In Silico Structure-Based Design of a Potent and Selective Small Peptide Inhibitor of Protein Tyrosine Phosphatase 1B, A Novel Therapeutic

Target for Obesity and Type 2 Diabetes Mellitus: A Computer Modeling Approach

http://www.jbsdonline.com

Abstract

Protein Tyrosine Phosphatase 1B (PTP1B) has been shown to be a negative regulator of insulin signaling by dephosphorylating key tyrosine residues within the regulatory domain of the β-subunit of the insulin receptor. Recent gene knockout studies in mice have shown the mice to have increased insulin sensitivity and improved glucose tolerance. Furthermore, these mice also exhibited a resistance to diet induced obesity. Inhibitors of PTP1B would have the potential of enhancing insulin action by prolonging the phosphorylated state of the insulin receptor. In addition, recent clinical studies have shown that the haplotype ACTTCAG0 of the PTPN1 gene, which encodes PTP1B, is a major risk contributor to type 2 diabetes mellitus (T2DM). Thus, there is compelling evidence that small molecule inhibitors of PTP1B may be effective in treating insulin resistance at an early stage, thereby leading to a prevention strategy for T2DM and obesity.

Based on the crystal structure of the complex of PTP1B with a known inhibitor, we have identified a tetrapeptide inhibitor with the sequence WKPD. Docking calculations indicate that this peptide is as potent as the existing inhibitors. Moreover, the peptide is also found to be selective for PTP1B with a greatly reduced potency against other biologically important protein tyrosine phosphatases such as PTP-LAR, Calcineurin, and the highly homologous T-Cell Protein Tyrosine Phosphatase (TCPTP). Thus the designed tetrapeptide is a suitable lead compound for the development of new drugs against type 2 diabetes and obesity.

Introduction

Protein Tyrosine Phosphatase 1B (PTP1B) has recently been receiving consider-able attention not only in the studies of pathophysiology of insulin resistance in diabetes mellitus, but also as a drug target for the management of insulin resistant states such as obesity and type 2 diabetes mellitus (T2DM) (1). PTP1B is encod-ed by the PTPN1 gene which is located in 20q13; this genomic region has been linked to T2DM in multiple genetic studies. A recent study aimed at investigating the association of PTPN1 gene polymorphisms with measures of glucose homeo-stasis was undertaken in 811 Hispanic American subjects (2). The results of eight SNP haplotype analyses demonstrated a variable correlation with Si (insulin sen-sitivity) and fasting plasma glucose, respectively. The haplotype ACTTCAG0 was significantly associated with lower Si (i.e., greater insulin resistance) and high fasting glucose, and the haplotype CTCCTGT0 was significantly associated with higher Si (greater insulin sensitivity) and lower fasting glucose. Most signifi-cantly, the haplotypes ACTTCAG0 and CTCCTGT0, in another study, have inde-pendently been shown to be associated with T2DM risk and protection respec-

Journal of Biomolecular Structure &Dynamics, ISSN 0739-1102Volume 23, Issue Number 4, (2006)©Adenine Press (2006)

Gita Subba Rao1,*

Manoj V. Ramachandran1

J. S. Bajaj2

1Department of BiophysicsAll India Institute of Medical SciencesNew Delhi – 110029, India2Chief Consultant and DirectorDiabetes, Endocrine and Metabolic MedicineBatra Hospital and Medical Research CentreNew Delhi, India

Presented as poster at the 14th Conversation, June 14-18, 2005.

377Email: gitarao@aiims.ac.in

Open Access ArticleThe authors, the publisher, and the right hold-ers grant the right to use, reproduce, and dis-seminate the work in digital form to all users.

tively (3). Indeed in this latter study wherein pooled data analyses involving 300 unrelated Caucasian T2DM patients with end-stage renal disease (ESRD) and randomly ascertained Caucasian control subjects without known diabetes, along with similar data from 275 T2DM subjects from Diabetes Heart Study (DHS), showed that the haplotype ACTTCAG0 was a major risk contributor for T2DM indicating population-attributable risk (PAR) of 17-20%, based on different mod-els and assumptions. These recent clinical studies lend considerable credence to the earlier experimental studies wherein mice lacking the PTP1B gene exhibited increased insulin sensitivity and improved glucose tolerance. In addition PTP1B deficient mice were also found to be resistant to diet induced obesity (4, 5). As PTP1B has been shown to be a negative regulator of insulin signaling by dephos-phorylating key tyrosine residues within the regulatory domain of the β-subunit of insulin receptor (6), inhibitors of PTP1B would have the potential of prolonging the phosphorylated state of the insulin receptor and, hence, enhancing the down-stream metabolic events in the expression of biological effects of insulin. The above evidence also strongly suggests that selective, small molecule inhibitors of PTP1B may be effective in treating insulin resistance at an early stage thereby leading to a preventive strategy for T2DM, and obesity.

Several PTP1B inhibitors have been developed to date using a structure-based design approach (7, 8). Amongst these, one of the most active compounds is 3-({5-[(N-Acetyl-3-{4-[([(carboxycarbonyl)(2-carboxyphenyl)amino]-1-naphthyl}-L-alanyl) amino]pentyl}oxy)-2-naphthoic acid (compound 23) (8) (Fig. 1). The crystal structure of PTP1B with the inhibitors (Figs. 2 and 3) reveals that in addition to the phosphotyrosine binding site (catalytic site) (residues Cys215-Arg221) there is a second binding site (Site 2) (Arg24 and Arg254). Inhibitors that bind to both the sites are found to be highly potent with activities in the nanomolar range. Recently, a third binding site (Site 3) (residues Tyr46-Asp48) was also found to contribute to the potency and selectivity of inhibitors (9).

All the inhibitors that have been developed so far are either non-peptidic or pep-tidomimetic in nature. Our aim in this work is to design a small peptide inhibitor, which has interactions with all the three binding sites, has potency comparable with that of the known inhibitors and is selective for PTP1B as compared to the closely related PTP’s, such as TCPTP, PTP-LAR, and Calcineurin, a potent ser-ine/threonine phosphatase.

Materials and Methods

The starting point for the modeling studies was the crystal structure of the com-plex of PTP1B with a diaryloxaminic acid based inhibitor, compound 23 (8) (PDB 1D 1NNY) (10). We chose this complex for two reasons. Firstly, the inhibitor is amongst the most potent known to date (Kd = 22nM) and secondly, the WPD loop of PTP1B (Trp179-Ser187) has an open conformation similar to that of the unoccupied enzyme.

Molecular modeling studies were performed using SYBYL 6.81 (Tripos Inc.) run-ning on a Silicon Graphics O2 Workstation, and version 3.0 of AutoDock (11).

Ligand Setup

Using the SEARCH PROTEIN DATABASE facility in the BIOPOLYMER module of SYBYL6.81, a peptide of the desired sequence was searched for. The resulting fragments were superimposed, and a conformation from the largest cluster was selected. The selected peptide was modeled with a protonated amino group and a deprotonated carboxyl group. Gasteiger-Marsili charges were assigned to the atoms, and a conformational search was carried out using a simulated annealing molecular dynamics protocol. The ligand was heated to a temperature of 700K and

378

Subba Rao et al.

Figure 1: Chemical structure of the known diarylox-aminic acid based inhibitor, compound 23 (8).

379Selective Small Peptide

Inhibitor of PTP1B

then annealed to 200K. Fifty such cycles were carried out. The conformation obtained at the end of each cycle was further subjected to local energy minimiza-tion using the Powell method available in the MAXIMIN2 procedure of SYBYL6.81 until the r.m.s. gradient was less than 0.05 kcal/mol/Å2. The 50 energy-minimized structures were then superimposed and the conformation occur-ring in the major cluster was taken to be the most probable conformation.

Docking

Docking of PTP1B with the ligands was carried out with version 3.0 of the program AutoDock (11). This program combines a rapid energy evaluation through pre-calculated grids of affinity potentials with a variety of search algorithms to find suitable binding positions for a ligand on a given protein. The program allows torsional flexibility in the ligand but the protein is kept rigid. The energy function is reported to have a residual standard error of about 2 kcal/mole, corresponding to a factor of about 30 in relative affinity (Kd1/Kd2).

A standard protocol with an initial population of 50 randomly placed individuals, a maximum of 1.5 × 106 energy evaluations, a mutation rate of 0.02, a crossover rate of 0.80, and the elitism value of 1 were applied (12). Proportional selection was used, where the average of the worst energy was calculated over a window of the previous 10 generations. For the local search, the so-called pseudo-Solis and Wets algorithm was applied using a maximum of 300 iterations per local search. The probability of performing local search on an individual in the population was 0.6, and the maximum number of consecutive successes or failures before doubling or halving the local search step size was 4. Fifty independent docking runs were car-ried out for each ligand. Results differing by less than 1.5 Å in positional root-mean-square deviation (rmsd) were clustered together and represented by the result with the most favorable free energy of binding.

With the ligands kept fixed in their bound conformation as determined by the AutoDock computations, the docked complex was further refined in SYBYL 6.81 in order to relax the restriction of keeping the protein fixed. The SYBYL docking energy is the energy of interaction between the ligand and the enzyme within a box surrounding the ligand. The Tripos force field with Kollman united atom charges for the protein, Gasteiger-Marsili charges for the ligand, and a distance-dependent dielectric constant was used for calculating the docking energy. The refinement was done by first optimizing the hydrogen positions and then minimizing the docking energy with the help of the MINIMIZE_DOCK procedure, with the side chains of the protein kept flexible, until the r.m.s. gradient was less than 0.1 kcal/mol/ Å2.

Molecular Dynamics

In order to check whether the designed inhibitor remains bound in the presence of explicit solvent, a molecular dynamics simulation using the ADVANCED COMPUTATION module of SYBYL6.81 was carried out on a fully hydrated, three-layer water model that fully enclosed the final docked complex.

The dynamics protocol was similar to that used by Glover and Tracey (23) to study the inhibition of PTP1B by sulfotyrosine peptides. The first step was the energy minimization of the hydrated model during which the backbone of the enzyme and the ligand and the water molecules in the outer layer were kept fixed (to prevent evaporation of water during the simulation), until the r.m.s. gradient was less than 0.1 kcal.mol/Å2.

The minimized, hydrated complex was then subjected to a molecular dynamics simulation in three stages. In all the stages the SHAKE algorithm was applied to all bonds connected to hydrogen atoms and a time step of 2 fs was used. The back-

bone of the enzyme and the outer layer of water molecules were kept fixed. In the first stage the temperature of the system was raised from 0 to 300 K over 15 ps of simulation time. Next, the system was equilibrated over 20 ps, and finally the production run was carried out over another 100 ps.

Results and Discussion

Modeling of the Complex of PTP1B with a Known Diaryloxaminic Acid Based Inhibitor [Compound 23 (8)]

The AutoDock 3.0 methodology was first tested on a known inhibitor, which was docked in the region encompassing the three binding sites of PTP1B, after extracting the ligand from the crystal structure. Figure 4 shows the superimposition of the docked structure (in blue) over the crystal structure (in red). It can be seen that the two structures overlap very well with a positional root mean square deviation (rmsd) of 2.4Å.

The activity (Kd value) as predicted by AutoDock 3.0 was found to be 21nM which agrees well with observed Kd value of 22nM. Thus the AutoDock 3.0 methodology gives results that are consistent with the observations.

Design of the Peptide Inhibitor and Modeling of the PTP1B Inhibitor Complex

On the basis of crystallographic, kinetic, and peptide binding studies involving phosphotyrosine (pTyr) – containing peptides, it was found that D/E-pY-pY-K/R is a consensus substrate sequence for PTP1B (13). This peptide sequence contains both negatively and positively charged residues. In addition, almost all of the known, potent inhibitors contain aromatic rings. We, therefore, designed several tetrapeptide sequences containing an aromatic residue and Lys, a Pro residue for constraining the conforma-tion and a charged/uncharged residue at the C-terminal end. Each of these tetrapeptides was docked after a conformational energy search as described in the methods and the most suitable one (in terms of the docking energy and the interactions) was found to have the following sequence.

NH3+-WKPD-COO-

The final docked position of the designed inhibitor is shown in Figure 5, and the list of contacting residues (up to 4Å) is given in Table I. It can be seen from the Figure that the inhibitor makes hydrogen bonds with residues Ser216, Ala217, and Arg221 in the catalytic site, with Tyr46 and Asp48 in

380

Subba Rao et al.

CATALYTIC SITE

BINDING SITE 3

BINDING SITE 2

Figure 2: Overall structure of the catalytic domain of PTP1B. The secondary structure elements are shown in blue (helices), yellow (β-strands), and pink (loops and turns). The three binding sites are shown in green ball-and-stick rendering. The figure was produced using MOLSCRIPT (19) and rendering was done using RASTER 3D (20).

Figure 4: Docked position of compound 23 (10) (ligand in blue) superimposed upon the crystal structure position (ligand in red).

Figure 3: X-ray crystal structure of the complex of PTP1B with a known inhibitor [compound 23 (10)]. The inhibitor is shown in red. The hydrogen bonded residues of PTP1B are shown green and the other contacting residues are shown in blue.

W179 R221 C215

K120

Q266

T263

Y20

G259

R254

M258

R24

G28 D29

Q262

I219V49

D48

G220 G218

A217

S216

Y46

Wat

binding site 3, and has hydrophobic interactions (up to 4 Å) with Arg24 in the binding site 2. In addition, there is hydrogen bonding with Lys116. The ligand also has hydrophobic interactions with several other residues in the binding sites (colored blue in Fig. 5). A comparison with the list of contacts of compound 23 (see Table I) shows a similar binding region for the designed ligand. The AutoDock 3.0 results gave a frequency of the largest cluster (which was also the top ranked cluster) of 20/50, with a free energy of binding ΔG = -12.26 kcal/mole and a corresponding Kd value of 1.03nM. Thus the designed peptide has potency comparable with that of the most potent known inhibitors.

Molecular Dynamics Simulation of a Fully Hydrated Model of the Final Docked Complex

The effect of solvent on the binding of the designed peptide inhib-itor was studied by a molecular dynamics simulation of a fully hydrated model as described in the methods. During the produc-tion phase of 100 ps following the initial heating and equilibration phases, the total energy and the simulation temperature were found to remain steady with little fluctuation. The snapshots of the dynamics trajectory at 0, 25, 50, 75, and 100 ps of the production run are shown in Figure 6, while the corresponding interaction energies and sets are given in Table II.

381Selective Small Peptide

Inhibitor of PTP1B

K116

E115

K120

Wat

R221P3

K2

W1

D4

P3

K2

W1

A217

Q266

K116

R221 Y46

D48

Q262

S50

M258

G220

Q262

G259

R24D29

M258

I219

G218

Y46

D48

V49

S50

2.5Å 2.5Å3Å

2.5Å

3.5Å 2.7Å

2.3Å

3.2Å

2.5Å

2.5Å

2.8Å

2.5Å

2.8Å

3.2Å

R45

Figure 5: Final docked complex of PTP1B with designed peptide inhibitor. The color coding is the same as in Figure 2. Hydrogen bonds are shown as black dashed lines together with distances.

Figure 6: Molecular dynamics trajectory for the docked complex. Snapshots of the designed peptide and selected PTP1B active site residue conformaers extracted from the production dynamics trajectory at time intervals of 0, 25, 50, 75, and 100 ps. The peptide is shown in thick stick rendering and is color-coded by atom type; PTP1B residues are shown as thin sticks and colored magenta.

It can be seen that the positions of the N- and C-termini and of K2 and P3 remain stable and restrained by the enzyme over the course of the dynamics simulation. The N-terminus and K2, which were initially hydrogen-bonded to D48 of the enzyme (Fig. 5) continue to form hydrogen bonds with D48 throughout the simulation. This is also the case with the C-terminus which remains hydrogen-bonded to Y46. The backbone of the ligand, which was allowed to vary in the simulation, is also seen to remain fairly stable with hydrogen-bonding to residues S216 and A217 being retained throughout.

Fluctuations during the dynamics run are observed in residues W1 and D4 of the ligand. W1, which initially has van der Waals contacts with R24 of the enzyme, is seen to fluctuate towards a fluctuating M258, resulting in the formation of a hydro-gen bond. An even more dramatic effect is observed with the fluctuating ligand residue D4, which is initially hydrogen-bonded to R221 of the enzyme (Fig. 5). It can be seen that D4 and R221 mutually move away from each other, while D4 and K116 move towards each other resulting in the formation of a salt bridge between D4 and K116. The fluctuations result in a more stable position of the ligand in the enzyme binding site, with a significant lowering in the docked energy (Table II).

Docking results using AutoDock 3.0 give a free energy of binding in the hydrated system of DG = -14.55 kcal/mole, with a corresponding Kd value of 0.002 nm, which is significantly lower than that obtained without the explicit solvent (DG = -12.26; Kd = 1.03 nm). Also, the peptide inhibitor is found to bind in almost the same position in both the presence and the absence of the explicit solvent with a backbone r.m.s. deviation of 0.4 Å.

Thus, we have shown that the docked structure and the strong binding are retained in the presence of explicit solvent.

Selectivity Against Other Closely Related PTPases

The designed ligand was docked with T-Cell Protein Tyrosine Phosphatase (TCPTP) (PDB ID 1L8K) (14,18), PTP of LAR (PDB ID 1LAR) (15), and

382

Subba Rao et al.

Calcineurin (PDB ID 1MF8) (16). The AutoDock 3.0 results are shown in Table III. It can be seen that the designed peptide is 800- fold selective over TCPTP (Kd = 0.83 μM). This is much higher than the best selectivity of 10-fold achieved so far with the existing inhibitors (17). The designed peptide is also seen to have high selectivity over LAR and Calcineurin (Table III).

Conclusion

Using an in silico structure-based approach, we have designed a small peptide inhibitor of PTP1B. Docking studies show that the designed peptide has potency comparable to that of the known non-peptidic and peptidomimetic inhibitors. Dynamics simulations on a fully hydrated model show that the potency and strong binding are retained even in the presence of explicit solvent. Docking calculations with closely related TCPTP show a very high selectivity of 800 for PTP1B, as also high selectivity over PTP-LAR and Calcineurin.

The above results suggest that the designed tetrapeptide is a potent and selective inhibitor of PTP1B, and is a suitable lead compound for the development of new drugs against PTP1B.

Although peptides are generally known to have undesirable pharmacokinetic proper-ties, yet they have provided novel lead compounds and, in several cases, modified peptide analogs have been developed as drugs. With recent drug delivery techniques, the opportunities for peptide drug development have been significantly enhanced.

A similar approach has been used by us earlier in designing peptide inhibitors of HIV-1 integrase (21) and HIV-1 reverse transcriptase (22).

References and Footnotes

383Selective Small Peptide

Inhibitor of PTP1B

1.2.

3.

4.

5.

6.7.8.9.

10.

11.

J. L. Evans and B. Jallal. Exp. Opin. Invest. Drugs. 8, 139-160 (1999).N. D. Palmer, J. L. Bento, J. C. Mychaleckyj, C. D. Langefeld, J. K. Compbell, J. M. Norris, S. M. Haffner, R. N. Bergman, and D. W. Bowden. Diabetes 53, 3013-3019 (2004).J. L. Bento, N. D. Palmer, J. C. Mychaleckyj, L. A. Lange, C. D. Langefeld, S. S. Rich, B. L. Freedman, and D. W. Bowden. Diabetes. 53, 3007-3012 (2004).M. Elchebly, P. Payette, E. Michaliszyn, W. Cromlish, S. Collins, A. L. Loy, D. Normandin, A. Cheng, J. Himms-Hagen, C. C. Chan, C. Ramachandran, M. J. Gresser, M. L. Tremblay, and B. P. Kennedy. Science 283, 1544-1548 (1999).L. D. Klaman, O. Boss, O. D. Peroni, J. K. Kim, J. L. Martino, J. M. Zabolotny, N. Moghal, M. Lubkin, Y. B. Kim, A. H. Sharpe, A. Stricker-Krongrad, G. I. Shulman, B. G. Neel, and B. B. Kahn. Mol. Cell Biol.20, 5479-5489 (2000).J. C. H. Byon, A. B. Kusari, and J. Kusari. Mol. Cell. Biochem. 182, 101-108 (1998).T. O. Johnson, J. Ermolieff, and M. Jirousek. Nat. Rev. Drug Discovery 1, 696-709 (2002) G. Liu and J. M. Trevillyan. Curr. Opin. Invest. Drugs 3, 1608-1616 (2002). G. Liu, B. G. Szczepankiewicz, Z. Pei, D. A. Janowick, Z. Xin, P. J. Hajduk, C. Abad-Zapatero, H. Liang, C. W. Hutchins, S. W. Fesik, S. J. Ballaron, M. A. Stashko, T. Lubben, A. K. Mika, B. A. Zinker, J. M. Trevillyan, and M. R. Jirousek. J. Med. Chem. 46, 2093-2103 (2003)B. G. Szczepankiewicz, G. Liu, P. J. Hajduk, C. Abad-Zapatero, Z. Pei, Z. Xin, T. Lubben, J. M. Trevillyan, M. A. Stashko, S. J. Ballaron, H. Liang, F. Huang, C. W. Hutchins, S. W. Fesik, and M. R. Jirousek. J. Am. Chem. Soc. 125, 4087-4096 (2003).G. M. Morris, D. S. Goodsell, R. S. Halliday, R. Huey, W. E. Hart, R. K. Beleu, and A. J. Olson. J. Comput. Chem. 19, 1639-1662 (1998).C. A. Sotriffer, H. Ni, and J. A. McCammon. J. Med. Chem. 43, 4109-4117 (2000).A. Salmeen, J. N. Andersen, M. P. Meyers, N. K. Tonks, and D. Barford. Mol. Cell. 6, 1401-1412 (2000).

384

Subba Rao et al.

Md. J. Ibarra-Sanchez, P. D. Simonic, F. R. Nestel, P. Duplay, W. S. Lapp, and M. L. Trembley. Semin. Immunol. 12, 379 (2000).A. Cheng, N. Dube, F. Gu, and M. L. Trembley. European J. Biochem. 269, 1050 (2002).C. R. Kissinger, H. E. Parge, D. R. Knigton, C. T. Lewis, L. A. Pelletier, A. Tempczyk, V. J. Kalish, K. D. Tucker, R. E. Showalter, E. W. Moomaw, L. N. Gastinel, N. Habuka, X. Chen, F. Maldonado, J. E. Barker, R. Bacquet, and J. E. Villafranca. Nature 378, 641-644 (1995).K. Shen, Y.-F. Keng, L. Wu, X.-L. Guo, D. S. Lawrence, and Z. Y. Zhang. J. Biol. Chem. 276, 50: 47311-47319 (2001).H. M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H. Weissig, I. M. Shindyalov, and P. E. Bourne. Nucl. Acids Res. 20, 235-242 (2000)P. J. Kraulis. J. Appl. Crystal. 24, 946-950 (1990).E. A. Merrit and D. J. Bacon. Meth. Enzymol. 277, 505-524 (1997)G. Subba Rao, S. Bhatnagar, and V. Ahuja. J. Biomol. Struct. Dynam. 20, 31-38 (2002).G. Subba Rao and S. Bhatnagar. J. Biomol. Struct. Dynam. 21, 171-178 (2003).N. R. Glover and A. S. Tracey. Biochem. Cell Biology 77, 469-486 (1999).

Date Received: June 8, 2005

Communicated by the Editor Ramaswamy H Sarma