Abstract
Venous Thromboembolism (VTE) remains a significant cause of morbidity
and mortality worldwide. Rivaroxaban, a direct oral factor Xa inhibitor,
mediates anti-inflammatory and cardiovascular-protective effects besides
its well-established anticoagulant properties; yet, these remain poorly
characterized. Extracellular vesicles (EVs) are considered
proinflammatory messengers regulating a myriad of (patho)physiological
processes and may be highly relevant to the pathophysiology of VTE. The
effects of Rivaroxaban on circulating EVs in VTE patients remain
unknown. We have established that differential EV biosignatures are
found in patients with non-valvular atrial fibrillation anticoagulated
with Rivaroxaban versus warfarin. Here, we investigated whether
differential proteomic profiles of circulating EVs could also be found
in patients with VTE.
We performed comparative label-free quantitative proteomic profiling of
enriched plasma EVs from VTE patients anticoagulated with either
Rivaroxaban or warfarin using a tandem mass spectrometry approach. Of
the 181 quantified proteins, 6 were found to be either exclusive to, or
enriched in, Rivaroxaban-treated patients. Intriguingly, these proteins
form a cluster tightly involved in negative feedback regulation of
inflammatory and coagulation pathways, suggesting that EV proteomic
signatures may reflect both Rivaroxaban’s anti-coagulatory and
anti-inflammatory potential. These findings may be of translational
relevance towards characterizing the emerging anti-inflammatory and
cardioprotective mechanisms associated with this therapy.
Venous thromboembolism (VTE), comprising deep vein thrombosis (DVT) and
pulmonary embolism (PE), affects nearly 10 million people worldwide each
year, with an estimated fatality rate of nearly 10%. [1] Numerous
risk factors are associated with increased risk of developing VTE such
as increasing age and obesity. [2], [3] Intriguingly, VTE
patients present with a systemic pro-inflammatory state, which
potentiates the risk of recurrence. [4] systemic anticoagulation is
the standard of care, both to treat acute VTE and to reduce the risk of
recurrence. Historically, vitamin K antagonists such as warfarin or
parenteral anticoagulation were the first in line treatment of choice.
In the last decade, however, direct oral anticoagulants (DOACs) have
emerged and are preferred due to their more predictable
pharmacokinetics, fixed dosing and decreased interactions with food and
co-prescribed medications. [5], [6] The direct factor Xa
inhibitor Rivaroxaban, has been found to be as effective/superior to
vitamin K antagonists in preventing recurrent VTE, with similar to
reduced rates of clinically significant bleeding events.
[7]–[9] Intriguingly, Rivaroxaban also demonstrates
anti-inflammatory properties beyond its anticoagulant effects; to date
this is primarily attributed to its inhibitory effects on
protease-activated receptors (PARs). [10]–[13]
Extracellular vesicles (EVs) are mediators of intercellular
communication, regulating a plethora of biological processes via
transfer of bioactive molecules such as proteins, lipids and miRNAs
[14]. EVs are highly implicated in pro-inflammatory diseases
[15]–[19] and can further augment inflammatory signalling by,
for example, activating the complement system or shuttling cytokines.
[15], [20], [21] EVs can also exacerbate endothelial
dysfunction, by inducing the expression of adhesion molecules and
inflammatory cytokines, collectively facilitating leukocyte recruitment
to the endothelium. [22]–[24] Importantly, the underlying
pro-inflammatory phenotype of VTE patients also manifests in increased
levels of circulating EVs [25]–[31] and levels of plasma EVs
have been implicated as diagnostic or prognostic biomarkers for VTE.
[32], [33] Despite the recognised importance of EVs in VTE, no
proteomic studies investigating differential EV cargoes have been
performed to date. The effects of anticoagulation with Rivaroxaban
compared to warfarin on proteomic signatures of circulating EVs
following acute VTE and during secondary prevention are currently
unknown.
Given this well-established pro-inflammatory state of VTE patients,
reflected by heightened levels of circulating EVs together with the
systemic anti-inflammatory properties reported for Rivaroxaban, we
hypothesised that Rivaroxaban’s anti-inflammatory properties may be
reflected in the proteomic profiles of circulating EVs. Here, we used
LFQ-proteomic profiling to compare the protein content of circulating
EVs in patients with single episode VTE anticoagulated with either
Rivaroxaban (n=6) or sex-, age-, and BMI-matched warfarin controls
(n=6).
We used LFQ-proteomic profiling to assess differences in the vesicular
proteome from VTE patients treated with Rivaroxaban and warfarin. Plasma
from 6 sex-, age- and BMI-matched individual biological donors for each
treatment were enriched for EVs by sucrose cushion ultracentrifugation
at an average 120,000 xg and 4 °C for 6 hours. 2 individual donors per
treatment cohort were randomly chosen to validate successful EV
enrichment by immunoblotting. 20 µg protein from these enriched EV
fractions were lysed and resolved on a 10% SDS gel. The presence of
transmembrane and soluble EV-associated proteins was investigated
according to the guidelines of the International Society of
Extracellular Vesicles (ISEV). [36] For a detailed description of
the methods, please refer to the online data supplement. Identification
of tetraspanins CD63 and CD81 as well as the soluble protein HSP70
confirmed successful enrichment of vesicles (Figure 1A). Detection of
albumin indicated co-isolation of plasma proteins (Figure 1A), although
study by Tóth et al. recently postulated that albumin could be
part of the newly established EV protein corona. [37] GO cellular
compartment analysis of all identified proteins using FunRich
furthermore revealed a highly significant association with the term
“exosomes” (p =1.74x10-27; Supplementary Table
1), further substantiating successful enrichment of circulating EVs.
Vesicle preparations were lysed, proteins precipitated, sequentially
digested with Lys-C and trypsin and analysed in technical duplicate in a
QExactive mass spectrometer. The raw data was searched against a human
FASTA using MaxQuant. For the search, a minimum of two peptides per
protein needed to be identified, minimal peptide length was set to seven
amino acids and a maximum of two miscleavages were allowed. The false
discovery rate for peptide and protein identifications in the initial
search was set to 0.01. Data filtering included removal of proteins of
the reverse data base, proteins only identified by site and common
contaminants. For statistical analysis of differential protein
expression, data was further filtered to only include proteins that were
identified in at least 50% of the patients in at least one treatment
cohort. Adopting this approach, 181 proteins were robustly identified
across at least 50% of patients in at least one treatment group
(Supplementary Table 2). Pearson correlation analysis of the protein
expression revealed robust correlation of the protein LFQ intensities
between the patients, averaging at 0.922 ± 0.036 and 0.948 ± 0.023 for
Rivaroxaban- and warfarin-treated patients, respectively (Supplementary
Table 3), suggesting low inter-donor variability in the protein
expression levels within our EV preparations.
Within the 181 identified proteins, one protein, vitamin K-dependent
protein Z (PROZ), was found exclusive to the Rivaroxaban-treated cohort
(Figure 1B; Table 2), an intriguing finding as low plasma levels of PROZ
have been associated with an increased risk of coronary and peripheral
artery disease. [38]–[41] Statistical analysis of the
expression level of the other shared 180 proteins using a student’st- test with a false discovery rate of 5% and a minimal fold
change (S0) of 0.1 (indicated by the black hyperbolic
lines in Figure 1C) revealed differential protein quantification for 5
proteins (Table 2). The expression level of the residual 180 proteins
remained statistically unchanged (Figure 1C, grey). The functional
activity of coagulation factors II (F2), X (F10), Protein S (PROS1) and
Protein Z (PROZ) requires vitamin K-dependent post-translational
γ-carboxylation, which is inhibited by warfarin. [42] In line with
the known mechanisms associated with warfarin therapy [43], proteins
upregulated in EVs from Rivaroxaban-treated patients were significantly
annotated to the GO biological pathway terms “Gamma-carboxylation of
protein precursors” (p =1.62x10-8),
“Gamma-carboxylation, transport, and amino-terminal cleavage of
proteins” (p =6.31x10-8), and
“Post-translational modification: gamma carboxylation and hypusine
formation” (p =8.51x10-4; Supplementary Table
1).
4 of the 5 proteins significantly upregulated in the Rivaroxaban cohort
(Figure 1C, red; Table 2), namely F2, F10, SERPINA10 and PROS1, are
involved in the regulation of coagulation. Warfarin as a vitamin K
antagonist inhibits the posttranslational modification of vitamin
K-dependent proteins such as F2, F10, PROS1 and PROZ. Although this
predominantly affects the functional activity of these proteins
[43], it might not be surprising that these proteins indicate higher
expression in Rivaroxaban-treated patients. SERPINA10 levels, however,
are not vitamin K-dependent. Strikingly, PROZ and SERPINA10 deficiencies
have been linked with an increased risk of developing venous thrombosis,
making our clinical proteomic findings incredibly relevant. Although
some human studies did not find a link, augmented arterial and venous
thrombosis was found in both PROZ- as well as SERPINA10-deficient mice.
[44]–[46] We further performed STRING network analysis. It
revealed that 5 of our 6 proteins found exclusive to or increased in
Rivaroxaban patients (F2, F10, PROS1, PROZ, and SERPINA10) formed a
distinct network collectively involved in blood coagulation (Figure 1D,
red, p =7.11x10-5). Physiologically, SERPINA10
alone is a potent inhibitor of FXIa [47] and complex formation of
SERPINA10 with its cofactor PROZ additionally initiates FXa inhibition.
[48] Our proteomic changes may therefore reflect the favourable
bleeding and anti-thrombotic mechanisms of rivaroxaban reported in
clinical trials. [7]–[9]
Besides their anticoagulant effects, PROS1, SEPRINA10 and PROZ can also
potentiate anti-inflammatory signalling. For instance, SERPINA10 was
recently identified to function as an acute phase protein [49] and
subsequently shown to downregulate the expression of pro-inflammatory
cytokines (TNFα, IL-1β, IL-6 and CCL3) independent of its anticoagulant
effects [50], however, the underlying mechanism remain to be
characterised. In line with these findings, low plasma levels of PROZ
correlated with increased levels of inflammatory cytokines (CRP and
IL-6) in a mouse model of sepsis [51], and in patients with acute
myocardial infarction and rheumatoid arthritis (RA) [52], [53],
respectively, potentially indicating an anti-inflammatory role for PROZ.
Furthermore, PROS1 mediated anti-inflammatory signalling via TAM (Tyro3,
Axl, Mer) receptors, negative regulators of inflammation. [54] PROS1
deficiency, on the other hand, increased lung inflammation in a rodent
model of lung cancer, which was rescued by addition of exogenous PROS1
in an NF-κB dependent manner. [55] Crucially, Rivaroxaban has
repeatedly been shown to inhibit pro-inflammatory signalling. While its
inhibitory effects have frequently been attributed to the direct
inhibition of FXa and downstream PAR signalling [56]–[60],
several studies indicated PAR-independent inhibition of NF-κB activation
[61], [62] and other pro-inflammatory signalling pathways.
[63]–[65] Collectively, our observed increases in PROS1,
SERPINA10 and PROZ levels in Rivaroxaban-treated patients may indicate a
pharmacologically ameliorated underlying pro-inflammatory state in VTE
patients.
In conclusion, we have used label-free quantification proteomic
characterisation to establish that patients with single-episode VTE
anticoagulated with Rivaroxaban compared with warfarin demonstrate
altered circulating EV profiles with distinct proteomic signatures. The
observed differences may indicate a pharmacologic reduction in the
pro-thrombotic state in combination with more stable coagulation in
Rivaroxaban-treated patients relative to warfarin. These results may be
of translational relevance towards characterising the emerging
anti-inflammatory and cardiovascular-protective characteristics
associated with Rivaroxaban therapy relative to warfarin.
[7]–[9]