|Home | About | Journals | Submit | Contact Us | Français|
The evolutionarily conserved cullin family proteins can assemble as many as 400 distinct E3 ubiquitin ligase complexes that regulate diverse cellular pathways. CUL4, one of three founding cullins conserved from yeast to humans, uses a large β-propeller protein, DDB1, as a linker to interact with a subset of WD40 proteins that serve as substrate receptors, forming as many as 90 E3 complexes in mammals. Many CRL4 complexes are involved in chromatin regulation and are frequently hijacked by different viruses.
The ubiquitin pathway regulates a wide range of diverse cellular processes by covalently attaching ubiquitin to specific substrate proteins either to alter their functions non-proteolytically or more commonly to promote their degradation by the 26S proteosome . Ubiquitylation begins with the ATP dependent activation of ubiquitin by the E1 enzyme, and is followed by the subsequent transfer of ubiquitin to an E2 ubiquitin conjugating enzyme, and finally an E3 ubiquitin ligase is responsible for recognizing a specific substrate and promoting ubiquitin ligation. Over a thousand distinct E3 ligases, belonging to two major families, the HECT family and the RING family, are present in higher eukaryotes. Human cells contain as many as 28 HECT proteins and most, if not all, are believed to function as E3 ligases. More than 450 RING proteins are expressed in human cells and E3 ligase activity has been experimentally demonstrated for many of them.
Although not containing a RING domain themselves, members of the evolutionarily conserved cullin family can bind with a small RING protein, either ROC1 or ROC2 (for RING of Cullins; also known as Rbx and Hrt1), to constitute a large family of cullin-RING E3 ligases (CRLs). There are three cullins in yeast, six in Caenorhabditis elegans and Drosophila melanogaster (CUL1-6), up to nine in Arabidopsis thaliana, and six in humans (CUL1, 2, 3, 4A, 4B and 5). In addition, three other proteins, CUL7 and CUL9 (also known as PARC) in mammals and APC2 (anaphase promoting complex subunit 2) in all eukaryotes, contain significant sequence homology to cullins over a ~180 residue region and bind ROC or a homologous small RING protein, APC11. Unlike other RING E3 ligases, cullins do not bind substrates directly, but rather rely on substrate recruiting receptors that are typically joined to the cullin complex by a linker protein (Fig. 1). Remarkably, each cullin can associate with a different family of substrate receptors, leading to the assembly of as many as 400 distinct CRLs. A recent estimate suggests that 20% of all proteins subject to proteasomal degradation are targeted by CRLs .
The founding cullin gene, Cul1(also called Cdc53), was first identified in budding yeast where its mutation blocks the G1-to-S transition  and leads to an accumulation of G1 cyclin , and was separately identified in C. elegans where loss of Cul1 function caused hyperplasia of multiple tissues . These diverse or even seemingly irreconcilable phenotypes were subsequently explained by the discovery that CUL1 can assemble into distinct E3 complexes and thereby ubiquitylate different substrates. CUL1 binds with a small protein, SKP1 (S-phase kinase-associated protein 1), which in turn binds with a conserved protein motif, the F-box, which is present in many different proteins . Through additional protein–protein interaction modules, individual F-box proteins can then recruit different substrates to the CUL1– ROC1 – E2 catalytic core . In this modular SKP1– CUL1– F-box (SCF) assembly, CUL1 indeed acts as a scaffold, SKP1 as a linker and F-box proteins as substrate receptors (Fig. 1).
To bring specific substrates to CUL2- and CUL5-based ligases, a heterodimeric linker complex containing elongins B and C binds to an analogous N-terminal domain in CUL2 and CUL5 and to two similar protein motifs, the VHL-box and SOCS box. VHL (von Hippel-Lindau) and SOCS (suppressor of cytokine signaling) proteins, via their additional protein–protein interaction modules, target various substrates differentially to the CUL2–ROC1– E2 or CUL5–ROC2– E2 catalytic cores . Omitting a linker, CUL3 utilizes its N-terminal domain to bind to a conserved 100-residue protein motif known as a BTB domain, which then target substrates to the CUL3–ROC1– E2 catalytic core  (Figure 1).
In this article, we will discuss the genetic function, molecular assembly and substrates of CUL4– RING ligases (CRL4s). We will compare CUL4’s pleiotropic functions across several species, and review the current model for CUL4 assembly with its linker protein DDB1 (for damaged DNA binding) and its family of WD40 containing substrate receptors. Finally, we will highlight recent studies expanding CRL4CDT2 substrates on chromatin, providing new details on the structure of CRL4s, and analyzing new Cul4A mouse models. Additionally, we will discuss outstanding questions remaining in the field.
CUL4 presents as a single gene in Schizosaccharomyces pombe, C. clegans, Drosophila, and Arabidopsis whereas mammalian cells express two closely related paralogues, CUL4A and CUL4B. Two closely related CUL4 genes are also present in zebrafish and frog, but not in sea urchin, suggesting that the CUL4 gene duplication is probably unique to vertebrates. Characterizations of null mutation or reduced expression of CUL4 in these organisms have revealed a wide range of cellular and organismal defects, including many associated with deregulated chromatin.
Cul4 (also known as pcu4) deletion in fission yeast results in severely growth retarded, but viable cells that are extremely elongated with decondensed chromosomes . The defect in chromosome condensation can be attributed in part to a role for Cul4 in maintaining the heterochromatin formation through its interaction not with Ddb1, but instead with Rik1, a protein that is distantly related to mammalian DDB1. Using Rik1 as linker protein, Cul4 binds the Clr4 histone H3K9 methyltransferase and is required for Clr4 localization to heterochromatin [11–13].
Inactivation of cul-4 in C. elegans by RNA interference (RNAi) results in a developmental arrest at the L2 larval stage and causes massive DNA re-replication. In cul-4 RNAi animals, there is a sustained accumulation of the DNA replication-licensing factor, CDT-1 (chromatin licensing and DNA replication factor 1), after S-phase completion whereas wild-type cells express undetectable CDT-1. Additionally, the nuclear export of the other replication licensing factor, Cdc6, is inhibited indirectly due to a failure to ubiquitylate the p21 homolog CKI-1 and contributes to the dramatic re-replication phenotype. Removal of one genomic copy of cdt-1 suppresses the cul-4 re-replication phenotype. These results demonstrate a function of CUL-4 in maintaining genome stability in part by facilitating CDT-1 degradation during the cell cycle and thus preventing aberrant re-initiation of DNA replication .
Arabidopsis CUL4 is expressed abundantly and broadly in almost all tissues examined, including inflorescences, stems, roots, siliques, both light- and dark-grown seedlings and leaves. Arabidopsis expresses two different CUL4 isoforms from a single gene, with one (CUL4-L) containing an additional 50 amino acid sequence at the N-terminus than the other (CUL4-S). Reduced CUL4 expression in Arabidopsis by transgene-mediated co-suppression or RNAi results in pleiotropic cop-like (constitutive photomorphogenesis) phenotypes and widespread developmental defects in lateral roots, abnormal vascular tissue, and stomatal development [15, 16], providing genetic evidence supporting the notion that CUL4 can assemble multiple distinct CRL4 complexes and regulate many different substrates.
In mammals, genetic analysis of CUL4 function is complicated by the existence of two closely related genes, CUL4A and CUL4B, which encode proteins sharing 80% identity with CUL4B containing an extended N-terminus (Figure 2A). A Cul4A mutant mouse created by deleting exon 1 (Cul4AΔ1/Δ1, Figure 2B) was reported to cause early embryonic lethality and wide spread defects in many organs and animal death in adult mice when conditionally deleted . This phenotype led to the early impression that the two CUL4 genes in mammals are functionally distinct. Recently, a different Cul4A mutant mouse strain targeting exons 17-19 (Cul4AΔ17-19/Δ17-19) which encode the ROC binding region and Nedd8 modification site unexpectedly showed no apparent phenotype . This significant discrepancy was explained by the inadvertent disruption in the Cul4AΔ1/Δ1 mutant strain of a very close neighboring gene, Pcid2, which encodes an uncharacterized protein with homology to essential proteasome subunits . Supporting this explanation, deletion of exons 4-8 (Cul4AΔ4-8/Δ4-8) which encode for a portion of the DDB1 binding domain resulted in only a mild decrease in the proliferation in mouse embryonic fibroblasts (MEFs) and viable mice [, Pradip Raychaudhuri, personal communication].
Despite their high degree of sequence homology, wide expression and apparent compensation of Cul4A loss by Cul4B in the mouse, the two CUL4 genes are not entirely functionally redundant. Deletion of exons 17-19 resulted in an increase in stability of several CRL4 substrates, including DDB2, xeroderma pigmentosum complementation group c (XPC) and p21, and enhanced DDB2-dependent global genomic repair activity of UV-damaged DNA in MEFs, and led to resistance to UV-induced skin cancer in mice . These findings, which are consistent with the enhanced resistance to UV-induced carcinogenesis in DDB2 transgenic mice , point to a role for CUL4A in restricting, rather than promoting, the activity of DNA damage responsive proteins by promoting DDB2 degradation, a function that is apparently not fully compensated by CUL4B. Although a Cul4B mutant mouse has not been reported, several familial mutations in CUL4B have been identified in association with X-linked mental retardation (XLMR) in humans [21–23]. Supporting a distinct genetic function of CUL4B gene, the N-terminus of CUL4B, but not CUL4A, has been shown to uniquely binds to the dioxin receptor (AhR), assembling a CUL4B specific ubiquitin ligase complex that targets estrogen receptor α (ER-α) for degradation . DDB1, the linker that is commonly used by both CUL4A and CUL4B for interacting with different substrate receptors, was identified, along with CUL4B in the AhR-immouncomplex, but its function in promoting CUL4B-mediated substrate degradation remains unclear as it was not apparently required for substrate recruitment.
Notably, S. cerevisiae, which contains CUL1 and CUL3 orthologues as judged by both sequence conservation and binding with SKP1 and BTB proteins, respectively, does not contain a CUL4 ortholog. Instead, budding yeast express a unique cullin, Cul8 (also known as Rtt101), that, although not sharing primary sequence homology with CUL4, appears to assemble CRL4-like complexes and performs similar functions as CUL4 in other organisms. The cul8Δ cells are viable, but have a slower growth rate, are delayed in anaphase progression, are sensitive to fork arrest induced by DNA alkylation and accumulate DNA damage, suggesting a function for Cul8 in maintaining genomic integrity [25, 26]. Cul8 interacts with Mms1, a protein that was initially identified by its function in protecting against the DNA-alkylating agent methyl methanesulfonate (MMS) as well as DNA damage during replication. Mms1 protein shares low, but significant primary sequence similarity (18% with human DDB1) with DDB1, which is absent in budding yeast, and could fold into a DDB1-like β-propeller structure. Cul8 binds Mms1 through an N-terminal sequence analogous to the CUL4 DDB1-binding domain and at the same time Mms1 can recruit additional proteins such as Crt10 and Mms20 that might function as substrates or substrate receptors .
A direct DDB1–CUL4 association was first noted in a DDB2 immunocomplex . The significance of this binding was not immediately clear as nearly all studies on DDB1 at that time focused on its role in DNA repair and no study had linked DDB1 to protein ubiquitylation. Later, genetic analyses of DDB1 mutants in S. pombe, C. elegans, Arabidopsis, Drosophila and mouse (Box 1), revealed broad functions for DDB1 beyond DNA repair. Three subsequent studies firmly established a functional role of DDB1 in CUL4-mediated ubiquitylation thereby leading to the eventual discovery of DDB1-mediated assembly of multiple CUL4-based E3 complexes.
Deletion of cul4 in S. pombe (also known as pcu4) resulted in severe growth retardation, but viable cells that are extremely elongated with decondensed chromosomes . ddb1Δ fission yeast cells are viable, but are hypersensitive to various of DNA damage agents, are delayed in DNA replication progression, accumulate DNA damage, and show elongated phenotypes, abnormal nuclei and retarded growth [75, 88]. Many of these phenotypes are similar to those observed in cul4Δ cells , Notably, S. pombe contains an additional protein, Rik1 that shares 21% identity with DDB1 and also assembles E3 ligases with Cul4.
Unlike other organisms, Arabidopsis contains two DDB1 genes, AtDDB1A and AtDDB1B, complicating the assessment of the functional dependency of CUL4 on DDB1. AtDDB1B disruption is lethal whereas the only effect of homozygous loss of AtDDB1A is decreased UV tolerance for a short period of time immediately following UV irradiation [89, 90]. The inability of AtDDB1A to complement AtDDB1B is puzzling given that AtDDB1A is 91% identical to AtDDB1B over the entire 1088 residues, also binds to AtCUL4 , and is widely and concurrently expressed in many tissues; however AtDDB1B is often expressed at higher levels . This puzzle lacks a clear solution.
Ddb1 is allelic to a previously defined locus termed piccolo (pic) . Semi-lethal piccolo mutants were originally characterized based on shared irregularities in bristle, wing, and body segment growth . Ddb1 mutants growth arrest at the second instar stage, thus failing to develop completely .
In the mouse, Ddb1 deletion results in early embryo lethality before E12.5 . Conditional deletion of Ddb1 in the central nervous system (CNS) leads to neuronal and lens degeneration, brain hemorrhages, and neonatal death with strong selective elimination of nearly all proliferating neuronal progenitor cells and lens epithelial cells by apoptosis . Accumulation of DNA damage and selective killing of proliferating cells were also observed in epidermal cells in which Ddb1 was conditionally deleted .
DDB2 and CSA, encode two related WD40 proteins which control two different pathways of nucleotide excision repair, global genome repair (GGR) and transcription-coupled repair (TCR), and when mutated they lead to two distinct hereditary diseases, xeroderma pigmentosum and Cockayne syndrome, respectively. In a search for the molecular basis underlying the function of DDB2 and CSA in DNA repair, it was surprising to find that they assemble into similar complexes that contain DDB1, CUL4A, ROC1 and all 8 subunits of COP9 signlaosome . Separately, an investigation of human DET1, the homolog to Arabidopsis DET1a, a negative regulator of plant light responses, identified an E3 complex that contains CUL4A, ROC1, COP1 and DDB1 and promotes ubiquitylation and degradation of the proto-oncogenic transcription factor c-Jun . Furthermore, analyses of endogenous ROC1 and CUL4A complexes identified DDB1 as a near-stoichiometric CUL4A binding partner. In this study, in vivo and in vitro ubiquitylation assays established DDB1 as an essential component for the CDT1 ubiquitylation by the CUL4A– ROC1– E2 catalytic core . Together, these studies provided the basis for the subsequent discovery that DDB1 links multiple WD40 proteins to the CUL4– ROC1– E2 catalytic core.
When compared with both SKP1 (160 residues), elongin B (161 residues) and elongin C (112 residues), a notable feature of DDB1 (1140 residues) as the linker for CUL4 is its large size. This difference could reflect the need of DDB1 to interact with more divergent substrate receptors (WD40 proteins) than other linkers. It also raises the possibility that DDB1 might interact with additional groups of substrate receptors or substrates. The presence of additional DDB1-related proteins potentially expands the versatility of CUL4 in the assembly of E3 complexes. Computational analyses suggested that two mRNA processing factors, SAP130 (spliceosome-associated protein 130) and CPSF160 (cleavage and polyadenylation specificity factor 160), share significant primary sequence homology with DDB1 (16% and 20%, respectively), contain multiple WD40-like repeats (21 in SAP130 and 18 in CPSF160) and could fold into three β-propeller domains . Very little is known about CPSF160. SAP130, by contrast, has been identified in two related GCN5 histone acetyltransferase-containing transcription complexes, the STAGA and TFTC complexes [33, 34]. Recently, a direct connection of SAP130 with the cullin family was made by the demonstration of its binding with CSN and, surprisingly, with multiple cullins (CUL1, 2 and 4A) . SAP130 binds to both C-terminal and N-terminal sequences in CUL1, similar to another cullin regulator, CAND1 (cullin-associated and neddylation-dissociated 1), and can form a complex with the linkers and substrate receptors of both CUL1 (SKP1–SKP2) and CUL2 (elongin B and VHL). This finding, together with the presence of DDB1 in the SAP130-containing STAGA and TFTC complexes , suggests that SAP130, unlike DDB1, is not a CUL4-specific linker.
Fission yeast Rik1, initially identified by its function in maintaining gene silencing and heterochromatin formation, shares 21% identity with DDB1 and was co-purified with Cul4, Roc1, as well as Clr4 histone H3K9 methyltransferase and Lid2 H3K4 demethylase [11, 12, 36, 37]. Fission yeast Ddb1 was not found in Rik1-containing complexes, as expected if Rik1 replaces Ddb1 as a linker for Cul4 in these complexes. A WD40 protein, Clr8 (also called Dos1 or Raf1), co-purifies with the Rik1–Cul4 complex, is required for heterochromatin formation, and appears to function as a structural component in a manner analogous to a substrate receptor. Thus far, Rik1 is the only protein besides DDB1 that is known to function as a linker for CUL4.
Following the discoveries that CUL1, CUL2 and CUL3 each interacts with multiple substrate receptors and that CUL4 performs pleiotropic functions, it was anticipated that CUL4 very likely would also interact with a protein motif present in multiple proteins. Elucidating the CUL4-interacting motif, however, proved to be challenging and was made possible by the discovery of DDB1 as a major functional partner of CUL4. Taking different approaches -- proteomic, bioinformatic and structural analyses -- four independent studies collectively identified and experimentally demonstrated 50 different DDB1 binding WD40 proteins, referred to variously as DWD (DDB1-binding WD40), DCAF (DDB1–CUL4 associated factors) or CDW (CUL4–DDB1–associated WDR) proteins [38–41] and reviewed in . It was also clear that only a subset of, but not all WD40 proteins, interact with DDB1 and CUL4. DDB1-binding proteins contain a common motif, variably defined as the “double DxR box” with two DxR motifs located at the end of two consecutive WD40 repeats , the “DWD box”, a 16-residue stretch that correspond to the second half of a WD40 repeat and an Arg residue at position 16 following the WD dipeptide , the “DXXXR/KXWDXR/K” motif as the subdomain of WD40 repeats , or simply as the “WDXR” motif to emphasize the Arg residue following the WD dipeptide.
Three common features of DDB1-binding WD40 proteins emerge from these analyses. (1) Some DWD/DCAF/CDW proteins contain one, most contain two and a few contain three such motifs. (2) The Arg residue following the WD dipeptide is critically important and point mutation of this single residue in some DWD/DCAF/CDW proteins is sufficient to disrupt its binding with DDB1. Notably, point mutations of this Arg residue, Arg273, in the DWD box of DDB2 is found in several human patients with xeroderma pigmentosum group E (XP-E) and disrupts DDB2–DDB1binding . Unexpectedly, when the co-crystal structure of human DDB1 and zebrafish DDB2 revealed that the corresponding Arg residue in zebrafish DDB2 (Arg309) did not contact DDB1 , leaving the exact contribution of this Arg to DDB1 binding unclear. (3) Additional residues besides the Arg are required for binding with DDB1. For example, a 6-amino acid sequence in the DWD/DCAF/CDW protein WDR21 which lies outside the WD40 repeats is important for binding with DDB1 . This region was identified by sequence homology in some, but not all, DWD/DCAF/CDW proteins. Although the exact residues that contact DDB1 are not yet precisely defined, the current definition of the DWD/DCAF/CDW motif is highly predictive of and likely structurally important for DDB1 binding proteins.
How many DWD proteins are there? The human genome encodes about 320 unique WD40 proteins. Database searches using the DWD box predicted that as many as 90 DWD proteins, or a third of total WD40 proteins, could bind to DDB1 . Similar searches predicted 33 DWD proteins in fission yeast, 36 in C. elegans, 75 in Drosophila, 78 in rice and 85 in Arabidopsis [38, 46]. When 11 predicted Arabidopsis DWD proteins were tested for their direct interaction with DDB1 by yeast two-hybrid assay, all were found to be positive , suggesting that most of these predicted DWD proteins indeed bind to DDB1. Notably, 11 human proteins contain both an F-box and WD-40 repeats (FBXW proteins), including 8 with a predicted DWD box. The experimental demonstration that FBW5, as well as two additional FBXW proteins, APG16L and KATNB1 , can bind with DDB1 and CUL4 suggests that an individual substrate receptor (a FBXW protein) might target a substrate to two different E3 ligases, potentially further expanding the reach of CRLs.
At present, nearly two dozen proteins are reported to be degraded by a CRL4 (Table 1), including several that are not ubiquitylated by CRL4 during normal cell growth, but are diverted to CRL4 by a virus (Box 2). Evidence is solid for some and incomplete or correlative for others, and among the CRL4 substrates, CDT1 currently is the best characterized. Several interesting features unique to CRL4 emerged from these studies.
Many viruses are known to exploit the ubiquitylation pathway to evade innate cellular antiviral mechanisms or otherwise benefit viral propagation . CRL4s are targeted by several viruses including members of the paramyxovirus, herpesvirus, lentivirus, and hepadnavirus families. Although these diverse viruses commonly target DDB1, they appear to disrupt CRL4s in different ways. Viruses from the paramyxovirus family including simian virus 5, mumps virus and human parainfluenza virus type 2, use CUL4–DDB1 to degrade signal transducer and activator of transcription (STAT) proteins that respond to interferon signaling and initiate cellular antiviral responses [79, 80]. The crystal structure of simian virus 5 V (SV5-V) protein in complex with DDB1 showed that despite almost no primary sequence homology with DDB2, SV5-V protein similarly inserts a helix between β-propellers BPA and BPC . Hepatitis B virus X protein (HBx) also binds to DDB1; this binding is thought to facilitate viral replication [97, 98], possibly by causing an extended S-phase . Binding assays show that SV5-V and HBx bind DDB1 in a mutually exclusive manner . Furthermore, structural analysis revealed that HBx, as well as SV5-V contains primary sequence homology to DDB2 helix h1 that inserts into a cavity formed by BPA and BPC in DDB1 . Collectively, these structural analyses suggest that HBx and SV5-V viral proteins, although lacking WD40 repeats and sharing little primary sequence homology with each other, might form a similar structural motif and bind to a region in the BPA-BPC pocket of DDB1 in a manner similar to that of DDB2. The cellular protein(s) targeted by HBx to DDB1– CUL4 ligase remains unknown, raising the possibility that HBx facilitates viral propagation by interfering with the ubiquitylation of a host protein by, instead of targeting a novel host protein to, the DDB1– CUL4 ligase. This could also be the case for the murine gamma herpesvirus 68 M2 protein that binds to CUL4–DDB1 and impairs DNA damage response and inhibits apoptosis through unknown cellular mechanisms . Most recently, several groups have established that HIV-1 Vpr and the related simian immunodeficiency virus protein Vpx bind to CRL4VprBP. Though it remains unclear how Vpr–VprBP association promotes viral replication, the hijacked E3 ubiquitin ligase complex degrades uracil-DNA glycosylase UNG2, a substrate not targeted by CRL4VprBP in the absence of Vpr . Indeed, mechanisms for how CRL4 is exploited by so many divergent viruses remains to be shown in detail, but clearly this broadly expressed ubiquitin ligase is an attractive target.
A unique feature of CRL4CDT2 mediated CDT1 ubiquitylation is the requirement of proliferating cell nuclear antigen (PCNA) [47–50], a cofactor of DNA polymerases that forms a trimeric ring to encircle DNA and make a sliding clamp . The function of PCNA in CRL4CDT2-mediated ubiquitylation remains unclear, but is not limited to CDT1 alone. Recently, three additional CRL4CDT2 substrates, both human and C. elegans p21, a CDK inhibitor, [52–54], C. elegans polymerase η , and Drosophila E2f1 , have all been reported to be degraded in a PCNA-dependent manner. All four CRL4CDT2 substrates contain a conserved PCNA-interacting motif (PIP box) and bind PCNA directly, and this direct binding is required for their ubiquitylation. Although the CDT1–PCNA interaction is necessary to recruit the CRL4CDT2 E3 complex to chromatin , this effect has not been clearly demonstrated for the other three CDT2 substrates. Humans express 47 PIP box proteins, and most have chromatin-associated functions . It will be interesting and important to determine whether and how many other PIP box proteins are also ubiquitylated by CRL4CDT2. Further, no direct interaction between PCNA and CDT2 has been reported, leaving open the possibility that the function of PCNA could extend beyond CDT2 to other CRL4 ligases. It is tempting to speculate that PCNA provides the biochemical basis for the strong links between CUL4–DDB1 and chromatin biology .
Several studies have observed CRL4DDB2-dependent ubiquitylation of core histones and repair factor XPC during DNA repair [58, 59] as well as ubiquitylation of DDB2 by DDB1–CUL4 [60–63]. The consequences of these ubiquitylation events to the substrates and to the DNA repair process remains unclear. One model posits that via ubiquitylating histones and presumably loosening up chromatin , DDB2 is ubiquitylated and degraded, thereby facilitating the access of other repair factor, such as XPC, to the DNA lesion. XPC ubiquitylation, by contrast, enhances its DNA binding affinity, but does not promote its degradation . The view that DDB2 ubiquitylation promotes the binding of other factors to the lesion is challenged by evidence that loss of DDB2 ubiquitylation enhances rather than inhibits DNA repair activity as was seen in the absence of c-Abl  and more recently in cells lacking Cul4A . Perhaps, DDB2 ubiquitylation functions to restrict repair activity to the lesion instead of spreading to undamaged DNA. More intriguingly, it remains unclear how one E3 ligase (DDB1– CUL4– ROC1) apparently causes monoubiquitylation of some substrates (H2A, H3 and H4) and polyubiquitylation of others that leads to degradation in one case (DDB2), but not in another (XPC). It will be important to determine whether different ubiquitin lysine residues are linked to different substrate and if so, whether ROC1 can itself switch between different E2s or if additional factors control the ROC1–E2 binding to conjugate different ubiquitin lysines to different substrates.
Not all CRL4 substrates identified to date are associated with chromatin, as illustrated by the ubiquitylation of cytoplasmically localized TSC2 (tuberous sclerosis 2) by CRL4AFBW5 , membrane associated protein Merlin by CRL4AVprBP [66, 67], and ER-α in vitro by CRL4BAhR . Another issue related to CRL4 substrate that should be mentioned is that a DWD/DCAF/CDW protein might itself be a substrate, as has been suggested for DDB2 . In the case of SCF E3 ligases, several F-box proteins, including Cdc4 and Grr1 , are degraded by the CUL1–ROC1–E2 core after the substrate they recruited is degraded, presumably to ensure the release and sustain the supply of the CUL1–ROC1 core for the assembly with other F-box proteins into a new SCF E3 complex.
Given the number of demonstrated and predicted DWD/DCAF/CDW proteins and the number of proteins interacting with CUL4 and/or DDB1, there is good reason to anticipate that a potentially large number of CRL4 substrates will be identified in the near future. We will also likely know the answers to three challenging issues related to CRL4. What is the biochemical function of PCNA in promoting CRL4CDT2 mediated substrate ubiquitylation? How does a single DDB1– CUL4– ROC1 E3 complex differentially cause monoubiquitylation as well as non-proteolytic and proteolytic polyubiquitylation? What are the biochemical mechanisms by which CRL4s contribute to gene silencing, heterochromatin formation and chromosome condensation? Answers to these issues will also have broad implications beyond CRL4 biology and should help to advance the fields of ubiquitin and chromatin biology.
We thank Xing-Wang Deng, Robert Durionio, Edward Kipreos, Andrew Neuwald, Pradip Raychaudhuri, Dana Schroeder, Cang Yong and Pengbo Zhu for insightful discussion and sharing unpublished information during the preparation of this manuscript, and past and current members of Xiong lab for their contribution over the years. This study is supported by an NIH grant (GM067113) to Y. X.