NMR ANALYSIS OF SYNTHETIC HUMAN SERUM ALBUMIN ALPHA HELIX 28 IDENTIFIES STRUCTURAL DISTORTION UPON AMADORI MODIFICATION

The non-enzymatic reaction between reducing sugars and long lived proteins in vivo results in the formation of glycation and advanced glycation end products (AGEs) that alter the properties of proteins, including charge, helicity, and their tendency to aggregate. Such protein modifications are various pathologies the general aging Alzheimer’s and the long-term of diabetes. it has been that glycation and AGEs protein structure and helicity, structural data and information currently exists on whether or not glycation does indeed influence change local protein secondary structure. We this model peptide human serum We in the presence of 50 mM glucose and at 37 o C, one of the lysine the di-lysine preferentially

The non-enzymatic reaction between reducing sugars and long lived proteins in vivo results in the formation of glycation and advanced glycation end products (AGEs) that alter the properties of proteins, including charge, helicity, and their tendency to aggregate. Such protein modifications are linked with various pathologies associated with the general aging process, Alzheimer's and the long-term complications of diabetes. Although it has been suggested that glycation and AGEs altered protein structure and helicity, little structural data and information currently exists on whether or not glycation does indeed influence or change local protein secondary structure. We have addressed this problem using a model helical peptide system containing a di-lysine motif derived from human serum albumin. We have shown that in the presence of 50 mM glucose and at 37 o C, one of the lysine residues in the di-lysine motif within this peptide is preferentially glycated. Using NMR analysis we have confirmed that the synthetic peptide constituting this helix does indeed form a α α α αhelix in solution in the presence of 30% trifluoroethanol. Glycation of the model peptide resulted in the distortion of the α α α α-helix, forcing the region of the helix around the site of glycation to adopt a 3 10 helical structure. This is the first reported evidence that glycation can influence or change local protein secondary structure. The implications and biological significance of such structural changes on protein function are discussed.
Non-enzymatic glycation describes the initial products arising from the formation of Maillard reaction adducts due to the reaction between primary amino groups on a protein surface and reducing sugars such as glucose and fructose (1)(2)(3)(4). These non-enzymatic reactions are initiated with the reversible formation of a Schiff's base adduct that undergoes rearrangement to form a more stable Amadori product (for glucose) (5) or Heyns product (for fructose) (6). The Amadori compound may then undergo a series of poorly understood rearrangements and reactions to yield protein adducts collectively termed advanced glycation end products (AGEs) (4). AGEs are 'naturally' formed in vivo on a variety of proteins and are implicated in the pathologies associated with aging, atherosclerosis, Alzheimer's, and long-term diabetic complications (7). More recently glycation has been shown at the N-terminus of the pathogenic prion protein in Transmissible Spongiform Encephalopathies, a group of transmissible neurodegenerative diseases that are characterized by the accumulation of abnormally folded prion protein (8) and has been implicated in food allergies (9).
The association of AGEs with a variety of pathologies has resulted in much scientific interest in the role played by glycation products and AGEs in the pathology of these disease states. For example, in recent years the receptor for advanced glycation end products (RAGE) has been described, and it has since been reported to be a member of the immunoglobulin superfamily of cell surface proteins (10). The expression of RAGE has also been implicated as a developmental factor in several pathologic conditions including chronic inflammation, cancer and Alzheimer's disease (10). Furthermore, a number of AGE products and cross-links have now been described (11), however despite much attention, the exact role that these modified protein forms play in the associated disease states remains to be fully elucidated.
The process of glycation and the formation of AGEs is known to promote protein aggregation and insolubilization (12). Furthermore, protein glycation and the formation of subsequent AGE products are thought to be involved in structural and functional changes in vivo in proteins during aging and the long-term complications of diabetes (12,13). Although glycation has been shown to inactivate a number of enzymes (12), little, if any, information currently exists about the effects these modifications have on protein secondary or tertiary structure. This is largely due to the problem of obtaining sufficient homogeneous material for structural studies, as glycation usually occurs at one or more residues on a protein structure and gives rise to multiple AGEs.
Notwithstanding these apparent problems, Blakytny et al used mass spectrometry and NMR to study the effect of glycation with galactose on the C-terminal extension of α-crystallin, investigating both the intact protein and a synthetic C-terminal peptide (12). Although these studies categorically identified the sites and level of glycation, both the intact protein and synthetic peptide exhibited great conformational freedom and adopted no preferred structure in solution and therefore no conclusions could be drawn about the effect of glycation on the secondary and tertiary structure of α-crystallin (12). NMR has also been used recently to detect the presence of glycated protein in the saliva of patients with diabetes (14) while others have used it to study the effect of glycation on a tetra-peptide and its N-terminal amide bond stereochemistry and tautomeric distribution (15).
Despite these studies, little information currently exists on whether or not glycation does indeed influence or change local protein secondary structure. A previous study on human serum albumin suggested that glycation does indeed result in conformation change (16), although the nature of this change was undetermined. Others have reported that glycation and AGE product formation alter the helicity of proteins (11). In the present study we have addressed this problem using a model peptide system derived from human serum albumin. Human serum albumen (HSA, ExPASY Primary Accession Number P02768) is a helical protein that is the main protein constituent of plasma. Although the exact function of HSA is unknown, its primary function is thought to be to regulate the colloidal osmotic pressure of blood (17). The primary site of glycation in human albumin, both in vitro and in vivo, is known to be Lys 525 (16). However, several other lysine residues in albumin are known to be glycated both in vitro and in vivo, including Lys 548 as shown in a study of bovine albumin (18). Further, the lysine residue Lys 548 resides in the same helix and in the same peptide motif in both human and bovine serum albumin. Lys 548 is of particular interest as it lies in a di-lysine motif (Lys 548 Lys 549 ) that is part of a helix on the surface of the molecule (helix 28), which is held in place relative to the rest of the molecule via a salt bridge. Further, glycation is thought to be accelerated at di-lysine motifs due to local acid-base catalysis.
The model peptide used in this study therefore corresponds to helix 28 of HSA, corresponding to residue numbers 543-557 in accordance with Swiss-Prot numbering. We have shown that in the presence of glucose, one of the lysine residues in the di-lysine motif (Lys 548 Lys 549 ) within this peptide is preferentially glycated. As in the intact protein, the lysine equating to Lys 548 was preferentially glycated in the model peptide system. Further, using NMR analysis we have confirmed that the synthetic peptide constituting this helix (543-557) does indeed form a α-helix in solution in the presence of 30% trifluoroethanol (TFE). Glycation of the model peptide resulted in the distortion of the α-helix, forcing the region of the helix around the site of glycation to adopt a 3 10 helical structure. This is the first reported evidence that glycation can indeed influence or change local protein secondary structure. The implications for such changes on protein function are discussed.
All materials were of analytical reagent grade or better and purchased from Sigma Aldrich unless otherwise stated.

Synthesis and Purification of Peptide HSA543-557
The peptide RERQIKKQTALVELV was synthesized with an acetylated N-terminal amino group using a Shimadzu PSSM-8 Multiple Peptide Synthesizer and an fmoc/HBTu synthesis strategy. The resulting peptide was then purified by reverse phase HPLC using a preparative C 18 10 x 250 mm column linked to a Hewlett Packard 1100 series HPLC. The sample to be analyzed was injected onto the column and salts washed out with an isocratic gradient of 2% acetonitrile containing 0.05% TFA. The peptide was then eluted from the column using a linear gradient from 2 to 40% acetonitrile (containing 0.045% TFA) over 50 minutes. The peptide peak was collected and then freeze-dried over night. Multiple runs were combined in order to purify all of the synthesized peptide. The authenticity of the purified peptide was then confirmed by mass spectrometry. Mass spectra were recorded in the positive ion mode using the extended mass range (m/z 250-4000) on a Finnigan MAT LCQ ion-trap mass spectrometer.

Generation and Isolation of Glycated Peptide HSA543-557
In order to generate glycated peptide, the peptide was dissolved in sterile PBS containing 50 mM glucose at a concentration of 5 mg/mL. The resulting solution was then incubated at 37 o C for 1 week. Prior to further experiments, salts and sugar components were then removed from the peptide solution by reverse phase HPLC as described above. Glycated peptide was then isolated from non-glycated peptide for NMR and mass spectrometric analysis using commercially available Glyco Gel columns (1 mL bed volume) from Pierce (Rockford, IL, USA). The incubated peptide sample from which the salt/sugar components had been previously removed by reverse phase chromatography were then redissolved in 0.05 M HEPES (pH 8.5) and loaded onto the columns pre-equilibrated in the equilibration/wash buffer provided by Pierce. The column was then washed (5 column volumes) and the glycated peptide eluted (in 3 column volumes) using the procedure previously described by Zhao et al (19). The glycated fraction was then desalted by reverse phase chromatography as described above and freeze-dried. Control, non-glycated peptide for NMR analysis was generated by heating the synthesized and purified peptide at 5 mg/mL for 1 week at 37 o C in PBS (i.e. in the absence of glucose), desalted by reverse phase HPLC and freeze dried. Both the glycated and non-glycated HSA543-557 peptides were then subjected to electrospray mass spectrometry (as described above) and NMR analysis.

NMR Sample Preparation
All NMR samples were prepared to a final volume of 300 µL for use in a Shigemi BMS005V NMR tube by dissolving purified, freeze-dried peptide to provide a final concentration of 2 mM peptide in phosphate buffered saline (PBS) at pH 6.4 (phosphate concentration 25 mM and saline concentration of 100 mM). To this system, trifluoroethanol-d 3 (TFE) was added as a helix stabilizer to provide a final concentration of 30% (v/v) in the system. TFE was used following AGADIR (20,21) analysis of the peptide that identified the helical propensity of the standard peptide as being 0.21. Peptides with this level of propensity will only highlight regions of helical tendency with the addition of TFE (20). TFE 30% (v/v) was considered the lowest proportion of TFE that could be used to give rise to a helical structure. This was confirmed from the NMR NOE contacts observed for the standard helix. In order to make a meaningful comparison, all Amadori modified peptide data was obtained under identical conditions to those used for the standard peptide.

NMR Spectroscopy
All experiments were recorded at 10°C on a Varian UnityINOVA 600 MHz NMR spectrometer with a z-shielded gradient triple resonance probe using standard procedures. For each peptide sample, a two-dimensional nuclear Overhauser effect spectroscopy (NOESY) and total correlation spectroscopy (TOCSY) experiment were recorded with mixing times of 250 and 68.4 ms respectively. These experiments were collected with 512 and 1024 complex points with acquisition times of 64 and 128 ms in the indirectly and directly acquired 1 H dimensions respectively. The maximum theoretical NOESY enhancement was estimated to be between 70-90% of the theoretical maximum whereas our projected ROESY enhancements were expected to be 60%. Therefore, NOESY experiments were chosen in this analysis and validated by obtaining build-up curves that also confirmed 250 ms as the most appropriate mixing time. In addition, a twodimensional double-quantum-filtered correlated spectroscopy (DQFCOSY) experiment was collected for each peptide, with 1024 and 2048 complex points with acquisition times of 128 and 256 ms in the indirectly and directly acquired 1 H dimensions respectively. Amides in slow exchange and deemed capable of being hydrogen bond donors were identified from a NOESY experiment that was collected with 256 and 1024 complex points with acquisition times of 32 and 128 ms in the indirectly and directly acquired 1 H dimensions respectively, obtained from a peptide sample resuspended in 2 H 2 O. Data processing and analysis were carried out on Sun Blade 100 and Transtec X2100 Linux workstations using NMRPipe (22) to process and NMRView (23) to view processed data.

Structural Calculations and Analysis
All structural calculations were obtained using the Crystalography and NMR System (CNS) version 1.1 (25) running on Silicon Graphics O2+ and Transtec X2100 Linux workstations. CNS parameter files were modified to incorporate the topology of the modified lysine residue for calculations. All NOE contacts were classified into one wide classification between 1.8-5.0Å with final structures calculated from extended coordinates using the standard CNS NMR anneal protocol with sum averaging for dynamic annealing with NOE's from both extended and folded precursors (24). A final structural ensemble of 40 structures for each peptide was produced with all structures used to produce statistical energy and root mean square (r.m.s.) deviation structural information. Backbone and heavy atom r.m.s. deviation values were obtained using MOLMOL version 2k.2 (25) on a PC running Microsoft Windows 2000. The structural integrity of each ensemble was evaluated using PROCHECK-NMR (26) run on a Transtec X2100 Linux workstation. Energy comparisons between structures created from the NMR constraint data and the equivalent human serum albumen helix were made using GROMOS96 43Bl parameter set (27) within DEEPVIEW version 3.7 (28).

Peptide Synthesis, Glycation and Purification
The peptide sequence corresponding to HSA αhelix 28 (residues 543-557) was successfully synthesized on a Shimadzu PSSM-8 Multiple Peptide Synthesizer as determined by mass spectrometry (Table I). The peptide sequence was identical to that found in the native HSA molecule except that the N-terminal lysine residue was replaced with an arginine residue to prevent glycation on the lysine amino side chain of this residue (Lys 543 ). The peptide was synthesized with the N-terminal NH 2 amino group blocked with an acetyl group to prevent potential glycation at this site in the synthetic peptide. Following incubation with glucose and purification of the glycated peptide from non-glycated peptide using a Glyco Gelcolumn, mass spectrometry analysis confirmed that the peptide was glycated with one glucose residue per peptide molecule (i.e. was not di-glycated), as indicated by an increase in the [M+H] + ion by 162 Da (Table I). From this data it was not possible to confirm if glycation had occurred entirely at one lysine residue, preferentially at one of the two-lysine residues in the di-lysine motif or if there was an equal distribution between the two. This was resolved during the NMR analysis (see below). We note that it has previously been shown that the sugar moieties in glycated human serum albumin occur as an equilibrium of the beta-pyranose (59%), alpha-furanose (19%) and beta-furanose (24%) anomers (29). Phenylboronate purification of Amadori products selectively binds beta-furanose sugar anomers, however as the confirmations are in rapid equilibrium the yield of glycated peptide is likely to be close to quantitative.

NMR Resonance Assignments
Spin systems were identified by analysis of twodimensional DQFCOSY and TOCSY NMR spectra and all the observed 1 H chemical shifts are listed in Table II. Assignments for the majority of 1 H spin systems were possible for both the standard and Amadori modified peptide with the exception of amino acid Ile 547 in the standard peptide. The H α shift of the N-terminal amino acid Arg 543 was not observed in either standard or modified peptide. The assignment of the modified Amadori lysine was possible following the observation of a duplicated set of resonances in the H N region of the TOCSY spectrum for the modified lysine side chain H β , H γ , H δ and H ε . The first set of resonances are correlated from the back bone H N of the lysine but chemical modification transforms the side chain lysine NH 3 + group to a H N group (H ζ ) and provides a second H N correlation and duplication of the lysine side chain 1 H resonances. Resonances of the Amadori peptide were achieved from both the DQFCOSY and TOCSY data together with H η protons identified by the NOESY through-space correlation to H ζ .

Structural Assignments and Additional Restraints
Through-space assignments were achieved using two-dimensional NOESY spectra of both standard and Amadori modified peptide (see Figure 1). Amides in slow exchange and deemed capable of being hydrogen bond donors were identified from a NOESY experiment obtained from a peptide sample re-suspended in 2 H 2 O. Additional φ restraints were obtained from application of the Karplus relationship to 3 J HNHα that were obtained from high-resolution DQFCOSY spectra. 3 J HNHα values less than 5 Hz were used to constrain φ for that residue to -60° ± 30°. A cut off value of 5 Hz was used to allow for the fact that 3 J HNHα values obtained by DQF-COSY are always larger than those obtained by more accurate heteronuclear NMR methods (30).
With the exception of Ile 547 in the standard peptide where an NOE distribution was not observed, the NOE distribution was uniform across all residues in both peptides. A total of 23 NOE's were found between the modified Lys 548 and residues Arg 545 , Gln 546 and Ile 547 that were crucial in defining the structural changes upon modification. A summary of the number of contact types and additional restraints are shown in Table III with the distribution of restraints across each peptide shown in Figure 2(a) and 2(b). NOE contact types observed in Figure 2(a) for the standard peptide support a standard helix conformation with NOE's observed between H α :i and H N :i+3 as well as H α :i and H β :i+3. Additionally, slow H N exchange and 3J HNHα values less that 5 Hz were observed for the majority of residues within the standard peptide, with the exception of Ile 547 , as the resonances from this residue were not observed in our system. The resonances from Ile 547 were difficult to assign due to placement of the H α resonance directly over the water resonance. This was confirmed by observing and assigning Ile 547 resonances in the absence of TFE (data not shown). Figure 2(b) highlights the NOE contact types for the Amadori modified peptide and defines a region from Lys 549 to Glu 555 that shows similar helix NOE contacts to that observed throughout the standard peptide. This is also confirmed from the observed H N exchange and 3J HNHα values obtained for the modified peptide. The N-terminal region of the Amadori peptide gave rise to reduced H α :i and H N :i+3 but increased H α :i and H β :i+3 together with H α :i and HN:i+2 and i+3 contacts. In this region there was a single retarded H N exchange and no 3J HNHα values less than 5 Hz.

Structure Calculations and Analysis
All structural data was determined using CNS as described in the experimental procedures. No calculated structure gave violations greater than 0.2Å or bond angle violations greater than 5° from the restraint data when all 40 structures were used to compute the ensemble average structural set. R.m.s. deviation statistics over all backbone atoms for residues 544-556 for the 40 structure ensembles for both the standard and Amadori modified peptide were found to be 0.37Å and 0.48Å respectively. The backbone ensembles for both peptides are shown in Figure 3(a) and Figure  3(b) respectively, together with the ensemble average structures and key residues shown as ribbon schematics in Figures 3(c)-3(f). Corresponding r.m.s. deviations over all heavy atoms for each 40 structure ensemble of the standard and Amadori modified peptides were found to be 1.29Å and 1.45Å respectively. PROCHECK-NMR analysis of the Ramachandaran plot for the 40-structure ensemble of the standard peptide identified that 100% of all residues fell in either the most favored or additionally allowed regions of the α-helix. Equivalent analysis of the Amadori modified peptide identified 84.6% of all residues falling in the most favored or additionally allowed regions, with the remaining 15.4% falling into the generously allowed regions of the α-helix. Ensemble average energies for each peptide are shown in Table IV together with the equivalent energies created from helix 28 extracted directly from the HSA X-ray structure.

Discussion
Incubation of the model HSA peptide in glucose resulted in the glycation of Lys 548 , as determined by NMR spectroscopy (see below), in preference to Lys 549 . Although Lys-Lys sequences are known to be more reactive toward protein glycation due to local acid-base catalysis, it is also thought that preferential protein glycation at one lysine residue in a di-lysine motif is, at least partially, due to the relative accessibilities of the two lysine side chains in question (2,31). However, the model peptide used in this study exhibited no measurable secondary structure in aqueous PBS solution in the absence of TFE and therefore stabilized structural aspects would not appear to account for the exclusive glycation of Lys 548 in preference to Lys 549 . Preferential glycation at Lys 548 could be due to local acid-base catalysis and the immediate neighboring amino acids resulting in the amino group of Lys 548 being more nucelophilic than that of Lys 549 , however it is clear that long-range structural affects must be involved since the adjacent I 547 is unlikely to affect the pKa of K 548 . We suggest it is possible that, while the mean structure is random in aqueous PBS solution, the chemical modification of K 548 is catalyzed when the random coil approximates the helical structure in the protein. This may occur when the side chain of R 545 approaches that of K 548 , and the charge on R 545 suppresses the charge on (lowers the pKa of) K 548 . However, it is also likely that E 544 may be in a position to catalyze the Amadori rearrangement on K 548 . In this respect, it is interesting that the position of E 544 is significantly altered in the glycated peptide relative to the non-glycated (standard) peptide (see Figure 4).
The H α chemical shift differences between the standard and modified peptide were found to be moderately small, except for residues Glu 544 and Lys 549 as indicated in Figure 4. This suggests that similar local conformation exists across the entirety of both peptides in accordance with known observations of H α chemical shift in the prediction of secondary structure elements (32). Upon initial inspection, one may be lead to believe that Lys 549 is the modified peptide on the basis of the H α shift difference alone, due to the large difference in H α chemical shift. We know from both NOESY and chemical shift assignments, as outlined in the results section, that Lys 548 is modified in the glycated peptide. This is also confirmed by analysis of the H ε chemical shift values for Lys 548 and Lys 549 in both the standard and modified peptide. The 1 H chemical shift difference between the standard and modified peptide for Lys 548 H ε is 0.272 ppm compared to 0.058 ppm observed for Lys 549 . As the H ε is in close proximity to the modification, it stands to reason that the greatest shift change will be observed by the H ε of the lysine that is modified; in this case, Lys 548 .
The standard peptide in 30% (v/v) TFE folds to form an α-helical peptide structure ( Figure 3). All NOE contact data, hydrogen exchange and 3J HNHα data supports this observation and the ensemble of 40 structures provides a model whereby 100% of modeled amino acids in the ensemble fall inside the allowed regions for an α-helix in a Ramachandaran plot. Figures 3(c) and 3(e) show a ribbon form of the ensemble average structure that adopts a helix with key residues Arg 545 and Lys 548 on one face of the helix and Gln 546 rotated on the top face of the peptide. The ribbon view highlights an exaggerated tightening of the helix at both Nand C-termini that is most likely due to the CNS modeling. As both terminal regions have no additional constraints from hydrogen bonding to assist in the calculation and modeling of the peptide helical structure, these regions will tend to tighten slightly as observed within the calculation. Narrower helical φ angle constraints could have been used at the termini to reduce this effect, but it was considered important not to over constrain any aspect of the data and risk providing structural effects that could not be explained.
NMR data of the Amadori modified peptide in 30% (v/v) TFE defines a fold that forms a more complicated structure compared to the standard peptide. Figure 2(b) and Figures 3(b, d and f) show the C-terminal region of the peptide from Gln 550 /Thr 551 towards the C-terminus forming a standard helix that does not differ greatly from the standard peptide helix in this region. This is illustrated by the fact that the backbone r.m.s. by guest on March 23, 2020 http://www.jbc.org/ Downloaded from deviation for residues Gln 550 and Leu 556 for an 80structure ensemble (40 standard structures and 40 Amadori modified structures) is only 0.25Å. As with the standard peptide, the NOE contact data, hydrogen exchange and 3J HNHα data supports the conclusion that the region from Gln 550 -Leu 556 is helical. Figure 5 shows the detailed arrangement of Arg 545 , Gln 546 , Ile 547 and the modified lysine, Lys *548 in the Amadori modified peptide in the ensemble average model that is supported by 23 NOE interactions between residues 545-547 and the modified Lys *548 . Ile 547 H N and H α have contacts to Lys *548 H N , H α and H β . Gln 546 has NOE contacts from its side chain atoms H β and H γ to Lys *548 H β , H γ , H δ and H ε as well as the modification protons H ζ and H η . Arg 545 has NOE contacts from its side chain atoms H β , H γ and H δ to protons H ι , H κ , H λ and H µ in the Amadori structure that can be seen clearly associated as a contact core in Figure 5. This overall appearance is observed in all 40structures of the ensemble with the modified carbohydrate chain on the lysine parallel to the side chain of Arg 545 . This is confirmed in a quantitative manner by calculation of an r.m.s. deviation of 0.36Å for all the side chain carbon atoms from both Arg 545 and Lys *548 over the 40structure ensemble of the modified peptide. It is likely that the interaction is stabilized by hydrogen bonding between protons on the Arg 545 H ε groups and hydroxyl oxygen atoms attached to C µ of the modified Lys *548 . This arrangement would also be stabilized by an electrostatic attraction between Arg 545 side chain NH 2 ε groups and electronegative oxygen atoms attached to both C µ and C λ atoms in the modification. This association and contact between the modified, glycated chain on Lys *548 and Arg 545 twists the C-terminal region of the helix that is observed in the standard peptide. Interestingly, there still appears to be some helical character to this region as seen from the presence of H α :i and H β :i+3 NOE contacts that in this region appear more prevalent than H N :i and H α :i+3 contacts. This suggests that this region adopts a conformation more in line with a 3 10 helix as H α :i and H β and H N :i+2/:i+3 contacts that are stronger in these systems are observed (33).
Despite the model's suggestion that the Amadori modification places structural strain on the peptide, PROCHECK analysis for the ensemble of 40 structures showed that 84.6% of the modeled amino acids resided inside the favored-allowed regions for α-helix in a Ramachandaran plot. It would appear that there is strain in φ/ψ for Gln 546 and Ile 547 and that this is created by the interaction of the Amadori Lys *548 side chain with Arg 545 because the 15.4% that fell inside generously allowed regions consisted entirely of φ/ψ distributions from Gln 546 and Ile 547 . These two residues (Gln 546 and Ile 547 ), together with Arg 545 , provided virtually all of interactions with the modified Lys 548 Amadori side chain.
Analysis of the pdb coordinated for HSA shows that in the intact protein structure, two disulphide bonds exist between Cys 538 :Cys 583 and Cys 582 :Cys 591 that hold helices 28, 29 and 30 together. This stabilization is further supported by an electrostatic attraction between the side chains of Arg 545 and Glu 580 . However, unlike much of the HSA structure, helices 28-30 at the C-terminus are not attached to the bulk of the protein by a disulphide bond (only to each other), but are held relative to the bulk structure via a salt bridge created between the side chains of Lys 543 of helix 28 and Asp 207 in helix 10. This salt-bridge arrangement is depicted in Figure 6. Close analysis of the modified structures shown in Figures 3(e) and 3(f) confirmed an anticlockwise 90° twist in the orientation of the N-terminus of the peptide upon modification that would remove the correct orientation of Lys 543 with respect to its salt bridge with Asp 207 in the intact protein. If disruption of this salt bridge was to occur in the intact protein, this would result in helices 28-30 becoming more flexible and mobile, loosen the protein structure, exposing hydrophobic residues and increasing the susceptibility of the three helices to hydrolysis and being cleaved from the structure. The reorientation of residues 543-548 would also remove the electrostatic interaction between Glu 580 and Arg 545 and destabilize the turn regions of helices 28-30. Previous NMR investigations on proteins with disulphide bonds have shown degrees of conformational flexibility around these bonds (34,35) and any loss of nearby electrostatic stabilization will increase the conformational flexibility in the region around the disulphide bond, in this case around Cys 538 and Cys 583 . This would further destabilize the structure in this region of HSA. Further, We have clearly shown that glycation of a model peptide system can result in disruption of local secondary structure and alter helicity as suggested by Sell and Monnier (11). To our knowledge, this is the first example that definitively shows that glycation can influence and change secondary structural elements. Although we have not determined this change in helical structure in the intact HSA protein molecule upon glycation of Lys 548 , we predict that the few interactions holding this part of the molecule together are likely to be disrupted by such a modification. This prediction disagrees with a previous study by Coussons et al who concluded that glycation had minimal effects on the folded structure of HSA (36) however, this observation was based on far-UV circular dichroism (CD) measurements which will only detect gross changes in secondary structure. The disruptions described and predicted here are unlikely to have been detected by far-UV CD. Further, helix 28 is in the region of drug-binding site 2 in HSA and previous investigations have suggested that modification with methylglyoxal of an arginine residue in close proximity to I 547 modifies the ligand binding and enzymatic activity of HSA domain 3A (37). We therefore suggest that glycation of K 548 is also likely to change the ligand binding and enzymatic activity of this domain of HSA.
A similar disruption of secondary structure in other proteins could potentially affect protein function by, for example, changing enzymic activity, binding affinities or exposing hydrophobic patches leading to protein aggregation. In this way protein glycation and the formation of an Amadori product may be involved in accelerating protein aggregation even before the formation of AGE crosslinks, which accumulate much more slowly. As such the formation of an Amadori product alone could be envisaged as playing a role in pathologies such as those seen in transmissible neurodegenerative diseases characterized by the accumulation of an abnormally folded prion protein whereby posttranslational glycation is known to occur (8).        Table II. NMR assignment list of observed 1 H chemical shifts for the standard and Amadori modified peptide in PBS/30%(v/v) TFE at 10°C. All chemical shifts are referenced externally to a 100µM solution of dimethylsilapetane sulphonic acid (DSS) in PBS/30%(v/v) TFE.