Thrombomodulin (p.Cys537Stop) is released from cells by an unusual membrane insertion/leakage mechanism

Thrombomodulin (TM) is a type-1 transmembrane glycoprotein expressed on the surface of vascular endothelial cells, consisting of an amino-terminal lectin-like domain, followed by 6 epidermal growth factor-like domains, a serine threonine-rich region containing chondroitin sulfate (CS) glycosaminoglycan addition sites, a transmembrane domain (TMD) and a short cytoplasmic tail.¹ 

TM mediates endothelial anticoagulant function: it binds thrombin (Factor IIa [FIIa]) to form TM-FIIa, which prevents FIIa from converting fibrinogen to fibrin and enables it to activate protein C (PC). Active PC degrades coagulation cofactors Va and VIIIa, thereby reducing FIIa generation. Binding of PC to the endothelial cell PC receptor (EPCR) amplifies and concentrates PC activation at the surface of the vessel wall.² Interestingly, TM and EPCR have been localized in lipid rafts of endothelial cells plasma membrane.^3-5 TM also mediates anti-inflammatory⁶ and antifibrinolytic processes.^1,7 

Low concentrations (<10 ng/mL) of soluble forms of TM (sTM) are present in the circulation of healthy individuals.⁸ Increased circulating levels are considered an indication of endothelial damage.⁹ The mechanisms underlying TM release have been reviewed.¹⁰ Briefly, microvesicles derived from endothelial cells have been shown to carry TM.^11,12 Consequently, increased release of microvesicles raises plasma levels of TM. It has been suggested that TM release from endothelial cells involves proteolysis by matrix metalloproteinases (MMPs), a disintegrin and metalloproteinases (ADAMs),^13,14 and neutrophil serine proteases.^15,16 TM cleavage by the intramembrane rhomboid protease RHBDL2 has been described,¹⁵ but its occurrence in endothelial cells remains uncertain.

Mutations in the THBD gene (encoding TM), which is responsible for high plasma levels of sTM, have been identified in patients with bleeding phenotypes. The frameshift variant c.1487delC (p.Pro496Argfs∗10) results in the synthesis of a secreted TM (TM_FS) lacking cytoplasmic tail and TMD.¹⁷ The variant c.1611C>A (p.Cys537Stop) produces a truncated TM (TM₅₃₆, ending at Leu536) lacking cytoplasmic tail and with the TMD shortened by 3 amino acids.^14,18-23 High levels of sTM in the plasma of these patients have been shown to be responsible for increased PC activation, leading to reduced FIIa generation,^17,20 and have therefore been suggested as the cause of the bleeding phenotype.

Despite its short intramembrane domain (IMD), TM₅₃₆ is still expressed on the cell surface,¹⁴ and the mechanisms contributing to its high plasma levels are poorly understood. This work elucidates the mechanisms responsible for the excessive release of TM₅₃₆ and characterizes the regulation of its efficacy as a cofactor in PC activation.

Human umbilical venous endothelial cells (HUVECs) were isolated and cultured as described.²⁴ HEK293 cells (GripTite 293 MSR; Thermo Fisher Scientific) were cultured as described by the manufacturer. Plasmid transient transfection was performed using PolyJet reagent (SignaGen Laboratories).

Giant plasma membrane vesicles (GPMVs) were prepared as described.²⁵ Transfected HEK293 cells were incubated for 2 hours in vesiculation buffer (20 mM HEPES [N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid; pH 7.4], 150 mM NaCl, 2 mM CaCl₂, 1.9 mM dithiothreitol, and 27.6 mM formaldehyde). The supernatant was centrifuged (500 g; 5 minutes) to remove detached cells, and GPMV was concentrated by centrifugation (25 000g; 35 minutes) and resuspended in a culture medium.

Transfected HEK293 cells were harvested by scraping, pelleted by centrifugation (300g; 3 minutes), resuspended in 10 mM HEPES (pH 7.8), 250 mM sucrose, 1 mM EGTA, and 2.5 mM KCl sonicated (3 bursts of 5 seconds), and centrifuged (12 000g; 15 minutes). The supernatant containing microsomes was supplemented with CaCl₂ and MgCl₂ (1 mM), incubated with Concanavalin-A Sepharose (45 minutes; 4°C), and centrifuged (500g; 3 minutes) to remove glycoprotein-rich membranes (plasma membrane). Supernatant containing plasma membrane–depleted microsomes was centrifuged (100 000g; 2 hours). Pelleted microsomes were disrupted by 30-minute incubation in Na₂CO₃ (100 mM, pH 11.5) or saponin (0.32%) solution, and then centrifuged (100 000g; 2 hours) to separate intravesicular proteins from vesicle membranes.

HEK293 cells expressing His-TM were incubated overnight in a serum-free medium. The culture medium was supplemented with protease inhibitors, centrifuged (100 000g; 1 hour), concentrated by filtration (30 kilodalton [kDa] molecular weight cutoff), and then diluted in binding buffer (Na₃PO₄, 50 mM [pH 8], NaCl 300 mM, Tween 20 0.01%). Soluble His-TM was isolated by immobilized metal affinity chromatography using a Dynabeads His-tag isolation kit, as described by the manufacturer (Invitrogen).

Cells were incubated with the membrane-impermeable biotinylation reagent Sulfo-NHS-SS-Biotin (0.25 mg/mL; 30 minutes; 4°C), washed with Tris 50 mM (pH 8) to stop the reaction, and then lysed. Biotinylated proteins were isolated using NeutrAvidin Agarose resin and analyzed by immunoblot.

Cells in a 48-well plate were incubated with PC (200 nM; 30 minutes; 25°C) then washed to remove unbound PC and incubated with 50 μL of FIIa solution (0.02 U/mL in Hanks balanced salt solution; 45 minutes; 37°C). The incubation medium was transferred to a 96-well plate, FIIa was inhibited with antithrombin (0.5 μM) and heparin (10 U/mL) addition, then the colorimetric substrate of activated PC (S-2366) was added. Absorbance was read at 405 nm every 5 minutes for 50 minutes. In some experiments, PC activation was performed in vitro using soluble TM.

Cells were lysed in the presence of protease inhibitors. Identical amounts of proteins were heat-denatured, reduced, and then separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) using NuPAGE gels (Thermo Fisher Scientific) and MES (2-(N-morpholino) ethanesulfonic acid) migration buffer, and then transferred to a polyvinylidene fluoride membrane for immunoblot detection.

The cell-surface expression of TM and EPCR was analyzed using flow cytometry (BD Accuri C6; BD Biosciences). The cells were gated on the forward and side scatter to exclude dead cells, debris, and aggregates.

All data were analyzed using GraphPad Prism software, and individual statistical 2-sided tests used are identified in the figure legends. P values ≤ .05 were considered statistically significant.

A 22-year-old French woman was referred to the “Center for Hemorrhagic and Thrombotic Diseases” at Timone Hospital (Marseilles, France) because of a personal and family history of unexplained bleeding after trauma or surgery (Figure 1A; supplemental Table 1). The patient showed no abnormalities in coagulation, fibrinolysis (supplemental Table 2), platelet function, or the expression of major platelet membrane glycoproteins (supplemental Tables 3 and 4). Plasma from the patient, her father, grandfather, and first son showed higher concentrations of sTM (Figure 1A) and lower FIIa generation than healthy subjects (Figure 1B; supplemental Table 5). High concentrations of FIIa were found in the serum derived from their coagulated plasma (Figure 1A). The sTM concentration in patients' plasma was not reduced by microvesicle removal (supplemental Table 6), indicating that sTM is not carried by microvesicles. Sequencing of the THBD gene revealed the heterozygous c.1611C>A mutation, which led to the synthesis of TM₅₃₆ (Figure 1A; supplemental Table 7).

We investigated the behavior of wild-type TM (TM_WT) and TM₅₃₆ overexpressed in HEK293 cells. TM₅₃₆ accumulated more in cell culture media (CM) than in TM_WT, although their cellular amounts were similar (Figure 2A). Flow cytometry analysis revealed that TM₅₃₆ exhibited lower cell-surface expression than TM_WT (Figure 2B).

As the length of a protein TMD determines its distribution in lipid rafts,²⁶ we analyzed the repartition of TM_WT and TM₅₃₆ in the detergent-resistant membrane fraction (DRM). TM_WT was detected both inside and outside the DRM, whereas TM₅₃₆ was mainly detected outside the DRM (Figure 2C). The presence of the entire IMD in TM₅₃₉ (supplemental Table 7) prevented its release (supplemental Figure 1A) but did not restore its normal distribution in the DRM (supplemental Figure 1B), indicating that the high release of TM₅₃₆ is not due to its altered membrane distribution and that TM distribution in lipid rafts involves factors other than the TMD length alone.

TM_WT was detected by immunoblot as a single band above 70 kDa, whereas TM₅₃₆ appeared as 2 bands above and below 70 kDa (Figure 2D). TM detection in cell-surface–biotinylated proteins (Figure 2D) and in proteins from the plasma membrane model GPMV (supplemental Figure 2A) showed that bands >70 kDa represent cell-surface–exposed TM_WT and TM₅₃₆.

As expected, cell surface TM_WT migrated slightly slower in SDS-PAGE than the 38 amino acid shorter TM₅₃₆ (Figure 2D; supplemental Figure 2A).

Treatment of cell lysates with Endoglycosidase H (EndoH), which removes N-linked glycans from glycoproteins that have not trafficked to the cis-medial Golgi, altered the migration of TM₅₃₆ detected <70 kDa, indicating that it localizes in the ESP. The high proportion of EndoH-sensitive TM₅₃₆ compared with TM_WT (Figure 2E) indicates its preferential accumulation in the ESP.

The form of TM detected at ∼150 kDa (Figure 2D) likely corresponds to SDS-resistant dimers (supplemental Figure 2).

Endogenous TM expression and release were characterized in HUVEC from control patients (HUVEC) and the proband (HUVEC₅₃₆), who had given birth to a son carrying the TM₅₃₆ mutation. TM messenger RNA levels were similar in both HUVEC types (supplemental Figure 3A), but HUVEC₅₃₆ released more sTM (Figure 3A). sTM contained in the CM of HUVEC₅₃₆ and sTM₅₃₆ produced by overexpression in HEK293 cells showed similar mobility on SDS-PAGE (Figure 3B), suggesting that both cell types release the same form of sTM. HUVEC₅₃₆ expressed less TM at their surface than HUVEC (Figure 3C). TM was detected as a single band >70 kDa in HUVEC lysates and as 2 bands in HUVEC₅₃₆ (Figure 3D), with a migration profile similar to that of TM and TM₅₃₆ coexpressed in HEK293 cells (supplemental Figure 3B). Similarly, TM was detected as 2 bands in the patient’s endothelial colony forming cells lysate (supplemental Figure 3C). TM detected <70 kDa in HUVEC₅₃₆ lysates was sensitive to EndoH deglycosylation, indicating its ESP localization (Figure 3E). The coexpression of TM_WT and TM₅₃₆ in HUVEC₅₃₆ and the impossibility of specifically detecting TM₅₃₆ prevented the characterization of TM₅₃₆ distribution in the DRM (supplemental Figure 3D).

In summary, TM₅₃₆ expression results in high sTM release and low TM cell-surface expression. TM₅₃₆ is not incorporated into lipid rafts and its intracellular trafficking is impaired.

Treatment of HEK293 cells with MMPs and ADAMs inhibitors strongly decreased the release of overexpressed TM_WT (by up to 80%) and only slightly decreased that of TM₅₃₆ (by up to 20%) (Figure 4A; supplemental Figure 4) without altering their cellular content. This result indicates that MMPs or ADAMs are not involved in the high release of sTM₅₃₆. Overexpression of RHBDL2 increased the release of sTM_WT and sTM₅₃₆ (Figure 4B). However, RHBDL2-cleaved sTM₅₃₆ migrated faster in SDS-PAGE than sTM₅₃₆ released without RHBDL2 action (Figure 4C), ruling out the involvement of RHBDL2 in the excessive basal release of TM₅₃₆. To facilitate visualization of the difference in mobility between RHBDL2-cleaved and uncleaved sTM₅₃₆, TM₅₃₆ lacking CS addition sites was used (supplemental Table 7). Interestingly, sTM₅₃₆ in the CM and TM₅₃₆ present at the cell surface (upper band detected in the cell lysate) showed similar mobility in SDS-PAGE (Figure 4D), suggesting the release of full-length TM₅₃₆ into the CM. Supporting this notion, mass spectrometry analysis of sTM₅₃₆ contained in CM identified a C-terminal tryptic peptide ending at Leu536 (supplemental Figure 5).

Soluble TM₅₃₆ in CM was deglycosylated by PNGaseF but not by EndoH (Figure 5A), indicating that it follows the canonical endoplasmic reticulum (ER)/Golgi-dependent secretion pathway.

If a fraction of TM₅₃₆ behaved like a “conventional” secreted protein, it would not be inserted into the membrane and would be found as a full-length protein in the CM. This full-length sTM₅₃₆ would escape proteolysis by RHBDL2, which only cleaves membrane-anchored substrates. Consequently, full-length sTM₅₃₆ and RHBDL2-cleaved sTM₅₃₆ coexisted in the CM. Immunoblot analyses of sTM₅₃₆ contained in the CM revealed a different situation: upon RHBDL2 overexpression, the band corresponding to full-length sTM₅₃₆ disappeared, whereas that corresponding to RHBDL2-cleaved sTM₅₃₆ appeared (Figure 4C), indicating that all synthesized TM₅₃₆ was inserted into the ER membrane. Therefore, the basal release of full-length TM₅₃₆ is likely due to spontaneous leakage from the membranes.

We tested the hypothesis that the IMD of TM₅₃₆ is not sufficiently hydrophobic to confer stable membrane insertion. Substitution of Cys527 (in the middle of the IMD) with Ser or Ile to reduce or increase IMD hydropathy (supplemental Table 7) resulted in the increased and reduced release of TM₅₃₆, respectively (Figure 5B). These data support the idea that the high release of TM₅₃₆ is due to the low hydropathy of its IMD. TM_536C527S, with the least hydrophobic IMD, was no longer detected in the CM of RHBDL2-overexpressing cells (Figure 5C), indicating that it follows the same secretion mechanism as TM₅₃₆.

To identify the membrane compartment from which TM₅₃₆ is released, we first measured TM liberation from GPMV derived from cells overexpressing TM_WT or TM₅₃₆. Both types of GPMV exhibited similar TM release capacities (Figure 5D), suggesting that TM₅₃₆ does not escape from the plasma membrane. We then investigated the presence of sTM₅₃₆ inside the vesicles of the secretory pathway (supplemental Figure 6). The proportion of intraluminal TM₅₃₆ was higher than that of TM_WT (Figure 5E), implying that sTM₅₃₆ exists in the lumen of the ER and/or Golgi.

sTM₅₃₆ contained in the CM was found to be associated with the ER protein chaperone heat shock protein family A (Hsp70) member 5 (HSPA5) (Figure 6A), and this interaction was barely detected with RHBDL2-cleaved sTM₅₃₆, which lacks IMD. An enzyme-linked immunosorbent assay specifically detecting the TM-HSPA5 complex (supplemental Figure 7A) revealed its increased presence in plasma from TM₅₃₆ patients compared with healthy patients (Figure 6B; supplemental Figure 7B).

The interaction between TM₅₃₆ and HSPA5 supports the notion that a fraction of TM₅₃₆ with its HSPA5-coated IMD is present in the ER lumen, suggesting that TM₅₃₆ in the ER lumen triggers the unfolded protein response (UPR). Indeed, overexpression of TM₅₃₆ significantly increased messenger RNA levels of the ER stress markers HSPA5 and the spliced form of XBP1 compared with overexpression of TM_WT (supplemental Figure 8). Proteasome inhibition by MG132 revealed a low-molecular-weight form of TM₅₃₆ (Figure 6C) with SDS-PAGE mobility similar to that of EndoH-deglycosylated TM₅₃₆ (Figure 6D), suggesting that it accumulates in the ER. Such a form of TM was barely detected with TM_WT. Similarly, proteasome inhibition resulted in TM accumulation in the ER of HUVEC₅₃₆ but not in HUVEC (Figure 6E). These results indicate that TM₅₃₆ is more likely than TM_WT to induce UPR and undergo proteasome-dependent ERAD.

Although TM₅₃₆ is sensitive to ERAD, a significant amount escapes it and accumulates in the CM. We hypothesized that ERAD stimulation would reduce the release of TM₅₃₆. We have previously shown that claramine (CL) enhances the ERAD-dependent elimination of misfolded proteins.²⁷ 

Treatment of HEK293 cells with CL reduced the amount of TM₅₃₆ accumulated in the ER upon proteasome inhibition (Figure 6F), indicating that CL increased TM₅₃₆ degradation via a pathway that bypasses the proteasome. Consistently, the release of TM₅₃₆ was reduced by CL (50%), whereas that of TM_WT was minimally affected (Figure 6G; supplemental Figure 9A). Other molecules intended to modulate ER protein quality control systems did not reduce TM₅₃₆ release (supplemental Figure 9B). Readthrough-promoting molecules did not rescue full-length TM expression in TM₅₃₆-expressing cells or decrease TM₅₃₆ release (supplemental Figure 9C-D).

These results indicate that the ER protein quality control system can be targeted to reduce TM₅₃₆ release.

We investigated whether the reduced amount of TM on the surface of TM₅₃₆-expressing cells and the poor integration of TM₅₃₆ in lipid rafts influenced FIIa-dependent PC activation.

HUVEC and HUVEC₅₃₆ expressed similar amounts of EPCR on their surfaces (supplemental Figure 10A). However, FIIa-dependent activation of EPCR-bound PC was lower in HUVEC₅₃₆ than in HUVEC (Figure 7A; supplemental Figure 10B), which is consistent with TM expression on the surface of these cells (Figure 3C). The role of TM localization in lipid rafts on PC activation was analyzed using HEK293 cells overexpressing EPCR and TM_WT or TM₅₃₉ (poorly integrated in lipid rafts but not released from cells; supplemental Figure 1). Disruption of lipid rafts by methyl-β-cyclodextrin increased PC activation on TM_WT cells, but not on TM₅₃₉ cells (Figure 7B; supplemental Figure 10C), suggesting that TM localization in lipid rafts plays a role in regulating its cofactor activity in PC activation.

Consequently, TM cofactor activity during PC activation is determined by its expression on the cell surface and its localization in lipid rafts.

Overexpression of secreted TM₅₁₄ (supplemental Table 7) in HEK293 cells produced a soluble form that migrated on SDS-PAGE as 2 bands: 1 smeared ∼100 kDa and the other thin, slightly >70 kDa (Figure 7C). The smeared band corresponded to CS-modified TM₅₁₄ because it was not present upon substitution of the CS attachment sites (supplemental Table 7; Figure 7C). The proportion of CS-modified full-length sTM₅₃₆ was lower than that of TM₅₁₄ and the naturally secreted variant TM_FS (Figure 7C), suggesting that the addition of CS to the soluble forms of TM depends on their secretion mechanisms.

The presence of CS in TM is known to enhance its binding to FIIa,²⁸ we therefore examined the impact of this modification on the efficacy of soluble TM to mediate FIIa-dependent PC activation in vitro (supplemental Figure 11A). As expected, TM₅₁₄ and sTM₅₃₆ provided greater PC activation than same amounts of their CS-devoid counterparts (Figure 7D; supplemental Figure 11B-C). TM_FS, which is more CS-modified than TM₅₃₆, was more effective than the latter as a cofactor in PC activation (Figure 7E; supplemental Figure 11D) and in supporting thrombin activatable fibrinolysis inhibitor cleavage by FIIa (supplemental Figure 12).

PC activation mediated by CS-devoid TM₅₃₆ was barely different from that mediated by TM₅₁₄ lacking CS and IMD (Figure 7F; supplemental Figure 11E), highlighting the modest role of the IMD in this activation.

Therefore, the efficacy of sTM as a cofactor in PC activation is determined by the proportion of its CS-modified form.

Patients with the TM₅₃₆ mutation have high plasma concentrations of TM, making them prone to bleeding. In this work, we have elucidated the mechanism responsible for the increased release of TM₅₃₆ from cells and proposed a strategy to reduce it. We have shown that TM₅₃₆-expressing cells have a reduced capacity to promote PC activation by FIIa and that the efficiency of soluble TM₅₃₆ as a cofactor in PC activation is lower than that of TM_FS. Some features of TM₅₃₆ inferred from cellular overexpression models were confirmed by observations in HUVEC and endothelial colony forming cells obtained from a patient carrying the TM₅₃₆ mutation.

The clinical features of newly identified patients with TM₅₃₆ mutation were similar to those of previously studied patients,^19,20,22,23 with sTM plasma levels ∼40 times higher than normal and reduced FIIa generation. Our results further show that the high TM levels in the plasma of TM₅₃₆ patients are not due to TM-carrying microvesicles.

Cells expressing TM₅₃₆ release more sTM than TM_WT-expressing cells¹⁴ (and this work) and this release occurs via the conventional secretory pathway. Mass spectrometry analyses revealed that TM₅₃₆ is released from cells with its entire IMD, ruling out the possibility that its substantial release results from increased proteolysis. Consistent with this, TM₅₃₆ release was minimally reduced by inhibitors of MMPs and ADAMs compared with TM_WT. This result contrasts with that of a previous study showing that the MMP inhibitor GM6001 was very effective in reducing sTM₅₃₆ release.¹⁴ However, this study did not examine the effect of GM6001 on the cellular levels of TM, which may have biased the interpretation of the results. The high release of TM₅₃₆ is also not due to its increased cleavage by RHBDL2. However, the fact that TM₅₃₆ without a cytoplasmic tail is cleaved by RHBDL2 contradicts a previous study concluding that the TM cytoplasmic tail is required for its cleavage by RHBDL2.²⁹ However, in that study, the cytoplasmic tail of TM was replaced by that of other proteins and not completely deleted, as in the case of TM₅₃₆.

The detection of sTM₅₃₆ in the lumen of ESP vesicles suggests that either the protein enters the ER lumen as a secreted protein or is first inserted into the ER membrane before leaking from the membranes of the ESP vesicles. Because RHBDL2 cleaves only membrane-anchored substrates³⁰ and overexpression of RHBDL2 increases the amount of cleaved sTM₅₃₆ at the expense of full-length sTM₅₃₆, we conclude that all synthesized TM₅₃₆ is inserted into the ER membrane. Therefore, full-length sTM₅₃₆ present in CM originates from the spontaneous leakage of TM₅₃₆ from the membranes of ESP vesicles. A hydrophobic mismatch between the short IMD of TM₅₃₆ and the lipid bilayer could explain the nonpermanent nature of TM₅₃₆ insertion in membranes, leading to its release. Supporting this interpretation, the substitution of an amino acid in the IMD of TM₅₃₆ to increase or reduce its hydropathy reduces or increases the release of sTM₅₃₆. We identified dimers of TM_WT and TM₅₃₆ in cells. Because oligomerization of transmembrane proteins is a process that attenuates the hydrophobic mismatch between intramembrane α-helical peptides and the lipid bilayer,³¹ it is possible that the fraction of TM₅₃₆ that remains inserted into membranes consists of TM₅₃₆ dimers.

The high spontaneous release of TM₅₃₆ from membranes is definitely a prerequisite for increasing sTM₅₃₆ concentration in CM or plasma. However, it is conceivable that some specificities of intracellular trafficking and/or degradation of TM₅₃₆ or plasma clearance of sTM₅₃₆ may contribute secondarily to the increased sTM₅₃₆ concentrations.

We have shown that the ER chaperone/UPR regulator HSPA5 binds to the IMD of sTM₅₃₆ and detected the TM-HSPA5 complex in the plasma of TM₅₃₆ patients. Such an interaction indicates the presence of sTM₅₃₆ in the ER lumen. HSPA5 binds to hydrophobic patches in misfolded proteins,³² so its interaction with the IMD of sTM₅₃₆ in the ER is not surprising. A limited number of situations in which HSPA5 is secreted have been documented.³³ We propose that the ER retention sequence of HSPA5 is masked during its interaction with TM₅₃₆, allowing the HSPA5-TM₅₃₆ complex to be secreted.

Unlike TM_WT, TM₅₃₆ is not contained in lipid rafts. However, this specific membrane distribution does not explain its high release. Indeed, TM₅₃₉ (with the whole IMD but without the cytoplasmic tail) is not present in lipid rafts and yet is not released from cells.

Overexpressed soluble forms of TM have been shown to be more CS-modified than transmembrane TM.³⁴ Our results further demonstrate that the proportion of sTM with CS depends on the release mechanism: conventionally secreted TM_FS is more CS-modified than full-length sTM₅₃₆, which has escaped from the ER membrane. The presence of CS moiety in TM enhances its binding to FIIa.²⁸ In line with this, we show that the absence of CS decreases the ability of sTM to support in vitro PC activation. With regard to natural soluble TM variants, TM_FS, which is more CS-modified than TM₅₃₆, is also more effective in PC activation.

Our results demonstrate that TM₅₃₆-expressing cells expose less TM on their surface than TM_WT-expressing cells, in agreement with previous data showing reduced TM expression in the vessels of ovarian tissue from a TM₅₃₆ patient.²⁰ The low TM expression on the HUVEC₅₃₆ surface is consistent with the reduced PC activation measured on these cells. Furthermore, the localization of TM in lipid rafts likely has an effect on its efficacy as a cofactor in PC activation. Indeed, lipid raft disruption increases PC activation in TM_WT-expressing cells but not in cells expressing TM excluded from lipid rafts. Consequently, it is likely that the anticoagulant potential of TM₅₃₆-expressing cells is not optimal.

The exclusion of TM₅₃₆ from lipid rafts could have consequences beyond coagulation. Indeed, TM_WT, EPCR, and protease-activated receptor 1 (PAR-1) have been colocalized in lipid rafts of the endothelial cell plasma membrane.⁵ In these membrane microdomains, the TM-FII complex activates EPCR-bound PC, leading to PAR-1 activation and triggering cytoprotective signaling. Consequently, TM₅₃₆, which is excluded from lipid rafts, would not be optimally effective in triggering PAR-1-dependent cytoprotective signaling.

The low TM expression on the surface of TM₅₃₆-expressing cells is probably only partially attributable to the high release of TM₅₃₆. Indeed, as expected for an aberrant protein, TM₅₃₆ is highly degraded by ERAD, which may contribute to its low cell-surface expression. However, the release of sTM₅₃₆ associated with HSPA5 indicates that a substantial amount of this protein still escapes ERAD.

Interestingly, the release of overexpressed TM₅₃₆ is reduced by CL, an aminosterol previously shown to promote ERAD processes.²⁷ This finding proposes ERAD activation as an approach to reduce sTM release in TM₅₃₆ patients. The possibility of activating ERAD (mainly using proteasome activators) to degrade misfolded proteins to restore protein homeostasis is still at an exploratory stage and has been tested almost exclusively in the context of neurodegenerative diseases.^35,36 

In summary, our study demonstrates that TM₅₃₆ is inserted into the ER membrane. A fraction of TM₅₃₆ follows the conventional trafficking pathway as a membrane-inserted protein and is exposed at the cell surface with an abnormal plasma membrane localization (outside lipid rafts). Another fraction extracts from the ER membrane. Part of this fraction is degraded by ERAD. The remaining fraction follows a conventional secretory pathway and is released into the extracellular environment. Activation of ERAD by CL reduces the proportion of TM₅₃₆ released by the cells. Additionally, cells expressing TM₅₃₆ exhibit reduced surface expression of TM, which impairs their ability to support the FIIa-dependent activation of PC.

The authors thank Karim Fallague (Aix-Marseille University, INSERM, INRAE, C2VN, Marseille, France) for isolating human umbilical venous endothelial cells. The expression vectors for wild-type thrombomodulin, active and inactive RHBDL2, and EPCR were provided by Y. Dargaud (Claude Bernard University, Lyon, France), M. Freeman (Sir William Dunn School of Pathology, University of Oxford, Oxford, United Kingdom), and L. Mosnier (Department of Molecular Medicine, Scripps Research, La Jolla, California), respectively.

This work was supported by INSERM and the Aix Marseille University.

Contribution: C.B., A.P., and C.R. performed most of the experiments; N.H. performed thrombin generation measurement; V.E. and C.F. were in charge of patient care; S.S. performed endothelial colony forming cells isolation and culture; R.L. performed the analysis of thrombomodulin on plasma microvesicles; P.-E.M. was involved in patient management and manuscript preparation; and F.P. designed the study, supervised the work, analyzed the data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Franck Peiretti, Aix-Marseille University, INSERM1263, Center for CardioVascular and Nutrition Research, C2VN, Faculté de Médecine Timone, 27 Bd Jean Moulin, Marseille 13385, France; email: franck.peiretti@univ-amu.fr.