TRC051384

Hsp70 protects from stroke in atrial fibrillation patients by preventing thrombosis without increased bleeding risk

ABSTRACT
Aims: Atrial fibrillation (AF) is a major risk factor for cardioembolic stroke. Anticoagulant drugs are effective in preventing AF-related stroke. However, the high frequency of anticoagulant-associated major bleeding is a major concern. This study sought to identify new targets to develop safer antithrombotic therapies.Methods and Results: Here, microarray analysis in peripheral blood cells in eight patients with AF and stroke and eight AF subjects without stroke brought to light a stroke related gene expression pattern. HSPA1B, which encodes for heat-shock protein 70 kDa (Hsp70), was the most differentially expressed gene. This gene was downregulated in stroke subjects, a finding confirmed further in an independent AF cohort of 200 individuals. Hsp70 knock-out mice subjected to different thrombotic challenges developed thrombosis significantly earlier than their wild-type (WT) counterparts. Remarkably, the tail bleeding time was unchanged.Accordingly, both TRC051384 and tubastatin A, i.e. two Hsp70 inducers via different pathways, delayed thrombus formation in WT mice, the tail bleeding time still being unaltered. Most interestingly, Hsp70 inducers did not increase the bleeding risk even when aspirin was concomitantly administered. Hsp70 induction was associated with an increased vascular thrombomodulin expression and higher circulating levels of activated protein C upon thrombotic stimulus.Conclusions: Hsp70 induction is a novel approach to delay thrombus formation with minimal bleeding risk, and is especially promising for treating AF patients and in other situations where there is also a major bleeding hazard.

INTRODUCTION
Atrial fibrillation (AF) is the most common cardiac arrhythmia, with an overall prevalence of 5.5%, which increases to 17.8% in individuals over 851. AF is a major risk factor for cardioembolic stroke since it increases the overall risk five-fold2, 3. Properly dosed anticoagulation with warfarin or other anti-vitamin K drugs is effective in preventing AF- related stroke, reducing the risk by two thirds compared with no therapy4, 5. However, the high frequency of major bleeding associated with warfarin use is a major concern1. The new thrombin/factor X activated (FXa) direct inhibitors, similar to warfarin in preventing cardioembolic stroke, raised great expectations of an improvement in the control of the anticoagulant regimen in AF patients. However, their associated major bleeding complications are still considerable, and remain the biggest challenge, particularly when we consider that the duration of anticoagulant treatment in AF patients is indefinite1.Furthermore, one third of anticoagulated AF patients are concomitantly treated with platelet antiaggregant drugs and thus are exposed to an additionally increased major bleeding risk6. Additionally, the threat of major bleeding forces physicians to discard, or stop, any anticoagulant treatment in one third of AF patients, who are thus exposed to an unacceptably high risk of stroke7. Thus, the identification of new molecular targets whose manipulation dissociates thrombosis from bleeding is warranted to develop new antithrombotic drugs with minimal bleeding risk. Here, we report that heat shock protein 70 (Hsp70) plays an unexpected antithrombotic role in AF patients. Most interestingly, we provide evidence in favour of using Hsp70 inducers to prevent thrombotic events with minimal bleeding risk.

For the initial study, i.e. the discovery population, subjects were recruited from the Morales Meseguer Hospital (Murcia, Spain). Demographic variables at the moment of blood withdrawal included age, sex, AF starting date, stroke date, CHAD (CHADS2 index minus stroke punctuation), total leukocyte, neutrophil, monocyte, lymphocyte, platelet counts, and international normalized ratio (INR). A total of 16 non-valvular AF patients were recruited for the study, among whom 8 had suffered a cardioembolic stroke between 2 and 11 years previously (stroke group). Stroke and non-stroke patients were matched by CHAD index and sex. All patients were on anticoagulant treatment with acenocoumarol, with an INR between 2 and 3 at the time of blood withdrawal. AF was diagnosed by electrocardiography and lasted more than three months. Cardioembolic stroke was diagnosed clinically by imaging techniques (magnetic resonance imaging or X-ray computed tomography).Patients were excluded from the study if they met any of the following criteria: carotid artery lesion occluding more than 50% of the lumen vessel diameter in the side of the infarction, cancer in progress, leukocytosis (more than 7,000 cells/ml), leukopenia (less than 3,500 cells/ml), history of venous thromboembolism in the last three months, acute coronary syndrome, infection, autoimmune disease or surgery. Renal failure (creatinin value more than double the normal value), oral contraceptive use, hormonal therapy, and corticoid consumption were also exclusion criteria.

A second, larger, population of non-valvular AF patients was used to validate the results found in the discovery cohort. In order to obtain reliable results, we ensured that the sample size of the validation population was large enough. For this purpose, we took as our basis the data obtained in the initial expression array: when we compared extreme tertiles, we observed an odds ratio (OR) for stroke of 0.25. Thus, being conservative and assuming an OR of 0.25 for the comparison between extreme quartiles, a proportion of exposure among controls of 0.25, a 1:1 case:control ratio and a statistical power of 0.8, we would need 180 participants for the independent AF validation cohort. Due to the limited number of subjects in the discovery cohort, we increased the sample size of the validation cohort to 200 participants. Thus, 204 consecutive patients with permanent non-valvular AF were recruited from the University Hospital of Salamanca, Spain, between May 2009 and February 2010. Inclusion and exclusion criteria were the same as those used for the initial study. 102 patients had a history of one episode of cardioembolic stroke while 102 did not. Four subjects (one case and three controls) were discarded because we failed to obtain RNA from them. The study was approved by the Institutional Review Boards of the Hospital of Salamanca and University of Navarra, all patients gave informed consent to the study and was performed conform the declaration of Helsinki.

Whole blood was collected from each subject into a PAXgene tube (BD, Franklin Lakes, NJ, USA) and RNA was isolated according to the manufacturer´s protocol using the PAXgene blood RNA kit (Pre-AnalytiX, Feldbachstrasse, Switzerland). Microarray Data Analysis 2 μg of RNA from the patients in the discovery population were labeled, hybridized, and scanned according to standard protocols (Affymetrix, Santa Clara, CA, USA). Affymetrix Human U133 Plus 2.0 Arrays were used. Microarray data normalization was performed using Robust Multi-array Average software (RMA)8, and after quality assessment and outlier detection with R/Bioconductor, two non- stroke patients were considered outliers and discarded. A filtering process was performed to eliminate low expression probesets. Applying a criterion of expression value greater than 16 in 22% of the samples, 41,658 probesets were selected for statistical analysis. Linear Models for Microarray Data9 was used to find out the probesets with significant differential expression. In the linear model several factors were taken into account, including the group of the samples (stroke or non-stroke), the hybridization batch effect and the percentage of polymorphonuclear cells. Probesets were selected as significant using a p-value cut off p<0.05. Prediction Analysis for Microarray10 was used to determine the minimum number of the differentially expressed genes that classified stroke and non-stroke patients correctly. Differentially regulated functions were analyzed by Ingenuity Pathways Analysis (IPA 7.6; Ingenuity Systems Inc., Mountain View, CA, USA) (see expanded Methods section in the Supplementary material online). We deposited our data in the NCBI Gene Expression Omnibus (GEO) database (www.ncbi.nlm.nih.gov/geo), which are available under accession number GSE66724. Analysis of HSPA1B expression One µg of each blood RNA sample was reverse-transcribed and the PCR was performed utilizing an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA, USA), according to the manufacturer’s instructions. The PCR primers and probe for HSPA1B were from Applied Biosystems (TaqMan, Hs01040501_sH). The relative quantity of mRNA was normalized to the relative quantity of the endogenous control, which was mitochondrial 2.4-dienoyl-CoA reductase (TaqMan, Hs00154728_m1) one of the most suitable house-keeping genes for samples from human peripheral blood cells11. The conditions for the RT-PCR thermal cycling were preheating at 60 ºC for 20 seconds (s) followed by 40 cycles of shuttle heating at 95 ºC for 1 s, then annealing at 60 ºC for 20 s. All samples were run in triplicate. All samples exhibited cycle threshold (Ct) differences lower than 0.5 cycles among triplicates. When we measure HSPAIB expression in cell cultures, human glyceraldehyde-3-phosphate dehydrogenase (Hs03929097_g1, Applied Biosystems) was used to normalize results (see expanded Methods section in the Supplementary material online). All mouse procedures were approved by the Institutional Animal Care and Use Committee (133/12) at the University of Navarra. Knock-out mice for HSPA1A and HSPA1B (B6;129S7-Hspa1a/Hspa1btm1Dix/Mmcd) (HSPA1A/B KO) were purchased from the Mutant Mouse Regional Resource Centers (University of California, Davis, CA, USA). B6129S2F1 mice (WT) (Harlan Interfauna Iberica S.A., Barcelona, Spain), belonging to the strain used to originate the HSPA1A/B KO mice were used for control purposes. CD-1 mice (Harlan) were used for the assay of different compounds. 7-8 week-old female mice were used for all the assays. All animal experiments were performed conform the guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes.Rose Bengal-laser and ferric chloride induced carotid artery thrombosis models were performed as described elsewhere with minimal changes12 (see expanded Methods section in the Supplementary material online).The lung thromboembolism model was performed in anaesthetized mice by an i.v. injection of a mixture of 0.8 mg/Kg collagen (Roche, Mannheim, Germany) and 60 µg/Kg epinephrine (Sigma-Aldrich). Animals that remained alive 30 min after challenge were considered survivors.Mice were anaesthetized as indicated in murine models of thrombosis and the distal 3-mm segment of the tail was removed with a scalpel. Bleeding was monitored by absorbing the bead of blood with a filter paper at 15 s intervals without touching the wound until haemorrhage cessation. Bleeding was stopped manually if it continued for more than 30 min. Compounds administration TRC051384 9.00 mg/Kg in saline solution, dose pattern chosen according to our TRC051384-dependent induction experiments in culture and mice as well as to previous literature13 were i.p. administered 3 and 1 hours before the start of the experiments. 20 mg/Kg tubastatin A (Sigma-Aldrich) in saline solution (10% DMSO final concentration), dose pattern chosen to have a two-fold increase of in vivo aortic HSPA1B induction (n =5 mice) relative to non-treated animals (n=6) [median (interquartile range), 1.91 (1.54-3.15) RQ], similar to that obtained with TRC051384. Aspirin (Sigma-Aldrich) (300 mg/Kg in 5% sodium bicarbonate) and rivaroxaban (Selleckchem, Houston, TX, USA) (3 mg/Kg in DMSO) or their appropriate vehicles were administered 24 hours i.p. and 1 hour i.v. respectively before the start of the tail bleeding time assay. The expression of the genes coding Hsp70 (HSPA1B and HSPA1A), endothelial protein C receptor, thrombomodulin, tissue factor, tissue factor pathway inhibitor, tissue-plasminogen activator, plasminogen activator inhibitor-1 and endothelial nitric oxide synthase was analyzed by RT-PCR using mRNA from murine aortic vessel tissue(see expanded Methods section in the Supplementary material online).Hsp70 and thrombomodulin analyses in protein extracts by immunoblotting Proteins were obtained from cell culture or aortic vessel lysates (see expanded Methods section in the Supplementary material online). 10 µg proteins were subjected to Western blot analysis using a rabbit anti-Hsp70 polyclonal antibody (pAb) (Thermo Fisher Scientific, Waltham, MA, USA; PAI-37842) or a rat anti-thrombomodulin pAb (R&D Systems, Minneapolis, MN, USA; MAB3894), and the appropriate secondary antibodies [goat anti-rabbit pAb (Dako, Glostrup, Denmark; P0448) and rabbit anti-rat pAb (Dako; P0450)]. Rabbit anti-β-actin pAb (R&D Systems; 4967) and its secondary goat anti-rabbit pAb (Dako; P0448) were used for loading control purposes.The effect of Hsp70 induction on the in vivo APC generation ability was studied as described elsewhere.14 In brief, since APC circulating levels in the absence of stimuli inducing thrombin generation are too low for inter-group comparison purposes, mice were i.v. injected 352 pmol/Kg FXa and 541.5 nmol/Kg phosphatidylcholine-phosphatidylserine (PCPS) vesicles to allow thrombin formation and subsequent protein C activation. 10 min thereafter, blood was withdrawn, and plasma immediately obtained to measure APC levels using a specific chromogenic assay. Blood was withdrawn from mice treated with vehicle or TRC051384 (n=18 each group), and pools from three different animals were made. Each pool was diluted 2:1 (v/v) with platelet additive solution SSP+ (MacoSpania, Madrid, Spain). Platelet-rich plasma (PRP) was obtained and light transmittance aggregometry was performed using a standard aggregometer (Aggrecorder II, Menarini Diagnostics, Florence, Italy) set at 37 ºC and 1,000 rpm. The agonists and doses used were 10 µM ADP (Menarini Diagnostics), 10 µg/ml collagen (Menarini Diagnostics), and 200 µM mouse-specific protease-activated receptor-4 (PAR4) activating peptide (Sigma-Aldrich).Blood from mice was extracted by cardiac puncture, mixed with sodium citrate, and centrifuged at 3,000 rpm for 10 min to obtain plasma. Protrombin time (PT) and activated partial thromboplastin time (APTT) were determined using an automated blood coagulation system (BSC Coagulation System, Siemens, Berlin, Germany). In the validation study, comparison between stroke and non-stroke patients for continuous and categorical variables was performed using the Student t-test for independent samples and the Chi square test, respectively. A non-conditional logistic regression model was used to evaluate the risk of stroke associated with HSPA1B expression levels. We tested the goodness of fit with the Hosmer-Lemeshow goodness-of-fit statistic. The main independent variable was the level of HSPA1B categorized into quartiles according to the distribution in the control, i.e. non-stroke, group. We fitted univariate and multivariate models, adjusting for traditional risk factors for cardioembolic stroke. To assess the p value for linear trend, the quartile specific median was assigned to each quartile and the resulting variable was treated as quantitative. Product-terms were introduced in the non-conditional logistic models to analyze interaction (effect modification) and their statistical significance was assessed with the likelihood ratio test.To assess the differences in the occlusion times, bleeding time, aortic tissue gene expression and apoptosis, the two side Mann-Whitney U test was used. The bleeding time in mice treated with ASA, vehicle, ASA plus TRC051384 or rivaroxaban was analysed using the two-side Kruskal-Wallis test followed by the Dunn post hoc test. The Kaplan–Meier curves significance was calculated by using the log-rank test. Correlation between HSPA1A and HSPA1B expression was examined by Spearman’s rho. All statistical analyses were performed with Stata 12.0 software (StataCorp, College Station, TX, USA). RESULTS We performed a discovery study in 16 non-rheumatic AF patients, eight of whom had suffered a cardioembolic stroke (see Supplementary material online, Table S1). Using RNA expression microarray technology we found that 282 probesets (214 genes) were able to differentiate stroke from non-stroke patients (see Supplementary material online, Table S2). The molecular signature was confirmed by hierarchical clustering (Figure 1). On the other hand, an Ingenuity Pathway Analysis (IPA)-based functional analysis showed that “Cardiovascular system development” and “Nervous system development” among physiological functions and “Cardiovascular disease” and “Neurological disease” among diseases were overrepresented (see Supplementary material online, Figure S1). These findings are consistent with the clinical features of patients, i.e. neurological and cardiovascular disorders, suggesting that the selection had been performed correctly and that it might therefore have biological relevance. High levels of HSPA1B are associated with low risk of stroke.Among the most differentially expressed probesets, HSPA1B showed the highest statistical significance (see Supplementary material online, Table S2). Our next aim was to confirm its differential expression using a larger, independent validation cohort of AF patients with or without cardioembolic stroke (see Supplementary material online, Table S3). We observed that HSPA1B expression levels were inversely associated with stroke in a dose-response manner after adjusting for CHAD (Congestive heart failure, Hypertension, Age, Diabetes), sex and leukocyte counts (Table 1). HSPA1A/B KO mice are prone to thrombosis Since HSPA1A and HSPA1B genes encode Hsp70, we considered that HSPA1A/B KO mice constitute a useful tool to study the influence of Hsp70 in thrombosis.16-18 These animals are healthy, reproduce normally and did not show any gross abnormality.18 HSPA1A/B KO animals and their WT counterparts were subjected to three different thrombosis models. First, the time until a thrombus formed in the laser light-exposed carotid artery of Rose Bengal- administered mice was significantly shorter in animals lacking Hsp70 (Figure 2A). When thrombosis was triggered by ferric chloride, vessel occlusion time was again significantly shorter in these animals (Figure 2B). Finally, when collagen and epinephrine were i.v. administered to cause a massive pulmonary thromboembolism HSPA1A/B KO mice displayed higher mortality rates (Figure 2C).In order to understand why HSPA1A/B KO mice were prone to thrombosis we explored their haemostatic system. Surprisingly, both HSPA1A/B KO (n = 4) and WT animals (n = 5) displayed similar PT [10.7 s (9.6-11.1) vs. 11.3 s (10.1-12.7), P = 0.289] and APTT [44.4 s (37.3-48.3) vs. 39.6 s (38.1-59.7), P = 0.999]. Furthermore, HSPA1A/B KO and WT mice did not show any difference in the tail bleeding time test (Figure 2D). Thus, it appears that HSPA1A/B KO animals do not display an obvious haemostatic alteration. Induction of Hsp70 in mice protects from thrombosis without increasing bleeding risk TRC051384 is a compound that induces HSPA1A/B expression and subsequent Hsp70 production via heat shock factor-1 (HSF1) activation.13 We first ensured that TRC051384 efficiently induced HSPA1B overexpression in cells of endothelial as well as of leukocyte origin (see Supplementary material online, Figure S2A-C). The doses needed to achieve this goal guided us to adjust the dose pattern in mice to obtain an increase in HSPA1B mRNA and Hsp70 levels in the vessel tissue (see Supplementary material online, Figure S2D-E). Since both HSPA1A and HSPA1B are equally induced by TRC051384 (see Supplementary material online, Figure S2F), we decided to measure HSPA1B from now on, assuming that expression of both genes would evolve in parallel. Overexpression of HSPA1B induced by TRC051384 significantly delayed the time to thrombus formation or death in the three thrombosis models (Figure 3A-C). Furthermore, TRC051384 was unable to delay the time to thrombus formation in HSPA1A/B KO mice (Figure 3D), suggesting that Hsp70 is directly involved in preventing thrombosis. Due to the possibility that Hsp70 may be expressed by human platelets, we decided to study the effect of the Hsp70 inductor TRC051384 on platelet aggregation and on coagulation tests. Remarkably, when we compared the haemostatic phenotype between TRC051384-treated and vehicle-treated mice we did not detect any significant difference either in coagulation or in platelet aggregation tests (Figure 4A-E). Most interestingly, there were no differences between TRC051384 and vehicle-treated mice in the tail bleeding time (Figure 4F). Since the bleeding risk becomes higher when patients on anticoagulant therapy are subjected to antiplatelet therapy, we explored the effect of Hsp70 induction on the bleeding tendency induced by aspirin. As expected, tail bleeding dramatically raised when aspirin and the FXa inhibitor rivaroxaban were administered simultaneously. However, TRC051384 did not lengthen the bleeding time further than aspirin did when these two drugs were given together (Figure 4G). To obtain additional confirmation of this unexpected role played by Hsp70, i.e. prevention of thrombus formation with no concomitant bleeding risk, we induced its expression using an alternative pathway. Tubastatin A promotes HSPA1A/B expression via inhibition of histone deacetylase 6 (HDAC6).19-21 Once we observed that tubastatin A stimulated Hsp70 production in our model (Figure 5A), we administered it to mice which were subsequently exposed to the same thrombotic stimuli used above. Tubastatin A exerted an antithrombotic effect, similar to that observed with TRC051384, in the three thrombosis models (Figure 5B- D). Importantly, tubastatin A treatment did not increase bleeding either, as the tail bleeding time was again unaltered upon treatment (Figure 5E).Thrombomodulin and APC are involved in the antithrombotic role of Hsp70.We then focused on searching for clues about the pathways through which Hsp70 exerts its antithrombotic effect. We studied the effect of Hsp70 induction by TRC051384 treatment on a series of haemostasis-related genes, whose expression does not influence the coagulation tests that we had used. Interestingly, we found no differences in vascular tissue expression of any of these genes except THBD (Figure 6A and see Supplementary material online Figure S3). The product of THBD, i.e. thrombomodulin, was also increased in vascular tissue (Figure 6B). Interestingly, HSPA1A/B KO mice did not display THBD induction upon TRC051384 treatment (Figure 6C), which suggests that Hsp70 was ultimately responsible for the effect. Since thrombomodulin is needed for APC generation, it was not surprising that the animals treated with TRC051384 did exhibit significantly higher circulating APC levels upon prothrombotic stimulus exposure (Figure 6D). DISCUSSION AF patients are at risk of suffering a stroke, and require preventive treatment.22 New anticoagulants, i.e. thrombin and FXa direct inhibitors, have undoubtedly improved the efficiency of antithrombotic therapy.23 However, the risk of severe bleeding complication remains the main caveat associated with antithrombotic drugs.24 Thus new treatments are needed to help fill this gap. Using a gene expression-based strategy, we identified a unique genetic profile among AF patients which differentiates those who have suffered a stroke from those who have not. HSPA1B was among the most differentially expressed genes, its lower levels in AF patients who had suffered a stroke being established in a large validation population.A cardioembolic stroke takes place when a thrombus formed in the left atrium travels to the brain and occludes a brain vessel.25 We therefore investigated the role of the product of HSPA1B gene, i.e. Hsp70, in thrombus formation.26 Since HSPA1A and HSPA1B genes encode Hsp70, we considered that HSPA1A/B KO mice constituted the best tool to study the influence of Hsp70 in thrombosis.16-18 Because the time needed to generate thrombosis upon a variety of stimuli is shortened in HSPA1A/B KO mice, ours is the first study showing that Hsp70 plays an unsuspected role in thrombosis prevention. The link between low levels of Hsp70 and thrombosis is particularly sound since its deficiency accelerates thrombus formation upon a wide variety of stimuli, in contraposition to many other molecules recently identified as potentially antithrombotic, which worked with one prothrombotic stimulus but not with others. For instance, peptidylarginine deiminase 4 deficiency protects against laser but no in ferrum chloride induced thrombosis. Our findings in the murine models of thrombus formation are consistent with those observed in patients. We have to mention that, recently, rats submitted to a heat shock were subsequently protected against thrombosis. This observation was consistent with ours, and suggests that Hsp70 plays a role in preventing thrombus formation.28 The retrospective nature of the human study precludes definitive conclusions to establish that Hsp70 downregulation actively plays a role in the pathogenesis of cardioembolic stroke. Nevertheless, we consider that the consistency of the results in humans and mice suggests a direct relationship between Hsp70 and cardioembolic stroke. Surprisingly, the classical coagulation tests are not influenced by the expression level of Hsp70. Most interestingly, the bleeding trend does not display an obvious alteration upon the presence or absence of Hsp70, since the canonical test to assess it, i.e. the tail bleeding time, was not influenced by Hsp70. These two findings prompted us to explore the possibility of developing an antithrombotic strategy with minimal bleeding risk based on the induction of Hsp70 expression.Hsp70 induction by TRC051384 protected mice from thrombosis in all cases when the same stimuli used above were applied. Importantly, the fact that tubastatin A, another Hsp70 inducer,19-21 also succeeded in delaying vessel occlusion further supports the involvement of Hsp70 in protection against thrombosis. Indeed, the promise of managing HSPA1A/B induction as a safer antithrombotic strategy is based on the fact that Hsp70 overexpression alters neither the haemostatic status nor the bleeding tendency. Anticoagulants increase the risk of bleeding, especially when administered with antiplatelet agents.5, 6 As expected, bleeding increased dramatically when mice were simultaneously given aspirin and anticoagulant treatment with the FXa inhibitor rivaroxaban. However, when concomitantly administered with aspirin, Hsp70 inducer TRC051384 did not lengthen the tail bleeding time more than aspirin alone did. These findings position Hsp70 induction as a priority in the search for safer long term antithrombotic treatments for AF patients which could be compatible with antiplatelet drugs. It had been reported that an increase in HSF1 activity in vascular endothelial cells corresponds to raised levels of thrombomodulin.29 It is therefore no surprise that, in our study, TRC051384, which is a HSF1 activator,13 increased the expression of THBD and, subsequently, the thrombomodulin levels in vascular tissue. Indeed, what is interesting is that TRC051384 was unable to exert such an effect in HSPA1A/B KO mice, thus positioning Hsp70 as likely responsible for increasing THBD expression. Thrombomodulin on the endothelial cell surface permits thrombin to activate protein C efficiently.30 It was therefore hardly surprising that, as a consequence of the increase in thrombomodulin, mice enhanced their ability to generate APC in the presence of stimuli leading to thrombin formation. APC is a molecule that exerts cytoprotective, antiapoptotic and anticoagulant functions.31 Furthermore, thrombomodulin has also been described as exerting direct antiinflammatory actions.32 Thus, the Hsp70-induced increase in thrombomodulin and APC could contribute to distort the crosstalk between coagulation and inflammation that facilitates thrombus formation. All these results encourage to study whether the currently available anticoagulants and antiplatelets agents modulate Hsp70 levels. We have not found any sound literature directly addressing this question and in future studies we should address it.This study has shown an unanticipated association between HSPA1B expression and cardioembolic stroke in AF patients. We suggest that HSPA1A/B product, Hsp70, actively contributes to prevent thrombus formation without increasing bleeding. As a consequence, the pharmacological development of agents to induce Hsp70 overexpression may be a good strategy to reach AF patients and others in need TRC051384 of a safe long term anticoagulant treatment.