Two novel, putative mechanisms of action for citalopram-induced platelet inhibition

Citalopram, a selective serotonin reuptake inhibitor (SSRI), inhibits platelet function in vitro. We have previously shown that this action is independent of citalopram’s ability to block serotonin uptake by the serotonin transporter and must therefore be mediated via distinct pharmacological mechanisms. We now report evidence for two novel and putative mechanisms of citalopram-induced platelet inhibition. Firstly, in platelets, citalopram blocked U46619-induced Rap1 activation and subsequent platelet aggregation, but failed to inhibit U46619-induced increases in cytosolic Ca2+. Similarly, in neutrophils, citalopram inhibited Rap1 activation and downstream functions but failed to block PAF-induced Ca2+ mobilisation. In a cell-free system, citalopram also reduced CalDAG-GEFI-mediated nucleotide exchange on Rap1B. Secondly, the binding of anti-GPVI antibodies to resting platelets was inhibited by citalopram. Furthermore, citalopram-induced inhibition of GPVI-mediated platelet aggregation was instantaneous, reversible and displayed competitive characteristics, suggesting that these effects were not caused by a reduction in GPVI surface expression, but by simple competitive binding. In conclusion, we propose two novel, putative and distinct inhibitory mechanisms of action for citalopram: (1) inhibition of CalDAG-GEFI/Rap1 signalling, and (2) competitive antagonism of GPVI in platelets. These findings may aid in the development of novel inhibitors of CalDAG-GEFI/Rap1-dependent nucleotide exchange and novel GPVI antagonists.

to the small GTPase Rap1, a process catalysed by the calcium and diacylglycerol guanine nucleotide exchange factor-1 (CalDAG-GEFI, also known as RAS guanyl-releasing protein 2) [15][16][17] . Rap1-GTP mediates the transition of integrin α IIb β 3 (also known as glycoprotein (GP) IIb/IIIa) to a high-affinity state, thereby facilitating platelet aggregation through fibrinogen crosslinks 18,19 . Therefore, our search for novel mechanisms for citalopram focussed on GPVI and Ca 2+ signalling. We utilised the selective GPVI agonist, collagen-related peptide (CRP), and also investigated Ca 2+ signalling in neutrophils, which share similar Ca 2+ -dependent mechanisms of activation 20,21 and are also inhibited by SSRIs in vitro 22 . Our data reveal two novel, putative inhibitory mechanisms of action for citalopram.
To confirm that Fura-2-loading had no functional effect on platelets, on one occasion, U46619-induced aggregation was measured in the same preparation of Fura-2-loaded platelets used for Ca 2+ measurements. Fura-2-loaded platelets aggregated normally in response to 0.2 µM U46619 (Fig. 1e) and citalopram (200 µM) inhibited this response, as expected. However, as reported above, the increase in [Ca 2+ ] cyt was unaffected by citalopram.
Thus, in citalopram-treated platelets, U46619 induced a normal Ca 2+ response, but no aggregation, suggesting that citalopram inhibits aggregation at a point downstream of Ca 2+ mobilisation. This conclusion was supported by the observation that citalopram also inhibited platelet aggregation induced by the Ca 2+ ionophore, ionomycin (0.5 µM), in a concentration-dependent manner: pIC 50 = 3.98 ± 0.09 (N = 4 blood donors) ( Supplementary  Fig. S2).
Citalopram inhibits Rap1 activation in platelets. U46619-induced platelet activation was inhibited by citalopram, despite preserved Ca 2+ store release (Fig. 1). We therefore aimed to identify where citalopram exerts its inhibitory effects downstream of Ca 2+ release.
Further experiments were conducted to investigate whether citalopram suppresses CalDAG-GEFI-mediated nucleotide exchange of isolated Rap1, using purified recombinant CalDAG-GEFI and Rap1B, the predominant Rap1 isotype in platelets 17,23 . Nucleotide exchange was monitored by detecting fluorescent BODIPY-FL-conjugated GDP as described in the Methods. Citalopram inhibited CalDAG-GEFI-mediated BODIPY-FL-GDP exchange onto Rap1B in a concentration-dependent manner (Fig. 2b,c). Peak increases in fluorescence intensity (ΔF.I.) were fitted to the four-parameter logistic (4PL) model, with the Max parameter constrained to the basal response, indicated by the ΔF.I. observed when no CalDAG-GEFI was added. The pIC 50 value was 3.67 ± 0.32 (N = 4 experiments).
Citalopram inhibits Rap1 activation in neutrophils. The results above indicate that citalopram inhibits CalDAG-GEFI-dependent Rap1B nucleotide exchange and imply that other cells expressing CalDAG-GEFI/Rap1 would also be inhibited by citalopram. Similarly to platelets, neutrophils express both CalDAG-GEFI and Rap1 (the predominant form is also Rap1B 24 ), which mediate agonist and Ca 2+ -dependent activation 21,25 . Signalling and functional studies were therefore conducted on isolated neutrophils to determine the effects of citalopram treatment.
Citalopram inhibits neutrophil function. Rap1 regulates the transition of α M β 2 integrin (Mac-1, CD11b/18) to a high-affinity binding state in macrophages 26 . α M β 2 is a cell surface adhesion receptor for fibrinogen 27 and CalDAG-GEFI-deficient neutrophils stimulated with PAF show impaired α M β 2 -dependent adhesion to fibrinogen 21 . Experiments were performed to determine if citalopram could inhibit activation of neutrophil integrin α M β 2 . Neutrophils were pre-incubated with citalopram (0, 5, 10, 20, 50, 100, 200 & 500 µM) for approximately 5 min, followed by PAF stimulation (1 µM). Representative flow cytometry histograms show that citalopram inhibited Inhibition of neutrophil signalling and function by citalopram closely matched that observed in platelets. Therefore, these data support the hypothesis that citalopram inhibits CalDAG-GEFI-dependent Rap1 nucleotide exchange. MECHANISM 2: GPVI antagonism. Inhibition of CRPXL-induced Ca 2+ signalling by citalopram ( Fig. 1a,b) is consistent with our previous observations that GPVI-mediated tyrosine phosphorylation of PLCγ2 is inhibited by citalopram 10 . It also suggests that citalopram has a mechanism of action distinct from the inhibition of CalDAG-GEFI-dependent Rap1 nucleotide exchange. Given that citalopram also reduces phosphorylation of FcRγ chain and Src family kinases (SFKs) 10 we hypothesised that citalopram may have a direct effect on GPVI structure and/or function. Citalopram inhibits the binding of GPVI antibodies. GPVI is expressed on the surface of platelets in both monomeric and dimeric conformations, although the dimeric form is thought to be particularly important in collagen binding and subsequent platelet activation 13,28 . We hypothesised that citalopram may disrupt the dimeric structure of GPVI, thereby preventing collagen-and CRPXL-induced responses. Experiments were conducted using antibodies that selectively detect either dimeric GPVI (204-11 Fab fragments) or total (dimeric and monomeric) GPVI (HY-101) to determine whether citalopram altered GPVI-dimer expression. Citalopram reduced the fluorescence intensity (F.I.) of unstimulated platelets labelled with 204-11 Fab fragments and HY-101 antibodies in a concentration-dependent manner (Fig. 4). M.F.I. from platelet samples were fitted to the 4PL model, with the Max parameter constrained to the F.I. of the isotype control (Fig. 4b,d: pIC 50 (204-11) = 4.16 ± 0.03; pIC 50 (HY-101) = 3.93 ± 0.07 (N = 6 blood donors)). These data suggest that citalopram either reduces total GPVI surface expression or blocks the binding of GPVI antibodies to GPVI. The reduction in HY-101 antibody binding suggests that the effect of citalopram is not specific to dimeric GPVI.
Citalopram-induced inhibition of GPVI-mediated platelet activation is fully reversible. We next investigated whether the impaired GPVI antibody binding caused by citalopram was due to a functionally irreversible mechanism of action such as receptor shedding or internalisation. Platelets were pre-incubated for approximately 5 min with citalopram (0 & 100 µM), which was subsequently removed by pelleting and resuspending platelets in fresh calcium-free Tyrode's (CFT) containing no citalopram (Fig. 5a). Platelets were then stimulated with CRPXL, with or without citalopram, under standard aggregometry conditions. As expected, citalopram (100 µM) inhibited CRPXL-induced platelet aggregation (Fig. 5c,d). Resuspension of citalopram-treated platelets in fresh CFT restored CRPXL-induced aggregation and the resuspended control and Citalopram rapidly inhibits CRPXL-induced platelet aggregation in a competitive manner. As shown above, citalopram inhibited the binding of anti-GPVI antibodies and platelet stimulation by CRPXL. However, this latter effect was fully reversible, suggesting that citalopram may bind reversibly to GPVI, thereby preventing binding of the anti-GPVI antibodies and CRP. Such a mechanism would be rapid in onset and competitive in character. We therefore performed additional experiments to investigate the kinetics of onset of platelet inhibition by citalopram, and whether it exhibited a competitive or non-competitive pattern of inhibition.
CRPXL-induced (1 µg mL −1 ) platelet aggregation was completely inhibited by citalopram (100 µM) following either short pre-incubation times (30, 60 seconds) or on simultaneous addition with CRPXL (i.e., 0 seconds pre-incubation) (Fig. 6a,b). The same result was observed with collagen as an agonist ( Supplementary  Fig. S6). Following 5 min pre-incubations, citalopram inhibited CRPXL-induced platelet aggregation in a concentration-dependent manner (Fig. 6c,d). Pre-incubating platelets with 20 µM and 50 µM citalopram caused 2.1-fold and 5.3-fold rightward shifts in agonist-response curves, respectively (Fig. 6d). Schild analysis  (Fig. 6d). Inclusion of this concentration into the Schild analysis increased the Schild slope to 1.60 ± 0.05 (Fig. 6e). In three experiments, at 200 µM citalopram, there was no response to CRPXL (highest concentration tested = 20 µg mL −1 ). These data suggest that at concentrations up to approximately 50 µM, citalopram inhibits CRPXL-induced platelet aggregation in a manner consistent with a competitive mechanism of action. Above this concentration, this pattern breaks down as may be predicted since, at these higher concentrations, citalopram will also exert inhibitory effects via its action on CalDAG-GEFI/Rap1.

Discussion
We have previously shown that citalopram-induced inhibition of platelet function is not caused by blockade of SERT-dependent 5-HT uptake into platelets 10 . The aim of this study was to identify putative SERT-independent mechanisms of platelet inhibition by citalopram. Specifically, we have characterised the effects of citalopram on two distinct processes: (1) Rap1 activation and (2) GPVI receptor function. In platelets and neutrophils, activation of both TP and PAF receptors respectively induces Ca 2+ release from intracellular stores via G protein-mediated activation of phospholipase Cβ (PLCβ) [29][30][31] . In both cell types, citalopram failed to inhibit either U46619-or PAF-induced Ca 2+ release indicating that the signalling pathways from receptor to elevated [Ca 2+ ] cyt were unaffected by the drug. By contrast, citalopram did block downstream Rap1 activation and cell function. Similarly, Tseng et al. 14 reported that ADP-induced platelet aggregation was inhibited by citalopram, but not ADP-induced Ca 2+ signalling. Notably, Ca 2+ -dependent Rap1 activation in both platelets and neutrophils is mediated by CalDAG-GEFI 15,21 . Our results from an in vitro fluorescence-based binding assay show citalopram inhibits CalDAG-GEFI-mediated nucleotide exchange of Rap1B. The recovery of CRPXL-, collagen-and U46619-induced aggregation after washing out citalopram (Fig. 5, Supplementary Figs S4 and S5) indicates that this inhibition by citalopram is reversible. We therefore propose that citalopram binds directly and reversibly to either CalDAG-GEFI, Rap1 or a complex of both, thereby inhibiting Rap1 activation.
Comparatively few studies have reported the in vitro effects of SSRIs on neutrophils. Although fluoxetine has previously been shown to inhibit some neutrophil functions 22 , we believe that ours is the first report of citalopram inhibiting human neutrophil function. Unlike platelets, neutrophils do not express SERT 32 . Therefore, our results provide further confirmation of a direct and SERT-independent mechanism of action of citalopram in neutrophils and, by extension, platelets.
We have previously reported that citalopram inhibited collagen-induced aggregation and phosphorylation of molecules in the GPVI signalling pathway 10 . We now report that citalopram also inhibits platelet aggregation induced by CRPXL, a GPVI-selective agonist, and reduces the binding of anti-GPVI antibodies to unstimulated platelets. One possible explanation for these results is a reduction in surface receptor number, either by shedding or internalisation. However, for a full agonist, a reduction in receptor number is predicted to reduce the observed potency of the agonist 33 , and this has previously been demonstrated for CRPXL-induced aggregation in platelets with 50% levels of GPVI 34 . Thus, the similarity in CRPXL responsiveness of untreated resuspended platelets (Fig. 5f, condition (3)) and citalopram-pre-treated resuspended platelets (Fig. 5f, condition (4)) suggests that little if any GPVI was lost from the platelet surface as a result of citalopram treatment. Moreover, our data show that inhibition of CRPXL-induced platelet aggregation by citalopram is both instantaneous in onset and fully reversible. Taken together, these data strongly support a reversible, competitive mechanism of action for citalopram, rather than a reduction in surface receptor expression. We therefore propose that citalopram binds directly to GPVI-FcRγ chain complex, thereby preventing collagen-and CRPXL-induced platelet activation.
Our proposal that citalopram exerts two distinct mechanisms of action is further supported by the observed inhibitory potencies of citalopram in our studies. The Schild analysis indicates that citalopram binds to GPVI/FcRγ chain with a K d of approximately 16 µM. This is wholly consistent with data reported in our previous study 10 showing that 20 µM citalopram caused an approximate 2-fold rightward shift of the collagen concentration-response curve but had no discernible effect on U46619-induced aggregation. pIC 50 values for inhibition of GPVI-independent functions: aggregation induced by U46619 (4.15 ± 0.27) and ionomycin (3.98 ± 0.09); PAF-induced activation of neutrophil α M β 2 (4.02 ± 0.15); adhesion of platelets (4.00 ± 0.07) and neutrophils (3.88 ± 0.04) to fibrinogen; and CalDAG-GEFI-dependent Rap1B activation (3.67 ± 0.32), are all consistent with citalopram binding to and inhibiting CalDAG-GEFI/Rap1B at a concentration of approximately 100 µM. This is further reflected by the Schild analysis showing a rightward shift in the CRPXL concentration-response curves consistent with competitive antagonism at citalopram concentrations up to 50 µM, whereas at higher concentrations the shift was greater (Fig. 6) caused by the combination of the two distinct inhibitory mechanisms outlined above.
Citalopram may inhibit platelets and neutrophils through other unidentified mechanisms that are distinct from SERT blockade. For example, Bonnin et al., (2012) 35 have proposed that (R)-citalopram, the lower potency isomer 36 , may act via the orphan sigma-1 receptor. However, as we have previously noted, this is unlikely to be the mechanism responsible for the action of citalopram in platelets 10 .
In summary, we propose a model (Fig. 7) in which citalopram binds to two distinct molecular targets: (1) GPVI/FcRγ chain (K d ≈ 16 µM) and (2) CalDAG-GEFI/Rap1B (K d ≈ 100 µM). This model predicts that citalopram would selectively disrupt GPVI-dependent platelet activation at concentrations between 20 and 50 µM, and above 50 µM, it would also inhibit Ca 2+ -dependent functions mediated through CalDAG-GEFI. These two novel, putative and distinct inhibitory mechanisms of action: (1) competitive antagonism of GPVI-FcRγ chain inhibition and (2), inhibition of CalDAG-GEFI-mediated nucleotide exchange of Rap1B, both need much higher concentrations of citalopram than are required to inhibit SERT, its primary mechanism of action. Hence, these effects are unlikely to be of clinical significance 10 , either as a cause of reported bleeding complications [37][38][39][40][41] , or as a strategy for reducing cardiovascular disease. Further studies will be required to confirm these putative mechanisms. However, if confirmed, citalopram may prove to be a useful investigative tool for the study of CalDAG-GEFI, Rap1, and GPVI signalling, as well as a practical chemical starting point for the discovery of more selective and potent inhibitors. A potent, selective GPVI antagonist could be a potentially useful anti-thrombotic agent and inhibitors of CalDAG-GEFI/Rap1 may have a wide range of uses in haematopoietic cells. Washed platelet preparation. Citrated blood was centrifuged (500 × g, 5 min) to obtain platelet-rich plasma (PRP). Following addition of PGE 1 (final concentration of 1 μM), PRP was centrifuged (900 × g, 15 min) and the resulting platelet pellet resuspended in a modified calcium-free Tyrode's buffer (CFT; 137 mM NaCl, 11.9 mM NaHCO 3 , 0.4 mM NaH 2 PO 4 , 2.7 mM KCl, 1.1 mM MgCl 2 , 5.6 mM glucose; pH = 7.4). Platelet counts were adjusted to 2 × 10 8 mL −1 using a Z2 Coulter particle counter (Beckman Coulter, High Wycombe, U.K.).
Neutrophil preparation. Citrated blood was mixed 2:1 with a saline solution of dextran-500 (final concentration of 1% [w/v]) and left for 30 min to allow red blood cell (RBC) sedimentation, whilst retaining white blood cells (WBCs) within the PRP. WBC-rich PRP was aspirated and layered over a discontinuous density gradient of Percoll ® (1.5 mL of 1.088 g mL −1 Percoll ® , carefully layered on top of 1.5 mL of 1.100 g mL −1 Percoll ® ). Samples were centrifuged (600 × g, 20 min) to separate granulocytes from the lower-density platelets, lymphocytes and monocytes. The isolated granulocyte band was aspirated, washed with phosphate-buffered saline (PBS), centrifuged (300 × g, 5 min) and resuspended in CFT. The cell concentration was adjusted to 1 × 10 6 mL −1 using a Z2 Coulter particle counter (Beckman Coulter, High Wycombe, U.K.).
Platelet aggregometry. Platelet aggregation was measured by turbidimetric aggregometry as previously described 10 . Samples were centrifuged (8,000 × g, 1 min) and washed 3 times with 400 µL LBW, before the addition of 50 µL Laemmli buffer. Samples were then centrifuged (8,000 × g, 2 min) through the spin cups to obtain bead-free samples for Western blot analysis. Total Rap1 and Rap1-GTP samples underwent SDS PAGE and Western blot analysis. Samples were added to 10 well 4-12% pre-cast NuPage Bis-Tris gels (Invitrogen, Paisley, U.K.), followed by SDS/PAGE separation and transfer to PVDF membranes (Millipore, Watford, U.K.). Membranes were incubated with Rap1 primary antibodies (Thermo Fisher Scientific, Loughborough, U.K) followed by incubation with HRP-conjugated secondary antibodies (Dako, Ely, U.K.). Enhanced chemiluminescence (ECL) and X-ray hyperfilm ® (Amersham Biosciences, Buckinghamshire, U.K.) were used to detect protein bands. Developed X-ray films were scanned, and unprocessed images analysed using ImageJ (v1.50) as follows: identical areas (height (100) × width (150) = 15,000 pixels) were drawn around each protein band and the density of all pixels (scale 0-255 each) summed. The density of each area was therefore quantified on a scale from 0 (totally white) to 3,825,000 (totally black). Values are presented as % black. Uncropped images of X-ray films used for quantification of Rap1-GTP are shown in Supplementary Figs S7  Rap1B and CalDAG-GEFI were cloned from human genes into a protein expression vector p15LIC2 6xHis, which was purified in E. coli, as previously described 17  (CD11b). Neutrophils (100 µL, 1 × 10 6 mL −1 ) were incubated with the active CD11b antibody in the absence of light (5 min, 4 °C). Samples were fixed with 2% [v/v] paraformaldehyde and neutrophils gated by forward scatter (FSC), side scatter (SSC) and CD15 + criteria. The median fluorescence intensity (M.F.I.) of activated CD11b from 30,000 gated events was then determined for each sample using an Accuri TM C6 flow cytometer (BD Bioscience, Oxford, U.K.).
Lactate dehydrogenase cytotoxicity assay. LDH release from neutrophils was measured to determine drug-induced cytotoxicity and cytolysis, using a Pierce LDH Activity Assay Kit (Thermo Fisher Scientific, Loughborough, U.K.). Neutrophils (250 µL, 1 × 10 6 mL −1 ) were centrifuged (8,000 × g, 1 min) and 50 µL of supernatant was aliquoted into wells of an Immulon-2HB 96-well flat-bottom plate. 50 µL of the proprietary reaction mixture (Thermo Fisher, product code: 1862887) was added to each well for 30 min. 50 µL of the proprietary stop solution (Thermo Fisher, product code: 1862880) terminated the reaction and background absorbance at 680 nm was subtracted from the absorbance at 490 nm. Measurements were made using a Sunrise TM plate reader (Tecan, Reading, U.K.).
Data and statistical analysis. Concentration-response data were modelled using a four-parameter logistic (4PL) model 43