Novel Cul3 binding proteins function to remodel E3 ligase complexes

Tandem affinity purification of in vivo Cul3 complexes

In this study, a combination of tandem affinity purification (TAP) and MudPIT mass spectrometry were used to identify new members of the BCR ubiquitin ligase complex. To obtain highly purified Cul3 complexes, an expression cassette encoding a FLAG-TEV-CBP-Cul3 fusion protein was constructed so that the Cul3 complex could be purified using anti-FLAG Sepharose beads and calmodulin-conjugated beads (Figure 1A). The size and expression levels of the FLAG-TEV-CBP-Cul3 fusion protein were verified by immunoblotting and probing against either anti-Cul3 antibody or anti-FLAG antibody (Figure 1B). As the proportion of neddylated to unneddylated Cul3 was roughly the same in endogenous Cul3 and overexpressed Cul3 (Figure 1B, bottom panel, lane 1 and 3), we reasoned that these purified Cul3 complexes would likely to be part of a functional BCR ubiquitin ligase complex rather than a free Cul3 monomer, and thus we proceeded with our experiment. The Cul3 complexes were purified with anti-FLAG Sepharose beads and TEV protease cleavage, where the lack of protein bands in the FLAG immunoblot confirmed that the FLAG-tag had been cleaved off the eluted Cul3 complexes (Figure 1C). Subsequently, the CBP-Cul3 complexes were purified with calmodulin-conjugated beads (Figure 1D).

Data analysis on the MS/MS spectra and potential binding proteins of the BCR ubiquitin ligase complex

Identification of leucine-rich repeats and WD40 domain-containing proteins as Cul3 binding proteins

Proteins that corresponded to the identified peptide spectra were subjected to the domain prediction program CDART and categorized based on conserved domains 17]. When compared with the list of proteins identified from purified actin complexes, we observed enrichment for proteins that contain WD40 domains and leucine-rich repeats (LRR) in the purified Cul3 data set with the percent enrichment scores of 73.3% of WD40 domain and 90.9% for LRR domain-containing proteins in the Cul3-bound complex in contrast to 26.7% of WD40 domain and 9.1% of the LRR domain-containing protein in the actin-bound complex (Additional file3: Tables S3 and Additional file4: Table S4). Because domain enrichment maybe indicative of preferential binding between the BCR complex and proteins with LRR/WD40 domains, we chose to further explore the role of these LRR/WD40 domains in the mammalian BCR complex.

The leucine-rich repeats (LRR) and WD40 domain-containing proteins (CLWs) bind Cul3

Three leucine-rich repeat proteins (LRRs, LRR1, LRR3 and LRR5 (fibromodulin)) and two WD40 domain-containing proteins (W14 and W16) were cloned from the list of 32 MudPIT-identified C ul3-bound proteins with L RR or W D40 domains (CLWs) to verify their binding to Cul3. The cDNAs were fully sequenced and cloned into mammalian expression vectors in frame with the HA-tag. The proteins were co-expressed with FLAG-tagged Cul3 (FLAG-Cul3). Immunoprecipitation assays revealed that FLAG-Cul3 was able to specifically bind the HA-tagged CLWs with either the LRR (Figure 2A) or the WD40 domains (Figure 2B), which further confirms the results from the proteomic analysis.

Cul3 mutants that cannot bind BTB domain-containing proteins have enhanced binding to LRRs

To clarify how other Cul3 binding proteins affect or regulate the binding of LRRs to Cul3, we tested a variety of Cul3 mutants that were unable to bind various Cul3 binding partners. Mutants that could not be neddylated (K712R) or could not bind RBX1 (F665D) were able to bind the LRR proteins in an identical fashion to wild type Cul3 (not shown). Surprisingly, mutants that could not bind BTB domain-containing proteins (Δ51-67, Figure 2C) showed enhanced binding to both LRR1 and LRR5 (Figure 2D). This observation led us to investigate a possible interaction between LRRs and BTB domain-containing proteins.

LRRs bind to BTB domain-containing proteins

Based on our observation that removal of the BTB binding region of Cul3 enhanced LRR binding we hypothesized that they may either have overlapping binding sites on Cul3 or interact with each other. To test this hypothesis, we examined Ctb73 18], also called BTBD1 19, 20] (Figure 3A), which is a BTB domain-containing protein previously shown to mediate substrate binding to Cul3, for its ability to bind to LRR1 and LRR5 (Figure 3B). In both cases, Ctb73 bound specifically to the LRRs. To further map the regions of interaction, we deleted the BTB domain on Ctb73. Unexpectedly; this form of Ctb73 showed enhanced binding to the LRRs. We confirmed this enhanced binding with a different combination of LRRs and BTB domain-containing protein mutants (Additional file1: Figure S1). These results imply that there is a secondary interaction region outside the BTB domain that mediates binding between LRRs and BTB domain-containing proteins.To further dissect the binding interaction between Ctb73 and LRRs, we created several Ctb73 deletion mutants, one of which has a deleted region N-terminal to the BTB domain (Ctb73Δ2-61), which is extremely proline rich (20%) (Figure 3C). We also examined binding between LRR5 and the Ctb73 deletion mutants without the BTB domain, the BACK domain and the Kelch domain. Though each domain deletion failed to completely disrupt the binding, we observed slight reduction between the binding of the Ctb73Δ2-61 mutant and LRRs (Figure 3C, lane 2). We speculated that the reason we could not completely eliminate binding between Ctb73 and LRR5 was that they both were able to bind Cul3 and therefore bridging of Cul3 caused the remaining binding.

Cul3 forms a complex with a BTB/Kelch domain-containing protein and a LRR domain-containing protein

Based on our observations that LRRs bind to Cul3 and BTB domain-containing proteins, as well as prior knowledge that Cul3 binds BTB domain-containing proteins, we wanted to determine if the three proteins co-existed within the same complex. FLAG-Cul3 with a TEV protease cleavage site (FLAG-TEV-Cul3) was co-expressed with MYC-Ctb73 and HA-LRR5. FLAG-beads were used to purify the Cul3 complex, which were subsequently treated with TEV protease to release the protein complex from the beads. The resulting soluble proteins were subjected to a second round of immunoprecipitation with MYC antibody to pull down MYC-Ctb73. The resulting protein complexes were analyzed by Western blot analysis to detect the presence of the third protein, HA-LRR5. Two sequential affinity purifications of FLAG-Cul3 and MYC-Ctb73 pulled down significant amounts HA-LRR5 only in the presence of all three proteins (Figure 3D, top panel). These data verify our hypothesis that Cul3, a BTB-domain containing protein and an LRR could co-exist within the same complex.Because our results in Figures 2 and3 suggested a potential for overlapping or co-regulation of binding between BTB domain-containing proteins and LRRs on Cul3, we also compared complex formation on wild type Cul3 versus the mutant that cannot bind BTB domain-containing proteins (FLAG-Cul3Δ51-67). We observed a slight increase in the binding between the LRRs and the Cul3 mutant when compared to wild type (Figure 3E lanes 2 and 4), which further supports a overlapping binding model.

A B/C-like box sequence mediates LRR binding to Cul3 complexes

As several LRRs are very small and the LRR domain dominates the coding sequence, we began determining the regions on the LRRs that mediate interactions with Cul3 and/or the BTB domain-containing proteins. In this case, the deletion of LRR domain would not be informative because the remaining portion would be short peptides that were unlikely to fold properly. We could not find any conserved regions outside of the LRR domains that were shared by the identified LRR domain containing proteins. Further examination however revealed a sequence adjacent to the LRR domain that strongly resembles a B/C box (Figure 4A). To determine if this sequence may be playing a role in the interaction between the LRR domain containing proteins and Cul3 complexes, we created I330G C334A mutations in the putative B/C-like box of the LRR5 (HA-LRR5-B/C) and tested its binding to Cul3 (Figure 4B) and Ctb73 (Figure 4C). When compared with wild-type LRR5, we observed that the LRR5-B/C mutant showed enhanced binding to wild type Cul3 (Figure 4B, lanes 1 and 4). However, the LRR-B/C interaction with Cul3Δ51-67 mutant that could not bind BTB domain was completely eliminated, though this form of Cul3 mutant showed enhanced binding to wild type LRRs (Figure 4B, lanes 2 and 5). Additionally, the binding between the LRR5-B/C mutant and various Ctb73 deletion mutants were examined. We observed that LRR5-B/C mutant maintained the same binding interaction with wild-type Ctb73 and most of the deletion mutants (Figure 4C). However, the binding interaction between the LRR5-B/C mutant and the Ctb73Δ2-61 was eliminated (Figure 4C, lane2), indicating that this B/C-like box is the major mediator of binding the LRR domain-containing protein to the BTB domain-containing protein. This B/C-like box binds the N-terminal sequence (residues 2–61) on the BTB domain containing protein. In addition, the residual binding between Ctb73Δ2-61 and wild type LRR5 was due to bridging via Cul3. We speculated that the enhanced binding of the LRR-B/C-like box mutant to wild type Cul3 was due to the reduced interaction between the LRR-B/C mutant and a BTB domain-containing protein in the complex, thus eliminating the interaction we have previously shown to regulate LRRs binding to Cul3. Consistent with this model, LRR5-B/C does not show enhanced binding to the version of Ctb73 that is missing its BTB domain, unlike wild-type LRR5 (Figure 4D, lane 2 vs. lane 5). A graphical summary of these interactions is shown in Figure 4E.

FMOD (LRR5) causes FAZF to become degraded

In order to determine the potential cellular role for these complexes that contain a BTB domain-containing protein, an LRR domain-containing protein and Cul3, we performed two hybrid screens using LRR5 as bait. After screening greater than a million clones we identified only one strong binding partner, FAZF. FAZF is a BTB domain-containing protein that has been shown to be important in transcriptional regulation of B-cell differentiation 21]. This was a surprise since we have seen LRR5 bind several BTB domain-containing proteins, and the library we used contains many BTB domain-containing proteins 18]. In order to determine specificity and verify these two proteins bind in mammalian cells LRR5 was co-expressed with either FAZF or the highly similar protein PLZF 22] (Figure 5A). Interestingly, LRR5 bound FAZF and not PLZF and also appeared to cause a reduction in expression levels of FAZF (Figure 5A). We therefore felt that the binding interaction was potentially biologically significant. FAZF overexpression has been shown to cause G1 arrest and apoptosis in B cell lines 21], in addition LRR5 (FMOD) is highly overexpressed in some types of leukemias 23]. Enforced down regulation of LRR5 in these leukemias also results in apoptosis 24]. Based on this observation and on our previous data regarding the relationship between LRR5 binding Cul3 and BTB domain-containing proteins, we postulated that LRR5 might be involved in regulating the levels of the BTB domain-containing protein FAZF. We reasoned that a complex of all three might convert the BTB domain-containing protein to a substrate instead of a substrate adaptor by modulating how it binds to Cul3. Consistent with the proposed hypothesis, the reduced levels of FAZF caused by coexpression of both LRR5 and FAZF could be restored by inhibiting the proteasome with MG-132 (Figure 5B). We then focused on a potential for LRR5 to modulate expression levels of endogenous FAZF using a well-characterized antibody against FAZF 21]. We found that high expression of LRR5 dramatically reduced endogenous levels of FAZF (Figure 5C). As LRR5 levels were decreased using coexpression of shRNA against LRR5, the levels of FAZF increase demonstrating an inverse relationship between expression of LRR5 and FAZF. An analogous experiment in which LRR1 was coexpressed with shRNA against LRR5 showed no change in endogenous FAZF levels (Figure 5D, note: these cells do not express detectable LRR5, not shown). From these data we conclude that LRR5 negatively regulates levels of endogenous FAZF, similar to the regulation we saw with coexpression.

A method to characterize potential substrates of the BCR ubiquitin ligase complex by mass spectrometry

The identification of known substrates of the BCR ubiquitin ligase complex, such as Mei1 and GluR1, by our MudPIT analysis led to the hypothesis that some of the newly identified Cul3-binding proteins are protein substrates targeted for ubiquitination by the BCR ubiquitin ligase complex. A large number of Cul3-binding proteins in this study made it impossible to individually analyze these potential substrates of the BCR ubiquitin ligase complex. Therefore, we searched for a screening method that would identify a large number of potential substrates in a short period of time. We, and others, have previously observed that Cul3 and protein substrates of the BCR ubiquitin ligase complex could be co-purified by an immunoprecipitation assay 12, 25], which led us to speculate that some substrates may remain bound to the Cul3 complex in their ubiquitinated form. In this analysis, purified Cul3 complexes were digested with trypsin, yielding a mixture of ubiquitinated and non-ubiquitinated precursor peptides that were analyzed by the mass spectrometer. Trypsin also cleaved conjugated ubiquitin molecules after the C-terminal lysine 74, leaving two glycine residues attached to the protein substrate on the side chain of the ubiquitinated lysine. Therefore, the mass of a tryptic peptide with a ubiquitinated lysine residue would also include the mass of two glycine residues from the C-terminal end of the conjugated ubiquitin, which would allow us to distinguish these potential substrates of the BCR ubiquitin ligase complex among hundreds of Cul3-binding proteins.

To verify this algorithm for predicting potentially real ubiquitination sites, we mutated the predicted site on Cul3 (K414R) based on our MudPIT data indicating that the K414 residue on Cul3 contains a ubiquitination site based on the extra mass that is equivalent to two glycine residues (Table 2). Biochemical analysis demonstrated that K414 was a genuine ubiquitination site for Cul3 in cells (Additional file1: Figure S2).

LRR domain-containing proteins regulate binding between Cul3 and BTB-domain containing proteins

We discovered that the LRR domain-containing proteins modulated the binding of the BTB domain-containing proteins binding to Cul3 such that they became substrates instead of substrate adaptors. We found that a region on the BTB domain-containing protein just N-terminal to the BTB domain appears to be mediating the interaction with LRR domain-containing proteins. This region on Ctb73 is extremely proline-rich; it is interesting to note that other BTB domain proteins that have been shown to bind Cul3 also have similar proline rich regions in their N-terminus, like actinfilin, BTBD2, and FAZF (PLZF does not have this region and does not bind LRR5). Thus, this region may play a role in a number of other BTB-domain protein/LRR-domain protein interactions. In addition the LRR domain-containing proteins contain a sequence that resembles a B/C box that mediates binding to the BTB domain-containing proteins.

Based on the identification of the WD40 domain in substrate adaptor subunits of the Cul4 complex, we first proposed that our MudPIT-identified WD40 domain-containing proteins also behave similarly in the Cul3 complex. However, mutational analysis revealed that the interaction site for the WD40 domain-containing proteins lies within the conserved cullin domain in C-terminal region of Cul3, suggesting that their binding interactions did not resemble those found on the LRRs (not shown). Therefore, the WD40 domain-containing proteins may function in a completely different way than the LRRs.

Potential substrates of the BCR ubiquitin ligase complex identified by a method that involves an altered SEQUEST algorithm

To distinguish novel Cul3-binding proteins from potential substrates of the BCR ubiquitin ligase complex, we adjusted the parameters in the SEQUEST algorithm to recognize the ubiquitin-modified lysine residue in the peptide spectra. Manual validation of the mass spectra revealed that the altered SEQUEST parameters could be used to identify ubiquitinated precursor peptides that are derived from the ubiquitin-conjugated proteins. This technique for the identification of ubiquitin-conjugated proteins using MudPIT mass spectrometry is an approach that can be applied to the analysis of ubiquitinated proteins bound to other ubiquitin ligase complexes. Using this technique, our list of MudPIT-identified Cul3-binding proteins also includes GluR-1, Mei1 and p60/katanin (Additional file1: Table S1), which are known substrates of the BCR ubiquitin ligase complex 18, 33, 34], further supporting the validity of this method for substrate identification.

Cul3 mediated the ubiquitination of lysine residues on its substrates based on the proximity of its location rather than selecting residues based on specific sequences

In this analysis, we have identified the position of lysine residues on which ubiquitin molecules were conjugated to target proteins, as well as the minimum number of ubiquitination sites that are found on each potential substrate. As a heterogeneous population of both ubiquitinated and non-ubiquitinated lysine residues were observed in identical tryptic peptides of Cul3-binding proteins (Table 2), we believe that the BCR ubiquitin ligase complex exhibits some flexibility in the selection of its ubiquitin conjugation site. Furthermore, a significant number of potential substrates also contain more than one ubiquitination site. Additionally, lysine residues that are within 10 residues of each other are often ubiquitinated at both positions. These findings suggest that the mechanism of ubiquitination by the BCR complex is not based on a specific protein sequence adjacent to the ubiquitinated lysine residue, but may be approximated by the distance of the bound substrate from the ubiquitin-conjugating enzyme. This hypothesis is partly based on a study, which indicates that the initial selection for the ubiquitination site on protein substrates is determined by a tertiary complex of E2-E3-substrate 35].

Model of the BCR^LRR complexes

The SCF model for cullin-based E3 ligase stated that the cullin provides a scaffold to assemble a ubiquitin-conjugating enzyme (E2), Skp1 and an F-box containing protein. The F-box containing protein recognizes substrates and recruits them to the E3 complex (Figure 6A). Previous models of the BCR ubiquitin ligase complex proposed that a BTB domain-containing protein serves the combined functions of both Skp1 and the F-box protein by binding to Cul3 via its BTB domain and using its secondary protein-protein interacting domain to target its bound substrates for ubiquitination (Figure 6B). This model is supported by the identification of substrates of the Cul3 ubiquitin ligase complex and their adaptor proteins, such as Nrf2 and Keap1, GluR1 and actinfillin, Dishevelled and KLHL12, Aurora B and KLHL11/KLHL9 and p60/katanin and KLHDC5 18, 36–39]. In this report, we have identified an additional group of Cul3-binding proteins that contain either LRR domains or WD40 domains. Incidentally, these some of these LRR domain-containing proteins also interact with the BTB domain-containing proteins. We propose that the B/C-like box domain on the LRRs and a unique, proline rich domain on the BTB domain containing protein Ctb73 is responsible for the binding between the BTB domain-containing protein, the LRRs and Cul3. This binding causes the BTB domain-containing protein to become a substrate, possibly by changing how it binds to Cul3 (Figure 6C).

Plasmids

Full-length cDNA sequences of Cul3, BTBs, and CLWs were cloned from a human testis cDNA library and inserted in-frame into the 3XFLAG-24 (Sigma-Aldrich), CS2 + MT, and CS2 + HA vectors, respectively. All clones were fully sequenced. Mutations in the Cul3 protein were made using the QuikChange site-directed mutagenesis kit (Stratagene). All plasmids were sequenced after mutagenesis. Expression plasmids for MudPIT analysis were constructed by cloning the full-length Cul3 sequence into the 3XFLAGNEO-26 plasmid (Sigma-Aldrich) in-frame with the FLAG tag. A TEV recognition sequence and a calmodulin-binding protein sequence were inserted in-frame between the 3XFLAG epitope and the Cul3 sequence, creating the 3XFLAG-TEV-CBP-Cul3 plasmid that was used for tandem affinity purification. As a control, a 3XFLAG-TEV-CBP-actin plasmid was created by cloning the full-length actin sequence from a human testis cDNA library and inserting it in-frame into the 3XFLAG-TEV-CBP plasmid.

Cell culture and protein expression

HEK293 cell lines were maintained in DMEM media supplemented with 10% fetal bovine serum (FBS). For transient protein expression, HEK293 cells were transfected with either 3XFLAG-TEV-CBP-Cul3 or 3XFLAG-TEV-CBP-actin plasmids as described previously 42].

Two-hybrid screening

pGildaLRR5 (FMOD) was screened using the Matchmaker LexA two hybrid system (Clontech). A human testis library was used and 2.7 million clones were screened. The only interacting clone identified was FAZF.

Immunoblotting, immunoprecipitation and immunofluorescence

Transfected HEK293 cells were harvested in 2.7 mM EDTA in phosphate-buffered saline and cell lysates were prepared by sonication for 10 seconds in radioimmunoprecipitation assay (RIPA) buffer (1% NP-40, 1% sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 0.01 M sodium phosphate buffer pH 7.2 and 2 mM EDTA). For immunoprecipitation assays, the clarified lysates were incubated with primary antibody, followed by protein A-Sepharose beads for 1 hr each at 4°C. Beads were spun down and washed twice with RIPA buffer. Protein samples were separated by electrophoresis on 10% polyacrylamide gels, transferred to PVDF membranes, and incubated overnight in primary antibodies. Proteins were visualized using enhanced chemiluminescence followed by digital acquisition using an AlphaInnotech Fluorochem SP system. The following antibodies were used in these experiments: anti-FLAG M2 monoclonal antibody (Sigma-Aldrich), anti-actin polyclonal antibody (Sigma-Aldrich), anti-HA monoclonal antibody (Covance), anti-MYC sc-789 polyclonal antibody (Santa Cruz Biotechnology), and anti-Cul3 antibody which has been previously described 43]. The anti-FAZF antibody was a kind gift of Dr. Maureen Hoatlin (OHSU). Immunofluorescence was performed as previously described 18].

FLAG immunoprecipitation and TEV protease cleavage

Clarified cell lysates were incubated with anti-FLAG M2 agarose slurry (Sigma-Aldrich) at 4°C for 1 hour. The protein-bound beads were washed twice in RIPA buffer and three times in tobacco etch virus (TEV) buffer (10 mM HEPES-KOH pH 8.0, 150 mM NaCl, 0.1% NP-40, 0.5 mM EDTA and 1 mM DTT). The FLAG beads were then incubated with the AcTEV protease enzyme (Life Technologies) in TEV buffer overnight at 4°C with constant agitation. After TEV cleavage was complete, the supernatant containing Cul3 complexes was collected and the FLAG beads were washed three times with TEV protease buffer and twice with calmodulin-binding buffer (10 mM 2-mercaptoethanol, 10 mM HEPES-KOH pH 8.0, 150 mM NaCl, 1 mM MgOAc, 1 mM Imidazole, 0.1% NP-40, and 2 mM CaCl₂) to recover the Cul3 complexes that were cleaved but remained bound to the beads. A sample of FLAG beads prior to TEV cleavage and protein samples from each eluted fraction were immunoblotted and probed with the appropriate antibodies to determine the efficiency of TEV cleavage and to estimate the protein yield for the next purification step.

Purification with calmodulin-conjugated beads

Collected protein samples from the FLAG immunoprecipitation and the TEV protease cleavage were supplemented with 1 M CaCl₂ solution at a 25:1 ratio, followed by incubation with Calmodulin Sepharose™ 4B beads (Amersham Biosciences) for 90 minutes at 4°C. The beads were spun down at 750 × g and washed three times with the calmodulin-binding buffer and twice with the calmodulin-rinsing buffer (50 mM NH₄HCO₃ pH 8.0, 75 mM NaCl, 1 mM MgOAc, 1 mM imidazole, and 2 mM CaCl₂). The bound protein complexes were then eluted from the calmodulin beads using calmodulin-eluting buffer (50 mM ammonium bicarbonate, pH 8.0 and 25 mM EGTA) and collected in separate fractions. A sample of protein-bound calmodulin beads and protein samples of each eluted fraction were immunoblotted and probed with the appropriate antibodies to estimate the relative protein yield of each eluted fraction. The fractions with the most amount of protein were selected for further analysis using MudPIT techonology.

Trypsin digestion and peptide desalting

An equal volume of 8 M urea, pH 8.3, was added to the eluants from the calmodulin bead purification and the mixture was denatured at 95°C for 5 minutes. After the mixture had cooled to room temperature, trypsin (Promega) was added to purified proteins in a 1:40 w/w ratio and the mixture was incubated overnight at 37°C. Once the digestion was complete, the peptide mixture was diluted with 0.1% acetic acid, adjusted to pH 3, and then loaded onto a C18 reverse phase peptide macrotrap cartridge (Michrom Bioresources) that had been equilibrated with 0.1% acetic acid. The cartridge with the bound protein was washed twice with 0.1% acetic acid and the peptides were eluted with 70% acetonitrile in 0.1% acetic acid. The peptide mixture was dried using a SAVANT SC110A SpeedVac centrifuge with an RVT400 refrigerated vapor trap (Thermo Electron Corp.) and stored at -80°C.

Strong cation exchange chromatography

The dried peptides were reconstituted in 0.1% acetic acid to a concentration of 0.5 mg/ml. They were subsequently loaded onto a strong cation capillary column (home-packed on a pressure bomb) with an inner diameter of 247.0 μm (Polymicro technology), 20 cm in length, and fritted by M-520 (Upchurch) containing polysulfoethyl A™ bulk material of 5 μm diameter and 300 Å pore size (The Nest Group, Inc.). The samples were eluted with a 12-step salt gradient between buffer C (30% acetonitrile, 5 mM phosphate buffer pH 3) and buffer D (30% acetonitrile, 5 mM phosphate buffer pH 3, and 500 mM ammonium acetate) with 0.01 ml/min flow rate for 10 minutes in each gradient step. Each solution step consists of a different ratio of buffer C to buffer D (C/D): step 1 = 100/0, step 2 = 98/2, step 3 = 96/4, step 4 = 94/6, step 5 = 92/8, step 6 = 90/10, step 7 = 88/12, step 8 = 86/14, step 9 = 84/16, step 10 = 80/20, step 11 = 50/50, and step 12 = 0/100. The eluted peptides from each salt gradient step were collected separately, dried and stored at -80°C until the next step.

On-line reverse phase column and tandem mass spectrometry

In preparation for analysis by mass spectrometry, the dried peptide samples from each salt step fraction were reconstituted in 0.1% acetic acid. The peptides were loaded onto a pre-column (360 μm outer diameter × 75 μm inner diameter) containing 2 cm of 5 μm Monitor C18 resin (Column Engineering). They were then eluted into the mass spectrometer (LTQ-FTICR, Thermo Electron Corp.) through an analytical column (360 μm outer diameter × 75 μm inner diameter fused silica packed with 12 cm of 5 μm Monitor C18 particles with an integrated ESI emitter tip having a 4 μm opening fritted with 3 μm silica particles; Bangs Labs, Fishers, IN) by using a linear HPLC gradient ranging from 0-70% acetonitrile over 30 minutes using buffer A (0.1 M acetic acid) and buffer B (0.1 M acetic acid in acetonitrile) with peak parting. The eluted peptides were then processed by an LTQ-FTICR mass spectrometer (data-dependent scanning with 1 MS scan followed by 5 MS/MS scans). MS/MS spectra were generated and automatically searched against the human NCBI non-redundant protein database using the SEQUEST algorithm provided with Bioworks 3.2 SR (Thermo Electron Corp.). The list of predicted proteins was imported into a custom-made FileMaker Pro relational database for data analysis.

Data analysis and domain prediction

MS/MS spectra with mass error of less than 20 ppm and high Xcorr value (charge 1 ≥ 2, charge 2 ≥ 2.5, and charge 3 ≥ 3) were chosen for further analysis. The exact predicted peptide sequences were searched against sequences in the NCBI protein-protein database to uncover potential Cul3-binding proteins 44]. All corresponding protein sequences were then scanned for predicted domains with the NCBI conserved domain architecture retrieval tool (CDART) 45]. MudPIT mass spectrometry was also performed on tandem affinity purified actin complexes to eliminate proteins that may be mistaken for binding partners. These may be contaminants from the purification that bind non-specifically to the CBP tag or to the surface of large globular proteins, such as actin. Predicted protein candidates that were identified in both the actin and the Cul3 complexes were regarded as false positives and removed from the list of potential Cul3-binding proteins. The remaining proteins were then categorized based on their conserved domains and their roles in cellular pathways.

Availability of supporting data

The data sets supporting the results of this article are included within the article (and its additional supplemental files).