Volatile Anesthetics, Not Intravenous Anesthetic Propofol Bind to and Attenuate the Activation of Platelet Receptor Integrin αIIbβ3

Background In clinical reports, the usage of isoflurane and sevoflurane was associated with more surgical field bleeding in endoscopic sinus surgeries as compared to propofol. The activation of platelet receptor αIIbβ3 is a crucial event for platelet aggregation and clot stability. Here we studied the effect of isoflurane, sevoflurane, and propofol on the activation of αIIbβ3. Methods The effect of anesthetics on the activation of αIIbβ3 was probed using the activation sensitive antibody PAC-1 in both cell-based (platelets and αIIbβ3 transfectants) and cell-free assays. The binding sites of isoflurane on αIIbβ3 were explored using photoactivatable isoflurane (azi-isoflurane). The functional implication of revealed isoflurane binding sites were studied using alanine-scanning mutagenesis. Results Isoflurane and sevoflurane diminished the binding of PAC-1 to wild-type αIIbβ3 transfectants, but not to the high-affinity mutant, β3-N305T. Both anesthetics also impaired PAC-1 binding in a cell-free assay. In contrast, propofol did not affect the activation of αIIbβ3. Residues adducted by azi-isoflurane were near the calcium binding site (an important regulatory site termed SyMBS) just outside of the ligand binding site. The mutagenesis experiments demonstrated that these adducted residues were important in regulating integrin activation. Conclusions Isoflurane and sevoflurane, but not propofol, impaired the activation of αIIbβ3. Azi-isoflurane binds to the regulatory site of integrin αIIbβ3, thereby suggesting that isoflurane blocks ligand binding of αIIbβ3 in not a competitive, but an allosteric manner.


Introduction
General anesthesia during surgery is induced and maintained by administration of inhalational (volatile) and/or intravenous anesthetic drugs. While anesthetic drugs primarily act on neuronal cells in the central nervous system [1], thereby inducing general anesthetic states, the report that halothane impairs adenosine diphosphate (ADP)-induced platelet aggregation by Ueda [2] triggered subsequent studies on the effect of hemostasis. Clinical observational investigations into the effects of anesthetics on hemostasis during surgery [3,4,5,6,7,8,9,10,11] point to an intriguing trend that intra-operative bleedings were less severe in anesthesia with the intravenous anesthetic propofol than volatile anesthetics isoflurane and sevoflurane [3,4,5,9,10] (Table 1). However, in vitro mechanistic investigations into the direct effects of propofol [12,13], isoflurane [10,12,14,15,16], and sevoflurane [12,14] [16] on platelet aggregation, a critical step in hemostasis have shown mixed results thus far.
aIIbb3 is the most abundant receptor in platelets and plays a critical role in platelet aggregation and clot stability through the interaction with its Arg-Gly-Asp (RGD)-motif -containing ligands fibrinogen, von Willebrand factor and fibronectin [17,18,19,20,21,22]. aIIbb3 is a member of the adhesion molecule family integrins, and is composed of non-covalently linked a/b heterodimers, with each subunit consisting of multiple wellcharacterized domains [23]( Figure 1A). Only upon activation, aIIbb3 undergoes the conformational changes referred to as ''the hybrid domain swing-out'', which induces the ligand binding site to the high-affinity state [24] ( Figure 1B). Three metal binding sites (metal-ion dependent adhesion site (MIDAS), SyMBS, and ADMIDAS) located on the top of the b3 I domain differentially regulate the activity of integrin aIIbb3 during this conformational change. The MIDAS directly binds to the RGD motif of ligands, while SyMBS and ADMIDAS take indirect roles in ligand binding by modulating metal coordinations at the MIDAS [24,25,26]. The study by Horn et al. demonstrated that sevoflurane, even at subanesthetic concentrations, significantly abolished the activation of aIIbb3 in whole blood [27]. Inspired by Horn et al, and building on our previous studies on the effects of volatile anesthetics to leukocyte integrins [28,29,30,31], here we tested the hypothesis that isoflurane and sevoflurane, not propofol directly interacted with platelet integrin aIIbb3and interfered with its activation.
PAC-1 binding assay using aIIbb3 transfectants CHO-K1 cells transfected with aIIbb3 were detached in HEPES-buffered saline (HBS)/10 mM EDTA and washed three times with HBS. Cells were incubated with 10 mg/ml PAC-1 (BD Biosciences) in HBS containing 1 mM MgCl 2 /CaCl 2 (inactivating condition) or HBS Containing 1 mM MnCl 2 /0.4 mM CaCl 2 (activating condition) in the presence of various concentrations of isoflurane, sevoflurane or propofol. Isoflurane and sevoflurane were administered to cells in the closed chamber using a Fluotec vaporizer, and their concentrations were measured using infrared spectroscopy. Goat anti-mouse IgM-FITC (Santa Cruz biotechnology Inc.; Santa Cruz, CA, USA) was used as a secondary antibody. Cells were analyzed with a FACScan. In addition, the cell surface expression of aIIbb3 was probed with AP3 antibody (Immune Disease Institute; Boston, MA, USA). PAC-1 binding % was calculated as [(MFI of sample at various concentrations of anesthetics)-(MFI of isotype control sample)]/(MFI of sample without anesthetics)-(MFI of isotype control sample)]6100%.

Protein expression and purification
The purification of full-length ectodomain and headpiece aIIbb3 was previously described [36,37]. Integrin aIIbb3 purified from human platelets was purchased from EMD Millipore (Billerica, MA, USA).

PAC-1 binding to the extracellular portion of aIIbb3
Capturing antibody AP3 was coated on 96 wells overnight. Wells were blocked with 2% BSA and then incubated with recombinant aIIbb3 (full length or headpiece). Following washing, wells were incubated with PAC-1 in the presence of various concentrations of isoflurane, sevoflurane or propofol containing 1 mM MgCl 2 /CaCl 2 or 1 mM MnCl 2 /0.4 mM CaCl 2 for 1 hour. Isoflurane and sevoflurane were administered to wells in Prospective, randomized study Less bleeding in propofol group [9] Head and neck surgery Isoflurane (n = 20) versus propofol (n = 18) Prospective, randomized Blood loss in isoflurane group tended to be slightly higher.

Photolabeling experiments
Photolabeling experiments were performed using azi-isoflurane, isoflurane with a diaryzinyl moiety. The details of experiment have been previously described [38,39]. Briefly, full-length ectodomain aIIbb3 or aIIbb3 purified from human platelets was incubated with or without 1 mM azi-isoflurane in quartz cuvettes for 15 minutes, and then exposed to 300 nm UV light for 15 minutes. The protein was separated using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Bands corresponding to the protein of interest were excised, trypsinized and submitted for nano liquid chromatography (LC)/mass spectrometry (MS) analysis. LC was performed using a 10 cm C18 capillary column at 200 nl/min for 60 minutes with gradient elution. MSdetected peptides were searched for adducts of the appropriate mass (196 Da) and then further fragment patterns (MS/MS) were searched using Sequest software to determine the adduct attach-ment sites. Mass spectrometry work was performed at the University of Pennsylvania Proteomics Core Facility.

Point mutagenesis and transfection
Alanine scanning mutagenesis was performed using Quikchange XL kit (Stratagene; La Jolla, CA, USA). DNA sequence was confirmed. Transfection was performed using Lipofectamine 2000 (Invitrogen; Carlsbad, CA, USA) per company protocol.

Statistical significance
Data were analyzed using an analysis of variance (ANOVA) with Tukey post hoc pairwise comparisons or student's t-test as indicated in corresponding figure legends. Statistical significance was defined as P,0.05. Statistical analysis was performed using PRISM 5 software (GraphPad Software; La Jolla, CA, USA).

Results
Isoflurane and sevoflurane, but not propofol attenuated PAC-1 epitope exposure in ADP stimulated platelets, but propofol did not From clinical observational studies on hemostasis as summarized in Table 1, we hypothesized that isoflurane and sevoflurane would attenuat the activation of integrin aIIbb3, but propofol would not. In fact, volatile anesthetic sevoflurane attenuated the activation of aIIbb3 on platelets stimulated by ADP as demonstrated by Horn et al. [27]. We demonstrated that another volatile anesthetic isoflurane at a clinically relevant concentration (2%)   40,41,42]. Propofol (50 mM) did not attenuate the activation of aIIbb3 on platelets ( Figure 2). These results supported our hypothesis. To assess the direct interaction of volatile anesthetics with aIIbb3, we examined the effect of anesthetics using CHO cells stably transfected with aIIbb3 and purified proteins in the following sections.
Isoflurane and sevoflurane attenuated the activation of wild type aIIbb3 on cells First, we evaluated the effect of anesthetics on aIIbb3 activation in CHO transfectants. PAC-1 contains the Arg-Tyr-Asp (RYD) sequence that is analogous to the RGD sequence in the complementarity determining region 3 of the heavy chain. This region is speculated to interact with the activated MIDAS [33]. We tested PAC-1 binding either in a resting condition (1 mM Mg 2+ /Ca 2+ ) or an activating condition (1 mM Mn 2+ /0.4 mM Ca 2+ ). In 1 mM Mg 2+ /Ca 2+ , PAC-1 does not binds to aIIbb3 WT ( Figure 3A). On the other hand, PAC-1 binds significantly to aIIbb3 WT in 1 mM Mn 2+ /0.4 mM Ca 2+ ( Figure 3A). Isoflurane and sevoflurane diminished PAC-1 binding to WT in 1 mM Mn 2+ /0.4 mM Ca 2+ ( Figure 3B and C), while they did not alter the expression of aIIbb3 WT ( Figure 4A). This suggested that isoflurane and sevoflurane either attenuated the activation of aIIbb3 WT or directly interacted with PAC-1 binding sites. The b3-N305T mutant was previously designed to introduce N-   . Data was shown as [(MFI of aIIbb3 exposed to anesthetic)/(MFI of aIIbb3 of sample not exposed to anesthetic)]6100%, and expressed as mean +/2 S.D. of three independent experiments. Isoflurane, sevoflurane, and propofol used were 5%, 4%, and 100 mM, respectively. doi:10.1371/journal.pone.0060415.g004 Volatile Anesthetics Inhibit Integrin aIIbb3 PLOS ONE | www.plosone.org glycosylation by changing amino acid sequences of the b3 subunit from N 303 -I 304 -N 305 to N 303 -I 304 -T 305 [32]. b3-Asn305 is located on the bottom of the I domain, at the interface with the hybrid domain. The introduction of N-glycan at this site opened up the interface between the I domain and the hybrid domain, mimicking the hybrid domain swing-out motion and making this mutant constitutively active [32] (Figure 3A). Both isoflurane and sevoflurane failed to modulate PAC-1 binding to b3-N305T mutant ( Figure 3B and C). Further, exposure to isoflurane and sevoflurane did not alter the expression level of aIIbb3 on the b3-N305T mutant ( Figure 4B), suggesting that these volatile anesthetics did not directly interact with PAC-1 binding sites on aIIbb3. Taking these results together, isoflurane and sevoflurane attenuated the activation of aIIbb3 WT. In contrast, the intravenous anesthetic propofol failed to modulate PAC-1 binding to aIIbb3 WT ( Figure 3D), indicating that propofol did not inhibit the activation of aIIbb3.

Isoflurane and sevoflurane attenuated the activation of aIIbb3 protein
We demonstrated that isoflurane and sevoflurane attenuated the activation of aIIbb3 in cell-based assays using CHO transfectants. Anesthetics are well appreciated as promiscuous molecules [43], and thus we cannot conclude if the results reflected the direct effect of the anesthetics on aIIbb3, or the indirect effect (for example, the effect on the plasma membrane or intracellular proteins). We examined the effect of anesthetics on the activation of purified aIIbb3 protein in cell-free ELISA type assay, which excluded the components of the plasma membrane and intracellular proteins. Both isoflurane and sevoflurane impaired the activation of aIIbb3 ( Figure 5A and B). Interestingly, there was no difference in the degree of inhibition between headpiece and full-length aIIbb3, suggesting that isoflurane and sevoflurane interacted with the headpiece portion ofaIIbb3. Propofol did not affect the activation of aIIbb3 in this cell-free assay ( Figure 5C), as predicted by the previous result ( Figure 3D).

Azi-isoflurane bound to the bI domain
Our cell-free and cell-based assays strongly suggested that isoflurane and sevoflurane directly interacted with aIIbb3 and attenuated its activation. Previously we reported that a novel photoactivatable compound, azi-isoflurane [38], reliably probed isoflurane binding sites on apoferritin, integrin aL I domain [39], and LFA-1 [31]. Thus, we used azi-isoflurane to reveal isoflurane binding sites on both full-length recombinant aIIbb3 and purified aIIbb3 from platelets. Azi-isoflurane bound to the I domain at Asp-158 and/or Lys-159 (Table 2, Figure 6, 7A and 7B) in both samples. Asp-158 and Lys-159 are close to the calcium binding site as shown in Figure 7B. This calcium binding site is called the synergistic metal binding site (SyMBS) or the ligand associated metal binding site (LIMBS) [26]. The adducted residues were in the headpiece region of aIIbb3, which was in line with our result of cell-free assays. The epitope mapping of PAC-1 by Puzon-McLaughlin et al. showed that they were within residues 156-162 and 229-230 of the aIIb subunit and residues 179-183 of the b3 subunit ( Figure 7C) [44]. Our adducted residues did not belong to these residues, which suggested that volatile anesthetics did not compete with PAC-1 directly as we indicated based on the results of cell-based assays. Unfortunately, a photoactive version of sevoflurane is not currently available, and we were not able to explore the binding site of sevoflurane using this approach. However, sevoflurane and isoflurane have similar physicochemistry, and we strongly suspect that sevoflurane interacts with the same site.

D158A mutant reduced the activation of aIIbb3
To confirm the functional role of the azi-isoflurane adducted residues (Asp-158 and Lys-159), we made b3-D158A and -K159A mutants to alter the chemical texture of this site. As shown in Figure 8A, b3-D158A mutant completely abolished the activation of aIIbb3 integrin in activating conditions (1 mM Mn 2+ / 0.4 mM Ca 2+ ), indicating the importance of this residue. This was consistent with the previously reported results of the mutants of the other SyMBS forming residues [26]. The b3-K159A mutant significantly diminished the cell surface expression of aIIbb3   ( Figure 8B), suggesting that Lys-159 was a critical residue for expression rather than activation.

Discussion
In this study, we demonstrated that isoflurane and sevoflurane, but not propofol, attenuated the activation of integrin aIIbb3. Furthermore, the photolabeling experiment using azi-isoflurane suggested that isoflurane bound to the residues around the SyMBS of the I domain of the b subunit. That these two findings were linked was suggested by the mutagenesis experiments, which indicated the importance of this site for expression and activation.
With the appreciation of its profound effect on platelet aggregation, aIIbb3 has been an attractive therapeutic target to prevent platelet aggregation in specific disease states. For example, peptides containing the RGD sequence competitively prevent aIIbb3 from binding to its ligands [22], and have thus served as a basis for antagonist design [17]. Currently, abciximab (Reo-Pro, Eli Lilly, Indianapolis, IN), eptifibatide (Integrelin, Cor therapeutics, Cambridge, MA) and tirofiban (Aggrastat, Merck, Whitehouse Station, NJ) are approved for clinical usage to reduce ischemic events in patients with acute coronary syndrome undergoing percutaneous coronary intervention [45,46]. When these drugs were developed, there was no structural information how these compounds interacted with aIIbb3. Now we know that the majority of aIIbb3 small molecule antagonists including eptifibatide and tirofiban bind to a small pocket on the top of the aIIbb3 head formed by loops from the aIIb b-propeller and the bI domain [24]. These compounds interact with the MIDAS Mg 2+ ion of the I domain via one of the oxygen atoms in the compound's Asp carboxyl or an equivalent carboxyl [24,25]. The exception is abciximab, the b3 specific-7E3 Fab, which blocks ligand binding by binding to residues in the specificity determining loop (SDL) [47] (Figure 7A). Macromolecules such as fibrinogen recognize a rather larger area at the interface between the b-propeller domain of the aIIb subunit and the I domain of the b subunit, and interact with the b3 SDL and aIIb b-propeller loops. Therefore, blocking SDL wth abciximab prevents fibrinogen binding [24]. Surprisingly, the adducted residues of photolabeling experiments were located around the SyMBS, which was not at the binding pocket of the aforementioned aIIbb3 small molecule antagonists and Fab. The SyMBS coordinates Ca 2+ and allosterically activates integrins for ligand binding by stabilizing the MIDAS site [36]. The side chain carboxyl of b3-Glu-220 coordinates the SyMBS Ca 2+ and MIDAS Mg 2+ at the same time [36] ( Figure 7B). Therefore, any alteration of residues surrounding the SyMBS could influence the orientation of MIDAS, and therefore aIIbb3 activation. The result of b3-D158A mutant supported this idea. Also, the SyMBS coordinates with the SDL and disruption of this interaction resulted in impaired activation, as shown by blockade of ligand binding by abciximab [48]. Thus, allosteric inhibition of activation via binding to SyMBS is feasible. However, azi-isoflurane is structurally altered from isoflurane, and it is possible that the sites reported could be different from isoflurane binding site(s). However, the crystallographically proven identity of azi-isoflurane and isoflurane protein binding sites in our previous reports argues against this possibility. Co-crystallization of isoflurane with aIIbb3 may answer this question in the future. Additionally, we cannot exclude the existence of other binding site(s) on regions of the protein that we were not able to detect using mass spectrometry.
Interestingly, we found the adducted residues only on the I domain of the b3 subunit in aIIbb3 with two different preparations. In addition to the aIIb subunit, the b3 subunit couples with the aV subunit to form integrin aVb3. The number of aVb3 copies on platelets is small compared with that of aIIbb3 [20], but aVb3is highly expressed on endothelial cells. Both aIIbb3 and aVb3integrins bind to fibrinogen, but at different sites, forming a cooperative interaction between aIIbb3 and aVb3 that allows the platelet thrombus to be anchored on the endothelium through aVb3 [49,50]. It is possible that sevoflurane and isoflurane also impair the activation of aVb3 as well to diminish the anchoring of platelets on the endothelium, which will be an additional effect to impair hemostasis by volatile anesthetics.
Clinical significance of functional alternation in aIIbb3 is apparent from a familial bleeding disease, Glanzmann thrombas- thenia. Bleeding in this disorder derives from the failure of platelet aggregation due to reduced or absent aIIbb3 [51]. Therefore, the fact that sevoflurane and isoflurane directly modulate the activation of aIIbb3 can be clinically significant. Our results are entirely consistent with this and with the previous clinical reports of endoscopic sinus surgeries. Although many studies have been performed in this surgical population, the numbers of patients enrolled in each study are not large ( Table 1). Future clinical investigation will be extremely important, particularly on cases at high risk of bleeding such as scoliosis and vascular surgeries. Since blood products are not unlimited resources and not entirely riskfree, this is an important health care consideration. The choice of anesthetic drugs may need to be considered from hemostasis standpoint.
In conclusion, we have demonstrated that the inhalational anesthetics isoflurane and sevoflurane, not but the intravenous anesthetic propofol, impairs the activation of integrin aIIbb3 via a direct novel allosteric mechanism.