Accelerated endoplasmic reticulum and coagulation factor processing in platelets are linked with the hypercoagulable state of patients with lung cancer

The risk of venous thromboembolism in cancer is nine times higher than in the general population and the second leading cause of death in these patients. Platelets play a key role in tumour growth, metastasis, and cancer-associated thrombosis. Despite this widely observed functional role of platelets in the prothrombotic state of certain cancers, the underlying molecular mechanisms are largely unknown. are signicantly linked to the plasmatic haemostasis factors D-dimer and brinogen. no prospective systemic study investigating the respective platelet proteome have been performed so far. To address this, we investigated the platelet proteome in patients with two different types of cancer to identify potential mechanisms of increased risks for VTE and/or mortality and possible molecular pattern that are specic for the cancer type. Our data demonstrate that several proteins and/or corresponding proteoforms were signicantly changed in the platelet proteome of patients with lung cancer, but not in brain cancer. Ten µL containing As for specicity irreversible transglutaminase inhibitor T101 µM nal used. The enzyme kinetic were at 37°C using Varioskan LUX microplate Fisher excitation and kinetic mode, absorbance was read every 36 s for min and the change of absorbance between 0 and 15 min were detected. F13A1 activity determined on the basis of a standard curve constructed with washed human platelets. Microsoft excel further processing of enzyme kinetic data.


Introduction
Patients with cancer have an increased risk to develop venous thromboembolism (VTE) (1). This complication occurs in up to 20% of affected individuals per year (2) and is associated with high morbidity and mortality (3,4). Malignancies with the highest incidence of VTE are brain and lung cancer (5). Platelets are thought to play an important role in the prothrombotic state and the development of cancer-associated VTE as well as arterial thromboembolism (6). Besides their crucial functions in primary haemostasis, platelets are also important players in the progression of VTE. Various platelet associated parameters were already identi ed to be linked with increased risk of VTE in patients with cancer such as high platelet counts (7) and exposed phosphatidylserine on platelets (8). High plasma levels of soluble P-selectin (sCD62P) indicate cancer-associated platelet activation (9). In a previous study of patients with different types of cancer, we found a statistical association of increased VTE risk and poor survival rate with lower surface expression of receptors important for platelet activation, namely CD62P and glycoprotein IIb/IIIa, a complex of integrin alpha IIb (ITGA2B) and beta 3 (ITGB3) (5). The decline of these platelet receptors is supposed to be caused by shedding during continuous platelet activation, stronger ongoing in cancer patients with increased For platelet isolation, blood was drawn from an antecubital vein into 3.5 mL vacuum tubes containing CTAD (0.129 mM trisodium citrate, 15 mM theophylline, 3.7 mM adenosine, 0.198 mM dipyridamole; Greiner Bio-One, Kremsmünster, Austria) as anticoagulant. The rst tube drawn was discarded to avoid any contaminations. Immediately following blood draw, CTAD tubes was centrifuged at 120 xg for 20 min without brake at room temperature. Obtained platelet rich plasma (PRP) was carefully transferred into a fresh tube containing prostacyclin I 2 (0.8 µM, PGI 2 ; Sigma-Aldrich, St. Louis, MO, USA) to prevent platelet aggregation and degranulation during the following washing steps. PRP was pelleted by centrifugation for 3 min at 3000 xg at room temperature and the protein pellets were washed twice in phosphate-buffered saline (w/o: Ca 2+ and Mg 2+ ) containing PGI 2 (0.8 µM). Before the last centrifugation step, platelet count was determined with a Sysmex XN-350 hematocytometer (Sysmex, Kobe, Japan). After the last centrifugation step the supernatant was completely taken off and the platelet pellet was snap-frozen in liquid nitrogen and stored at -80°C until processing.
For preparation of plasma blood was drawn into 3.5 mL vacuum tubes containing sodium citrate (0.129 mM citrate; Greiner Bio-One, Kremsmünster, Austria). The blood was centrifuged at 2500 xg for 15 min at 15°C to separate the cellular fraction and the plasma supernatant was stored at -80°C.

Platelet preparation for 2D-DIGE analysis
The frozen platelet protein pellets were resolubilized in urea-sample buffer (7 M urea, 2 M thiourea, 4% CHAPS, 20 mM Tris-HCl pH 8.68) and incubated for 2 h at 4°C under agitation (800 rpm). Protein quantitation of individual samples was done in triplicate with a Coomassie brilliant blue protein assay kit (Pierce, Thermo Scienti c, Rockford, IL, USA). The internal standard (IS) was made by pooling the same protein amounts from each platelet samples of all included study participants. Platelet protein samples and IS were aliquoted and stored at -80°C.

2D-DIGE image analysis
For protein spot detection, 2D-DIGE gels were scanned with three different wavelengths of the particular CyDye at a resolution of 100 µm using a Typhoon 9410 Scanner (GE Healthcare, Uppsala, Sweden). Gel images were analysed via the DeCyder™ software (version 7.2, GE Healthcare, Uppsala, Sweden). Spots were matched to a master 2D-DIGE gel (a representative pH 4-7 platelet protein map of the IS images). On average, 500 protein spots were matched manually to the master gel using the DeCyder™ software. Afterwards an automatic spot match was used which achieved an average of 4310 matched spots per gel. The standardised abundance (SA) of every protein spot was calculated by the DeCyder™ software with two normalisation steps. The rst step is the in-gel normalisation by dividing each spot with the center volume of the corresponding spot map and a the second one by dividing each normalised spot volumes against the corresponding spot normalised spot value of the IS. (22).
Detailed information about the image analysis was published by Winkler et al. (23).
Protein identi cation via mass spectrometry For MS-based identi cations, 250 µg unlabelled proteins were separated by 2D-DIGE and proteins were visualized by MScompatible silver staining (24). Protein spots of interest were excised manually from the gels, de-stained, disul de were reduced as well as derivatized with iodoacetamide and the proteins were tryptically digested. Subsequently, these peptide samples were applied onto a Dionex Ultimate 3000 RSLC nano-HPLC system (Thermo Scienti c) and afterward directly subjected to a QqTOF mass spectrometer oTOF compact (Bruker Daltonics) equipped with nano-ow CaptiveSpray ionisation device. Detailed analytical conditions were previously described (25). The protein identi cation was obtained with database searches against UniProtKB/Swiss-Prot (2020-06) using Mascot v2.7 server.
For two-dimensional Western blot (2-D WB) analysis, 36 µg Cy2-labeld platelet proteins were separated by IEF on either a 7 cm pH 3-10 or a 24 cm pH 4-7 IPG strip as described for 2D-DIGE gels, and subsequently transferred onto a NC-or PVDF membrane (75 V, 90 min). Afterward, the membrane was blocked with 5% non-fat dry milk (BioRad, Hercules, CA, USA) in 1x PBS containing 0.3% Tween-20 (PBS-T) over night at 4°C under gentle shaking. On the next day, membranes were washed (3x with PBS-T for 5 min, each). For detection, following primary antibodies were used in the corresponding dilutions by incubation for 2 h at room temperature (180 rpm) in PBS-T containing 3% non-fat dry milk: monoclonal anti-Factor F13A1 (ab1834; Abcam, USA) 1:250, washing (3x with PBS-T for 5 min, each), the membranes were incubated with a DyLight 650 conjugated secondary antibody (Novus Biologicals, Littleton, CO, USA), diluted 1:500 or with a horse-radish peroxidase (HRP)-conjugated secondary antibody, diluted 1:20,000 in PBS-T containing 3% non-fat dry milk for 1.5 h in the dark at room temperature (65 rpm). After washing (3x with PBS-T for 5 min, each), the membranes were incubated for 1.5 h in the dark at room temperature (65 rpm). After washing (2x with PBS-T, 1x with 1x PBS for 5 min, each), the antibody uorescence signals were detected with a Typhoon FLA 9500 imager (GE Healthcare, Uppsala, Sweden) at a resolution of 100 µm. The HRP signal was detected using an Enhanced Chemiluminescent (FluorChem® HD2, Alpha Innotech, CA USA). The 1-D WB antibody signal of F13A1 were normalized by the RuBPS uorescence signal from the 40 kDa to 100 kDa bands and quanti ed with ImageQuant 8.0 (GE Healthcare, Uppsala, Sweden).

Measurement of haemostatic biomarkers in plasma
FXIII activity in citrate plasma was measured by Berichrom Factor XIII chromogenic determination (Siemens Healthcare) according to manufacturers' instructions. Factor VIII activity was measured on a Sysmex CA 7000 analyzer using factor VIIIde cient plasma (Technoclone) and APTT Actin-FS (Dade Behring). D-dimer levels were measured by a quantitative latex assay (STA-LIAtest D-DI; Diagnostica-Stago, Asnieres, France) on a STAR analyser (Diagnostica-Stago) according to the manufacturer's instructions. Fibrinogen was routinely measured in platelet poor plasma according to Clauss (STA Fibrinogen; Diagnostica Stago, Asnieres, France; normal range: 180-390 mg/dL). C-reactive protein (CRP) was measured with a fully automated particle enhanced immunonephelometry (N high sensitivity CRP, Dade Behring, Marburg, Germany). For sCD62P, to obtain de nitely platelet-free plasma a second centrifugation step (Eppendorf) at 13,400 xg for 2 min was performed. These plasma aliquots were also stored at − 80°C, until the determination of sCD62P plasma levels in series. Soluble CD62P levels were measured using a human sCD62P immunoassay (R&D Systems, Minneapolis, MN) following the manufacturer's instructions.
Fluorogenic F13A1 activity assay from platelet samples Washed platelets from 42 patient´s samples were lysed by 1% Triton X-100 (Sigma-Aldrich, St. Louis, MO, USA) and sonicated for 1 min at 4°C. After centrifugation (15,000 xg, 4°C) lysed platelets were transferred to a fresh tube. Protein concentration in the lysate was determined by Coomassie. F13A1 activity in platelets was performed by a uorogenic FXIIIA enzyme activity kit (Zedira GmbH, #F001, Darmstadt, Germany) essentially according to the manufacturers' protocol with minor modi cations.
Measurements were performed in triplicate in 96-well at bottom black microplates (Thermo Scienti c, #137101, Denmark). Ten µL of lysed samples was mixed with 90 µL reagent mix solution (modi ed peptide, 100 NIH Units thrombin, Tris buffer pH 7.5 containing CaCl 2 , NaCl, polyethylene glycol (PEG), glycine methyl ester, clot inhibitor peptide, Heparin antagonist (hexadimethrine bromide) and sodium azide), creating a nal volume of 100 µL/well. As a control for speci city of enzymatic F13A1 activity, the irreversible transglutaminase inhibitor T101 (50 µM nal concentration) was used. The enzyme kinetic were recorded at 37°C using a Varioskan LUX microplate reader (Thermo Fisher Scienti c) with excitation at 313 nm and emission at 418 nm in the kinetic mode, absorbance was read every 36 s for 15 min and the change of absorbance between 0 and 15 min were detected. F13A1 activity was determined on the basis of a standard curve constructed with washed human platelets. Microsoft excel was used for further processing of enzyme kinetic data.

Biological pathway analysis
Biological data base analysis was made from the fteen lung cancer-related platelet proteins. The data source for the proteinprotein interaction (PPI) networks was the protein query of the STRING database (26), with the settings (active interaction sources: experiments and databases; score = 0.4; maximal additional interactors = 0). For the functional enrichment, the Gene Ontology Biological Processes and KEEG pathway analysis were used for the PPI networks with a speci c colour for each biological process and KEGG pathway. The STRING Version 11.0b was used.
In addition, we applied the NetworkAnalyst platform (27,28) for further analysis. The differentially expressed proteins were uploaded to the web-platform (www.networkanalyst.ca) and generic protein-protein interaction (PPI) network analysis was performed with the Imex interactome (29) resulting in a rst-order network with 722 nodes and 889 edges comprising the 15 lung cancer proteins as seeds. This network was downloaded as graphML-le and imported into the Cytoscape 3.8.2 program. The stringApp was used to "STRINGify" the network, followed by functional enrichment of pathways and gene ontologies, which were sorted according to p-values. Relevant pathways were selected and coloured using the bypass function for node colours.

Statistics
For statistical analysis, from each 2D-DIGE image only protein spots are included which could be matched by the IS spot map with more than 95% of all 2-D platelet proteome maps of this study. This quality selection limits resulting protein spots to 617 from in average 2720 to the master gel matched spots. One-way and two-way analysis of variance (ANOVA) were calculated for these 617 highly reliable matched spots between the three study groups (healthy controls, patients with lung and brain cancer).
Resulting p values of the one-way ANOVA were false discovery rate corrected (FDR) by Benjamini Hochberg for multiple comparisons (30). FDR-corrected one-way ANOVA and unadjusted two-way ANOVA were calculated with the Extended Data Analysis (EDA) module of the DeCyder™ software (version 7.2, GE Healthcare, Uppsala, Sweden). Supervised principal component analysis (PCA) of one-way ANOVA signi cant spots was also performed by this EDA module to control sample clustering. To assess the effect of mortality on the examined platelet proteom between cancer type groups, the two-way ANOVA main effects "mortality", "cancer type group" and interaction term "cancer type group * mortality" were included. Signi cant differences between control group and each cancer group were analysed by planned post-hoc contrasts analysis with unpaired Student´s t-Test in SPSS and corrected for multiple comparison by the online FDR calculator (31). For a direct comparison of different parameters the effect size was calculated by Cohen's D = (mean 1 -mean 2 )/standard deviation pooled ). Graphs were created with GraphPad Prism 6 (GraphPad Software, Inc. San Diego California, USA).

Patient characteristics
To investigate if and how the platelet proteome is affected in patients with cancer and high thrombosis risk, in total 82 study participants were included in this study: Twenty-two patients with brain cancer, 19 patients with lung cancer and 41 healthy age and sex-matched healthy controls.
During the course of the study, two patients with brain (9.1%) and two patients with lung cancer (10.5%) had a VTE. These thrombotic events included pulmonary embolism (PE; 7.3%) and deep vein thrombosis (DVT; 2.4%). Eight patients with brain cancer (36.4%) and ten patients with lung cancer (52.6%) died during the follow up. Detailed characteristics of the study populations are presented in Table 1.

Platelet proteome of patients with brain and lung cancer compared to controls
For the identi cation of differentially regulated platelet proteins in patients with brain and lung cancer and matched healthy controls, the platelet proteome of these samples were analysed by 2D-DIGE in the pH range 4-7 ( Fig. 1). After the application of protein spot quality selection criteria (de ned in the material and methods section) a total of 617 protein spots were included in statistical analysis.
Platelet counts of patients with lung cancer were signi cantly higher compared to healthy controls and patients with brain cancer (Table 1). Hence, it may be supposed that changes of the platelet proteomes of patients with lung cancer can also be associated to increased platelet counts. To prove this possible bias of a platelet proteomics study, platelet counts were correlated with the 617 2D-DIGE quanti ed platelet protein spots from the whole study cohorts. No signi cant correlation could be detected by these statistical control evaluations. This assessment ensured that cancer-related protein pro les in platelets were not biased by differences in platelet counts Signi cant differences in platelet protein levels between the two cancer patients groups with high thrombosis risk and healthy controls were determined by one-way ANOVA. Multiple comparison of variance in the platelet proteome revealed 36 signi cant protein spot changes between the three study groups (FDR-adjusted P < 0.05; Fig. 1; Table 2). Identi cation of these 36 protein spots by MS showed that these proteoforms correspond to 19 different proteins ( Table 2). To get a graphical overview of these cancer-related proteome data, PCA was performed to visualize the overall difference of the 36 signi cantly altered platelet protein spots from every study participant (Fig. 2a). Furthermore, spots from each study group can be summarized in a graph to enable a straightforward comparison between the three study cohorts (Fig. 2b). PCA analysis revealed a lower variance between patients with brain cancer and healthy participants, but a stronger difference between lung cancer patients and other cohorts.
Adjusted post-hoc contrast calculations of these 36 one-way ANOVA-ltered proteoforms between each cancer type and matched controls con rmed the graphical PCA overview. These further statistical evaluations revealed 27 signi cant lung cancer-related protein spot changes but no signi cant brain cancer-related difference ( Table 2). Identi cation of the protein spots by MS showed that the 27 lung cancer related proteoforms belonged to 15 different proteins (Table 2). Therefore, further investigations focused on the lung cancer-related platelet protein changes.

Pathway analysis of lung cancer-related platelet proteins
To assess coherent protein networks and functional relationships of lung cancer related platelet protein changes, biological database analysis was conducted. Unbiased STRING protein interaction analysis allocated lung cancer-related changes mainly to proteins primary responsible for thrombosis. Based on the altered levels of ITGA2B, ITGB3, TLN1, F13A1, TF and HSPA5 the highest signi cant functional enrichment was found for the biological process of platelet degranulation (p = 3.19 − 11 ) and for platelet activation in KEGG pathways (p = 4.47 − 5 ). Additionally, a signi cantly altered KEEG pathway was protein processing in the endoplasmic reticulum (ER) (p = 4.74 − 6 ) (Fig. 3). These ndings were further supported by analyses using the NetworkAnalyst approach (27), where differentially altered genes or proteins are compared with known networks such as the human protein interactome, followed by the computation of a rst-order network, which is then queried against different databases, including KEGG and Reactome of gene ontology databases. This leads to a very robust identi cation of signi cantly altered pathways or biological processes (Additional le 1: Fig. S1).
Based on these ndings and literature recherché the main stabilizer of the brin clot, F13A1, and the ER chaperones P4HB, CALR and HSPA5, important for folding and glycosylation of secretory proteins such as coagulation factors (32, 33), were chosen for deeper characterisations and functional explorations. F13A1 protein processing is changed in platelets of patients with lung cancer The abundance of a 55 kDa fragment (pI 5.00) from F13A1 was signi cantly increased in platelets of patients with lung cancer (Table 2; Fig. 1). The brin crosslinking protein F13A1 is a transglutaminase with a molecular weight of 83 kDa.
MS-based identi cation of this 55 kDa fragment of F13A1 was validated by 2-D WB. An antibody, raised against the full length F13A1, recognised the respective 55 kDa proteoform. In addition, the F13A1 antibody bound to three spots with a MW of 83 kDa and pI between 5.85 and 5.65. Subsequent MS analysis con rmed these spots as F13A1 proteoforms (Fig. 4a). Another two F13A1 spots (12 and 12d) were identi ed by MS in the near neighbourhood (Additional le 2: Fig. S2b). The abundance of all these full-length F13A1 spots were also found to be increased in platelets of patients with lung cancer compared to matched healthy controls, except one spot with a pI of 5.85 (spot 12c; Table 3). Increased levels of the 55 kDa F13A1 proteoform in platelets of patients with lung cancer were con rmed by uorescence 1-D WB with a signi cant correlation (r s = 0.843; p = 2.0 − 6 ) to the corresponding 2D-DIGE signals (Additional le 3: Fig. S3a). The abundance of the full-length F13A1 proteoforms were visible as a single band but the lung cancer-related increase of the 83 kDa proteoforms between the pI 5.75 to 5.60 could not be detected by 1-D WB. These unchanged quantitative 1-D pro les of the full-length F13A1 in patients with lung cancer may be based on the signi cant negative (opposite) correlation of the abundance from the most alkaline F13A1 spot (pI 5.85; spot 12c) to the levels of all other 83 kDa proteoforms from F13A1 (Additional le 4: Table S1). Interestingly, this F13A1 spot 12c with the pI 5.85 was also in the opposite direction changed in patients with lung cancer compared to all other F13A1 proteoforms (Table 3).
To investigate functional outcomes of these altered levels of F13A1 proteoforms in patients with lung cancer, its enzymatic activity was measured in platelets and plasma of the particular study samples. No change of enzymatic F13A1 activity was detectable in platelets of patients with lung cancer (Fig. 4f). However, enzymatic F13A1 activity signi cantly correlated with the levels of F13A1 spots 12b (pI 5.75), 12 (pI 5.60) and 12a (pI 5.65), abundances of which were signi cantly increased in patients with lung cancer (Table 3). Once again, the F13A1 level of spot 12c (pI 5.85) attracted attention with an opposite direction by a weak trend of a negative correlation to enzymatic F13A1 activity (Table 3). These observations indicated that a pI shifting PTM regulates the enzymatic activity of these 83 kDa F13A1 proteoforms in platelets, whereby the most alkaline one (spot 12c; pI 5.85) seems to be the inactive variant. The 55 kDa spot 11 of F13A1 did not correlate with enzymatic activity of this transglutaminase (Table 3).
Generally F13A1 is activated by thrombin with the enzymatic removal of a short (2-37) activation peptide (34,35), which results into the 79 kDa activated F13A1 proteoform. Whereas a 55 kDa fragment of F13A1 is observed to be produced after the enzymatic inactivation of activated F13A1 by further cleavage with e.g. thrombin (36), chymase (37) and plasmin (38). MS analysis of the lung-cancer-related 55 kDa F13A1 cleavage product detected peptides between the amino acid sequence 13 and 492 (Additional le 5: Fig. S4a). For inactivation of F13A1, thrombin cuts at K514 (36), chymase at F574 (37) and plasmin at R492 (38). Accordingly, several of these serine proteases can be responsible for this 55 kDa F13A1 cleavage product in platelets (Additional le 5: Fig. S4b). Interestingly, MS analysis of this platelet 55 kDa F13A1 fragment identi ed also peptides from the amino acid sequence 13 to 38, which is within the amino sequence of the activation peptide. Thus, the present MS data indicated that this 55 kDa fragment was not cleaved from the activated proteoform of F13A1 in platelets (Additional le 5: Fig. S4a).
A long-standing hypothesis is that activated F13A1 is inactivated by the brinolytic system (38). To evaluate these in-vitro observations in the clinical situation of this study, the levels of the 55 kDa fragment of F13A1 in platelets of patients with lung cancer and matched controls were correlated with corresponding plasma levels of the brinolysis product, D-dimer. This degradation product of cross-linked brin is a routine established indicator for VTE and also re ects cancer-associated thrombosis risk (39). The haemostatic biomarker D-dimer correlated signi cantly (r s = 0.599; p unadj = 1.40 − 4 with the 55 kDa fragment of F13A1 in the platelets of patients with lung cancer and matched healthy controls (Fig. 4d). However, D-dimer levels were signi cantly increased in patients with brain and lung cancer by an effect size (Cohen's D) of 1.02 and 0.94 respectively.
Interestingly, the lung-cancer-related increase of the 55 kDa degradation product from F13A1 had a stronger effect size with 1.15 compared to the plasma thrombose biomarker D-dimer with 0.94.
Because in clinical studies F13A1 are more frequently investigated in plasma, its enzymatic activity was also analysed in the plasma of recruited cancer and matched healthy controls. Plasma F13A1 activity was unchanged in patients with lung cancer and was signi cantly decreased in patients with brain cancer (Fig. 4e).
Chaperones from the KEEG pathway "protein processing in endoplasmic reticulum" are elevated in platelets of patients with lung cancer One-way ANOVA-based differential 2D-DIGE analysis identi ed signi cantly changed abundance of two P4HB, ve CALR spots, and one spot of HSPA5 in platelets between patients with lung and brain cancer and matched healthy controls (Table 2). Posthoc contrast analysis showed that all these ER chaperones are increased in platelets of patients with lung cancer. Twodimensional WB analysis con rmed MS identi cations of these CALR and HSPA5 spots and identi ed three more CALR and two more HSPA5 proteoforms (Fig. 5a), which were also all increased in patients with lung cancer. The ER proteins CALR and HSPA5 are essential chaperones for glycosylation and secretion of proteins and this function is described for the nal biosynthesis of coagulation factors (33,46,47). Besides F13A1, platelets are also a potential source for coagulation factors e.g. FV (43) and FVIII (48-50). Hence, the FVIII activity, an important activation marker of the haemostatic system, was signi cantly increased in plasma of patients with lung cancer (Table 1; FC = 1.75; p = 0.0002) and to a lesser extent in patients with brain cancer (Table 1;  to the interactome pathway "protein processing in ER" (Fig. 6d). Remarkably, the substrate for F13A1, plasma brinogen, had the strongest correlation with the 55 kDa fragment of brin stabilising factor F13A1 of the platelet proteome (Fig. 6e). The platelet 55 kDa fragment of F13A1 had also one of the strongest correlation with the acute phase proteins CRP (Fig. 6c), which indicates the strong interplay of in ammation with essential steps of haemostasis, which nally may result into immunothrombosis. In contrast to the strong association of plasma FVIII with the pathway "protein processing in ER" in platelets, D-dimer, brinogen and CRP correlated signi cantly with proteins from platelet degranulation and platelet activation pathways such as lamin A Page 10/27 (FLNA) and the integrins ITGA2B and ITGB3 (Fig. 6a, 6c and 6e). No strong correlations of the platelet proteome were found with the plasma laboratory values of prothrombin fragment 1 + 2, antithrombin III and plasminogen activator inhibitor 1.
For an additional functional proof, for the speci city of lung-cancer-related platelet protein pro les to platelet activation, the 617 2D-DIGE quanti ed platelet proteoforms were also correlated to plasma levels of sCD62P, reliable biomarker for platelet activation in-vivo (55). Again, platelet degranulation and the ER protein processing pathways were signi cantly enriched via the sCD62Pcorrelating platelet proteins by STRING interactome analysis (Fig. 6b). All platelet proteins correlating signi cantly with haemostatic plasma laboratory parameters are indicated in additional le 10: Fig. S8.
The MS identi cations of ITGA2B, ITGB3 and TLN1 were also veri ed by 2-D WB analysis. These immunological platelet proteome analyses revealed also several additional proteoforms of the respective proteins (Additional le 6: Fig. S5).
Associations in the platelet proteome with lung cancer and risk of death SERPINB1 abundance was independent of mortality in patients with brain cancer (Additional le 7: Fig. S6).
For a small explorative follow-up case-report, we performed 2D-DIGE analysis of platelet samples from four patients with lung cancer after three and six months to evaluate how the identi ed changes of protein pro les in lung-cancer are associated with patient outcome. These patients were given chemotherapy within 3 months of the rst blood draw. One of the traced patients had a PE at two months after the baseline and died 5.5 months after study inclusion.
The platelet SERPINB1 level of this deceased patients stayed decreased (SA = 1.10) after three months compared to the corresponding baseline value (SA = 1.17) and the mean baseline value of all deceased patients (SA mean = 1.10; n = 11) with lung cancer (Additional le 8: Fig. S7). After three months, the SERPINB1 levels decreased also in the surviving patients but returned almost to baseline values after six months and thus almost aligned with the mean values of healthy controls.
Quite similar patterns in the follow up tracing were observed for the lung cancer related proteins ITGA2B, PH4B and 55 kDa F13A1. Thus, these lung cancer related alterations were most pronounced in the deceased patient, who suffered from PE two weeks before the second blood sampling (3 months value). In the surviving patients, the lung cancer-related changes were also forti ed after 3 months and returned almost to baseline values after 6 months (Additional le 8: Fig. S7). However, exclusively the abundance of SERPINB1 was already signi cantly associated with patient outcome at baseline.

Discussion
Continuous platelet activation may contribute to increased VTE risk in patients with brain and lung cancer. To the best of our knowledge, this is the rst study comparing the platelet proteome of these patients and matched healthy controls to elucidate prothrombotic pathways. These investigations showed a predominant in uence of lung cancer on the peripheral platelet proteome compared to brain tumours. Altered platelet proteins levels of patients with lung cancer could be linked to enhanced platelet degranulation, changes in conversion of F13A1, and upregulated protein processing in ER presumably re ected by increased plasma levels of FVIII. The identi cation of these protein changes in platelets expands our knowledge of differential molecular mechanisms for the prothrombotic state in lung and brain cancer.
Interestingly, no brain cancer-related changes were detected by the 2D-DIGE analysis in the platelet proteome. However, in a previous study we identi ed that high podoplanin expression of brain tumours is functionally linked with increased VTE risks of this cancer type (56). For the functional examination of this assumption, it is shown that podoplanin on primary human glioblastoma cells induces platelet activation via the speci c binding to their C-type lectin receptor type 2 receptors. Since increased intra-tumoral platelet aggregates correlates with VTE events in patients with brain cancer (56), it can be reasoned that platelets are caught by podoplanin of the brain tumour and related proteome changes are not detectable in peripheral blood platelets.
For pathophysiological explorations of platelet proteome changes in patients with lung cancer we concentrated on four candidates described to be involved in the processing of coagulation proteins, these are, F13A1 and the ER proteins P4HB, CALR and HSPA5. An altered conversion of the nal clotting factor F13A1, allocated to the pathway "platelet degranulation", was detected by elevated levels of a 55 kDa cleavage product from F13A1 in the platelets of patients with lung cancer. The brin stabilizing factor F13A1 is well-known to circulate in the plasma as a heterotetrameric protein complex, consisting of a dimer of the catalytic F13A1 subunits and a dimer of the carrier/inhibitory FXIIIB2 subunits. The transglutaminase F13A1 is a critical determinant of venous thrombus characteristics such as thrombus size, stability and red blood cell retention (57,58). Platelets contain only F13A1. Although, the concentration of F13A1 is around 150-fold higher in platelets compared to plasma (59) its functions in platelets are quite unknown. Only recently it was shown that platelet surface exposed F13A1 contributes to thrombus stability whereas the secreted F13A1 from platelets is functionally irrelevant (60). However, the enzymatic activity of F13A1, exposed on platelets, is temporally limited with a maximum after 5 minutes and completely declines after one hour (60). The However, no signi cant change of F13A1 enzymatic activity could be measured in platelets and plasma of patients with lung cancer. Although F13A1 abundance was unchanged in platelets of patients with brain cancer, a signi cant decrease of enzymatic F13A1 activity was found in the plasma of patients with brain cancer. Decreased plasma levels of F13A1 were previously observed in plasma of patients with thrombosis history with or without cancer. It is assumed that this decrease is caused by an increased consumption of F13A1 due to its inactivation (64). In the actual study the increased levels of the precursor and the 55 kDa fragment from F13A1, may also indicate an increased turnover of F13A1 in the platelets of patients with lung cancer.
Platelets are also of functional importance for the progression of cancer and metastasis formation (65). The observed increased levels of F13A1 and changed processing of F13A1 in platelets of patients with lung cancer may also play a role in metastasis. In Elevated levels of ER chaperones P4HB, CALR and HSPA5 indicated the induction of the ER unfolded protein response (UPR) in platelets of patients with lung cancer. An increased abundance of these ER proteins was already previously observed in platelets of LA patients with thrombosis history (16).
Interestingly, the UPR in ER is also associated with malignant transformation of cancer cells (70) as well as with a prothrombotic in uence of pancreatic cancer. In this particular study, it was shown that proteins of the UPR in cancer are released from pancreatic cancer cells by extracellular vesicles. Increased levels of the same UPR protein pro les are also detected in the plasma of patients with pancreatic cancer and thrombosis history. These observations indicate a mechanistic link between tumour progression and cancer-associated thrombosis (71). That group also showed that inhibiting enzymatic PDI activity in plasma and platelets by the avonoid isoquercetin reduced the hypercoagulability of advanced cancer patients (72) as well as that of LA patients (43). Remarkably, in patients with lung cancer as well as in LA positive patient with thrombosis history (16) we identi ed a highly signi cant increase of the PDI family member, P4HB, in platelets, which also manifest this ER chaperone as a highly purposive antithrombotic drug target.
The ER proteins HSPA5 and CALR mediate to a great part glycosylation and folding of secretory proteins. Interestingly the function of these ER proteins and the induction of the UPR in the ER are extensively investigated and described for the industrial and therapeutic production of FVIII (46, 73,74). Increased levels as well as increased activity of FVIII in plasma are generally associated with high VTE risk (75) and predict VTE in cancer patients (51). In the current study FVIII was also signi cantly increased in the plasma of patients with lung cancer and showed a predominant correlation with the ER proteins HSPA5 and CALR in comparison to all other proteomic investigated platelet proteins. This induction of the UPR and correlations of FVIII with P4HB, HSPA5 and CALR levels may also take place systemically in patients with lung cancer. Thus, platelets may be a peripheral indicator for enhanced glycosylation and secretion from e.g. liver sinusoidal endothelial cell (76). The lung-cancer-related upregulation of ER chaperones in platelets corroborate a functional association of the UPR with FVIII production and thrombosis risks.
The abundance of both integrins of the brinogen receptor integrin αIIbβ3, ITGA2B and ITGB3, were signi cantly reduced in platelets of patients with lung cancer. These glycoproteins are the two integral parts of the key adhesion receptor of platelets and play a crucial role in brinogen-mediated thrombus formation by promoting platelet adhesion. Diminished levels of these integrins in platelets do not functionally correspond to an expected increased adhesion and aggregation of platelets in patients with high thrombosis risks. The decrease of ITGA2B and ITGB3 levels in platelets during this prothrombotic condition of patients with lung cancer and LA with thrombosis history (16) may be caused by continuous shedding of extracellular vesicles, which was recently also described as "membrane pearling" of platelet pseudopodia (77). Platelet extracellular vesicles have in relation to their whole protein content a higher load of ITGA2B and ITGB3 than their corresponding resting platelets (78). Thus, these integrin levels may be declined in exhausted patient platelets, because of an extensive release of extracellular vesicles. In fact, an increase of platelet released extracellular vesicles is observed in plasma of patients with cancer (79) and LA (80).
Accordingly, lower levels of ITGA2B and increased levels of the ER proteins P4HB, CALR and HSPA5 in patients with lung cancer and LA suggest some common affected pathways in their prothrombotic pathology. In contrast to that, the altered proteoforms of F13A1 in platelets appeared to be speci c for patients with lung cancer. The data of these two platelet proteomics studies (16) are highly comparable since exactly the same methodology of platelet isolation and 2D-DIGE proteome analysis were used.
Altogether, our study has both limitations and strengths. The selected proteomics technology of 2D-DIGE in the pH range 4-7, does not cover the whole possible pH range of 3-10. Low-abundance and high molecular weight proteins are not captured by 2Dbased analysis meaning a lower sensitivity as compared to MS-based shotgun analyses. Hence, the probability is high that we did not detect all protein changes in platelets of patients with lung and brain cancer. However, there is not a single proteomics technology nowadays that can analyse the whole proteome including all PTMs from a biological sample. PTMs are particularly essential for proteomic changes in haemostasis, as primary and secondary haemostasis are tightly coordinated by the PTM "enzymatic cleavage" of coagulation and other factors. Thus, two-dimensional gel electrophoresis should be more versatile to study the prothrombotic phenotype of platelets than the currently mainly applied "bottom-up" shotgun proteomics. Conventional MS-based shotgun proteomics approaches cannot capture the information of regulatory variation of different proteoforms to each other, due to the necessary pre-analytical proteolytic digestion of biological samples (81). Accordingly, 2D-DIGE analysis had an analytical advantage in the concrete clinical proteomics study. For example, it would not have been possible to reveal the lung-cancer-related changes of F13A1 processing by MS based shotgun analysis, because the preanalytical digestion of the sample would have eliminated the information of different pIs and molecular weights of the respective F13A1 proteoforms.
From the clinical perspective, our study is rather explorative with the inclusion of a relatively small number of patients and matched healthy controls. Due to the limited sample size and number of VTE events it was not possible to perform well-founded statistically subgroup analysis to evaluate comprehensively the in uence of VTE or mortality on the investigated platelet proteome.

Conclusions
Our study delivers novel insights into the differential behaviour of platelets in patients with different types of cancer. These data provide the rst evidence that alterations of the platelet proteome are more pronounced in lung cancer compared to brain cancer.
Lung-cancer related changes as increased platelet levels of P4HB from the PDI family underscores the importance of this antithrombotic drug target for the hypercoagulable state in certain cancer types [81]. F13A1 might be also involved in development of VTE, as well as in lung cancer progression, and especially this patient group may bene t in multiple ways from antithrombotic treatment such as FXIII inhibitors [82]. Our clinical platelet proteomics study con rms also previously characterised deregulations of excessive malignancies-related platelet activation in lung cancer and furthermore provides novel insights in the mechanisms of thrombosis in lung cancer and its correlation with in ammation. These potential antithrombotic drug targets in platelets should be further investigated in cancer research. VTE during follow up, n (%) n.a. n.a n.a 2 (9.1) 2 (10.5) PE n.a n.a n.a 1 (5) 2 (10.5) DVT n.a n.a n.a 1 (5) n.a.
Deaths during follow up, n (%) n.a n.a n.a 8 (36.4) 10 (52.6) Table 2 Page 20/27   Figure 1 2D-DIGE-based proteome analysis of platelets from patients with brain and lung cancer compared to controls. This representative 2D-DIGE image shows all one-way ANOVA ltered signi cantly altered protein spots between patients with lung (n = 19) and brain (n = 22) cancer compared to healthy controls (n = 41). A total of 36 µg (12 µg sample Cy3-, 12 µg sample Cy5-, and 12 µg IS Cy2labeled) platelet protein extracts were separated according to the isoelectric point (pI) in the pH 4-7 range (separation distance 24 cm) and the molecular weight (MW, separation distance 20 cm). Protein spots identi ed by MS are circled and labeled with their corresponding gene name and spot numbers are given in Table 1. Protein spots of interest were selected according to (a) protein spots matched >95% of all 2D-DIGE gels and (b) FDR-corrected one-way ANOVA p-value <0.05. Abbreviations: MW -molecular weight; kDa -kilodalton; pI -isoelectric point; 2D-DIGE -two-dimensional differential in-gel electrophoresis; IS -internal standard; MS -mass spectrometry Functional association analysis of lung cancer-related platelet proteins. Network and enrichment analysis shows top pathways obtained upon entering the set of signi cantly lung cancer-related platelet proteins in STRING database analysis tool [25]. The type of interaction is indicated by colored linear slopes; pink: experimentally determined, blue: from curated databases. The enrichment graphs depict the most signi cantly enriched GO Biological Process with platelet degranulation in red and two most signi cant KEGG pathways, in cobalt blue, protein processing in endoplasmatic reticulum, and in green, platelet activation. The proteins highlighted in grey were not signi cantly associated to these quoted functional networks. association was assessed by a Spearman´s Rank correlation coe cient (rs). (e) F13A1 activity levels in plasma and (f) platelets in patients with lung and brain cancer and matched healthy controls. Protein levels were depicted as single values and mean.*p<0.05, ***p<0.001. Abbreviations: 2D-DIGE -two-dimensional differential in-gel electrophoresis; WB -western blot; pIisoelectric point; MW -molecular weight; kDa -kilodalton; con dence interval -CI; BC -patient with brain cancer; LC -patient with lung cancer; C -healthy control . The solid straight line represent best tted model with 95% con dence interval (CI) as dashed lines. Abbreviations: WB -Western blot; pI -isoelectric point; MW -molecular weight; kDa -kilodalton; con dence interval -CI; BC -patient with brain cancer; LC -patient with lung cancer ; Chealthy control Figure 6 Functional relationships of the platelet proteome with haemostatic plasma laboratory parameters. The standardised abundance of the 617 included platelet protein spots from 2D-DIGE platelet proteome analysis were correlated with the plasma levels of (a)