The critical role of dimer formation in monosaccharides binding to human serum albumin

Prapasiri Pongprayoon *abc and Toshifumi Mori *de
aDepartment of Chemistry, Faculty of Science, Kasetsart University, Chatuchak, Bangkok, 10900, Thailand. E-mail:; Fax: +66-2579-3955; Tel: +66-2562-5555
bCenter for Advanced Studies in Nanotechnology for Chemical, Food and Agricultural Industries, KU Institute for Advanced Studies, Kasetsart University, Bangkok 10900, Thailand
cComputational Biomodelling Laboratory for Agricultural Science and Technology (CBLAST), Kasetsart University, Bangkok 10900, Thailand
dInstitute for Molecular Science, Myodaiji, Okazaki, Aichi 444-8585, Japan. E-mail:; Fax: +81-564-53-4660; Tel: +81-564-55-7305
eSchool of Physical Sciences, The Graduate University for Advanced Studies, Okazaki, Aichi 444-8585, Japan

Received 15th September 2017 , Accepted 30th October 2017

First published on 30th October 2017

Human serum albumin (HSA) is the most abundant transport protein found in human blood. HSA is known to bind a wide range of drugs and monosaccharides, but where and how these molecules bind are largely unknown. Recently, a crystal structure of glycated HSA has been reported, and interestingly, in that structure two glucose molecules have been located in pyranose (GLC) and open chain (GLO) forms bound in the same binding pocket (Sudlow site I). Molecular simulations also proposed two binding modes of GLC and GLO (binding two ligands either in a distant location or in close contact). Yet, how HSA binds sugars in general is poorly understood. To this end, here we study the mechanism of binding glucose and its epimer galactose to HSA using alchemical free energy perturbation calculations and molecular dynamics simulations, and show why two sugar molecules appear in the bound state. We find that HSA does prefer glucose over galactose, in line with experiments, by binding glucose deeper in the pocket. Furthermore, out of the two possible binding modes suggested previously, the binding becomes tighter when the two sugars are in contact; this is achieved by a hydrogen bond connecting the two sugars and filling the large cavity of Sudlow site I as a dimer. We also find tight hydrogen bonds between open chain glucose/galactose and HSA, which includes the possible glycation site K199, while the pyranose form does not interact strongly with any characteristic residues. Thus the current result highlights the importance of dimeric structures of glucose/galactose for binding to HSA and triggering glycation/galactation.


Human serum albumin (HSA) is the most abundant transport protein in human blood. HSA carries metabolites and a wide range of drugs (e.g. ibuprofen, indomethacin, and warfarin).1–3 Changing pathological and physiological conditions can cause changes in protein conformation and ligand binding efficiency, resulting in functional impairment of HSA.4,5 Glycation is one of the important impairments found in many proteins including HSA and amyloid β.6,7 The glycation products can lead to diseases such as cardiovascular disease, Alzheimer's disease, and renal dysfunction.8 The level of glycated HSA (GHSA) concentration can also be used to indicate glycemic states; 1–10% of HSA becomes non-enzymatically glycated in healthy persons whereas this percentage is dramatically increased in diabetic patients.9,10 As a result, several studies were devoted to explore the feasibility of GHSA as a diabetes biomarker.11–14

HSA consists of 585 amino acids and forms a monomeric heart-like structure (Fig. 1A). It has 3 domains (I, II, and III), and each domain is divided into 2 subdomains (A and B). HSA can bind drugs at Sudlow sites I (warfarin–azapropazone binding site) and II (indole–benzodiazepine binding site) located inside domains IIA and IIIA, respectively. In the presence of glucose, a free amino group of lysine in HSA reacts with an aldehyde group in glucose to form an Amadori product.9,15,16 Earlier studies with mass spectrometry suggested many glycation sites on HSA (K12, K51, K93, K195, K199, K205, K233, K276, K281, K286, K378, K414, K439, K525, K538, K545).17–19 K195 and K199 were also found computationally to interact with glucose.5,20 In addition to glucose, galactose and fructose can also bind to HSA.21–23 In particular, glucose and galactose are hexose monosaccharides where galactose is a C4 epimer of glucose (Fig. 1C) and share several glycation sites and a similar glycation mechanism in HSA.24,25

image file: c7cp06324e-f1.tif
Fig. 1 (A) Cartoon representation of human serum albumin (HSA). Colors describe different subdomains. (B) HSA with two bound sugar molecules; initial glucose (λ = 0) is colored based on atom types and final galactose (λ = 1) is superimposed in blue. Two possible sugar orientations (Sim1 and Sim2) inside Sudlow site I are shown in red boxes. (C) Chemical structures of glucose and galactose in pyranose (GLC and GAC) and open chain (GLO and GAO) conformations, respectively. Atoms which are different between glucose and galactose are highlighted in red.

Recently, the first structure of HSA with bound glucose was crystalized.26 Interestingly, two isomers of glucose molecules (closed-chain (GLC) and open-chain (GLO) forms in Fig. 1B) were found to sit in the same Sudlow site I, and glycation of K195 by GLO was suggested.26 Other studies have also shown that the sugar binds to Sudlow site II.23,27,28 In our previous molecular dynamics (MD) simulations of pre-glycated HSA,5 we found two possible binding orientations of the two glucose molecules (GLC and GLO): bound either in a distant location or in close contact (Sim1 and Sim2 in Fig. 1B). An increase in protein flexibility at Sudlow site II was also observed. These structures indicate that GLO is more mobile inside the binding pocket, yet molecular insight into the binding affinity and specificity remains largely unknown.

In contrast to the glycation of HSA, studies on galactation have been limited to exploring possible galactation sites,8,22,25,29,30 and no galactose-bound HSA structure has been reported to date. Thus the microscopic understanding of galactose inside HSA is lacking. To this end, here we employ free energy perturbation (FEP) calculations and MD simulations to study the binding affinity and molecular structures of glucose and galactose inside HSA. Starting from the glucose–HSA complex proposed in our recent work,5 we explore the two possible binding modes in more detail and investigate the energetic and structural properties of the sugar–HSA complex. By comparing the glucose– and galactose–HSA complexes in molecular detail, this work allows us to better understand the diverse characteristics of HSA and the mechanism of sugar binding to HSA from a broader perspective.

Materials and methods

Preparation of sugar–protein complexes

The initial glucose–HSA complexes are prepared using the method reported in ref. 5. In ref. 5, Awang et al. found two possible binding structures for two glucose isomers. In both structures, two glucose molecules, in pyranose (GLC) and open chain (GLO) forms, were located in Sudlow site I (Fig. 1B), which is also the binding site observed in the X-ray structure.26 The isomers are located either in a distant position or in close proximity in the two structures. In the current work, the complex structures of the two binding modes are used as initial structures for the glucose-bound system. The two modes having GLC/GAC and GLO/GAO in a distant location and in close proximity are hereafter denoted as SimGLU/GAL1 and SimGLU/GAL2, respectively (Fig. 1B). For the galactose–HSA complex, the initial orientations of the two galactose molecules (pyranose (GAC) and open chain (GAO) galactose) follow those of the glucose systems (SimGLU1 and SimGLU2), and only the hydroxyl groups on the C4 atom (on both GAO and GAC) are modified via alchemical transformation in FEP. We thus have 4 systems in total (2 glucose (SimGLU1 and SimGLU2) and 2 galactose (SimGAL1 and SimGAL2)). Note that each system contains 2 bound sugar molecules; GLO and GAO stand for open chain glucose and galactose, while GLC and GAC represent pyranose glucose and galactose, respectively. Suffixes 1 and 2 describe that the sugars are from Sim1 and Sim2, respectively. The Amber99SB-ILDN force field31 is used for the protein, and the charges of the sugar molecules are determined with the AM1-BCC method using AmberTools1432 (Fig. S1, ESI). Each complex is placed in a cubic box of 9 × 10 × 10 nm and solvated by 26504 TIP3P water molecules; counterions are then added to neutralize each system. An energy minimization of 1000 steps is performed to remove bad contacts using the steepest descent algorithm, and a 10 ns equilibration under the constant pressure and temperature (NPT) condition with restraints on the heavy atoms is performed after heating the system to 310 K. To perform the FEP calculations, the ligands in solution are also prepared following the same protocol using 599 TIP3P waters. Long-range electrostatic interactions are treated using the particle mesh Ewald (PME) method33 with a short-range cutoff of 1 nm. The time step is set to 2 fs.

Our study is divided into 2 parts. First, alchemical FEP calculations are conducted to evaluate the relative binding free energies between the two sugars and to obtain a galactose-bound HSA complex; 200 ns MD simulations on the sugar–protein complexes are then performed to explore the interactions that are critical for binding. The details of FEP and MD simulations are described below.

Free energy perturbation (FEP)

The theoretical background of FEP can be found in a review paper.34 Here, we perform alchemical transformations of two glucose molecules (GLC and GLO) into galactose molecules (GAC and GAO). The relative binding free energies (ΔΔG) between the two bound glucose (GLU) and galactose (GAL) molecules are calculated following the thermodynamic cycle displayed in Fig. 2. The relative binding free energy is equal to ΔGbind(GLU) − ΔGbind(GAL) where the subscript “bind” denotes the protein–ligand complex, and “GLU” and “GAL” denote that two glucose and galactose molecules are bound, respectively.
image file: c7cp06324e-f2.tif
Fig. 2 Thermodynamic cycle for current ligand binding. P:GLU and P:GAL denote HSA–glucose (GLU) and HSA–galactose (GAL) complexes, respectively, and P + GLU, P + GAL states are unbound GLU and unbound GAL. Vertical and horizontal arrows represent ligand binding and alchemical transformations of two ligands (2GLU → 2GAL), respectively. Subscripts “prot” and “solv” represent the alchemical transformation in HSA and in solution. The difference in binding free energy is thus given by ΔΔG = ΔGprot – ΔGsolv = ΔGbind(GLU) − ΔGbind(GAL).

The free energy change for this FEP is calculated using the Bennett acceptance ratio method:35

image file: c7cp06324e-t1.tif
where λ is the coupling parameter that varies from initial GLU (λ0) to final GAL (λ1) states. image file: c7cp06324e-t2.tif represents the free energy change associated with the transition from state λi to λi+1. The n equidistant λ-states including the physical end points (λ0 = 0 and λ1 = 1) are used with an increment of 0.1.

The FEP calculations from GLU to GAL are conducted in three steps: (i) removing the charges of the hydrogen and hydroxyl groups on the C4 atom of GLC and GLO, (ii) changing the van der Waals (vdW) potentials as well as atom types, and (iii) recovering the charges of the hydrogen and hydroxyl groups on the C4 atom of GAC and GAO. Note that in step (ii) the whole molecule is transformed from GLU to GAL, i.e. charges of the atoms other than the hydrogen and hydroxyl groups on the C4 atom in the molecules are perturbed. Soft-core potential algorithm is used to change the vdW potentials. The NPT MD simulations at 310 K of the ligand and ligand–protein complexes are conducted for 10 ns, where the first 3 ns in each window is considered as equilibration and discarded in the analysis.

Molecular dynamics simulations

Simulations of the glucose–protein and galactose–protein complexes are carried out using the GROMACS 5 package.36 All simulations are performed in the NPT ensemble at 310 K. The pressure is coupled using the Berendsen algorithm at 1 bar with a coupling constant of τp = 1 ps. 200 ns production runs are performed for each system. Molecular graphic images are prepared using VMD.37

Results and discussion

Protein–ligand binding affinities

First, the binding affinities of the 2 monosaccharides are studied from FEP calculations. The free energies for the alchemical transformations from glucoses to galactoses for the two binding modes (Sim1 and Sim2) are summarized in Table 1. dG/dλ and its standard deviation at each λi are displayed in Fig. S2 (ESI) to show convergence of each step. The relative binding free energies from the two binding modes (Sim1 and Sim2) have similar values of 6.72 and 5.36 kcal mol−1, respectively (Table 1), indicating that HSA binds more strongly to glucose than to galactose. Note that the free energy differences between sugars in solution (Table 1) are not identical due to differences in the initial sugar distances (separated or in contact), but the result is only marginally affected (∼0.5 kcal mol−1). In contrast, differences in binding modes are found to have larger effects in the complexes. The electrostatic and vdW energies contribute to ΔΔG differently in the two binding modes, implying the flexibility and complexity of the protein–sugar interactions.
Table 1 Free energy changes (in kcal mol−1) for the alchemical transformations from GLC and GLO to GAC and GAO. Steps 1, 2, and 3 are the perturbation steps for removing charges (−Q), changing vdW potentials and atom types (vdW), and recovering charges (+Q), respectively
Step In solution Bound to HSA ΔΔG (kcal mol−1)
Sim1 1(−Q) −0.44 ± 0.01 0.80 ± 0.02
2(vdW) 0.29 ± 0.01 3.40 ± 0.02
3(+Q) −2.17 ± 0.02 0.19 ± 0.02
Total −2.33 ± 0.02 4.39 ± 0.03 6.72 ± 0.04
Sim2 1(−Q) −1.95 ± 0.01 1.34 ± 0.02
2(vdW) 1.20 ± 0.01 −2.00 ± 0.02
3(+Q) −2.09 ± 0.02 3.18 ± 0.01
Total −2.84 ± 0.02 2.52 ± 0.03 5.36 ± 0.04

Protein–ligand interactions and dynamics

To explore the dynamics of sugars and HSA in more detail, 200 ns-long MD simulations of the four sugar–HSA complexes are performed (SimGLU1, SimGLU2, SimGAL1, and SimGAL2). First, the structural drift and fluctuation of HSA are analyzed from the root mean-square deviations (RMSDs) and fluctuations (RMSFs) (Fig. 3). RMSDs are calculated using the coordinates at t = 0 ns as the reference in each system. Fig. 3A shows that the overall fluctuation of HSA appears to be similar in all cases. Among the subdomains, the sugar binding subdomains, IIA and IIIA, are less mobile compared to other subdomains (Fig. 3B, E and G). In contrast, subdomains IA and IIIB, located at the top of the heart-like structure, fluctuate quite largely, as large as 0.4 nm in RMSD (Fig. 3C and H).
image file: c7cp06324e-f3.tif
Fig. 3 RMSDs of HSA using Cα atoms for (A) the whole protein and (C–H) subdomains. RMSDs are computed using the initial structure at 0 ns as the reference. (B) RMSFs of the residues over the 200 ns trajectories. Black, red, green, and blue lines represent SimGLU1, SimGLU2, SimGAL1 and SimGAL2, respectively. The cartoon view of each domain is shown as an inset in (E).

Principal component analysis (PCA) further reveals the coupled motion distributed mainly about subdomains IA and IIIB. In particular, the first principal component (PC) mode (PC1), which dominates the motion of HSA (Fig. S3, ESI), shows a scissor-like mode mainly appearing at subdomain IIIB and at the end of subdomain IA in all four simulations (Fig. 4). The second PC mode also involves large motions about subdomains IA and IIIB, but is more system dependent compared to PC1 (not shown).

image file: c7cp06324e-f4.tif
Fig. 4 Cartoon view of the time-dependent motion of HSA calculated from the first principal component mode for (A) SimGLU1, (B) SimGLU2, (C) SimGAL1, and (D) SimGAL2. Principal component analysis is performed using only Cα atoms. The colors are in the RWB format where the displacements for t = 0 to t = 200 ns change from red to blue.

Next we follow the time trace of the sugar molecules inside the pocket. As shown in Fig. 5, similar starting sugar orientations appear to behave similarly in the 200 ns simulations; i.e. in SimGLU1 and SimGAL1, both open chain glucose and galactose (GLO1 and GAO1) located near the gate of the pocket escape the binding site (green molecules in Fig. 5A and C). GAO1 seems to have weaker protein contacts than GLO1 and is released out to the bulk quickly; yet interestingly, we find that GAO1 rebinds to the gate of the pocket (Fig. 5C and Fig. S4, ESI). The quick escape of GAO1 may facilitate protein flexibility about the binding pocket, and indeed unbinding of GAC1 is found. In contrast, GLC1 remains inside the pocket even after GLO1 has escaped.

image file: c7cp06324e-f5.tif
Fig. 5 Cartoon representations of the glucose-bound (A and B) and galactose-bound (C and D) HSA. The time traces of sugar molecules are colored as a function of time in the licorice format. The colors change from cyan → iceblue → blue for GLC/GAC and yellow → orange → red for GLO/GAO every 70 ns (see color bar on left). Two boxes on the right of each protein system represent the top views of key amino acids around the glucose and galactose molecules inside the pocket. Subdomains are also colored in (D) for clarification.

image file: c7cp06324e-f6.tif
Fig. 6 Protein–sugar hydrogen bonds as a function of time for (A) SimGLU1, (B) SimGLU2, (C) SimGAL1, and (D) SimGAL2. Left and right figures show the plots for pyranose and open chain molecules, respectively. Each black band represents the existence of a hydrogen bond at a given time. Hydrogen bonds are counted as 1 when the hydrogen–donor–acceptor angle is below 30° and the donor–acceptor distance is shorter than 0.35 nm and 0 otherwise. The inset in the top-right figure is to remind the initial position of the sugar molecules.

In SimGLU2 and SimGAL2, each close-packing pair of sugars is buried inside subdomain IIA throughout the simulations. In comparison to SimGLU1 and SimGAL1, the close proximity of 2 sugars may be essential for strong sugar–HSA binding. Such close contacts of 2 isomers in SimGLU2 and SimGAL2 are stabilized by strong hydrogen bonds between the sugars (see below) and are indeed found in the crystal structure.26

Inside a pocket, all sugars are solvent-accessible and can interact with various amino acids. This is due to a large cavity in Sudlow site I as observed previously.5Fig. 6 summarizes the hydrogen bonds formed between the sugar molecules and HSA during the 200 ns-long trajectories. While pyranoses (GLC and GAC) are found to be more successful in binding (3 out of 4), surprisingly no permanent hydrogen bonds between pyranoses and HSA are found. For instance, GLC1 preferentially interacts with E153 and E285 while GLC2 prefers Y150 and S287. Note that after ∼80 ns, GLC2 starts to form a hydrogen bond with I290 (but loses contact with Y150 and S287) due to slight relocation of GLC. GAC1 leaves the pocket, and GAC2 interacts with R218 and R222, which are residues located near the gate of the pocket. Overall, GLC appears to bind more frequently to HSA than GAC, and GAC is more exposed to the solvent (Fig. S5, ESI). This may be another reason why GAC1 escaped the binding pocket. In addition, pyranoses interact very rarely with the lysine residues K195 and K199, which have been suggested as glycation sites.38 This supports the idea that pyranose may not be a suitable form for glycation.

In contrast, when the open chains are bound (GLO2 and GAO2), more frequent and stronger hydrogen bonds with HSA are formed. Common hydrogen bond partners, e.g. K199 (possible glycation site), H242, and Q196/E153, are found to interact strongly with both GLO2 and GAO2. Note that Q196 and E153 are located in close proximity and seem to compensate for each other (Fig. S6, ESI). In addition, hydrogen bonds between the sugars, i.e. GLC2–GLO2 and GAC2–GAO2, are found quite frequently when GLO2 and GAO2 are bound. This suggests the importance of the dimeric structure of the sugar molecules to be able to bind to the large cavity of Sudlow site I.

To analyze these dimeric structures and their interactions with the surrounding residues in more detail, Fig. 7 summarizes the probabilities of finding the hydrogen bonds between each pair of atoms in pyranose and open chain glucose/galactose as well as between the atoms in glucose/galactose and HSA. The figure shows that the difference in the location of the O4 atom has a large impact on the dimeric structure of glucose and galactose; GLC2 and GLO2 interact through O6 atoms as well as through O1 atoms of GLC2, whereas GAC2 and GAO2 are connected at O1 and O2 atoms. The O4 atoms in GLC2 and GLO2 point outwards and are found to interact with I290, S287, K199, and Q196, whereas only O4 of GAO2 interacts with E153 and K199, possibly due to the difference in their orientations.

image file: c7cp06324e-f7.tif
Fig. 7 Percentage of hydrogen bonds formed between the atoms in pyranose and open chain forms of (A) glucose and (B) galactose during the near-contact simulations (Sim2). The typical structures found in the trajectories of glucose and galactose are given in (C and D), respectively. The probabilities of forming hydrogen bonds between the oxygen atoms in glucose/galactose and the surrounding residues are shown for (E) GLC2, (F) GAC2, (G) GLO2, and (H) GAO2.

We also note that during unbinding of the sugars, no common residue appears to play a role in translocating the sugar from the pocket to the bulk. We do, however, find that in SimGAL1, GAO1 returns to the entrance of Sudlow site I (residues P339, Y341, V343, and P451) after interacting with D107, N109, and K466, which are located at the front of subdomain IB. How sugar binds to HSA is another interesting problem, and the current result may shed light on the possible binding pathways; yet this is beyond the scope of the current study.

Overall, our results clearly highlight that sugar–sugar contacts are essential for the sugar molecules to bind efficiently to Sudlow site I of HSA. The sugar dimers are found to bind tightly inside the pocket, but pyranoses (GLC2 and GAC2) seem to interact rather randomly with polar and charged residues. This implies that pyranose is not a preferable substrate for HSA. In contrast, long chain glucose and galactose (GLO2 and GAO2) stay in a similar location with common hydrogen bond partners and, importantly, are in direct contact with a glycation site K199. Thus the open chain form is expected to be the target for glycation in HSA. Although pyranose does not appear to be a favorable form for glycation, we emphasize that the binding of pyranose is still important. In particular, the cavity in Sudlow site I is too large to fit either open chain glucose or galactose alone efficiently. The insertion of a pyranose ring induces the dimeric structure to help fill the space and to enforce the orientation of the open chain sugar to be suitable for glycation/galactation.


In this work we have investigated the non-covalent binding of glucose and galactose to HSA from energetic and dynamic perspectives. Based on the two possible glucose-bound HSA complexes proposed previously, alchemical free energy perturbations have been performed to calculate the relative binding free energies between glucose and galactose as well as to obtain the galactose-bound structures. Glucose was found to bind more strongly to HSA than galactose, but the sugar–HSA interactions were dependent on initial structures, indicating non-specific interactions between the sugar and HSA.

Four 200 ns-long MD simulations have been conducted to understand the sugar–HSA interactions in detail. The results show that when pyranose and open chain sugars are bound in a distant location, the open chain sugars easily unbind from the binding pocket within 100 ns. Pyranose molecules also become flexible, and indeed galactose (GAC1) has been found to unbind. In contrast, when the pyranose and open chain sugars are placed nearby, both glucose and galactose sugars are found to bind tightly to HSA. Open chain sugars hydrogen bond with similar residues, e.g. K199 (glycation site), H242, and Q196/E153. In contrast, pyranose sugars do not interact with common residues, thus do not appear to bind strongly to HSA. Nevertheless, strong hydrogen bonds between the pyranose and open chain sugars have been found, and the two sugars in close proximity fill the large cavity of Sudlow site I as a dimer to enable tighter sugar binding. Furthermore, this closely packed conformation allows the open chain glucose/galactose to point towards K199, which may promote glycation/galactation at this site. This result is in line with those of earlier studies which indicated that open chain sugar is required for glycation.5,26 We also note that fructose is also found to bind to HSA as a dimer and in close proximity,26 which further supports the idea that monosaccharides bind tightly to HSA as a dimer.


HASHuman serum albumin
GLOOpen-chain glucose
GLCClosed-chain glucose
GAOOpen-chain galactose
GACClosed-chain galactose
FEPFree energy perturbation

Conflicts of interest

There are no conflicts to declare.


We are grateful to Shinji Saito for stimulating discussions. This work was financially supported by the grants from the Kasetsart University Research and Development Institute (KURDI) (S-K1.56 and S-K161.59) and the Thailand Research Fund (TRG5880230) to P. P., and Grant-in-Aid for Young Scientists (B) (JP15K17813) and Scientific Research on Innovative Areas (JP16H00856) from JSPS to T. M. The calculations are partly carried out at the Research Center for Computational Sciences in Okazaki.


  1. M. Fasano, S. Curry, E. Terreno, M. Galliano, G. Fanali, P. Narciso, S. Notari and P. Ascenzi, IUBMB Life, 2005, 57, 787–796 CrossRef CAS PubMed .
  2. C. E. Petersen, C. E. Ha, M. Mandel and N. V. Bhagavan, Biochem. Biophys. Res. Commun., 1995, 214, 1121–1129 CrossRef CAS PubMed .
  3. I. Petitpas, A. A. Bhattacharya, S. Twine, M. East and S. Curry, J. Biol. Chem., 2001, 276, 22804–22809 CrossRef CAS PubMed .
  4. P. Pongprayoon and M. P. Gleeson, J. Mol. Graphics Modell., 2014, 54, 164–173 CrossRef CAS PubMed .
  5. T. Awang, N. Wiriyatanakorn, P. Saparpakorn, D. Japrung and P. Pongprayoon, J. Biomol. Struct. Dyn., 2016, 2017(35), 781–790 Search PubMed .
  6. S. Menon and N. Sengupta, ACS Omega, 2017, 2, 2134–2147 CrossRef CAS .
  7. A. K. Jana, K. B. Batkulwar, M. J. Kulkarni and N. Sengupta, Phys. Chem. Chem. Phys., 2016, 18, 31446–31458 RSC .
  8. A. Szkudlarek, A. Sulkowska, M. Maciazek-Jurczyk, M. Chudzik and J. Rownicka-Zubik, Spectrochim. Acta, Part A, 2016, 152, 645–653 CrossRef CAS PubMed .
  9. N. Shaklai, R. L. Garlick and H. F. Bunn, J. Biol. Chem., 1984, 259, 3812–3817 CAS .
  10. E. Bourdon, N. Loreau and D. Blache, FASEB J., 1999, 13, 233–244 CAS .
  11. A. Mohamadi-Nejad, A. A. Moosavi-Movahedi, G. H. Hakimelahi and N. Sheibani, Int. J. Biochem. Cell Biol., 2002, 34, 1115–1124 CrossRef CAS PubMed .
  12. H. V. Roohk and A. R. Zaidi, J. Diabetes Sci. Technol., 2008, 2, 1114–1121 CrossRef PubMed .
  13. A. Raghav and J. Ahmad, Diabetes, Metab. Syndr. Obes.: Targets Ther., 2014, 8, 245–251 CrossRef PubMed .
  14. H. Cao, T. Chen and Y. Shi, Curr. Med. Chem., 2015, 22, 4–13 CrossRef CAS PubMed .
  15. J. Rohovec, T. Maschmeyer, S. Aime and J. A. Peters, Chemistry, 2003, 9, 2193–2199 CrossRef CAS PubMed .
  16. P. J. Thornalley, A. Langborg and H. S. Minhas, Biochem. J., 1999, 344(Pt 1), 109–116 CrossRef CAS PubMed .
  17. A. Lapolla, D. Fedele, R. Reitano, N. C. Arico, R. Seraglia, P. Traldi, E. Marotta and R. Tonani, J. Am. Soc. Mass Spectrom., 2004, 15, 496–509 CrossRef CAS PubMed .
  18. C. Wa, R. L. Cerny, W. A. Clarke and D. S. Hage, Clinica Chimica Acta, 2007, 385, 48–60 CrossRef CAS PubMed .
  19. O. S. Barnaby, R. L. Cerny, W. Clarke and D. S. Hage, Clinica Chimica Acta, 2011, 412, 277–285 CrossRef CAS PubMed .
  20. R. Nasiri, H. Bahrami, M. Zahedi, A. A. Moosavi-Movahedi and N. Sattarahmady, J. Biomol. Struct. Dyn., 2010, 28, 211–226 CAS .
  21. G. Suarez, R. Rajaram, A. L. Oronsky and M. A. Gawinowicz, J. Biol. Chem., 1989, 264, 3674–3679 CAS .
  22. I. Syrovy, J. Biochem. Biophys. Methods, 1994, 28, 115–121 CrossRef CAS PubMed .
  23. A. Szkudlarek, M. Maciazek-Jurczyk, M. Chudzik, J. Rownicka-Zubik and A. Sulkowska, Spectrochim. Acta, Part A, 2016, 153, 560–565 CrossRef CAS PubMed .
  24. L. Frost, M. Chaudhry, T. Bell and M. Cohenford, Anal. Biochem., 2011, 410, 248–256 CrossRef CAS PubMed .
  25. J. C. Urbanowski, M. A. Cohenford and J. A. Dain, J. Biol. Chem., 1982, 257, 111–115 CAS .
  26. Y. Wang, H. Yu, X. Shi, Z. Luo, D. Lin and M. Huang, J. Biol. Chem., 2013, 288, 15980–15987 CrossRef CAS PubMed .
  27. A. Szkudlarek, A. Sulkowska, M. Maciazek-Jurczyk, M. Chudzik and J. Rownicka-Zubik, Spectrochim. Acta, Part A, 2016, 152, 645–653 CrossRef CAS PubMed .
  28. J. Baraka-Vidot, C. Planesse, O. Meilhac, V. Militello, J. van den Elsen, E. Bourdon and P. Rondeau, Biochemistry, 2015, 54, 3051–3062 CrossRef CAS PubMed .
  29. D. A. Armbruster, Clin. Chem., 1987, 33, 2153–2163 CAS .
  30. J. Aring, J. Schlepper-Schaefer, V. Burkart and H. Kolb, Biochim. Biophys. Acta, 1989, 1010, 140–144 CrossRef CAS .
  31. K. Lindorff-Larsen, S. Piana, K. Palmo, P. Maragakis, J. L. Klepeis, R. O. Dror and D. E. Shaw, Proteins, 2010, 78, 1950–1958 CAS .
  32. D. A. Case, T. E. Cheatham, T. Darden, H. Gohlke, R. Luo, K. M. Merz, A. Onufriev, C. Simmerling, B. Wang and R. J. Woods, J. Comput. Chem., 2005, 26, 1668–1688 CrossRef CAS PubMed .
  33. T. Darden, D. York and L. Pedersen, J. Chem. Phys., 1993, 98, 10089–10092 CrossRef CAS .
  34. P. Kollman, Chem. Rev., 1993, 93, 2395–2417 CrossRef CAS .
  35. C. H. Bennett, J. Comput. Phys., 1976, 22, 245–268 CrossRef .
  36. E. Lindahl, B. Hess and D. van der Spoel, J. Mol. Model., 2001, 7, 306–317 CrossRef CAS .
  37. W. Humphrey, A. Dalke and K. Schulten, J. Mol. Graphics, 1996, 14, 33–38 CrossRef CAS PubMed .
  38. Y. Wang, H. Yu, X. Shi, Z. Luo, D. Lin and M. Huang, J. Biol. Chem., 2013, 288, 15980–15987 CrossRef CAS PubMed .


Electronic supplementary information (ESI) available. See DOI: 10.1039/c7cp06324e

This journal is © the Owner Societies 2018