Serpin structure, mechanism, nomenclature and evolution

There are 36 protein-coding serpin family members in humans, of which 30 can inhibit proteinases [5]. Serpins are subdivided into ‘clades’ — A through to I — based on sequence similarity (Table 1). Outside of these groups, serpins exhibit surprisingly little sequence similarity, except for some conserved residues (e.g. Ser⁵³ and Ser⁵⁶ ) within the ‘shutter region’ in the core of the protein. This is an area of critical importance for the serpin mechanism, which is highlighted by mutations in these positions leading to several serious human serpinopathies; a collection of diseases resulting from serpin dysfunction [6]. However, despite only low sequence similarity, serpins are remarkably similar structurally; usually consisting of 7–9 α-helices surrounding 3 β-sheets and a large flexible region above the body of the protein, known as the reactive centre loop (RCL).

Table 1.

The protein-coding serpins in humans

	Standardised nomenclature	Historical name(s)	RCL sequence (P4–P3–P2–P1⇣P1′–P2′–P3′–P4′)	Described proteinase target(s)
Clade A	SerpinA1	Alpha-1 antitrypsin; alpha-1 proteinase inhibitor	AIPM⇣SIPP	Neutrophil elastase; proteinase-3
	SerpinA2	Alpha-1 antitrypsin-related protein	EKAW⇣SKYQ
	SerpinA3	Alpha-1 antichymotrypsin	ITLL⇣SALV	Cathepsin G
	SerpinA4	Kallistatin	IKFF⇣SAQT	Tissue kallikrein
	SerpinA5	Protein C inhibitor	FTFR⇣SARL	Activated protein C
	SerpinA6	Corticosteroid-binding globulin		Non-inhibitory
	SerpinA7	Thyroxin-binding protein		Non-inhibitory
	SerpinA8	Angiotensinogen		Non-inhibitory
	SerpinA9	Centerin	FIVR⇣SKDG	—
	SerpinA10	Protein-Z dependent proteinase inhibitor	ITAY⇣SMPP	Factor Xa; Factor XIa
	SerpinA11	—
	SerpinA12	Vaspin	TLPM⇣ETPL	Kallikrein 7
Clade B	SerpinB1	Leukocyte elastase inhibitor (LEI); monocyte/neutrophil elastase inhibitor (MNEI)	ATFC⇣MLMP	Neutrophil elastase, cathepsin G, proteinase-3, cathepsin L
	SerpinB2	Plasminogen-activator inhibitor (PAI-2)	MTGR⇣TGHG	uPA
	SerpinB3	Squamous cell carcinoma antigen 1 (SCCA1)	GFGS⇣SPTS	Cathepsin K, cathepsin L, cathepsin S
	SerpinB4	Squamous cell carcinoma antigen 2 (SCCA2)	VVEL⇣SSPS	Cathepsin G, chymase
	SerpinB5	Maspin		Non-inhibitory
	SerpinB6	Placental thrombin inhibitor (PTI); cytoplasmic antiproteinase (CAP)	MMMR⇣CARF	Thrombin, plasmin, chymotrypsin, cathepsin G
	SerpinB7	Megsin	IVEK⇣QLPQ
	SerpinB8	Cytoplasmic antiproteinase 2 (CAP2)	RNSR⇣CSRM	Furin, thrombin, subtilisin A
	SerpinB9	Cytoplasmic antiproteinase 3 (CAP3)	VVAE⇣CCME	Granzyme B, caspase 1, subtilisin A
	SerpinB10	Bomapin	IDIR⇣IRVP	Thrombin
	SerpinB11	Epipin	IAVK⇣SLPM
	SerpinB12	Yukopin	VSER⇣SLRS	Trypsin, plasmin
	SerpinB13	Headpin	FTVT⇣SAPG
Clade C	SerpinC1	Antithrombin	IAGR⇣SLNP	Thrombin; Factor Xa
Clade D	SerpinD1	Heparin cofactor II	FMPL⇣STQV	Thrombin
Clade E	SerpinE1	Plasminogen-activator inhibitor 1	VSAR⇣MAPE	Plasminogen activators (tPA; uPA)
	SerpinE2	Protease nexin-1	LIAR⇣SSPP	Plasminogen activators (tPA; uPA), thrombin
	SerpinE3	—	LLKR⇣SRIP	—
Clade F	SerpinF1	Pigment epithelial-derived factor		Non-inhibitory
	SerpinF2	Alpha-2 antiplasmin	AMSR⇣MSLS	Plasmin
Clade G	SerpinG1	C1-inhibitor	SVAR⇣TLLV	C1 proteinase
Clade H	SerpinH1	HSP-47		Non-inhibitory
Clade I	SerpinI1	Neuroserpin	AISR⇣MAVL	Plasmin, plasminogen activators (tPA; uPA)
	SerpinI2	Myoepithelium-derived serine proteinase inhibitor (MEPI); Pancpin	IPVI⇣MSLA

For inhibitory serpins, the reactive centre loop (RCL) sequences are defined as P4–P4′, where (⇣) is the cleavage site for the target proteinase. Previously identified targets are also listed. List modified from [1,5,11] and RCL sequences confirmed using the UniProt Knowledgebase [46].

Inhibitory serpins have an elegant mode of inhibition. Monomeric, serpins exist in a constrained (metastable, M-state) conformation until they interact with a proteinase. Often described as ‘molecular mousetraps’, the RCL of inhibitory serpins acts as a ‘bait region’ which contains a sequence of amino acids targeted by specific proteinase(s) [7]. Upon initiation of cleavage, the serpin undergoes a rapid switch, whereby the RCL is inserted into a β-sheet within the main body of the serpin [8–10]. As this occurs prior to the ‘deacylation’ step of proteolysis, the proteinase remains covalently attached to the serpin as an acyl-enzyme intermediate. This mode of suicide inhibition involves a huge conformational change which moves the proteinase ∼70 Å to the other side of the serpin protein [9] and renders a hyperstable serpin : proteinase complex (relaxed; R-state). The molecular dynamics of the serpin mechanism has both perplexed and fascinated the structural biology field for decades, and has been investigated and reviewed extensively elsewhere (see [11,12]). Interestingly, evolutionary analyses suggest that different serpins have emerged through gene duplication events, evolving by speciation to perform particular physiological roles, rather than specific inhibition of a proteinase [1,6]. The crystal structures in Figure 1 depict the important regions of the serpin and the structural changes which occur upon complex formation with a target proteinase.

Figure 1.

Despite significant differences in sequence, the serpin superfamily display remarkable structural similarity. Serpins have 2–3 β-sheets surrounded by 7–9 α-helices and a reactive centre loop (RCL), which acts as a bait region for inhibitory serpins. Some conserved residues do exist amongst family members, for example, in the shutter region. Serpin inhibition begins with proteolytic attack of the RCL by a proteinase. Under normal circumstances, cleavage by a serine proteinase involves the formation of a covalent acyl-enzyme intermediate, followed by a deacylation step to release the cleaved products. In serpin inhibition, initial interaction with a proteinase is followed by the rapid and significant conformational of the serpin and distortion of the enzyme active site, such that deacylation cannot occur. This results in a hyperstable, covalent serpin:proteinase complex (right). Figure generated in PyMol, using structures imported from the Protein Data Bank — ID 1QLP [12], left; 1EZX [10], right.Serpin structure and inhibitory mechanism.

Cartilage destruction in osteoarthritis

Cartilage is a tissue which covers the end of long bones, lacks any vasculature or innervation, and has a single cell type — the chondrocyte. Composed of an organised extracellular matrix (ECM), it consists predominantly of type II collagen providing structural integrity, and the proteoglycan aggrecan, which provides compressive strength through osmotic water retention. Central to osteoarthritis (OA) is the destruction of this tissue and the exposure of the underlying bone, a process driven by metalloproteinases [13]. In particular, a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS) enzymes drive the initial loss of cartilage aggrecan, while matrix metalloproteinase (MMPs) can together degrade all the components of cartilage. Of these, in particular, it is the soluble collagenases, such as MMP-13, which are responsible for pathological collagen cleavage. Metalloproteinases are inhibited by a cognate family of inhibitors, the tissue inhibitor of metalloproteinases (TIMPs), of which there are four in humans (TIMP-1, -2, -3 -4) and provide a delicate balance between synthesis and degradation.

More recently important roles for a different family of extracellular proteinases — the serine proteinases — has emerged in OA. Serine proteinases can activate proMMPs, cleave cellular receptors and cytokines as well as destroy the ECM directly [14–18]. The largest family of inhibitors of these proteinases are the serpins, which until now have not been interrogated collectively in this tissue. This review is not exhaustive, but will cover important studies relating to this unique superfamily of proteins in cartilage biology and OA. The currently identified roles for serpins in this context are summarised in Figure 2.

Figure 2.

Abbreviated common names are shown in brackets for reference: SerpinA1 — AAT (alpha-1 antitrypsin); SerpinA9 (Centerin) — SerpinA3 — ACT (alpha-1 antichymotrypsin); SerpinB2 — PAI-2 (plasminogen Activator Inhibitor-2); SerpinC1 — AT (antithombin); SerpinD1 — HC-II (heparin cofactor II); SerpinE1 — PAI-1 (plasminogen activator inhibitor-1); SerpinE2 — PN-1 (protease nexin-1); SerpinF1 — PEDF (pigment epithelium-derived factor); SerpinF2 — A2AP (alpha-2 antiplasmin); SerpinH1 — HSP47 (heat shock protein 47).The putative roles and regulation of serpins in the joint.

Clade A — the ‘antitrypsin-like’ serpins

This clade forms the largest extracellular serpin sub group; containing 11 genes and 2 pseudogenes [1]. Perhaps the most studied members of this clade are SerpinA1 (also known as alpha-1 antitrypsin or alpha-1 proteinase inhibitor) and SerpinA3 (alpha-1 antichymotrypsin). These serpins inhibit several proteinases but association kinetics are most favourable for the neutrophil serine proteinases (NSPs) — neutrophil elastase and proteinase-3 in the case of SerpinA1, and cathepsin G for SerpinA3 [19]. Both serpins are described as ‘acute phase’ proteins, which are synthesised upon initiation of systemic inflammation — predominantly by the liver — and function to down-regulate the inflammatory response [20]. Interestingly, both SerpinA1 and SerpinA3 are abundantly expressed by chondrocytes [21]. Perhaps underscoring the importance of SerpinA1 and SerpinA3 in cartilage biology, both are also markedly induced during chondrogenic differentiation from mesenchymal stem cells, and have been previously earmarked as putative differentiation markers [22]. In addition, SerpinA9 (centerin) is also induced when chondrogenesis is stimulated by kartogenin ex vivo [23]. It is not yet clear what if any role these serpins have in chondrogenesis and whether this is due to their inhibitory function. Indeed, both SerpinA1 and SerpinA3 have important functions outside of their inhibitory activity which could be important within the context of cartilage biology [24,25].

As early phase proteins, it is perhaps unsurprising that levels of SerpinA1 and SerpinA3 have been shown to be increased in the serum and synovial fluid of rheumatoid arthritis patients [26] and SerpinA1 expression is increased in response to inflammatory cytokines in chondrocytes [27]. However, the expression of this serpin appears to be reduced in OA cartilage, and lower levels have been observed in OA synovial fluid compared with non-OA controls [22,28]. Intraperitoneal injection of SerpinA1 protects against destruction in the collagen-induced arthritis model of inflammatory arthritis [29], and this serpin can block collagen release from cytokine-stimulated bovine cartilage ex vivo [30]. The localisation of both SerpinA1 and SerpinA3 within the cartilage may also provide insights as to their function in the tissue. A cross-sectional quantitative proteomic study of cartilage observed that both serpins were located predominantly in the superficial cartilage layer [31]. This perhaps reflects a well-positioned defence against damaging neutrophil proteinases, which through knock-out studies have been shown to be essential for the progression of collagen-induced arthritis, a model of rheumatoid arthritis [32]. Emerging evidence suggests proteinases such as neutrophil elastase may also have a role in OA [33–35] meaning changes in levels of these serpins might be of particular importance in this context.

Clade B — the intracellular serpins

Even amongst the serpin superfamily, the clade B serpins are unique. Lacking a signal peptide, they form an intracellular clade, which likely predates extracellular serpins [1]. Despite this, several studies have reported clade members extracellularly, suggesting other modes of secretion into the extracellular space, beyond the classical secretory pathway [36]. They show remarkable similarity in gene structure across the clade (8 exons, 7 introns) and are present in mammals and birds, but not in earlier model organisms such as Caenorhabditis elegans or Drosophila melanogaster , leading to suggestions the clade is at least 300 million years old [36]. Members of this clade have diverse roles including regulating cell growth, apoptosis, immunity and protecting cells from intracellular proteolysis [37]. Few studies have investigated B clade serpins in cartilage, although it has been shown that SerpinB2 (plasminogen activator inhibitor (PAI)-2), like SerpinA9, is markedly induced during cartilage formation following MSC treatment with kartogenin [23]. Interestingly, proteomic analyses of urine suggest that SerpinB1 (leukocyte elastase inhibitor; LEI) and SerpinB3 were both significantly down-regulated in older patients with OA compared with urine from younger healthy controls without OA [38].

Clade C — antithrombin

This clade has just one member, SerpinC1 (commonly known as antithrombin), which acts a major inhibitor of coagulation enzymes and has a circulating blood concentration of 0.15 g/l [39]. SerpinC1 is poorly expressed by chondrocytes [21], but this serpin is detectable in synovial fluid with increased SerpinC1 : proteinase complexes observed in both the OA and RA patients [40]. Interestingly, SerpinC1 requires heparin for maximal inhibitory activity, which is commonly administered as an antithrombotic. This glycosaminoglycan (GAG) increases the association constant approximately 300-fold, resulting from a shift in the main sheets within the serpin core, and an extrusion of the RCL, making it more exposed to proteolytic attack [41]. The potential significance of cartilage-derived GAGs in regulating serpin activity is explored further in section ‘ECM binding’.

Clade D — heparin cofactor II

This clade also only contains one member — SerpinD1 (also known as Heparin Cofactor II) which is rapid inhibitor of thrombin, but like SerpinC1, only in the presence of GAGs; including heparin, heparan sulfate or dermatan sulfate. Despite sharing 30% sequence identity with SerpinC1, little is known about its physiological function, although a role in the response to vascular injury is likely [42,43]. SerpinD1 has been shown to bind strongly to the small leucine-rich proteoglycans decorin and bigylcan [44], both of which are abundant in cartilage. Proteomic analysis of synovial fluid has demonstrated that SerpinD1 is elevated in OA compared with healthy controls [45].

Clade E — plasminogen activator inhibitor-1 and proteinase nexin-1

This clade is home to three serpins, only two of which are well described — SerpinE1 (plasminogen activator inhibitor-1) and SerpinE2 (protease nexin-1). These serpins share 40% sequence identity and significant structural overlap [46]. Like many serpins, they target trypsin-like serine proteinases due to an arginine residue in the P1 position in their RCL [5] but they also share an additional commonality in other RCL amino acids (Table 1). SerpinE1 is a well-established regulator of fibrinolysis through inhibition of urokinase and tissue-type plasminogen activators (uPA and tPA, respectively; [47]). SerpinE2 is also a potent inhibitor of uPA, as well as the coagulation proteinase, thrombin [48]. It is important to recognise that despite the biochemical similarities, the physiological functions of these serpins display marked differences, perhaps best evidenced by phenotypes of mice deficient for these serpins. SerpinE1^−/− mice display altered clot lysis [49], whilst SerpinE2^−/− mice display neurological abnormalities and males are infertile [50,51]. As with the clade A serpins, both SerpinE1 and SerpinE2 are induced by pro-inflammatory cytokines in chondrocytes [52–54] and SerpinE1 has been shown to be induced by both mechanical loading and fluid shear stress [55,56]. In OA, some studies have observed decreased levels of both SerpinE1 [57] and SerpinE2 [21] in the cartilage, perhaps suggesting an increased proteolytic load. Indeed, we have demonstrated that both serpins are able to protect against cartilage collagen breakdown in an ex vivo model of cartilage destruction, likely through the inhibition of proteolytic activators of MMPs (as yet unpublished observations). Intra-articular SerpinE2 administration was also able to protect against joint destruction in a rabbit model [58]. Santoro and colleagues demonstrated that SerpinE2 can block interleukin (IL)-1 induced expression of MMP-13 in chondrocytes, in a mechanism which appears to involve ERK, NFκB and AP-1 [53] with similar observations demonstrated in the cartilage of the intervertebral disc [59]. Neither study investigates the involvement of serine proteinases in these observations, and it is possible that SerpinE2 may have chondroprotective roles outside of proteinase inhibition. Interestingly, an inverse observation has been made for SerpinE1, as murine chondrocytes from SerpinE1^−/− mice show a reduction in IL-1 induced Mmp13 expression [60]. These mice also display accelerated subchondral osteopenia compared with wild-type mice following surgical induction of OA in ovariectomized female mice [61].

The effect of clade E serpin genes on joint pathology is not exclusively at the protein level. In 2019, Shen and colleagues identified a circular RNA (circSERPINE2) which is down-regulated in OA and appears to regulate catabolic factors in the cartilage, acting as a sponge for microRNAs. The authors demonstrated that adenoviral overexpression of circSERPINE2 was able to protect cartilage in a rabbit OA model [62].

Clade F — pigment derived epithelial factor and alpha-2 antiplasmin

This clade consists of 2 members SerpinF1 and SerpinF2. SerpinF1 is perhaps better known as pigment epithelium-derived factor (PEDF). Devoid of inhibitory activity, this serpin has anti-angiogenic, neurotrophic and differentiation-inducing properties, and numerous studies have described its role cancer [63]. SERPINF1 acts predominantly through the cognate receptor PEDFR, and its action appears to be highly regulated by binding to ECM components [64–66]. In the joint, the role of SerpinF1 is catabolic, promoting cartilage destruction. Indeed, SerpinF1 has been reported to be increased in OA cartilage [45], and a recent study demonstrated that SerpinF1 deficiency reduces cartilage damage in an age-dependent manner in the murine monoiodoacetamide OA model, and overexpression in chondrocytes enhances cytokine-induced expression of catabolic MMPs [67]. In support of these observations, a different study demonstrated that stimulation of chondrocytes with recombinant SERPINF1 led to a catabolic phenotype and chondrocyte terminal differentiation [68].

The other serpin in this clade, SerpinF2 (also known as alpha-2 antiplasmin) is the primary inhibitor of plasmin, a proteinase which functions to cleave fibrin to promote clot disruption. Plasmin can also activate MMPs and contribute to ECM turnover [14]. In haemophilic mice, it has been demonstrated than fibrinolytic proteinases are liberated from the synovium during hemoarthrosis [69] and that intra-articular SerpinF2 administration was able to reduce cartilage damage and synovitis [70]. Any protective effect of this SerpinF2 in the OA joint has not yet been investigated, however.

Clade H — HSP47 — an essential collagen chaperone

The only serpin in its clade, SerpinH1 (also known as heat shock protein (Hsp)47) is a non-inhibitory serpin which has a primary function as a collagen chaperone within the endoplasmic reticulum (ER). Unlike other chaperones, this serpin appears to have only one client protein, and binds at regular sites along the collagen triple helix [71]. Cells deficient for this protein show accumulation of collagen aggregates in the ER [72]. It is, therefore, essential for new collagen synthesis and normal cartilage development and homeostasis. Indeed, homozygous missense mutations in the SERPINH1 gene result in osteogenesis imperfecta [73] and mice deficient for SerpinH1 die shortly after birth, exhibiting generalised chondrodysplasia and bone abnormalities [74]. Phylogenetic analysis demonstrates the SERPINH1/HSP47 gene to be present in cartilaginous fish such as the Japanese lamprey, dating back approximately 500 million years [75].

A role for SerpinH1 in OA is not yet clear. Changes in levels of collagen synthesis during different stages of cartilage catabolism in OA are, however, well described [76]. The corollary of which is that an essential collagen chaperone will also have an important role in such changes, but this has not yet been determined experimentally.

ECM binding

Cartilage ECM is rich in proteoglycan; predominantly aggrecan, but also more minor components such as biglycan, decorin and perlecan. GAG chains provide the correct osmotic potential in the cartilage to retain water and resist compression, but they also act as a major regulators of cartilage ECM proteins, including both metalloproteinases and their inhibitors [77,78]. GAGs can not only bind and sequester serpins, but for some also enhance their inhibitory activity, by causing large changes in conformation, making proteolytic attack of the RCL more favourable (see [79]). Many serpins are recognised as binding to GAGs and protein structures are often crystallised in complex with them (usually heparin; [48,80]). It is plausible that serpin inhibitory activity can be enhanced in the presence of GAGs within cartilage, and as our understanding of the role of serpins in this tissue improves, so too will their association with, and the relevance of, particular cartilage GAG chains. Sulphation patterns of GAGs are critical to the binding of serpin exosites [81]. In cartilage, sulphation has been shown to be important for GAG binding of ADAMTS-5 and TIMP-3 [78], and the sulphation pattern of important cartilage GAGs changes in OA, which may impact on their function in the tissue [82,83].

Endocytosis

Endocytosis is becoming established as a mode of metalloproteinase regulation in cartilage and OA. The endocytic receptor LRP-1 is a large transmembrane protein which plays a critical role in the regulation of MMPs, ADAMTS proteinases and TIMPs by binding and removing them from the extracellular space [84–87]). Indeed, there is evidence that LRP-1 and GAGs compete for the binding of TIMP-3 [78]. Serpin binding to LRP-1 is well established; initial observations suggested the involvement of a ‘serpin/enzyme complex (SEC) receptor’ which was later identified to be LRP-1 [88]. Many serpin : proteinase complexes have been identified as LRP-1 ligands (see [88,89]). It has been suggested that for most serpins, little LRP-1 binding of monomeric serpin (native or cleaved) occurs, whereas those in complex with a proteinase are readily endocytosed [88], implying a common mode of clearing complexes from the extracellular milieu. LRP-1 shedding is increased in OA cartilage [90], but it is not yet clear how this affects serpin levels within the tissue.

Proteolytic inactivation by MMPs

Interaction between serine and metalloproteinase families is well established as an important factor in the breakdown of cartilage in arthritis [14], and serpin inactivation by MMPs is another example. Cleavage occurs within the serpin RCL, removing the bait region and rendering the serpin inactive. In vitro , SerpinA1 and SerpinA3 are inactivated by MMP1, MMP-2 and MMP-3 [91,92], while SerpinA1 has also been demonstrated to be inactivated by MMP-9 in vivo [93]. Furthermore, SerpinF2 is rapidly inactivated by MMP-3 [94] and we have recently demonstrated that MMP-13 — the major OA collagenase — is also able to rapidly inactivate SerpinA1 (as yet unpublished observations). It is not yet clear what purpose this regulation serves, but it is likely to have a significant impact on the proteolytic burden in tissues where serine and metalloproteinase activity is of particular importance, such as the OA joint. Indeed, proteolytic inactivation of serpins in synovial fluid from arthritis patients has been observed [95,96] but it is unclear the degree to which serpin ‘disarming’ exacerbates catabolism. The development of neo-epitope antibodies specific for MMP-cleaved serpins could provide the foundation for understanding how serpins control the proteolytic environment and how this changes with the progression of joint destruction in OA.