Comparative proteomics reveals unexpected quantitative phosphorylation differences linked to platelet activation state

There is a need to assess platelet activation in patients with thrombotic disorders. P-selectin and activated integrin αIIbβ3 are usually quantified by flow cytometry to measure platelet activation. Monitoring changes in vasodilator-stimulated phosphoprotein (VASP) phosphorylation is an established method to determine the platelet-reactivity status. To study disruptions of platelet reactivity more comprehensively, we compared the human non-secretory platelet proteome after in-vitro -activation and –inhibition with their respective untreated controls using unbiased fluorescence two-dimensional differential in-gel electrophoresis. The non-secretory platelet proteome was more severely affected during inhibition than during activation. Strikingly, while VASP reached a 1.3-fold increase in phosphorylation levels in inhibited platelets, other protein kinase A targets showed several-fold stronger inhibition-induced phosphorylation levels, including LIM and SH3 domain protein 1 (6.7-fold), Src kinase-associated phosphoprotein 2 (4.6-fold), and Ras-related protein Rap1b (4.1-fold). Moreover, phosphorylation of integrin-linked protein kinase (ILK) and pleckstrin (PLEK) species was associated with P-selectin surface expression. The discrimination power between activation and inhibition was more pronounced for dephosphorylated ILK (3.79 Cohen’s d effect size) and phosphorylated PLEK (3.77) species than for P-selectin (2.35). These data reveal new insights into the quantitative changes of the platelet reactivity proteome and suggest powerful alternatives to characterise their activation and inactivation potential.

counteract the activation of platelets 4 . Platelet activation and thrombus formation can, therefore, be regarded as the predominance of either pro-thrombotic factors or a reduction of endogenous inhibitors 5 . An impairment 6 of platelet-inhibition mechanisms, or resistance 7 to them, may result in hyper-reactive platelets and spontaneous generation of platelet-dependent thrombosis 8,9 . Accordingly, typical platelet activation markers like P-selectin (CD62P) and activated integrin α IIb β 3 (CD41/CD61) are significantly changed under these conditions and are thus used to assess the platelet activation status [10][11][12] . However, whether increased platelet activation results from an endogenous activation or from an insufficient inhibition remains largely unknown.
To investigate how the platelet proteome is affected by activation and inhibition in a comprehensive and unbiased manner, proteomics may be applied. Several in-vitro platelet proteomic studies have shed light on the complexity of signal propagation induced upon activation with thrombin receptor activating peptide-6 (TRAP-6) 13 , ADP 14 , collagen-related peptide 15 or via the C-type lectin-like receptor 2 16 . Although an insufficient inhibition of platelets is also a significant feature in platelet-related thrombotic disorders, the first proteomic studies addressing this issue have been published just recently 17,18 . There, the dynamics of the phospho-proteome was analysed after short-term inhibition via selective stimulation of the sGC/cGMPIß (cyclic guanosine monophosphate/protein kinase GIß) pathway 18 , cyclic adenosine monophosphate/protein kinase A (cAMP/PKA) pathway 17 or with simultaneous activation by ADP 19 . Nevertheless, a direct comparison of the platelet proteome response upon both activation and inhibition might uncover differential pathophysiological signs for hemostatic disorders. Protein profiles that reflect the activation status may enable us to target platelets pharmaceutically according to the underlying perturbations and thus to identify new therapeutic options.
Therefore, the aim of the present platelet proteome study was to reveal specific features of activated, as well as inhibited platelets, compared to their untreated controls within one biochemical method using fluorescence two-dimensional differential gel electrophoresis (2D-DIGE). This proteomics technique enables reliable quantitative results on differential protein expressions by displaying thousands of proteins, their isoforms and post-translational modifications at the same time. Previously, the terminology "protein species" has emerged to describe proteins which arise from the same amino acid sequence but have been post-translationally modified. The related term "proteoform" defines distinct protein forms originating from a single gene 20 . Especially, the generation of protein species may be of great interest in the proteome of anucleate platelets during activation and inhibition in the circulation.
Importantly, the concentrations of platelet activators and inhibitors needed to address these two main regulatory conditions were selected from typical protocols of functional platelet studies [21][22][23][24] . For platelet activation TRAP-6 (15 µM), was chosen to provoke a strong PAR1-mediated activation response while ADP (5 µM) induced a moderate one. For inactivation we used the unstable PGI 2 (0.4 µM) and the more stable CTAD formulation (15 mM theophylline, 3.7 mM adenosine, 0.198 mM dipyridamole). Both inhibitory treatments especially increase cAMP to prevent platelet activation 25 .
Proteomic snapshots were taken after 15 minutes of activation/inhibition treatments to reveal long-lasting changes in platelet protein profiles, which may be useful for translation of these biomarkers into diagnostic applications. Notably, instead of the usual centrifugation step, proteins of in-vitro treated gel-filtered platelet suspension were extracted by TCA-precipitation 26 to "freeze" them in their particular activation status. It was thanks to this method that we quantified the non-secretory platelet proteome changes since the granule release was not separated by this protein extraction protocol. Using 2D-DIGE, most of the affected proteoforms were identified to be specific to the either activated or inhibited platelet status. By quantifying their abundances, some of these proteins were found to be more powerful than established biomarkers in defining platelet reactivity.

Results
Quality control of platelet in-vitro models and 2D-DIGE analysis. To assure that our platelet proteome analysis was assessed from well-defined in-vitro activated and inhibited platelets, CD62P, a surface marker specific for platelet activation, was quantified by flow cytometry. A strong platelet activation (>90% CD62Ppositive platelets) was induced by TRAP-6 (15 µM) and a moderate response was provoked by ADP (5 µM, on average 50% CD62P-positive platelets), which is comparable to published data of platelet activation studies 27,28 . Accordingly, we found a 7.36-fold and 2.53-fold increase of CD62P-positive platelets, as compared to untreated controls, for TRAP-6 (p = 0.009) and ADP (p = 0.031), respectively (Fig. 1). These increases imply that platelet functionality was maintained after isolation. For the assessment of platelet inhibition, CD62P surface expression was also evaluated after CTAD and PGI 2 (0.4 µM) treatment. Inhibited gel-filtered platelets showed a reduced CD62P positivity with 0.67-fold after CTAD-(p = 0.005) and of 0.83-fold after PGI 2 treatment (p = 0.016) compared to untreated gel-filtered platelets.
We also monitored the quality of our 2D-DIGE proteome analysis regarding the biological and technical coefficient of variation (CV tot = CV bio + CV tech ) 29 . In the present study, we determined a median CV tot of 12.3% in the untreated control platelet proteome (5,906 spots) from all included volunteers, plus a single CV tot of 5.7% for 14-3-3γ, a low biological variation protein, confirming a solid-technical quality when compared to our previous platelet proteomic studies 30 . Inhibition shows stronger changes than activation in the 2D-DIGE platelet proteome. The proteome of these activated and inhibited platelets was compared with untreated controls to gain a deeper insight into phenotypic changes reflecting their activation status. Short-time acting PGI 2 and the more stable, long lasting CTAD mixture were applied to inhibit platelets via the cAMP/PKA pathway. TRAP-6 and ADP were added for platelet activation. Platelet protein extraction was performed by TCA-precipitation of the gel-filtered platelet suspension to prevent any additional activation during pelleting by centrifugation. This precipitation step, however, prohibited the separation of the platelet secretome. Therefore, our experiment focused on events leading to changes within the non-secretory platelet proteome. To yield a high electrophoretic resolution quality these samples were separately investigated at the pH ranges of 4-7 and 6-9. Protein abundances were evaluated by DeCyder-image analysis software and quantified as fold changes compared to untreated controls. These analyses revealed 73 differentially regulated protein spots according to our statistical selection criteria. Upon PGI 2 addition, 65 protein spots were changed, 45 were changed by adding CTAD, with a broad overlap (44 spots) ( Fig. 2A). In contrast, six protein spots were regulated differently following TRAP-6 treatment compared to seven ADP-related protein spots (Fig. 2B). In summary, ten times more protein spots were affected by inhibition with a median abundance difference of 11% (+1.45-vs. +1.34-median FC, p = 0.03 Fig. 2A,B). Hence, in the non-secretory proteome inhibition footprints were qualitatively and quantitatively more prominent than activation ones, a finding that encouraged us to investigate the molecular characteristics for this difference. platelet proteome changes after inhibition/activation are mainly related to phosphorylation.
To uncover the plethora of proteome changes seen in inhibited platelets compared to activated ones, the differentially regulated protein spots were identified by mass spectrometry. These results showed that several identified proteins appeared in more than one spot, giving a total of 31 distinct proteins (Tables 1 and 2, Fig. 2A2,B2). These proteoforms, originating from a single protein, had the same molecular weight, however, migrated across the pI range, thereby displaying a chain-like spot pattern (Fig. 3, Supplementary Fig. S4). For example, five different proteoforms with the same molecular weight but different pIs ranging from 7.19 to 9.05 were identified as VASP (Fig. 4A). A similar pattern was observed for LASP1, ZYX, and SKAP2 ( Fig. 4B-D).
Subsequently, biological database analysis (Reactome) was applied to identify the most probable underlying biological processes. Knowledge-based Ingenuity Pathway Analysis alongside literature research revealed that one key node of this network was PKA upon inactivation. Since almost all proteins with the strongest changes in abundance (>1.5-fold; LASP1, SKAP2, NAP1L1, RAP1B, CALD1, ZYX, and VASP) belonged to the cAMP/ PKA-pathway ( Supplementary Fig. S5), an involvement of post-translational modifications such as phosphorylation was anticipated. To provide biochemical evidence, protein lysates of PGI 2 and CTAD-treated platelets were treated with λ-phosphatase (λ-PPase) 31 . Phosphorylated proteins lose their phospho-groups after λ-PPase treatment (treated + PPase/treated < 0.8), thereby causing a pI-shift towards an alkaline pH accompanied by an increase in the protein abundance of the corresponding "native" protein spots (treated + PPase/treated > 1.2). Using this approach, we were able to identify phosphorylation of protein species due to pI-dependent abundance changes (Table 1). Following the addition of λ-PPase, phosphorylation patterns were confirmed for VASP (pI 7.19-9.05; Fig. 5C1), LASP1 (pI 6.34-6.83; Fig. 5C2), alongside other potential PKA substrates such as RAP1B and ZYX ( Supplementary Fig. S6).
Protein spot identities and phosphorylation profiles of VASP and LASP1 were further characterised using pan and phosphosite-specific antibodies by 2D Western blots. The pan VASP antibody detected spots at five pI The CD62P-positive platelets of each individual are depicted as relative fold changes to untreated gel-filtered platelets (dashed line at y = 1). For each comparison, the average fold change (FC) and the significance level are given in the graph. The null hypothesis was tested by a one-sample t-test with an α-level set at 0.05. n = 6 (PGI 2 , CTAD), n = 6 (ADP, TRAP-6). *p < 0.05 **p < 0.01. PGI 2 -prostacyclin, CTAD -citrate, theophylline, adenosine, dipyridamole, ADP -adenosine-diphosphate, TRAP-6 -thrombin receptor-activating peptide-6.
positions in the untreated control (Fig. 5A1) and at seven pI positions in the Western blot from PGI 2 -treated platelets (Fig. 5B1). Thus, all MS identifications were verified and additional PGI 2 -inducible VASP spots located at pI 6.52 and 6.79, were immunologically identified. Moreover, the PGI 2 induced VASP protein profile displayed several spots with molecular weight shifts at the particular pIs between 6.52 and 8.45. The pVASP-Ser 239 as well as pVASP-Ser 157 antibody recognised pre-phosphorylated VASP spots at the pIs 7.19, 7.73 and 8.45 in the control proteome and detected additional phosphorylated protein species, visible by more acidic spots at the pIs 6.52 and 6.97 in the PGI 2 treated samples ( Fig. 5B1; white arrows). The two VASP spots with the most akaline pIs (8.76 and 9.05) were not detected by any of the phosphosite-specific VASP antibodies ( Fig. 5B1; black arrows), which confirmed their unphosphorylated state as already indicated by 2D-DIGE analysis. These spots displayed a decreased abundance after PGI 2 treatment and an increased abundance after λ-PPase treatment (Table 1). A molecular and (B) activation (ADP/TRAP-6) compared to untreated control. In total, 66 spots for inhibition and 7 spots for activation were significantly altered. These spots were deduced from a total of 5,906 detected protein spots by applying the following criteria: [(a) protein spots matched >90% of all 2D-DIGE gels, (b) protein abundance changes >20% between treatment group (ADP, TRAP-6, CTAD or PGI 2 ) and untreated control and (c) FDRcorrected ANOVA p-value < 0.05]. (1) Absolute numbers of separately and commonly altered protein spots upon inhibition (Σ = 66) and activation (Σ = 7). (2) Altered protein spots upon inhibition (Σ = 66) and activation (Σ = 7), that originated from distinct proteins: in sum, 31 proteins were identified (26-inhibition; 5-activation) of which 17 proteins were sensitive to phosphatase-treatment (15-inhibition; 2-activation) (3) The direction and magnitude of regulation is given for commonly and uniquely altered protein spots of each subgroup: CTAD = 45 (1 unique /44 common), PGI 2 = 65 (21/44), ADP = 7 (1/6) and TRAP-6 = 6 (0/6). Relative fold changes [treated vs. untreated] are depicted and compared to the respective control (dashed line y = 1) on a logarithmic scale. (4) The absolute amount (|x|) of altered protein spots, regardless of direction, is represented as positive fold changes [treated vs. untreated] compared to the respective control with its corresponding medians. The median abundance difference was determined by a Mann-Whitney-U test with an α-level set at p < 0.05. PGI 2 -prostacyclin, CTAD -citrate, theophylline, adenosine, dipyridamole, ADPadenosine-diphosphate, TRAP-6 -thrombin receptor-activating peptide-6. www.nature.com/scientificreports www.nature.com/scientificreports/ shift from 46 kDa to 50 kDa is described for VASP after phosphorylation at Ser157 32 . Thus, the pVASP-Ser 157 antibody should only recognize VASP signals at the molecular weight of 50 kDa. To proof these observations using the current samples, 2D Western blots were repeated with 12.5% SDS-PAGE in the second dimension to achieve a better resolution at this MW region ( Supplementary Fig. S8). These 2D Western blots clearly showed separated pVASP-Ser 157 antibody spot signals only at 50 kDa with different pIs ( Supplementary Fig. S8) indicating a sequence and different combinations of multi-site phosphorylation events. A hypothesis for VASP multisite phosphorylation chronology is delineated in Supplementary Fig. S9.
Platelet activation by ADP and TRAP-6 induced significant abundance changes in MYL9, ILK, and PLEK. A pathway analysis allocated these proteins to the phospholipase C/protein kinase C (PKC-) pathway ( Supplementary Fig. S5). Thus, potential PKC-mediated phosphorylation of these spots could also be identified by λ-PPase treatment ( Table 2).
In conclusion, these experiments confirm that phosphorylation is a predominant proteomic feature in defining platelet activation status and especially their inhibition. Beyond that, these findings propose more powerful protein biomarker candidates compared to the established platelet reactivity marker pVASP.
Finally, we were also interested in identifying platelet protein markers for activation and comparing them to the established activation marker CD62P.
The platelet activation marker CD62P is strongly associated with phosphorylated species of PLEK and ILK. After ADP or TRAP-6-mediated platelet activation, a considerably lower number of seven significantly changed protein spots were detected in the non-secretory proteome compared to platelet inhibition (Fig. 2). However, PLEK, the major substrate of PKC, was found to be significantly upregulated with two different protein spots (pI 6.52 and pI 6.60) and λ-PPase treatment confirmed this phosphorylation (Table 2). Interestingly, ILK, as well as PLEK, were the only two proteins that were affected by both, activation and inhibition of platelets. Values of each protein spots are given as relative fold changes between treated gel-filtered platelets (PGI 2 •, CTAD ○) and untreated control (dashed line y = 1) with their means and 95% confidence interval (n = 6 in each subgroup). Statistical significance was set at p < 0.05 and was determined by post-hoc testing between subgroups (PGI 2 , CTAD) and untreated controls. PGI 2 -and CTAD-dependent protein profiles of VASP were also validated by flow cytometry (Supplementary Fig. 11). PGI 2 -prostacyclin, CTAD -citrate, theophylline, adenosine, dipyridamole, pI -isoelectric point. www.nature.com/scientificreports www.nature.com/scientificreports/ PLEK phosphorylation at pI 6.60 significantly increased upon activation and in addition, showed a reduced spot signal upon inhibition when compared to the untreated controls (Table 2, Fig. 7A). Inversely, ILK at pI 7.10 was dephosphorylated upon activation and phosphorylated after inhibition (Fig. 7B). To assess whether the phosphorylation profiles of PLEK and ILK might be related to the so far used platelet activation marker CD62P, we evaluated its association. Such a correlation could be confirmed as illustrated in Fig. 7C,D. Moreover, despite the potent effect size of 2.35 Cohen's d for CD62P, the quantitative discrimination power between activation and inhibition was even more pronounced for the phosphorylated PLEK (3.77 Cohen´s d; pI 6.60) and ILK (3.79 Cohen's d; pI 7.10) protein spots. These results indicate that phosphorylated ILK and PLEK species can distinguish more effectively between activation and inhibition when compared to the commonly used platelet activation marker CD62P.

Discussion
Increased platelet reactivity may result in thrombosis and other vascular diseases. By investigating the platelet proteome after in-vitro activation and inhibition, we attempted to uncover discriminative signs for these modulations in platelet activation state. Here, we expose for the first time that platelet inhibition was associated with more pronounced alterations in the non-secretory platelet proteome as compared to activation. Moreover, we showed that these changes were mainly caused by phosphorylation, suggesting a predominant role of this post-translational modification in maintaining platelets in their resting state. Novel diagnostic options were revealed, since PGI 2 or CTAD inhibited platelets showed stronger phosphorylation changes in PKA targets than the benchmark platelet reactivity marker pVASP. Additionally, phosphorylated species of PLEK and ILK were associated with CD62P surface expression and showed an even larger effect size than this established activation marker.
In our representative 2D-DIGE platelet proteome analysis, comprising 5,906 protein spots, 66 and 7 protein spots were significantly changed in response to 15 minutes inhibition and activation stimuli, respectively. However, changes caused by platelet protein secretion or protein dynamics in the early phase of inhibition, as previously shown 17 , are not captured based on our study design. On a physiological basis, the disparity between inhibition and activation induced changes in the proteome could be due to platelets being more responsive to inhibitory stimuli. Potent platelet inhibition is constantly needed in vascular system due to the exposure to shear stress, acting as a strong platelet activator 33 . Shear stress is permanently present under physiological conditions and is much more pronounced in pathophysiological conditions, e.g. carotid or coronary vessel stenosis 34,35 . Accordingly, the present findings demonstrate that the addition of PGI 2 or CTAD, to obtain resting platelets, strongly influences the non-secretory platelet proteome. Thus, after inhibition, 15 out of 26 proteins were altered in their phosphorylation levels. Knowledge-based pathway analysis demonstrated that the unbiased 2D-DIGE proteomics technology specifically accentuated the cAMP-induced PKA-signaling pathway in PGI 2 or CTAD-treated platelets. Finally, we characterised multiple proteoforms at seven pI positions of the well-known PKA phosphorylation target VASP -a negative regulator of platelet activation 36 and a clinically established platelet reactivity marker 28,37,38 . Remarkably, CTAD-treated platelets did not show a significant phosphorylation in any 2D-DIGE detected VASP species. Intriguingly, other spots showed substantially stronger abundance changes in their phosphorylation levels than phosphorylated VASP species: LASP1 6.7-fold, RAP1B 4.1-fold, SKAP2 4.6-fold, ZYX 1.7-fold, and RGS18 1.5-fold compared to VASP with 1.3-fold. Previously, a 6.7-fold stronger phosphorylation of RAP1B compared to VASP was reported in platelets treated with cAMP-elevating reagents 39 , which supports the reliability of our data. This latter study showed that VASP phosphorylation was reversible with maximum phosphorylation after 10 minutes, dropping to baseline after 60 minutes. Conversely, RAP1B phosphorylation was irreversible with constant levels over 2 hours 39 . Beyond RAP1B, the stability of PGI 2 -and CTAD-induced changes in LASP1 and SKAP2 phosphorylation was confirmed for at least 40 minutes in our study (Supplementary Fig. S13). This long-lasting stability of these phosphorylation events is of additional interest, since it is generally accepted that after about 25 minutes at room temperature the biologically active PGI 2 is degraded and the inhibitory influence of PGI 2 comes to an end.
The phosphorylated protein species of LASP1, RAP1B and RGS18 40 have already been described as targets for PKA in the inhibitory pathways of platelets 41,42 . For ZYX, phosphorylation after stimulation of the cAMP/PKA pathway by forskolin has recently been shown in endothelial cells 43 . Until now, SKAP2 is not shown as a target of PKA, but its phosphorylation upon iloprost and PGI 2 treatment in resting platelets has been reported 17,44 . These actual novel findings in cAMP-mediated quantitative differential phosphorylation degree of PKA targets suggest that e.g. phosphorylated LASP1, RAP1B, and SKAP2 could be technically more robust biomarkers to measure the platelet activation status than phosphorylated VASP. A previous proteome analysis of inactivated platelets employing gel-free MS analysis from TiO 2 -enriched phospho-peptides revealed 300 proteins with differential phosphorylation patterns between 10 to 60 seconds after stimulation with the stable PGI 2 analogue iloprost 17 . However, a stronger upregulation in phosphorylation of LASP1, ZYX, SKAP2, RGS18, and RAP1B in comparison to a weaker phosphorylation of VASP was not reported. The authors just stated that pVASP-Ser 239 had the highest degree of phosphorylation with a 7.1-fold up-regulation after 60 s platelet inhibition in the presence of protein spots are recognised by the pan as well as phospho-site specific antibodies. (A) Platelet protein extracts from citrate anticoagulated blood constitute the untreated control. (B) PGI 2 -treatment shifts and increased protein spot abundances towards the acidic pH-range with a concomitant reduction of the corresponding alkaline spot abundances. (C) Λ-PPase treatment removes the phosphorylation of respective protein species thereby shifting its pI towards the alkaline direction. The LASP1 images (2) overlap (vertical solid line) between pH 6-7 due to its wide-spreading across the pI-range (pH 6-7), resprective non-cropped 2D images are displayed in Supplementary Fig. S7 www.nature.com/scientificreports www.nature.com/scientificreports/ 5 nM Iloprost 17 . The PKA phosphorylation targets pLASP1-Ser 146 and pSKAP2-Ser 283 or Ser 286 were found to be less phosphorylated (both 3.6-fold upregulation) in comparison to phosphorylated pVASP-Ser 239 . Phosphorylated peptides from RAP1B were not detected and RGS18 phosphorylation decreased. However, these discrepancies regarding the phosphorylation degree of these PKA targets may be related to their earlier measuring points (10, 30 and 60 s) in addition to the completely different methodology employed. Compared to the number of 26 inhibition-affected platelet proteins detected in our study, a considerably higher number with 300 altered phosphorylated proteins from iloprost-inactivated platelets was described in the study by Beck et al. 17 . This higher count of identified potential PKA targets was achieved since the used TiO 2 affinity-based method enriches phospho-peptides and enables the detection of phosphorylated low-abundant proteins 17 . With this enrichment step, however, a direct quantitative assessment of phosphorylation events compared to the global platelet proteome is lost 17 . Our 2D-DIGE snapshot, on the other hand, displayed the relationship of protein phosphorylation to the overall proteome and, thereby also highlights the specificity of platelet inhibition pathways.
Despite a prominent CD62P surface upregulation, the actual 2D-DIGE based analysis revealed unexpected weaker changes in the non-secretory proteome of activated platelets compared to inhibited platelets. In contrast to previous studies investigating the activated platelet proteome by two-dimensional gel electrophoresis 13,24,45,46 , our experimental design could only measure non-secretory proteome changes but cannot dissect alterations based on granula and microvesicle release. However, our statistical stringent platelet proteome analysis by Benjamini-Hochberg correction showed in total only seven protein spots changed after platelet activation. Without correction for multiple comparisons (FDR), we would have detected 32 additional proteins upon ADP-and TRAP-6 activation. Nevertheless, these changes in spot counts are quite comparable to the explorative 2D-DIGE based proteomics study of gel-filtrated platelets from Schulz et al. with 13 differentially regulated protein spots after activating the glycoprotein VI receptor 47 . In this study, the degree of platelet stimulation was also quantified by CD62P upregulation 47 and is comparable to our in-vitro platelet activation CD62P profiles. Evidence for the specific stimulation of platelet activation pathways in our study is the phosphorylation of well-known PKC substrates such as PLEK and MYL9, proteins that were also detected in most of the published 2D gel-based proteomics studies of in-vitro activated platelets 13,24,47,48 . Recently, Beck et al. 19 characterized 302 phosphorylated proteins following ADP-treatment with kinetics between 10-and 60 seconds. Despite the high sensitivity of TiO 2 enrichment, some important PKC substrates in platelet activation, such as MYL9, have not been detected 19 . These differences in results indicate that each technical proteomic approach has distinct strengths and limitations. For the present proteomics study by 2D-DIGE analysis the following limitations need to be considered: under-representation of low abundant proteins, very hydrophobic proteins and proteins with a high molecular weight. However, 2D-DIGE analysis also provides several advantages like highly reliable qualitative and quantitative accurate profiling of differentially abundant proteins by the inclusion of an internal standard which enables a low technical variation with a CV of 7% 29 . Finally, only protein-based "top-down" proteomics as 2D-DIGE can provide reliable quantitative results from different proteoforms 49 . For LC-MS/MS mass spectrometry-based proteome analysis tryptic digestion of the protein samples is necessary but causes a loss of information for the correct assignment of peptides and post-translational modifications to the appropriate proteoform 49 .
In the present study, PLEK (pI 6.60) and ILK (pI 7.10) species were the only two proteins that were altered by both platelet activation and inhibition. Apart from that, a pronounced overlap in the affected proteome was not found. This seems plausible since ADP and thrombin inhibit adenylyl cyclase activity via G i , the central enzyme in the inhibitory cAMP/PKA pathway 50 . PKC-dependent phosphorylation of PLEK is a sensitive marker for platelet activation 51 . Although ILK is described as a substrate for PKC and PI3K 52 to induce platelet adhesion 53 , we found Figure 6. Platelet inhibition causes stronger phosphorylation changes in LASP1, SKAP2 and RAP1B than in VASP. The acidic protein spot abundances of VASP, LASP1, SKAP2, RAP1B, and ZYX are depicted as fold changes between treated gel-filtered platelets after (A) PGI 2 • and (B) CTAD ○ treatment with their means, 95% confidence interval (CI) and isoelectric point (n = 6 in every group). Statistical significance was set at p < 0.05 and determined by a repeated ANOVA with single post-hoc contrast compared to VASP. Statistical significance is marked and represents the respective acidic protein abundance compared to VASP abundance. *p < 0.05 **p < 0.01 ***p < 0.001 ****p < 0.0001.
www.nature.com/scientificreports www.nature.com/scientificreports/ an explicit dephosphorylation pattern for ILK (pI 7.10) in activated platelets that has not been reported before. Despite a potent effect size of 2.35 (Cohen's d) for the established platelet activation marker CD62P, the quantitative discrimination power between activation and inhibition was even more pronounced for phosphorylated PLEK (3.77 Cohen's d; pI 6.60) and ILK (3.79 Cohen's d, pI 7.10) species. As platelet CD62P expression may decrease in severe clinical states of platelet activation 54 , potentially due to receptor-and microvesicle-shedding and/or platelet adhesion to endothelial cells and leukocytes, phosphorylated ILK and PLEK species may represent further more robust diagnostic targets for platelet activation.
These stronger phosphorylation changes of the non-secretory proteome from inhibited compared to activated platelets may provide fundamental new insights into quantitative functional regulations of platelets. From there, new and more robust biomarkers for platelet inhibition and activation as well as new therapeutic options could follow, which may enable to treat abnormal platelet function more targeted in accordance with underlying pathophysiological mechanisms. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods platelet preparation and in-vitro conditions for platelet inhibition and activation. The study included twelve healthy volunteers aged 31.7 ± 7.1 years (mean ± SD), who had not taken any antiplatelet medication for at least 14 days prior to blood sampling. Individuals had a 1:1 gender ratio and were all normal weight (BMI 23.6) non-smokers. The study was approved by the Ethics Committee of the Medical University of Vienna (1076/2016). Each participant was advised of the purpose of the study and written informed consent was obtained prior in accordance with the principles of the Declaration of Helsinki. Blood was drawn from the antecubital vein (21 G needle) without stasis into sodium-citrate tubes (0.129 mM citrate) as well as CTAD tubes (0.129 mM trisodium citrate, 15 mM theophylline, 3.7 mM adenosine, 0.198 mM dipyridamole), a mixture of cAMP-elevating agents. These commercially available CTAD blood tubes are used to prevent platelet activation during blood sampling and subsequent processing steps 51 . Blood tubes were centrifuged at 120 xg for 20 minutes to obtain platelet-rich plasma (PRP). A graphical experimental design for in-vitro treatments is displayed in the Supplementary  Fig. S1. Briefly, for inhibition, PRP was incubated for 5 minutes with PGI 2 (0.4 µM final at RT) before gel-filtration, while PRP from CTAD blood tubes was left untreated. A PGI 2 concentration of 0.1-0.5 μM is normally used to generate washed platelets 21,23,24 . To gently purify platelets from plasma proteins by size-exclusion chromatography 25 , 1 ml PRP was applied onto a 12 cm sepharose-2B packed column pre-equilibrated and eluted in PBS buffer. Gel-filtered platelets were caught at the end of the column and precipitated by mixing 3:1 (v/v) with 6.1 N TCA solution (containing 80 mM DTT). Thus, the inhibition of platelets by PGI 2 was stopped after altogether 15 minutes and by CTAD after 40 minutes. Alternatively, PGI 2 was added not to the PRP, but to the gel-filtered platelet suspension, which induced comparable changes in the proteome as seen in PRP + PGI 2 ( Supplementary Fig. S2). For activation, gel-filtered platelets derived from PRP were incubated either with ADP (5 µM) or TRAP-6 (15 µM) as described previously 20,27 to evoke a moderate or strong activation response, respectively. Gel-filtration constitutes a gentle method to isolate platelets 52 , however, activated platelets cannot pass the sepharose column and get stuck, resulting in a diminished platelet yield ( Supplementary Fig. S3). Therefore, we decided to activate platelets after gel-filtration. Stimulations were stopped by TCA/DTT precipitation after 15 minutes at RT. For baseline values, corresponding citrated PRP was gel-filtered and immediately precipitated. After one hour at 4 °C, the TCA-extracted platelet proteins from all conditions were pelleted at 20,000 × g for 10 minutes at 4 °C and washed four times (20,000 × g, 10 minutes, 4 °C) with acetone containing 20 mM DTT.

Platelet proteome analysis by fluorescence two-dimensional gel electrophoresis (2D-DIGE).
Resolubilisation of precipitated proteins, determination of protein concentration, preparation of the internal standard and labelling with fluorescent cyanine dyes (Cy2, Cy3, Cy5) were performed as described previously 29 . Briefly, the TCA-precipitated and acetone-washed platelet proteins were re-solubilised in urea-sample buffer (7 M urea, 2 M thiourea, 4% CHAPS, 20 mM Tris-HCl pH 8.5) and fluorescence-labeled prior to 2D-separation. For the normalisation process, an internal standard (IS) of platelet proteins from all treatment conditions was prepared. This IS, labelled in Cy2 and added to every 2D-gel, enables a highly accurate qualitative and quantitative comparison of all samples and treatments 28,53 . The 2D-DIGE analysis was performed in the pH-ranges 4-7 and 6-9, respectively. For the acidic side, 36 μg protein (2 × 12 μg sample +1 × 12 μg internal standard) were passively rehydrated on 24 cm pH 4-7 IPG-Dry-Strips in a solution containing 7 M urea, 2 M thiourea, 4% CHAPS, 70 mM DTT and 0.5% ampholyte pH 4-7. For the alkaline side, 24 cm pH 6-9 IPG-Dry-Strips (GE Healthcare, Uppsala, Sweden) were soaked in a rehydration solution containing 7 M urea, 2 M thiourea, 4% CHAPS, 150 mM DTT and 2% ampholytes pH 6-9 prior to isoelectric focusing. Afterward, 3 × 12 μg of labeled protein samples were applied by "cup-loading" via the acidic side and the isoelectric focusing was carried out until 30 kVh was reached. The following SDS-PAGE was performed with an 11.5% acrylamide gel at 35 V for 1 hour, 50 V for 1.5 hours and 110 V for 16.5 hours at 10 °C in an Ettan DALTsix electrophoresis chamber (GE Healthcare, Uppsala, Sweden). The gels were scanned afterward with a resolution of 100 μm using a Typhoon 9410 imager (GE Healthcare, Uppsala, Sweden). The experimental study design of the 2D-DIGE analysis is delineated in Table S1.
Detailed description of blood sampling, platelet isolation, in-vitro platelet inhibition and activation, 2D-DIGE image analysis, flow cytometry analysis, VASP phosphorylation assay, biological pathway analysis and Western blot analysis is provided in the Supplemental Material and Methods.
Statistics. Fold changes (FC) of standardised protein values (treatment/IS) were calculated between intervention groups (PGI 2 , CTAD, ADP, TRAP-6) and the untreated control group. Protein spots were considered relevant if 1) spots were matched in more than 90% of all 2D images (n = 24) and 2) if FC were >20% as detected between untreated control and at least one of the intervention groups PGI 2 , CTAD, ADP, TRAP-6. After this filtering, 495 of 5,906 protein spots in the pH range 4-9 were considered relevant. Log-transformed data of 495 spots were evaluated by a repeated ANOVA conducted as a linear mixed model with "group" (untreated control, PGI 2 , CTAD, ADP, TRAP-6) as repeated factor. The multiplicity was controlled by the Benjamini-Hochberg procedure (FDR < 0.05). For a direct comparison of different parameters the effect size was calculated by Cohen's D = (mean 1 -mean 2 )/standard deviation pooled . SPSS Statistics 20.0 (IBM, Chicago, USA) was used for the statistical analysis. Unless stated otherwise, all fold changes represent means.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding authors on reasonable request.