- Open Access
Modulation of p53 binding to MDM2: computational studies reveal important roles of Tyr100
© Dastidar et al; licensee BioMed Central Ltd. 2009
Published: 3 December 2009
The tumor suppressor protein p53 is regulated by the ubiquitin ligase MDM2 which down-regulates p53. In tumours with overexpressed MDM2, the p53-MDM2 interaction can be interrupted by a peptide or small molecule to stabilize p53 as a therapeutic strategy. Structural and biochemical/mutagenesis data show that p53 has 3 hydrophobic residues F19, W23 and L26 that embed into the ligand binding pocket of MDM2 which is highly plastic in nature and can modulate its size to accommodate a variety of ligands. This binding pocket is primarily dependent on the orientation of a particular residue, Y100. We have studied the role of the dynamics of Y100 in p53 recognition.
Molecular dynamics simulations show that the Y100 side chain can be in "open" or "closed" states with only the former enabling complex formation. When both p53 and MDM2 are in near native conformations, complex formation is rapid and is driven by the formation of a hydrogen bond between W23 of p53 and L54 of MDM2 or by the embedding of F19 of p53 into MDM2. The transition of Y100 from "closed" to "open" can increase the size of the binding site. Interconversions between these two states can be induced by the N-terminal region of MDM2 or by the conformations of the p53 peptides.
Molecular dynamics simulations have revealed how the binding of p53 to MDM2 is modulated by the conformational mobility of Y100 which is the gatekeeper residue in MDM2. The mobility of this residue can be modulated by the conformations of p53 and the Nterminal lid region of MDM2.
We have used molecular dynamics simulations to examine three different conformations of MDM2: (i) 'Open' state of the conformation taken from the wild type (WT) complex (1YCR) with the Y100 side chain pointing away from the binding pocket; (ii) 'closed' state - the conformation of MDM2 taken from the complex with α-helical P27S mutant of p53 with the Y100 side chain pointing towards the binding pocket; (iii) The 'apo' state - the unliganded state of MDM2, obtained from the ensemble of NMR structures  where Y100 is in a deeply buried position (see Figure 2).
Other approaches to investigate such phenomena are Brownian Dynamics simulations , Replica Exchange methods [14–17] and the more recently developed accelerated molecular dynamics methods . While the Brownian dynamics methods are well suited for studying protein-protein associations where proteins are depicted in a reduced representation , the replica exchange methods can be used to examine pathways of folding ; the accelerated molecular dynamics methods have been very successful in examining long time scale processes . In this study, we are attempting to understand atomic level details of the process prior to the embedding of p53 into MDM2. Aspects of this process involve a coupling between folding and binding, at least of the p53 peptide . We have chosen to run classical molecular dynamics simulations with several different starting conditions and as will be seen, these capture local folding/unfolding events through extensive surface rearrangements. A major factor that limits the usage of replica exchange methods here is that the current system is comprised of around 100 amino acids and replica exchange methods, particularly in explicit solvent, can only meaningfully (exhaustively) be applied to peptides that are up to 40 amino acids long [21–24] and hence are not suitable for the sort of extensive surface rearrangements that we sample here. Moreover, the MDM2 binding site is highly hydrophobic and there is the possibility that it will rapidly unfold in the REMD methods. For the same reason, the accelerated molecular dynamics method was not used.
We first outline briefly the currently accepted picture of the mechanism of interaction between p53 and MDM2 in their bound state. Crystallographic, biophysical and computational studies have traditionally shown that F19, W23 and L26 are the three critical residues of the transactivation domain (TA) of p53 which largely determine the stability of its complex with MDM2 [10, 25]. The residues F19 to L25 form an α-helical segment which has a hydrophobic face that subtends the side chains of the three hydrophobic residues F19, W23 and L26 which get embedded in the binding pocket of MDM2. In addition to the hydrophobic interactions between these 3 residues and MDM2, the W23 side chain also makes an HB with the backbone of L54 of MDM2, and this is very critical for the stability of the complex [7, 26]. More recently, it has been demonstrated that other parts of both p53 and MDM2 are involved in modulating these interactions [7, 12, 27]. One residue whose dynamics appear to potentiate this binding is Y100 of MDM2 which "gates" the conformation of the p53 peptide. This residue is conserved across species (see Supporing Information in ). If the p53 peptide is extended at its C-terminus (as seen in the crystal structure 1YCR), the side chain hydroxyl of Y100 is involved in hydrogen bond (HB) with the backbone of either E28 or N29 of the p53 peptide; however the p53 peptide can also adopt an α-helical conformation at its C-terminus and this is potentiated by the Y100 hydroxyl forming an HB with the backbone of L26 which provides a 'cozier' fit between the p53 peptide and MDM2 . We set out to investigate the mechanism that governs the development of these interactions as p53 approaches MDM2; we are particularly interested in examining the modulation of the conformational activity of Y100. Recently, a study has investigated such a process using targeted simulations of the binding pathway under a coarse grained description . In contrast, here we examine the process of binding, without directing the binding, and in an all-atom model with explicit solvation. The steady behavior of the root mean squared deviations (RMSD) of MDM2 in the various situations shows that the simulations are stable (see Figure S1 in Additional File 1).
p53 and 'open' MDM2
List of trajectories.
RMSD 3 Å
RMSD 6 Å
RMSD 3 Å
RMSD 3 Å
RMSD 6 Å
simulation started from the snapshot taken at 7.5 ns of Mc6A
Apo, model 1
RMSD 4 Å
Apo, model 1
RMSD 6 Å
Taken from M1n4A at the end of 20 ns
RMSD ~6 Å
Apo, model 1
Apo, model 2 (Residue 1-119)
Apo, model 4 (Residue 1-119)
In contrast, when p53 approaches the "open" state of MDM2 from a separation of ~6 Å (Trajectory Mo6A), it manages to reach the surface of MDM2 in the vicinity of the binding pocket (Figure S1 in Additional File 1) within 0.5 ns. However the interactions are nonspecific and p53 never manages to embed into MDM2 completely. The fact that it reaches the surface originates in the long range electrostatic fields that will no doubt steer the two molecules . The W23 side chain forms interactions with the side chain of F55 which appears to displace p53 to a position slightly away from its native (or crysatllographically observed) location. At the same time, Y100 flips in towards the binding pocket; this presumably happens to minimize the exposure to solvent of a very hydrophobic binding pocket of MDM2. Again the mobility of Y104 is correlated with that of Y100 (Figure S2 in Additional File 1).
p53 and 'closed' MDM2
p53 and 'apo' MDM2
To examine these later stages of binding we take the last (at a time of 20 ns) snapshot of the above trajectory and replace p53 (which has been somewhat distorted conformationally from its native bound conformation) by p53 taken from the crystal structure. We find that despite p53 being in its optimal conformation, in these simulations Y100 (which points in toward the binding pocket) does not enable p53 to embed and indeed pushes it away. In addition, the side chains of H96 and R97 obstruct the C-terminus of p53. Once again, Y100 is constrained by the presence of the C-terminal end of the TA of p53 (as we saw earlier in the case of p53 at 3 Å from the closed state) and cannot flip out to create space for W23 to embed. This indicates that prior to the binding of p53, the opening of the hydrophobic pocket of MDM2 by transition of Y100 to its 'open' conformation appears to be a key step to facilitate this binding. The interesting feature is that the mobility of Y104 is now dependant upon the position of the lid. The movies in the additional files 9, 10, 11 show that that once the lid opens, Y100 goes into an open state but Y104 is still constrained by the lid and does not open.
The TA of p53 is known to be largely intrinsically disordered with some parts of it adopting a helical conformation upon binding to MDM2. This brings three side chains (F19, W23 and L26) of p53 to be displayed on the same face of the helix thus enabling them to embed into a hydrophobic binding pocket of MDM2. Experimental data show that the unliganded state or the apo-state of MDM2 (derived using NMR, ) is quite different from its complexed state (derived using crystallography, ). What is not understood well is the process of binding of p53 and MDM2 to each other, mediated by these conformational states: is it pre-organization prior to binding or is it reorganization after binding (or induced fit). To test these, we have carried out a series of MD simulations that mimic the approach of p53 to MDM2. We avoided the simulation of folding/unfolding transitions of p53 during the simulations  and assumed that when p53 is very close to MDM2, stereo-chemical constraints would demand that the helical conformation of p53 be dominant in its interactions. We have mostly focused on the plasticity of MDM2 and the dynamics of Y100. The structural data shows that the plasticity of the binding pocket of MDM2 is mainly determined by the orientation of Y100 (Figures 1, 2), with spatially contiguous Y104 correlated with Y100 in mobility. When p53 is in its native (or crystallographically observed) state and MDM2 is "open", binding occurs with only small local reorganizations as side chains reorient minimally to maximize interactions. The size of the binding site, as determined from the solvent accessible surface area (data not shown) varies from ~2250 Å2 to ~2450 Å2 as the MDM2 transits from a relatively closed apo state to the p53 bound state. The general fluctuations of MDM2 (Figure S3 in Additional File 1) are conserved in pattern across the various simulations but the magnitudes vary. This is understandable because the peptide and MDM2 modulate each other and this will certainly change depending on their relative orientations, with the largest mobility witnessed in the presence of the lid (but absence of peptide). The one outlying feature is the high mobility of the 30-45 region when p53 is actually binding to MDM2 in approximately the crystallographically observed mode. While this region is distal to the p53 binding side, nevertheless the fact that it is high compared to the equilibrium dynamics  suggests that real equilibrium has not been achieved (this is ok for the purposes of our current study which only aims to examine the processes that occur as p53 approaches MDM2). When p53 is distant from MDM2, Y100 can assume both "in" and "out" states, with only the "out" state enabling binding. The binding process requires an initial encounter complex that is driven by nonspecific forces where electrostatic steering plays a major role. This is then accompanied by either the embedding of Phe19 or W23, that act as anchors across which the other two residues can be embedded. Both these residues, W23 and L26 require that Y100 is in an "out" conformation. This would enable the sidechain of W23 to make an HB with the backbone of L54 (which otherwise makes an HB with the hydroxyl of Y100). When Y100 is "in", it also occludes L26 from embedding. At the same time, when Y100 is "in", the C-terminal region of the TA of p53 also plays a critical role in modulating the dynamics of Y100 by preventing its transition to "open" states (at least in the timescales of these simulations).
Molecular dynamics simulations have revealed how the binding of p53 to MDM2 is modulated by the conformational mobility of Y100 which is the gatekeeper residue in MDM2. They also reveal how the mobility of this residue itself can be modulated by the conformations of p53 and by the conformations of the Nterminal lid region of MDM2.
The 'open' state of MDM2 (residues 25-109) was obtained from the crystal structure of the MDM2-p53 complex (RCSB entry 1YCR, resolved at 2.6 Å) . The 'closed' state of MDM2 (residues 25-109) was obtained from our previous work . The structure of the 'apo' state was chosen as Model 1 of the NMR ensemble  and although this structure consists of residues 1-119, we chose only residues 25-109 to be consistent with the other structures of MDM2 that have been used in the current and previous work. In order to examine the dynamics of the approach of p53 and MDM2 towards each other, we carry out all-atom simulations with MDM2 and p53 separated by varying distances (Table 1). We have also carried out simulations of MDM2 alone with its N-terminal 1-24 residues included as well. The CHARMM22 force field  was used to represent the systems. Each system was solvated using TIP3P water molecules and neutralized using counter-ions as required. After brief energy minimizations, each system was heated to 300 K followed by equilibration under constant pressure and temperature. Then molecular dynamics (MD) simulations at constant temperature and volume were carried out on each system for periods varying between 5-40 ns, yielding a total sampling time of ~200 ns. SHAKE  was applied to freeze the vibration of bonds involving hydrogen atoms, thus enabling a 2fs integration time step to be used. The Berendsen thermostat  was applied to keep the temperature constant. Simulations were carried out using the CHARMM package. Enthalpy was calculated using the MMGBSA approximations, with the GBSW [41, 42] implicit solvent model. Figures were prepared using PYMOL  and movies were generated using VMD .
Other papers from the meeting have been published as part of BMC Genomics Volume 10 Supplement 3, 2009: Eighth International Conference on Bioinformatics (InCoB2009): Computational Biology, available online at http://www.biomedcentral.com/1471-2164/10?issue=S3.
This work was supported by the Biomedical Research Council (Agency for Science, Technology and Research), Singapore. We thank Ted Hupp of Edinburgh University for discussions.
This article has been published as part of BMC Bioinformatics Volume 10 Supplement 15, 2009: Eighth International Conference on Bioinformatics (InCoB2009): Bioinformatics. The full contents of the supplement are available online at http://www.biomedcentral.com/1471-2105/10?issue=S15.
- Vogelstein B, Lane D, Levine AJ: Surfing the p53 network. Nature 2000, 408: 307–310. 10.1038/35042675View ArticlePubMedGoogle Scholar
- Bottger A, Bottger V, Sparks A, Liu WL, Howard SF, Lane DP: Design of a synthetic Mdm2-binding mini protein that activates the p53 response in vivo. Curr Biol 1997, 7(11):860–869. 10.1016/S0960-9822(06)00374-5View ArticlePubMedGoogle Scholar
- Moll UM, Petrenko O: The MDM2-p53 interaction. Mol Cancer Res 2003, 1(14):1001–1008.PubMedGoogle Scholar
- Sakurai K, Schubert C, Kahne D: Crystallographic analysis of an 8-mer p53 peptide analogue complexed with MDM2. J Am Chem Soc 2006, 128(34):11000–11001. 10.1021/ja063102jView ArticlePubMedGoogle Scholar
- Vassilev LT, Vu BT, Graves B, Carvajal D, Podlaski F, Filipovic Z, Kong N, Kammlott U, Lukacs C, Klein C, et al.: In vivo activation of the p53 pathway by small-molecule antagonists of MDM2. Science 2004, 303(5659):844–848. 10.1126/science.1092472View ArticlePubMedGoogle Scholar
- Zhong H, Carlson HA: Computational studies and peptidomimetic design for the human p53-MDM2 complex. Proteins 2005, 58(1):222–234. 10.1002/prot.20275View ArticlePubMedGoogle Scholar
- Dastidar SG, Lane DP, Verma CS: Multiple peptide conformations give rise to similar binding affinities: molecular simulations of p53-MDM2. J Am Chem Soc 2008, 130(41):13514–13515. 10.1021/ja804289gView ArticlePubMedGoogle Scholar
- Uhrinova S, Uhrin D, Powers H, Watt K, Zheleva D, Fischer P, McInnes C, Barlow PN: Structure of free MDM2 N-terminal domain reveals conformational adjustments that accompany p53-binding. J Mol Biol 2005, 350(3):587–598. 10.1016/j.jmb.2005.05.010View ArticlePubMedGoogle Scholar
- Buolamwini JK, Addo J, Kamath S, Patil S, Mason D, Ores M: Small molecule antagonists of the MDM2 oncoprotein as anticancer agents. Curr Cancer Drug Targets 2005, 5(1):57–68. 10.2174/1568009053332672View ArticlePubMedGoogle Scholar
- Kussie PH, Gorina S, Marechal V, Elenbaas B, Moreau J, Levine AJ, Pavletich NP: Structure of the MDM2 oncoprotein bound to the p53 tumor suppressor transactivation domain. Science 1996, 274(5289):948–953. 10.1126/science.274.5289.948View ArticlePubMedGoogle Scholar
- Keskin O, Gursoy A, Ma B, Nussinov R: Principles of protein-protein interactions: what are the preferred ways for proteins to interact? Chem Rev 2008, 108(4):1225–1244. 10.1021/cr040409xView ArticlePubMedGoogle Scholar
- Macchiarulo A, Giacche N, Carotti A, Baroni M, Cruciani G, Pellicciari R: Targeting the conformational transitions of MDM2 and MDMX: insights into dissimilarities and similarities of p53 recognition. J Chem Inf Model 2008, 48(10):1999–2009. 10.1021/ci800146mView ArticlePubMedGoogle Scholar
- Rojnuckarin A, Kim S, Subramaniam S: Brownian dynamics simulations of protein folding: access to milliseconds time scale and beyond. Proc Natl Acad Sci USA 1998, 95(8):4288–4292. 10.1073/pnas.95.8.4288PubMed CentralView ArticlePubMedGoogle Scholar
- Habeck M, Nilges M, Rieping W: Replica-exchange Monte Carlo scheme for bayesian data analysis. Phys Rev Lett 2005, 94(1):018105. 10.1103/PhysRevLett.94.018105View ArticlePubMedGoogle Scholar
- Nguyen PH, Mu Y, Stock G: Structure and energy landscape of a photoswitchable peptide: a replica exchange molecular dynamics study. Proteins 2005, 60(3):485–494. 10.1002/prot.20485View ArticlePubMedGoogle Scholar
- Nguyen PH, Stock G, Mittag E, Hu CK, Li MS: Free energy landscape and folding mechanism of a beta-hairpin in explicit water: a replica exchange molecular dynamics study. Proteins 2005, 61(4):795–808. 10.1002/prot.20696View ArticlePubMedGoogle Scholar
- Zhang W, Wu C, Duan Y: Convergence of replica exchange molecular dynamics. J Chem Phys 2005, 123(15):154105. 10.1063/1.2056540View ArticlePubMedGoogle Scholar
- Hamelberg D, Mongan J, McCammon JA: Accelerated molecular dynamics: a promising and efficient simulation method for biomolecules. J Chem Phys 2004, 120(24):11919–11929. 10.1063/1.1755656View ArticlePubMedGoogle Scholar
- De Rienzo F, Gabdoulline RR, Menziani MC, De Benedetti PG, Wade RC: Electrostatic analysis and Brownian dynamics simulation of the association of plastocyanin and cytochrome f. Biophys J 2001, 81(6):3090–3104. 10.1016/S0006-3495(01)75947-4PubMed CentralView ArticlePubMedGoogle Scholar
- Chen HF, Luo R: Binding induced folding in p53-MDM2 complex. J Am Chem Soc 2007, 129(10):2930–2937. 10.1021/ja0678774PubMed CentralView ArticlePubMedGoogle Scholar
- Chebaro Y, Mousseau N, Derreumaux P: Structures and thermodynamics of Alzheimer's amyloid-beta Abeta(16–35) monomer and dimer by replica exchange molecular dynamics simulations: implication for full-length Abeta fibrillation. J Phys Chem B 2009, 113(21):7668–7675. 10.1021/jp900425eView ArticlePubMedGoogle Scholar
- Nymeyer H: Energy landscape of the trpzip2 Peptide. J Phys Chem B 2009, 113(24):8288–8295. 10.1021/jp806749bView ArticlePubMedGoogle Scholar
- Singh G, Brovchenko I, Oleinikova A, Winter R: Demixing Transition of the Aqueous Solution of Amyloidogenic Peptides: A REMD Simulation Study. J Phys Chem B 2009. ASAP Article ASAP ArticleGoogle Scholar
- Takeda T, Klimov DK: Probing the effect of amino-terminal truncation for Abeta1–40 peptides. J Phys Chem B 2009, 113(19):6692–6702. 10.1021/jp9016773PubMed CentralView ArticlePubMedGoogle Scholar
- Bottger A, Bottger V, Garcia-Echeverria C, Chene P, Hochkeppel HK, Sampson W, Ang K, Howard SF, Picksley SM, Lane DP: Molecular characterization of the hdm2-p53 interaction. J Mol Biol 1997, 269(5):744–756. 10.1006/jmbi.1997.1078View ArticlePubMedGoogle Scholar
- Zondlo SC, Lee AE, Zondlo NJ: Determinants of specificity of MDM2 for the activation domains of p53 and p65: proline27 disrupts the MDM2-binding motif of p53. Biochemistry 2006, 45(39):11945–11957. 10.1021/bi060309gView ArticlePubMedGoogle Scholar
- Schon O, Friedler A, Bycroft M, Freund SM, Fersht AR: Molecular mechanism of the interaction between MDM2 and p53. J Mol Biol 2002, 323(3):491–501. 10.1016/S0022-2836(02)00852-5View ArticlePubMedGoogle Scholar
- Archakov AI, Govorun VM, Dubanov AV, Ivanov YD, Veselovsky AV, Lewi P, Janssen P: Protein-protein interactions as a target for drugs in proteomics. Proteomics 2003, 3(4):380–391. 10.1002/pmic.200390053View ArticlePubMedGoogle Scholar
- Baker NA, McCammon JA: Electrostatic interactions. Methods Biochem Anal 2003, 44: 427–440. full_textPubMedGoogle Scholar
- Carotti A, Macchiarulo A, Giacche N, Pellicciari R: Targeting the conformational transitions of MDM2 and MDMX: Insights into key residues affecting p53 recognition. Proteins 2009.Google Scholar
- Pazgier M, Liu M, Zou G, Yuan W, Li C, Li J, Monbo J, Zella D, Tarasov SG, Lu W: Structural basis for high-affinity peptide inhibition of p53 interactions with MDM2 and MDMX. Proc Natl Acad Sci USA 2009, 106(12):4665–4670. 10.1073/pnas.0900947106PubMed CentralView ArticlePubMedGoogle Scholar
- Showalter SA, Bruschweiler-Li L, Johnson E, Zhang F, Bruschweiler R: Quantitative lid dynamics of MDM2 reveals differential ligand binding modes of the p53-binding cleft. J Am Chem Soc 2008, 130(20):6472–6478. 10.1021/ja800201jView ArticlePubMedGoogle Scholar
- Worrall E, Wawrzynow B, Worrall L, Walkinshaw M, Ball K, Hupp TR: Regulation of the E3 ubiquitin ligase activity of MDM2 by an N-terminal pseudo-substrate motif. J Chem Biol 2009, 2(3):113–129. 10.1007/s12154-009-0019-5PubMed CentralView ArticlePubMedGoogle Scholar
- McCoy MA, Gesell JJ, Senior MM, Wyss DF: Flexible lid to the p53-binding domain of human Mdm2: implications for p53 regulation. Proc Natl Acad Sci USA 2003, 100(4):1645–1648. 10.1073/pnas.0334477100PubMed CentralView ArticlePubMedGoogle Scholar
- Henzler-Wildman KA, Lei M, Thai V, Kerns SJ, Karplus M, Kern D: A hierarchy of timescales in protein dynamics is linked to enzyme catalysis. Nature 2007, 450(7171):913–916. 10.1038/nature06407View ArticlePubMedGoogle Scholar
- Candeias MM, Malbert-Colas L, Powell DJ, Daskalogianni C, Maslon MM, Naski N, Bourougaa K, Calvo F, Fahraeus R: P53 mRNA controls p53 activity by managing Mdm2 functions. Nat Cell Biol 2008, 10(9):1098–1105. 10.1038/ncb1770View ArticlePubMedGoogle Scholar
- Ferreon JC, Lee CW, Arai M, Martinez-Yamout MA, Dyson HJ, Wright PE: Cooperative regulation of p53 by modulation of ternary complex formation with CBP/p300 and HDM2. Proc Natl Acad Sci USA 2009, 106(16):6591–6596. 10.1073/pnas.0811023106PubMed CentralView ArticlePubMedGoogle Scholar
- MacKerell AD Jr, Bashford D, Bellott M, Dunbrack RL Jr, Evanseck JD, Field MJ, Fischer S, Gao J, Guo H, Ha S, et al.: All-Atom Empirical Potential for Molecular Modeling and Dynamics Studies of Proteins. Journal of Physical Chemistry B 1998, 102: 3586–3616. 10.1021/jp973084fView ArticleGoogle Scholar
- Ryckaert JP, Ciccotti G, Berendsen HJC: J Comput Phys. 1977, 23: 327–341. 10.1016/0021-9991(77)90098-5View ArticleGoogle Scholar
- Berendsen HJC, Postma JPM, Van Gunsteren WF, Dinola A, Haak JR: Molecular-Dynamics with Coupling to an External Bath. J Chem Phys 1984, 81(8):3684–3690. 10.1063/1.448118View ArticleGoogle Scholar
- Im W, Feig M, Brooks CL 3rd: An implicit membrane generalized born theory for the study of structure, stability, and interactions of membrane proteins. Biophys J 2003, 85(5):2900–2918. 10.1016/S0006-3495(03)74712-2PubMed CentralView ArticlePubMedGoogle Scholar
- Im W, Lee MS, Brooks CL 3rd: Generalized born model with a simple smoothing function. J Comput Chem 2003, 24(14):1691–1702. 10.1002/jcc.10321View ArticlePubMedGoogle Scholar
- DeLano WL: The Pymol molecular graphics system. In The Pymol molecular graphics system. San Carlos, CA USA: DeLano Scientific; 2002.Google Scholar
- Humphrey W, Dalke A, Schulten K: VMD-visual molecular dynamics. Journal of Molecular Graphics 1996, 14: 33–38. 10.1016/0263-7855(96)00018-5View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.