The biological properties of a protein depend on its conformation, and the protein conformation can be influenced by physical or chemical factors, for example ligand binding. Upon interacting with fatty acids, the HSA molecule undergoes dramatic conformational changes, according to X-ray crystallographic studies. The changes are described as a significant left-shifting of subdomains I and III relative to the central subdomain II, as well as rotations of side chains of amino acid residues [6]. Nevertheless, detailed mechanisms for the conformational changes, as well as the physiological implications, are still unclear, apparently because of the limited structural information obtained. It has been suggested that approaches other than X-ray crystallography and NMR are necessary in order to provide better structural elucidation pertinent to the conformational changes [14].

MS in conjunction with chemical cross-linking has emerged as a sensitive tool for probing the tertiary structure of proteins and protein–protein interactions [15–24]. Chemical cross-linkers introduce new covalent bonds into the 3-D (three-dimensional) structure of the protein [25]. The modified protein is digested and subjected to analysis by MS. Cross-linked peptides are identified by comparing the mass map of the digests from the cross-linked protein with that of the unmodified protein. Fragmentation by MS/MS (tandem MS) confirms the peptide sequence and localizes the cross-linked amino acid residues. Based on the cross-linked sites and the spacer arm lengths of the cross-linkers, low-resolution distance constraints can be obtained to assist the construction of the 3-D structure of a protein. This approach is particularly advantageous, since analysis time as well as the amount of protein sample required are significantly reduced in comparison with the traditional X-ray crystallography or NMR analysis. Moreover, there is no limitation to the molecular size of a protein of interest. Furthermore, like NMR, this method can be applied in solution, and therefore shows promise for monitoring the conformational changes of proteins due to protein–ligand or protein–protein interactions under native biological conditions.

The ε-amino groups of two lysine residues reacted with BS³, DSS and DSG, resulting in a net mass increase of 138, 138 and 96 Da respectively, whereas modification of a single lysine residue (end-capping) increased the mass by 156, 156 and 114 Da respectively [15,22]. The cross-linked peptides were identified by comparing the mass spectra of tryptic digests obtained from unmodified and cross-linker-modified HSA. Furthermore, tryptic digestion in H₂¹⁸O confirmed a through-space cross-linked pair between two peptide segments by a characteristic 8 Da mass shift, due to the complete incorporation of two ¹⁸O atoms for each C-teminus [21,22]. Cross-linking with BS³ generated eight new peaks, which are marked with asterisks (*) in Figure 2(b). For example, a cross-linked peptide with a mass value of 3552.8 Da, which was reconstructed from a quadruply charged ion at m/z 889.2 (bottom left of Figure 2), emerged in BS³-modified sample (Figure 2b). The mass of the peptide shifted to 3560.8 Da when digested in H₂¹⁸O (results not shown), indicating involvement of a through-space cross-linking. With non-cross-linked control HSA, the peak was absent (Figure 2a). The identity of the cross-linked peptide was revealed by MS/MS, as shown in Figure 3. The data clearly revealed that this peak resulted from through-space cross-linking between two peptide segments, Gln-390–Arg-410 and Ala-539–Lys-545, with Lys-402 located in subdomain IIIA cross-linked to Lys-541 from subdomain IIIB, via C₈H₁₀O₂ (derived from C₈H₁₂O₂, the cross-linking arm of BS³, with two hydrogens subtracted due to cross-linking). The lysines participating in the cross-linking reaction were not hydrolysed by trypsin. The N-terminal b ions and C-terminal y ions resulting from the amide bond cleavages are labelled in Figure 3, in accordance with the nomenclature suggested by Schilling et al. [30]. In addition to b and y ions, the neutral loss of NH₃ (−17 Da) from b ions, probably from side chains of Gln-390 or Asn-391, was also observed. Ions b₅ and b₆ included C₈H₁₀O₂ with a net m/z increase of 138. The accurate mass measurement showed that the mass of the peptide was 3552.798 Da, which represented an error within 8 p.p.m., further confirming the assignment of the cross-linking pair.

The spectra of tryptic digests obtained from modified HSA before and after fatty acid binding were different from each other, as shown in Figures 2(c)–2(e). In addition to the peaks observed in the BS³ control (Figure 2b), a new peak with a mass of 4505.2 Da, reconstructed from the quadruply charged ion at m/z 1127.3 and the quintuply charged ion at m/z 902.0 (bottom right of Figure 2), were detected upon binding of AA (Figure 2d) or DHA (Figure 2e) to HSA. MS/MS analysis indicated that this peak originated from the through-space cross-linking between Cys-200–Arg-209 and Met-446–Arg-472, with Lys-205 of subdomain IIA linked to Lys-466 of subdomain IIIA. The peak was not seen in the spectrum obtained from OA-bound HSA (Figure 2c). In addition, a considerable increase in the intensity of the cross-linking peptide between Lys-402 of IIIA and Lys-541 of IIIB with mass of 3552.8 Da, discussed above, was observed in all three fatty acid-bound HSA samples.

The changes in the peak intensities were further confirmed by the ¹⁸O labelling technique. The BS³-modified fatty acid-free HSA sample digested in H₂¹⁸O was mixed with the fatty acid-bound HSA sample digested in ¹⁶O water. The mixture was subjected to mass spectrometric analysis. For example, Figure 4(a) shows the reconstructed mass spectrum of the ¹⁸O-labelled fatty acid-free sample at a mass range of 3500–4600 Da. A cross-linked peptide peak with a mass of 3872.0 Da resulting from the cross-linking between Lys-402 and Lys-525 was identified. In addition, a weak peak with mass of 3560.8 Da due to the cross-linking of Lys-402 to Lys-541 was detected. Figure 4(b) shows the spectrum obtained from the mixture of DHA-bound sample digested in ¹⁶O water and the ¹⁸O-labelled fatty acid-free sample (mixed at a ratio of 1:2). The peak height ratio of 3864.0 Da (corresponding to the tryptic peptide cross-linked between Lys-402 and Lys-525) over 3872.0 Da was approx. 1:2. Even though the ¹⁸O-labelled fatty acid-free sample was twice concentrated, the peak with mass of 3560.8 Da was hardly detected in the mixture because of the dilution. However, the presence of the peak at 3552.8 Da was obvious. The peak with a mass of 4505.2 Da was also identified only in the mixture, whereas the corresponding ¹⁸O-labelled peak at 4513.2 Da was not present in the ¹⁸O-labelled control sample (Figure 4a).

Modification by DSS generated similar results. A total of 12 cross-linked pairs were identified in the DSS-modified control. Eight of them were identical to the pairs seen with BS³. Although DSS generated more cross-linked pairs, no further changes in cross-linked pairs were observed due to fatty acid binding other than the ones observed with BS³.

DSG was used to further refine the distance constraints for the cross-linked lysine pairs. Although three of the cross-linked pairs found in DSS-modified samples were not detected in the DSG-modified samples, the rest of the cross-linked peptides were consistent with those obtained from the DSS-modified counterparts. Figure 5 shows the confirmation of the cross-linked pair of Lys-205/Lys-466 induced upon AA or DHA binding, by LC/MS/MS, which allowed more reliable identification of the cross-linked peptides due to the separation of tryptic peptides prior to the mass spectrometric detection. The ion chromatograms for the peptide pair Lys-205/Lys-466 cross-linked with DSG (Figure 5, left) or DSS (Figure 5, right) are shown using quintuply charged ions of m/z 893.6 or 902.0 for a molecular mass of 4463 or 4505 Da respectively. The chromatographic peak eluting at approx. 43.5 min was observed in the digests obtained from AA- or DHA-bound HSA (Figures 5c and and5d),5d), whereas it was not present in the digests obtained from OA-bound (Figure 5b) or control HSA (Figure 5a). The MS/MS spectrum obtained from the ion at m/z 902.0 was virtually identical to that from the ion at m/z 893.6. The fragments detected in the spectrum (results not shown) revealed that this peptide resulted from the through-space cross-linking between the peptide segments Met-446–Arg-472 and Cys-200–Arg-209, with Lys-205 (subdomain IIA) linked to Lys-466 (subdomain IIIA), via C₅H₄O₂, derived from the cross-linking arm of DSG. The DSS-modified peptide showed a mass of 4505 Da (as seen from BS³-modified samples in Figure 2), an increase of 42 Da compared with its DSG-modified counterpart, because DSS and BS³ contain longer spacer arms (C₈H₁₂O₂) than DSG (C₅H₆O₂).

The hydrophobic environment may prevent nearby lysine residues from reaching the hydrophilic BS³ for cross-linking. On the other hand, the hydrophobic DSS can access lysine residues in hydrophobic surroundings. However, all surface lysine residues that are exposed to solvents should be available to both DSS and BS³. This view is consistent with the cross-linking results shown in Table 1. For example, Lys-73 and Lys-240 are surrounded by a hydrophobic environment [22], explaining the absence of Lys-64/Lys-73 and Lys-212/Lys-240 pairs when BS³ was used as the cross-linker. Eight cross-linked lysine pairs observed in both DSS- and BS³-modified control (fatty acid-free) HSA suggested that these lysine residues are exposed to the surface. The cross-linking results suggested that the distances between all cross-linked lysine pairs in fatty acid-free HSA were within 11.4 Å (between ε-amino groups) or 24 Å (between α-carbons, considering the maximum distance derived from the two side-chain lengths plus the spacer arm length of BS³ or DSS). Using DSG, nine cross-linked lysine pairs were identified; further refining 7.7 Å (between ε-amino groups) or 20 Å (between α-carbons) distance constraints for these lysine pairs. The results were consistent with the distances between α-carbons of lysine pairs determined by X-ray crystallography in most cases (Table 1). However, our results concerning the distance constraints of the ε-amino groups were not consistent with X-ray crystallographic data.

The cited maximum span lengths of the cross-linkers are based on a fully extended conformation of both the cross-linking arm of a cross-linker and the reactive side chains of two lysine residues in a perfect co-operative orientation [31]. In addition to the lysine–lysine distance from each α-carbon, the cross-linking would depend critically on the orientation of the lysine side chains in the most energetically favourable conformation in solution. Our data showed that three cross-linked peptides, including Lys-212/Lys-351, Lys-414/Lys-541 and Lys-276/Lys-286, were observed with DSS but not with DSG, even though the crystal structure indicated that the distances between their α-carbons were within the maximum distance allowed by DSG (20 Å). These data suggested that the orientation of the lysine side chains was such that the distance between the two ε-amino groups in the cross-linked lysine pair is within 11.4 Å but beyond 7.7 Å in solution. However, the distances between the ε-amino groups of these cross-linked pairs were determined to be 16.4, 16.4 and 14.9 Å respectively by X-crystallography (PDB 1ao6) (Table 1). The distance constraints for most ε-amino groups deduced from our data also differed from those obtained from X-ray crystallographic studies. Since the side chains are much more flexible than the bonds between α-carbons, it is possible that the cross-linkers could instantaneously trap the protein in a short-lived, less populated conformational state. Consequently, the most energetically favoured conformation of side chains may not always be represented. Alternatively, the discrepancy in the spatial distances may reflect a different conformational state of HSA in solution in comparison with the crystal structure.

The 11 cross-linked peptides derived from fatty acid-free HSA were cross-linked between lysine pairs within an individual domain. Included were intra-subdomain cross-linking pairs Lys-212/Lys-240 in subdomain IIA, Lys-276/Lys-286 in IIA, Lys-351/Lys-359 in IIB, Lys-64/Lys-73 in IA, Lys-313/Lys-317 in IIB, Lys-541/Lys-545 in IIIB and Lys-436/Lys-439 in IIIA, and the inter-subdomain cross-linking of Lys-402 of IIIA to Lys-525 of IIIB, Lys-212 of IIA to Lys-351 of IIB, and Lys-414 of IIIA to Lys-541 of IIIB. These cross-linkings were also observed when OA, AA or DHA was bound to HSA, with intensities similar to those observed for the fatty acid-free form, suggesting that no major distortion occurs in general within an individual domain upon fatty acid binding. In this regard, the cross-linking data appear to be consistent with the results obtained from X-ray crystallographic studies [14].

The inter-domain cross-linked peptide involving Lys-205 of subdomain IIA and Lys-466 of subdomain IIIA was identified only upon binding of AA or DHA to HSA. The fact that the cross-linking was observed using DSG indicated that the distance between the reactive amino groups in the side chains of the lysine pair was within 7.7 Å. In the fatty acid-free HSA molecule, the cross-linking was not detected even with DSS or BS³, despite the single modification (end-capping) for both Lys-205 and Lys-466, implying that the side chains of the lysine pair were separated by more than 11.4 Å, which is consistent with the distance determined by X-ray crystallography. It was therefore concluded that the side chains of Lys-205 and Lys-466 moved towards each other upon AA or DHA binding due to a conformational change of HSA after binding to the long-chain polyunsaturated fatty acids, as depicted in Figure 6. In the crystal structures of HSA complexed with OA or AA (PDB 1gni and 1gnj respectively), the distances between the α-carbons of Lys-205 and Lys-466 are 16.6 and 17.7 Å respectively, whereas the distances between the ε-amino groups of the lysine residues are not available due to the lack of the ε-amino group in Lys-466 in the crystal structure. Our data indicating that the cross-linking between Lys-205 and Lys-466 was not detected in OA-bound HSA molecules suggested a distinct difference in the method of binding of monounsaturated fatty acid to HSA in comparison with polyunsaturated fatty acids.

NMR studies have shown that polyunsaturated fatty acids exhibit binding behaviour quite distinct from that of saturated and monounsaturated fatty acids when interacting with HSA [10,32]. Similarly, in an investigation using X-ray crystallography, Peptipsa et al. [8] reported that AA and OA showed different patterns of binding to HSA. It was suggested that, at least at two sites, the presence of four double bonds in AA can cause a notable difference in the bound conformations of the fatty acid methylene tail. One site is located in the interface between subdomains IA and IIA, and the other is located in subdomain IIIA [6]. This conformational change induced by the multiple double bonds in AA or DHA is consistent with the relative movement of Lys-205 located at the beginning of IIA closer to Lys-466 in IIIA suggested by our data.

The observation of the significant enhancement of cross-linking between Lys-402 (IIIA) and Lys-541 (IIIB) upon HSA binding to all three fatty acids suggested that a local conformational change occurred in domain III. It is likely that the side chains of the lysine residues rotated upon accommodating incoming fatty acids, resulting in a conformation in which the side chains of the lysines pointed in the same direction, allowing the cross-linking to occur (Figure 6).