Research ArticleCOAGULATION

Targeting STUB1–tissue factor axis normalizes hyperthrombotic uremic phenotype without increasing bleeding risk

See allHide authors and affiliations

Science Translational Medicine  22 Nov 2017:
Vol. 9, Issue 417, eaam8475
DOI: 10.1126/scitranslmed.aam8475

Striking a balance for anticoagulation

Patients with chronic kidney disease can present a therapeutic conundrum because they are not only at increased risk of blood vessel thrombosis but also more likely to experience bleeding complications when treated with anticoagulants. Shashar et al. examined the mechanism of thrombosis in mouse models of renal disease and found a potential therapeutic target in a protein called STUB1. STUB1 is a ubiquitin ligase that interacts with tissue factor, a vascular wall protein that triggers the coagulation signaling cascade. The authors demonstrated that increase in STUB1 can prevent thrombosis but does not prolong bleeding in mouse models of kidney disease, suggesting that this may be a viable approach to antithrombotic management of patients.


Chronic kidney disease (CKD/uremia) remains vexing because it increases the risk of atherothrombosis and is also associated with bleeding complications on standard antithrombotic/antiplatelet therapies. Although the associations of indolic uremic solutes and vascular wall proteins [such as tissue factor (TF) and aryl hydrocarbon receptor (AHR)] are being defined, the specific mechanisms that drive the thrombotic and bleeding risks are not fully understood. We now present an indolic solute–specific animal model, which focuses on solute-protein interactions and shows that indolic solutes mediate the hyperthrombotic phenotype across all CKD stages in an AHR- and TF-dependent manner. We further demonstrate that AHR regulates TF through STIP1 homology and U-box–containing protein 1 (STUB1). As a ubiquitin ligase, STUB1 dynamically interacts with and degrades TF through ubiquitination in the uremic milieu. TF regulation by STUB1 is supported in humans by an inverse relationship of STUB1 and TF expression and reduced STUB1-TF interaction in uremic vessels. Genetic or pharmacological manipulation of STUB1 in vascular smooth muscle cells inhibited thrombosis in flow loops. STUB1 perturbations reverted the uremic hyperthrombotic phenotype without prolonging the bleeding time, in contrast to heparin, the standard-of-care antithrombotic in CKD patients. Our work refines the thrombosis axis (STUB1 is a mediator of indolic solute–AHR-TF axis) and expands the understanding of the interconnected relationships driving the fragile thrombotic state in CKD. It also establishes a means of minimizing the uremic hyperthrombotic phenotype without altering the hemostatic balance, a long-sought-after combination in CKD patients.


Chronic kidney disease (CKD/uremia) imposes a strong and independent risk for both venous and arterial thrombosis in addition to conventional risk factors (14). CKD-associated thrombotic propensities introduce variability, which is not accounted for in the clinical thrombosis risk assessment nor targeted by contemporary antithrombotic/antiplatelet therapies. This variability contributes to their suboptimal efficacy in several clinical postinjury arterial thrombosis scenarios such as angioplasty, stenting, or vascular surgeries in CKD patients (2, 57).

Vessel wall factors are critical triggers for the postvascular injury thrombosis, where denuded endothelium and exposed vascular smooth muscle cells (vSMCs) create a highly reactive vascular bed. Tissue factor (TF), a potent procoagulant protein and the driver of postinjury thrombosis models, is two- to threefold higher in vSMCs in the uremic milieu and enhances thrombosis (8, 9). Retention of a distinct set of metabolites characterizes the state of uremia, including indolic solutes such as indoxyl sulfate (IS), which are particularly vasculotoxic (10). They enhance TF expression by activating the aryl hydrocarbon receptor (AHR) pathway, and AHR antagonists destabilize and down-regulate TF in the uremic milieu (11). Although IS and AHR are emerging regulators of TF, their contribution to the thrombotic uremic milieu and the mechanism in CKD warrant further elucidation, because better understanding will help design improved approaches to minimize CKD-specific thrombosis risk.

It is also imperative to weigh the antithrombotic benefit of the approach to bleeding risk in CKD patients, because uremia is a state of bleeding diathesis (12). This risk is further exacerbated by current antithrombotics (13), which target the hemostatic defenses in the blood. Even newer antithrombotics that are deemed safer, although not tested specifically in the uremic milieu, may function suboptimally in CKD, because none target CKD-specific risk factors (14) and may paradoxically enhance thrombosis due to altered baseline platelet reactivity in CKD (15). Targeting CKD-associated thrombotic factors is likely to lower the thrombosis risk to non-CKD range and may create a milieu more conducive to current antithrombotics/antiplatelet agents. Such a strategy is also likely to be safer, because it will leave the hemostatic defense in blood intact.

We demonstrate IS as an AHR-dependent mediator of the hyperthrombotic uremic milieu all across the CKD spectrum and show that AHR regulates TF through the STIP1 homology and U-box–containing protein 1 (STUB1), a ubiquitin ligase. We also demonstrate that the perturbation of STUB1 reverts the CKD-associated thrombosis risk to nonuremic range without altering hemostasis.


IS mediates a hyperthrombotic uremic phenotype in an AHR- and TF-dependent manner across the spectrum of CKD

To examine the mediators of the hyperthrombotic uremic milieu in vivo, we considered different animal models of CKD (16). The renal damage inflicted in all of these CKD models results in retention of a whole host of uremic solutes and precludes probing of a thrombosis axis initiated by a specific uremic solute. Furthermore, contrary to human CKD, 5/6 nephrectomy model of CKD failed to show enhanced thrombosis (17). Therefore, we set out to create a solute-specific animal model that faithfully recapitulates the hyperthrombotic uremic phenotype. Because IS increases TF (8, 11, 18), we hypothesized that IS will enhance thrombosis. Toward that end, we developed an animal protocol to increase the amount of IS similar to that seen in patients with advanced CKD [CKD stage 5/end-stage renal disease (ESRD)] by administering IS and inhibiting its excretion through the organic anion transporter (OAT) channel using probenecid (19). Of the different tested protocols, a combination of IS (4 mg/ml) in water given ad libitum and probenecid (150 mg/kg, intraperitoneally) twice a day increased IS concentrations beyond those of ESRD patients and was selected for further experiments (Fig. 1A and fig. S1A). Probenecid used in this model inhibited renal excretion of IS. Given the ubiquitous expression of OAT channels (19), probenecid may also compromise the entry of IS into the cells. However, we posited that high blood concentration of IS resulting from the above protocol will ensure sufficient intracellular IS concentrations to activate AHR, even in the presence of probenecid. Toward that end, we examined AHR activation in the vessels of mice that received IS + probenecid, using an AHR decay assay (fig. S1, B and C). The assay is based on the observation that AHR activation with an agonist ligand such as IS eventually results in the degradation of AHR protein (11). Thus, AHR expression is likely to be lower in the vessels of animals exposed to IS + probenecid if intracellular concentration of IS rises in the presence of probenecid. The aortae of IS + probenecid mice harvested at different time points showed a significant (P values ranging from 0.003 to <0.001) and persistent reduction in AHR throughout the duration of exposure, as compared to probenecid controls (fig. S1, B and C). In addition, a striking increase in TF was observed in their aortae (fig. S1D). Together, these data show that the increase in IS concentrations in the vessel walls is sufficient to activate AHR-TF signaling in mice exposed to IS + probenecid.

Fig. 1. IS mediates a hyperthrombotic uremic phenotype in an AHR-dependent manner across CKD stages.

(A) Mean blood concentrations of IS in 10- to 14-week-old male and female C57BL/6 mice (n = 5 per time point) are shown. The P values correspond to an increase in IS concentrations compared to the probenecid group. No differences in blood concentrations of IS were noted between male and female mice. The dashed line represents the average concentration of IS in ESRD patients (11, 47). Student’s t test was performed. Data are shown as means ± SD. (B) Representative traces of carotid artery flow in C57BL/6 animals before and 20 min after the application of 10% FeCl3 (n = 8 per group). The time for the blood flow to drop below 0.299 ml/min was considered as TtO (arrow). (C) Representative images from hematoxylin and eosin–stained carotid arteries collected after the FeCl3 procedure. Representative images from six vessels per group are shown. Scale bars, 25 μm. RT, right; LT, left. (D) Mean TtO (n = 8 per group) after 5 days of exposure to IS and IS + CH223191. Student’s t test was performed. Data are shown as means ± SD. (E) Mice were exposed to probenecid + IS for 5 days and then subjected to the FeCl3 thrombosis assay. Thirty minutes before the assay, control antibody or rat anti-mouse TF neutralizing antibody was administered. Mean TtO (n = 6 per group) is shown. Data are shown as means ± SEM. (F) Mean TtO in the photochemical thrombosis model in probenecid and probenecid + IS mice (n = 6 per group). Student’s t test was performed. Data are shown as means ± SEM. (G) Relationship between TtO and different concentrations of IS in 16 animals. IS concentrations from different stages of human CKD are shown (3, 11). A Pearson correlation analysis was performed.

After confirming the activation of the AHR-TF axis in the vessel walls of these animals, we examined thrombogenicity in the carotid artery with the FeCl3 postinjury thrombosis model (20). The primary end point was time to occlusion (TtO), which corresponds to the carotid blood flow between 0 and 0.2 ml/min (Fig. 1B and fig. S1E) (20) and was validated by the presence of an occlusive thrombus at the site of injury (Fig. 1C). From a range of FeCl3 concentrations tested, 10% FeCl3 resulted in the smallest SD and was adopted for subsequent experiments (fig. S1F).

Under these optimized conditions, IS significantly (P =0.001) reduced TtO (Fig. 1D), which was reversed with the AHR antagonist CH223191 (P = 0.005) (11) or previous infusion of a prevalidated anti-mouse anti-TF neutralizing antibody (P = 0.021; Fig. 1E) (21). To further confirm the thrombogenicity of IS, we induced thrombosis by photoactivated rose bengal dye, which produces singlet oxygen to damage endothelial cell membranes (22). Consistent with the FeCl3 model, IS-treated animals showed a significant reduction in TtO after photochemical injury (P = 0.017; Fig. 1F). Whereas these two models examined the prothrombotic properties of IS at concentrations corresponding to advanced CKD, we next probed its role in early stages of CKD, because patients with mild CKD are also predisposed to enhanced thrombosis (11, 23). IS concentrations corresponding to the different stages of CKD were achieved in the blood of animals by titrating the IS concentration in water. Resultant blood concentrations of IS were determined and correlated to the TtO. The data showed a significant inverse correlation between TtO and IS concentrations corresponding to all CKD stages (P < 0.0001; Fig. 1G). Overall, the above results support IS as an AHR- and TF-dependent mediator of the hyperthrombotic uremic milieu across the entire CKD spectrum.

AHR regulates TF through STUB1, a ubiquitin ligase for TF

Because the above data implicated IS as a strong contributor to the hyperthrombotic uremic phenotype, we further probed the mechanism of IS-induced changes in AHR/TF, which promote thrombogenicity. We found that STUB1, a ubiquitin ligase and an AHR interactor, mediated the effect of IS/AHR on TF (24). STUB1 silencing in primary human aortic vSMCs and human umbilical vein endothelial cells (thrombosis-relevant cell types) showed significantly increased TF expression (P = 0.001) (Fig. 2A and fig. S2A) and TF activity in both uremic and nonuremic milieu (Fig. 2B). In line with STUB1’s ubiquitin ligase function (24, 25), STUB1-silenced vSMCs showed a significantly (P < 0.001) prolonged TF half-life under the uremic condition, from 3.25 hours to more than 8 hours (Fig. 2, C and D), and reduced ubiquitination (fig. S2B). CB7993113, a competitive AHR antagonist (11), reduced TF expression by 80% (from 1.0 to 0.21), whereas it was only reduced by 30% (from 2.7 to 1.98) with a concomitant STUB1 silencing (Fig. 2E). STUB1 silencing also abrogated CB7993113-mediated TF ubiquitination in the uremic milieu (Fig. 2F). Consistent with the above results, mesenchymal embryonic fibroblasts (MEFs) from STUB1 knockout (KO) animals (25) showed a significant increase in TF expression (P = 0.001) and activity (P = 0.003) and reduced TF ubiquitination (fig. S2, C to E). Ubiquitin ligase–dependent STUB1 regulation of TF was further demonstrated using a ubiquitin ligase–deficient H260Q STUB1 mutant (25). Compared to STUB1 wild-type (WT), the H260Q mutant showed little effect on TF ubiquitination (fig. S2F). The above results suggest that STUB1 mediates TF ubiquitination and that AHR antagonist augments TF down-regulation through STUB1.

Fig. 2. STUB1 destabilizes and ubiquitinates TF and mediates TF regulation by AHR.

(A) Lysates from primary human aortic vSMCs transfected with control (Csi) or STUB1 silencing oligonucleotides (STUB1si) were probed for TF and normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) using ImageJ. Equal amounts of lysates were separately probed for STUB1 to avoid stripping of blots for the proteins with molecular weights in close range, a strategy also applied to other figures. Expression of TF and STUB1 normalized to the loading control is shown below this and subsequent Western blots. A representative figure from four independent experiments is shown. (B) vSMCs pretransfected with Csi and STUB1si were stimulated with 5% of the indicated type of serum for 24 hours. A mean TF activity of two independent experiments performed in duplicates is shown. Student’s t test was performed. Number sign (#) corresponds to TF activity in Csi vSMCs treated with uremic compared to control serum. Other P values correspond to STUB1-silenced compared to Csi cells. Data are shown as means ± SD. (C) vSMCs pretransfected with Csi and STUB1si grown in 5% uremic serum for 24 hours and then treated with cycloheximide (80 μM) to inhibit protein translation for the indicated time. Equal amounts of lysates were probed separately to confirm STUB1 silencing. TF expression normalized to the loading control is depicted below the blot. A representative figure from four independent experiments is shown. (D) Densitometry of normalized TF expression is represented as the percentage of TF at time 0. The time to reach 50% of initial TF was considered as the half-life of TF. An average of four experiments is shown. Data are shown as means ± SD. (E) vSMCs pretransfected with Csi and STUB1si were treated with 5% uremic serum with or without the AHR antagonist CB7993113 (20 μM) or vehicle for 24 hours. A representative of two independent experiments performed in duplicate is shown. (F) vSMCs pretransfected with Csi and STUB1si were treated with 5% uremic serum + 20 μM CB7993113 (12 hours) and 10 μM MG132 (4 hours) before harvest. The lysates immunoprecipitated with anti-TF antibody were probed with anti–ubiquitin (Ub) antibody. Five percent of lysates are shown as inputs. A representative of three independent experiments is shown. IP, immunoprecipitation; WB, Western blotting.

Uremia-dependent STUB1-TF interaction is dynamic

Ubiquitination is a nonlinear process consisting of a complex interdigitated set of regulated steps, which include ubiquitin activation, conjugation, and transfer of the ubiquitin moiety to the target molecule (26). This process requires a precise and dynamic interaction between a ligase and its putative target. Reciprocal immunoprecipitation assays in vSMCs demonstrated that anti-TF antibodies coimmunoprecipitated STUB1 and vice versa (fig. S3A). An in vitro binding assay demonstrated a direct binding of recombinant STUB1 to purified human TF protein (Fig. 3A). Immunofluorescence studies showed colocalization of STUB1 and TF predominantly in the cytosol of vSMCs in normal human artery and cultured vSMCs with Pearson’s correlation coefficients of 0.50 and 0.82, respectively (fig. S3, B to F) (27). Although these data indicated a constitutive interaction between STUB1 and TF, the STUB1-TF interaction was dynamic and dependent on the uremic status. STUB1-TF interaction reduced substantially with IS (Fig. 3B) and increased within 20 min of AHR antagonist treatment. This rapidity of increased STUB1-TF interaction induced by AHR antagonist is consistent with the previously observed restoration of TF ubiquitination and shortened TF half-life with AHR antagonist (11). Together, the above data suggest that uremia reduces STUB1-TF interaction and stabilizes TF and that AHR antagonists rapidly restore this interaction.

Fig. 3. A dynamic STUB1-TF interaction depends on the uremic status.

(A) Recombinant GST-tagged STUB1 protein was immobilized on glutathione S-transferase (GST) beads and treated with lipidated human recombinant TF (rhTF) protein for 4 hours. Eluents were probed for bound TF. Five percent of recombinant GST-tagged STUB1 stained with Coomassie and recombinant TF are shown as inputs. Representative immunoblots from four independent experiments are shown. (B) vSMCs were pretreated with 10 μM IS (24 hours) and 20 μM CH223191 (20 min) followed by immunoprecipitation using anti-STUB1 antibody, and the eluents were probed for TF. The stripped blot was reprobed for STUB1. Five percent of cell lysates are shown as inputs. Representative blots from three independent experiments are shown. DMSO, dimethyl sulfoxide. (C) Human embryonic kidney (HEK) 293T cells stably expressing FLAG-tagged WT TF (TF-WT) or TF C-terminal truncation (TFdelC) were transfected with myelocytomatosis (MYC)–tagged STUB1. Cell lysates were immunoprecipitated with FLAG or MYC antibodies, and coimmunoprecipitated proteins were detected using MYC and FLAG antibodies. Five percent of lysates were probed as inputs. A representative of two independent experiments done in duplicates is shown. (D) HEK293T cells coexpressing FLAG-tagged WT TF and empty vector (control) or MYC-tagged STUB1 plasmids were treated with cycloheximide for the indicated amount of time. The TF bands were normalized to GAPDH. Equal amounts of lysates were probed for MYC-tagged STUB1. A representative of three independent experiments is shown. (E) HEK293T cells coexpressing FLAG-tagged TFdelC and empty vector (control) or MYC-tagged STUB1 were processed as in (D). A representative from three independent experiments is shown. (F) HEK293T cells coexpressing FLAG-tagged WT TF (WT) or TFdelC (delC) along with MYC-tagged STUB1 were treated with 10 μM MG132 for 16 hours. The lysates were immunoprecipitated with anti-TF antibodies and probed for ubiquitin. The stripped blot was reprobed with FLAG. Five percent of the cell lysates are shown as inputs. Representative immunoblots from three independent experiments are shown. (G) Confocal images of paraffin-embedded sections of an explanted AVF from a 42-year-old male with stage 5 CKD stained with anti-TF and anti-STUB1 antibodies are shown. In different areas of the same AVF, cells with lower STUB1 and higher TF expression are marked by asterisks (*), and cells showing the opposite pattern are marked by crosses (+). The images shown are representative of eight immunofluorescence images acquired from four CKD/ESRD patients. Scale bar, 100 μm. (H) Pipeline of object recognition algorithm developed to correlate cell-level intensity distributions of STUB1 with TF (see Materials and Methods for more details). (I) Eight immunofluorescence images from four explanted AVFs from CKD/ESRD patients were analyzed using an object recognition algorithm. The intensities of TF and STUB1 within image objects containing vSMCs (average of 187 cells per image; total, 1501 cells) were quantified and averaged for each image. Data are shown as means ± SD. (J) Confocal images of paraffin-embedded sections of an explanted AVF from a 47-year-old CKD patient (uremic) and a popliteal artery from a 53-year-old male with normal renal function (nonuremic) (table S1) stained with anti-TF and anti-STUB1 antibodies. Two representatives from a total of eight uremic and nonuremic images are shown. L, lumen; S, subendothelium; M, medium. Asterisks (*) mark areas of cells with low STUB1 and high TF expression similar to those in (G). Scale bars, 100 μm. (K) Confocal images acquired from four explanted AVFs from CKD/ESRD patients and popliteal arteries from patients with normal renal function were analyzed using a pixel-level colocalization algorithm. Percentage colocalization of TF-STUB1 was defined as the fraction of pixels with intensity values greater than image-specific thresholds for STUB1 and TF and was compared between the two groups. Data are shown as means ± SD.

Furthermore, the site of interaction of STUB1 and TF corroborated the above binding pattern. Because STUB1 is a cytosolic protein (28) and TF has a cytosolic C terminus (fig. S3G) (29), STUB1 is likely to target the C-terminal tail of TF. Domain mapping was performed by comparing the WT TF (TF-WT) with a TF truncation lacking its C terminus (TFdelC). STUB1 interacted with TF-WT but not TFdelC (Fig. 3C) and down-regulated it (but not TFdelC) by 70% (time, 0; Fig. 3, D and E). STUB1 also significantly shortened the half-life of TF-WT (P = 0.006) (fig. S3H) and increased its ubiquitination (Fig. 3F). STUB1 did not destabilize (Fig. 3E and fig. S3I) or ubiquitinate TFdelC (Fig. 3F), indicating that STUB1 targets the C-terminal tail of TF for ubiquitination and degradation.

STUB1-TF relationship in human tissue is uremia-dependent

Dynamic interaction with and regulation of TF by STUB1 were further substantiated in human vascular tissue using immunofluorescence studies of explanted arteriovenous fistulae (AVFs), unique vascular conduits required in patients with advanced CKD, which are frequently prone to thrombosis (30). The vSMCs within the walls of AVF explants expressed both STUB1 and TF (Fig. 3G). On the basis of the above data (Fig. 2, A, C, and E), we posited an inverse relationship between STUB1 and TF in the uremic milieu. However, the precise demonstration of a relationship between proteins requires their quantification, which is challenging in heterogeneous human tissues using conventional methods (31). Therefore, we developed an object-level intensity estimation algorithm to quantify TF and STUB1 expressions. The results showed a strong inverse correlation between TF and STUB1 expression in the vSMCs in the walls of AVFs (Fig. 3, H and I, and table S1). We further posited that reduced STUB1-TF interaction in vSMCs induced by IS (Fig. 3B) is likely to reduce STUB1-TF colocalization in the uremic vessel wall. To this end, we compared the STUB1-TF colocalization in human AVF (uremic vessel) and vessels from non-CKD patients (nonuremic control) using a customized colocalization algorithm with pixel-level analysis (table S1). The results showed an almost 50% reduction in the colocalization of STUB1-TF in uremic compared to nonuremic human vessels (Fig. 3, J and K). Together, these results indicate STUB1’s direct and dynamic interaction with and regulation of TF.

STUB1 modulation regulates postinterventional thrombosis

TF from exposed vSMCs is a critical trigger of postinjury thrombosis, especially in the uremic milieu (8, 9). We examined the effect of STUB1 on thrombosis using the flow-loop system (32), a validated model of postinterventional thrombosis (Fig. 4, A and B). It specifically examines the blood and vessel wall factor interactions under humanized rheological conditions and can recapitulate thrombosis in the uremic milieu (8). The flow loops with STUB1 KO MEFs showed more clot formation (Fig. 4C) and a significantly higher clot burden [increased hemoglobin (Hb), P = 0.007; lactate dehydrogenase (LDH), P = 0.002; Fig. 4D). Next, the effect of STUB1 up-regulation on thrombosis was examined using 2-(4-hydroxy-3-methoxyphenyl)benzothiazole (YL-109) (Fig. 4E). YL-109, an analog of 2-(4-amino-3-methylphenyl)benzothiazole, is a preclinical compound with tumor-suppressive properties (33, 34). Being a ligand of AHR, YL-109 binding to AHR elicits its nuclear translocation to induce the transcription of STUB1 gene (33). In our model, YL-109 significantly (Hb, P = 0.023; LDH, P = 0.017) inhibited IS-induced thrombosis in the flow loops (Fig. 4F). Similarly, we observed a dose-dependent reduction in TF expression and activity along with doubling of STUB1 in vSMCs (Fig. 4, G and H, and fig. S4). Because YL-109 increases STUB1 expression, it is likely to enhance TF ubiquitination, which was examined using coexpression of ubiquitin and TF in the presence of uremic serum. YL-109 treatment increased the higher–molecular weight polyubiquitinated TF in vSMCs (Fig. 4I). Together, these genetic and pharmacological manipulations of STUB1 in vSMCs lining the flow loops consistently demonstrated a regulation of thrombosis by STUB1 in a cell type–specific manner.

Fig. 4. STUB1 modulation regulates TF activity and thrombosis in a postvascular interventional model.

(A) Scheme of flow-loop preparation. (B) The flow-loop system consists of silastic loops loaded on rotor stages and driven by motors and motion controllers. The tubes injected with the human blood are subjected to coronary-like flow pattern until clotting appears. The wall motion creates bidirectional flows that are measured via onboard, extracorporeal flow probes built into the rotor stages. (C) A representative tube from six independent flow loops seeded with STUB1 WT and KO MEFs is shown (n = 6 per group). (D) Average Hb and LDH from clotted flow loops lined with STUB1 WT and KO MEFs (n = 6 per group). Student’s t test was performed. Data are shown as means ± SEM. (E) Top: Structure of YL-109 (33). Bottom: Primary human aortic vSMCs seeded on fibronectin-coated tubes were stimulated with IS (10 μM) with or without YL-109 (25 μM) for 24 hours before loading. A representative tube from six independent flow loops in each group is shown. (F) Average Hb and LDH from clotted flow loops pretreated with IS with or without YL-109 (n = 6 per group). Data are shown as means ± SEM. (G) Lysates of vSMCs pretreated with 5% uremic serum along with the indicated amount of YL-109 for 24 hours were probed for TF and GAPDH. An equal amount of lysates was probed for STUB1. Representative blots from three independent experiments are shown. Student’s t test was performed, and asterisk (*) indicates significant changes in TF and STUB1 expressions compared to 5% uremic serum (P < 0.05). (H) A serum-starved confluent monolayer of vSMCs was treated with 5% uremic serum and YL-109 at different concentrations, and TF activity was measured in picomolar and normalized to 103 cells. An average of two independent experiments performed in duplicates is shown. A log dose of YL-109 was plotted against TF activity. Data are shown as means ± SD. (I) Primary human aortic vSMCs cotransfected with the FLAG-tag TF and hemagglutinin (HA)–tag ubiquitin plasmids were treated with IS (50 μM) with or without YL-109 (20 μM) for 24 hours and with MG132 (5 μM) for 16 hours before harvest. The lysates were immunoprecipitated using anti-FLAG antibody and probed with anti-HA antibody. Five percent of whole-cell lysates were separately probed for STUB1. A representative blot of three independent experiments is shown.

IS-AHR-STUB1 axis modulation reverses the hyperthrombotic uremic phenotype to non-CKD range without altering hemostasis

We next examined whether augmented STUB1 suppressed the hyperthrombotic uremic milieu in vivo using the indolic solute–specific animal model (Fig. 1). The effects were compared to heparin, which is a standard-of-care antithrombotic in CKD patients (Fig. 5A). Compared to the IS group, YL-109 significantly (P = 0.005) prolonged the TtO (Fig. 5, B and C) with no differences in IS concentration between these groups (fig. S5A). Because STUB1-TF is a CKD-specific thrombotic pathway, we posited that its targeting by YL-109 is likely to revert the thrombogenicity to the non-CKD range. Therefore, we compared the effect of YL-109 to the probenecid (non-CKD) controls. The results showed no difference in TtO with IS + YL-109 compared to probenecid controls. On the other hand, heparin at a dose considered therapeutic in humans (35) significantly (P = 0.001) prolonged TtO compared to both IS and probenecid controls (Fig. 5C). YL-109–treated animals showed a significant increase in STUB1 (P = 0.004) and decrease in TF (P = 0.001) in their aortae (Fig. 5D and fig. S5B). To further probe a quantitative link between changes in TF expression and thrombogenicity in vivo, we correlated the reduction of TF in the aortic lysates of individual mice exposed to IS + YL-109 to the prolongation of TtO. We hypothesized that both these parameters will inversely correlate should the reduction in TF expression in the vessel wall reduce the thrombogenicity in the uremic milieu. A significant negative correlation (Spearman rho = −0.833 and P = 0.02) between the reduction in TF within the vessel wall and prolongation of TtO was observed in mice treated with YL-109 (Fig. 5E). Together, these data strongly suggest that reduction in TF by STUB1 regressed the hyperthrombotic uremic phenotype to the non-CKD range.

Fig. 5. Perturbations of AHR-STUB1-TF axis normalize the hyperthrombotic uremic phenotype without altering the bleeding risk.

(A) The experimental design assessing the effect of YL-109 on the IS-specific uremic thrombosis model. A dose of heparin that increases activated partial thromboplastin time (aPTT) to the therapeutic range in mice (35) was administered 2 hours before the procedure. IP, intraperitoneal; PO, per os. (B) Representative traces of carotid artery blood flow showing TtO (arrow) in different groups. (C) An average TtO from each group of animals from three independent experiments is shown. The numbers of animals used include probenecid (n = 9), probenecid + IS (n = 12), probenecid + IS + YL-109 (n = 8), and probenecid + IS + heparin (n = 6). No difference in TtO was observed between probenecid and probenecid + IS + YL-109 (P = 0.65). Heparin significantly prolonged TtO compared to both IS and probenecid control groups (P = 0.021). Data are shown as means ± SD. (D) Mean normalized TF and STUB1 expression in the aortae of five mice injected with IS with or without YL-109 is shown. Student’s t test was performed. Data are shown as means ± SD. (E) Changes in TF expression in the individual aortae of IS + YL-109–treated animals and their respective TtO are shown. A Spearman correlation analysis was performed. (F) IS concentrations in the blood of mice exposed to 0.25% adenine diet for 2 weeks and animals on regular chow (control) (n = 5 per group). Data are shown as means ± SD. (G) An average TtO in control (regular diet) and adenine-induced CKD mice with AHR inhibitor (CH223191) or STUB1 enhancer (YL-109) (n = 5 per group) is shown. #P value compares the adenine and control groups. *P value compares CH223191 and YL-109 groups with the adenine group. Data are shown as means ± SEM. (H) Average tail vein bleeding times for each group (n = 5 per group). Data are shown as means ± SD. (I) Differences in the average tail vein bleeding times (y axis) between probenecid control (non-CKD) and other groups (CKD). The solid line represents the average bleeding time of the control group, and the dotted lines represent two SDs of bleeding time from the control group. The shaded areas are the regions beyond 2 SD.

Because the solute-specific animal model partially recapitulates the uremic phenotype, these results warranted further validation in an established model of CKD. Among the different animal models of CKD, the proteinuric CKD model and hypertensive model are likely to confound the thrombosis assay due to loss of anti- and prothrombotic factors in the urine (36) and increased endothelin, respectively, in these models (37). Therefore, we used the adenine-induced renal injury model, a well-established model of uremia induced by extensive tubulointerstitial fibrosis (fig. S5C) (38). Within 2 weeks of 0.25% adenine diet, the animals showed significant (P < 0.001) increases in IS concentrations and blood urea nitrogen, corresponding to advanced CKD patients (Fig. 5F and fig. S5D). A significant (P = 0.01) reduction in TtO was observed in the adenine-induced CKD animals compared to animals on a regular chow diet (Fig. 5G), further supporting CKD as a hyperthrombotic environment. In the adenine-treated group, the TtO was significantly prolonged with CH223191 (P = 0.004) or YL-109 (P = 0.050), compared to the group treated with adenine alone (Fig. 5G), despite no differences in renal function between these three groups (fig. S5D). Furthermore, similar to the indolic solute–specific model (Fig. 5C), no differences in TtO were noted in the CH223191-treated (P = 0.374) and YL-109–treated (P = 0.983) groups when compared to the animals on a regular chow diet (non-CKD animals). All these data obtained from two independent models strongly argue for CKD as a hyperthrombotic milieu and indicate that uremia-induced thrombogenicity is normalized to a non-CKD range with the modulation of AHR-STUB1 axis.

The antithrombotic effects of the above agents were weighed against the bleeding risk using the standard tail vein transection model (20). Because STUB1 modulation normalized the thrombotic risk to non-CKD range, we compared YL-109–mediated alteration in bleeding time to probenecid (non-CKD) controls. The data showed no reduction in bleeding time with IS compared to probenecid controls (Fig. 5H). Compared to the probenecid controls, bleeding time was not different from YL-109 (Fig. 5H) and fell within the two SD ranges of the non-CKD controls’ bleeding time (Fig. 5I, clear area). In contrast, heparin significantly (P = 0.003) prolonged the bleeding time beyond two SDs of probenecid controls (Fig. 5I, shaded area). Overall, these data indicate that, unlike heparin, STUB1 modulation reverted the IS-mediated hyperthrombotic phenotype to the nonuremic range without enhancing the bleeding risk.


CKD patients are prone to thrombosis and are specifically sensitive to the relatively CKD-nonspecific antithrombotic and antiplatelet agents (12, 13). CKD patients are at fourfold increased risk of bleeding, which is further augmented by 50% with every 30 ml/min decrease in creatinine clearance (39, 40). To some, this implies that a CKD patient represents an extreme case of classic vascular disease, and to others, this implies that the CKD milieu is distinct and unique. Our data support the latter and implicate IS in particular as a mediator of the hyperthrombotic uremic state through modulation of STUB1-TF interaction. Here, we show that reduced STUB1-TF interaction in vSMCs in uremia stabilizes TF and augments thrombosis upon exposure to vascular injury. Conversely, restoration of STUB1-TF interaction by suppression of uremic effect by AHR antagonist or STUB1 inducer augments TF ubiquitination and degradation and inhibits thrombosis. These data help discern the antithrombotic mechanism of AHR antagonism and add to the sophistication of the complex vascular biology of CKD while also highlighting the role of dynamic protein-protein interaction influenced by the uremic state, which we further substantiate in human vessels. We also demonstrate the prominent role of STUB1-TF axis in mediating CKD-specific thrombotic risk, which reverses to non-CKD range upon pharmacological targeting.

Although there can be different explanations to the dynamic interaction of STUB1 and TF, rapid modulation of this interaction with AHR status argues for the role of AHR and posttranslational modifications (PTMs) in this process. Because STUB1, AHR (24), and TF bind to each other (11), it is likely that they form a multiprotein complex, which may undergo stoichiometric or conformational changes upon AHR activation (IS treatment) or AHR suppression (AHR antagonist treatment), altering the interactions between the components (24). The kinetics of STUB1-TF interaction support PTMs in either or both partners. Similar PTMs can modulate the interaction of STUB1 with other targets (41). This rapid STUB1-TF interaction is followed by ubiquitination and proteasomal degradation of TF, such that TF protein reaches its half-life within 40 min in vSMCs (11). The involvement of ubiquitination in regulating TF imparts specificity and efficiency to TF biology. This system is built for amplification, such that the target (TF), the ubiquitin ligase (STUB1), and the modifying protein (ubiquitin) are all governed by complex autoregulatory and pararegulatory loops (26). It is such a system that can regulate the biology of a central stimulus like TF, which is released and/or exposed on vSMCs early after vascular injury and especially within the context of the complex uremic milieu (8, 11, 18). Such biochemistry also explains why modest changes in any of the involved elements induce large, nonlinear effects on outcomes and a strong in vivo correlation between the reduction in thrombogenicity and changes in TF within the vessel wall in response to pharmacological manipulation of STUB1.

Even minor changes in TF expression are expected to have a substantial effect on thrombosis, because vessel wall TF activation is only a primary trigger, followed by a multitude of nonlinear and interrelated events, such as the extrinsic coagulation cascade and platelet aggregation (42), all of which amplify the primary trigger of TF activation to produce robust thrombus formation. Thus, perturbation of TF through the AHR-STUB1 axis represents an effective therapeutic strategy in CKD patients, because none of the current antithrombotics or the investigational agents such as factor XII and XI inhibitors target CKD-specific thrombosis (14, 15). Their suboptimum efficacy is likely to be exacerbated by poor thrombus control from a persistent nidus of exposed vSMCs secondary to compromised reendothelialization of the vascular wound in the uremic milieu (43). Compounding the poor efficacy profile is their enhanced bleeding risk due to altered pharmacokinetics in CKD and perturbation of critical hemostatic defenses (12, 13).

Two complementary in vivo thrombosis models used here, involving damaged endothelium, established the relevance of our in vitro studies. The added value of a solute-specific animal model in the context of vascular injury was in allowing us to directly address CKD-associated risk assessment. In addition to avoiding confounders such as hypertension in other CKD models (16), this animal model allows titration of a specific uremic solute to examine its effect at different stages of CKD. Although this animal model should be a valuable tool for mechanistic probing and is potentially useful in preclinical drug development, especially having been validated with an established adenine-induced CKD model, it does not emulate the human arterial rheological patterns or the flow characteristics that are critical in postinterventional thrombosis (44). The flow-loop system used herein (32) is a model of postinterventional thrombosis combining a humanized system of human cells and human blood simulating human coronary flow patterns. Clot formation in this system is defined by three discrete components, which provide a different view of the net extent of thrombosis. Hb and LDH measurements represent red cell content and active remodeling of the clot, respectively, and visual inspection assesses overall clot burden. The results from the flow loops tracked well with the genetic loss of function of STUB1 and with pharmacological up-regulation of STUB1. Interrogating the AHR-TF-STUB1-TF axis using orthogonal models has strengthened the translational relevance of this study.

We identified two discrete mechanisms of TF down-regulation in the uremic milieu, namely, increasing STUB1 concentration and enhanced STUB1-TF interaction, supporting the development of direct STUB1 activators and selective AHR modulators (sAHRMs) as antithrombotics (45). Because the effects of these agents are likely to be influenced by the state of AHR signaling, a biomarker based on the concentrations of uremic solutes (11) should further refine the risk-benefit profile for antithrombotics in CKD patients. Moreover, combining therapy with agents that target disease-specific mechanisms to normalize the uremic hyperthrombotic milieu together with the current antiplatelet/antithrombotic agents might enhance therapeutic efficacy and safety.

We recognize the study’s limitations, knowing that thrombosis is a dynamic process orchestrated by several different cell types and mediators, and almost all of these components are altered in CKD (3, 4). Although vessel wall factors and platelets in IS-mediated thrombosis are being examined (11, 18, 46), the roles of other metabolites (termed “thrombolome”) (3, 11), microvesicles, etc. remain to be probed to fully characterize the hyperthrombotic uremic milieu. Although the current report demonstrates the role of IS-AHR-STUB1 axis in suppressing TF, a potential ability of this axis to regulate other members of the coagulation cascade or platelets cannot be ruled out. Finally, YL-109 was developed as an sAHRM (33, 45), but it may have additional effects on thrombosis through pathways other than AHR-STUB1-TF, and this warrants further investigation.

Several dynamic events influence thrombosis and hemostasis processes, and their precise quantification is required to determine the global thrombosis risk and allow individualization of therapy. CKD-specific thrombosis risk assessment can be performed by integrating these uremic solutes with conventional thrombotic markers (3, 11, 12). A combination of such a risk panel with a means of targeting CKD-specific thrombotic mediators is likely to improve the antithrombotic management in these patients. Our study focused on postinterventional thrombosis in CKD, because these patients carry high cardiovascular disease burden requiring complex vascular procedures and are also at risk of thrombosis of dialysis access, a lifeline of CKD patients. The signaling axis we identified needs further probing in other models of thrombosis.


Study design

The number of animals was guided by the expected effect size and SD of the readout (TtO) to obtain 90% power at an α of 0.05. Male and female mice were randomized to different groups in these experiments. Although the administration of compounds in animal experiments was not performed in a blinded manner, an investigator blinded to the experimental groups performed the thrombosis assays.

Statistical analysis

Summary statistics are presented as the mean and SD or SEM. Analysis of variance (ANOVA), paired or unpaired Student’s t test, or Wilcoxon rank sum test was used to compare the groups as appropriate, and the test for each experiment is indicated in the figure legends. A Spearman or Pearson correlation was performed to analyze the correlation between two variables, as appropriate. P < 0.05 was considered statistically significant.


Materials and Methods

Fig. S1. An indolic solute–specific animal model.

Fig. S2. STUB1 regulation of TF protein.

Fig. S3. STUB1 interaction with TF.

Fig. S4. STUB1 expression in the YL-109–treated flow loops.

Fig. S5. AHR or STUB1 manipulation in uremic animal model.

Table S1. Demographic and clinical characteristics of patients included in the study to examine the vascular expression of STUB1 and TF.


Acknowledgments: We acknowledge N. Mackman (University of North Carolina, Chapel Hill) for his guidance in TF activity assay and D. Salant and N. Rahimi (Boston University School of Medicine) for sharing resources and cell lines. We thank C. Patterson (University of North Carolina, Chapel Hill) for STUB1 KO and knockin (KI) MEFs and WT and H260Q STUB1 constructs. We also thank F. García Polite, D. Gómez Jiménez, and F. Lozano (Universitat Ramon Llull, Barcelona, Spain and Massachusetts Institute of Technology) for their technical assistance in flow-loop experiments and D. Mun for the help in obtaining patient data. We also thank M. Kirber at the Boston University Medical Center (BUMC) Core facilities for his assistance in confocal microscopy. Portions of this work were presented at the 2015 and 2016 American Society of Nephrology meetings. Funding: This work was funded in part by NIH R01 HL132325 and R01 CA175382; Evans Faculty Merit award (V.C.C.); NIH R01 HL080442 (K. Ravid); National Institute of General Medical Sciences (NIGMS) R01 GM49039 (E.R.E.); P42 ES007381 and the Art BeCAUSE Breast Cancer Foundation (D.H.S.); Hariri Research Award (no. 2016-10-009) from the Hariri Institute of Computing, Boston University and American Heart Association’s Scientist Development grant no. 17SDG33670323 (V.B.K.); Sharon Anderson Research Fellowship grant award from the American Society of Nephrology (M.S.); AHA Fellow to Faculty Transition grant 12FTF12080241 (K.K.); the grant SAF2013-43302-R from Spain Ministerio de Economía y Competitividad and Fundació Empreses IQS (M.B.); T32 training in renal biology T32 DK007053-44 (K. Rijal); and T32 training grant in cardiovascular biology T32 HL007224-40 (J.W.). A part of this work was funded by the Thrombosis and Hemostasis Affinity Research Collaborative program from the Department of Medicine, BUMC. Author contributions: V.C.C. and M.S. designed the research and experiments, with contributions from K. Ravid; M.S., J.W., S.R., F.A., K. Rijal, and S.R. performed the cell-based experiments; J.M.H. and M.E.B. performed immunofluorescence studies; M.S., M.E.B., S.M., J.W., S.R., and F.A. performed the animal experiments; J.F., J.M.H., and A.G. performed the analysis on AVF and immunofluorescence data; K.K. and M.B. designed and performed the flow-loop experiments; V.B.K. performed quantitative histology and image processing; M.O. and K.N. synthesized YL-109; D.K. provided anti-TF neutralizing antibody; R.R. assisted in the photochemical thrombosis injury model; V.C.C. and M.S. reviewed and analyzed the final data; M.S., M.E.B., and V.C.C. prepared the figures and manuscript; D.K., D.H.S., J.F., S.M., J.M.H., K.K., V.B.K., R.R., E.R.E., and K. Ravid contributed conceptually to different degrees and edited the manuscript. Competing interests: The authors declare that they have no competing interests.

Stay Connected to Science Translational Medicine

Navigate This Article