Exposure to diisocyanates in the workplace is one of the leading causes of occupational asthma. It is hypothesized that isocyanate acts as a hapten by reacting with protein carriers via nucleophilic attack; however, the ultimate form of these protein-isocyanate conjugates that functions as allergens in vivo
is, as yet, unknown. The diverse functional groups present in proteins (amines, amides, thiols, alcohols, carboxylic acids) present a large number of potential reaction sites for the diisocyanate (dNCO). However, previous studies have suggested that under physiological conditions, these are limited to N-terminal α
-amines, the sulfhydryl group of cysteine, the hydroxyl groups of serine and tyrosine, the ε
-amine of lysine, and the secondary amine of the imidazole ring of histidine [76
]. Understanding the products formed by reaction of allergenic dNCOs such as methylene diphenyldiisocyanate (MDI) and toluene diisocyanate (TDI) with biological molecules is critical to understanding the mechanisms by which these chemicals affect living systems. Tandem mass spectrometry performed on a quadrupole time-of-flight (qTOF) mass spectrometer [77
] is particularly well suited for the characterization of chemically modified proteins. Proteins of interest may be digested with a proteolytic enzyme (such as trypsin) and the resultant peptides analyzed with high sensitivity and mass accuracy. Because covalent modification of an amino acid residue results in a change in that residue's mass, accurate mass determination of the fragment ions produced by collision-induced dissociation [78
] allows unambiguous assignment of the site of modification. Such experiments have become routine for the analysis of posttranslational modifications such as acetylation, glycosylation, and phosphorylation, among many others [79
Recent efforts in our laboratory [81
] and others [80
] have begun to focus on harnessing the power of tandem mass spectrometry to determine how and where dNCOs modify model peptides and proteins. Hydrolysis of the isocyanate functional group to a primary amine is a competing reaction under aqueous conditions. These hydrolyzed isocyanate amines may then undergo nucleophilic addition to another dNCO molecule. Therefore, conjugation products observed upon reaction of dNCOs with model peptides and proteins in vitro
results in a complex variety of different reaction products, including intra- and intermolecular crosslinking, dNCO self-polymerization, and dNCO hydrolysis. Our initial study focused on determining the site of dNCO modification on model bioactive peptides conjugated under aqueous conditions [81
]. Analysis of these conjugates by tandem mass spectrometry revealed that the dNCO was bound preferentially to the N-terminal amine of each of the peptides examined. Furthermore, when a peptide with an N-terminal residue containing a side chain amine (lysine, arginine) was reacted with dNCO, intramolecular crosslinking with the side chain amine becomes competitive with hydrolysis, however, the reactivity decreases as the residue is displaced further from the N-terminus. The results of this peptide study suggested agreement with a long-held hypothesis that the N-terminal amine of protein chains is a likely target for isocyanate conjugation [76
]. Studies of the kinetics of isocyanate binding with protein functional groups [83
] determined that at pH 7, reaction with an N-terminal amine should proceed approximately 100 times faster than the ε
-amine of the lysine side chain. The difference is due to the relative pKa
of the two functional groups (α
~ 9 versus Lys ε
~10.5), as dNCO conjugation proceeds through the neutral –NH2
rather than the charged –NH3+
Although tandem MS studies of dNCO-conjugated peptides are useful for determining the diverse chemical species produced when dNCO binds amino acids, they do not produce the unique chemical microenvironments presented by the complex three-dimensional structure of proteins. In order to understand how dNCOs react in these complex environments, studies on model proteins are essential. Serum albumin is an appropriate model protein, as it is monomeric and its sequence and three dimensional structure have been well defined. It is naturally abundant (35–50
mg/mL in serum), found in most tissues, and has been identified as a target of dNCO binding in vivo
]. Recently, our laboratory completed an extensive analysis of the binding sites of TDI on human serum albumin [90
]. At high (40
protein) ratios, near-stoichiometric binding was observed; TDI binds at thirty-seven sites on the protein, including the N-terminal amine on aspartic acid at position one and the side chain of thirty-four lysine residues. At lower conjugation ratios (1
protein), a small subset of these thirty-seven sites is conserved, with binding observed at the N-terminus and four lysine residues, suggesting these sites are preferred binding sites. Kristiansson and coworkers [91
] determined that at a tenfold molar excess, HHPA bound to thirty-seven sites on human serum albumin, including the N-terminal aspartic acid and thirty-six lysine residues.
Interestingly, all 59 lysine residues of human serum albumin have been determined to be solvent accessible, but only 37 are reactive toward TDI, while 19 are reactive toward MDI (Hettick and Siegel, unpublished data). Human serum albumin is a highly charged molecule, in part accounting for its high solubility. For example, provides two views of human serum albumin based on its crystal structure [92
]. In , the lysine residues that bind MDI and TDI are highlighted, whereas in , all 59 lysine residues are highlighted. Steric effects are insufficient to explain the difference between accessibility and observed binding. It is therefore likely that the microenvironment of the binding site(s) determines whether or not a certain lysine residue is reactive toward isocyanate. Lysine 199, which is known to bind hydrophobic anions such as aspirin and benzyl penicillin, was also determined to be a predominant binding site for both TDI and MDI. Gerig and Reinheimer [93
] determined that the pKa
of the aspirin binding site (later determined to be Lys199) of albumin was 7.9. These authors hypothesized based on the reactivity of human serum albumin with dinitrofluorobenzene that there exist two lysines on HSA that have a pKa
as low as 7.9. In addition, Lys199 has been shown by molecular dynamics calculations to be predominantly uncharged, undergoing proton transfer with the nearby Lys195 [94
]. This study elegantly suggests the reason we observed dNCO bound to Lys199, and not the nearby the Lys195. Other sites noted to be abundant binders of dNCO, such as Lys439 and Lys525, have been observed to undergo nonenzymatic glycosylation in vivo
. Glycosylation is generally observed to occur at lysine residues located near another amino group, presumed to be charged [95
]. Lys439 is located in a region with two other nearby lysine residues (Lys432 and Lys436) and Lys525 is part of a dilysine motif.
(a) Lysine residues observed bound to MDI and TDI on human serum albumin. (b) All lysine residues on human serum albumin.
More recently, we have begun an investigation to compare the binding sites of TDI and MDI on serum albumin under identical conditions (40
, pH 7.9). Under these conditions, MDI is observed to bind to a subset of 19 of the 37 sites observed for TDI (see ). Although it would be tempting to attribute differences in binding between TDI and MDI to steric effects, as seen in , many of the lysine residues that are reactive toward TDI but not MDI are open and highly accessible. We therefore hypothesize that the difference in observed binding between TDI and MDI is attributable to a combination of steric effects and the increased reactivity of TDI. The electron withdrawing character of the second N=C=O group on the aromatic ring of TDI significantly increases the reactivity of the first isocyanate. In contrast, the reactivity of the isocyanate functional group(s) on MDI is lower because the p-ethyl phenylisocyanate substituent is less electron-withdrawing. Wisnewski and coworkers also examined the reaction products between MDI and human serum albumin by HPLC-MS/MS. Their data indicated 14 binding sites on albumin, including 12 lysine and 2 asparagine residues, in relatively good agreement with the results presented in . In addition, these authors suggested that the four “dilysine” (KK) motifs in human serum albumin are important binding sites, and that MDI shows reactive specificity for the second lysine. As discussed previously, the ability of a lysine residue to transfer its proton to a nearby lysine or histidine residue may, in fact, lead to increased reactivity toward dNCOs. However, as dilysine motifs account for 8 of 36 TDI binding sites and 6 of 19 MDI binding sites, it is clear that two lysine residues in a “KK” arrangement are not essential to binding.
Lysine residues reactive to TDI but not MDI.