Expression and purification of functional recombinant CUL2•RBX1 from E. coli

Cullin-2 (CUL2) based cullin-RING ligases (CRL2s) comprise a family of ubiquitin E3 ligases that exist only in multi-cellular organisms and are crucial for cellular processes such as embryogenesis and viral pathogenesis. CUL2 is the scaffold protein that binds one of the interchangeable substrate receptor modules, which consists of adaptor proteins and the substrate receptor protein. The VHL protein is a substrate receptor known to target hypoxia-inducible factor α (HIF1α) for ubiquitination and degradation. Because of its critical role in the ubiquitination of important cellular factors such as HIF1α, CRL2s have been investigated for their biological functions and the development of novel therapeutics against diseases. Given the importance of CRL2s in biological and biomedical research, methods that efficiently produce functional CUL2 proteins will greatly facilitate studies on the mechanism and regulation of CRL2s. Here, we report two cost-effective systems for the expression and purification of recombinant human CUL2 from E. coli cells. The purified CUL2 proteins were ~ 95% pure, could bind their substrate receptor modules, and were enzymatically active in transferring ubiquitin or ubiquitin-like protein to the corresponding substrate in in vitro assays. The presented methodological advancements will help advance research in CRL2 function and regulation.

www.nature.com/scientificreports/ first known structures of CUL2 in complex with RBX1, EloB/C, and VHL 33,34 . Although the baculovirus-infected insect cells can produce properly folded full-length CUL2, the production of baculovirus vectors can be time consuming, and strict, costly cell culture conditions are usually required for optimal protein yield from insect cells 35,36 . Thus, a more cost-effective method for recombinant CUL2 protein expression and purification is desirable to facilitate studies of CRL2.
Here, we report novel methods for generating human CUL2•RBX1 proteins from E. coli cells. With a single plasmid, the recombinant CUL2•RBX1 can be generated by either co-expressing the full-length CUL2 with RBX1 or using the "Split-n-Coexpress" strategy 35,37 . In our in vitro assays, the recombinant CUL2•RBX1 purified from either expression system can catalyze the conjugation of NEDD8 to CUL2 and can bind VHL•EloB/C to form CRL2 VHL that ubiquitinates the peptide substrate derived from the degron motif of HIF1α. In summary, we have established and optimized an efficient system for bacterial expression and purification of functional recombinant human CUL2•RBX1 proteins.

Results
Expression and purification of full-length human CUL2 from E. coli. Previous trials of expressing human CUL2 protein have generated the N-terminal segment of CUL2 (CUL2  ) from E. coli cells 38 , and the full-length CUL2 in complex with RBX1 from insect cells 33,34 . The low protein solubility is a common limitation for high-level expression of full-length cullin proteins in bacteria 35 , and this limitation was overcome by deleting two short unstructured regions when full-length human CUL1•RBX1 was produced in E. coli 39 . We learnt from the successful experience for CUL1•RBX1 expression and we used computational tools to predict disordered regions in the CUL2 sequence (Fig. 1A). We then found that deleting two segments in CUL2 (Fig. 1A, shaded sequences), which are invisible in the CUL2 crystal structure 33 , eliminated all disordered regions predicted in silico. Therefore, we tested if deleting these segments would yield soluble CUL2 when CUL2 was co-expressed with RBX1 in E. coli cells. Furthermore, because MsyB, a hyper-acidic bacterial protein, has been shown to improve protein solubility and assist correct protein folding in E. coli 40 , we fused it to the N-terminus of CUL2 and placed a TEV protease cutting site after it. Key components in plasmids co-expressing His6 RBX1 and His6 M-syB-StrepII CUL2 are illustrated in Fig. 1B. The His6 tag and StrepII tag were added for purification purposes. The deleted segment was marked as Δ (Fig. 1B).
Expression and purification of human CUL2 from E. coli cells via "Split-n-Coexpress". After successful expression and purification of RBX1•CUL2 FL , we then sought to improve the yield of the recombinant protein. Besides deleting the disordered region in the CUL2 NTD (CUL2Δ), we also tried codon optimization of the CUL2 coding sequence and the "Split-n-Coexpress" strategy. "Split-n-Coexpress", which splits the cullin protein into halves and co-expresses them with RBX1, was firstly reported for expressing RBX1•CUL1 in E. coli cells 35 . In search of a way to split CUL2, we aligned crystal structures of CUL1 and CUL2 ( Fig. 2A), and we found that the counterpart for the CUL1 splitting site is after the 14th helix in CUL2 ( Fig. 2A). Thus, we generated a construct to co-express GST RBX1, StrepII CUL2 2-380 , and CUL2 381-745 (Fig. 2B). We then used excess amounts of Strep-Tactin resin to extract different variants of RBX1• StrepII CUL2 expressed in E. coli, including RBX1• StrepII CUL2 FL , RBX1• StrepII CUL2Δ, RBX1• StrepII CUL2 opFL with CUL2 codon optimized, and RBX1• StrepII CUL2 opSplit with CUL2 split and codon optimized. The results showed that while deleting the disordered segment at NTD or codon optimization alone increased the amount of extractable StrepII CUL2, "Split-n-Coexpress" offered a significant improvement (Fig. 2C). We thus expressed and purified the RBX1• StrepII CUL2 Split through glutathione affinity chromatography (Fig. 2D, #1), thrombin cleavage (Fig. 2D, #2), Strep-Tactin affinity chromatography (Fig. 2D, #3), and size exclusion chromatography (Fig. 2D, #4). We obtained RBX1• StrepII CUL2 Split that was over 98% pure with a yield of ~ 0.3 mg protein per liter of culture, and its retention volume on the size exclusion chromatography ( Fig. 2E) was almost the same as for RBX1• StrepII CUL2 FL (Fig. 1E).

Activities of the recombinant RBX1•CUL2 purified from E. coli.
To determine if the recombinant RBX1•CUL2 purified from E. coli cells retained their activity, we first tested the binding of VHL•EloB/C, a substrate receptor module for CUL2, with our recombinant RBX1•CUL2. Our pulldown assay with the StrepII tag showed that RBX1• StrepII CUL2 FL , RBX1• StrepII CUL2Δ, and RBX1• StrepII CUL2 Split all bound to VHL•EloB/C at similar degrees (Fig. 3A). Furthermore, previous studies have shown that the third amino acid in CUL2, a Leucine (L3), is important for EloC binding 38 . Thus, we deleted L3 from our CUL2 FL (FL ΔL3 ) and CUL2 Split (SP ΔL3 ) protein, and consistent with the previous report, we found that the loss of L3 significantly reduced the binding of CUL2 to VHL•EloB/C (Fig. 3B)    www.nature.com/scientificreports/ lyzed the protein mixture on size exclusion chromatography, similarly to that performed before 33,41 . As shown in Fig. 3C, the protein mixture was eluted as two major peaks (green lines). The first peak was eluted earlier than RBX1•CUL2 (orange lines), and it contained both RBX1•CUL2 and VHL•EloB/C (Fig. S1), suggesting the formation of the pentameric CRL2 VHL complex. The second peak contained VHL•EloB/C (Fig. S1), and it showed the same retention volume as VHL•EloB/C (purple lines) with a reduced peak area. Because peak areas represent amounts of the detected analyte, by quantifying the reduction in the VHL•EloB/C peak area, we estimated that 79%-83% of RBX1•CUL2 in the protein mixture was in complex with VHL•EloB/C (Fig. 3C). Taken together, these results demonstrated that our recombinant RBX1•CUL2 can efficiently recruit the substrate receptor module to form a CRL2 complex. We then tested if the recombinant RBX1•CUL2 was enzymatically active. Conjugating the small protein NEDD8 to cullin is one of the key mechanisms for CRL activation 2,42,43 . This process, referred to as neddylation, is achieved by recruiting E2 ~ NEDD8 via RBX1 and subsequently transferring NEDD8 to cullin CTD 2,42,43 . We performed in vitro neddylation of our recombinant RBX1•CUL2, including RBX1•CUL2 FL , RBX1•CUL2Δ, and RBX1•CUL2 Split . We found that the protein bands representing CUL2 FL , CUL2Δ, or CUL2 381-745 (CUL2 CTD ) shifted upwards on the SDS-PAGE gel after incubating with the neddylation enzymes (Fig. 4A), suggesting that each variant of CUL2 was fully conjugated to NEDD8. We further tested if the neddylated RBX1•CUL2, once associated with VHL•EloB/C, can ubiquitinate its substrate, a peptide derived from the amino-terminal oxygen-dependent degradation (NODD) motif of HIF1α 44 . This peptide substrate, which differs from the CODD degron sequence characterized previously 33  (green). The same trace for VHL•EloB/C was used in all three plots. Based on the reduction of the VHL•EloB/C peak area, percentage of RBX1•CUL2 assembled into CRL2 VHL complex was estimated as 79% for CUL2 FL , 82% for CUL2Δ, and 83% for CUL2 Split . Protein species under the main peaks were confirmed by SDS-PAGE (Fig.  S1).  47 , and a TAMRA fluorophore at the C-terminus for detection. Besides the three variants of RBX1•CUL2 we purified from E. coli, we also included RBX1•CUL2 purified from insect cells as a positive control, and a no RBX1•CUL2 mixture as a negative control. As shown in Fig. 4B, fluorescence scan for TAMRA signal revealed that when no RBX1•CUL2 was present, the peptide substrate was unchanged. In contrast, when each type of RBX1•CUL2 was present, TAMRA labeled molecules with higher molecular weights appeared and accumulated over time. To further access if the enzymatic activities differ among different versions of RBX1•CUL2, we repeated the substrate ubiquitination assay with more time points and shorter time periods (Fig. 4C, left panel). By comparing the rates at which ubiquitinated substrates were generated, we found that RBX1•CUL2 purified from E. coli cells displayed activities similar to, or slightly greater than, RBX1•CUL2 purified from insect cells (Fig. 4C, right panel). Based on these results, we conclude that all forms of the recombinant RBX1•CUL2 we purified from E. coli cells were enzymatically active.

Discussion
We have developed new cost-effective systems to generate recombinant human RBX1•CUL2. We found that the addition of an MsyB fusion protein helped producing soluble full-length CUL2 in complex with RBX1 from bacterial cells. This is likely because the MsyB protein assists protein folding specifically in E. coli cells by reducing aggregation of the targeted protein 40 . In addition to the MsyB fusion protein, we also found that deleting one unstructured region at the CUL2 NTD (RBX1•CUL2Δ) increased the yield of soluble protein.  www.nature.com/scientificreports/ After successfully producing the full-length CUL2 protein in E. coli, we further increased our protein yield with codon optimization of the CUL2 coding sequence and the "Split-n-Coexpress" strategy. In the "Split-n-Coexpress" system, the cullin protein is divided into two halves, which allows for the full expression of two smaller subunits. When folding, the two subunits dock into each other at a hydrophobic interface to assemble the functional full protein 35,37 . With the structural alignment of CUL1 and CUL2, we determined the split site for CUL2, and we successfully produced RBX1•CUL2 Split via the "Split-n-Coexpress" system. We found that both codon optimization and "Split-n-Coexpress" were beneficial to our protein production, while the "Split-n-Coexpress" offered a greater improvement.

A B
After purifying our RBX1•CUL2 FL , RBX1•CUL2Δ, and RBX1•CUL2 Split proteins with two affinity chromatography, site-specific protease cleavage for tag removal, and size exclusion chromatography, we then tested their biochemical activities. We first demonstrated that all versions of the recombinant RBX1•CUL2 could efficiently form CRL2 through recruiting its substrate receptor module VHL•EloB/C, unless the L3 at CUL2 NTD was intentionally deleted. This result is consistent with the previous finding that the L3 of Cul2 is important for the binding of EloC 38 , and it shows that our recombinant RBX1•CUL2 can detect changes in the affinity of CRL2. We then showed that all forms of our recombinant RBX1•CUL2 could be fully conjugated to NEDD8, a key modification of cullin that promotes CRL activities 2,4,43,48 . Lastly and importantly, we confirmed that our purified RBX1•CUL2 proteins were capable of ubiquitinating its substrate. Because HIF1α is a well-studied CUL2 substrate 4,49 , we used a hydroxylated peptide that corresponds to a degron motif of HIF1α as the substrate for our RBX1•CUL2 in the in vitro ubiquitination assay. As a positive control, we also included full-length RBX1•CUL2 produced from insect cells, which was used in previous structural studies of CRL2 33,34 . We showed that in the presence of RBX1•CUL2 produced from either bacteria or insect cells, the peptide substrates were ubiquitinated over time and increasing amounts of higher molecular weight products were detected. Of note, we found that when mixed with slight excess amount of VHL•EloB/C, around 80% of bacteria produced RBX1•CUL2 could form the CRL2 VHL complex. While this percentage of complex assembly is lower than RBX1•CUL2 purified from insect cells 33,41 , all forms of bacteria produced RBX1•CUL2 showed ubiquitination activities equivalent to RBX1•CUL2 produced from insect cells.
In conclusion, our study provides novel and efficient methods for producing fully functional recombinant RBX1•CUL2 in E. coli. We expect that the new expression systems we established will facilitate future research on CRL2, such as structural or mechanistic studies of factors/mutations that alter CRL2 assembly or activity. Further, in vitro ubiquitination assays similar to what we reported here will help validate novel natural or neosubstrates of CRL2s, study kinetics of substrate ubiquitination, as well as characterize regulators or PROTACs that modulate the enzymatic activity of CRL2s.

Constructs.
For generating the construct co-expressing RBX1 and CUL2, a pGEX vector (Sigma-Aldrich) was firstly edited using the Q5 Site-Directed Mutagenesis Kit (NEB BioLabs), to replace the GST coding sequence with DNA sequence encoding a His6 tag followed by a TEV protease cutting site (ENLYFQS) and a few restriction enzyme cutting sites. The following DNA fragments were sequentially inserted after the TEV site: RBX1 coding sequence (with NdeI/NotI sites), T7 promoter (with NotI/NcoI sites), MsyB coding sequence followed by a TEV site (with NcoI/NheI sites), and StrepII CUL2 coding sequence (with NheI/SalI sites). Codon optimized CUL2 coding sequence was synthesized (Gene Universal) and inserted with NheI/SalI sites. Loop deletion(s) of CUL2 were introduced using the Q5 Site-Directed Mutagenesis Kit (NEB BioLabs). The construct expressing RBX1• StrepII CUL2 Split was generated through modifying the pCool vector (Addgene Plasmid #29,519, a gift from Ning Zheng). First, the CUL1 coding sequence was replaced by DNA sequence encoding codon optimized CUL2 381-745 (with NcoI/NotI sites). Then the sequence of codon optimized StrepII CUL2 2-380 with the RBS sequence preceding it was inserted via the NotI site. Sequences of all the constructs were confirmed by Sanger sequencing.
Expression and purification of VHL•EloB/C. VHL 54-213 •EloB/C complex was expressed by co-transforming BL21 (DE3) E. coli with XLB250 and XLB192 and inducing overnight at 16°C 50,51 . It was then purified on glutathione resin followed by on-column digestion with thrombin overnight at 4 °C. Protein released from the glutathione resin to the supernatant was concentrated via a centrifugal filter unit (10 kD cutoff, Millipore-