that consists of eight subunits named CSN1 to CSN8, in order of decreasing size (Deng et al., 2000). With its discovery in Arabidopsis thaliana and purification from cauliflower, the complex was originally described as a regulator of light-dependent growth (Wei et al., 1994; Chamovitz et al., 1996). Subsequent studies employing multiple model systems indicated that CSN impinges on numerous signaling pathways involved in diverse cellular and developmental processes (reviewed in Cope and Deshaies, 2003; Serino and Deng, 2003; Harari-Steinberg and Chamovitz, 2004).

CSN is biochemically linked to ubiquitin-dependent protein degradation, and many of the effects of csn mutations can be explained by the role of the CSN in regulating multiple cullin-based E3 ubiquitin ligases (Azevedo et al., 2002; Liu et al., 2002; Schwechheimer et al., 2002; Feng et al., 2003). CSN possesses deneddylation activity toward the cullin subunit of E3 ligases, mediated by CSN5 (reviewed in Cope and Deshaies, 2003). CSN also suppresses cullin activity through recruitment of the deubiquitylating enzyme Ubp12p (Zhou et al., 2003). These contradictory activities have been reconciled in a model wherein CSN is a platform that stabilizes assembling SCF complexes (Wolf et al., 2003; He et al., 2005; Wee et al., 2005; Cope and Deshaies, 2006). Other data point to a scaffolding role for the CSN, mediating between kinases, ubiquitin ligases, the proteasome and their substrates (such as p53, cJun and p27^kip1, among many others) (reviewed in Harari-Steinberg and Chamovitz, 2004).

All CSN subunits share sequence homologies with subunits of the 26S proteasome lid subcomplex and the translation initiation complex eIF3 (reviewed in Kim et al., 2001). While these three protein complexes have independent biochemical functions, they intersect with each other at several levels. Various biochemical studies indicate interactions between subunits of the three complexes (Hoareau Alves et al., 2002), either direct complex–complex associations (Peng et al., 2003) or interactions between one complex and subunits of the other complexes (Kwok et al., 1999; Yahalom et al., 2001; Kim et al., 2004). Genetic studies further support a biological significance to these interactions (Yen et al., 2003; Kim et al., 2004).

At the level of overall complex organization, each of the three complexes contains two subunits that carry an MPN (for Mov34, Pad1 N-terminal) sequence motif and six subunits that bear a PCI (for Proteasome, COP9 signalosome, and Initiation factor 3) motif. These motifs, based on sequence sets, are found almost exclusively in subunits from these complexes and hint at conserved structural folds. Biochemical and structural studies suggest that certain versions of the MPN domain confer metalloprotease activity that cleaves Nedd8 or ubiquitin from proteins (Cope et al., 2002; Maytal-Kivity et al., 2002a). PCI is an ∼200–amino acid motif, not well conserved in its primary sequence, usually located near the C terminus of the protein (Aravind and Ponting, 1998; Hofmann and Bucher, 1998).

Various studies suggested that the PCI motif mediates and stabilizes protein–protein interactions within the complexes (Kapelari et al., 2000; Tsuge et al., 2001). For example, the predicted PCI region of Drosophila melanogaster CSN7 was sufficient for mediating the interaction with other CSN subunits in a yeast two-hybrid interaction assay (Freilich et al., 1999). In Saccharomyces cerevisiae, at least three subunits of the CSN-like complex that contain a PCI motif interact with each other, probably cooperating in complex formation (Maytal-Kivity et al., 2002b). However, despite bioinformatic and genetic studies, the PCI still defies rigorous biochemical description.

To define the structure of the PCI motif and to study its functional significance, we used Arabidopsis CSN7 as a model PCI protein. CSN7 is encoded by the FUSCA5 locus, and mutations in this gene lead to a constitutive photomorphogenic/deetiolated/fusca (cop/det/fus) phenotype (Karniol et al., 1999). CSN7 is detected in both the nucleus and the cytoplasm (Yahalom et al., 2001), as both a monomer and as an obligate subunit of the CSN. It contains 225 residues with a 26-kD molecular mass. Indeed, csn7^null mutants show no evidence for CSN complex formation (Gusmaroli et al., 2007). Furthermore, CSN7 from several organisms has been shown to interact with other CSN subunits (Freilich et al., 1999; Kapelari et al., 2000). We describe biochemical and structural studies that have defined the CSN7 PCI domain. In addition, we address how the PCI works in the context of the whole protein and how an additional structural element, a C-terminal tail, enables a heterotypic interaction with an MPN protein, CSN6. We further explored assembly of ternary complexes by in vitro reconstitution and directly investigated the biochemical functional correlations using transgenic plants.

RESULTS

Limited Proteolysis Identifies a Stable CSN7 Core

Previous bioinformatic analyses, motivated by investigation of the structure and function of the 19S proteasome regulatory lid, delimited a PCI sequence motif (Aravind and Ponting, 1998; Hofmann and Bucher, 1998). This motif is predicted to be highly helical with low primary sequence similarity, distributed unevenly. Thus, the PCI C-terminal half is more conserved, while its N-terminal half is much less so, with ill-defined N-terminal borders, making sequence and structural prediction of the N-terminal half challenging.

To obtain an experimentally defined domain, we overexpressed and purified to homogeneity full-length Arabidopsis CSN7 and subjected it to limited proteolysis (Figure 1). The proteolyses, performed using various proteases including the highly nonspecific papain (see Supplemental Figure 1 online), presented us with a resistant fragment. The protease Glu-C produced a particularly homogeneous product, but due to Glu-C's specificity, we chose to challenge this digestion product further with both amino and carboxypeptidases. While aminopeptidase did not further digest the Glu-C product, carboxypeptidase did succeed in trimming the Glu-C fragment. Both of these proteolytic products were then analyzed by mass spectrometry and N-terminal sequencing, enabling unambiguous definition of the molecules. The initial product produced by digestion with Glu-C spans residues 1 through 182 (CSN7¹⁸²), while the carboxypeptidase product spans residues 1 through 169 (CSN7¹⁶⁹). The biochemically defined domain includes the original PCI motif and an additional 40 residues. We call this the CSN7 core.

Figure 1.

Limited Proteolysis of Full-Length CSN7.

Full-length CSN7 was incubated with Glu-C (V8) protease at a ratio of 150:1 on ice. Aliquots were taken from the reaction at the indicated times, SDS sample buffer was added, and samples boiled and analyzed later by SDS-PAGE and Coomassie blue staining. The same conditions were used for the digestion reaction of the Glu-C digested CSN7 with carboxypeptidase Y. The digestion products are defined on the right.

Crystal Structure of the CSN7 Core

To obtain high-resolution structural information, we crystallized and determined the structure of CSN7 core (CSN7¹⁶⁹) to a resolution of 2.1 Å. Selenomethionine protein was prepared and used in a multiwavelength anomalous diffraction (MAD) experiment, whose details are described in Table 1. Additional diffraction measurements of a native protein crystal extended to a resolution of 1.5 Å. These data have been used for model refinement to R_free= 19.9% and include residues 4 through 164, and 244 water molecules, with excellent stereochemistry.

Table 1.

Crystallographic Data and Refinement Statistics

		Selenomethionine
Data Statistics	Native	λ1	λ2
Wavelength (Å)	0.934	0.97944	0.97967
Space group	C2221	C2221
Unit cell parameters (Å)
a	57.1	57.1	57.1
b	85.8	86.2	86.2
c	72.3	72.5	72.5
α = β = γ (°)	90	90	90
Unit-cell volume (Å3)	352,201.3	356,608.1	356,608.1
Resolution range (Å)	50–1.5 (1.55–1.5)	50–2.1 (2.2–2.1)	50–2.1 (2.2–2.1)
Total reflections	464,906	95,650	95,873
Unique reflections	28,844	20,019	20,118
Completeness (%)a	99.5 (97.1)	99.3 (100.0)	99.1 (99.1)
Rmerge (%)a,b	3.5 (26.9)	5.2 (15.4)	4.6 (13.0)
I/σa	47.9 (6.1)	23.57 (8.8)	27.7 (11.5)
f’/f’’		−7.8/4.8	−9.0/2.62
Phasing power (anomalous)		1.75	1.37
Figure of merit		0.46
Beamline (ESRF)c	ID14-1	BM16	BM16

Refinement statistics
No. of reflections (working/test)	27,494/1,186
dmin (Å)	1.5
Rwork/Rfree (%)	16.4/19.9
RMSD from idealityd
Bond lengths	0.01
Bond angles	1.54
B factors (Å2) (RMSD of bonded atoms-main/side chain)	2.0/3.4
Averaged B factor (Å2)	17.7
No. of protein atoms/solvent	1,268/244

^aValues of the highest resolution shell are given in parentheses.

^bR_merge = Σ_hklΣ_i |I_hkl,i − < I>_hkl| / Σ_hklΣ_i|I_hkl,i|, where I_hkl is the intensity of a reflection and <I>_hkl is the average of all observations of this reflection.

^cESRF, European Synchrotron Radiation Facility.

^dRMSD, root mean square deviation.

CSN7¹⁶⁹ is comprised of two subdomains that are intimately connected (Figures 2A and 2B), an N-terminal helical bundle (HB) subdomain and a C-terminal winged helix (WH) subdomain. Two pairs of antiparallel helices with pseudotranslational symmetry (helices 1/2 and 3/5) mark the center of the HB (Figure 2C). We consider the two short helices, 3 and 4, to be one longer helix to simplify our description. These pairs of helices are identified as HEAT repeats, structural motifs with a fundamental unit constituting a pair of antiparallel helices (Groves and Barford, 1999). Helices 3 to 5 may be identified as an ARM repeat, a highly related motif that has three helices. ARM repeats do not have a readily identifiable signature sequence but are known for assembling into stacks (Andrade et al., 2001).

Figure 2.

Structure of the CSN7 PCI Domain.

(A) and (B) Cartoon (A) and topology depictions (B) of the CSN7 PCI domain. The 1.5 Å resolution structure of CSN7's PCI domain reveals two subdomains: the HB subdomain (blue) and the WH subdomain (red). α, α′, and β denote α-helix, 3₁₀ helix, and β strand, respectively.

(C) A view down the helix axes of the two HEAT/ARM-like repeats in the HB subdomain. Green and orange electron densities indicate intra- and inter-repeat interactions, respectively.

(D) A magnified axial view of α7 shows how it serves as the interface between the two subdomains.

Each repeat is held together by intra-unit van der Waals interactions made by aliphatic residues (Figure 2C). Side chain pairs L14-L29 and A18-L26 anchor unit one, while L46-L66 and V51-L63 bind unit two. The two HEAT/ARM repeats assemble due to inter-unit van der Waals interactions, such as I11- I45, V15-L48, I30-L66, and L26-L54. Beyond these hydrophobic contacts, salt bridges, like R17-E32, give specificity to each antiparallel helix pair.

The WH subdomain includes helix 7, as discerned by structural alignment with a variety of previously reported WH domains. It contains the canonical helix-turn-helix, along with an electrostatic potential similar to many WH nucleic acid binding proteins (see Supplemental Figure 2 online). Helix 7 acts as the interface between the two subdomains, with both of its faces apolar, except its ends (Figure 2D). Consistent with this pivotal architectural role, helix 7 is highly conserved among CSN7 orthologs (Figure 3A) and is largely buried. Notably, the partially buried R160 mediates between the two subdomains, orienting α5 and α6 by making hydrogen bonds with the backbone carbonyls of F69 and G72 in the HB subdomain and the hydroxyl of T102 in the WH subdomain. Given these observations and the proteolysis data, we conclude that the HB and WH subdomains in the context of CSN7¹⁶⁹ cannot be autonomous folding units.

Figure 3.

Multiple Sequence Alignment and Secondary Structure of CSN7.

(A) Sequences of Arabidopsis CSN7, Danio rerio CSN7, Fugu rubripes CSN7, Anopheles gambiae CSN7, Drosophila melanogaster CSN7, Mus musculus CSN7b, Rattus norvegicus CSN7b, Pan troglodytes CSN7a, Homo sapiens CSN7a, and Homo sapiens CSN7b aligned. Cylinders, arrows, and waved ribbon represent helixes, β strands, and 3₁₀ helixes, respectively, with nomenclature and color scheme as in Figure 2B. Dark-blue residues are conserved, while light blue are variable. Relative solvent accessibility is represented by a black histogram.

(B) CSN7 sequence conservation. Scores account for evolutionary distance and are normalized to units of standard deviation. Zero represents the average evolutionary rate, and <0 indicates increasing conservation. The scores were then averaged in a window of ±9 residues (i.e., 19 total residues) for every residue and plotted as seen in the histogram. Two blocks of significant conservation show alignment with the two subdomains (red and blue bars) determined by the x-ray studies. The reference sequence for numbering is Arabidopsis CSN7.

(C) CD spectra of CSN7 core orthologs from Arabidopsis (At), Drosophila (Dm), and mouse (Mus). Each curve is labeled as per the legend. Deconvolution analysis calculates 72% helix, 5% strand, 23% turns, and coil for the Arabidopsis CSN7 core.

The CSN7 Core (CSN7¹⁶⁹) Is a Bona Fide PCI Domain

Amino acid identity shared among CSN7 orthologs is not high. For example, CSN7 from Arabidopsis and human share 34% amino acid identity (Figure 3A). However, closer analysis of sequence conservation using Conseq (Berezin et al., 2004) shows a good correlation between the structurally defined subdomains and amino acid conservation (Figure 3B). We then experimentally corroborated the sequence findings. The orthologous CSN7 proteins from Drosophila and mouse were expressed and purified. The Drosophila protein was subjected to limited proteolysis (see Supplemental Figure 3 online), while CSN7¹⁸² cores from the three species were studied by circular dichroism (CD) spectroscopy (Figure 3C). The spectra indicate a highly helical protein, consistent with the crystal structure. Both analyses revealed that the Drosophila and mouse CSN7 predicted PCI domains exhibit a close match to the Arabidopsis CSN7 core, strongly suggesting very high structural homology. Together with the comparison to eIF3k discussed below, we thus conclude that the CSN7 core solidifies a structural definition for the PCI domain.

CSN7 and CSN7¹⁶⁹ Interacts with CSN1 and CSN8

To date, the only known biochemical function for the PCI motif is to facilitate protein–protein interactions. To establish an in vitro binding assay, we used the purified recombinant CSN1 and CSN8, known to associate with CSN7 (Fu et al., 2001; Serino et al., 2003), in an analytical size-exclusion chromatography (SEC) assay for binding to full-length CSN7. As seen in Figure 4A, when CSN7 is first equilibrated with CSN8 and then run on an analytical size-exclusion column, the major peak elutes at a smaller volume than either protein alone, suggesting complex formation between CSN7 and CSN8. The same assay was used for CSN7 and CSN1 (Figure 4B) with a similar result. Notably, these interactions are specific since CSN8 and CSN1 do not form a complex (Figure 4C), consistent with the established subunit interaction network reported in the literature (Fu et al., 2001; Serino et al., 2003).

Figure 4.

PCI Subunits Associate in a Specific Manner.

(A) to (D) SEC elution profiles of CSN1, CSN7^full, CSN7¹⁶⁹, and CSN8 (alone or in a mixture) show a smaller elution volume for CSN7^full/CSN8, CSN7¹⁶⁹/CSN8, and CSN7^full/CSN1 mixtures indicating complex formation. In (C), an open arrow indicates the expected elution volume for a 1:1 CSN1-CSN8 complex.

(E) ITC data for a solution of 500 μM CSN7¹⁶⁹ titrated into 50 μM of CSN8. The top panel shows the raw data and the bottom panel the integrated data with the continuous lines representing the fit to a one-site binding model. The inset shows the equivalent experiment performed when the CSN8 was omitted from the solution (same scale). The calculated association constant was 1.4 × 10⁶ M⁻¹ (±76,000); n (stoichiometry) = 1.26 (±0.004); ΔH = −6460 ± 28.6 cal/mol; ΔS = 5.4 cal/mol/deg.

Having defined a bona fide PCI domain, we tested whether it indeed is responsible for interaction with other proteins. We therefore checked for interactions between CSN7¹⁶⁹ and CSN8 using the SEC assay. Figure 4D shows complex formation, as with full-length CSN7. The interaction was quantified using isothermal titration calorimetry (ITC), yielding a K_d of 0.71 ± 0.04 μM for the CSN7¹⁶⁹–CSN8 interaction (Figure 4E). ITC measurements of the interaction between CSN7¹⁶⁹ and CSN1 show a slighter lower affinity with a K_d of 2.2 ± 0.4 μM (see Supplemental Figure 4 online). Additional studies using yeast two-hybrid analysis show interaction of CSN7 with subunits CSN1, 4, and 8, all mediated by the PCI domain (i.e., CSN7¹⁶⁹) (Table 2).

Table 2.

Yeast Two-Hybrid Experiments

	CSN7 Clones in pEG202
CSN Subunits in pJG4-5	CSN7full	CSN7169	CSN7170-225
CSN1	+	+	−
CSN2	−	n.t.	n.t.
CSN3	−	n.t.	n.t.
CSN4	+	+	−
CSN5a	−	n.t.	n.t.
CSN6a	+	−	+
CSN7	−	n.t.	n.t.
CSN8	+	+	−

CSN7^full, CSN7wt; CSN7¹⁶⁹, CSN7 PCI domain; CSN7^170-225, CSN7 C-terminal tail. +, interaction; −, no interaction; n.t., not tested.

We then asked what region of the PCI domain provides the protein–protein interface. Mutations at five disparate positions on the surface of CSN7 were engineered: D12A, E44A, H71A, K144A, and E153A, the former three mapping to the HB subdomain, while the latter two map to the WH subdomain (Figure 5B). All mutant proteins were expressed, purified, and analyzed with CD spectroscopy to confirm structural integrity (see Supplemental Figure 5 online). These mutants were then tested for binding to CSN8 using our SEC assay (Figure 5A). Mutants D12A and E153A show no effect on binding, while E44A, H71A, and K144A are less effective in associating with CSN8. We also constructed the double mutant E44A,H71A, and this displayed markedly lower affinity for CSN8. We conclude that one face on the CSN7 HB subdomain and another from the WH subdomain form contact surfaces for CSN8. Using these same mutants and assay, we checked the interaction with CSN1. Association of CSN1 was not noticeably perturbed (see Supplemental Figure 6 online).

Figure 5.

Mutation Analysis Using the SEC Binding Assay.

(A) CSN7^E44A, CSN7^H71A, CSN7^K144A, and the double mutant CSN7^E44A,H71A showed weaker interaction to CSN8 as demonstrated by the presence of an unbound fraction that does not exist in the wild-type protein. Solid line indicates CSN7 mutant protein, dotted line indicates CSN8, and the dashed line indicates mixture. All experiments were performed with an analytical Superdex 200 column, except CSN7^K144A, which ran on an analytical Superdex 75 column.

(B) A ribbon/sticks representation of the positions of the mutated amino acids. Color scheme as in Figure 2.

Structure and Function of the CSN7 C-Terminal Tail: CSN6 Binding

The full-length CSN7 protein comprises a PCI domain, extending from residue 1 to 169, and a C-terminal tail, from residues 170 to 225. What might be the structure of the C-terminal tail? The proteolysis results presented earlier imply that it may have limited secondary and tertiary structure. Thus, we expressed the C-terminal tail segment as a glutathione S-transferase (GST) fusion protein, cleaved it from the GST, and further purified it to homogeneity. The CD spectrum strongly suggests that the CSN7 C-terminal tail is predominantly a random coil or disordered structure, although it may have small fractions of helical and extended conformation (Figure 6A). We also examined its conformation in the context of the full-length protein by calculating a difference CD spectrum of full-length CSN7 and CSN7¹⁶⁹ (Figure 6A). The results remain the same; the C-terminal tail is disordered. We further examined the C-terminal tail's hydrodynamic properties by analytical SEC. Using globular calibration standards, the C-terminal tail runs as a 12-kD protein, significantly larger than the expected 7 kD mass (see Supplemental Figure 7 online). We attribute the large apparent mass to the C-terminal tail's disorder, as unstructured proteins are known to have anomalously large Stokes radii (Martenson, 1978; Weinreb et al., 1996). Bioinformatic analysis of the C-terminal tail is equivocal, with one algorithm predicting some secondary structure and others predicting disorder.

Figure 6.

Structure-Function of the CSN7 C-Terminal Tail.

(A) CD analysis reveals an unordered structure for the CSN7 C-terminal tail (residues 170 to 225). The ordinate is delta epsilon units (liter·mol⁻¹·cm⁻¹). The solid line represents the CSN7^full spectrum, the dotted line represents CSN7¹⁶⁹, the dashed line is the calculated difference spectrum that represents the CSN7^tail whose calculation is described in Methods, and the alternating line represents the spectrum of a purified CSN7^tail.

(B) CSN7^full forms a complex with the MPN subunit, CSN6, as seen by the SEC elution profile of coexpressed CSN7^full-CSN6. CSN7^full is seen to coelute with CSN6 at a significantly smaller elution volume compared with monomeric CSN7^full. On the right, SDS-PAGE followed by Coomassie blue staining of the samples loaded onto the column is shown.

(C) Limited proteolysis performed with carboxypeptidase Y (75:1, total protein:protease ratio by mass) on CSN7-CSN6 complex versus CSN7 alone. Protection of CSN7 digestion is detected beginning at 4 h.

(D) Truncation analysis of CSN7 C-terminal tail and its effect on CSN6 association. Coomassie blue–stained SDS-PAGE gel of His-tagged CSN7 versions coexpressed with CSN6 and then precipitated by Ni-NTA beads. CSN6 clearly binds CSN7^full and CSN7²⁰². The CSN7 variants in this panel migrate more slowly compared with the other panels in this figure due to the addition of the His-tag (28 residues). CSN6 expression was confirmed for all experiments.

The structural attributes of the CSN7 C-terminal tail suggest that while alone it is largely disordered, it nevertheless may provide a target for protein interactions that would perforce induce structure upon binding. Using our defined C-terminal tail, we used yeast two-hybrid assays to detect interactions with different proteins previously identified as interacting with CSN7. Of the known CSN7-interacting proteins, CSN6 (an MPN domain protein) interacts specifically with the C-terminal tail of CSN7 (Table 2). The CSN7 C-terminal tail alone is both necessary and sufficient for this interaction as the CSN7¹⁶⁹ does not bind CSN6.

We confirmed that the C-terminal tail interaction with CSN6 is direct by coexpression and purification of a coeluting CSN7/CSN6 complex using SEC (Figure 6B) that also is consistent with a ratio of 1:1. Moreover, only coexpression of CSN6 with CSN7 facilitated its expression in Escherichia coli. No soluble CSN6 expression was obtained in the absence of CSN7 after repeated attempts. Comparison of carboxypeptidase digestions of CSN7 in the presence of CSN6 versus its absence indicates protection of CSN7 (Figure 6C), buttressing the notion that this interaction is mediated by CSN7's C-terminal tail. We then mapped the CSN7 C-terminal tail region required for CSN6 binding. A truncation of CSN7 to residue 202 still supports stoichiometric CSN6 association, while truncations to residues 182 or 169 reduce the amount of CSN6 bound dramatically (Figure 6D).

Our characterization of CSN6 association led us to investigate this subunit's behavior compared with CSN7. The above results suggest that CSN6 may be an obligate subunit with CSN7. We further tested this idea by subjecting the purified CSN7/CSN6 complex to stringent solution conditions to separate the two subunits using SEC. The complex is stable at 2 M NaCl and 2 M urea. Only at pH 10.0 do the subunits begin to separate. Thus, under physiological conditions, we conclude that CSN6 is only stable as a bound partner to CSN7.

Given the findings with CSN6, we proceeded to investigate whether we could assemble ternary complexes of CSN7/CSN6 with CSN1 or CSN8. Once again, using SEC, we show in Figures 7A and 7B that the respective ternary complexes are formed. We also asked if we could form a ternary complex between CSN7, 8, and 1. Upon incubating the three components, we were unable to obtain ternary complexes (Figure 7C), indicating that the CSN7 interacting surface with CSN1 and CSN8 partially overlap. The various permutations for CSN7 partner binding are summarized in Figure 7D.

Figure 7.

Ternary Complex Formation.

(A) SEC elution profiles of CSN1, CSN6, and CSN7 (alone or in a mixture) as indicated. SDS-PAGE gel stained by Coomassie blue is shown for the peak fraction as denoted by the arrow.

(B) SEC elution profiles of CSN6, CSN7, and CSN8 (alone or in a mixture) as indicated. SDS-PAGE gel stained by Coomassie blue is shown for the peak fraction as denoted by the arrow.

(C) CSN7, CSN8, and CSN1 do not form a ternary complex. SEC elution profile of CSN1, CSN7^full, and CSN8 (alone or in a mixture) shows no peak that corresponds with a ternary complex. The corresponding SDS-PAGE gel stained by Coomassie blue is shown below the chromatogram. Peaks labeled 1 to 3 are comprised of multiple fractions, shown on the gel. Each elution profile is labeled as per the legend. The outlined arrow indicates the expected elution volume of a CSN1-CSN7-CSN8 ternary complex.

(D) Schematic summary of CSN7 complex forms.

The CSN7 PCI Domain Does Not Incorporate into the CSN Complex in Planta

While the above studies indicated that CSN7 interacts with other PCI CSN subunits through its PCI domain, we wished to determine if the CSN7 PCI domain is sufficient for complex assembly and function in vivo in plants. Toward this end, transgenic plants were constructed that express various CSN7 transgenes under the control of the cauliflower mosaic virus 35S promoter. Three different transgenes were used: full-length CSN7 (35S:CSN7^full), CSN7-PCI as defined above (35S:CSN7¹⁶⁹), and CSN7-PCI plus an additional 13 amino acids (35S:CSN7¹⁸²). A HA-tag was cloned in frame C-terminal to each transgene to allow detection of the transgene product. Several independent transformants of each construct were chosen for analysis.

To determine if the transgene products can incorporate into the CSN, we used SEC to fractionate native protein extracts from transgenic seedlings. Wild-type (Columbia [Col-0]) extracts were analyzed as a control. Anti-HA antibodies recognize the transgene products, while anti-CSN7 recognizes both the endogenous and recombinant proteins. Lines were chosen whose transgene expression levels were comparable (see Supplemental Figure 8 online). As seen in Figure 8A, both endogenous and transgenic CSN7 are detected in all transgenic plants, while only the endogenous protein is seen in the wild type. In plants overexpressing the full-length protein (35S:CSN7^full), the transgene product is clearly detected in fractions corresponding to the native CSN complex, while extending to smaller complex and/or monomeric fractions. By contrast, CSN7¹⁶⁹ from two independent lines was detected almost exclusively in fractions corresponding to the monomeric form of CSN7¹⁶⁹ (fractions 16 to 21). CSN7¹⁸² from two independent lines showed an intermediate profile, with a majority of the transgene product detected in monomeric fractions, but with some detected also in fractions corresponding to the intact CSN. These results suggest that the CSN7 PCI domain is not sufficient for in vivo complex formation and that the C-terminal tail has a role in CSN7 incorporation into or assembly of the complex.

Figure 8.

CSN7 PCI Domain Does Not Incorporate into the CSN in Vivo.

(A) SEC analysis of crude protein extracts. Soluble protein (400 μg) from Col-0, 35S:CSN7^full, 35S:CSN7¹⁸², and 35S:CSN7¹⁶⁹ seedlings were separated over a Superose 6 column (GE Healthcare). Fractions (0.5 mL) were collected, concentrated, separated on 12.5% SDS-PAGE, and probed with anti-HA (top panels) and anti-CSN7 (bottom panels). The histograms to the right of the immunoblots show relative intensity of the CSN7 bands, quantified by densitometry, with the strongest band in each blot set to 100%.

(B) Coimmunoprecipitation analysis. Equal amount of total soluble protein from the indicated strains was incubated with anti-HA, and the bound proteins were separated on 12.5% SDS-PAGE. The blot was then probed with anti-CSN5. The input indicates the source of total extract.

To further confirm that the high molecular fractions do indeed represent an intact CSN and are not simply aggregation of overexpressed protein, coimmunoprecipitation analyses were done to identify in vivo interactions between the transgene products and endogenous CSN5 subunit, a CSN subunit that does not directly interact with CSN7 (Table 2) (Serino et al., 2003). The recombinant protein was immunoprecipitated with anti-HA, and the precipitate was probed with anti-CSN5. As seen in Figure 8B, endogenous CSN5 clearly immunoprecipitates with recombinant full-length CSN7, but not with CSN7¹⁶⁹. A small amount of CSN5 is identified in the immunoprecipitate from CSN7¹⁸² consistent with CSN7¹⁸² incorporating into the CSN at low efficiency.

Hypocotyl lengths of the transgenic plants were measured to determine if overexpression of the CSN7 transgenes influences an established CSN-dependent process, light inhibition of hypocotyl elongation. While both light-grown 35S:CSN7^full and 35S:CSN7¹⁶⁹ seedlings had longer hypocotyl lengths than did control Col-0 or 35S:green fluorescent protein (GFP) seedlings (Figure 9A), 35S:CSN7^full seedlings had significantly longer hypocotyl lengths than did 35S:CSN7¹⁶⁹ seedlings, even though we documented considerable levels of transgene expression in all lines (Figure 9B). No difference was detected between dark-grown control and transgenic seedlings (see Supplemental Figure 9 online). Thus, overexpressed full-length CSN7 is significantly more active than overexpressed CSN7¹⁶⁹.

Figure 9.

Overexpression of CSN7 Affects Hypocotyl Lengths.

(A) Two independent transformant lines from 35S:CSN7¹⁶⁹ and 35S:CSN7^full, together with Col-0 and 35S:GFP control seedlings, were grown in white light (150 μE) for 5 d, and hypocotyls measured. Both 35S:CSN7^full lines were significantly longer than both 35S:CSN7¹⁶⁹ lines (P < 0.001, student's t test) and the two control lines (P < 0.001).

(B) Immunoblot analysis of the transgene products, using anti-HA antibodies, showing transgene expression levels in each strain. The same membrane was reacted with anti-Actin antibodies (bottom panel) to control for sample loading.

The lack of integration of CSN7¹⁶⁹ into the CSN complex could be a result of competition with the endogenous CSN7. To determine if CSN¹⁶⁹ can integrate into the CSN in the absence of endogenous CSN7, the transgenes were checked for complementation of the csn7 null mutation in fus5 (csn7 null) mutants. The transgenic plants were crossed to Fus5⁺/fus5⁻ plants. Complementation, as reduction in number of fus seedlings, was monitored in T2 progeny of Fus5⁺/fus5⁻ transgenic T1 plants. Two different positive controls were employed, Fus5⁺/fus5⁻ plants (the trangene acceptor line) and Fus5⁺/fus5⁻ heterozygotes harboring a nonrelated β-glucuronidase (GUS) transgene to control for any effect of the transgene on segregation frequencies. As seen in Table 3, the number of fus progeny was <25% in both control lines (∼14%), probably owing to limited embryo lethality of some of the fus5 homozygotes. Complementation of the fus phenotype was observed for both CSN7^full and CSN7¹⁸² as fus frequency in progeny of Fus5⁺/fus5⁻, 35S:CSN7^full or Fus5⁺/fus5⁻, 35S:CSN7¹⁸² plants was only ∼4%. Analysis of two independent transgenic lines for each construct yielded essentially identical results. On the other hand, segregation frequencies for progeny of Fus5⁺/fus5⁻, 35S:CSN7¹⁶⁹ plants were not significantly different from the two controls. These results strongly indicate that the two transgene products that showed detectable integration into the CSN, CSN7^full and CSN7¹⁸², can complement a null mutation in endogenous CSN7, while CSN7¹⁶⁹ cannot complement the mutation. Thus, the CSN7 PCI is insufficient for biological activity in vivo.

Table 3.

35S:CSN7^full and 35S:CSN7¹⁸² but Not 35S:CSN7¹⁶⁹ Can Partially Rescue the fus5 Phenotype

Parental Strain	Total	Non fus	No. of fus	fus (%)
CSN7 Transgenic Lines in Fus5-1/fus5-1
Fus5-1/fus5-1	229	196	33	14.4%
full-7-8	373	357	16	4.3%
full-7-5	338	324	14	4.1%
182-4-14	355	339	16	4.5%
182-12-3	337	323	14	4.2%
169-4-10	350	301	49	14.0%
169-2-9	363	318	45	12.4%
TG control	292	249	43	14.7%

35S:CSN7^full, 35S:CSN7¹⁸², and 35S:CSN7¹⁶⁹ T2 pollen grains (Col-0) were used to fertilize Fus5⁺/fus5⁻ heterozygous plants (Landsberg erecta). F1 seedlings were first screened for kanamycin resistance and then for Fus5⁺/fus5⁻ heterozygosity by restriction fragment length polymorphism. F2 seeds were sown on Murashige and Skoog (MS) plates in the light and the number of fus seedlings counted. Two controls were used, Col-0 and TG (unrelated transgenic line expressing glucuronidase, p:GUS), crossed to Fus5⁺/fus5⁻, and were analyzed as described above. Each transgenic line is independent and was checked for its ability to express the exogenous protein (see Supplemental Figure 12 online). The CSN7^full and CSN7¹⁸² lines were significantly (Fisher's exact test, P <0.0001) different from the heterozygote and TG control lines and from the CSN7¹⁶⁹ lines. The CSN7¹⁶⁹ lines were not significantly different from the control lines nor were the CSN7^full lines different from the CSN7¹⁸² lines (Fisher's exact test, P > 0.1).

To further study the effect of the truncation of CSN7 in CSN7¹⁸², complemented seedlings were monitored through adult development. As seen in Figure 10A, fus5/fus5, 35S:CSN7^full plants are essentially indistinguishable from wild-type plants, indicating that the transgene can fully restore CSN function. fus5/fus5, 35S:CSN7¹⁸² plants on the other hand, were severely stunted in growth (Figure 10B). Many did not survive early seedling growth, and those that did showed a gradient of phenotypic severity. One line (Figure 10B-1) remained in the rosette stage and never bolted, while other lines (Figures 10B-2 and 10B-3) were severely dwarfed and partially sterile, dependent on the severity of the phenotypes. Sterility resulted from abnormal anthers (Figure 10C-3). No fus5/fus5 35S:CSN7¹⁶⁹ plants were isolated, consistent with the lack of complementation described above. Taken together, the SEC, coimmunoprecipitation, and complementation analyses indicate that the CSN7 PCI is not sufficient for in vivo complex assembly and suggests that the C-terminal tail plays an essential role in complex stability and/or assembly.

Figure 10.

Phenotypic Analysis of 35S:CSN7, fus5/fus5 Complemented Plants.

(A) Adults from five independent 35S:CSN7^full, fus5/fus5 lines 9 weeks after germination.

(B) 1 to 3: Three independent 35S:CSN7¹⁸², fus5/fus5 plants 8 weeks after germination. 4: Comparison between line 2 (arrow) and one of its 35S:CSN7¹⁸², fus5⁺/fus5⁻ siblings.

(C) Flowers (1), cauline leaves (2), stamens (3), and siliques (4) isolated from 35S:CSN7^full, fus5/fus5 (left) and 35S:CSN7¹⁸², fus5/fus5 line 2 (right) plants 6 weeks after germination.

DISCUSSION

The PCI motif was originally identified and defined by bioinformatic analysis based upon a shared motif with weak primary sequence conservation. Because of the weak amino acid homology, the exact definition of a PCI motif was variable. The biochemical and structural studies presented here now support a rigorous definition for the PCI that comprises two structural motifs intrinsically linked to build a complete and autonomous domain. The two subdomains, an N-terminal HB fold with two HEAT/ARM repeats and a C-terminal WH fold, are intimately connected through α7, and attempts to biochemically separate them were unsuccessful. Extrapolating to other PCI domains, we foresee a general architecture that includes these two elements with the HB subdomain varying between the different PCI family proteins. The structural variation will center on the different number of HEAT/ARM repeats present, as described below.

Several lines of evidence strongly indicate that CSN7 can serve as a model PCI protein and that the structure described here will be instructive for other PCI proteins. First, similar CSN7 cores were defined by limited proteolysis on proteins from diverse organisms, suggesting a conserved overall structure. Second, CD spectra of the CSN7 PCI domain from these organisms were highly similar. Third, the structure of the CSN7 PCI domain is similar to eIF3k.

eIF3, as does the CSN, contains parallel PCI and MPN proteins. The structure of eIF3k, a subunit of the eIF3 complex not originally identified as a PCI-containing protein, has been described (Wei et al., 2004). The eIF3k structure contains both HB and WH subdomains. We calculated superpositions between CSN7¹⁶⁹ and eIF3k using both subdomains individually and the complete molecule (see Supplemental Figure 10 online). The orientation of the two subdomains is somewhat changed between the two molecules, but the topology and interface is conserved. eIF3k bears an additional HEAT repeat in its HB subdomain and has a helix after its WH subdomain that folds back onto the HB subdomain. Structure-based sequence alignments using the above superposition templates indicate no sequence conservation between the two proteins (sequence identity <6%), despite their statistically significant structural homology (P < 2.0 × 10⁻⁶). These comparisons underscore the biochemical PCI domain definition where HB and WH subdomains are intrinsically linked. Such a structural template may be a more effective paradigm in the search for other PCI domains.

What can we infer from the structural findings that provides insight into the functional significance of the PCI domain? Previous studies, primarily based on yeast two-hybrid assays, intimated an essential role in PCI complex assembly, where different PCI proteins interact through their PCI domains. We confirmed this hypothesis in vitro by showing that the CSN7 PCI domain interacts specifically with either CSN1 or CSN8. CSN1 and CSN8 themselves do not interact, confirming ours and others' two-hybrid studies showing PCI specificity. These results are not surprising since both the HB and WH subdomains enable protein–protein interactions.

These two structural motifs, employed elsewhere in different contexts, function also as modules for nucleic acid binding. At least two observations are consistent with the possibility that nucleic acid binding may be a molecular function for PCI domains in general, CSN7 as a monomer, and the CSN. First, CSN7 (Yahalom et al., 2001) and the CSN (Chamovitz et al., 1996) are found predominantly in the nucleus. Second, three studies have clearly demonstrated that CSN PCI proteins associate with chromatin (Groisman et al., 2003; Menon et al., 2007; Ullah et al., 2007). The postulated nucleic acid binding could include DNA or RNA, as the WH binds to both classes. To test the structural plausibility of this hypothesis, we docked the CSN PCI domain with duplex DNA using the RFX-1-DNA complex as a template (see Supplemental Figure 11A online). The CSN7 WH fits nicely with DNA in the model with structural elements, such as basic side chains well positioned to interact with the phosphate backbone. Notably, the PCI domain has structural homology with linker histones, specifically H5, a molecule that employs a WH architecture. Alternatively, we have docked the PCI domain with a folded RNA molecule using as a template the SelB-RNA complex (see Supplemental Figures 11B and 11C online). This proposed biochemical activity (i.e., nucleic acid binding) could partially explain CSN's role in transcriptional regulation (Ma et al., 2003; Oron et al., 2007). We do note that the CSN7 PCI domain, despite maintaining basic residues appropriate for nucleic acid interaction, does not show the same very strong electropositive surface potential as found for proteins used as our templates. Hence, the potentially low relative affinity for nucleic acid in one PCI domain may be dramatically enhanced by the multiple WH domains extant in the CSN.

The CSN7 PCI domain is responsible for binding to other PCI proteins. Point mutants in both the HB and WH subdomains described here affect this interaction. Therefore, our current estimation is that both subdomains will be required for the relatively high affinity interaction between PCI domains. Other structural elements are not required. This observation led us to prepare transgenic plants that bear just the CSN7 PCI domain to test if indeed the PCI domain interactions are sufficient for full function. However, despite the strong in vitro binding of CSN7 PCI domain to CSN1 and CSN8, it is not sufficient to efficiently mediate CSN complex assembly in vivo in plants. CSN7¹⁶⁹ did not efficiently incorporate into the CSN in otherwise wild-type plants, nor could it complement the fus5 mutation, suggesting that the C-terminal tail of CSN7 is essential for complex assembly. A principal reason for this could be that the CSN7 C-terminal tail directly interacts with CSN6.

The CSN7 C-terminal tail, while lacking considerable intrinsic structure, does serve as the moiety for binding at least CSN6, an MPN motif subunit. Moreover, the C-terminal tail is essential for incorporation of CSN7 into the CSN complex as determined from the in planta studies. We note that CSN7¹⁸², which contains some of the C-terminal tail sequence, particularly its more conserved end, has significant residual ability in incorporating into the full complex that enables complementation of the fus phenotype of the csn7/fus5 mutants. Thus, we speculate that the disordered tail folds onto its targets when interacting, a mechanism widely applicable in the proteome. Additionally, this characteristic might facilitate more efficient assembly of CSN and/or CSN targets/substrates in the way of a fishing rod reeling mechanism (Sigler, 1988; Shoemaker et al., 2000). Finally, other PCI domain proteins may employ this type of functional architecture whereby the PCI domain interacts with other PCI domains (homotypic) and N or C termini tails interact with the MPN proteins.

However, the rest of CSN7's tail is necessary for normal function as the CSN7¹⁸² complemented plants displayed numerous phenotypes consistent with earlier reports of partial loss-of-CSN function plants. For example, plants with csn6 or csn5 downregulated using antisense strategies have similar adult phenotypes, as do plants missing one of the two copies of CSN5 in the Arabidopsis genome (Peng et al., 2001; Schwechheimer et al., 2001; Gusmaroli et al., 2007). These phenotypes have been attributed primarily to the effect of the CSN on SCF^Tir1 in mediating auxin signaling (Schwechheimer et al., 2001). As CSN7 does not directly bind to CSN5, and as CSN7¹⁸² incorporates inefficiently into the CSN complex, we predict that the phenotypes seen in the CSN7¹⁸² complemented plants arise from an overall reduction in functional CSN levels in these plants.

Surprisingly, a partial PCI domain is sufficient for mediating CSN1 incorporation into the CSN complex (Tsuge et al., 2001; Wang et al., 2002). A CSN1 truncation containing only the predicted PCI domain, but missing the N-terminal half of the protein, could integrate into the CSN complex, but only partially rescue CSN activity in a csn1/fus6 mutant. Thus, Wei and colleagues proposed that the CSN1 N-terminal half harbors the domain conferring most of the CSN1 transcription repression function, while the PCI mediates complex assembly (Tsuge et al., 2001). The N-terminal extension of CSN1 may be analogous to the CSN7 C-terminal tail.

Our dissection of CSN7 provides a glimpse into how the COP9 signalosome may be assembled, based on in vitro reconstitution data. We conclude that CSN7 is paired with CSN6 where this interaction is mediated by the CSN7 C-terminal tail, leaving the CSN7 PCI domain to interact with other PCI subunits. Moreover, this pairing is essential for CSN7 incorporation into the CSN in vivo, as evidenced by the differential incorporation of CSN7 truncations, which was generally correlated with the behavior of the truncations in vitro as well. We successfully assembled ternary complexes involving CSN7, CSN6, and, alternatively, CSN1 or CSN8. Detection of a quaternary complex was beyond the resolution of our SEC assay. Nonetheless, while we cannot formally exclude its formation, our inability to assemble ternary complexes of CSN7 with CSN1 and CSN8, coupled with the knowledge that CSN6 binds a domain that does not mediate PCI interactions, leads us to think that such a complex may not form. This possibility implies that either there is more than one copy of CSN7 in the complex, each copy affording direct interactions with CSN1 and CSN8, or that the intact CSN assembles with CSN7 interacting with either CSN1 or CSN8 exclusively. That CSN7 interacts in vitro with both may hint at the formation of CSN7 alternative complexes with various CSN partners. Indeed, complexes containing only a subset of CSN subunits have been observed, although their complete composition is not known (Fukumoto et al., 2005). In vivo, no fus5/fus5 35S:CSN7¹⁶⁹ plants were isolated, consistent with the lack of complementation described above. The results with the three analyzed CSN7 forms are summarized in Figure 11 and lead us to conclude that the CSN7 PCI domain is not sufficient for in vivo complex assembly with the CSN7 C-terminal tail playing an essential role for function.

Figure 11.

Summary of the Functions for the Different CSN7 Forms, CSN7¹⁶⁹, CSN7¹⁸², and CSN7^full, Analyzed in Transgenic Plants.

The phenotypic data in the first three columns relate to the phenotypes of fus5 plants expressing the indicated transgenes, while the last column summarizes the SEC data of wild-type plants expressing the indicated transgenes. CSN7¹⁸² and CSN7^full seedlings undergo normal skotomorphogenesis and photomorphogenesis in the dark and light, respectively, while CSN7¹⁶⁹ plants maintain a photomorphogenic phenotype in both dark and light conditions.

In conclusion, we provide a comprehensive structural and functional examination of a PCI domain protein. The results permit us to discern a common structure-function thread that unifies the disparate multiprotein complexes in which PCI domains are found. That thread fuses two structural motifs previously known to impart protein–protein and protein–nucleic acid interaction function. PCI domain complexes, acting in regulated protein degradation (proteasome and CSN) and protein synthesis (translation factor eIF3), will likely use such functionality.

METHODS

Subcloning, Expression, and Purification

The following versions of Arabidopsis thaliana CSN7 were subcloned into pET-28a (Novagen): CSN1, CSN7 ^full, CSN8, CSN7¹⁸², CSN7¹⁶⁹, and CSN7^mut, where mut indicates specific point mutants. Point mutants were generated by overlapping PCR. CSN7 from Drosophila melanogaster (Freilich et al., 1999) and mouse were also cloned into pET-28a. In addition, CSN7^full and CSN6a were subcloned into coexpression vector pET-Duet (Novagen) where CSN7 ^full had a 8xHis tag fused to the N terminus. PCR was used to engineer a TEV protease cleavage site as described (Dessau et al., 2006). Digested PCR product was ligated into doubly digested pET-28a or pET-Duet vectors (EcoRI/NotI for the CSN7 constructs; BamHI/NotI for CSN1 and CSN8; NdeI/KpnI for CSN6). In the CSN7 C-terminal tail truncation analysis, C-terminal truncations of CSN7 (CSN7¹⁶⁹, CSN7¹⁸², and CSN7²⁰²) and CSN7^full were subcloned into a pET-Duet vector in multiple cloning site I using BamHI and NotI, thereby creating an N-terminal His-tag for all the CSN7 versions. CSN6 was inserted into multiple cloning site II using NdeI and KpnI. All primers are listed in Supplemental Table 1 online. Subsequently, positive clones were identified by colony PCR and, where appropriate, sequenced.

Proteins were expressed in Escherichia coli Tuner strain (Novagen), bearing the RIL Codon Plus plasmid (Stratagene). Transformed bacteria were used to inoculate sequentially increasing volumes of Luria-Bertani (LB) growth media containing 50 μg/mL kanamycin and 34 μg/mL chloramphenicol at 37°C. The cells were grown for 2 to 3 h in a final volume of 4 L. Upon reaching OD₆₀₀ = 0.3, the temperature was lowered to 16°C. At culture density of OD₆₀₀ = 0.5 to 0.6, protein expression was induced by 200 μM isopropylthio-β-galactoside. Cells were harvested after 12 to 16 h by centrifugation (9700g) for 10 min at 4°C. The bacterial pellet was stored at −80°C.

Cells were suspended at 1:10 (w/v) ratio with lysis buffer, buffer L (50 mM sodium phosphate, pH 8.0, 300 mM NaCl, 0.1% Triton X-100, 10 mg lysozyme, 1500 units DNase, and 5% glycerol). Immediately after lysis of the bacterial cell suspension by microfluidizer (Microfluidics), phenylmethylsulfonyl fluoride (PMSF) was added to the lysate (0.5 mM). Cell debris were removed by 1 h centrifugation (20,000g) at 4°C, and the soluble fraction was loaded onto a preequilibrated metal chelate column (Ni-CAM; Sigma-Aldrich) in buffer A (50 mM sodium phosphate, pH 8.0, 300 mM NaCl, and 5% glycerol) at a flow rate of 1 mL/min. The column was washed with buffer A, containing 5 mM imidazole, until a stable base line was achieved. After a step elution by buffer A supplemented with 50 to 125 mM imidazole, the fractions containing the CSN subunit were pooled and subjected to TEV protease, prepared as described (Opatowsky et al., 2003). An additional step of a hydroxylapatite column was performed for CSN1 before TEV protease digestion. CSN1 containing fractions were pooled and diluted five times with distilled water and were loaded onto a preequilibrated hydroxyapatite column (10 mM sodium phosphate buffer, 150 mM NaCl, and 5% glycerol). Protein was eluted using a shallow gradient (100 mL) of sodium phosphate (10 to 200 mM). TEV proteolysis of all proteins was performed for 12 to 14 h at 19°C while dialyzing against 4 liters of buffer A containing no imidazole. Optimal digestions used a protein-to-protease ratio of 50:1 (w/w). Subsequently, the sample was loaded again onto a Ni-CAM column with unbound fractions collected and pooled. For CSN1 and CSN8, 5 mM imidazole was needed to elute the digested protein due to nonspecific binding of these proteins to the column. All proteins were concentrated using spin concentrators (Vivascience). SEC was then performed using a preequilibrated Superdex200 HiLoad prep grade column (GE Healthcare). Protein was eluted with buffer G (20 mM Tris, pH 7.1, 100 mM NaCl, and 5 mM β-mercaptoethanol [β-ME]). Pooled fractions were concentrated to various concentrations between 15 mg/mL (CSN1) and 50 mg/mL (CSN7^full), divided into aliquots, and flash frozen in liquid N₂.

CSN7 C-Terminal Tail Preparation

CSN7's C-terminal tail (residues 170 to 225) was subcloned into pGEX-4T (GE Healthcare) containing a TEV recognition site upstream of the MCS. It was than expressed as a GST fusion protein in E. coli Tuner strain as described. The GST fusion protein was purified using standard affinity chromatography with glutathione beads (GE Healthcare) and then was cleaved with TEV protease. Digested sample was loaded onto a SEC column (analytical Superdex 75; GE Healthcare) preequilibrated with buffer G. Fractions containing CSN7's tail were pooled and dialyzed against 5 mM phosphate buffer, pH 7.5.

Selenomethionine Protein Expression and Purification

Selenomethionine-substituted CSN7¹⁶⁹ was prepared and crystallized for MAD. Protein was produced in E. coli Tuner (DE3) as described, by inhibition of the Met synthesis pathway (Van Duyne et al., 1993). An overnight starter culture was grown from a single transformed colony in 10% LB medium. LB medium was removed prior to introduction to 2 liters of New Minimal Media (Budisa et al., 1995), fortified with Kao and Michayluk vitamin solution (Sigma-Aldrich), 50 μg/mL kanamycin, and 34 g/mL chloramphenicol. Cells were grown at 37°C to OD₆₀₀ = 0.25, whereupon the temperature was lowered to 16°C. Lys, Phe, and Thr (100 mg/mL), Ile, Leu, and Val (50 mg/mL), and dl-selenomethionine (50 mg/mL) were added 30 min prior to induction (OD₆₀₀ = 0.6). Expression was induced by 200 μM isopropylthio-β-galactoside over a 12- to 14-h period. Purification of protein was as before, except that 5 mM β-ME was added to all solutions to prevent selenium oxidation.

Limited Proteolysis of CSN7

Glu-C protease (1 mg/mL) (Roche Applied Science) was added to 2.5 mg/mL of CSN7 protein 1:150 (w/w) protease to protein. For proteolysis with papain (Sigma-Aldrich), the ratio was 1:1500. CSN7¹⁸² was digested with carboxypeptidase Y (1 mg/mL) (Worthington) at a ratio of 1:150 (w/w) protease to protein. Reactions were performed on ice. Proteolysis at different time intervals was monitored by SDS-PAGE analysis and Coomassie Brilliant Blue staining. Proteolytic products were purified for further analysis by HPLC reverse phase chromatography using a C4 column (Vydac) with a shallow acetonitrile gradient extending from 10 to 80% (both solvents were supplemented with 0.05% trifluoroacetic acid). Protein elution was monitored by absorbance at 280 nm. Samples were lyophilized and sent for mass spectroscopy analysis and N-terminal sequencing at the Weizmann Institute Biological Services Facility or to the Maiman Proteomics Center at Tel Aviv University.

CSN Subunit Interaction Assays

Potential interacting proteins were incubated in 1:1 molar ratio (26 nmol) for 1 h at room temperature in buffer G. Sample was loaded onto either a preequilibrated analytical Superdex 200 or Superdex 75 SEC column (GE Healthcare) running buffer G. Proteins were monitored by absorbance at 280 nm. Individual protein loading concentrations were 0.26 mM, and a 100-μL sample loop was used. Elutions were all performed at room temperature.

For the CSN6-CSN7 complex, association was assessed using a preequilibrated Superdex 200 column with varying conditions. For high pH, we used 20 mM CAPS, pH 10.0, 150 mM NaCl, and 5 mM β-ME. To assess various ionic strengths, 20 mM Tris, 7.5 and 0.15 M, or 0.5 or 2 M NaCl and 5 mM β-ME were used. Pull-down assays were used to examine the complex in the presence of urea. CSN6-CSN7 complex was incubated with preequilibrated metal chelate beads (Ni-CAM) with buffer A for 1 h at 4°C. The beads were then washed with buffer A with increasing concentrations of urea (0.5, 1, 1.5, and 2 M). After each wash, beads were removed for SDS-PAGE analysis. Buffer A supplemented with 250 mM imidazole was used for elution.

In the CSN7 C-terminal tail truncation analysis, bacterial cell lysates were incubated in batch with metal chelate beads (Ni-NTA; Qiagen) for 1 h at 4°C. Gentle centrifugation was used to sediment the beads followed by supernatant removal. The beads were then washed once with buffer A supplemented with 5 mM imidazole and 0.1% Triton X-100. Beads were then boiled in sample buffer and the eluate analyzed by SDS-PAGE and Coomassie Brilliant Blue staining.

Yeast Two-Hybrid Assays

The coding region of CSN7¹⁶⁹ and CSN7^170-225 was cloned individually into the EcoRI-BamHI site of pEG202 to make an in-frame fusion with LexA. Primers are listed in Supplemental Table 1 online. The construct containing full-length CSN7 in pEG202 was described (Karniol et al., 1999). Constructs CSN1 (Kwok et al., 1998), CSN2 (Serino et al., 2003), CSN3 (Peng et al., 2001), CSN4 (Serino et al., 1999), CSN5a (Kwok et al., 1998), CSN6a (Peng et al., 2001), CSN7 (Karniol et al., 1999), and CSN8 (Kwok et al., 1998) in pJG4-5 were described (Serino et al., 2003). The reporter plasmid used was pSH18-34 (Ausubel, 1999). Transformations and plate assays were performed as described (Ausubel, 1999).

ITC

Calorimetry was performed with a MicroCal VP-ITC instrument. All samples were degassed after dilution prior to the experiment. A solution of CSN7¹⁶⁹ (500 μM) was titrated into CSN8 or CSN1 (50 μM). All protein samples were in 20 mM Tris, pH 7.4, 150 mM NaCl, and 5 mM β-ME, and measurements were performed at 10°C. The titration experiment consisted of 29 8-μL injections, where each injection took 16 s. Data were fit to a single-site binding model using Origin 7.0 software (MicroCal). Details on the equations used for this model are provided in Supplemental Methods online.

CD Spectroscopy

All CD measurements were performed with an Aviv CD spectrometer model 202. Spectra were measured over the range of 180 to 260 nm at scan rate of 1 nm/s. For all measurements, a cell with 0.1-mm path length was used except those of CSN7^170-225, where 1-mm path length cell was used. The raw data were corrected by subtracting the contribution of the buffer to CD signal. Data were smoothed and converted to molar ellipticity or Δε units. The measurements were taken at a constant temperature of 4°C with protein concentrations of 25 to 40 μM, except CSN7^170-225, where the concentration was 15 μM. A difference signal of CSN7 C-terminal tail (residues 170 to 225) was calculated using the signal of CSN7 and CSN7 core in the equation: Δε·n_CSN7 − Δε·n_CSN7core = Δε·n_CSN7tail, where Δε·n_CSN7 is the experimental CD signal for full-length CSN7, Δε·n_CSN7core is the experimental CD signal for CSN7¹⁶⁹, and Δε·n_CSN7tail is the calculated difference CD signal. To normalize the signal, we divided the experimental signal by the number of residues of each protein (Figure 6). For the difference calculation, we measured the protein concentration within the experimental cell using the absorbance at 205 nm (Scopes, 1974). All other experiments used protein concentration measured by absorbance at 280 nm. Deconvolution analysis was performed using CDNN (Bohm et al., 1992).

Crystallization and Data Collection

Native crystals for x-ray diffraction experiments were obtained and prepared as previously published (Dessau et al., 2006). Synchrotron data collection was conducted at the European Synchrotron Radiation Facility (ESRF Grenoble, France), beamlines BM-16 and ID14-1, using MAR CCD and Quantum-210 detectors, respectively. All diffraction data were processed with HKL (Otwinowski and Minor, 1997).

Structure Determination

Selenomethionine-substituted CSN7¹⁶⁹ protein was crystallized at 19°C by sitting drop vapor diffusion with conditions near those of native protein. Equal volumes (1 to 2 μL) of protein (5 to 10 mg/mL) were mixed with reservoir solution containing the above conditions plus 5 mM β-ME. Crystals with native morphology appeared rapidly and allowed to grow for 72 to 96 h and were then cryoprotected (reservoir conditions with 15% glycerol). The crystals were mounted in cryoloops and flash frozen with liquid N₂.

Diffraction data for the selenomethionine protein crystals were measured at 110 K and processed as before. A two-wavelength MAD experiment was performed on a single crystal (Table 1). The anomalous absorption peak followed by its inflection point were chosen for the wavelengths. Scaled data sets for each wavelength were then rescaled by local scaling and two selenium sites located using CNS (Brunger et al., 1998). The scaled data and sites were input into SHARP (Bricogne et al., 2003) to calculate experimental phases. Density modification was performed using SOLOMON (Abrahams and Leslie, 1996), whose output gave a 2.1 Å electron density map of high quality. A model was built using ARP/wARP (Morris et al., 2003) and Coot (Emsley and Cowtan, 2004). We refined that model against a native 1.5 Å data set (Table 1) using REFMAC5 (Murshudov et al., 1997). For superposition analysis and molecular graphics, we used LSQMAN (Madsen and Kleywegt, 2002) and Pymol (http://pymol.sourceforge.net), respectively. Secondary structure elements were assigned with DSSPcont (Andersen et al., 2002). Relative solvent accessibility was calculated with NACCESS (http://wolf.bms.umist.ac.uk/naccess/).

Conservation Analysis

Arabidopsis CSN7 amino acid sequence was used as query to search the National Center for Biotechnology Information nonredundant database using BLAST. A manual search of individual sequenced genomes (Ensemble databases) was performed also. Eleven CSN7 ortholog sequences that were found and sorted manually were submitted to ClustalW on the EMBL-EBI server (http://www.ebi.ac.uk/clustalw) for multiple sequence alignment. This multiple sequence alignment was then submitted to the Conseq server (Berezin et al., 2004), which uses the Rate4Site algorithm (Pupko et al., 2002) to compute sequence conservation scores for conservation analysis. Each amino acid received a conservation score (the more negative, the more conserved), but to calculate the conservation score of a particular residue in its specific position, we averaged the surrounding residues around each amino acid in a 19-residue window.

Plant Experimentation

Subcloning for Plant Transformation

cDNA of At CSN7 from the start to the 507 (CSN7¹⁶⁹), 546 (CSN7¹⁸²), and 675 (CSN7^full) bp (without the stop codon) was amplified by PCR with forward primer 41 (see Supplemental Table 1 online) and reverse primers 42, 43, or 44. The protein coding region was digested with SmaI and inserted into pGEM-T, TA cloning vector (Promega), and then sequenced. The insert was then cut with SmaI (Fermentas) and ligated into pPZP (Hajdukiewicz et al., 1994), a binary vector with a HA-tag at the protein-coding C terminus, and checked for insert orientation. Vectors were transformed using Agrobacterium tumefaciens by the floral dip method (Clough and Bent, 1998), and transformants were selected on MS agar plates containing 50 mg/L of kanamycin.

General Growth Conditions for Arabidopsis

Plants were grown under a light-and-dark cycle with 16 h of white light at 75 μmol m⁻² s⁻¹ and 8 h of darkness. Seedlings were grown on agar plates containing 1× strength MS salts, pH 5.7, and 1% sucrose.

Strains Used

All strains are in the Col-0 background, which was used as the wild-type control. The fus5-1 strain was described by Karniol et al. (1999). The 35S:GFP line was provided by E. Sadot (Golomb et al., 2008). The p:GUS line, provided by A. Epel from Tel Aviv University, contains 538 bp upstream of the eIF3e (AT3g57290) start codon fused to GUS in the plasmid pCambia1300.

Plant Protein Extraction and Immunoblot Analysis

Ten-day-old Arabidopsis seedlings were batch-frozen in liquid N₂ and ground into powder using a bead beater (Qiagen). The powder was subsequently resuspended in extraction buffer (100 mM Tris, pH 7.2, 10% sucrose, 5 mM MgCl₂, 5 mM EDTA, 40 μM 2-mercaptoethanol, 50 μM proteasome inhibitor MG-132 [Calbiochem], 1 mM PMSF [Sigma-Aldrich], and protease inhibitor cocktail [Roche]). Insoluble material was pelleted at 20,800g for 30 min in 4°C. The supernatant was collected and used for immunoblot analysis. Proteins separated by SDS-PAGE were transferred to Immobilon-P membranes (Millipore) and probed with the following antibodies: mouse α-HA (1:5000) (Covance), rabbit α-CSN7 (1:5000) (Karniol et al., 1999), rabbit α-CSN5 (1:10,000) (BioMol), and mouse α-actin (1:2000) (Santa Cruz). Bound antibody was detected with alkaline phosphatase (anti-rabbit, 1:10,000) (Sigma-Aldrich) or horseradish peroxidase–coupled (anti-rabbit or anti-mouse,1:10,000) (Jackson Immuno Research) secondary antibodies. Densitometry was performed using ImageMaster 1D Image Analysis (GE Healthcare).

SEC of Plant Homogenate

Total homogenates were prepared in SEC buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl₂, 10 mM MgCl₂, 5 mM EDTA, 5% glycerol, 50 μM proteasome inhibitor MG-132 [Calbiochem], 1 mM PMSF [Sigma-Aldrich], and protease inhibitor cocktail [Roche]). The total homogenate was microcentrifuged for 30 min at maximum speed, and the supernatant was filtered through a 0.45-μm filter (Millex-Millipore). Approximately 400 μg of total soluble protein was fractionated with an analytical Superose 6 HR column (GE Healthcare), previously equilibrated with SEC buffer at a flow rate of 0.3 mL min⁻¹. All fractionations were performed at 4°C. Fractions (0.5 mL) each were collected and concentrated by Strata clean resin beads (Stratagene). The protein standards were as follows: thyroglobulin (669 kD), apoferritin (443 kD), catalase (232 kD), aldolase (158 kD), and BSA (66 kD).

RFLP

One leaf was batch-frozen in liquid N₂ and ground into powder using bead beater (Qiagen). The powder was subsequently resuspended in 100 μL DNA extraction buffer (Kasajima et al., 2004). To remove insoluble material, the resulting extracts were centrifuged at 20,800g for 1 min. The extracted DNA (5 μL) was used for PCR reaction with the CSN7/fus5 primers 45 and 46. The 300-bp product was cut with MvaI restriction enzyme (Fermentas) for 3 h and separated on a 3% TAE agarose gel.

Immunoprecipitation

Total plant extract was cleared of aggregates by centrifugation for 30 min at 4°C at 20,800g. Total protein (400 μg) was incubated with α-HA antibodies in SEC buffer (see SEC section) for 3 h at 4°C on an Orbiton rotator (Boekel Industries). The tube content was transferred to new tubes containing 40 μL of washed protein G Sepharose 4 Fastflow beads (GE Healthcare). The tubes were incubated for an additional 5 h on a rotator at 4°C. The tubes were spun at 5000g for 1 min, and the protein G beads were washed five times with 1 mL of SEC buffer. Proteins were eluted by boiling for 5 min in SDS sample buffer, separated by SDS-PAGE, and blotted for antibody detection.

Hypocotyl Length Measurement

T2 seeds were sown on MS with 1% sucrose plates in an array and grown under white light in 16-h-light/8-h-dark cycles for 5 d. The hypocotyl length was measured on millimetric paper under a binocular microscope (Olympus SZX7).

Accession Numbers

Coordinates and structure factors have been deposited in the Protein Data Bank under code 3CHM. Arabidopsis CSN subunits: CSN1 (AT3G61140), CSN2 (AT2G26990), CSN3 (AT5G14250), CSN4 (AT5G42970), CSN5a (AT1G22920), CSN6a (AT5G56280), CSN7 (AT1G02090), and CSN8 (AT4G14110). Drosophila and mouse CSN7 subunits: Dm CSN7 (Q9V4S8), Mus CSN7a (NP_036133). The eIF3k PDB code is 1RZ4. Sequences for alignment are as follows: Arabidopsis CSN7 (AF063852.1), Danio rerio CSN7 (XM_691997.1), Fugu rubripes (SINFRUG00000151241), Anopheles gambiae CSN7 (XM_320160.2), Drosophila CSN7 (NM_136535.2), Mus musculus CSN7b (NM_172974.1), Rattus norvegicus CSN7b (343614.3), Pan troglodytes CSN7a (ENSPTRG00000004586), Homo sapiens CSN7a (AF193844.1), and H. sapiens CSN7b (NM_022730.1).

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure 1. Limited Proteolysis of Arabidopsis CSN7 with Papain.

• Supplemental Figure 2. Surface Electrostatics Potential of CSN7.

• Supplemental Figure 3. Limited Proteolysis of Drosophila CSN7.

• Supplemental Figure 4. CSN1-CSN7¹⁶⁹ ITC Binding Assay.

• Supplemental Figure 5. CD Spectra of CSN7 Mutants.

• Supplemental Figure 6. Mutation Analysis of CSN7–CSN1 Association Using the SEC Binding Assay.

• Supplemental Figure 7. Analytical SEC of CSN7 C-Terminal Tail.

• Supplemental Figure 8. Transgene Protein Levels in Different Transgenic Plant Lines.

• Supplemental Figure 9. Overexpression of CSN7 Does Not Affect Hypocotyl Length in Dark-Grown Arabidopsis Seedlings.

• Supplemental Figure 10. Superposition of CSN7¹⁶⁹ with eIF3k.

• Supplemental Figure 11. Models of CSN7 PCI Domain Docked to Nucleic Acids.

• Supplemental Figure 12. Transgene Protein Levels in Lines Used for the fus5 Complementation Study.

• Supplemental Table 1. Oligonucleotide Primers Used.

• Supplemental Methods.