Structure of Human MDM4 N-Terminal Domain Bound to a Single-Domain Antibody

J. Mol. Biol. (2009) 385, 1578–1589

doi:10.1016/j.jmb.2008.11.043

Available online at www.sciencedirect.com

Structure of Human MDM4 N-Terminal Domain Bound to a Single-Domain Antibody Grace W. Yu†, Marina Vaysburd†, Mark D. Allen, Giovanni Settanni and Alan R. Fersht⁎ Centre for Protein Engineering, Medical Research Council, Hills Road, Cambridge CB2 0QH, UK Received 18 August 2008; received in revised form 19 November 2008; accepted 19 November 2008 Available online 30 November 2008

The N-terminal domain of MDM4 binds to the N-terminal transactivation domain of the tumor suppressor p53 and is an important negative regulator of its transactivation activity. As such, inhibition of the binding of MDM4 to p53 is a target for anticancer therapy. The protein has not been crystallized satisfactorily for structural studies without the addition of an N-terminal p53 peptide. We selected a single-domain antibody (VH9) that bound to the human domain with a dissociation constant of 44 nM. We solved the structure of the complex at 2.0-Å resolution. The asymmetric unit contained eight molecules of VH9 and four molecules of MDM4. A molecule of VH9 was located in each transactivation domain binding site, and the four nonMDM4-bound VH9 domains provided additional crystal contacts. There are differences between the structures of human MDM4 domain bound to VH9 and those of human and zebra fish MDM4 bound to a p53 peptide. Molecular dynamics simulations showed that the binding pocket in the three MDM4 structures converged to a common conformation after removal of the ligands, indicating that the differences are due to induced fit. The largest conformational changes were for the MDM4 molecules bound to p53. The simulated and observed structures should aid rational drug design. The use of single-domain antibodies to aid crystallization by creating a molecular scaffold may have a wider range of applications. © 2008 Elsevier Ltd. All rights reserved.

Edited by I. Wilson

Keywords: MDM4; MDMX; p53; crystallization; domain antibody

Introduction The proteins MDM21 and MDM42,3 are key regulators of the stability and activity of the tumor suppressor p53.4–6 The N-terminal domains of human MDM4 (also known as MDMX) and MDM2 are homologous and bind to the N-terminal transactivation domain (TAD1, residues 19–27) of p53 with similar affinities.7–9 MDM2 is an E3 ligase that is under the transcriptional control of p53. MDM2, via its RING domain, ubiquitinates p53 so that it undergoes proteasomal degradation. MDM4 is not under the transcriptional control of p53, and, even though MDM4 has a RING domain, it is not an E3 ligase but functions as a tight-binding competitive *Corresponding author. E-mail address: [email protected]. † G. W. Y. and M. V. contributed equally to this work. Abbreviations used: TAD, transactivation domain; sdAb, single-domain antibody; MD, molecular dynamics; CDR, complementarity-determining region; HRP, horseradish peroxidase.

inhibitor of other proteins that bind to TAD1 of p53. Unlike that of MDM2, the expression of MDM4 does not change after UV irradiation of cells. Amplification and overexpression of MDM2 and MDM4 are frequently associated with tumorigenesis.4 MDM4 and MDM2 are targets for anticancer therapy to increase the activity of p53 and its stability against proteasomal degradation.6 MDM2 has a well-defined binding cleft for TAD1, in which residues 19–26 become α-helical.10 The small molecule nutlin binds tightly in this cleft, inhibits the binding of p53, and has strong antitumorigenic activity; in addition, it is very useful for modulating the activity of p53 in cell-based systems.11 Surprisingly, nutlin binds only weakly to MDM4, despite its high sequence identity in the putative binding site.7,12–14 Information on the structure of the N-terminal domain MDM4 could possibly aid in the rational design of specific inhibitors that might have antitumorigenic activity and be of use in cell-based analysis. So far, the structure of human MDM4 and a “human mimic” zebra fish homolog had been solved bound to a p53 peptide.9,15

0022-2836/$ - see front matter © 2008 Elsevier Ltd. All rights reserved.

Human MDM4 N-Terminal Domain Structure

After improving the solubility of the MDM4 N-terminal domain by binding to a p53-derived peptide, we selected a single-domain antibody (dAb), VH9, against the protein–peptide complex as a means of obtaining good diffracting crystals of MDM4. Here, we report the crystal structure of the protein–VH9 complex at 2-Å resolution. The VH9 bound specifically to the cleft of MDM4, with an affinity of 44 nM, displacing the TAD1 peptide. We report molecular dynamics (MD) simulations started from this and other available high-resolution structures of MDM4 to evaluate structural changes on removal of the ligands, and we also discuss the flexibility of the binding cleft and the implications for the binding mechanism to the p53 peptide.

Results Selection of p53-derived peptide to improve MDM4 solubility The N-terminal domain of human MDM4 tends to aggregate in isolation, while it becomes more

1579 Table 1. Binding affinity of the p53-derived peptides to lipoyl-MDM4 N-terminal domain as measured by fluorescence anisotropy p53-derived peptide 17–26 15–29 1–29 1–57 S15P S20P S15/20P T18P

MDM4 N-terminal domain (Kd, μM) 0.09 ± 0.006 0.12 ± 0.007 0.19 ± 0.008 0.25 ± 0.04 0.29 ± 0.01 0.24 ± 0.01 0.23 ± 0.01 8.40 ± 0.25

soluble when bound to p53-derived peptides.9 We performed binding activity assays to select a p53derived peptide for improving MDM4 solubility. The lipoyl-tagged MDM4 was used throughout the assays. The wild-type peptides used for anisotropy experiments were derived from the N-terminal domain of p53 from amino acids 1–57, 1–29, 15–29, and 17–26. Several residues were selectively phosphorylated in the wild-type peptide 15–29 to study the effects of binding to MDM4. Lipoyl-MDM4 bound tightly to the wild-type p53 peptides, as measured by fluorescence anisotropy (Fig. 1), and the binding affinity was in good agreement with earlier studies7,9 (Table 1). These results showed that the most critical residues of p53 N-terminal domain are located within residues 17–26. Additionally, the p53 peptide (15–29) bound more tightly to MDM4 (Kd = 0.12 μM) than it did to MDM2 (Kd = 0.6 μM).9,16,17 The p53 binding affinity to MDM4 was weakened by phosphorylation at positions S15 and S20, while phosphorylation at T18 weakened binding to MDM4 by 40-fold. This phenomenon is similar to MDM2 binding to phosphorylated p53 peptides, 8,17,18 indicating that T18 phosphorylation could be a major candidate among p53 N-terminal modifications that negatively regulate p53–MDM4 interaction. For the abovementioned reasons and since peptide 15–29 has been widely used for the crystallization of MDM210 and MDM4,9 we chose it to improve the solubility of MDM4 for antibody selection. Selection of the MDM4-specific sdAb

Fig. 1. Anisotropy measurements of the p53-derived peptide binding to the lipoyl-MDM4 N-terminus. The phosphorylated residues are marked with “P.”

To select the MDM4-specific antibodies, we used a synthetic single-domain VH library with extensively randomized complementarity-determining regions (CDRs) and residue 30 of framework 1 within the single human framework VH DP47.16 Germ-line segment DP-47 is the most frequently used one within the VH3 family. It has superior properties with respect to, thermodynamic stability, and expression yield.17,18 Owing to the difficulty in isolating dAbs against the aggregation-prone MDM4 N-terminal domain, the more stable complex of MDM4 and p53-derived peptide (residues 15–29) was used as an antigen. The complex was captured on a surface of streptavidin-coated magnetic beads

Human MDM4 N-Terminal Domain Structure

1580

Fig. 2. ELISA analysis of the phage library before and after two rounds of selection against the MDM4–p53 complex. Phages were incubated with biotinylated MDM4–p53 complex or lysozyme immobilized on streptavidin plates. Bound phages were detected with antiM13-HRP.

through a biotin linker. We used a low molar excess of the activated biotin linker over the MDM4–p53 complex to avoid substantial chemical modification of the antigen surface. The resulting complex, analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, contained no more than four biotin groups. The library was selected by two rounds of bacteriophage panning against the MDM4–p53 peptide complex. During successive rounds of panning, we increased washing stringency while simultaneously decreasing the concentration of the complex and employing a preadsorption step to remove nonspecific binders. Following two rounds of selection, an increase in bacteriophage titer was observed, suggesting positive enrichment within the population. ELISA analysis of the library before and after each round of selection indicated an increase in MDM4–p53 peptide complex-specific clones compared with the negative control antigen (Fig. 2). Screening was performed using 12 nM MDM4–p53 peptide complex. For further analysis, the entire VH repertoire after two rounds of selection was cloned into the expression vector pET 12/25. Individual dAbs were expressed and analyzed by ELISA for binding to the MDM4–p53 peptide complex. Eleven clones

Fig. 3. Soluble VH fragments were tested by ELISA for binding to decreasing amounts of biotinylated MDM4– p53 complex immobilized on streptavidin plate. Bound antibody fragments were detected with anti-HSV-HRP.

that showed the highest binding signal were analyzed by sequencing. The selected dAb pool was dominated by one sequence, VH1 (Table 2). Seven dAb clones had identical sequences, whereas VH6 had the same CDR3 but different CDR1 and CDR2. The rest of the dAbs had unique sequences. For further analysis, proteins VH1, VH7, VH9, and VH12 were expressed and purified from the supernatant by binding to protein A agarose. All proteins except for VH7 were efficiently expressed in Escherichia coli (yields = 6–20 mg/l), and their comparative binding to the decreasing amount of MDM4–p53 peptide complex was analyzed by ELISA. VH9 bound the strongest (Fig. 3), and so it was selected for structural investigation. Interaction of antibody VH9 with MDM4 and MDM2 We used isothermal titration calorimetry to characterize the interaction between VH9 and MDM4 and to compare VH9 binding to MDM2. The equilibrium dissociation constant (Kd) for the MDM4– VH9 complex was 44 ± 9 nM, with 1:1 stoichiometry, whereas antibody bound 140-fold less tightly to MDM2 (Kd = 6 ± 0.9 μM) (Fig. 4).

Table 2. CDR sequences of human VHs selected against the MDM4–p53 complex

VH1 VH2 VH4 VH5 VH8 VH10 VH11 VH6 VH7 VH9 VH12

CDR1

CDR2

CDR3

GSYTMN GSYTMN GSYTMN GSYTMN GSYTMN GSYTMN GSYTMN GKYGMT KDYAMS EEYAML ENYDMQ

AINNTGQLTYYADSVKG AINNTGQLTYYADSVKG AINNTGQLTYYADSVKG AINNTGQLTYYADSVKG AINNTGQLTYYADSVKG AINNTGQLTYYADSVKG AINNTGQLTYYADSVKG GISSDGRTTYYADSVKG TISSAGNLTYYADSVKG GINARGYTTYYADSVKG SITSKGQSTYYADSVKG

DSRRGL———FDY DSRRGL———FDY DSRRGL———FDY DSRRGL———FDY DSRRGL———FDY DSRRGL———FDY DSRRGL———FDY DSRRGL———FDY VAAGFSLG———FDY PWYPFMASKGSEFDY AWTAVQD———FDY

Human MDM4 N-Terminal Domain Structure

1581

Fig. 4. Typical isothermal titration calorimetric19 measurements of the MDM4–VH9 and MDM2–VH9 interactions. The upper panel shows the original raw data, whereas the lower panel shows the fit after integration. MDM4 (a) and MDM2 (b) N-terminal proteins (25 μM) were loaded into the reaction cell and titrated with 200 μM VH9. The experiments were carried out at 15 °C in a buffer of 25 mM NaPi, pH 7.5, 150 mM NaCl, and 5 mM DTT.

Crystallization of MDM4 with peptide and VH9 Attempts to crystallize a mixture of MDM4, VH9, and a p53 TAD peptide resulted in crystals consisting only of a complex of MDM4 and VH9 antibody. Initially, crystals containing a selenomethioninelabeled MDM4 that diffracted to 3.0 Å were used to solve the structure of the complex. Analysis of the maps revealed the presence of eight molecules of VH9 and four molecules of MDM4 in each asymmetric unit. Molecules of an dAb [based on the Protein Data Bank (PDB) structure 1dql and MDM4 (based on MDM2, PDB structure 2gv2) were placed into the maps, and the structure was refined. Subsequently, a 2.0-Å native data set was obtained, and the refined complex was used as a search model for molecular replacement into the native data set. The 3.0-Å selenomethionine and 2.0Å native data sets were both obtained from crystals with a P21 space group. Cell constants, crystallographic data, and details of the refined models are shown in Table 3. Analysis of the asymmetric unit reveals that all four molecules of MDM4 bind primarily to VH9 using the same interface as that used by MDM2 to bind to TAD1. Each of the VH9 molecules bound to MDM4 also has a significant secondary contact with a non-MDM4-bound VH9 domain with a buried surface of ∼ 1534 Å2 that is stabilized by several hydrogen bonds. The non-MDM4-bound VH9 do-

mains also interact with another non-MDM4-bound VH9 domain (to form VH9 dimers) using aromatic residues in the CDR3 loop Y101–F103 (residues Y97 and F99 according to Kabat et al.21). As such, an extensive network of interdomain contacts that are essential for maintaining the crystal integrity can be seen. MD simulations The structure of human MDM4 from the complex with VH9 has been used as starting point for several MD simulations. Similarly, the structures of human15 and humanized zebra fish9 MDM4 from the complexes with the p53 peptide have been used to start two other sets of simulations. In all three simulated systems, a conformational relaxation occurred after removal of the ligand partner, with the Cα root-mean-square deviation (RMSD) reaching values around 2.0 Å. Atomic root-mean-square fluctuations (RMSFs) indicated a larger conformational heterogeneity in the zebra fish variant than in the human variants (Fig. 5a). Removal of p53 TAD peptide from the human MDM4 caused slightly larger RMSF amplitudes than removal of VH9 (Fig. 5a). The more flexible regions in the two variants coincide and involve the turn segments connecting secondary structural elements, which, instead, are relatively rigid, as expected. Notably, the residues in the binding pocket

Human MDM4 N-Terminal Domain Structure

1582 Table 3. Crystallographic data collection and refinement statistics Data set Wavelength (Å) Resolution range (Å) Unique reflections Completeness (%)a Rmergeb Multiplicityc I/σIa No. of Se sites permonomer Space group and unit cell dimensions

Peak

Inflection

Remote

0.9797 0.9799 0.9686 95.0–3.0 45.0–3.0 95.0–3.0 34,327 34,259 34,540 99.1 (98.9) 99.0 (98.7) 99.0 (98.9) 0.075 (0.148) 0.065 (0.129) 0.069 (0.157) 7.3 (7.4) 3.7 (3.7) 3.7 (3.7) 19.4 (9.7) 11.4 (6.1) 8.5 (4.1) 5 P21, a = 79.793 Å, b = 114.512 Å, c = 99.308 Å, β =.105.32°

Model refinement Resolution range (Å) No. of reflections (working/free) No. of residues No. of water molecules no. of ligand molecules Rwork/Rfreed (%) B averagee Geometry bonds/anglesf Ramachandran plotg PDB IDh

Native 0.9537 30.0–2.0 116,657 99.6 (99.6) 0.086 (0.272) 3.9 (4.0) 10.4 (4.3) P21,a = 79.823 Å, b = 111.200 Å, c = 99.368 Å, β = 104.86°

25.0–2.025 .0 – 2.0 106,975/5655 A–D, 25–109; E, F, H, J, K, and L, 1–124; G, 1–123; I, 1–125 752 4 SO4 0.207/0.248 28.43 Å2 0.008Å/1.086° 92.9%/0.2% 2vyr

The first three columns refer to the Se derivative, whereas the fourth column refers to the native protein. a Values within parentheses are for the highest-resolution bin. b Rmerge = ShSi|I(h,i) − I(h)|/ShSiI(h,i) where I(h,i) refers to symmetry-related intensities and I(h) is the mean intensity of the reflection with unique index h. c Multiplicity values for unique reflections for multiwavelength anomalous diffraction data sets I(+) and I(−) are kept separate. d Five percent of reflections was randomly selected for determination of the free R-factor, prior to any refinement. e Temperature factors averaged for all atoms. f RMSDs from ideal geometry for bond lengths and restraint angles.20 g Percentage of residues in the most favored region of the Ramachandran plot and percentage of outliers (PROCHECK). h PDB identifiers for coordinates.

showed generally low RMSFs, apart from residues V92 and K93, which, especially in the zebra fish variant, were rather flexible (Fig. 5a). The conformational spaces sampled by the binding pocket region along the simulations of the three protein systems showed a degree of overlap. Namely, at a later stage of one of the simulations of the zebra fish variant, the binding pocket reached conformations similar to those sampled by the human variants (Fig. 5b). The latter sampled similar conformations over the whole trajectory (Fig. 5b). On the other hand, the heterogeneity of the conformational spaces of the variants originally bound to the p53 peptide was confirmed by the pairwise all-atom RMSD of the binding pocket residues, which reached more than 3.0 Å in some cases (Fig. 5b). Conformational cluster analysis of the binding pocket was used to group similar conformations from the simulations into homogeneous sets. It revealed that one of the largest clusters spanned conformations sampled by human and zebra fish protein variants (Fig. 5c), confirming the similarities mentioned above.

Discussion Initial attempts to crystallize the N-terminal domain of MDM4 or its complex with a tight binding peptide from TAD1 of p53 yielded poorly diffract-

ing crystals only, the best of which diffracted to 4.0Å resolution. In contrast, crystals of a complex of MDM4 bound to a specific dAb, VH9, diffracted significantly better and resulted in a 2.0-Å resolution crystal structure. The asymmetric unit was composed of four MDM4–VH9 complexes with four additional molecules of VH9, forming crystal contacts such that the asymmetric unit was composed of four molecules of MDM4 and eight molecules of VH9. Structure of human MDM4 Human MDM4 adopts an overall fold very similar to that of human MDM2. Superposition of the backbone atoms for MDM4 residues 25–109 and human MDM2 (IYCR) residues 26–110 gave an RMSD value of 1.29 Å (Fig. 6a). In MDM4, α1 consists of residues 30–40; α2, residues 49–64; α1′, residues 80–85; α2′, residues 96–105; β1, residues 26–28; β2, residues 73–75; β1′, residues 89–91; and β2′, residues 106–108. The residues in the hydrophobic core of MDM4 that differ from those of MDM2 may cause a slight distortion in the cleft. Val49 in MDM4 is less bulky than the equivalent Met50 in the α2 helix of MDM2, which may open the cleft by weakening the interactions between α2 and α2′. The distance between Val49 and Leu102 in MDM4 is 5.03 Å, and the distance between Met50 and Ile103 in MDM2 is 4.36 Å. The side chain of Met53 in MDM4 (Leu54 in MDM2) points toward

Human MDM4 N-Terminal Domain Structure

1583

Fig. 5. (a) Residue averaged RMSFs measured along the simulations of the human variant originally bound to VH9 (black line) and the zebra fish variant (red) and the human variant originally bound to p53 (green). At the top, the sequence, secondary structure, and solvent exposure of the residues of the two proteins are provided as a reference. Residues in the binding pocket are marked with a red letter code. The numbering of residues is according to the human variant. (b) Matrix of the RMSDs of the binding pocket region (heavy atoms of residues M53, H54, L56, G57, I60, M61, Y66, Q71, H72, V74, F90, V92, K93, P95, backbone of residue L98, and Y99, according to human MDM4 sequence). Each conformation saved along the simulations of human and zebra fish variants is compared with all other conformations.22,23 Dots on the matrix are colored according to the RMSD between the structures sampled at the corresponding time points on the axes. The regions on the matrix marked by green ellipses point to the conformational similarities (blue dots) of the binding pocket observed between different MDM4 variants. The observed similarity exceeds the variability measured along the pair of simulations of MDM4 variants originally bound to p53 (red dots marked by red circles). (c) Structure of the binding pocket in the conformational cluster that spans the human (VH9) and zebra fish protein variants. Binding pocket residues of the structure at the center of the cluster are reported as liquorices, and other conformations of the side chains in the same cluster are reported as transparent tubes. Side chains of residues outside the binding pocket have been removed to simplify the picture. Labels indicate the three most variable side chains within the cluster.

the center of the cleft and forms multiple van der Waals contacts with residues on the opposite side of the cleft. These interactions result in a smaller p53binding cleft. Another distinct difference in amino acid sequence between MDM4 and MDM2 is a stretch of MDM4 residues P95, S96, and P97 in helix α2′, which are replaced by H96, R97, and K98 in MDM2. The presence of P95 in helix α2′ may be of

importance as the end of the helix is displaced outward from the cleft by approximately 4 Å compared with that observed in the MDM2–nutlin complex. P97 also causes deformation of the helix, causing the Tyr99 ring to move outward and away from Met53, and thus also contributes to the opening of the cleft. In contrast, Tyr100 in MDM2 is in close contact with Leu54 (3.91 Å), which pulls

1584

Human MDM4 N-Terminal Domain Structure

Fig. 6. (a) Superimposed structures of MDM4 (green) and MDM2 (gray). The major differences between MDM4 and MDM2 are located in their α2 and α2′ helices. The side chains of residues involved in cleft formation are colored. MDM2 residues Met50 and Leu54 are highlighted in salmon, and MDM4 residues Val49 and Met53 are highlighted in green. (b) Superposition of the α2′ helix of human MDM4 bound to VH9 (red) and human MDM4 (green) and zebra fish MDM4 (blue) bound to p53 peptide. The side chains of Y99 involved in ligand binding are indicated.

the two helices together to form a more compact hydrophobic core. Comparison of human and zebra fish MDM4 Human MDM4 bound to a 15-residue TAD peptide has been reported (3dab)15while we were preparing this article. Comparison of all the MDM4 structures reveals that both human MDM4 and zebra fish MDM4 adopt a similar overall fold. The backbone superposition of the human MDM4 structure (residues 25–109) bound to VH9 (2vyr) with the zebra fish variant (2z5s), residues 22–106, and the human variant bound to peptide (3dab), gives RMSD values of 0.63 and 0.7 Å, respectively. The most significant differences are in the α2′ helix (human MDM4 residues 96–105 and zebra fish MDM4 residues 93–102), where the orientation of Y99 in the human variant (3dab) and that of Y96 in the zebra fish variant (2z5s) involved in peptide binding are different from the orientation of Y99 involved in antibody binding (2vyr) (Fig. 6b). This might be a result of the conformational rearrangement upon binding to ligands. To clarify the situation, we conducted MD simulations to evaluate the

effects of removal of VH9 from the complex with human MDM4 and p53 peptide from both human MDM4 and zebra fish MDM4. All the species of MDM4 underwent conformational relaxation upon dissociation of their ligands. The conformational space sampled by the unbound zebra fish variant was more heterogeneous than the human variant originally bound to p53 as evidenced by RMSFs and pairwise RMSDs. Similarly, the human variant originally bound to p53 showed larger fluctuations than the variant originally bound to VH9. Notwithstanding the heterogeneity, in one of the simulations of the zebra fish variant, the binding pocket eventually relaxed to conformations very similar to those sampled by both human variants, as shown by pairwise RMSDs. A similar but faster relaxation is observed early in the trajectories of the human MDM4 originally bound to p53. This may indicate that the perturbation introduced by the binding of the p53 peptide to the binding cleft on MDM4 is larger than the perturbation introduced by binding of the VH9; thus, the relaxation process observed upon removal of the p53 ligand involves a wider exploration of the conformational space. The conformations of the binding pocket common to the trajectories of

Human MDM4 N-Terminal Domain Structure

the three MDM4 variants as identified by cluster analysis may be used for the screening of highaffinity compounds because they may summarize the characteristics relevant for binding to MDM4. Although the 4.0-Å resolution structure of the human MDM4–peptide complex is inadequate for side-chain analysis, it was useful for a comparison of polypeptide chains to see whether the binding of

1585 VH9 induces any conformational changes relative to the TAD1 peptide. We found that there was no observable difference in overall backbone chains between peptide-bound MDM4 and VH9-bound MDM4; however, due to the low resolution of the MDM4–peptide complex, we were unable to determine an accurate RMSD value. These data were consistent with the MD simulation.

Fig. 7. (a) Crystal structure of MDM4 (gray) in complex with dAb VH9 (green). The CDR3 loop of VH9 is shown in red, and the side chains of major residues [W100, Y101, F103, F111, and Y113 (residues W96, Y97, F99, F100g, and Y102, respectively, according to Kabat et al.21)] involved in binding are shown. Residues E110 and D112 (residues 100f and D101, respectively, according to Kabat et al.21), which form hydrogen bonds with MDM4, are indicated in cyan. The side chains of CDR1 residues E31 and Y32 (residues E31 and Y32, respectively, according to Kabat et al.21) are shown in magenta. (b) Stereoview of the antibody-bound structure of MDM4. Contact residues of the VH9 CDR3 loop (red) and CDR1 loop (magenta) are indicated. VH9 residues E110 and D112 (residues 100f and D101, respectively, according to Kabat et al.21) form hydrogen bonds with MDM4 residues Y99 and K50 (blue).

1586 Structure of MDM4–VH9 interface The overall structure of a dAb VH9 in complex with MDM4 is very similar to the unbound structures previously reported (1qhq and 1t2j), except within the CDR3 region. CDR3 consists of 15 residues (residues 99–113, which correspond to residues 95–103 according to Kabat et al.21) and is longer than the average human CDR3 loop (12 residues24). The extended CDR3 loop was implicated in improved solubility of the naturally occurring VH domains in Camelidae25 and isolated VH domains from the human synthetic library17 as it partially covers the hydrophobic region equivalent to the interface with VL in conventional antibodies.26 The CDR3 is the major contributor to the antigen binding as it accounts for 77% of the buried surface (1534 Å2; Fig. 7a). It protrudes into the p53 binding pocket of MDM4 and packs closely onto the α2 helix of MDM4. The most prominent structural feature of the MDM4–VH9 interface is the burial of several aromatic residues of the CDR3: W100, Y101, F103, F111, and Y113 (residues 96, 97, 99, 100g, and 102, respectively, according to Kabat et al.21). These are aligned along the hydrophobic surface of the α2 helix and make extensive contacts with the MDM4 residues (M53, L56, I60, M61, V49, and L102). The complex is further stabilized by several side chainmediated polar interactions. VH9 CDR3 residues Y101, E110, Y113, and D112 (residues 97, 100f, 102, and 101, respectively, according to Kabat et al.21) form hydrogen bonds with MDM4 residues Q71, Y99, E51, and K50, respectively (seen in PDB structure 2vyr, chains D and H). Additionally, CDR1 residues E31 and Y32 (residues E31 and Y32,

Human MDM4 N-Terminal Domain Structure

respectively, according to Kabat et al.21) also contribute to the stabilization of the complex by interacting with MDM4 residues M61, N58, and H54 and form intermolecular hydrogen bonds with Q58 (Fig. 7b). Overall, the epitope of human MDM4 recognized by the antibody overlaps significantly with the p53–MDM4 interface. A topographic correspondence can be established between antibody residues W100 and Y101 (residues 96 and 97, respectively, according to Kabat et al.21) and the residues of p53 peptide, W23, P19, and L26, all of which interact with the same region of MDM4. Thus, the VH9 antibody provides a molecular surface that is complementary to the p53 TAD binding cleft of MDM4, and its CDR3 aromatic residues mimic the spatial arrangement of the groups of the p53 peptide crucial for binding to MDM4. However, compared with the p53 peptide, VH9 binds to the human MDM4 by creating an additional layer of contacts (antibody residues E31, Y32, K98, Y113, and W114: residues 31, 32, 94, 102, and 103, respectively, according to Kabat et al.21) that extends the interaction surface beyond the boundaries of the p53 binding cleft (Fig. 8). This results in a larger interface area of the MDM4–VH9 complex compared with the MDM4–p53 complex (1534 versus 919 Å2). Taken together, crystal structure comparison and MD simulations of the free MDM4 may suggest that the improved affinity of the MDM4 for VH9 compared with the p53 peptide is a result of a larger interaction surface and a better surface complementarity that requires smaller induced-fit changes in MDM4 upon binding. The structure presented here and the conformations of MDM4 sampled using MD simulations may

Fig. 8. Superposition of the structures of the VH9–MDM4 and p53–human MDM4 complexes. Cα traces for p53 peptide (green) and VH9 contact residues (magenta) are shown. The determinants of p53 binding to MDM4 (Phe19, Trp23, and Leu26) are highlighted.

Human MDM4 N-Terminal Domain Structure

provide a template for rational design of inhibitors of p53–MDM4 interactions. The VH9 antibody represents a starting point for designing tighter binding dAbs. The technique introduced here, which exploits sdAb's to help protein crystallization, may have a wide range of applications.

Materials and Methods Cloning and purification of the lipoyl-MDM4 N-terminus (residues 16–116) The cDNA corresponding to the N-terminal domain of MDM4 (residues 16–116) was amplified by PCR from the human mdm4 cDNA using oligonucleotides 5′-CGGAATTCCTAAGCATCTGTAGTAGCAGTGGC-3′ and 5′CGGGATCCGCTTGCAGGATCTCTCCTGGAC-3′ and cloned into a modified pRSETa vector encoding an Nterminal 6× His tag, lipoyl domain, and TEV cleavage site to make the plasmid pHLT-MDM4 (residues 16–116). Expression and purification were performed as described previously.27 The selenomethionine-substituted protein was prepared in an identical manner except that the cells were grown in M9 minimal media supplemented with selenomethionine and amino acid substitutions. Purification of MDM4 N-terminal domain in complex with p53-derived peptides Purified lipoyl-MDM4 N-terminal domain was incubated with p53-derived peptides, SQETFSDLWKLLPEN, at 4 °C for at least 4 h. The complex was digested with TEV protease at room temperature. The digested protein complex was purified by a Ni–nitriloacetic acid Superflow column following size-exclusion chromatography. The purified protein complex was concentrated and used for antibody selection. Phage selection The synthetic sdAb repertoire used for selection was a mixture of dAb libraries (Domantis Ltd.) with different lengths of the CDR3 loop, varying from 7 to 15 residues. The phage library with a diversity of 2 × 109 clones was selected against biotinylated MDM4–p53 peptide complex immobilized on the surface of streptavidin-coated Dynabeads (Dynal Biotech). The selection protocol had been described elsewhere.28 Briefly, purified phage from the library (3 × 1012 TU/ml in phosphate-buffered saline and 5 mM DTT) was depleted from nonspecific binders and incubated for 1.5 h with MDM4–p53 peptide complexcoated beads. Bound phages were eluted with trypsin followed by infection of E. coli TG1 cells for phage amplification. Expression and purification of soluble dAbs The phage library after two rounds of selection was subcloned in a pET-based vector (Novagen) that is essentially a pET12a vector harboring the pET25a cassette inserted between the SalI and Bpu11021 cloning sites. This cassette provides HSV and hexahistidine tags for protein detection and purification. After transformation of E. coli C41 cells, dAbs were expressed and purified as described previously.18

1587 ELISA assays A total of 0.13 μg/ml of biotinylated MDM4–p53 peptide complex was immobilized on a streptavidin-coated plate (Roche). After blocking the wells with 4% phosphate buffered saline supplemented with marvel milk powder in the presence of 5 mM DTT, phage-containing supernatants or soluble dAb-containing supernatants in 2% phosphate buffered saline supplemented with marvel milk powder and 5 mM DTT were incubated for 1.5 h at 4 °C. Phages were detected using an anti-M13-horseradish peroxidase (HRP) monoclonal antibody (1:5000, Amersham), and soluble dAbs were detected with anti-HSV-HRP (1:3000, Bethyl Laboratories). Purification of the VH9 antibody The VH9 was expressed and purified as described above with minor modifications. Twelve liters of culture was concentrated via Crossflow (Sartorius) before applying to a column packed with rProtein A-Sepharose™ Fast Flow (GE Healthcare). The collected fractions were immediately subjected to dialysis with 25 mM Tris buffer, pH 7.4, and 150 mM NaCl. The antibody was further purified by gel filtration (20 mM Tris buffer, pH 7.4, 150 mM NaCl, and 10 mM β-mercaptoethanol). Purification of the MDM4 N-terminal domain in complex with VH9 Purification of the MDM4–VH9 complex was the same as that described above for the MDM4–peptide complex. The complex was eluted as a single peak in gel-filtration column. Crystallization and structure determination Human MDM4/VH9 antibody was crystallized at 11.7 mg/ml by the sitting-drop vapor method in 1.6 M ammonium sulfate, 0.5 M lithium chloride, and 1 mM Tris, pH 7.0, at 17 °C. Crystals used for phasing were grown in 20% PEG (polyethylene glycol) 3350, 0.2 M magnesium chloride, and 1 mM Tris, pH 7.0, using selenomethioninesubstituted human MDM4 bound to native VH9 antibody at the same concentration as that for the native proteins. The crystals were flash frozen in liquid nitrogen after the addition of glycerol to 20% while leaving the other components of the mother liquor at the same concentration. Data sets were collected at the European Synchrotron Radiation Facility on beamline ID14-2 (native and selenomethionine). All data were indexed and integrated with MOSFLM29 and further processed using the CCP4 package.30 Initially, 19 selenomethionine sites were found with SHELXD31 using the peak, inflection, and remote data sets (Table 3). The coordinates of these sites were imported into SHARP,32 and phases were calculated. The model was built manually with the program MAIN33 and refined using CNS34 and PHENIX.35 The starting models for refinement were based on the structures of MDM2 (PDB structure 2gv2) and an sdAb with a homologous sequence (PDB structure 1DQL). Fluorescence anisotropy All anisotropy measurements were performed as described previously27,36 in a buffer containing 25 mM NaPi,

1588

Human MDM4 N-Terminal Domain Structure

pH 7.5, 150 mM NaCl, and 5 mM DTT at 20 °C. Peptides were a generous gift from Dr. Daniel Teufel (Medical Research Council Centre for Protein Engineering). Peptide concentrations were estimated from absorption readings using ɛ324 = 12,000 M− 1 cm− 1.37 Dissociation constants were calculated by fitting the data (corrected for dilution) to a simple 1:1 equilibrium model.

Acknowledgements

Isothermal titration calorimetry

References

Isothermal titration calorimetry was performed as described previously27 at 15 °C in a buffer of 25 mM NaPi, pH 7.5, 150 mM NaCl, and 5 mM DTT, and the data were analyzed using Origin™ software.

1. Cahilly-Snyder, L., Yang-Feng, T., Francke, U. & George, D. L. (1987). Molecular analysis and chromosomal mapping of amplified genes isolated from a transformed mouse 3T3 cell line. Somatic Cell Mol. Genet. 13, 235–244. 2. Shvarts, A., Steegenga, W. T., Riteco, N., van Laar, T., Dekker, P., Bazuine, M. et al. (1996). MDMX: a novel p53-binding protein with some functional properties of MDM2. EMBO J. 15, 5349–5357. 3. Shvarts, A., Bazuine, M., Dekker, P., Ramos, Y. F., Steegenga, W. T., Merckx, G. et al. (1997). Isolation and identification of the human homolog of a new p53binding protein, Mdmx. Genomics, 43, 34–42. 4. Toledo, F. & Wahl, G. M. (2006). Regulating the p53 pathway: in vitro hypotheses, in vivo veritas. Nat. Rev. Cancer, 6, 909–923. 5. Toledo, F. & Wahl, G. M. (2007). MDM2 and MDM4: p53 regulators as targets in anticancer therapy. Int. J. Biochem. Cell Biol. 39, 1476–1482. 6. Vassilev, L. T. (2007). MDM2 inhibitors for cancer therapy. Trends Mol. Med. 13, 23–31. 7. Laurie, N. A., Donovan, S. L., Shih, C. S., Zhang, J., Mills, N., Fuller, C. et al. (2006). Inactivation of the p53 pathway in retinoblastoma. Nature, 444, 61–66. 8. Bottger, V., Bottger, A., Garcia-Echeverria, C., Ramos, Y. F., van der Eb, A. J., Jochemsen, A. G. & Lane, D. P. (1999). Comparative study of the p53–mdm2 and p53– MDMX interfaces. Oncogene, 18, 189–199. 9. Popowicz, G. M., Czarna, A., Rothweiler, U., Sszwagierczak, A., Krajewski, M., Weber, L. & Holak, T. A. (2007). Molecular basis for the inhibition of p53 by Mdmx. Cell Cycle, 6, 2386–2392. 10. Kussie, P. H., Gorina, S., Marechal, V., Elenbaas, B., Moreau, J., Levine, A. J. & Pavletich, N. P. (1996). Structure of the MDM2 oncoprotein bound to the p53 tumor suppressor transactivation domain. Science, 274, 948–953. 11. Vassilev, L. T., Vu, B. T., Graves, B., Carvajal, D., Podlaski, F., Filipovic, Z. et al. (2004). In vivo activation of the p53 pathway by small-molecule antagonists of MDM2. Science, 303, 844–848. 12. Patton, J. T., Mayo, L. D., Singhi, A. D., Gudkov, A. V., Stark, G. R. & Jackson, M. W. (2006). Levels of HdmX expression dictate the sensitivity of normal and transformed cells to Nutlin-3. Cancer Res. 66, 3169–3176. 13. Wade, M., Wong, E. T., Tang, M., Stommel, J. M. & Wahl, G. M. (2006). Hdmx modulates the outcome of p53 activation in human tumor cells. J. Biol. Chem. 281, 33036–33044. 14. Hu, B., Gilkes, D. M., Farooqi, B., Sebti, S. M. & Chen, J. (2006). MDMX overexpression prevents p53 activation by the MDM2 inhibitor Nutlin. J. Biol. Chem. 281, 33030–33035. 15. Popowicz, G. M., Czarna, A. & Holak, T. A. (2008). Structure of the human Mdmx protein bound to the p53 tumor suppressor transactivation domain. Cell Cycle, 7, 2441–2443. 16. Tomlinson, I. W. S., Ignatovich, O., Corbett, S. &

MD simulations MD simulations of the human MDM4 from the complex with VH9 (2vyr), of the human MDM4 from the complex with p53 peptide (3dab),15 and of the zebra fish MDM4 from the complex with p53 peptide (2z5s) 9 were performed based on the X-ray structures for the initial conformation where the ligands were removed. Simulations were carried out using the GROMACS package38 and the OPLSAA force field.39 The TIP4P model40 was used for the water. The molecules were immersed in a periodic dodecahedral box with about 5900 water molecules, and Cl− ions were added to neutralize the overall charge of the systems. The dimensions of the initial box allowed for at least 10-Å distance between the protein and the box boundaries in the human variant 2vyr. A similar box size was used for the other variants. A cutoff of 10 Å was used for the calculation of the nonbonded interactions, and particle mesh Ewald method41,42 was used for the treatment of the long-range electrostatic interactions. A time step of 2 fs for the integration of the equations of motion was used. The LINCS43 algorithm was used to constrain the distance of the hydrogen atoms. Temperature and pressure were controlled by the Berendsen algorithm 44 with 0.1- and 0.5-ps coupling constants, respectively. The pressure was kept constant at 1 atm. Before the production run, the system was minimized using steepest descent for 1000 steps and conjugate gradient for 1000 other steps. Then, relaxation of the solvent molecules was performed for 10 ps. The latter was followed by an equilibration of the whole system for 0.5 ns, letting the temperature gradually increase from 30 to 300 K. Eventually, atom coordinates and velocities from the last equilibration snapshot were used to start the production runs. The latter consisted of 2 × 10 ns of trajectory per variant at 300 K and 1 atm, where atom coordinates and velocities were saved every 10 ps. All analyses of the simulations were performed using the program WORDOM.45 Conformational clustering was done using the leader algorithm as in Refs. 46 and 47, and the dRMS of the heavy atoms in the residues of the binding pocket (M53, H54, L56, G57, I60, M61, Y66, Q71, H72, V74, F90, V92, K93, P95, backbone of residue L98, and Y99, according to the human MDM4 numbering) was used as the metric with a cutoff of 1.1 Å. The residues in the binding pocket were those that have heavy atoms within 5 Å of any atom of the peptide ligand TAD1 in the crystal structure of the zebra fish MDM4 variant.43 Accession number Coordinates and structure factors have been deposited in the PDB with accession number 2vyr.

We thank Drs. Andreas Joerger, Antonina Andreeva and Mark Bycroft for valuable discussions. We also thank Domantis for providing the phage library.

Human MDM4 N-Terminal Domain Structure

17. 18.

19.

20. 21.

22.

23.

24. 25.

26.

27.

28. 29.

30.

31. 32.

Winter, G. (1988). V BASE Sequence Directory, MRC Centre for Protein Engineering, Cambridge, UK. Ewert, T. H., Honneger, A. & Pluckthun, A. (2003). Biophysical properties of human antibody variable domains. J. Mol. Biol. 325, 531–553. Jespers, L., Schon, O., Famm, K. & Winter, G. (2004). Aggregation-resistant domain antibodies selected on phage by heat denaturation. Nat. Biotechnol. 22, 1161–1165. Lee, W., Harvey, T. S., Yin, Y., Yau, P., Litchfield, D. & Arrowsmith, C. H. (1994). Solution structure of the tetrameric minimum transforming domain of p53. Nat. Struct. Biol. 1, 877–890. Engh, R. A. & Huber, R. (1991). Accurate bond and angle parameters for X-ray protein structure refinement. Acta Crystallogr., Sect. A, 47, 392–400. Kabat, E. W. T., Perry, H., Gottesman, K. & Foeller, C. (1991). Sequences of Proteins of Immunological Interest. US Department of Health and Human Services, Bethesda, MD. Jemth, P., Gianni, S., Day, R., Li, B., Johnson, C. M., Daggett, V. & Fersht, A. R. (2004). Demonstration of a low-energy on-pathway intermediate in a fast-folding protein by kinetics, protein engineering, and simulation. Proc. Natl Acad. Sci. USA, 101, 6450–6455. Interlandi, G., Settanni, G. & Caflisch, A. (2006). Unfolding transition state and intermediates of the tumor suppressor p16INK4a investigated by molecular dynamics simulations. Proteins, 64, 178–192. Wu, T., Johnson, G. & Kabat, E. A. (1993). Length distribution of CDRH3 in antibodies. Proteins: Struct., Funct., Genet. 16, 1–7. Van der Linden, R. H., Frenken, L. G., de Geus, B., Harmsen, M. M., Ruuls, R. C., Stock, W. et al. (1999). Comparison of physical chemical properties of llama VHH antibody fragments and mouse monoclonal antibodies. Biochim. Biophys. Acta, 1431, 37–46. Desmyter, A., Spinelli, S., Payan, F., Lauwereys, M., Wyns, L., Muyldermans, S. & Cambillau, C. (2002). Three camelid VHH domains in complex with porcine pancreatic α-amylase. J. Biol. Chem. 277, 23645–23650. Yu, G. W., Rudiger, S., Veprintsev, D., Freund, S., Fernandez-Fernandez, M. R. & Fersht, A. R. (2006). The central region of HDM2 provides a second binding site for p53. Proc. Natl Acad. Sci. USA, 103, 1227–1232. Lee, C. M., Iorno, N., Sierro, F. & Christ, D. (2007). Selection of human antibody fragments by phage display. Nat. Prot. 2, 3001–3008. Leslie, A. G. W. (1991). Molecular Data Processing. In Crystallographic Computing 5 (Moras, D., Podjarny, A. D. & Thierry, J. C., eds), Oxford University Press, Oxford, UK. Collaborative Computational Project No. 4. (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr., Sect. D: Biol. Crystallogr. 50, 760–763. Sheldrick, G. M. (1991). In Crystallographic Computing 5 (Moras, D., Podjarny, A. D. & Thierry, J. C., eds), Oxford University Press, Oxford, UK. De La Fortelle, E. & Bricogne, G. (1997). Maximumlikelihood heavy-atom parameter refinement for

1589

33.

34.

35.

36.

37.

38. 39.

40.

41. 42. 43. 44.

45.

46. 47.

multiple isomorphous replacement and multiwavelength anomalous diffraction methods. Methods in Enzymology, vol. 276, pp. 472–494. Academic Press, San Diego, CA. Turk, D. (1992). Weiterentwicklung eines Programms für Molekülgraphik und Elektronenolichte manipulation und seine Anwerdung auf verschiedene Protein Strukturaufklärungen, Technische Universität. Brünger, A. T. (1998). Crystallography & NMR System: a new software suite for macromolecular structure determination. Acta Crystallogr., Sect. D: Biol. Crystallogr. 54, 905–921. Adams, P. D., Grosse-Kunstleve, R. W., Hung, L. W., Ioerger, T. R., McCoy, A. J., Moriarty, N. W. et al. (2002). PHENIX: building new software for automated crystallographic structure determination. Acta Crystallogr., Sect. D: Biol. Crystallogr. 58, 1948–1954. Teufel, D. P., Freund, S. M., Bycroft, M. & Fersht, A. R. (2007). Four domains of p300 each bind tightly to a sequence spanning both transactivation subdomains of p53. Proc. Natl Acad. Sci. USA, 104, 7009–7014. Knight, C. G., Willenbrock, F. & Murphy, G. (1992). A novel coumarin-labelled peptide for sensitive continuous assays of the matrix metalloproteinases. FEBS Lett. 296, 263–266. Lindahl, E., Hess, B. & van der Spoel, D. (2001). GROMACS 3.0: a package for molecular simulation and trajectory analysis. J. Mol. Model. 7, 306–317. Jorgensen, W. L., Maxwell, D. S. & Tirado-Rives, J. (1996). Development and testing of the OPLS all-atom force field on conformational energetics and properties of organic liquids. J. Am. Chem. Soc. 118, 11225–11236. Jorgensen, W. L., Chandrasekhar, J., Madurd, J. D., Impey, R. W. & Klein, M. L. (1983). Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 79, 926–935. Darden, T., York, D. & Pedersen, L. (1993). Particle mesh Ewald: an N∑log(N) method for Ewald sums in large systems. J. Chem. Phys. 98, 10089–10092. Essmann, U., Perera, L., Berkowitz, M. L., Darden, T., Lee, H. & Pedersen, L. G. (1995). A smooth particle mesh Ewald method. J. Chem. Phys. 103, 8577–8593. Hess, B., Bekker, H., Berendsen, H. J. C. & Fraaije, J. G. E. M. (1997). LINCS: a linear constraint solver for molecular simulations. J. Comput. Chem. 18, 1463–1472. Berendsen, H. J. C., Postma, J. P. M., van Gunsteren, W. F., DiNola, A. & Haak, J. R. (1984). Molecular dynamics with coupling to an external bath. J. Chem. Phys. 81, 3684–3690. Seeber, M., Cecchini, M., Rao, F., Settanni, G. & Caflisch, A. (2007). Wordom: a program for efficient analysis of molecular dynamics simulations. Bioinformatics, 23, 2625–2627. Settanni, G. & Fersht, A. R. (2008). High temperature unfolding simulations of the TRPZ1 peptide. Biophys. J. 94, 4444–4453. Settanni, G., Rao, F. & Caflisch, A. (2005). Phi-value analysis by molecular dynamics simulations of reversible folding. Proc. Natl Acad. Sci. USA, 102, 628–633.