Retrospective detection for V-type OPNAs exposure via phosphonylation and disulfide adducts in albumin

Organophosphorus nerve agents (OPNAs) that damage the central nervous system by inhibiting acetylcholinesterase activity, pose severe threats to human health and life security. Reliable biomarkers that quickly and accurately detect OPNAs exposure are urgently needed to help diagnose quickly and treat in time. Albumins that covalently bind to OPNAs could serve as important targets for retrospective verification of OPNAs exposure. The goal of this study is to explore the potential biomarkers in albumins with high reactivity and good stability and expand the group of potential biomarkers in different species for detecting the exposure of V-type OPNAs including O-ethyl S-(2-(diisopropylamino)ethyl) methylphosphonothioate (VX), O-isobutyl S-(2(diethylamino)ethyl) methylphosphonothioate (VR), and O-butyl S-(2-(diethylamino)ethyl) methylphosphonothioate (Vs). Taking human serum albumin (HSA), bovine serum albumin (BSA) and rabbit serum albumin (RSA) as the research objectives, multiple active sites including phosphonylation and disulfide adduct sites were observed in albumins from different species. Numerous phosphonylation sites labeled by all agents in one type of albumin were found. Among the different species, four shared phosphonylation sites with high reactivity include K499, K549, K249, and Y108. In addition, Y108 on ETY*GEMADCCAK, Y287 on Y*ICENQDSISSK, Y377 on TY*ETTLEK and Y164 on YLY*EIAR in HSA were stably phosphonylated by all agents in gradient concentration, making them stable and suitable potential biomarkers for V-type OPNAs exposure. Notably, Y108 on ETY*GEMADCCAK in HSA, on DTY*GDVADCCEK in RSA, and on ETY*GDMADCCEK in BSA were highly reactive to all V-type agents, regardless of species. It was also successfully labeled in HSA exposed to class V agents in gradient concentration. Y108 is expected to be used to screen and identify the exposure of V-type agents in the retrospective research. Disulfide adducts sites, consisted of four sites in HSA and two sites in BSA were also successfully labeled by V-type agents, and characteristic ion fragments from these disulfide adducts were also identified by secondary mass spectrometry. Molecular simulation of the stably modified sites were conducted to discover the promoting factors of covalent adduct formation, which help further clarify formation mechanism of albumin adducts at active sites.

safety 3,6,15,41,50,51 . The agents enter the body and attack albumin to form two types of protein adducts: phosphonylation adducts and disulfide adducts 40,45,52 . In order to expand the group of potential biomarkers in different species, the work utilized different albumins with similar amino acid sequences. The amino acid sequences in serum albumin from the UniProt database revealed that the sequence homologies of HSA with RSA and BSA were about 75.21% and 76.36%, respectively. Significant similarity among the active peptides indicated minimal differences in albumins from rabbits, bovines, and humans. Therefore, the modified sites shared by albumins from three species show the reliability of these adducts to a certain extent, providing reference data for finding potential biomarkers.
The active sites across all samples were modified by the V-type agents in two ways: phosphonylation and disulfide bond exchange. As shown in Tables 1, 2, 3, there were ten phosphonylation sites in BSA, eight in HSA and ten in RSA. These active sites were simultaneously labeled by the three poisons: Vs, VX and VR, indicating that the active sites had highly reactivity to these agents. However, it was different in the selectivity of OPNAs to active sites. For example, K438 on K*VPQVS*TPTLVEVSR in HSA could be modified by VX and Vs, while S443 could only be labeled with VR. K490 and S494 on LCVLHEK*TPVS*EK showed similar differences in selectivity.  The differences in selection did not affect the productive information provided by the modified peptides after OPNAs exposure. Thus, the targeted identification of these modified peptides and active sites showed great potential to detecting the exposure of V-type agents. Highly homologous sequences from the different species had the same phosphonylated peptides and active sites, even though some amino acid residues were different. As shown in Table 4, K499, K549, K249, and Y108 modified by poisons were shared active sites in HSA, BSA, and RSA. Remarkably, Y108 on ETY*GEMADCCAK in HSA, on DTY*GDVADCCEK in RSA and on ETY*GDMADCCEK in BSA, were phosphonylated by all the V-type agents. These active sites were deemed suitable as potential biomarkers for identifying V-type OPNAs exposure due to their low selectivity and high reactivity to class V toxicants. Furthermore, these sites showed no species differences. The other three modified sites exhibited different selectivity and reactivity to the OPNAs. These shared sites provided clear guides to act as biomarkers in the different species due to they could screen targeted V poisons. For example, K499 on VTK*CCTESLVNR could help screen the exposure of VR in BSA. Moreover, K249 on FPK*AEFVEVTK could be used to detect VR in BSA, and K549 on K*QTALVELVK could   www.nature.com/scientificreports/ be used to identify VX in HSA. When these sites were located in albumins from other species, they could be used to screen for all V-type poisons. The active sites of the phosphonylated adducts were mainly distributed in the various amino acid residues with active side chains, while the disulfide adduct sites were limited to cysteine residues without species-related differences. As shown in Table 5, a total of six disulfide adducts were identified between the V-type agents and HSA (four sites), or BSA (two sites). These peptides (VTKCCTESLVNR and ETYGEMADCCAK in HSA, ETYG-DMADCCEK in BSA) were labeled at cysteine residues, and were regarded as potential biomarkers for V-type agents exposure through phosphonylation identification previously. These modified sites further confirmed the reliability of the modified peptides as potential biomarkers. However, there is no recognized disulfide adducts as biomarkers at present. This may be due to the formation mechanism of disulfide adducts. The disulfide adducts formed via disulfide bonds exchange between the thiol groups from the leaving groups and cysteine residues, following the formation of phosphonylation adducts 39,53 . The data in Table 5 showed that the labeled peptides from the disulfide adducts were accompanied by phosphonylated sites, which further verified this conclusion.

MS/MS analysis of the modified peptides.
Mass spectrometry had served as a powerful tool for identifying protein adducts, and it can provide reliable evidence to further explore the adduct mechanism and discover more active sites 18,25,50,[54][55][56] . The OPNA-albumin adducts formed when small molecules attacked the active residues in the albumin. Some stable and reliable adducts were regarded as potential biomarkers that can be utilized in detecting OPNA exposure. The accurate molecular weight of the amino acid residues of trypsin-digested peptides was determined before and after OPNAs modification. Thus, the reactive sites were discovered by calculating the mass-to-charge ratio of b and y series ions. The representative MS/MS spectra of highly reactive peptides in albumins from different species were shown in Fig. 2, where the key peaks were marked by blue lines for y-ion series, red lines for b-ion series, and green lines for the characteristic precursor ions.  Table 6. The characteristic precursor ions were captured, including phosphoryl lysine imine ions losing ammonia, its fragment ion after losing the butyl group and the phosphoryl lysine imine ions, which provided direct evidence for phosphonylation at lysine residues. The characteristic ions fragments from albumin adducts at lysine were summarized in Table 7. Two lysine residues in the entire peptide were potential phosphonylation sites, but more effective signals are needed to determine the position of phosphonylated residue. The mass of the y1 ion equaled the mass of lysine residue indicating that it had not been modified and modified site may be at the other lysine residue. The y1-y8 ions did not show phosphonylation of the AEFAEVSK amino acid sequences, while the mass gap between y8 and y9 ions equaled the mass of the lysine residue plus one molecule of O-butyl methylphophonate. This demonstrated that Vs was bound to K249 on FPK*AEFAEVSK. This conclusion was further confirmed by the corresponding mass differences between the b2 and b3 ions.  (Fig. 2e), helped to confirm the phosphonylated peptides and modified small molecules, while no relevant information was found in Fig. 2f. The characteristic ion peaks of the methyl phosphonate tyrosine imine ion (214.0624 amu) indicated that this peptide was modified by V-type agents, and the O-alkylphosphonate tyrosine imine ion (270.1251 amu) provided the evidence of phosphonylated by Vs or VR 30,39 . The b and y series ions assisted in offering accurate information that further confirmed site modification. The mass gap between the b 3 and b 2 ions was equal to the mass of one tyrosine residue (163.0633 amu) plus the O-alkylphosphonate group. The mass of the peptide GEMADCCAK corresponded to the y 1 -y 9 ions, while the mass differences between the y 9 and y 10 ions equaled the b 3 ion minus the b 2 ion. It provided the evidence for phosphonylated site Y108 at the third residue from the C-terminus in the ETY*GDMADCCEK peptide. Figure 2g-h showed the disulfide adducts between the cysteine residues and Vs in BSA (Fig. 2g), VX in HSA (Fig. 2h). The characteristic ion peaks of the leaving group of VX (2(diisopropylamino) ethanethiol) (m/z 160.1156), trimethylamine (m/z 128.2388) generated by α-cleavage, and (E)-3-disulfanylacrylic acid (m/z 136.1832) or (m/z 136.1826) generated by β-elimination. The characteristic ions of albumin-adducts at www.nature.com/scientificreports/ endogenous cysteine residues after V-type OPNAs exposure were summarized in Fig. 3. The mass values of the b 1 -b 6 ions corresponding to ETYGEM from the N terminus, and y 1 -y 3 corresponding to CAK from the C terminus offered evidence that there were no modification sites. The mass gap of y 4 ions minus the y 3 ions equaled the molecular weight of the cysteine residue plus the corresponding dialkylamino ethylthiol group of VX or Vs. Therefore, these data suggested that C114 on peptide ETYGDMADC*CEK was modified by the class V agents with a disulfide bond. Liquid chromatography-tandem mass spectrometry was a commonly used method to detect modified peptides obtained by the enzymatic hydrolysis of protein adducts 9,26,28,48 . Analysis of the modified peptides was based mainly on the mass differences between the b and y series ions, and the characteristic fragment ions in secondary mass spectrometry. Characteristic ions consisting of the same structure, such as 2(diethylamino) ethanethiol) from VR or Vs, and isomers such as the O-butyl methylphophonate group from Vs and the O-isobutyl methylphophonate group from VR are not distinguished by mass spectrometry. In addition, according to the peak intensity of the disulfide adduct fragment ions and linker bond breakage energies, the possible characteristic fragment ions of the disulfide adducts were summarized in Fig. 3, and generated by disulfide bond breakage, α-cleavage and β-elimination.
Stably labeled HSA peptides by OPNAs. The ability of the OPNAs to bind to amino acids in albumin was affected by the number of active residues 13,18,25,54,56 . One molecule of OPNAs is covalently linked to one molecule of active amino acid. Using tyrosine as an example, 609 amino acid residues were found in one molecule of albumin, including nineteen tyrosine residues. Ideally, all 19 tyrosine residues would be modified by 19 molecules of OPNAs. With a constant concentration of albumin, augmenting the concentration of OPNAs would increase the binding rate between the OPNAs and active residues until the modified active sites were saturated. Thus, the lower the concentration, the fewer active residues were available for OPNAs modification. The residues continuously modified were identified as stable modification sites, suggesting the high selectivity for V-type poisons and the potential as biomarkers.
For HSA, seven peptides that could be stably modified by various concentrations of OPNAs were identified. As shown in Fig. 4, Gradient concentration of OPNAs refers to the final concentration of OPNAs in the solution system. The peak area of the phosphonylation sites was enhanced with increasing OPNAs concentration, and the increased signal intensities in the mass spectra indicated the binding rates of the OPNAs to these sites increased. As shown in Table 8, sustainable modification sites were observed in HSA exposed to V-type agents in a range of concentrations. Y108 on ETY*GEMADCCAK, Y287 on Y*ICENQDSISSK, Y377 on TY*ETTLEK, and Y164 on YLY*EIAR were stably phosphonylated by all of the poisons, providing evidence for identifying V-type OPNAs exposure as potential biomarkers.

Molecular simulation and structural analysis of the Vs adducts. Molecular simulation was used
to study the protein adducts between the stable modification sites (K588, Y287, Y108, K130, Y164, Y377, and C77) and Vs by (1) calculating the solvent accessibility (SAS) and solvent accessible area ratios (% SAS) of the active sites (residue exposure in HSA when %SAS > 10) to determine the positions of modified residues on HSA, (2) predicting the interactions between the active sites and surrounding residues, and inferring the possible influencing factors to promote adduct formation, (3) predicting the covalent interactions between the potential biomarker and Vs, and understanding the influence of spatial effect of the active sites on HSA modification. Table 9 showed that the five modified sites (K588, Y287, Y108, K130 and Y164) were in exposed state in the HSA structure, which increased the collision probability between these sites and Vs and provided better reactive  Table 7. Characteristic ions of albumin adducts at lysine residue.    www.nature.com/scientificreports/ conditions for adduct formation. The %SAS of the two residues Y377 and C77was approximately zero, which indicated they were buried in HSA. It is worth noting that the residues distributed on the surface of the HSA could increase the possibility of binding the small molecules of toxic agents, while the buried residues still had the ability to attack the phosphorus atoms in Vs. The interactions between the active sites and surrounding residues could create powerful conditions for adduct formation between HSA and Vs. Two types of modified site interaction were found, including hydrogen and disulfide bonds, as shown in Table 10. Hydrogen bonding forces increased with decreasing distance between the labeled sites and others. Generally, hydrogen bonding action is effective within a distance of 3.5 Å. We observed that Y287, K130, Y164, and Y164 formed strong hydrogen bonds with the corresponding residues as the hydrogen bond distances were all within 2.5 Å. The strongest interactions were found between the Y164 (acceptor) and R192 (donor). These interactions exhibited a certain hydrophobicity, which helped to stabilize the toxic molecules bound to the active sites. C77 also displayed a disulfide bond with C110 at a distance of 2 Å. The side chain of C77 will be in a bound state under normal conditions, and may become free during the experiment. Subsequently, disulfide bond was formed with Vs. C77 is a potential biomarker as it was stably modified under these conditions. The binding free energy of Vs to the active sites was also calculated. The lower binding free energy resulted in the stronger binding action between Vs and HSA, providing the greater probability of Vs binding to the active sites. The binding energies of these active sites were similar. K130 had the lowest binding energy with Vs, and creating advantageous conditions for subsequent covalent binding.
The ETY*GEMADCCAK peptide was determined to be stably modified by V-type agents, where Y108 on the peptide was regarded as a potential biomarker for detecting OPNA exposure. As shown in Fig. 5, covalent molecular docking was used to predict the covalent binding energy between Vs and active residue Y108 (Y84). The results showed lower binding energy (− 5.51 kcal/mol, less than standard value − 5 kcal/mol) between Vs and the labeled site, illustrating there is a certain binding ability and good matching degree. To obtain a binding pattern between Vs and HSA, these adducts were visualized with Pymol 2.1 software. According to the binding patterns, the amino acid residues in the HSA pocket and Vs were observed, as well as the interaction between Vs and other residues (V77, L80, P35, K41, and D38) adjacent to Y108 was found, as shown in Table 11. This adduct formed by phosphorus groups from Vs covalently linked to the hydroxyl group at Y108. The hydrophobic chains with certain hydrophobicity formed hydrogen bonds with V77, L80, P35, and K41 in the active pocket of albumin. Thus, the hydrophobic effect of hydrogen bonds played an important role in stabilizing the small molecules of the toxic agents in HSA. In addition, electrostatic repulsions between the sulfur atoms of Vs and www.nature.com/scientificreports/ the carboxyl groups of D62 also contributed to stabilizing Vs binding to the active site. In summary, Vs had a good binding ability and matching degree with Y108 in the HSA.

Conclusions
This work took HSA, BSA and RSA as research objectives, discovering the possible biomarkers and expanding the group of the possible biomarkers in different species, which provided a theoretical basis for retrospective identifying potential biomarkers to determine V-type OPNAs exposure. The possible active residues in albumins were examined according to the following two parallel experimental workflows: neat HSA, BSA, and RSA were sequentially exposed to V-type OPNAs in vitro, and pure HSA was subjected to class V agents in different concentrations. Numerous shared active sites were observed in albumins from different species including phosphonylation sites (four sites) and one disulfide adduct site. Notably, Y108 on ETY*GEMADC*CAK in HSA, DTY*GDVADC*CEK in RSA, and ETY*GDMADC*CEK in BSA were shared phosphonylation site labeled by all poisons tested. C114, a disulfide adduct site on these homologous sequences, was selectively modified by class V agents in both HSA and BSA. The results indicated that these sequences could be used as potential biomarkers to trace the exposure of V-type OPNAs. In particular, Y108 on ETY*GEMADC*CAK in HSA was stably phosphonylated by V-type agents in gradient concentration which further illustrated the reliability of this sequence for screening V-type OPNAs. Molecular docking showed the promoting factors for the formation of Vs adducts at Y108 in HSA consisted of low covalent binding energy, hydrogen bonding, and electrostatic interactions among the residues adjacent to Y108 and Vs. Y287 on Y*ICENQDSISSK, Y377 on TY*ETTLEK and Y164 on YLY*EIAR in HSA were also stable phosphonylated by all the poisons tested. This suggested they were potential biomarkers of V-type toxicant exposure as well.

Materials and methods
Caution: V-type OPNAs are highly toxic organophosphorus compounds. To ensure their safety, the operators were professionally trained and protected.  LC-MS analysis. The peptide mixtures were analyzed on an Orbitrap Fusion Lumos (Thermo Fisher Scientific) mass spectrometer interface using an Easy-nLC 1200 nanoflow liquid chromatography system (Thermo Fisher Scientific) with nanospray ionization (NSI) in positive ion polarity mode. The samples were dissolved in mobile phase A (0.1% formic acid aqueous solution) and mixed thoroughly. Then, Nano LC system (Thermo Fisher Scientific) with a 150 μm × 12 cm C18 reversed-phase chromatographic column (2 cm × 100 µm) was used to separate the samples. The mobile phase consisted of phase A (0.1% formic acid and 100% water) and phase B (0.1% formic acid, 80% acetonitrile and 20% water). The elution program was set to a 30 min linear gradient elution (25 min 11% B, 1 min 41% B and 4 min 100% B) and the flow rate was 600nL/min. Data acquisition adopt Orbitrap fusion Lumos mass spectrometer with nano-current electrospray ion source (NSI), the spray voltage of 2200 V and Ion transport capillary of 320 °C. Mass spectrometry data were collected in data-dependent acquisition (DDA) mode in the positive ion mode. A full scan (200-1500 m/z, 120,000 resolution) was conducted via an Orbitrap mass analyzer. Related parameters included the scanning range of 200-1500 m/z, the resolution of 120,000, the automatic gain control targets of 5e 5 ions and iron max injection time of 50 ms. The fragment ions were detected in the Orbitrap following fragmentation of the parent ions via 32% high-energy collision dissociation using two-dimensional mass spectrometry. Related parameters included the first mass of 50, the resolution of 150,000, the automatic gain control targets of 5e 5 ions and iron max injection time of 22 ms.

Materials. V-type
Qualitative analysis of protein. The original mass spectral data were searched with Sequest, which was embedded in Proteome Discoverer 2.4. HSA, RSA, and BSA sequence databases were obtained from Uniprot. The parameters deviations of the parent ion and fragment ion mass were set to 20 ppm and 0.05 Da, respectively. False positives were controlled by reverse sequence decoy strategy, and the results were verified using Percolator software.
Computer simulations. Molecular simulations were performed using Discovery Studio 4.5 molecular simulation software. The three-dimensional crystal structure of HSA was obtained from the protein crystal database (http:// www. rcsb. org/ pdb/, PDB number 1bm0). Residue solvent accessibility (SAS) and its percentage (%SAS) were obtained using the DelPhi method according to the electrostatic potential of the focused section. Then, the non-binding interaction monitor method was applied to predict non-binding interactions of the seven active residues with adjacent amino acids. A molecular docking technique was used to calculate the noncovalent binding energies between Vs and the modified residues in the HSA 28,33,39 . Covalent docking between Vs and Y108 on HSA was conducted on the molecular manipulation platform (MOE2019.01). Protein processing and molecular docking were performed using the structure preparation and docking modules respectively. Afterward, the interactions between the surrounding residues adjacent to Y108 and Vs were calculated afterward.

Data availability
All data supporting the findings of this study are provided within the manuscript.