|Home | About | Journals | Submit | Contact Us | Français|
The CUL4-DDB1 ubiquitin ligase machinery regulates diverse cellular functions and is frequently subverted by pathogenic viruses. Here we report the crystal structure of DDB1 in complex with a central fragment of hepatitis B virus X protein (HBx), whose DDB1-binding activity is essential for viral infection. The structure reveals that HBx binds DDB1 through an α-helical motif, which is also found in the unrelated paramyxovirus SV5-V protein despite their sequence divergence. Our structure-based functional analysis shows that, like SV5-V, HBx captures DDB1 to exploit the ubiquitin ligase activity of the CUL4-DDB1 E3. Based on the shared action mechanisms of the two viral proteins, we further identify the same α-helical motif in the substrate-recruiting subunits of the cellular E3 complex, DCAFs, which are functionally mimicked by the viral hijackers. Together, our studies reveal a common yet promiscuous structural element important for the assembly of viral and cellular substrate receptors into the core complex of the CUL4-DDB1 ubiquitin ligase.
Protein ubiquitination is a widespread post-translational modification that regulates the activities of myriad eukaryotic proteins in diverse cellular functions1. In order to conjugate ubiquitin to various protein targets with high specificity, eukaryotic cells have evolved a large number of enzymes, known as ubiquitin E3 ligases, that can each recognize one or a limited set of specific protein substrates and catalyze the ubiquitin transfer reaction together with ubiquitin-activating E1 and ubiquitin-conjugating E2 enzymes2. The cullin-RING ligases represent the largest super-family of multi-subunit E3 ubiquitin ligase complexes in eukaryotic cells3. Organized by a catalytic core consisting of a cullin scaffold and the RING domain protein Rbx1/Roc1, the cullin-RING complexes all feature interchangeable substrate receptor subunits, which are docked to the E3 ligase platform through an adaptor. By combining different substrate receptors with the same catalytic core, the cullin-RING E3 complexes greatly expand their substrate repertoire while maintaining high specificity. In humans, six closely related cullin proteins (CUL1, CUL2, CUL3, CUL4A, CUL4B, and CUL5) have been identified, each capable of assembling a distinct family of E3 complexes. Using Skp1 and Skp1-like adaptors, CUL1–3 and CUL5 share a common structural mechanism to nucleate different cullin-RING E3 complexes.
Recent studies have shown that the human CUL4A and CUL4B proteins can both interact with the evolutionally conserved DDB1 adaptor protein and assemble a unique family of cullin-RING ligase complexes, hereafter referred to as the CUL4-DDB1 E3s4–7. Cellular functions regulated by CUL4-DDB1 E3s include, but are not limited to, DNA repair8–14, DNA replication5,14–18, transcription19, and signal transduction20,21. Distinct from all other cullin adaptors with the Skp1/BTB fold, DDB1 is a large multi-domain protein consisting of three β-propeller domains (BPA to BPC) and a C-terminal α-helical fold22. In DDB1, the BPB domain interacts with the N-terminal end of the CUL4 scaffold, whereas the structurally coupled BPA-BPC double-propeller fold is responsible for docking a family of substrate receptor proteins, known as DCAFs (DDB1-CUL4-Associated Factors)4,5 or DWDs (DDB1-binding WD40 proteins)6. Although the cellular functions of many DCAFs remain poorly understood, most of them are characterized by a WD-repeat domain in their primary sequences. Analogous to the ones found in some of the SCF substrate receptor F-box proteins, such as β-TrCP and Fbw723,24, the WD-repeat domains of DCAFs potentially provide the substrate-binding sites on the CUL4-DDB1 E3 complexes. Lacking a conserved DDB1-binding motif outside the WD-repeat domain, however, how DCAFs are selectively recognized by DDB1 among all WD-repeat proteins remains unclear. A double DxR motif on the surface of the WD-repeat domains of DCAFs and a DWD motif within the DCAF WD-repeat sequences have been separately proposed as the signature motif for the CUL4-DDB1 substrate receptors4–6. Yet, neither motif is strictly shared among all experimentally identified DCAFs, suggesting that a more complex mechanism might underlie the specific docking of DCAFs to DDB1.
As widely expressed cellular ubiquitin ligases, the cullin-RING E3 complexes are frequently subverted by pathogenic viruses3. By bridging a cellular target to the adaptor or a substrate receptor of a cullin-RING E3, a number of viral proteins are able to hijack the cellular E3 machinery for ubiquitinating and degrading anti-viral or regulatory factors of the host. One of the best examples is the exploitation of the CUL4-DDB1 E3 by the paramyxovirus V protein, which can functionally mimic DCAFs to directly dock to DDB1 and simultaneously interact with one of the STAT proteins in the interferon signaling pathway25–27. Upon recruiting a STAT protein to the CUL4-DDB1 ligase complex, the V protein promotes the polyubiquitination and rapid turnover of the key signal tranducer, thereby, blocking the anti-viral response of the host cells. The hepatitis B virus X protein (HBx), which plays an essential role in viral replication in vivo28, is a second viral protein known to specifically interact with DDB129. Although the precise mechanism by which HBx mediates viral replication remains elusive, its normal interaction with DDB1 is essential for efficient viral infection as well as its reported activities in cultured cells, including stimulation of viral genome replication and induction of genetic instability and subsequent cell death30–34. In this study, we show that, despite sequence diversity, HBx anchors itself on DDB1 through an α-helical motif that is also used by the paramyxovirus V protein. Similar to the paramyxovirus V protein, HBx requires the intact CUL4-DDB1 complex, thereby, the ligase function of the E3 machinery, to retain its activities. Upon revealing the common structural element used by both viral proteins to functionally mimic DCAFs, we further identify a similar and previously unrecognized α-helical motif in the cellular substrate receptors of the CUL4-DDB1 ubiquitin ligase critical for the assembly of the modular E3 machinery.
The hepatitis B virus X protein (HBx) is a 17 kDa small regulatory protein conserved among mammalian hepadnaviruses29. Like HBx, the woodchuck hepatitis virus X protein (WHx) also shows the DDB1-binding activity, which is essential for efficient viral infection in vivo35,36. Previous studies have mapped a partially conserved short sequence motif in HBx and WHx important for DDB1 association37 (Fig. 1a). To unravel the structural basis of the interaction between DDB1 and the viral X proteins, we have determined the crystal structures of human DDB1 in complexes with peptides corresponding to these central fragments of HBx and WHx (Table 1). The DDB1-HBx complex structure reveals that the HBx peptide adopts a 3-turn α-helical conformation and binds to DDB1 at the large pocket enclosed by its BPA-BPC double propeller fold (Fig. 1b). With few contact to the DDB1 BPA domain, the HBx peptide predominantly interacts with the “top” surface of the DDB1 BPC domain. Most, if not all, DDB1-interacting residues of the viral peptide are located at the bottom side and the two ends of the α-helical structure. At the opening of the DDB1 double propeller pocket, the N-terminal end of the HBx peptide helix interacts with a DDB1 loop (4b–4c loop) projecting from the BPC domain. Deep inside the pocket, the viral sequence ends its helical conformation and points its very C-terminus back toward the entrance of the DDB1 pocket (Fig. 1b).
The WHx peptide adopts the same helical conformation as HBx does upon interacting with DDB1, albeit their rather divergent sequences (Fig. 1c,d,e). The two viral DDB1-binding sequences have only three invariant amino acids, all of which are found at the C-terminal end of the helical motif (Fig. 1a). Of these three conserved residues, the viral Arg residue (Arg96 of HBx and Arg94 of WHx) forms two hydrogen bonds with DDB1; the viral Leu residue (Leu98 of HBx and Leu96 of WHx) is accommodated by a hydrophobic patch on the DDB1 BPC domain formed among Leu328, Pro358, and Ala381, Phe382; and the viral Gly residue (Gly99 of HBx and Gly97 of WHx) terminates the helix (Fig. 1c). In both structures, this part of the interface is further strengthened by two DDB1 residues, Arg327 and Asn1005, each donating a hydrogen bond to a carbonyl group of the helical peptide backbone (Fig. 1c).
The sequences of the N-terminal halves of the HBx and WHx helical motifs are noticeably divergent, with no strictly conserved amino acid (Fig. 1a). Most DDB1-contacting residues in this part of the two X proteins, nevertheless, show overall conserved side chain properties and contribute to the DDB1 interaction in a conserved fashion. In particular, Phe87, Val88, and His91 of WHx and their corresponding HBx residues mediate the intermolecular docking by packing against several fixed residues on the “top” surface of the DDB1 BPC domain (Fig. 1d). Superposition analysis of the two structures, on the other hand, reveals unexpected structural differences in DDB1 upon interacting with the two different X proteins. In the DDB1-WHx complex, three hydrogen bonds are formed between the N-terminal end of the viral helix and the DDB1 4b-4c loop, two made by the first amino acid of the viral helical motif (Asn96), whereas in the DDB1-HBx structure, the DDB1 loops is pushed further out by the viral helix, making only hydrophobic and van der Waal contacts with the N-terminal end of viral peptide (Fig. 1d). Next to the middle part of the viral helices, another surface loop of the DDB1 BPC domain also adopts different conformation in the two complex structures (Supplementary Fig. S1). Thus, the combination of the hydrophobic surface properties and the intrinsic plasticity of the “top” surface of the DDB1 BPC domain allow it to accommodate the viral helical motifs with significant sequence variation. Overall, the two viral peptides fold into a similar short α-helix and anchor themselves deeply into the large DDB1 double propeller pocket, burying a total of ~800 Å2 surface area.
The helical DDB1-binding motif of the viral X proteins is reminiscent of the paramyxovirus SV5-V protein, which also contains an N-terminal helical sequence interacting with the DDB1 double propeller pocket22. Superposition analysis shows that the SV5-V N-terminal sequence and the two X protein peptides adopt essentially the same helical structure and occupy the same surface area on the DDB1 BPC domain (Fig. 1e). Strikingly, the amino acid sequence of the SV5-V helical motif is significantly different from the HBx helical motif (Fig. 1a). Consistent with the lack of detectable sequence homology between the unrelated viral V and X proteins, the DDB1-binding motifs of HBx, WHx, and SV5-V have no single amino acid in common when they are aligned based on the structures (Fig. 1a). None of the three invariant residues between HBx and WHx at the C-terminal part of the helical motif is conserved in SV5-V. Only when aligned in pairs does the helical motif of WHx show sequence similarity to each of the other two, but in non-overlapping positions (Fig. 1a).
Close examination of the interfaces in all three structures reveals several common key contacts made by the viral motifs through amino acids of the same or similar types. For instance, all the hydrophobic residues of the SV5-V helix (Val24, Phe27, and Val32) form hydrophobic interactions with the same DDB1 residues as their corresponding residues in HBx and WHx do. The same hydrogen bond network formed among the backbone groups of the WHx peptide and the surface residues of the DDB1 BPC domain is also found in the DDB1-SV5-V structure. Together, these analyses indicate that the completely unrelated paramyxovirus V and hepatitis virus X proteins share a common mechanism for DDB1 interaction, which involves the docking of a short α-helical viral motif formed by relatively variable sequences to DDB1.
To verify the crystallographic results in the context of the full-length proteins, we performed a series of structure-based mutation analyses of DDB1. We first investigated whether the binding of full-length HBx involves the structurally independent BPB domain of DDB1, which is closely connected to the BPC domain (Fig. 1b). Wild-type DDB1 and a deletion mutant lacking the entire BPB domain were fused in frame to the VP16 activation domain and tested in a yeast two-hybrid assay for interaction with HBx, WHx, and SV5-V linked to the DNA-binding protein RFX. Figure 2a shows that the VP16-DDB1 truncation mutant lacking the BPB domain activates an RFX-dependent lacZ reporter gene comparably to the wild type DDB1 protein when expressed in combination with RFX-HBx or RFX-WHx. The same is true when SV5-V is used as bait (Supplementary Fig. S2). These results suggest that HBx and WHx interact with only the BPA-BPC module of DDB1, as SV5-V does in the crystal structure of the DDB1-SV5-V complex22.
We next probed the importance of the “top” surface of the DDB1 BPC domain for holding the full-length viral X proteins in place. Our analysis focused on two DDB1 amino acids, Ala381 and Phe382, which pack against the HBx Leu98 residue deep inside the DDB1 double propeller pocket (Fig. 1c). As shown in Figure 2b, a DDB1 double mutant bearing charged substitutions at these two positions (A381E/F382D) effectively disrupts the binding of DDB1 to both full-length HBx and WHx, and also impairs DDB1 association with SV5-V (Fig. 2b and Supplementary Fig. S2). The DDB1 double mutant exhibits normal binding to a cellular DDB1 partner of unknown function4, Trpc4AP, which also interacts with the DDB1 double propeller (Fig. 2b and Supplementary Fig. S2). This control experiment indicates that mutations of the two DDB1 residues abolish viral protein binding without affecting the proper folding of DDB1. Consistent with these results, mutation of Leu98 within HBx has been shown to compromise its binding to DDB133. Taken together, the above mutational analyses underline the importance of the interface between the viral helical motif and the “top” surface of the DDB1 BPC domain for intact HBx-DDB1 complex formation.
To test the functional importance of the interaction between DDB1 and the viral X protein helical motif, we next examined whether HBx would retain deleterious activities in mammalian cells expressing the DDB1 mutant defective for HBx binding in place of endogenous DDB1. Because DDB1 is essential for viability of proliferating cells38,39, we first established that the DDB1(A381E/F382D) double mutant can functionally substitute for the wide type DDB1 for cell viability and proliferation. Figure 2c shows that silencing of DDB1 by transfection of an episomal vector directing the synthesis of a DDB1-specific small interfering RNA (siRNA) inhibits HeLa cell growth in a colony formation assay, as expected (Fig. 2c, upper panel). These cells can be largely rescued by transfection with an siRNA-resistant form (SiR) of wild-type DDB1 and, to a slightly lesser extent, with the DDB1(A381E/F382D) double mutant, but not by transfection with wild-type DDB1 (Fig. 2C, lower panel). Western blot analysis using HA-epitope tagged DDB1 to distinguish from the endogenous protein demonstrates that the siRNA-mediated knockdown of DDB1 is efficient and specific and that the DDB1SiR variants are expressed at close to normal levels (Supplementary Fig. S3).
To assess for HBx cytotoxic activity in the DDB1(A381E/F382D) mutant background, we generated HeLa cells depleted for DDB1 by siRNA and expressing the siRNA-resistant version of either wild-type DDB1 or the DDB1(A381E/F382D) double mutant. The transfected cells were then transduced with lentiviral vectors encoding GFP, GFP-HBx or the GFP-HBx(R96E) point mutant, which is defective for DDB1 binding as we have reported before33. Figure 2d demonstrates that transduction efficiency as determined 5 days later by FACS analysis for GFP fluorescence was high in all cases (Fig. 2d, left panel). A very similar FACS profile was obtained at 16 days after transduction with cells transduced with GFP or the GFP-HBx(R96E) mutant (Fig. 2d, right panel), indicating that expression of these proteins confers no major growth disadvantage to cells. By contrast, the proportion of GFP-HBx-expressing cells markedly decreased in control cells normally expressing endogenous DDB1 and in cells complemented with the wild-type DDB1SiR variant (Fig. 2d, right panel), consistent with HBx exerting deleterious activities in these cells. This, however, did not occur when cells were complemented with the HBx-defective DDB1(A381E/F382D)SiR mutant. In fact, most cells in this case remained GFP-positive at day 16 (Fig. 2d, right panel) and showed proliferation in a colony formation assay (Supplementary Fig. S4). Thus, HBx is expressed yet largely lacks cytotoxic activities in these cells. This result unequivocally demonstrates that HBx acts through its interaction with DDB1 and binding of the viral helical motif to the “top” surface of the DDB1 BPC domain is essential for HBx activities.
The common DDB1-binding motif shared between the hepatitis X protein and the paramyxovirus V protein suggests that HBx may function as SV5-V does by subverting the normal function of the cellular ubiquitin ligase complex. Alternatively, HBx could inhibit the activities of the CUL4-DDB1 E3 machinery to fulfill its role. To distinguish these two possibilities, we first probe whether HBx can physically integrate into the CUL4A-DDB1 complex. Figure 3a (lower panel) shows that upon co-transfection both HA-DDB1 and Myc-Cul4A can be coimmunoprecipitated efficiently with GFP-HBx but not with a GFP control from extracts of HeLa cells. Much lower amounts of myc-Cul4A and HA-DDB1 are recovered when the experiment is performed with the HBx(R96E) mutant defective for DDB1-binding.
Next, we generated a CUL4-binding defective mutant of DDB1 to test whether CUL4 and the rest of the E3 ubiquitin ligase complex are in fact required for HBx activities. Our previous structural studies revealed that CUL4A interacts with DDB1 by contacting both the “top” surface and one peripheral side of the DDB1 BPB domain22. Although simultaneous mutations of three critical residues (W561K, I587D and R589E) on the “top” surface of the DDB1 BPB domain was not sufficient to disrupt the binding of DDB1 to CUL4A (data not shown), introduction of a fourth amino acid substitution, A400D, at the peripheral side of the DDB1 BPB domain completely abolished complex formation as assessed by coimmunoprecipitation analysis (Fig. 3b). Because the resulting mutant, DDB1(m4), fails to substitute for endogenous DDB1 (data not shown), we tested the mutant for its ability to support HBx-induced cytotoxicity following a previously established method. Our previous studies have shown that a covalent link between HBx and DDB1, by acting as a “clamp” forcing the two protein together, can restore the activities of the HBx(R96E) DDB1-binding mutant40. Figure 3c shows that, with the same expression level, HBx inhibits cell growth in a colony-forming assay when fused to wild-type DDB1, as expected, but not when fused to the DDB1(m4) mutant. Hence, HBx cytotoxic activity requires DDB1 binding to CUL4A and, likely, the ubiquitin ligase activity of the cellular E3 complex.
To determine whether HBx has similar requirements for stimulation of HBV replication, we made use of an HBx-dependent replication system in which the HBx function can be provided in trans31. In this assay, human hepatoma cells are transfected with either a wild-type HBV genomic construct or a mutant lacking a functional HBx gene, with or without cotransfection of an HBx expression plasmid. The amount of viral DNA replicative intermediates recovered from purified cytoplasmic core particles is quantified by real time PCR. As seen in Figure 3c, replication of the HBx-deficient HBV genome is strongly reduced compared to the wild type. It is restored by cotransfection of HBx but not the HBx(R96E) DDB1-binding mutant. Remarkably, both wild-type HBx and the HBx(R96E) mutant exhibit stimulatory activities when fused to native DDB1. Yet, they are essentially inactive when linked to the CUL4A-binding defective DDB1(m4) mutant. The same was observed when the HBx fusions were tested for their ability to stimulate transcription of a luciferase reporter gene placed under control of the HBV enhancer I and associated promoter element (Supplementary Fig. S5). These results highlight the importance of DDB1 binding to CUL4A for HBx activities and implicate that HBx interacts with DDB1 to exploit the ligase function of the cellular E3 machinery instead of inhibiting it. The precise substrate ubiquitinated by the HBx-hijacked CUL4-DDB1 E3 complex awaits future investigation.
By reprogramming the cellular CUL4-DDB1 E3, the viral X and V proteins functionally mimic DCAFs, the modular substrate receptors of the E3 complex. The common DDB1-docking mode adopted by the two viral proteins raises the possibility that docking of DCAFs to DDB1 might involve the same helical motif, a structural mechanism also mimicked by the viral proteins. Our previous studies have shown that the WD-repeat-containing DCAF proteins interact with DDB1 on its BPA-BPC double propeller fold4. If DCAFs indeed bind to the “top” surface of the DDB1 BPC domain, the X protein peptides should be able to compete with DCAFs for DDB1 binding. We tested this in an in vitro pull-down assay with purified recombinant proteins. As seen in Figure 4a, DDB1 shows robust interaction with GST-fused DCAF9, a DCAF protein identified in our previous proteomic studies and also known as WDTC14, whereas the two proteins can no longer form a stable complex when DDB1 is preloaded with the α-helical DDB1-binding peptide of WHx. This result suggests that either the “top” surface of the DDB1 BPC domain is directly involved in DDB1-DCAF9 interactions, or the subtle conformational changes induced by the viral peptide binding might indirectly perturb the DDB1-DCAF9 interface.
The WD-repeat domains of most DCAFs are formed exclusively by β-strands, which fold into a globular β-propeller fold. It is, therefore, unlikely that the DCAFs use their WD-repeat domain to interact with the “top” surface of the DDB1 BPC domain as the viral helical sequences do. As DCAF9 has sequence extensions beyond both ends of its predicted WD-repeat domain, we hypothesized that these regions outside the WD repeats might be responsible for contacting the DDB1-BPC “top” surface. Indeed, binding analysis with a series of DCAF9 truncation mutants shows that removal of the N-terminal 26 amino acids is sufficient to abolish DCAF9-DDB1 interaction, whereas C-terminal truncation of DCAF9 has little effect (Fig. 4b,c). Intriguingly, amino acids 5 to 17 in the N-terminal sequence of DCAF9 is predicted to be an α-helix and shows moderate sequence similarity to the WHx DDB1-binding helix (Fig. 5b). These properties of the DCAF9 N-terminal sequence strongly suggest that the DCAF9 substrate receptor of the CUL4-DDB1 E3 might also use a short α-helical motif to dock to the DDB1 at the double propeller pocket.
In order to validate this possible structural role played by the DCAF9 N-terminal sequence, we determined the crystal structure of DDB1 in complex with a peptide corresponding to amino acids 5 to 17 of the DCAF protein (Table 1). Consistent to our binding analysis and the secondary structure prediction result, the DCAF9 peptide adopts a helical conformation and interacts with the “top” surface of the DDB1 BPC domain in exactly the same manner as the three viral proteins do (Fig. 5a,c). Despite considerable sequence variation in the N-terminal half, the short helical DDB1-binding motif of DCAF9 makes similar contacts with DDB1 as observed in the viral protein-DDB1 complexes. This result, once again, underscores the capability of the DDB1 BPC domain to accommodate variable helical sequences through its “top” surface. Importantly, it definitely shows that a DCAF protein binds DDB1 through an α-helical motif anchored at the DDB1 double propeller pocket and such a binding mode is mimicked by the viral hijackers. Given the α-helical structure of the motif and its presence in the HBx protein, we name it the H-box.
In the primary sequence of DCAF9, the H-box motif is located about 30 amino acids N- terminal to the first predicted β-strand of the WD-repeat domain. To reveal whether other DCAF proteins also have such a short motif, we scrutinized the sequences of other DCAFs in the similar region and searched for a short sequence that is predicted to adopt an α-helical secondary structure and shares a similar sequence pattern with the H-box motifs from DCAF9 and the three viral proteins. Based on the four available crystal structures, we outlined the H-box motif as a 13 amino acids sequence with a non-charged polar residue or a non-aromatic hydrophobic residue at position 1, generally hydrophobic residues at position 2, 3, and 6, and a Val/Leu/Ile residue at position 11. Furthermore, an Arg residue is preferentially found at position 9 and a Gly residue is preferentially located at either position 12 or 13. In six additional DCAFs (DDB2, DCAF4, DCAF5, DCAF6, DCAF8, and DCAF12), a potential H-box motif can be identified fitting the criteria above (Fig. 5b). Due to their sequence divergence and short length, these motifs were not recognized in previous analysis, although a truncation mutant of DDB2 lacking part of the predicted H-box motif has been previously shown to be important for DDB1-DDB2 interaction5. To verify that these predicted helical motifs in DCAFs indeed interact with DDB1 like the H-boxes in DCAF9 and the viral proteins, we determined the crystal structure of DDB1 in complex with a peptide corresponding to the predicted motif in DDB2. Previous studies have shown that an eleven amino acid region of DDB2 overlapping the predicted H-box motif is important for DDB2-DDB1 association5. As shown in Figure 5c, the DDB2 peptide is indeed α-helical and binds DDB1 in the same fashion as the ones from DCAF9, HBx, WHx, and SV5-V. We therefore conclude that at least a major fraction of the substrate receptors of the CUL4-DDB1 ubiquitin ligase uses the H-box motif to bind the DDB1 adaptor protein, a structural mechanism that is also employed by viral hijackers of the cellular E3 complex.
While our current studies have identified the H-box motif as a critical structural element used by both viral and cellular substrate receptors to bind DDB1, several lines of evidence suggest that it is not the only structural contact made by these substrate receptor proteins to dock to the CUL4 adaptor protein. In the crystal structure of the DDB1-SV5-V complex, the paramyxovirus V protein features a C-terminal globular zinc-binding domain that closely interacts with the DDB1 BPC domain outside its “top” surface22. Previous deletion analysis indicated that both the H-box motif and the zinc-binding domain of SV5-V are important for DDB1 binding41. Our studies suggest that such a bipartite binding mechanism might be also true for HBx, whose residues outside the H-box sequence also contribute to DDB1 association33. Similar to the viral hijackers, DCAFs might interact with DDB1 via multiple interfaces, which involve not only the H-box motif but also other parts of the polypeptides, most likely the common WD-repeat domain. In fact, our previously proposed tandem double DxR motif on the “bottom” surface of the WD-repeat domains of DCAFs might mediate such interactions, as point mutations in the double DxR motif in several DCAFs effectively abrogate DDB1 binding4. As shown in Figure 6, we propose that HBx and most WD-repeat domain-containing DCAF proteins interact with DDB1 through a bipartite interface. Inside the DDB1 double propeller pocket, the short and helical H-box motif of HBx and DCAFs docks to the “top” surface of the DDB1 BPC domain. Outside the DDB1 pocket, another domain of HBx and the WD-repeat domain of DCAFs might reinforce the assembly by anchoring to the DDB1 double propeller outside the pocket. Although we have found the H-box motif in a total of seven DCAF proteins with high confidence, it is very likely that such a motif also exists in other DCAFs and possibly involves even more divergent sequences. While the functional advantage for DCAFs to have such a bipartite interface with DDB1 remains to be understood, it might provide a unique mechanism for the CUL4-DDB1DCAF E3 complex to switch between productive and non-productive forms of a E3 machinery without completely disassembling the ubiquitin ligase complex.
Full-length human DDB1 was expressed and purified as described previously22. GST tagged full-length DCAF9/WDTC1 and its truncation mutants were expressed in E. coli and isolated from soluble cell lysate by glutathione affinity chromatography.
DDB1 was crystallized under the same condition as before22. Single DDB1 crystals were soaked with 1mM peptide in crystallization reservoir solution for three hours and subsequently frozen in a cryoprotectant solution in liquid nitrogen for data collection. The DDB1-peptide crystals have the same space group, unit cell, and number of molecules in an asymmetric unit as the crystals of DDB1 alone. With the published DDB1 structure as search model, structures of the DDB1-HBx, DDB1-WHx, DDB1-DCAF9, and DDB1-DDB2 complexes were determined by molecular replacement using the program Phaser42. The peptide models were built in with COOT or O 42,43. CNS and Refmac were used for model refinement44. The final parameters of the each structure are laid out comparatively in Table I.
GST tagged full-length DCAF9 and its truncation mutants were left on the glutathione sepharose beads after affinity purification from E. coli cultures. DDB1 was mixed with excessive amount of peptide WHx before passed through the beads. After extensive wash, proteins left were eluted from the beads and analyzed by SDS-PAGE with Coomassie staining.
GFP, GFP-HBx and GFP-HBx(R96E) expressed from the lentiviral vector pWPTS (Fig. 2d and S4) and from the episomal vector EBS-PL (Fig. 5a) have been described31. DDB1, the HA-tagged DDB1 version carrying a triple HA epitope at the N terminus32, and mutants thereof were produced from EBS-PL except in Figure 5b where the HA-DDB1 proteins were made from pSRαS. The DDB1-specific small interfering RNA (siRNA) was produced from the pLV-TH lentiviral vector carrying a GFP marker31 in Figure S4 and from EBB-SUP40 in all the other Figures. The myc-epitope tagged CUL4A in pcDNA has been described previously22. The HBx-DDB1 and HBx(R96E)-DDB1 fusions, all variants thereof, and the control HBx and HBx(R96E) point mutant were all expressed from EBS-PL. The replication-competent wild-type HBV genomic construct (payw1.2) and the HBx-deficient derivative (payw*7) used in Figure 5d has been previously published31. The luciferase reporter construct used in Figure S5 is driven by a 550-bp restriction fragment derived from payw1.2 and containing the HBV Enhancer I and associated X gene core promoter. The fragment was cloned upstream of the luciferase coding region into pGL3 (Promega) using a naturally occurring NcoI site overlapping the ATG initiator codon of the X gene. DDB1(ΔBPB) was generated by ligation of two PCR products to replace the entire BPB domain (from I396 to Q708) by 2 glycine residues linking BPA to BPC. DDB1(A381E, F382D) and the CUL4-binding defective DDB1(m4) mutants were produced by multistep PCR mutagenesis using partially overlapping primers. The siRNA-resistant (SiR) DDB1 versions used in Figure 2 contain three silent mutations in the region covered by the siRNA. All constructs were verified by partial sequencing. Details of the plasmid constructions are available upon request.
HeLa cells and the human hepatoma HepG2 cell line were grown exactly as described34. Cells seeded the day before at a density of ~105 total cells per 30-mm-diameter well (~8×105 HepG2 cells in the replication assay31) were transfected using FuGENE 6 (Roche), or JetPEI (PolyPlus Transfection) in Figures 5a and 5b, following the manufacturer’s instructions. When needed, a GFP expression plasmid was cotransfected (10% of total DNA) to assess transfection efficiencies. Total DNA amount was kept equal using empty vectors. Transfection efficiencies estimated 24-h later by fluorescence-activated cell sorting (FACS) analysis were generally 50–70% with HeLa cells and around 5% with HepG2 cells. Lentiviral transduction was with an MOI of 5 as described before31.
The coimmunoprecipitation experiments in Figures 5a and 5b were performed with whole-cell extracts prepared 24-h post-transfection from 2×105 HeLa cells collected with Versene (Invitrogen) as described34, except that glycerol was added at 17% final concentration after cell lysis. Western blot analyses were performed by transfer of proteins to PVDF Hybond-P membranes (Amersham). The membranes were probed with 1:1,000 rabbit anti-GFP polyclonal antibodies (Santa Cruz Biotechnology), 1:1,000 anti-HA monoclonal antibody (clone 16B12; Covance), 1:1,000 anti-myc monoclonal antibody (clone 9E10; Covance), 1:250 goat anti-DDB1 antibodies (Everest Biotech, Fig. 5c) or 1:500 anti-DDB1 monoclonal antibody (Zymed Laboratories, Fig. S3), and 1:2,000 anti-a-tubulin monoclonal antibody (Sigma-Aldrich). Horseradish peroxidase-conjugated donkey anti-goat IgG (Santa Cruz Biotechnology, 1:5,000), sheep anti-rabbit or anti-mouse IgG (Amersham Biosciences, 1:5,000) were used as secondary antibodies and detection was carried out with ECL (Pierce).
Singly or doubly transfected HeLa cells were re-plated at lower density 1 day after transfection and cultured in appropriate selection media containing 6 μg/ml blasticidin S (Invitrogen) and/or 200 μg/ml hygromycin B (Chemie Brunschwig). When needed, cells were re-plated after appropriate dilution to avoid cell death due to confluency. Drug- resistant cells were stained with crystal violet (Sigma) on the day indicated in the Figure legends. Note that in Figure 2c, control cells and cells complemented with DDB1(WTSiR) were re-plated six times, whereas DDB1(dmSiR) complemented cells were re-plated three times before staining.
Viral genome replication was assessed by determining the amount of cytoplasmic core particle-associated HBV DNA three days after transfection as previously described31, except that quantitation was performed by real-time PCR45. Each value is the mean of four separate PCR reactions performed with two primer pairs designed to amplify distinct regions within the HBV genome, one in the polymerase gene and the other within Enhancer I, and two dilutions of the template DNA. Shown are the mean values normalized to transfection efficiency.
All the proteins are encoded by single-copy plasmids marked with the TRP1, URA3 or ADE2 gene. VP16-DDB1, VP16-RFX, RFX-HBx, and SV5-V fused to RFX or overexpressed in its native form have been described33,40. VP16-DDB1(ΔBPB) and VP16-DDB1(A381E/F382D) were constructed by replacing the region encoding wild- type DDB1 in VP16-DDB1. RFX-WHx was generated by PCR amplifying the WHx coding region from a cloned woodchuck hepatitis B virus genome (J02442; kindly provided to us by Olivier Hantz, INSERM, U871, Lyon) and inserting the resulting fragment into RFX-HBx to replace the HBx coding sequence. RFX-Trpc4AP was constructed in a similar way using a Trpc4-Associated Protein isoform (a) (NM_015638) cDNA clone that was isolated in a yeast two-hybrid screen for DDB1 interactors (O. Leupin and M.S., unpublished). A fragment starting 24 nucleotides upstream of the AUG initiation codon and encompassing the entire Trpc4AP open reading frame was excised from the original clone and inserted in frame C-terminal to RFX, resulting in RFX- Trpc4AP. The β-galactosidase assays were performed in a yeast strain carrying an integrated RFX-dependent lacZ reporter gene as described previously40.
We thank ALS synchrotron beamline staff for assistance with data collection; all members of the Zheng lab for invaluable discussions; Wenqing Xu for help and support in our research; and Oliver Leupin for the siRNA-resistant DDB1 clone. N.Z. is a Pew scholar. This study is supported by the Howard Hughes Medical Institute and by BWF PATH award and NIH grant CA107134 to N.Z. and Swiss National Science Foundation 3100A0-100785 and 3100A0-112496 to M.S.
The Protein Data Bank accession numbers for the DDB1-HBx, DDB1-WHx, DDB1-DCAF9, and DDB1-DDB2 structures are WWWW, XXXX, YYYY, and ZZZZ.