Antithrombin: an anticoagulant, anti-inflammatory and anti-bacterial serpin

Anticoagulant function of antithrombin

In addition to being an anticoagulant serpin by inhibiting the proteolytic activity of coagulation proteases of both intrinsic and extrinsic pathways, antithrombin (AT) also exhibits potent anti-inflammatory activities [1–3]. The protease inhibitory activity of AT is mediated by binding to the active-site pocket of coagulation proteases through an exposed and mobile reactive center loop (RCL), and undergoing a large-scale conformational change that results in the trapping of the target protease in inactive stoichiometric and covalent complex [4,5]. The inhibitory function of AT is dramatically accelerated by heparin and this is the basis for the extensive use of therapeutic heparins for prophylaxis and treatment of thrombotic complications [6]. The mechanism by which therapeutic heparins promote the inhibition of coagulation proteases by AT has been extensively studied. It has been demonstrated that the binding of a distinct pentasaccharide (H5) sequence of heparin containing 3-O-sulfate (3-OS) modification to a basic helical motif referred to as D-helix induces a conformational change that activates AT, thereby improving its reactivity with vitamin K-dependent coagulation proteases [7–9]. Long chain high molecular weight heparin also contributes to rapid inhibition of coagulation proteases by a bridging mechanism [10]. This mechanism of the cofactor function for heparin in AT-mediated inhibition of coagulation proteases has been primarily established based on studies conducted in in vitro systems, nevertheless, it has been traditionally assumed that this can also occur in the in vivo system by AT binding to specific vascular heparan sulfate proteoglycans (HSPGs) lining the vessel wall. In support of this hypothesis, it has been demonstrated that a small population of vascular HSPGs (1–5%) has glycosaminoglycans (GAGs) containing H5 sequences with the same characteristic 3-OS modification that has been found in therapeutic heparins [11,12].

Anti-inflammatory signaling function of AT

The anti-inflammatory function of AT is mediated through the same D-helix-dependent interaction of AT with 3-OS containing vascular HSPGs [3,13). Thus, it has been found that the binding of AT to vascular GAGs induces prostacyclin (PGI₂) production that culminates in inhibition of NF-κB activation, down-regulation of the expression of vascular cell adhesion molecules and inhibition of synthesis of pro-inflammatory cytokines [1–3,13–15]. Results from several groups including ours have indicated that syndecan-4 (Synd-4) is an HSPG receptor that can bind AT to transmit the anti-inflammatory effect of the serpin in endothelial cells (Fig. 1) [3,16,17]. We have demonstrated that genetic knockdown of Synd-4 by the siRNA approach markedly inhibits the anti-inflammatory signaling function of AT [3]. The 3-OS moiety of GAGs in endothelial cells is added by the gene encoding heparan sulfate 3-O-sulfotransferase-1 (3-OST-1) [9]. To investigate the contribution of D-helix-dependent interaction of AT with 3-OS containing GAGs for the regulation of coagulation and inflammation under different pathophysiological conditions, several years ago, HajMohammadi et al. prepared mutant mice lacking the 3-OST-1 gene [9]. Unexpectedly, the 3-OST-1 deficient mice had normal hemostasis and exhibited comparable phenotypes as the wild-type mice in response to prothrombotic challenges [18]. However, 3-OST-1 deficient mice exhibited severe pro-inflammatory phenotypes in response to LPS challenge that was markedly reduced by treatment with plasma-derived AT [18], raising the possibility that interaction of AT with 3-OS containing vascular GAGs may actually be primarily responsible for the anti-inflammatory rather than the protease inhibitory function of AT. Further support for this hypothesis came from the observation that the genetic knockdown of 3-OST-1 (the primary isoform involved in 3-OS modification of vascular GAGs) by the siRNA approach [13], markedly inhibited anti-inflammatory signaling function of AT. Interestingly, when AT is cleaved by a protease or is converted to a latent form, the serpin loses its affinity for 3-OS GAGs and the D-helix-dependent binding of these low-affinity heparin conformers of AT to vessel wall HSPGs elicit antiangiogenic responses [19–21]. Thus, the significance and the extent of the contribution of D-helix-dependent interaction of AT with vascular GAGs to its anticoagulant (promotion of the protease inhibition), anti-inflammatory and anti-angiogenic functions under different pathophysiological conditions remains an open question that requires further investigation.

Figure 1.

A hypothetical model for the physiological functions of α and β isoforms of AT. AT circulates in plasma as α-AT (90–95%) and β-AT (5–10%) isoforms. Both isoforms bind heparin via their basic D-helix. The affinity of β-AT for 3-OS containing GAGs including heparin and/or those covalently attached to vascular HSPGs (i.e., Synd-4), is 5–10-fold higher than the affinity of α-AT for the same type of GAGs. Since only a small population of vascular HSPGs contain 3-OS modified GAGs, β-AT preferentially occupies these vascular sites, thereby initiating anti-inflammatory signaling responses through induction of PGI₂ synthesis by endothelial cells. Thus, this model predicts that the less abundant β-AT primarily functions as a signaling serpin in endothelial cells and the highly abundant α-AT primarily functions as a coagulation protease inhibitory serpin in circulation. The manuscript by Papareddy et al. has discovered that β-AT also preferentially binds to three cell surface receptor proteins; CD13, CD300f, and LRP-1, on monocytes to initiate anti-inflammatory responses. The mechanism by which β-AT interacts with these receptors has not been investigated. Moreover, since these receptors are glycoproteins, it cannot be ruled out that β-AT interacts via it’s D-helix with carbohydrates attached to these receptors by a mechanism similar to that observed in endothelial cells. Further investigation is required to answer this question. Monocytes are known to also express HSPGs including Synd-4 (not shown). See the text for more details. Figure was prepared by software provided by Biorender.com.

Anti-inflammatory function of AT in vivo

The physiological relevance of the anti-inflammatory signaling function of AT has been established in different animal models of injury, inflammation and severe sepsis. It has been found that AT exhibits potent anti-inflammatory effects in these models and that the serpin reduces mortality from severe sepsis in animal models [1,2,13,22–27]. However, in a recent randomized, placebo-controlled human clinical trial (KyberSept), AT administration did not show any beneficial effect on mortality in patients with severe sepsis, and the use of AT was associated with an increased risk of bleeding [28,29]. It should be noted that this study used a low dose of heparin concomitant with AT in the trial, and a post hoc analysis of a subgroup of patients receiving a high-dose AT without heparin showed a clear benefit in the 90-day mortality rate [29]. Thus, it is possible that the use of heparin had flawed the outcome of this study. Since the interaction of AT with HSPGs on the endothelium is necessary for the anti-inflammatory effect of AT, it is believed that heparin may have antagonized this effect in the septic setting. A quandary that has not been understood with respect to the anti-inflammatory function of AT in different animal studies is that the protective activity of AT required very high doses of the serpin. Noting that AT has a plasma concentration of ~2.5 μM and exhibits a K_D of 10–20 nM for H5 [30,31], and assuming that it has the same affinity for the 3-OST containing vascular HSPGs, it is unclear why the protective activity of AT requires a supraphysiological concentration of the serpin. A possible answer for this question that may be worth pondering is the observation that AT circulates in plasma in two α and β isoforms with α-AT representing the majority (90–95%) and β-AT representing only 5–10% of the total plasma AT [32,33]. β-AT lacks the carbohydrate at Asn-135 and thus binds heparin with 5–10-fold higher affinity [32,33]. To test the possibility that β-AT may be the primary AT isoform that mediates the protective anti-inflammatory function of the serpin, several years ago, we prepared recombinant equivalent of β-AT by substituting Asn-135 of AT with a Gln residue and evaluated the signaling function of the variant serpin in both cellular and animal models [13]. The results of both cellular and animal models demonstrated that the AT mutant lacking the Asn-135 carbohydrate side chain had markedly improved anti-inflammatory signaling function [13].

Anti-inflammatory receptors for β-AT on monocytes

In an interesting recent study, Papareddy et al. investigated the anti-inflammatory properties of AT during bacterial infection and found that circulating AT levels have been significantly decreased in infectious disease patients in particular in those who had been admitted to the intensive care unit [34]. Interestingly, their further analysis revealed that the extent of the decrease in β-AT is much more dramatic than α-AT in these patients, suggesting that β-AT is preferentially consumed to a greater extent under severe infectious disease conditions in plasma samples [34]. This observation motivated the authors to fractionate the plasma AT to their separate isoforms and look at their ability to interact with isolated human blood polymorphonuclear neutrophils and monocytes by FACS analysis. Although both isoforms bound these cells, nevertheless, β-AT exhibited more binding to both cell types. The authors then decided to use a TriCEPS™-based ligand-receptor capture assay system to identity proteins/receptor(s) on monocytes that may interact with β-AT. This method identified cathepsin G (CatG) and three membrane spanning receptor proteins; CD13 (also known as aminopeptidase N), CD300f (also known as LMIR3, CLM-1, IgSF13, IREM-1 or CMRF35-like molecule 1), and LRP-1 (lipoprotein receptor-related protein 1) that bind to β-AT and may be involved in monocyte cell signaling [34]. Using surface plasmon resonance (SPR) technology, the affinity of two isoforms of AT for interaction with CatG and recombinantly expressed extracellular parts of CD13 and CD300f was analyzed. It was found that while both isoforms of AT can bind with similar affinity to CatG, however, only β-AT could inhibit the proteolytic activity of this neutrophil-derived secreted serine protease. The SPR data also revealed that β-AT binds to both CD13 and CD300f with higher affinity than α-AT. Both of these receptors are known to be negative regulators of inflammatory signaling on monocytes with CD13 playing a role in internalization of Toll-like receptor 4 (TLR4) and CD300f being involved in blocking both MyD88 and TRIF-mediated TLR signaling through activation of Src homology region 2 domain-containing phosphatase 1 (SHP-1) [35–37]. Based on these observations, it has been hypothesized that β-AT contributes to modulation of innate immune responses during infection. To confirm this hypothesis, stimulation with β-AT, but not with α-AT could trigger the phosphorylation of SHP-1 in THP1 cells [34], an event that is known to be associated with mobilization of NF-kB [36,38]. Similarly, stimulation of THP1 cells with lipopolysaccharide (LPS) with β-AT, but not with α-AT blocked NF-kB activation, suggesting that β-AT elicits anti-inflammatory responses in stimulated monocytic cells through interaction with these receptors that are known to be negative regulators of cell signaling. As indicated above, it has been well established that AT through D-helix-dependent interaction with the 3-OS containing GAG moiety of HSPGs (i.e., Synd-4) elicits anti-inflammatory responses on both endothelial cells and monocytes [3,14–18]. These interesting results are the first to demonstrate that β-AT induces anti-inflammatory signaling in monocytes through cell surface CD13 and CD300f receptors (Fig. 1). Since both receptors are glycoproteins with CD13 containing 11 N-linked potential glycosylation sites [39,40], further studies will be required to understand whether these receptors, similar to Synd-4, contain high affinity AT-binding GAGs that transmit the signaling function of AT in monocytes (Fig. 1). Moreover, the primary difference between β-AT and α-AT is thought to be a difference in their affinity for GAGs with the former having 5–10-fold higher affinity for 3-OS containing heparan sulfates. Thus, both in vitro and in vivo studies have demonstrated that β-AT exhibits significantly higher anti-inflammatory effects than α-AT at identical doses [13]. Thus, it will be interesting to investigate whether higher doses of α-AT can also dampen inflammatory signaling CD13 and CD300f receptors or the effect is unique and specific for β-AT. It will be also interesting to determine whether this mechanism of AT-mediated CD13 and CD300f signaling is unique for monocytes or if it can also occur in endothelial cells. One would expect that D-helix of AT is involved in interaction with these receptors since as indicated above the only known difference between β-AT and α-AT is in their affinity for interaction with GAGs.

Anti-bacterial effect of β-AT

The manuscript by Papareddy et al. then demonstrates that AT can bind to both LPS and the Gram-negative bacteria E. coli and Pseudomonas aeruginosa [34]. Interestingly, it has been found that the binding of β-AT, but not α-AT, to LPS and bacteria leads to formation of aggregates and enhanced disruption of the bacterial cell wall. Furthermore, only the β isoform has been found to be effective in increasing the phagocytic activity of macrophages and subsequent clearance of bacteria as demonstrated by the uptake of fluorescent labelled E. coli particles by RAW264.7 cells. To investigate the specific anti-bacterial function of β-AT further, the authors decided to use CRISPER-Cas9 technology to overexpress the two human isoforms of AT in mice. It has been found that challenging these mice with LPS results in critically reduced levels of β-AT without a significant decrease in the α-AT protein level. Additionally, analysis of mouse blood revealed increased consumption and higher affinity binding of β-AT to inflammatory cells, a property that is also observed if human AT isoforms are incubated with human blood. Furthermore, mice overexpressing β-AT, but not α-AT, exhibited significantly reduced LPS-induced pulmonary lesions and inflammatory cytokine levels (including TNF-α) and improved survival rates. Surprisingly, normalization of the platelet-count and reduced liver damage was also observed only in mice overexpressing β-AT. Interestingly, incubation of E. coli in the blood of mice overexpressing β-AT led to bacterial lysis and subsequent lower bacterial count, providing further support for the observation that purified β-AT functions as an antimicrobial agent. By contrast, blood from mice overexpressing human α-AT showed higher number of bacterial counts when incubated with E. coli. Nevertheless, analysis of various organs such as the liver, kidney and spleen of the transgenic mice infected with E. coli indicated a significant decrease in the number of bacterial colonies in mice overexpressing both isoforms of AT.

Given these findings, the authors speculate that β-AT may have therapeutic utility as an antimicrobial drug in infectious disease settings, in particular in patients with sepsis. To provide support for this possibility, they compared the therapeutic effects of the two AT isoforms in normal mice challenged with lethal doses of either LPS or live E. coli. The administration of equal doses of either human α-AT or β-AT, 1 h to 5 h after challenge, markedly prolonged the survival time in both models for mice treated with β-AT but not with α-AT. The multiplex immunoassay analysis of plasma samples from infected mice revealed a pronounced reduction of inflammatory cytokines in mice treated with the β isoform compared to a modest reduction with α-AT. The growth of bacterial colonies in selected organs (lungs, livers and spleens) in mice treated with β-AT was lower than the mice treated with α-AT and associated pulmonary lesions were also much lower in mice treated with β-AT. Severe bleeding was detected in the lung tissue of control mice and mice treated with α-AT, surprisingly however, mice treated with β-AT exhibited no bleeding complications nor signs of lung damage. Bleeding in this model can be caused by consumptive coagulopathy, severe thrombocytopenia and/or direct inhibition of thrombin or its generation by factor Xa as the result of the administration of therapeutic AT isoforms. Since the latter function (the protease inhibitory function) is expected to be similar for both isoforms of AT, it is likely that the anti-inflammatory cell signaling by β-AT leading to prostacyclin expression in endothelial cells, inhibition of tissue factor expression on monocytes, normalization of platelet counts (see above) and the direct antimicrobial effect are primary contributing factors for the protective effects observed in mice treated with β-AT, but not with α-AT. Since the only known structural difference between the two isoforms of AT lies in β-AT lacking one of carbohydrate side chains near the heparin-binding D-helix of the serpin, it is not expected that the heparan sulfate-independent reactivity of the two isoforms with coagulation proteases to be significantly different. Thus, further studies will be required to understand the basis for functional differences between α-AT and the β-AT with respect to their effect on bleeding in the infectious mice model. Furthermore, all studies of the manuscript have compared the effect of equimolar concentrations of AT isoforms. Thus, analyzing the concentration-dependence of the protective effects with the two isoforms is required to determine whether the high affinity interaction of β-AT with the cellular receptors is responsible for the better treatment outcome or the effect is specific for β-AT and higher concentrations of α-AT will not recapitulate the results. The antimicrobial effect that is prominent in β-AT and missing in α-AT is also very intriguing since based on previous publications by the authors it appears that this effect is mediated through D-helix of the serpin [41]. This conclusion is based on the observation that a synthetic peptide derived from D-helix of either AT or heparin cofactor II, another heparin-binding serpin, also exhibits antimicrobial and anti-LPS effects [41–43]. In one of these previous studies, it was demonstrated that the antimicrobial activity of AT required extensive degradation of the serpin by bacterial and/or host enzymes and that the intact AT was devoid of this function [41]. Thus, further studies will be required to determine whether the antimicrobial effect of β-AT is mediated through a similar mechanism and if the presence of the N-linked carbohydrate side-chain of Asn-135 in α-AT interferes with this effect. It might also be worth preparing recombinant equivalence of α and β isoforms of AT and further compare their anticoagulant and anti-inflammatory activities in cellular and animal models. For studying the β-AT function, we have substituted Asn-135 of AT with a Gln and demonstrated that the resulting AT variant possesses significantly higher anti-inflammatory activities in both cellular and in vivo models [13]. It is also known that a Ser residue in the third position of the N-glycosylation consensus sequence of AT, is responsible for the partial glycosylation of AT and the minor β-AT isomer in plasma [44]. Thus, it has been possible to express the α-equivalence of AT by substituting this residue (Ser-137) with a Thr [44].

In summary, the studies of the manuscript by Papareddy et al. provide a framework for further evaluating the potential therapeutic utility of β-AT in treating infectious diseases including severe sepsis for which no approved pharmaceutical drug is available.