Glycated Albumin With Loss of Fatty Acid Binding Capacity Contributes to Enhanced Arachidonate Oxygenation and Platelet Hyperactivity: Relevance in Patients With Type 2 Diabetes

Many epidemiological studies have established an inverse relationship between the level of serum albumin and the risk of death (1,2). After adjustment for usual risk factors, this association was also found to hold not only for liver dysfunction (3) but also for cardiovascular disease and stroke (4–6). Albumin is involved in several biological mechanisms, including the regulation of oncotic pressure, and the binding and transport of a number of hydrophobic or amphipathic molecules (mainly nonesterified fatty acids [NEFAs]). In addition, albumin displays potent antioxidant and free radical scavenging activities through the redox cycling of its free thiol (Cys³⁴) (7) and its ability to bind metal ions (2,8,9). High levels of NEFAs are associated with type 2 diabetes (T2D) and metabolic syndrome (10,11), and they are known to mediate ventricular arrhythmia and sudden death (10,12). Hypoalbuminemia resulting from either very rare familial deficiency or increased clearance in kidney disease has been associated both with abnormally elevated NEFA concentrations in serum and with high cardiovascular risk (13), suggesting a putative role of serum albumin in NEFA metabolism. Low plasma concentrations of albumin are known to have adverse effects, and high levels of glucose and free radicals may impair the biological properties of serum albumin through the formation of harmful adducts (14). The nonenzymatic covalent attachment of glucose molecules to protein and the subsequent free radical–mediated oxidation give rise to advanced glycoxidation end (AGE) products, of which methylglyoxal (MGO) is considered a key component through the Amadori rearrangement (15). Several reports indicate that AGE products are able at least in vitro to induce activation of various cells including platelets (16,17). In addition to classical glycemic indices, several recent reports indicated that glycated albumin is a relevant and practical biomarker for the progression of diabetic atherosclerosis (18–20).

Although most circulating NEFAs are produced by the adipose tissue during lipolysis, phospholipases also contribute to NEFA production during the platelet activation process. Blood platelets are not only essential for primary hemostasis but also play a central role in the pathophysiology of atherothrombosis. Upon adhesion and activation by various stimuli (such as thrombin, ADP, and prostanoids), platelets undergo shape changes, express receptors, aggregate, and release their cytosolic granule content (21). Clinical and epidemiological studies provide strong evidence for an association between platelet hyperactivity and atherothrombosis (22). Numerous investigations have reported that risk factors for cardiovascular diseases such as tobacco smoking, hyperlipidemia, hypertension, and diabetes are associated with enhanced adhesion, as well as spontaneous and inducible platelet activation (23,24).

Arachidonate can bind to albumin, which can then modulate its conversion into active oxidized metabolites. However, current knowledge concerning the influence of modified versus native albumin (N-Alb) on blood platelets, in particular on their ability to release arachidonate metabolites, is scarce. In the first part of the current study, we investigated abnormalities in the structure and NEFA-binding property of albumin in plasma of T2D patients compared with healthy control subjects. In the second part, we set up an experimental protocol to reproduce in vitro the harmful derivatives of albumin that actually occur in the plasma of T2D patients. Finally, we demonstrated that unmodified albumin can exert a negative regulatory effect on platelet adhesion and aggregation. When glycoxidation of albumin occurs, its fatty acid binding capacity diminishes, and fewer arachidonate molecules are sequestered. Thus, more arachidonate is available for conversion into eicosanoids, which are potent platelet activators. This newly described mechanism might contribute to the elevated thromboembolic risk in patients with T2D.

Control subjects were healthy volunteers (n = 20) with no history of hyperlipidemia, coronary artery disease, or diabetes. Patients (n = 20) had T2D. The protocol was approved by the local medical ethics committee, and all participants gave written informed consent. The clinical and biological parameters of the subjects are described in Table 1. For the patients with T2D, 7 were received insulin treatment, 11 with sulfamide, and 11 with metformin. Five patients received a treatment with a fibrate, and 14 were treated with a statin for hyperlipidemia. One patient was administrated an antiplatelet agent. As shown by creatinine clearance, there was not any patient in a severe chronic renal insufficiency (clearance <30 mL/min), and three had a moderate chronic renal insufficiency (clearance between 30 and 60 mL/min).

Fasting blood samples were collected and specifically handled for each test. Plasma was obtained from citrate-containing glass tubes and separated by centrifugation (15 min, 4°C, 3,000g). Routine biochemistry analyses (plasma lipids, lipoproteins, and apolipoproteins) were performed according to standard protocols (25,26). Plasma albumin concentrations were measured by a standard procedure using bromocresol green with human albumin as the standard (27,28).

Blue Sepharose (Pharmacia Fine Chemicals) was used for albumin isolation by gel purification according to a previously published procedure (29). Briefly, the plasma (1.350 mL) was mixed with the gel (2.5 mL) equilibrated with starting buffer (50 mmol/L Tris/HCl [pH 7.4] and 0.5 mol/L NaCl). After centrifugation (4,000g, 5 min, 4°C), the supernatant was discarded. The gel was washed three times with starting buffer under the same conditions. Gel-bound albumin was eluted (0.650 mL) with 50 mmol/L Tris/HCl (pH 7.4) containing 1.5 mol/L NaCl.

Native BSA and human serum albumin (HSA), MGO- and glucose-treated albumin, and isolated albumin fractions from control subjects or T2D patients and protein pI standards (SERVA) were diluted in 125 μL hydration buffer (8 mol/L urea, 4% 3-[(3-Cholamidopropyl)dimethylammonio]-1-propanesulfonate [CHAPS], 20 mmol/L dithiothreitol, 0.2% Bio-Lyte 3/10). After overnight hydration of 7-cm-long ReadyStrip, pH 4–7 (Bio-Rad), at 50 V in a Protean IEF cell (Bio-Rad), isoelectrofocalization was conducted for 12 kV-h. Either the strips were stained with Coomassie Brilliant Blue G-250 or proteins were subjected to second-dimension SDS-PAGE (NuPAGE 4–12% bis-Tris gels; Invitrogen). Blots were analyzed with Image Laboratory software on a ChemiDoc device (Bio-Rad). Coomassie-stained spots of interest were manually excised for samples from six control subjects and six T2D patients. The gel pieces were then washed twice with 0.1 mol/L NH₄HCO₃ and 100% acetonitrile (ACN) for 10 min. Reduction/alkylation was achieved by incubating the excised spots successively in 10 mmol/L Tris(2-carboxyethyl) phosphine/0.1 mol/L NH₄HCO₃ for 30 min at 37°C and in 55 mmol/L iodoacetamide/0.1 mol/L NH₄HCO₃ for 20 min. Peptide fragments were obtained after digestion with a solution of 40 mmol/L NH₄HCO₃ and 10% ACN containing 5 ng/µL trypsin (Promega) for 3 h at 37°C. The resulting peptides were acidified with 0.1% formic acid. Extraction from the polyacrylamide gel piece was performed by two successive incubations for 5 min in ACN with sonication. Digests were concentrated by evaporation until 10 µL. The concentrate (0.5 μL) was deposited onto a Ground Steel target (Bruker Daltonics, Bremen, Germany), mixed with 1 μL matrix solution (3.5 mg/mL cyano-4-hydrocycinnamic acid [CHCA] in ACN 50% and TFA 0.2%). Analysis was conducted using a MALDI-TOF/TOF ultrafleXtreme (Bruker Daltonics) in the automatic mode operating at 1,000 Hz in the reflectron mode. Mass calibration was done using the peptide calibration standards from Bruker Daltonics. Proteins were identified within Swiss-Prot, restricting the taxonomy to Homo sapiens. Methionine oxidation and carbamidomethyl modification of cysteines were accepted as variable and global modifications, respectively. The mass deviation tolerance was 30 ppm in mass spectrometry, and only one miscleavage was suggested.

The antioxidant defense of the subjects was examined using a test based on in vitro free radical–induced blood hemolysis (27,30). Fasting blood samples were drawn using 10% (v/v) buffered sodium citrate as the anticoagulant. Hemolysis was started by adding 52.4 mmol/L 2,2′-azo-bis(2-amidinopropane) HCl and was assayed by monitoring absorbance at 450 nm (Dynatech MR5000) of diluted blood from the subjects. Results are expressed as 50% of maximal hemolysis time (HT50) (in min) and referred to as the susceptibility of blood to free radicals. In humans and other animals in which oxidative stress has been well documented, HT50 was shown to be representative of the total defenses against free radicals (31).

Plasma lipid peroxidation was measured as thiobarbituric acid reactive substances (TBARS) in the presence of 2,6-di-tertbutyl-p-cresol and expressed as malondialdehyde equivalents as previously described (30).

As we needed a robust tool to test the effects of various albumin preparations on platelet function, we chose to use rat platelets because, due to their well-known lower reactivity, they present a much higher reproducibility in terms of activity and signaling compared with those from humans (32).

Male white Sprague-Dawley rats (230–250 g), 7 weeks of age, purchased from Charles River (L'Arbresle, France), were used as blood donors. They were housed in an air-conditioned room that was kept at constant temperature and hygrometry and on a 12-h light-dark cycle. Water and food were supplied ad libitum. All of the experiments involving animals were approved by the ethics committee on the Use of Laboratory Animals of Université de Bourgogne (protocol no. 3508), and they were performed in accordance with institutional guidelines.

Blood was drawn as in previous studies (24,33,34) from the jugular vein of overnight fasted animals into plastic syringes with 1 v anticoagulant (38 mmol/L citric acid, 75 mmol/L sodium citrate, and 136 mmol/L glucose) for 4 v blood. The platelet-rich plasma was obtained after centrifugation (210g, 10 min). The platelets, which were isolated by low-speed centrifugation, were counted with a thrombocoulter (Coultronics) and resuspended in Ca²⁺-free Tyrode buffer, pH 7.4.

Aggregation experiments were performed as in previous studies (24,33,34). Briefly, platelet suspension (3–4 × 10⁸/mL) was stimulated with agonists in a computerized four-channel laser-light coagulo-aggregometer (Servibio) under steering (1,100 rpm, 37°C). The final thrombin, ADP, calcium ionophore A23187, and arachidonate concentrations were chosen so as to induce 30–40% of the maximal aggregation (100% = Tyrode buffer) and were usually 0.08 units/mL, 0.6 μmol/L, 0.1 μmol/L, and 50 μmol/L, respectively. The final results are expressed as the percentage of control values obtained without albumin.

The platelet adhesion test was carried out using the modified colorimetric method of Bellavite et al. (35) and Eriksson and Whiss (36). Briefly, after coating of a microtiter plate with collagen I (Sigma), native BSA (N-BSA), or modified BSA and thorough washing, diluted platelet-rich plasma was added to the wells. After three washes with PBS, adherent platelets were measured by the reaction between p-nitrophenyl-diphosphate and the intracellular enzyme acid phosphatase liberated by lysis with 0.2% Triton ×100. Absorbance readings were performed at 405 nm after background subtraction (630 nm) with a computerized microplate reader (Dynatech MR5000). Control values were obtained without coating (plastic alone). The results are expressed as percentages with the following formula: (sample − blank)/(total − blank) × 100.

Phospholipids of washed rat platelets were labeled as previously described (33) by incubation with [1-¹⁴C]arachidonic acid (AA) (60 mCi/mmol; Amersham) for 60 min at 37°C. After temperature stabilization, the platelets (2 × 10⁹ platelets/mL) were stimulated at 37°C for 2 min with thrombin (final concentration 0.2 units/mL; Sigma). The results for metabolites were expressed as dpm per 10⁸ platelets. Platelet synthesis of thromboxane A₂ was also determined by enzyme immunoassay of its stable hydrolysis product, thromboxane B₂ (Cayman Chemical kits), as previously described (33). The results were expressed as picomoles of thromboxane B₂ (TxB₂) per 10⁸ platelets for 5 min stimulation.

Calcium uptake by platelets was measured using radiocalcium ⁴⁵CaCl₂ (Amersham), following published methods (33,34). Briefly, after the calcium concentration was adjusted to 0.3 mmol/L, radiocalcium was added to the platelet suspension (5 × 10⁸/mL), 1 min before incubation with thrombin (0.08 IU/mL), and centrifuged (9,000g, 30 s) through a layer of separating oil. The platelet pellet was then treated with 0.5 mL 0.5% Triton X-100 and transferred to counting vials with scintillation fluid. The results of the thrombin-induced calcium uptake were expressed as picomoles of Ca²⁺ per 10⁹ platelets and compared with appropriate controls.

To measure intraplatelet reactive oxygen species (ROS), we used dichlorofluroscin diacetate (DCFH-DA) in the reduced nonfluorescent form, which is able to penetrate platelets as described in previous works (37). The results were expressed as picomoles of dichlorofluorescein (DCF) per 10⁷ platelets using a calibration curve constructed with pure DCF.

As the sulfhydryl (SH) group in most commercial BSA and HSA preparations is in the oxidized form (38), fatty acid–free BSA and HSA obtained from Sigma were first reduced by a 4-h treatment with 4.5 mmol/L dithiotreitol and preserved from light. After dialysis against 20 mmol/L PBS, pH 7.4, to remove excess dithiotreitol, stock solutions of BSA (N-BSA) and native HSA (N-HSA) were stored for no longer than 2 weeks at 4°C under argon. This reduction treatment allowed the partial recovery of free Cys³⁴ from albumin as assessed by the thiol value, which increases from ~0.13 to 0.72 SH/mol albumin (see below). These reduced BSA or HSA preparations were used as the starting material for all of our studies. Proteins were measured using the bicinchoninic acid technique (Pierce) (39).

Glycated BSA (G25-BSA) was obtained as previously detailed by Bourdon et al. (14). Briefly, BSA was incubated with 25 mmol/L glucose in sterile conditions for 30 days at 37°C under argon. Excess glucose was removed by extensive dialyses against PBS, pH 7.4. BSA was also modified with 2 mmol/L MGO for 24 h at 37°C. Excess MGO was removed by extensive dialyses against PBS to obtain MGO-modified BSA (MGO-BSA). Similar modifications were performed with HSA to get G25-HSA and MGO-HSA.

Thiol groups of N-HSA or modified HSA were measured according to the Ellman assay (40) using 5,5′-dithiobis, 2-nitrobenzoic acid. The free thiol concentration was calculated from absorbance data obtained at 410 nm using the molar extinction coefficient ε = 13,600 L × mol⁻¹ × cm⁻¹ and expressed as SH per mole of BSA.

Free amino groups of native or modified BSA/HSA essentially from lysine were measured by means of their reactivity with 2,4,6-trinitrobenzenesulfonic acid (41). Data from absorbance at 340 nm were calculated with the molar extinction coefficient ε = 14,500 L × mol⁻¹ × cm⁻¹, and Lys-NH₂ was thus expressed as a percentage of control values obtained with N-BSA and N-HSA.

As BSA contains two tryptophan residues (Trp²¹³ and Trp¹³⁴) and HSA a single tryptophan (Trp²¹⁴), we were able to evaluate the effects of albumin modifications by monitoring the molecular conformation change by assaying its intrinsic fluorescence. Fluorescence measurements of albumin preparations were monitored using a spectrofluorometer (LS 50B; PerkinElmer) at excitation and emission wavelengths of 293 nm and 340 nm, respectively, as in previous studies (14). Quenching of fluorescence was also measured as previously described using acrylamide in final concentrations up to 40 mmol/L. The results of triplicate assays were expressed as mean values of I₀/I (Stern-Volmer constant), where I₀ and I represent the fluorescence before and after addition of the quencher, respectively.

Molecular conformation changes of albumin preparations were also investigated by measuring the fluorescence of 8-anilinonaphthalene-1-sulfonic acid (ANSA), which evaluates variations in the probe’s accessibility to hydrophobic protein sites, in accordance with previous studies (42). After incubation for 10 min of 0.5 mg/mL native or modified albumin with 200 μmol/L ANSA in 0.15 mol/L NaCl, fluorescence was measured at excitation and emission wavelengths of 385 and 465 nm, respectively. The results are expressed as arbitrary units per mg protein and changes normalized to N-BSA or N-HSA.

Interactions between oleic acid and native and modified albumin were evaluated by measuring the maximal absorbance in the 250–300 nm range 10 min after the addition of increasing concentrations of sodium oleate (up to 20 μmol/L) solutions (pH 9) according to the methodology of Fang et al. (43). The fitting equation was Y = (B_max × X)/(K_d + X), where X was the oleate concentration in micromoles per liter, B_max the maximal binding as absorbance changes (at 292 nm), and K_d the apparent oleate concentration in micromoles per liter required to reach half of the maximal binding.

Data are expressed as means ± SD from at least three experiments performed in triplicate. Linear and nonlinear curve fitting as well as the statistical comparisons were evaluated using one-way ANOVA followed by the Bonferroni posttest performed with Prism software (GraphPad Software, Inc., San Diego, CA). ANCOVA was used to examine the effect of T2D on plasma albumin levels when age was controlled. Statistical analyses were performed with STATA software, version 11 (StataCorp).

Thirty spectra were available from the MALDI-TOF/TOF analysis, but six of these did not meet biological criteria and were removed from the analysis. The following steps were taken to process the raw data: denoising, baseline subtraction, and normalization. Peaks were identified and quantified by their intensities. Statistical analysis was thus performed on 24 spectra (8 healthy and 16 diabetic spectra, respectively, representing healthy and diabetic patients) for 231 variables. Data were preprocessed and analyzed using R software (44). Principal components analysis (PCA) was implemented to visualize the principal sources of variability in the data. The aim of PCA is to reduce the dimensionality of the data set by finding new components that maximize variance of the data. The new variables, called principal components, are linear combinations of all the original variables. A sparse partial least squares (SPLS) regression (45) was thus implemented on the data to select peaks that mostly contribute to discriminate between both groups. This method is derived from PLS (46). The aim of the SPLS is to reduce the dimensionality of the data set by finding new components that maximize the discrimination between diagnostic groups. SPLS simultaneously makes it possible to select variables that mostly contribute to the discrimination, thus facilitating the biological interpretation; 20 peaks were thereby selected by SPLS as those separating at best samples from control subjects and samples from patients with diabetes. It was shown that these 20 peaks were sufficient to optimize predictions.

Table 1 shows the clinical and lipid parameters of 20 patients with T2D and 20 matched control subjects. Fasting glycemia was significantly higher in T2D with a mean HbA_1c value of 9.3% compared with values <6% in control plasma. Triglyceridemia, plasma NEFAs, age, and BMI were significantly higher in T2D patients than in control subjects (P < 0.001, P < 0.05, P < 0.001, and P < 0.01, respectively). Although total cholesterol and LDL cholesterol did not differ significantly between the two groups, HDL cholesterol levels were significantly lower in T2D patients (P < 0.01). Plasma fibrinogen levels were higher (12%) and albumin values were lower (18%) in T2D patients than in control subjects, with threefold higher Glc-HSA in the former group (Table 1). This could not be ascribable to an impairment of the renal status, as creatinine values were normal (92.3 ± 29.9 μmol/L), as was the MDRD marker (64.2 ± 26.8 mL/min). Glycated LDL, TBARS, and oxidized LDL levels were significantly higher in T2D than in control subjects (P < 0.05 in all cases). Total free radical defense was measured on whole blood samples, and the HT50 parameter (i.e., the 50% maximal hemolysis time) was found to be significantly shorter in T2D than in control subjects (P < 0.01). These results are in favor of a general impairment of the redox status in T2D patients. This is in line with the reduced plasma thiol level in T2D patients compared with control subjects (P < 0.01) (Table 1).

Albumin was isolated from the plasma of control subjects and T2D patients. HSA fractions obtained for six control subjects and six T2D patients were subjected to two-dimensional SDS-PAGE, and the corresponding Coomassie-stained spots were analyzed by proteomics. After digestion with trypsin and extraction from gel pieces, peptide fragments were identified and quantitated after MALDI-TOF/TOF analysis. PCA enabled to show a clear separation between the two groups (Supplementary Fig. 1). SPLS (45) aimed to make a regression on components retaining only peaks contributing the most to discrimination. This method was implemented on the data, and 20 peaks were sufficient to separate the two groups (Fig. 1A). A detailed comparative analysis of the most important peaks of trypsin-treated HSA, highlighted by SPLS, indicated that intensities differed between the two groups, suggesting differently modified residues (especially Lys) (Supplementary Fig. 2). From these data, it can be concluded that albumin fractions isolated from the plasma of control subjects and T2D patients have different structural characteristics.

HSA fractions were also specifically assayed for their thiol content and antioxidant properties using a free radical–induced hemolysis test. We found a significantly lower level of total plasma thiols (−28%, P < 0.01), especially with a lower level of albumin-associated free thiols (−29%, P < 0.01), as well as shorter HT50 times (−30%, P < 0.01) in T2D patients than in control subjects (Table 1). These data illustrate the reduced antioxidant capacity of T2D albumin compared with control albumin. The fatty acid binding capacity of albumin from control subjects and T2D patients was evaluated through incubations with increasing concentrations of oleic acid. As shown in Fig. 1B, the B_max of isolated albumin was significantly lower in T2D than in control subjects (−32%, P < 0.001). The half-saturation value (apparent Sat_50%) of albumin was reached with a lower concentration of oleic acid for T2D patients than for control subjects (4.61 ± 1.05 vs. 6.60 ± 1.08 µmol/L, respectively; P < 0.001).

Taken together, these data come in support of reduced saturation of T2D albumin in the presence of oleic acid and suggest that the capacity of T2D albumin to bind unesterified fatty acids is blunted. This finding comes in addition to the lower albumin levels in T2D patients than in control subjects (Table 1).

N-HSA was treated in vitro with either 25 mmol/L glucose (G25-HSA) or 2 mmol/L MGO (MGO-HSA). The experimental conditions were identical to those previously described and specifically set up to reproduce in vivo modifications resulting from hyperglycemia (14). As expected from earlier data for glycation (14) and as illustrated in Fig. 2, changes in Trp emission fluorescence spectra were observed and plots of quenching experiments were drawn. The almost linear nature of the plots indicated that the quenching process followed a collision-type mechanism. The Stern-Volmer constants were calculated and showed significant differences between native and modified albumin (32.96 ± 1.25, 25.37 ± 1.26, and 19.84 ± 1.09 Ksv values for N-HSA, G25-HSA, and MGO-HSA, respectively, P < 0.01). As shown on spectra and Ksv plots (Fig. 2), albumin preparations from T2D patients (Diab-Alb) were found to behave similarly to MGO-HSA and G25-HSA, while, in contrast, albumin from control subjects did not differ from N-Alb.

We further aimed to characterize the structural modifications induced in albumin under our experimental conditions, as well as the putative changes in the properties of albumin. By using 5,5′-dithiobis, 2-nitrobenzoic acid, we found that fewer thiol groups were available in G25-BSA and MGO-BSA than in N-BSA (−59 and −31%, respectively, P < 0.01) (Table 2). In addition, there were significantly fewer free Lys-NH₂ groups in BSA incubated with 25 mmol/L glucose and BSA treated with 2 mmol/L MGO than in N-Alb (−29 and –33%, respectively, P < 0.01) (Table 2). As revealed by PAGE, our conditions did not result in extensive damage to the protein, with no detectable fragmentation or aggregation (data not shown). ANSA fluorescence data indicated that alterations in albumin conformation led to major decreases in the accessibility of hydrophobic domains. These observations were similar for MGO- and G25-BSA (Fig. 3A).

As it is well known that albumin binds and transports NEFAs, our results concerning ANSA binding prompted us to specifically investigate the interactions of fatty acids with regard to albumin modifications. After the incubation of distinct fatty acid–free albumin preparations with increasing amounts of oleic acid, maximal absorbance was measured in the 280–300 nm range according to previous works (43). Under our experimental conditions, we found a gradual increase in the absorbance at ∼292 nm, which could be fitted according to a binding hyperbole (Fig. 3B). The apparent K_d value calculated with N-BSA was 1.6 ± 0.4 μmol/L, whereas it was up to 4.5 times higher with modified albumin (P < 0.001) (Fig. 3B). These results indicate that, because of the modifications, albumin lost part of its affinity for NEFAs, since oleate binding is considered a good representative of other fatty acids (47). Based on curves obtained in Fig. 3B and compared with N-Alb, we found that albumin modified with G25 and MGO had a significantly decreased B_max for oleate.

Arachidonate-labeled platelets were used to monitor the effects of the arachidonate pathway on thrombin stimulation in the presence of increasing concentrations of N-BSA or modified BSA. We chose to use rat platelets because of their known easy handling that confer a better reproducibility than human platelets, especially for long-term incubation (32,33,48). Under our experimental conditions, most of the labeling material was recovered in the platelet phospholipids, with no difference in the relative distribution of the phospholipid fractions whether albumin was added or not (not shown). The thrombin-induced mobilization of radiolabeled AA preincorporated into phospholipids and the subsequent formation of labeled cyclooxygenase (COX) and lipoxygenase (LOX) products were measured. After only 2 min of stimulation, labeled AA was released from the phospholipids, and the main metabolites (i.e., the stable derivatives of thromboxane A₂, TxB₂, 12-hydroxy-heptadecanoic acid [12-HHT], and 12-hydroxy-eicosatetraenoic acid [12-HETE]) were detected in platelet extracts. The production of active metabolites of the COX pathway (HHT and TxB₂) and of the LOX pathway (HETE) was significantly higher with modified albumin (G25-BSA and MGO-BSA) than with N-BSA (Fig. 4).

These observations suggest that platelet-released arachidonate is sequestered by N-Alb and thus becomes less available for conversion into active metabolites by COX and LOX enzymes. In contrast, with glycoxidized albumin, higher amounts of phospholipid-derived arachidonate are available for conversion into oxidized derivatives, and concordant observations were made with G25-BSA and MGO-BSA.

First, the appropriate conditions to investigate the effect of albumin on platelet aggregation were set up. The results illustrated in Fig. 5A clearly indicate that increased concentrations of N-BSA led to an increased inhibitory effect of albumin on platelet aggregation (as induced by various agonists such as thrombin, ADP, calcium ionophore A23187, and arachidonate). In addition, the spontaneous adhesion of platelets to the microplates was found to be gradually inhibited when the microplates were coated with increasing amounts of N-BSA (Fig. 5B). In contrast, and in agreement with earlier observations (35), coating with collagen resulted in platelet adhesion as uncoated wells (Fig. 5B). To investigate further the inhibitory properties of albumin on platelet signaling, we also analyzed TxB₂ release by means of an ELISA and calcium fluxes in the presence of radiocalcium. Thrombin-induced TxB₂ was significantly reduced and in a concentration-dependent manner in the presence of increasing concentrations of N-BSA (Fig. 5C1). The basal calcium uptake, measured without thrombin, was not different whether N-BSA was added or not (data not shown). In contrast, the thrombin-induced calcium influx was significantly reduced with increasing concentrations of N-BSA (Fig. 5C2). To analyze whether N-BSA influenced platelet oxidative metabolism, platelets were probed with DCFH-DA. N-BSA significantly decreased arachidonate-mediated DCF production in a dose-dependent way (Fig. 5C3).

Again, albumin was modified by incubation with 25 mmol/L glucose (G25-BSA) and MGO (MGO-BSA), and preparations were tested for their effects on platelet function in comparison with N-BSA at identical final concentrations (2 mg/mL). The results for thrombin-induced platelet aggregation and platelet adhesion are shown in Fig. 6. We found that G25-BSA and MGO-BSA significantly lost their inhibitory capacity on platelet aggregation compared with N-BSA (Fig. 6A). Similarly, the inhibitory effect on platelet adhesion was greatly impaired when wells were coated with modified albumin (Fig. 6B). In addition, platelet signaling (Fig. 6C and D) and ROS production (Fig. 6E) stimulated with thrombin were also impaired after incubation with modified albumin preparations as shown by calcium uptake and DCF level.

Finally, platelets were incubated with albumin fractions isolated from control subjects and patients with T2D, and they were stimulated with thrombin. As for N-HSA, which was used in control samples, albumin isolated from healthy control subjects was able to markedly inhibit the thrombin-induced aggregation of platelets (Fig. 6F). Compared with albumin isolated from healthy control subjects, albumin from patients with T2D diabetes displayed an impaired capacity to block platelet aggregation (44.0% in T2D patients vs. 80.9% in healthy control subjects). These results are in line with observations made above with modified BSA.

In the present work, albumin abnormalities resulting from glycation and glycoxidation were characterized. They were found to produce significant impairment in the binding capacity of NEFAs, as found in patients with T2D, thus contributing significantly to platelet hyperactivity in this population.

Isolated albumin fractions from T2D patients were shown here to display reduced NEFA binding capacity. When albumin was incubated in vitro with increasing concentrations of oleic acid, the saturation curves obtained for T2D albumin were significantly different from those for albumin from control volunteers, as characterized by reduced half saturation and decreased B_max of NEFAs. Importantly, this impaired ability to bind NEFAs may lead to several biological consequences. First, relative amounts of unbound NEFAs may increase, thus raising their potential adverse effects through interactions with lipoproteins and cells. Second, NEFA binding and albumin oxidation of the thiol group are known to be intimately linked, and the more NEFAs bound to albumin, the more Cys³⁴ was oxidized (49). In comparison with albumin from control subjects, that from T2D patients displayed reduced levels of free thiols and Lys-NH₂ and decreased antioxidant activity (14,42). The present data are in line with an increase in the overall oxidative status of patients with diabetes as assessed by increased plasma TBARS and oxidized LDL (50,51). Significant structural differences were found between isolated albumin of control subjects and that of T2D patients as assessed by proteomics. Furthermore, conformation analyses of albumin from T2D patients carried out with an ANSA fluorescent probe or Trp fluorescence revealed similarities with albumin modified in vitro by glycation and MGO. Most importantly, the modifications of N-Alb obtained here in vitro led to drastic decreases in the accessibility of hydrophobic domains that constitute NEFA-binding sites. These data indicate that the abnormalities of T2D albumin can be reproduced in vitro. Although the current study involved reconstituted experimental media, both albumin and NEFAS were used in relative proportions that were similar to those found in vivo (in both cases, approximately one-twentieth of normal biological concentrations). In this respect, it could be considered of potential clinical relevance.

As interactions between NEFAs and platelets are of major importance in hemostasis, the effects of native and modified albumin on platelet function were compared. NEFAs, including AA, are well-known platelet stimulators (52). When platelets are activated, large amounts of AA are released from membrane phospholipids through the lipolytic activity of phospholipase A2, and AA is the initial substrate for the subsequent biosynthesis of prostanoids, hydoxyacids, and thromboxanes produced by LOX and COX (21,22). In the present work, we showed that platelets treated with Glc-Alb and MGO-Alb released significantly higher amounts of AA metabolites (HHT, HETE, and TxB₂) than did platelets incubated with N-Alb or those incubated in the absence of albumin. Albumin isolated from T2D patients presented a higher thrombin-induced aggregation than the albumin fraction of control subjects. Concomitantly, platelet adhesion and thrombin-induced aggregation were blocked by N-Alb but to a much lower extent by Glc-Alb and MGO-Alb. These findings support the hypothesis that the NEFA-binding property of albumin is essential to sequester platelet-derived NEFAs. This mechanism occurs to a significantly lower extent when the albumin has undergone glycoxidation. Beyond its role in the systemic, intravascular transport of adipocyte-derived NEFAs toward the liver (53), albumin also seems to be a key component in minimizing the available amounts of the AA precursor in the vicinity of platelets. This provides a new explanation for the known hyperreactivity of washed platelets, namely, platelets resuspended in buffer in the absence of plasma components (54). Given the impaired redox status that occurs in a context of hyperglycemia (55), the conditions required to obtain such modifications in the albumin molecule are likely to be encountered in T2D. In support of this view, we found in the current study that T2D patients cumulated the two conditions that are potentially harmful in terms of NEFA handling, i.e., concomitant reductions in both the concentration and the NEFA binding capacity of albumin. A number of pathological states, including lipid disorders, nephrotic syndrome, malnutrition, obesity, and metabolic syndrome in addition to diabetes, are associated with high thrombotic risk (22). The current study strongly suggests that, as well as hypoalbuminemia, which is known to occur in these pathological states (56,57), an abnormal reduction in the NEFA binding capacity of albumin might well contribute significantly to the disorders.

A high level of circulating NEFAs is predictive of sudden death, stroke, and ischemic heart disease as reported by a large number of population studies (10,58,59). The numerous deleterious effects such as cardiac arrhythmia and proinflammatory and prothrombotic states have been the subject of intense research, which has shown that some of the deleterious effects could be directly attributed to elevated plasma NEFAs (12,60). However, though the effects of high levels of total plasma NEFAs have been studied (10,11,61), the putative deterioration in the NEFA binding capacity of albumin was not addressed in a systematic way. In previous studies, it was reported that oxidative stress and glycation caused conformational changes in albumin structure (14,29,62) with possible effects in vitro on platelet function (17,63,64). The novelty of the current study is the demonstration that glycation probably impairs the binding properties of albumin of patients with T2D. When the NEFA binding capacity of albumin is blunted by glycoxidation, higher amounts of free arachidonate are available for the generation of prothrombotic oxidized metabolites in platelets, which become hyperactive.

The current study presents some limitations, in particular the enrollment of a relatively low number of patients with T2D from a single center. In addition, the patients had diabetes that was poorly controlled, with high HbA_1c levels, thus limiting the conclusions of the study to a transient time in the life of most patients with T2D. At this time, one cannot assess whether the observed changes also apply to patients with type 1 diabetes and whether they are due to insulin resistance, glycemia, BMI, or age. The present work constitutes a first line of investigation and warrants additional exploration in larger and specifically designed clinical trials to confirm that higher amounts of free arachidonate are made available for the generation of active metabolites in platelets when the NEFA binding capacity of albumin is blunted by glycoxidation.

Acknowledgments. The authors thank Philip Bastable (Research Unit, Dijon Centre Hospitalier Universitaire) for manuscript editing and D. Souyhel and N. Loreau for their help in this study.

Funding. This work was supported in part by INSERM, Nouvelle Société Française d’Athérosclérose, Conseil Régional de Bourgogne, Université de Bourgogne, and a French government grant managed by the French National Research Agency under the program Investissements d’Avenir (ANR-11-LABX-0021-LipSTIC).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. D.B. researched data and drafted, reviewed, and edited the manuscript. E.B. researched data and reviewed and edited the manuscript. P.S., G.L., and P.D. researched data for proteomics and statistics and reviewed and edited the manuscript. J.-M.P. and B.V. took part in providing clinical data and recruitment and reviewed and edited the manuscript. L.L. contributed to data interpretation and reviewed and edited the manuscript. D.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.