VHL: A very hip ligase

A deluge of recent evidence suggests that the abundance, and therefore activity, of oncogene and tumor suppressor gene products is controlled by the ubiquitin-proteasome degradation system. Cancer targets that are regulated by the ubiquitin system range from cell cycle regulators to signaling proteins to transcription factors (1). These and many other substrates are targeted for degradation upon the covalent attachment of the small protein ubiquitin, an event catalyzed by the now-famous cascade of ubiquitin transferase enzymes, E1 → E2 → E3 (1). Assembly of a polyubiquitin chain on the substrate leads to its capture and degradation by an abundant protease particle, the 26 S proteasome (2). The crucial substrate recognition step in ubiquitin-dependent proteolysis is mediated by the diverse family of E3 enzymes, also known as ubiquitin ligases.

In many instances, oncogenic mutations stabilize the encoded gene products, yet a component of the ubiquitin machinery has never been directly implicated in tumor suppression. But now reports by Iwai et al. in this issue of PNAS (3), and by Lisztwan et al. in a recent issue of Genes and Development (4), tightly link the notorious VHL tumor suppressor protein to ubiquitin ligase activity. In von Hippel-Lindau (VHL) syndrome, mutation of one germ-line copy of the VHL gene predisposes individuals to a wide range of tumors, including renal cell carcinoma, pheochromocytoma, cerebellar hemangioblastomas, and retinal angioma, which arise upon somatic loss of the remaining wild-type VHL gene, in accord with the classical two-hit definition of a tumor suppressor (5, 6). In addition, both somatic copies of the VHL gene are mutated in the majority of sporadic clear cell renal carcinomas (6). A hallmark of VHL^−/− tumors is their high degree of vascularization, which arises from constitutive expression of a suite of hypoxia-inducible genes, including the crucial vascular endothelial growth factor (VEGF) (6). VHL also is required for cell cycle exit upon serum withdrawal and so may serve as a gatekeeper in the proliferation of renal cells (7).

Hints that VHL might tie into the ubiquitin system have come at a furious pace over the past couple of years, mainly through structural analogies to a recently described class of ubiquitin ligases termed Skp1-Cdc53/CUL1-F-box protein (SCF) complexes. In SCF complexes, the Skp1 subunit links any one of a set of adaptor molecules called F-box proteins to a core ubiquitination complex, composed of the scaffold protein Cdc53/CUL1, the RING-H2 finger protein Rbx1/ROC1 and, typically, the E2 enzyme Cdc34 (see Fig. 1) (8, 9). An N-terminal ∼40-residue motif called the F-box serves as a Skp1 binding site within the large family of otherwise unrelated F-box proteins, which in turn recruit various substrates for ubiquitination through dedicated protein–protein interaction domains (8, 10). SCF complexes regulate diverse processes in the cell cycle, signaling and development in all eukaryotes, from yeast to humans. Known targets of SCF complexes in mammalian cells include the cytoplasmic anchor protein IκBα, the transcription factors β-catenin and E2F-1, and the cyclin-dependent kinase inhibitor p27^Kip1, all of which are targeted for degradation by substrate level phosphorylation (11, 12). The SCF family has diversified through evolution, as a number of Cdc53 homologs (called cullins in metazoans), Skp1 homologs, and indeed hundreds of F-box containing proteins can be found in sequence databases (8, 9).

Figure 1.

The SCF and VCB-like ubiquitin ligase complexes. Homologous components are indicated in the same color. Only two representative adaptor proteins for each complex are shown, and not all candidate substrates are shown for each adaptor. Each complex is able to use at least two different E2 enzymes, Cdc34 and Ubc5. Substrate interaction domains shown for VCB-like complexes have not been unequivocally demonstrated. The shaded patch marked C indicates a region of similarity shared by all cullins. F indicates F-box, S indicates SOCS box.

VHL interacts with the proteins ElonginC, ElonginB, and CUL2 in a complex referred to as VCB-CUL2 (see Fig. 1) (6). The first evidence of structural analogy between the VCB complex and SCF complexes came from the sequence similarity between Skp1 and ElonginC, and between Cdc53/CUL1 and CUL2 (8, 10). Like Skp1, ElonginC appears to bridge VHL to the cullin subunit (6). A crucial piece in the puzzle fell into place with the identification of a small RING-H2 finger protein Rbx1/ROC1 as both a VHL-associated protein and as an essential component of SCF complexes (9). Rbx1 interacts with at least five mammalian cullins, including CUL1 and CUL2. In SCF complexes Rbx1 appears to stabilize the cullin-E2 interaction and stimulate the activity of the E2 enzyme (9). A final loose structural analogy can be drawn between the F-box and the ElonginBC binding site in VHL, called the BC box, which is embedded within a larger motif, the SOCS box (13). The SOCS box was spotted in a family of proteins that regulate cytokine signaling, known as suppressors of cytokine signaling (SOCS), which are defined by a solitary Src homology 2 domain, followed by a C-terminal ∼50-residue SOCS box sequence (14). Missense mutations in the BC box disrupt both VHL- and SOCS box protein–ElonginBC interactions (15). The SOCS box motif occurs in numerous proteins of unknown function, many of which contain other protein–protein interaction domains (15, 16). Thus like SCF complexes, ElonginBC-CUL2-based complexes may have developed a repertoire of adaptor subunits to target a wide range of different substrates. Indeed, because the other main cell cycle ubiquitin ligase, the anaphase promoting complex/cyclosome (APC/C), also contains a RING-H2 subunit, a cullin subunit, and multiple adaptor subunits, it appears that the SCF, VCB-like, and APC/C complexes comprise an E3 superfamily (9).

The recently determined three-dimensional structure of the VCB complex provided several key insights into VHL function (13). VHL has two domains: a 100- residue N-terminal domain composed mainly of β-sheets (the β-domain) and a smaller C-terminal domain composed mainly of α-helices (the α-domain). The α-domain consists of three α-helices, including one derived from the BC box, which combine with an α-helix donated by ElonginC to form a stable intermolecular four-helical bundle at the interaction surface. The β-domain is on the opposite side of the α-domain, where it is presumably free to contact other proteins that may interact with the complex. Strikingly, the large collection of tumor-derived VHL mutations defines two prominent surface patches: one matches the predicted ElonginC interaction site within the SOCS box region, the other is a possible substrate interaction surface on the β-domain (13). Taken together, these clues suggested that mutations in VHL might uncouple it, or its substrates, from a core ubiquitin ligase complex and thereby cause the accumulation of cellular proteins that lead to transformation.

The structural analogies between SCF and VCB complexes set the stage for the discovery of VHL-associated ubiquitin ligase activity (3, 4). This task might have been nearly impossible in the absence of a relevant substrate but, fortunately, VHL brings its own substrates along for the ride. As measured against endogenous VHL-associated proteins, ubiquitin ligase activity can be captured from cell extracts with recombinant VHL fusion proteins, VHL immune complexes, and recombinant VHL-ElonginBC-CUL2 complexes purified from insect cells (3, 4). In each case, ubiquitination activity is manifest only in the presence of an E2 enzyme, either Cdc34 or Ubc5. Furthermore peptides that correspond to the BC box sequence, but not a mutant BC box peptide, are able to outcompete VHL protein for ElonginBC, CUL2, and ubiquitin ligase activity (4). In general, high molecular weight polyubiquitinated species are preferentially formed only by complexes captured with wild-type VHL protein and not with tumor-derived VHL mutants or truncated VHL proteins. As for SCF complexes, Rbx1 stimulates the activity of recombinant VCB-CUL2 complex, suggesting that Rbx1 (or its closely related homologs) may indeed be a common cofactor for all cullin-based ubiquitin ligase complexes (3, 9).

With VHL-associated ubiquitin ligase activity in hand, the game is afoot to identify the physiologically relevant VCB substrates, which presumably include the products formed during the in vitro ubiquitination reactions (Fig. 2). Tantalizingly, Iwai et al. (3) detect two candidate substrates, designated p100 and p220, among the reaction products of their assay and in far-Western blots with a radiolabeled VCB probe. The p100 and p220 signals are abrogated by VHL mutations that disrupt either the ElonginC interaction or the presumptive substrate binding patch on the β-domain. Currently, the most attractive candidate target of the VCB-CUL2 ubiquitin ligase is the hypoxia-inducible transcription factor HIF-1α, and its homolog HIF-2α, which together with the constitutively expressed subunit HIF-1β, form the HIF-1 transcriptional complex responsible for activation of genes involved in metabolism, angiogenesis, and apoptosis (6). The HIF-1α subunit is rapidly degraded by the proteasome under normoxic conditions, but is stabilized by hypoxia (17). Recently, it has been shown that HIF-1α binds to VHL and that HIF-1α persists at normal oxygen levels in VHL-deficient cells, making it a likely target of VHL (18). The mechanism that regulates HIF-1α stability appears to be linked directly to oxygen status by a ferroprotein because treatment of cells with iron chelators or iron mimetics such as cobalt stabilizes HIF-1α and eliminates its interaction with VHL (18). However, if HIF-1α is indeed targeted by VHL, regulation is apparently not at the level of VHL binding because VHL-HIF-1α complexes can be detected under both normoxic and hypoxic conditions (18). The predicted size of HIF-1α is close to that of p100, but as yet the presence of HIF-1α in the polyubiquitinated in vitro products has not been tested. Efforts to reconstitute ubiquitin ligase activity against HIF-1α are no doubt underway.

Figure 2.

VHL may target HIF-1α and other substrates for degradation under conditions of oxygen sufficiency. The HIF-1 transcriptional complex is composed of the hypoxia-inducible subunit HIF-1α and the constitutively expressed subunit HIF-1β, which activate hypoxia-inducible genes, such as VEGF, in conjunction with other transcription factors, including SP1. Blunt arrows indicate inhibition, pointed arrows indicate activation, and question marks indicate that substrate or significance is unknown. See text for details.

In keeping with the pleiotropic defects in VHL^−/− cells, the VCB-CUL2 ligase may target sundry other proteins for degradation. The persistence of HIF-1 activity under normoxic conditions may not be the only mechanism that deregulates hypoxia-inducible genes in VHL^−/− tumors as hypoxia-induced mRNAs such as VEGF also are more stable in VHL^−/− cells (6). VHL therefore might target RNA binding proteins that destabilize hypoxia-induced mRNAs. Alternatively, RNA binding proteins that stabilize hypoxia-induced transcripts may be under the direct control of HIF-1. Another possible VHL substrate is the extracellular matrix protein fibronectin, which interacts with VHL in a manner that is sensitive to VHL mutations (19). The VHL-fibronectin interaction appears to be physiologically relevant because VHL^−/− mouse embryos show defects in deposition of fibronectin into the extracellular matrix (19). Although fibronectin is of a similar size as p220, it remains to be seen whether the effects of VHL on fibronectin are directly mediated by ubiquitin ligase activity (3). Finally, because VHL^−/− cells fail to exit the division cycle in response to low serum levels, one or more substrates of VHL may prevent entry into the quiescent state (7).

Now that the VCB-CUL2 complex is a card-carrying ubiquitin ligase, it seems likely other SOCS box proteins that interact with the ElonginBC-CUL2 core complex will follow suit. Many interacting partners have been defined for SOCS family proteins, at least some of which may be substrates. The archetypal SOCS protein, SOCS-1, binds to JAK kinases and the Kit receptor tyrosine kinase through its Src homology (SH) 2 domain, to SH3-containing adapter proteins such as Grb2 and p85 though N-terminal proline-rich sequences, and to the hematopoietic specific guanine nucleotide exchange factor VAV through as-yet-undefined sequences (20). Intriguingly, SOCS-1 destabilizes VAV and, moreover, SOCS-1 suppresses the transforming activity of the truncated onco-VAV protein in a SOCS box-dependent manner (R.R., unpublished data). In another example, a SOCS family protein called CIS appears to suppress erythropoietin signaling by binding to the activated erythropoietin receptor and targeting it for ubiquitin dependent degradation (21). Whether or not these effects are actually mediated by the putative SOCS box protein-ElonginBC-CUL2 ligase complexes remains to be seen.

Returning to VHL itself, the newly discovered ubiquitin ligase activity associated with VHL suggests possible explanations for the intricate pathology of the VHL syndrome. Because VHL resides in a multiprotein enzyme complex, the genetic modifiers of the VHL phenotype (6) may arise from normally innocuous mutations in companion components of the complex, analogous to genetic interactions observed between SCF components (8). Allele-specific effects on different substrates also may help explain the association of certain VHL mutations with particular tumor phenotypes. One hopes the new insights into VHL function will suggest a means to combat the aggressive tumors that characterize the VHL syndrome. Inevitably, the issue will turn on the physiologically relevant substrates of the VCB-CUL2 ubiquitin ligase complex.