Skip to main content
NCBI home page
Glycobiology logoLink to Glycobiology
. 2016 Oct 18;26(10):1041–1047. doi: 10.1093/glycob/cww073

Estimating glycosaminoglycan–protein interaction affinity: water dominates the specific antithrombin–heparin interaction

Aurijit Sarkar 2,3,1, Wenbo Yu 4, Umesh R Desai 3, Alexander D MacKerell 4, Philip D Mosier 3,1
PMCID: PMC5072149  PMID: 27496757

Abstract

Glycosaminoglycan (GAG)–protein interactions modulate many important biological processes. Structure–function studies on GAGs may reveal probes and drugs, but their structural complexity and highly acidic nature confound such work. Productivity will increase if we are able to identify tight-binding oligosaccharides in silico. An extension of the CHARMM force field is presented to enable modeling of polysaccharides containing sulfamate functionality, and is used to develop a reliable alchemical free-energy perturbation protocol that estimates changes in affinity for the prototypical heparin–antithrombin system to within 2.3 kcal/mol using modest simulation times. Inclusion of water is crucial during simulation as solvation energy was equal in magnitude to the sum of all other thermodynamic factors. In summary, we have identified and optimized a reliable method for estimation of GAG–protein binding affinity, and shown that solvation is a crucial component in GAG–protein interactions.

Keywords: CHARMM, desolvation, free-energy perturbation, molecular dynamics, specificity

Introduction

Sulfated glycosaminoglycans (GAGs) are highly acidic polymeric molecules mostly found in the extracellular matrix. Their location and highly charged nature impart a central role in cellular signaling (Häcker et al. 2005), inflammation (Taylor and Gallo 2006), cancer (Sasisekharan et al. 2002), host–pathogen interactions (Aquino et al. 2010) and hemostasis (Huntington 2003). It would therefore be very useful for the research community to identify a reduced set of GAG structures that are selected for high affinity and specificity toward a particular protein target in a rational manner. Considering their high degree of structural complexity and synthetically challenging nature, a computational technique capable of reliably predicting GAG–protein binding affinity is required.

Computational evaluation of GAG–protein interactions is in its infancy. We have previously identified GAG species capable of binding various proteins using docking and simulation techniques (Raghuraman et al. 2006, 2010; Sankaranarayanan and Desai 2014; Joseph et al. 2015). Others have also used similar methods to study GAG–protein interactions (Mulloy and Forster 2008; Samsonov et al. 2011; Sapay et al. 2011; Gandhi et al. 2012; Mottarella et al. 2014). Such studies have often yielded experimentally validated results, but the ability to segregate high-affinity ligands with confidence has yet to be demonstrated. It is also extremely important to identify various factors that affect GAG–protein interactions. We have previously postulated that solvation and desolvation upon binding must play an important part (Mosier et al. 2012; Sarkar and Desai 2015). We showed that GAGs preferentially bind residues that exist within a region of high positive charge density manifested by nearby arginine or lysine residues, but possess lower desolvation penalties (Sarkar and Desai 2015). Similar observations about the importance of entropic and desolvation factors have been made previously for GAGs binding to platelet endothelial cell adhesion molecule 1 and annexin A2 (Gandhi and Mancera 2009). However, such studies would be strongly supported if we were to study solvation phenomena within the context of an appropriate, structure–function data set providing a continuum of data points. Antithrombin and the specifically binding, high-affinity heparin pentasaccharide (DEFGH) comprise one such system (Meagher et al. 1996; Kridel and Knauer 1997; Desai et al. 2000; Arocas et al. 2001), which we (Mosier et al. 2012; Sarkar and Desai 2015) have employed previously (see Figure 1). No other GAG–protein affinity data set has been studied to the same extent.‬‬‬‬‬‬‬‬‬.

Fig. 1.

Fig. 1.

Open in a new tab

(A) Structure of the DEFGH sequence for which antithrombin has specificity. (B) Interaction of antithrombin with a DEFGH analog (ball-and-stick representation), as exemplified by PDB entry 1TB6. Basic residues of the pentasaccharide binding site (PBS; K11, R13, R46, R47, K114, K125, R129) and extended heparin binding site (EHBS; R132, K133, N135, K136) are shown as capped sticks. Crystallographic water molecules near the PBS and EHBS (w1 and w2) are also shown, highlighting their involvement binding site conformation and in recognition of the ligand

Free-energy perturbation (FEP) is a technique extensively used to assess the free energy of binding involving interactions of biomolecules via a thermodynamic cycle involving the interacting species (Shirts and Mobley 2013). It is often called “alchemical” because of its ability to simulate physically unrealistic states that can be very useful in assessing free-energy changes as a function of changing molecular structure, a very useful concept in structure–function relationship studies. Figure 2 illustrates the cycle as applied to the DEFGH–antithrombin system employed here. These methods can accurately estimate drug–protein interaction binding free energies (Wang et al. 2015). Yet, the buried, largely hydrophobic nature of typical drug–protein interactions is very different from the surface-exposed, highly charged interactions between GAGs and proteins. Therefore, it is of interest to investigate whether this technology might be useful for GAG–protein interactions as well. Our data set (vide supra) gave us the opportunity to explore, evaluate and optimize an FEP methodology for use in GAG–protein interactions. It also allowed us to gain insight into how solvation effects contribute to GAG–protein binding.

Fig. 2.

Fig. 2.

Open in a new tab

Thermodynamic cycle employed to assess the binding affinity of mutants for DEFGH in the FEP procedure.

Here, we report that FEP simulations, as implemented in NAMD (Phillips et al. 2005), using newly derived simulation parameters for sulfamate linked to carbohydrates and the CHARMM Additive All-Atom Force Field (MacKerell et al. 1998; Guvench et al. 2009; Raman et al. 2010; Guvench et al. 2011; Mallajosyula et al. 2012) (see Supplementary data) are able to estimate how mutagenesis changes the affinity of antithrombin for DEFGH. We also discuss the various strengths and weakness of such simulations. Finally, we demonstrate that solvation energy is a significant contributor to binding free energy and that vacuum simulations are unlikely to produce good estimates of affinity.‬‬‬‬‬‬‬‬‬

‬Results and discussion

The antithrombin–DEFGH data set is a prototype for affinity and specificity

The effect of protein structure on binding free energy for the antithrombin–DEFGH system has been studied extensively (Meagher et al. 1996; Kridel and Knauer 1997; Desai et al. 2000; Arocas et al. 2001). Wild-type antithrombin binds DEFGH with high (i.e. low-nanomolar) affinity (Olson et al. 1992; Jin et al. 1997). Various site-directed mutants of antithrombin have been generated and tested for their ability to bind DEFGH in highly accurate enzyme kinetics-derived assays, giving a range of ~106-fold change in affinity (Table I). Most mutants (or double mutants) therein involve alteration of a basic amino acid residue to a polar uncharged or non-polar amino acid residue, to investigate the typically charged nature of GAG–protein interactions. The exception to this rule is the N135A mutation that eliminates differences in affinity of antithrombin for heparin between glycosylated and non-glycosylated forms and also enhances the affinity (Olson et al. 1997). This data set also provides us with an archetypal platform to judge whether we are able to segregate specific interactions from nonspecific ones. Wild-type antithrombin binds the DEFGH-containing anticoagulant heparin with a very high affinity, which leads to co-localization of heparin, antithrombin and thrombin. Ternary complex formation ensues, ultimately leading to accelerated deactivation of thrombin by antithrombin. Thus, specific recognition of DEFGH by antithrombin is necessary for effective inhibition of thrombin. The N135A-K114A and N135A-R129H mutations lose this specific recognition, as evidenced by a decrease in catalytic efficiency (Meagher et al. 1996; Kridel and Knauer 1997; Desai et al. 2000; Arocas et al. 2001). The N135A-K114A mutation, in particular, loses almost 106-fold affinity, bringing it closer to the prototypical nonspecific thrombin–heparin system (Xu and Esko 2014). It may therefore be generalized that loss of specific interactions in antithrombin-like systems will be highlighted by a precipitous decrease in affinity. Conversely, similar highly specific interactions will confer manifold higher affinity. Thus, the ability to assess the change in affinity for a GAG–protein system accurately is the de facto ability to segregate specific/nonspecific interactions in such systems. Furthermore, any factors that significantly affect binding are consequently essential for specificity.‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬.

Table I.

Experimental versus predicted change in free energy for the heparin pentasaccharide binding to antithrombin and mutants

ID Mutant ΔΔGOBS a ΔΔGaq b ΔΔGvac c ΔΔGsolv d
A wte 0 1.6 ± 1.4 –5.1 ± 1.6 6.7 ± 3.0
B K125Q 0.4 3.7 ± 1.1 –12.8 ± 0.9 16.5 ± 2.0
C K136T 0 –0.7 ± 0.7 –2.9 ± 0.5 2.2 ± 1.2
D N135A –0.6 –0.2 ± 0.2 4.0 ± 0.2 –4.2 ± 0.4
E N135A-K114A 7.6 11.4 ± 0.9 –4.5 ± 0.9 15.9 ± 1.8
F N135A-R129H 2.9 4.6 ± 2.4 –17.3 ± 1.0 21.9 ± 3.4
G R132M 1.6 0.9 ± 0.8 –8.1 ± 0.6 9.0 ± 1.4
H K133M 2 5.6 ± 1 11.9 ± 1.2 –6.3 ± 2.2
I N135A-R129Q 3.4 –1.6 ± 3.3f –21.9 ± 1.7 20.0 ± 5.0f
Open in a new tab

aΔΔGOBS is the experimentally measured change in binding affinity reported earlier (Sarkar and Desai 2015) based on literature (Kridel et al. 1996; Meagher et al. 1996; Desai et al. 2000; Arocas et al. 2001).

bΔΔGaq is the predicted free energy of binding under aqueous conditions.

cΔΔGvac is the free energy of binding under vacuum conditions.

dΔΔGsolv represents the solvation energy contribution to binding and is equal to ΔΔGaq – ΔΔGvac.

eThe wt-to-wt FEP simulation was performed by conducting a K125K mutation.

fFEP simulation did not converge in the aqueous phase.

All values are reported in kcal/mol units.

FEP calculations are able to reproduce experimentally observed changes in affinity

We conducted FEP calculations to assess the change in free energy of binding of DEFGH to antithrombin upon site-directed mutagenesis; these data are presented in Table I. Each simulation was conducted in forward and reverse paths, and the hysteresis between these paths, measured by Bennett Acceptance Ratio (BAR) analysis (Bennett 1976), provided an estimate of error. Details of the experimental design can be found in the “Materials and methods” section and Supplementary data. The direct measurement of GAG–protein affinity by decoupling DEFGH from the system was difficult due to the “wandering ligand” problem (Gumbart et al. 2013): Decoupling the highly charged DEFGH from the system caused the ligand to drift away from the binding site irreversibly. Because microreversibility is a requirement for any thermodynamic process, such simulations cannot be used for our purposes. However, we found that mutation of antithrombin bound to DEFGH can be simulated with relative ease and these simulations are much more stable.

As expected, the N135A-K114A double mutant demonstrated the highest change in affinity for DEFGH, while K125Q, K136T and N135A mutants demonstrate affinities similar to that of wild-type antithrombin (Table I). A good correlation was observed between experimental and predicted affinities for all mutants studied. We were able to estimate experimental free-energy changes to an root-mean-square error (RMSE) of 2.3 kcal/mol for all antithrombin mutants (Figure 3A), which roughly evaluates to an error of estimating affinity within 50-fold. One mutant from our data set, N135A-R129Q, could not be simulated to convergence (ΔΔGaq =–1.6 ± 3.3 kcal/mol, ΔΔGOBS = 3.4 kcal/mol), but even on including this data point our predictions fall within ~100-fold error (RMSE = 2.8 kcal/mol), allowing reasonably accurate and practically useful estimates of GAG–protein interactions. Further work is needed to confirm the ability of FEP to assess the effect of structural changes in GAGs. However, this report demonstrates that FEP can be developed to identify high-affinity GAG sequences.

Fig. 3.

Fig. 3.

Open in a new tab

Correlation between experimental and predicted free energies of binding. (A) FEP calculations run with 500 ps per FEP window were able to predict the change in free energy for the specific heparin pentasaccharide DEFGH binding with various antithrombin mutants with an RMSE of ~2.3 kcal/mol. (B) Simulations run with 240 ps per FEP window were significantly less accurate than those run at 500 ps per window, resulting in an RMSE of ~3.9 kcal/mol. Mutants are identified by the letter code given in Table I. The one-to-one line is shown as a solid line; dashed lines represent ±2 kcal/mol from the one-to-one line. (C) Predictions of binding free energy using vacuum simulations are compared to experimentally measured values. The importance of desolvation effects in antithrombin–DEFGH binding is evidenced by the observation that vacuum simulations cannot estimate the free-energy changes for DEFGH–antithrombin/mutant interactions. (D) The magnitude of desolvation energy (dark bars) and free energy of binding in vacuum simulations (light bars) are shown for wt antithrombin and its mutants. Desolvation contributes significantly to binding and its magnitude is comparable to that of all other factors combined.

The antithrombin–DEFGH system presents a unique opportunity to develop computational tools capable of segregating high affinity binding GAG sequences from those with low affinity. The availability of the antithrombin–thrombin–heparin ternary complex structure (Li et al. 2004) is fortuitous because antithrombin is present in its reactive center loop-expelled conformation, which is the biologically relevant GAG–bound equilibrium state. FEP calculations in this state are likely to yield good results. However, this methodology is unlikely to yield good results if GAGs induce large structural changes in a protein structure as significant enhancement of sampling will be required to reversibly simulate the system. Therefore, one limitation of the FEP methodology is that the protein should not undergo large structural changes upon binding GAGs. However, further development of the method may yield ways to overcome this problem.‬‬‬‬‬‬‬‬‬

Unlike antithrombin, other systems where specific binding is observed will often demonstrate far less significant differences in affinity between tight binders and average ones. Wild-type basic fibroblast growth factor (FGF2), for example, demonstrates a maximum of ~2 kcal/mol difference in binding free energy with mutants deficient in their ability to bind heparin (Thompson et al. 1994). Far more stringent FEP calculations will be required to identify specific sequences in such systems. On the other hand, slight increases in simulation length significantly reduce error in our simulations (Figure 3B). Therefore, it is likely that the length of simulations required to identify low-affinity-specific interactions, such as those of FGF2, is possibly within reach for most research-intensive institutions. Whether FEP methods can accurately predict changes in affinity for other systems as well as it did with antithrombin will become clear only upon obtaining considerable structure–function data and retrospective evaluation. Based on our current observations, however, we are fairly confident that FEP calculations are a good way of distinguishing tight binders from average at the very least. Hence, FEP calculations will likely become invaluable tools for glycobiologists and others searching for highly specific GAG sequences for use as chemical biology probes and drugs.‬‬‬‬‬‬‬‬‬

Substantial computational power is still required for such simulations. Our simulations took 2 weeks per mutation reported above on 64 CPU cores. Given that GAGs demonstrate a higher combinatorial diversity than proteins and DNA (Sasisekharan and Venkataraman 2000), FEP cannot be used to screen large GAG databases. Instead, docking-based methods such as our dual-filter CVLS approach (Raghuraman et al. 2006, 2010) are designed to address this GAG property. Based on our recent observations (Boothello et al. 2015), we also propose that a computational fragment-based design approach may prove to be fruitful, although specialized methods will need to be developed for this purpose.‬‬‬‬‬‬‬‬‬

(De)solvation energy is the dominant contributor towards binding

Traditionally, the scientific community has described GAG–protein interactions as mostly electrostatic, with significant hydrogen bond/nonionic contributions separating specific and nonspecific interactions (Mosier et al. 2012; Xu and Esko 2014). If electrostatic interactions between arginine or lysine residues and GAG sulfate groups are sufficient to explain changes in affinity, in theory we should be able to reproduce experimental free energies of binding using only vacuum simulations. This would be most desirable, because explicit solvent simulations take significantly longer to complete than vacuum simulations. Therefore, we also estimated the change in free energy of binding DEFGH for each antithrombin mutant in vacuum. This is done by using the same procedures as described above, but in the absence of solvent. The difference between free energy of binding in solution (ΔΔGaq) and the same in vacuum (ΔΔGvac) allows us to assess the role of solvent in binding (ΔΔGsolv). Table I and Figure 3C and D demonstrate our findings.

ΔΔGvac does not correlate well with experimentally obtained ΔΔGOBS values (Figure 3C, r2 ~0.05, RMSE ~13 kcal/mol). Affinity was overestimated significantly in most cases. Quite clearly, vacuum simulations cannot be used successfully to estimate the affinity of heparin–antithrombin interactions. Finally, because the magnitude of ΔΔGsolv is comparable to the sum of all other factors involved in binding comprising ΔΔGvac, (Figure 3D), desolvation energy is a significant factor that dominates heparin–antithrombin interaction specificity.

Because GAGs are typically found in the extracellular matrix and are very highly charged in nature, it was always probable that desolvation plays an important part in GAG–protein interactions. This study provides concrete evidence to support this notion in the heparin–antithrombin system. This also supports our earlier assertion (Sarkar and Desai 2015) that GAGs will preferentially bind in a manner that lowers desolvation penalties. Therefore, it is highly likely that desolvation energy will contribute just as much to GAG–protein specificity as electrostatic interactions. Further studies comparing nonspecific GAG–protein interactions with the specific antithrombin–DEFGH will reveal the true nature of GAG–protein interaction specificity. Our prototypical simulations can be used as a general starting point for studying these GAG–protein systems.

At this time it is clear that GAG–protein interactions are not just dependent on interactions of sulfate groups with arginine and lysine, but also on water. This strongly indicates that computational methods that do not consider water as a powerful contributor may not be reliable at all times. Such methods may provide raw, initial data, but must be fine-tuned by considering the role of solvent explicitly.

Materials and methods

FEP simulations

FEP molecular dynamics (MD) simulations were conducted using the CHARMM Additive All-atom Force Field (MacKerell et al. 1998; Guvench et al. 2009; Raman et al. 2010; Guvench et al. 2011; Best et al. 2012; Mallajosyula et al. 2012), from which all simulated atoms (i.e. protein, GAG, water and ions) were assigned parameters. To facilitate the modeling of heparin-like GAGs (e.g. DEFGH), an extension of the existing CHARMM force field to enable the modeling of sulfamate groups linked to carbohydrates was developed and is described in the Supplementary data. These parameters were derived using a well-established protocol that we have previously described, for example, to obtain parameters for phosphate and sulfate groups linked to carbohydrates (Mallajosyula et al. 2012).

Initial coordinates for antithrombin and the antithrombin-specific heparin pentasaccharide DEFGH were extracted from PDB code 1TB6 (Li et al. 2004). Because 1TB6 contains a heparin mimic, structural changes were made to obtain the pentasaccharide structure. The charged form of Arg, Lys, Asp and Glu residues and the neutral delta-protonated form of His were used. The protein–ligand complex was placed in a box of TIP3P water, padded in all directions by at least 15 Å. The solvated complex was then neutralized using sodium chloride and minimized for 100,000 steps.‬‬‬‬‬‬‬‬‬

The FEP implementation in NAMD (Phillips et al. 2005) and VMD (Humphrey et al. 1996) was used to set up and conduct the simulations. All simulations were conducted in forward and backward directions, and the BAR method (Bennett 1976), as implemented in VMD, was used to assess hysteresis between the two paths. Initial procedures consisted of 20 ps of equilibration for each FEP window followed by 60 ps of simulation. Appropriately sized periodic boundary conditions were used and pressure/volume control was exerted using Langevin dynamics algorithms. The Particle Mesh Ewald method was used to calculate electrostatic potentials in a grid (1 Å spacing was used). A cut-off of 11 Å was employed. Once the system appropriately equilibrated, 20 windows (0.5 ns each) were used for the FEP run starting from coordinates and velocities produced during equilibration. The “dual topology” approach was used to achieve mutations and the energy was recorded for each window. A soft core potential (Beutler et al. 1994; Zacharias et al. 1994) was used during the FEP run.‬‬‬‬‬‬‬‬‬

A separate simulation was run in the absence of water box, to assess binding interactions in vacuum. This ensured that not only the electrostatic effects of water are accounted for, but also the effects of water that are related to solute conformation and physical barriers to molecular motion. Similar parameters were used for both vacuum and solvated simulations, but without PBCs and pressure/temperature controls.‬

Supplementary data

Supplementary data for this article including the derivation of CHARMM sulfamate parameters are available online at http://glycob.oxfordjournals.org/.

Supplementary Data
supp_26_10_1041__index.html (1.1KB, html)

Acknowledgements

The authors would like to thank Dr. J. Neel Scarsdale for assistance in running simulations on the compute cluster.

Funding

This work was supported by the Samuel Waxman Cancer Foundation and National Institutes of Health (R01 GM070855 to A.D.M., P01 HL107152, R01 HL090586 and R25 HL128639 to U.R.D.) and the National Center for Research Resources (S10 RR027411 to Virginia Commonwealth University).

Conflict of interest statement

None declared.

Abbreviations

BAR, Bennett Acceptance Ratio; DEFGH, high-affinity heparin pentasaccharide; FEP, free-energy perturbation; FGF2, basic fibroblast growth factor; GAG, glycosaminoglycan; MD, molecular dynamics; RMSE, root-mean-square error.

References

  1. Aquino RS, Lee ES Park PW. 2010. Diverse functions of glycosaminoglycans in infectious diseases. Prog Mol Biol Transl Sci. 93:373–394. [DOI] [PubMed] [Google Scholar]
  2. Arocas V, Bock SC, Raja S, Olson ST Björk I. 2001. Lysine 114 of antithrombin is of crucial importance for the affinity and kinetics of heparin pentasaccharide binding. J Biol Chem. 276:43809. [DOI] [PubMed] [Google Scholar]
  3. Bennett CH. 1976. Efficient estimation of free energy differences from Monte Carlo data. J Comput Phys. 22:245–268. [Google Scholar]
  4. Best RB, Zhu X, Shim J, Lopes PEM, Mittal J, Feig M MacKerell AD Jr. 2012. Optimization of the additive CHARMM all-atom protein force field targeting improved sampling of the backbone ϕ, ψ and side-chain χ1 and χ2 dihedral angles. J Chem Theory Comput. 8:3257–3273. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Beutler TC, Mark AE, van Schaik RC, Gerber PR van Gunsteren WF. 1994. Avoiding singularities and numerical instabilities in free energy calculations based on molecular simulations. Chem Phys Lett. 222:529–539. [Google Scholar]
  6. Boothello RS, Sarkar A, Tran VM, Nguyen TKN, Sankaranarayanan NV, Mehta AY, Alabbas A, Brown S, Rossi A Joice AC, et al. 2015. Chemoenzymatically prepared heparan sulfate containing rare 2-O-sulfonated glucuronic acid residues. ACS Chem Biol. 10:1485–1494. [DOI] [PubMed] [Google Scholar]
  7. Desai U, Swanson R, Bock SC, Björk I Olson ST. 2000. Role of arginine 129 in heparin binding and activation of thrombin. J Biol Chem. 275:18976. [DOI] [PubMed] [Google Scholar]
  8. Gandhi NS, Freeman C, Parish CR Mancera RL. 2012. Computational analyses of the catalytic and heparin-binding sites and their interactions with glycosaminoglycans in glycoside hydrolase family 79 endo-β-d-glucuronidase (heparanase). Glycobiology. 22:35–55. [DOI] [PubMed] [Google Scholar]
  9. Gandhi NS, Mancera RL. 2009. Free energy calculations of glycosaminoglycan–protein interactions. Glycobiology. 19:1103. [DOI] [PubMed] [Google Scholar]
  10. Gumbart JC, Roux B Chipot C. 2013. Standard binding free energies from computer simulations: What is the best strategy. J Chem Theory Comput. 9:794–802. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Guvench O, Hatcher E, Venable RM, Pastor RW MacKerell AD Jr.. 2009. CHARMM additive all-atom force field for glycosidic linkages between hexopyranoses. J Chem Theory Comput. 5:2353. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Guvench O, Mallajosyula SS, Raman EP, Hatcher E, Vanommeslaeghe K, Foster TJ, Jamison FW II MacKerell AD Jr. 2011. CHARMM additive all-atom force field for carbohydrate derivatives and its utility in polysaccharide and carbohydrate–protein modeling. J Chem Theory Comput. 7:3162–3180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Häcker U, Nybakken K Perrimon N. 2005. Heparan sulphate proteoglycans: The sweet side of development. Nat Rev Mol Cell Biol. 6:530–541. [DOI] [PubMed] [Google Scholar]
  14. Humphrey W, Dalke A Schulten K. 1996. Vmd: Visual molecular dynamics. J Mol Graph. 14:33–38. [DOI] [PubMed] [Google Scholar]
  15. Huntington JA. 2003. Mechanisms of glycosaminoglycan activation of the serpins in hemostasis. J Thromb Haemostasis. 1:1535–1549. [DOI] [PubMed] [Google Scholar]
  16. Jin L, Abrahams JP, Skinner R, Petitou M, Pike RN Carrell RW. 1997. The anticoagulant activation of antithrombin by heparin. Proc Natl Acad Sci U S A. 94:14683. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Joseph PRB, Mosier PD, Desai UR Rajarathnam K. 2015. Solution NMR characterization of chemokine CXCL8/IL-8 monomer and dimer binding to glycosaminoglycans: Structural plasticity mediates differential binding interactions. Biochem J. 472:121–133. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Kridel SJ, Chan WW Knauer DJ. 1996. Requirement of lysine residues outside of the proposed pentasaccharide binding region for high affinity heparin binding and activation of human antithrombin III. J Biol Chem. 271:20935–20941. [DOI] [PubMed] [Google Scholar]
  19. Kridel SJ, Knauer DJ. 1997. Lysine residue 114 in human antithrombin III is required for heparin pentasaccharide-mediated activation. J Biol Chem. 272:7656–7660. [DOI] [PubMed] [Google Scholar]
  20. Li W, Johnson DJD, Esmon CT Huntington JA. 2004. Structure of the antithrombin–thrombin–heparin ternary complex reveals the antithrombotic mechanism of heparin. Nat Struct Mol Biol. 11:857. [DOI] [PubMed] [Google Scholar]
  21. MacKerell AD Jr, Bashford D, Bellott M, Dunbrack RL Jr., Evanseck JD, Field MJ, Fischer S, Gao J, Guo H Ha S, et al. 1998. All-atom empirical potential for molecular modeling and dynamics studies of proteins. J Phys Chem B. 102:3586. [DOI] [PubMed] [Google Scholar]
  22. Mallajosyula SS, Guvench O, Hatcher E MacKerell AD Jr. 2012. CHARMM additive all-atom force field for phosphate and sulfate linked to carbohydrates. J Chem Theory Comput. 8:759–776. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Meagher JL, Huntington JA, Fan B Gettins PGW. 1996. Role of arginine 132 and lysine 133 in heparin binding to and activation of antithrombin. J Biol Chem. 271:29353. [DOI] [PubMed] [Google Scholar]
  24. Mosier PD, Krishnasamy C, Kellogg GE Desai UR. 2012. On the specificity of heparin/heparan sulfate binding to proteins. Anion-binding sites on antithrombin and thrombin are fundamentally different. PLoS One. 7:e48632. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Mottarella SE, Beglov D, Beglova N, Nugent MA, Kozakov D Vajda S. 2014. Docking server for the identification of heparin binding sites on proteins. J Chem Inf Model. 54:2068–2078. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Mulloy B, Forster MJ. 2008. Application of drug discovery software to the identification of heparin-binding sites on protein surfaces: A computational survey of the 4-helix cytokines. Mol Simul. 34:481–489. [Google Scholar]
  27. Olson ST, Björk I, Sheffer R, Craig PA, Shore JD Choay J. 1992. Role of antithrombin-binding pentasaccharide in heparin acceleration of antithrombin–proteinase reactions. J Biol Chem. 267:12528. [PubMed] [Google Scholar]
  28. Olson ST, Frances-Chmura AM, Swanson R, Björk I Zettlmeissl G. 1997. Effect of individual carbohydrate chains of recombinant antithrombin on heparin affinity and on the generation of glycoforms differing in heparin affinity. Arch Biochem Biophys. 341:212–221. [DOI] [PubMed] [Google Scholar]
  29. Phillips JC, Braun R, Wang W, Gumbart J, Tajkhorshid E, Villa E, Chipot C, Skeel RD, Kalé L Schulten K. 2005. Scalable molecular dynamics with NAMD. J Comput Chem. 26:1781. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Raghuraman A, Mosier PD Desai UR. 2006. Finding a needle in a haystack: Development of a combinatorial virtual screening approach for identifying high specificity heparin/heparan sulfate sequence(s). J Med Chem. 49:3553. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Raghuraman A, Mosier PD Desai UR. 2010. Understanding dermatan sulfate–heparin co-factor ii interaction through virtual library screening. ACS Med Chem Lett. 1:281–285. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Raman EP, Guvench O MacKerell AD Jr. 2010. CHARMM additive all-atom force field for glycosidic linkages in carbohydrates involving furanoses. J Phys Chem B. 114:12981–12994. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Samsonov SA, Teyra J Pisabarro MT. 2011. Docking glycosaminoglycans to proteins: Analysis of solvent inclusion. J Comput-Aided Mol Des. 25:477–489. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Sankaranarayanan NV, Desai UR. 2014. Toward a robust combinatorial screening strategy for identifying glycosaminoglycan sequences that display high specificity for target proteins. Glycobiology. 24:1323–1333. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sapay N, Cabannes E, Petitou M Imberty A. 2011. Molecular modeling of the interaction between heparan sulfate and cellular growth factors: Bringing the pieces together. Glycobiology. 21:1181–1193. [DOI] [PubMed] [Google Scholar]
  36. Sarkar A, Desai UR. 2015. A simple method for discovering druggable, specific glycosaminoglycan–protein systems. Elucidation of key principles from heparin/heparan sulfate-binding proteins. PLoS One. 10:e0141127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Sasisekharan R, Shriver Z, Venkataraman G Narayanasami U. 2002. Roles of heparan-sulphate glycosaminoglycans in cancer. Nat Rev Cancer. 2:521–528. [DOI] [PubMed] [Google Scholar]
  38. Sasisekharan R, Venkataraman G. 2000. Heparin and heparan sulfate: Biosynthesis, structure and function. Curr Opin Chem Biol. 4:626–631. [DOI] [PubMed] [Google Scholar]
  39. Shirts MR, Mobley DL. 2013. An introduction to best practices in free energy calculations. Methods Mol Biol. 924:271–311. [DOI] [PubMed] [Google Scholar]
  40. Taylor KR, Gallo RL. 2006. Glycosaminoglycans and their proteoglycans: Host-associated molecular patterns for initiation and modulation of inflammation. FASEB J. 20:9–22. [DOI] [PubMed] [Google Scholar]
  41. Thompson LD, Pantoliano MW Springer BA. 1994. Energetic characterization of the basic fibroblast growth factor–heparin interaction: Identification of the heparin binding domain. Biochemistry. 33:3831. [DOI] [PubMed] [Google Scholar]
  42. Wang L, Wu Y, Deng Y, Kim B, Pierce L, Krilov G, Lupyan D, Robinson S, Dahlgren MK Greenwood J., et al. 2015. Accurate and reliable prediction of relative ligand binding potency in prospective drug discovery by way of a modern free-energy calculation protocol and force field. J Am Chem Soc. 137:2695–2703. [DOI] [PubMed] [Google Scholar]
  43. Xu D, Esko JD. 2014. Demystifying heparan sulfate–protein interactions. Annu Rev Biochem. 83:129–157. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Zacharias M, Straatsma TP McCammon JA. 1994. Separation-shifted scaling, a new scaling method for Lennard–Jones interactions in thermodynamic integration. J Chem Phys. 100:9025–9031. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Data
supp_26_10_1041__index.html (1.1KB, html)
supp_cww073_SarkarAetal_SupplementaryData_Glycobiology_20160701.pdf (290.8KB, pdf)

Articles from Glycobiology are provided here courtesy of Oxford University Press

RESOURCES