Structural basis for the recognition of hydroxyproline in HIF-1α by pVHL

Article metrics


Hypoxia-inducible factor-1 (HIF-1) is a transcriptional complex that controls cellular and systemic homeostatic responses to oxygen availability1. HIF-1α is the oxygen-regulated subunit of HIF-1, an αβ heterodimeric complex1. HIF-1α is stable in hypoxia, but in the presence of oxygen it is targeted for proteasomal degradation by the ubiquitination complex pVHL, the protein of the von Hippel–Lindau (VHL) tumour suppressor gene and a component of an E3 ubiquitin ligase complex2,3. Capture of HIF-1α by pVHL is regulated by hydroxylation of specific prolyl residues in two functionally independent regions of HIF-1α4,5,6,7. The crystal structure of a hydroxylated HIF-1α peptide bound to VCB (pVHL, elongins C and B) and solution binding assays reveal a single, conserved hydroxyproline-binding pocket in pVHL. Optimized hydrogen bonding to the buried hydroxyprolyl group confers precise discrimination between hydroxylated and unmodified prolyl residues. This mechanism provides a new focus for development of therapeutic agents to modulate cellular responses to hypoxia.


Ubiquitin-mediated degradation of specific proteins is central in the regulation of many cellular events8. Such processes require mechanisms for specific recognition of proteolytic targets by E3 ubiquitin ligases. Several multisubunit, cullin-containing, RING E3 ubiquitin ligases use phosphorylation to regulate capture of protein substrates9. In contrast, recognition of HIF-1α by VCB E3 ligase (pVHL, elongins B and C, Cul2 and Rbx1; ref. 10) requires hydroxylation of specific prolyl residues by a group of 2-oxoglutarate-dependent, non-haem dioxygenases11,12. In human HIF-1α, hydroxylation of two functionally independent prolyl residues, 402 and 564—which are within amino- and carboxy-terminal oxygen-dependent degradation (NODD and CODD) motifs7,13,14—causes direct and highly efficient interaction with pVHL. We here address how the insertion of a single oxygen atom into a prolyl residue governs selective recognition, by examining the crystal structure of human VCB in complex with a peptide containing the human HIF-1α CODD region (residues 549–582) that has a 4(R)-l-hydroxyproline (Hyp) as residue 564.

The structure was determined by molecular replacement using the apo form of VCB15 as a search model. Reflection data from 30 to 2.0 Å were used for model rebuilding and refinement. Electron density for HIF-1α peptide residues 560–577 is of good quality, with only the side chain of E560 poorly defined (Fig. 1). The peptide lies in an extended conformation across one side of the β-domain of pVHL (Fig. 2a, b), binding to one of the β-sheets as if it were a complementary β-strand, although there is no extensive main-chain hydrogen bonding to define it as such. The binding does not involve conformational changes to the main chain of pVHL (r.m.s. deviation of Cα atoms is 0.3 Å between the apo and peptide-bound forms of the β-domain). There is some rearrangement of the three VCB subunits compared to the apo structure (reflecting differing crystal contacts), but the alterations are distant from the CODD binding site.

Figure 1: Electron density for the bound CODD peptide.

The boomerang-shaped CODD peptide is shown as a stick model (where the carbon, nitrogen, oxygen and sulphur atoms are grey, blue, red and yellow, respectively). The difference electron density (contoured at 1.5σ) was calculated using phases derived from the model after refinement by simulated annealing, with CODD omitted to remove model bias.

Figure 2: Structure of VCB–CODD complex, and sequence alignments of HIF ODD regions and pVHL.

a, b, Surface rendering of the VCB–CODD complex with the secondary structures revealed underneath the semitransparent surface. Two regions seen in the current structure but not included in the previously reported model of apo VCB15 are circled in elongin B (residues 99–104) and pVHL (residues 205–210). The C-terminal tail of elongin B makes newly observed contacts with the α domain (V170 and N174) of pVHL. The still-unobserved region (residues 50–57) in elongin C is modelled as a dotted loop. c, d, Sequence alignments of the CODD-binding region of pVHL β domain (c) and of the ODD motifs of HIF homologues (d). Residues that form van der Waals (VDW) contacts are indicated by ‘v’. ‘w’ denotes single hydrogen bonds (H-bonds) formed with a key water molecule (see Fig. 3d), as opposed to those formed between pVHL and CODD (see Fig. 3a, b). Arrows denote β-strands. VHL disease missense mutations17 that occur in the CODD-binding region of pVHL are listed.

Within the CODD peptide, residues 560–567 and 571–577 form distinct binding sites (sites 1 and 2, respectively). Between these sites only the side chain of D569 contacts pVHL. The binding surface on pVHL spans residues 67–117 of the β domain (Fig. 2a–c). The buried surface area (1,850 Å2) and the shape complementarity16 (0.79) between CODD and pVHL are comparable to those of the inter-subunit interfaces in VCB (2,050 and 2,300 Å2; 0.74 and 0.71). The pVHL–peptide interactions are predominantly nonpolar but also include ten electrostatic bonds (Figs 2c, d, and 3a, b).

Figure 3: CODD–pVHL interactions.

a, b, Views of site 1 (a) and site 2 (b). CODD and CODD-contacting residues of pVHL are shown as stick models (colour scheme as in Fig. 1), whereas pVHL, elongin C and elongin B are rendered as van der Waals surfaces using the colour scheme of Fig. 2. Dotted lines represent electrostatic bonds. Residues for pVHL are labelled in white; those for CODD are in black. c, Views of the Hyp-binding pocket, illustrating its surface complementarity. The van der Waals surface of the ODD motif, LAHyp, is rendered solid and coloured by atom type (as in Fig. 1). The surface of pVHL is rendered as a mesh object with both the front and rear planes clipped away. Inset, side view of Hyp. d, The hydrogen-bonding network involved in binding of the Hyp564 hydroxyl group. The side chain of L562 (CODD) has been omitted for clarity. The red sphere in a, c and d represents a key water molecule. a has an approximately similar orientation to that of Fig. 2a, whereas the view in b is rotated from that in a by 70 degrees about the vertical axis.

Site 1 (EMLAHypYIP) is the primary binding site, contributing more than 60% of the total buried surface area and van der Waals contacts. It contains the leucine–alanine–proline (LAP) motif found in HIF α-subunits, where P is converted to Hyp4,5,6,7,12 (Fig. 2d). Two residues, HypY, contribute all five hydrogen bonds in this site. The peptide-binding cavity on pVHL is, in general, rather shallow, with LAHypYIP following the curvature of the binding surface (Fig. 3a). The binding surface for site 2 (DFQLRSF) is relatively flat and only one residue of the peptide makes specific side-chain interactions (Fig. 3b), arguing against extensive sequence specificity in this region. R575 does not bind pVHL, splitting site 2 into two subsites. Five electrostatic bonds are contributed by the subsite DFQL; two are salt bridges formed by D571, and the rest are hydrogen bonds involving the main-chain atoms of F572 and L574. FQL is conserved as a nonpolar–polar–nonpolar motif: LDF in the NODD motif and MQM in the Caenorhabditis elegans ODD motif (Fig. 2d).

Hyp564 is almost entirely buried in a deep pocket in the site 1 binding cavity (Fig. 3c), contributing 20% of the total van der Waals contacts between the peptide and pVHL. The shape of this pocket (the only side-chain-specific cavity in the CODD-binding region of pVHL) complements precisely the up-pucker conformation of the hydroxyproline ring. The pocket is lined by W88, Y98, S111, H115 and W117, which are invariant between human and C. elegans pVHL (except for S111, which is conservatively substituted by threonine) (Fig. 2c). Residues S111 and H115 are sunk within the β-domain, with the exposed S111 hydroxyl group and H115 imidazole amino group (atom N3/Nδ) forming the floor of the Hyp-binding pocket and serving as hydrogen-bonding partners to the Hyp564 hydroxyl group (Fig. 3a, c). These two hydrogen bonds are of optimal distances and geometry (Fig. 3d). The positions of the five Hyp-binding residues are conserved between the apo and peptide-bound structures, locked by interdependent side-chain packing. In addition, the side-chain orientation of H115 is fixed by van der Waals contacts with S65, Y112, R113 and G114. H115 is part of a water-mediated hydrogen-bonding network (Fig. 3d), involving a water molecule buried in a pocket neighbouring the Hyp-binding pocket (Fig. 3a, c). This water is also present in the apo structure. It coordinates the H115 imidazole amino group distal to Hyp564, the side chain of N67 (which is functionally conserved as a histidine in C. elegans pVHL; Fig. 2c), the main-chain amide nitrogen of R69 and the carbonyl oxygen of L562 (CODD). By hydrogen bonding to the opposed imidazole amino group of H115, the water molecule stabilizes the Hyp-binding imidazole amino group as a conjugate base. The resulting partial charge may strengthen the hydrogen bond, rendering H115 the hydrogen bond acceptor for Hyp564, with S111 acting as a hydrogen bond donor. The S111 hydroxyl group is in turn a hydrogen bond acceptor of the indole nitrogen of W117. Hence Hyp564 completes an elaborate network of hydrogen bonds, providing a chemical basis for efficient capture of HIF-1α after oxygen-dependent hydroxylation. Although the physiological importance of this process is established, its relationship to pVHL tumour suppressor function is less clear. It is noteworthy that all five pVHL residues lining the Hyp-binding pocket are affected by missense mutations in VHL disease17 (Fig. 2c), suggesting that failure to capture HIF-1α, and/or another hydroxylated target, is important to the tumour-promoting mechanism associated with VHL disease.

Surface plasmon resonance data were measured for VCB binding to a range of peptides (Table 1 and Supplementary Information). Hydroxylated human CODD and NODD, and C. elegans ODD peptides, bind VCB with similar kinetics and binding affinities (29–53 nM). Moreover, competition studies indicate that human CODD and NODD bind to the same site. Both hydroxylated human CODD and a short LAHypYIP peptide derived from site 1 inhibit binding of VCB to human NODD (half-maximal inhibitory concentration, IC50, about 1.0 and 1.3 µM, respectively). Peptides of human CODD and NODD containing just site 1 have similar binding kinetics and only slightly weaker binding than their counterparts containing sites 1 and 2, demonstrating that site 1 drives the binding, as expected from the crystal structure.

Table 1 Binding kinetics of VCB to HIF-1α peptides as assayed by surface plasmon resonance

The similar affinity of pVHL for the very different ODD peptides is consistent with only Hyp occupying a side-chain-specific pocket. Binding of both CODD and NODD to a single pVHL site is also consistent with in vivo observations that each pVHL binding motif on HIF-1α can function in isolation, and that overexpression of either peptide can overcome endogenous HIF-1α degradation (C. Willam and C.W.P., unpublished results).

No stable complexes could be detected between VCB and unhydroxylated CODD for protein concentrations up to 45 µM. The steady-state affinity constant was estimated to be 34 µM (see Supplementary Information), which is about 1,000-fold less than the affinity for hydroxylated CODD. In principle, this discrimination could derive from specific binding effects and/or from conformational changes in either the pVHL or peptides. The crystallographic results show no significant conformational changes in pVHL on binding the high-affinity peptide, and 1H-nuclear magnetic resonance (NMR) experiments on the hydroxylated and unmodified peptides indicate that they adopt similar conformations in solution (see Supplementary Information). Apart from minor entropic effects due to differences in the torsional properties of proline and hydroxyproline, all of the discrimination must arise, therefore, from the interaction with pVHL. The drastic decrease in binding affinity owing to the absence of a hydroxyl group translates to about 4 kcal mol-1 of free energy change, which is consistent with the loss of two hydrogen bonds. Indeed, Hyp binding is stabilized by two hydrogen bonds, which would be denied to proline, and although proline could fit into the Hyp-binding pocket, it would exclude a water molecule that hydrogen bonds to the rigidly positioned S111 and H115 in the ligand-lacking structure. There would therefore be a net loss of two hydrogen bonds on binding the unhydroxylated peptide compared with the hydroxylated peptide. Hence, the specificity for Hyp is based on the special positions and environment of its hydrogen-bonding partners. This mechanism is reminiscent of that used by tyrosyl transfer RNA synthetase to achieve similar discrimination between tyrosine and phenylalanine18.

The central role of the HIF/pVHL system in a range of responses to hypoxia, including angiogenesis and metabolic adaptation19, has provided a new impetus for the therapy of ischaemic diseases through the activation of HIF. Transgenic expression of stabilized HIF-1α molecules that lack the pVHL recognition motifs leads to hypervascularity without leakage or inflammation20, indicating that proteolytic escape of HIF-1α can drive a productive angiogenic response. The specificity intrinsic to the rigid hydroxyproline-binding pocket of pVHL revealed here suggests that it could be targeted for development of small molecules to block HIF-1α recognition by pVHL.


Peptide synthesis

Biotinylated and unmodified human CODD peptides (residues 549–582) were synthesized by Genemed Synthesis. The other peptides were synthesized by Biopeptide. Peptide purity was assessed by mass spectrometry and high-performance liquid chromatography.

Protein expression, purification and crystallization

The proteins pVHL, elongin B and elongin C were co-expressed in Escherichia coli using vectors provided by N. Pavletich. The protein complex was purified essentially as described previously15. Purified VCB was mixed with excess biotinylated and hydroxylated CODD peptide, and was further purified on a Superdex 200 size-exclusion column (Pharmacia) in buffer containing 5 mM HEPES (pH 7.4), 200 mM NaCl and 1 mM dithiothreitol. Crystallization of the heterotetrameric complex (at 8 mg ml-1) was carried out by sitting-drop vapour diffusion at 22 °C. The mother liquor consisted of 100 mM HEPES (pH 7.4), 40% polyethylene glycol 2000 monomethylester and 200 mM (NH4)2SO4.

Crystallographic data collection and structure refinement

Two native data sets were collected. The crystals were cooled in their mother liquor to 100 K. Crystallographic data were indexed and scaled using the HKL2000 package21. The crystals belonged to the space group P43212 with one complex per asymmetric unit. A data set, to 3.2 Å resolution, was collected with a Rigaku X-ray generator (100 mA and 50 kV) and a MAR345 imaging plate detector. Coordinates for the VCB complex (Protein Data Bank code 1VCB) yielded a molecular replacement solution (program EPMR22). A difference map revealed the peptide. Data to 2.0 Å resolution were subsequently collected on beamline ID14-2 at the European Synchrotron Radiation Facility using an ADSC Quantum 4 detector. Cycles of rebuilding and refinement were carried out using the programs O23 and CNS24. The current Rcryst and Rfree are 22.6% and 27.8%, respectively (r.m.s. deviations from ideal bond lengths and angles are 0.006 Å and 1.3°, respectively). Other statistics are listed in Supplementary Information. The following residues are missing in the current model: 105–118 in elongin B, 50–57 in elongin C, 54–60 and 211–213 in pVHL, 549–559 and 578–582 in CODD. The backbone geometry of the refined model is satisfactory, with 89.3% of the residues in favourable conformations (program PROCHECK25). Two residues, H10 and D47 of elongin B, are in the disallowed region. The electron density around these two residues is excellent and their unusual main-chain geometry is due to their structural role in stabilizing two β-turns.

Structural analyses and graphics presentation

Solvent-accessible surface areas were calculated by NACCESS26. Least-squares structural alignment was performed with LSQMAN27. All structural graphic presentations were rendered with PyMOL ( Intermolecular van der Waals contacts and hydrogen bonds were analysed by CONTACT, and surface complementarity was calculated by Sc16 (a score of 1 is assigned to perfect complementarity, and 0 to none); both programs are implemented in CCP428.

Surface plasmon resonance

Experiments were performed in duplicate at 25 °C on a BIAcore 2000 using streptavidin-coated sensor chips and analysed using the BIAevaluation software (version 3.0) (BIAcore AB). The data were collected at a rate of 10 Hz, and all analyses were performed with VCB complexes injected over biotinylated peptides immobilized on the sensor chips. Peptides for site 1 were synthesized with an extra N-terminal cysteine residue to which biotin tags were chemically attached using a biotin-PEO-maleimide linker (which contains a spacer arm of 29 Å; Perbio Science), whereas peptides containing both sites 1 and 2 were synthesized with biotin attached directly at the N terminus. The concentration of the VCB complex was estimated by absorbance at 280 nm, using the theoretical extinction coefficient 24,540 M-1 cm-1. Kinetics experiments used a high flow rate (100 µl min-1) and a low surface density of ligand to minimize mass transport limitation29. Each experiment was performed with five dilutions of the VCB complex: 4.88, 2.44, 1.22, 0.610 and 0.305 nM. Association and dissociation constants were obtained by global fitting of the data to a 1:1 Langmuir binding model. For the competition experiments, the VCB complex was pre-incubated briefly at room temperature with various concentrations of nonbiotinylated peptides. The relative rate of binding estimated from the initial slope of the binding curves was used as a measure of inhibition. For the steady-state binding experiment, responses of VCB binding to immobilized, unhydroxylated CODD were subtracted from binding to a blank cell over a range of concentrations.


  1. 1

    Wang, G. L., Jiang, B. H., Rue, E. A. & Semenza, G. L. Hypoxia-inducible factor 1 is a basic-helix-loop-helix-PAS heterodimer regulated by cellular O2 tension. Proc. Natl Acad. Sci. USA 92, 5510–5514 (1995)

  2. 2

    Maxwell, P. H. et al. The tumour suppressor protein VHL targets hypoxia-inducible factors for oxygen-dependent proteolysis. Nature 399, 271–275 (1999)

  3. 3

    Ohh, M. et al. Ubiquitination of hypoxia-inducible factor requires direct binding to the β-domain of the von Hippel-Lindau protein. Nature Cell Biol. 2, 423–427 (2000)

  4. 4

    Ivan, M. et al. HIFα targeted for VHL-mediated destruction by proline hydroxylation: implications for O2 sensing. Science 292, 464–468 (2001)

  5. 5

    Jaakkola, P. et al. Targeting of HIF-α to the von Hippel-Lindau ubiquitylation complex by O2-regulated prolyl hydroxylation. Science 292, 468–472 (2001)

  6. 6

    Yu, F., White, S. B., Zhao, Q. & Lee, F. S. HIF-1α binding to VHL is regulated by stimulus-sensitive proline hydroxylation. Proc. Natl Acad. Sci. USA 98, 9630–9635 (2001)

  7. 7

    Masson, N., Willam, C., Maxwell, P. H., Pugh, C. W. & Ratcliffe, P. J. Independent function of two destruction domains in hypoxia-inducible factor-α chains activated by prolyl hydroxylation. EMBO J. 20, 5197–5206 (2001)

  8. 8

    Hershko, A. & Ciechanover, A. The ubiquitin system. Annu. Rev. Biochem. 67, 425–479 (1998)

  9. 9

    Tyers, M. & Jorgensen, P. Proteolysis and the cell cycle: with this RING I do thee destroy. Curr. Opin. Genet. Dev. 10, 54–64 (2000)

  10. 10

    Kamura, T. et al. Rbx1, a component of the VHL tumour suppressor complex and SCF ubiquitin ligase. Science 284, 657–661 (1999)

  11. 11

    Bruick, R. K. & McKnight, S. L. A conserved family of prolyl-4-hydroxylases that modify HIF. Science 294, 1337–1340 (2001)

  12. 12

    Epstein, A. C. et al. C. elegans EGL-9 and mammalian homologs define a family of dioxygenases that regulate HIF by prolyl hydroxylation. Cell 107, 43–54 (2001)

  13. 13

    Pugh, C. W., O'Rourke, J. F., Nagao, M., Gleadle, J. M. & Ratcliffe, P. J. Activation of hypoxia-inducible factor-1; definition of regulatory domains within the α subunit. J. Biol. Chem. 272, 11205–11214 (1997)

  14. 14

    Huang, L. E., Gu, J., Schau, M. & Bunn, H. F. Regulation of hypoxia-inducible factor 1α is mediated by an O2-dependent degradation domain via the ubiquitin-proteasome pathway. Proc. Natl Acad. Sci. USA 95, 7987–7992 (1998)

  15. 15

    Stebbins, C. E., Kaelin, W. G. Jr & Pavletich, N. P. Structure of the VHL–ElonginC–ElonginB complex: implications for VHL tumour suppressor function. Science 284, 455–461 (1999)

  16. 16

    Lawrence, M. C. & Colman, P. M. Shape complementarity at protein/protein interfaces. J. Mol. Biol. 234, 946–950 (1993)

  17. 17

    Beroud, C. et al. Software and database for the analysis of mutations in the VHL gene. Nucleic Acids Res. 26, 256–258 (1998)

  18. 18

    Fersht, A. R. et al. Hydrogen bonding and biological specificity analysed by protein engineering. Nature 314, 235–238 (1985)

  19. 19

    Semenza, G. L. Regulation of mammalian O2 homeostasis by hypoxia-inducible factor 1. Annu. Rev. Cell Dev. Biol. 15, 551–578 (1999)

  20. 20

    Elson, D. A. et al. Induction of hypervascularity without leakage or inflammation in transgenic mice overexpressing hypoxia-inducible factor-1α. Genes Dev. 15, 2520–2532 (2001)

  21. 21

    Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326

  22. 22

    Kissinger, C. R., Gehlhaar, D. K. & Fogel, D. B. Rapid automated molecular replacement by evolutionary search. Acta Crystallogr. D 55, 484–491 (1999)

  23. 23

    Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M. Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110–119 (1991)

  24. 24

    Brunger, A. T. et al. Crystallography & NMR system: A new software suite for macromolecular structure determination. Acta Crystallogr. D 54, 905–921 (1998)

  25. 25

    Laskowski, R. A., Rullmannn, J. A., MacArthur, M. W., Kaptein, R. & Thornton, J. M. AQUA and PROCHECK-NMR: programs for checking the quality of protein structures solved by NMR. J. Biomol. NMR 8, 477–486 (1996)

  26. 26

    Hubbard, S. J. & Thornton, J. M. ‘NACCESS’, Computer Program Version 2.1.1 (Department of Biochemistry and Molecular Biology, Univ. Coll. London, London, 1996)

  27. 27

    Kleywegt, G. J. Use of non-crystallographic symmetry in protein structure refinement. Acta Crystallogr. D 52, 842–857 (1996)

  28. 28

    CCP4 The CCP4 suite: Programs for protein crystallography. Acta Crystallogr. D 50, 760–763 (1994)

  29. 29

    Myszka, D. G. Kinetic analysis of macromolecular interactions using surface plasmon resonance biosensors. Curr. Opin. Biotechnol. 8, 50–57 (1997)

Download references


We thank the staff at beamline ID14-2 of the ESRF, France. We are grateful to A. van der Merwe and A. Kearney for advice and assistance with the surface plasmon resonance experiments. L. M. McNeil assisted with the NMR analysis. This work was supported by Cancer Research UK and the Wellcome Trust. W.-C.H. and M.I.W. were recipients of a Human Frontiers long-term fellowship and a Natural Sciences and Engineering Research Council of Canada postdoctoral fellowship, respectively. The Medical Research Council supports C.W.P. and D.I.S., and Cancer Research UK supports E.Y.J.

Author information

Correspondence to E. Yvonne Jones.

Ethics declarations

Competing interests

The authors declare that they have no competing financial interests.

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.