2.1 Chemicals:
The ligand solutions were prepared using reagents purchased from Sigma-Aldrich: cobalt(II) chloride (CAS: 7646-79-9), gold(III) chloride trihydrate (CAS: 16961-25-4), copper(II) chloride (CAS: 7447-39-4), silver acetate (CAS: 563-63-3), and L-thyroxine sodium salt pentahydrate (CAS: 6106-07-6). Dansylsarcosine Piperidinium Salt (>95%) was obtained from RareChemicals GmbH. All ligands were diluted with MilliQ purified water, and their pH levels were measured before use. Human albumin Vialebex, 200 mg/mL, was utilized to test binding capacity.
2.2 Setting up the Serum Enhanced Binding (SEB) test
2.2.1. Patients
We collected residual serum samples from 90 patients admitted to Limoges University Hospital, following their consent for the use of residual biological materials, in accordance with local and French regulations (Code de la Santé Public, Art. L1211-2) and the declaration of Helsinki for experiments involving human subjects. The study was coordinated by the University Hospital of Limoges and the biocollection was authorized by the French Ministry of Health and registered under numbers DC 2010-1074 and AC-2016-2758, in accordance with the French Bioethics Act 2011-814 of July 7, 2011. Informed consent has been obtained from all patients.
Among the 90 collected samples, 45 were obtained from control patients without hepatic impairment, and the remaining 45 were from patients at various stages of cirrhosis. The serum samples, collected in dry tubes and sent to the laboratory for routine biochemical tests, were used in this study after the completion of routine analyses. Control patients were included when their clinical diagnosis did not indicate any liver dysfunction, and their levels of transaminases (AST, ALT), alkaline phosphatases (ALP), gamma-glutamyl transferases (GGT), total and conjugated bilirubin (BILIT, BILID), and lactate dehydrogenase (LDH) were within normal ranges. Cirrhotic patients, classified as cirrhosis A, B, or C using Child-Pugh scores, were diagnosed according to BAVENO VII recommendations 18.
2.2.2. Analytical procedures
2.2.2.1. Ligands optimization
We first evaluated the global capacity of serum to bind Cu, Au, L-thyroxine, Cd and dansylsarcosine in patients with no liver dysfunction. Each ligand was independently added in increasing concentrations to patient serum samples in order to obtain HSA/ligand theoretical ratios (mol/mol) of 1/1, 1/5, 1/10, 1/20, 1/50, 1/100, 1/500, and 1/1000 when possible. These theoretical ratios were calculated with 0.6 mM as an average concentration HSA in the serum.
Six different serum samples (from six different patients) per ligand and per ratio were used for this evaluation. After incubation for 30 min at 4°C, the serum samples were ultrafiltrated on Amicon® filters with a 30 kDa cut-off and 10 µL of the ultrafiltrate was then diluted in HNO3 0.1 M before analysis using a multi-element ICP-MS method for the determination of free (unbound) concentrations of Cu, Au, Cd, iodine (for L-Thyroxine) and sulfur (for dansylsarcosine). The bound fractions as well as the concentration ratios of HSA/bound ligand (mol/mol) were then calculated.
To confirm that the binding is only due to HSA and that there is no unspecific binding, we performed the same tests on a commercial human albumin solution Vialebexâ, 200mg/mL. This allowed us to determine for each ligand its maximum unspecific binding capacity. These thresholds were then set to best discriminate between serum samples containing mostly modified HSA or mostly native HSA.
Comparison of HSA binding capacities in patients with liver cirrhosis or no hepatic dysfunction:
After establishing the threshold for >90% albumin binding for each ligand, we proceeded to apply the Serum Enhanced Binding (SEB) test to a separate group of patients: 12 diagnosed with cirrhosis and 12 with no liver dysfunction. It's important to note that these patients were different from those used in the analytical development of the SEB test.
In brief, independent solutions of Cu and Cd at 1190 µM, Au at 11900 µM, and L-thyroxine at 75 µM were individually incubated with 200 µL of serum for Cd, Cu, L-thyroxine, and dansylsarcosine to achieve HSA/ligand theoretical ratios (mol/mol) of 1/5. For Au, 50 µL of serum was incubated to obtain an HSA/ligand ratio of 1/50. The same ligands and concentrations were also applied to rat plasma albumin in our animal models for comparison.
2.2.2.2. ICP-MS analysis:
Calibration curves were constructed for each element using a six-point calibration range of 10 to 100 µg/L for Cu, Cd, Au, and sulfur, and 1 to 20 µg/L for L-thyroxine. Sulfur calibration employed L-cysteine, while iodine calibration utilized L-thyroxine. The KED mode was applied with oxygen at a flow rate of 0.3 ml/min for both calibrators, controls, and ultrafiltrates. When necessary, ultrafiltrates were diluted with 0.1 M HNO3. Cu was measured at m/z 65, Cd at m/z 112, Au at m/z 197, iodine at m/z 127, and sulfur at m/z 48, following previously described methods19. To standardize SEB test results across samples, the concentration of each ligand in the ultrafiltrate was normalized by the total albumin concentration in the sample.
2.3 Animal experiments:
All animal care and experimental procedures were approved by the French Ministry of Higher Education, Research and Innovation (APAFIS reference APAFIS#20354-2019042414581742) and were performed in accordance with the guidelines for animal experimentation of the European Communities Council Directive (EU/63/2010). These experiments and procedures are reported according to the ARRIVE guidelines20,21 with the recommendations made by French Ministry of Higher Education. All methods are reported in accordance with ARRIVE guidelines (https://arriveguidelines.org).
For all included animals, weight was recorded daily throughout the experimental period until sacrifice. Animals were euthanized with intraperitoneal injection of pentobarbital (150 mg/kg). Blood was collected in Vacutainer® lithium heparin tubes (Beckton Dickinson, France) and centrifuged at 3000 rpm for 10 minutes. The resulting plasma samples were stored at -80 °C for analysis. The liver was promptly excised and fixed in formalin for histological analysis.
2.3.1. Induction of ethanol (EtOH) hepatotoxicity:
Six different groups of 6 rats were orally administered (by gavage) 2 mL of a 50% EtOH solution (equivalent to 0.4 g or 1.6 g/kg of body weight) prepared in physiological saline (0.9% NaCl). The groups were followed for 1, 3, 7, 10, or 14 days, respectively to evaluate the time-dependent changes in biochemical markers and histological liver injuries. The animals were sacrificed 24 hours after their last intake of EtOH. Control rats (n=6) received 0.9% NaCl by gavage for 14 days.
2.3.2. Induction of carbon tetrachloride (CCl4) hepatotoxicity:
Five groups of 6 rats were orally administered (by gavage) 1 mL/kg body weight of a 30% solution of CCl4 diluted in olive oil (equivalent to 1.594 g/kg of body weight). The groups were followed for 1, 3, 7, 10 days, respectively and the animals were sacrificed 24 hours after their last intake of CCl4. Control rats (n=6) received olive oil by gavage for 10 days.
2.3.3. Pathological analysis of the liver
After animal sacrifice, the liver was cut into sections of 1 to 1.5 cm perpendicular to the major axis to allow homogeneous fixation in a 4% formalin solution, and kept at ambient temperature for a maximum of 7 days. Samples were stained for light microscopy with hematoxylin, eosin and Masson's trichrome. The pathologist performed histological analysis blindly of the experimental groups.
2.3.4 Biochemistry analyzes
From the collected plasma, measurements of classic biochemistry parameters such as albumin (ALB), total (BILIT) and conjugated (BILID) bilirubin, aspartate aminotransferases (ASAT), alanine aminotransferases (ALAT) and alkaline phosphatases (PAL) were determined using a COBAS® 8000 system (Roche, Germany).
2.3.5. Characterization and quantification of albumin isoforms in rat plasma
Albumin isoforms were determined using the method described elsewhere 22. Briefly, 20 µL of plasma were diluted with 980 µL of an aqueous solution of 20mM ammonium formate with 0.1 % formic acid, vortex-mixed before filtration on a 0.22 µm cellulose filter and then injected on LC-HR-MS system (Nexera LC40 system coupled to a TripleTOF® 5600+, Shimadzu, Noisiel, France and Sciex, Concord, Canada). The LC-HRMS data were processed using PeakView® 2.2 and its Bio Tool Kit 2.2.0 (Sciex). The input MS spectra were filtered-in between 1300 to 1600 and then deconvoluted at low resolution (5000) between m/z 1,000 to 200,000.
2.4 Molecular dynamics simulations:
The initial HMA model was obtained from the protein data bank (PDB ID 5GIX) which has been co-crystallized with seven palmitate (PLM) molecules 23. Missing N- and C-terminal residues were added using the Modeller software 24. Protonation states of titratables residues (namely arginine, lysine, glutamate and aspartate) were determined using the H++ server assuming a physiological pH at 7.4. Since HMA was shown to be natively bound to PLM with a molecular ratio ranging from 0.1 to 2.0 in physiological conditions, we built the present HMA model with two PLM molecules, which were considered docked in fatty acid binding sites (FABS) 2 and 5 25,26. The apo form of HMA was also considered in the present investigations. Three models of palmitate-bound PTM albumins were also built in the present study, namely HNA1, HNA2 and N-truncated HMA. All systems were solvated in explicit water box for which minimal distance between atoms and box edges were set up at 10 Å. Systems were all neutralized considering physiological NaCl salt concentration ([NaCl] = 154 mM). FF14SB 27 and TIP3P 28,29 forcefields were used to respectively model protein residues and water molecules. Parameters from Joung and Cheatham 30 were used to model Na+ and Cl- counterions. Parameters for PLM, S-bonded cysteine were derived from Lipid17 31 and FF14SB force fields while those for cysteic acid residues were derived from amber 99SB-based parameters available in the literature 32. Each replica was minimized and then equilibrated for 10.25 ns. MD production run were performed for 2 ms. Simulations were analyzed using the CPPTRAJ package 33, and in house python scripts. Plots were obtained using the matplotlib v3.7.0 Python package 34. Rendering was prepared using VMD software (alpha-v1.9.4). Structural clustering was carried out using density peak algorithms 35, with inter-subdomain distances as metrics. Clusters representing more than 10% of the overall conformational space were considered. Allosteric communications were calculated with Allopath Tool 36.