The proteolytic degradation of fibrin clots is mediated by plasmin, an enzyme formed by the proteolytic action of urokinase-type plasminogen activator (uPA) and tissue-type plasminogen activator (tPA) on the zymogen plasminogen (Loskutoff et al, 1989). Plasminogen activator inhibitor-1 (PAI-1, also known as SERPINE1) is a member of the serine protease inhibitor (SERPIN) superfamily and is the primary physiological regulator of uPA and tPA activity (Schleef et al, 1989). PAI-1 is a 47-kDa glycoprotein that is synthesized and secreted by various types of cells. A quarter century has passed since the discovery of PAI-1 (van Mourik et al, 1984) and its cDNA (Ny et al, 1986). Since then, a number of studies have been conducted to elucidate PAI-1 functions, both in vivo and in vitro. Although the principal function of PAI-1 is the inhibition of fibrinolysis, PAI-1 possesses pleiotropic functions besides haemostasis (Fig 1). This review will focus on the studies performed to investigate the multifunctional aspects of PAI-1, especially focussing on its clinical relevance.
Plasminogen activator inhibitor-1 (PAI-1, also known as SERPINE1) is a member of the serine protease inhibitor (SERPIN) superfamily and is the primary physiological regulator of urokinase-type plasminogen activator (uPA) and tissue-type plasminogen activator (tPA) activity. Although the principal function of PAI-1 is the inhibition of fibrinolysis, PAI-1 possesses pleiotropic functions besides haemostasis. In the quarter century since its discovery, a number of studies have focused on improving our understanding of PAI-1 functions in vivo and in vitro. The use of Serpine1-deficient mice has particularly enhanced our understanding of the functions of PAI-1 in various physiological and pathophysiological conditions. In this review, the results of recent studies on PAI-1 and its role in clinical conditions are discussed.
PAI-1 in haemostasis
Upon initiation of thrombus formation as a result of a vascular insult, PAI-1 is primarily released from platelets (Brogren et al, 2004) and also secreted by endothelial cells (Hakkert et al, 1990). Under normal physiological conditions, the release of PAI-1 attenuates tPA-mediated plasminogen activation and contributes to the stabilization of the thrombus for proper maintenance of wound healing and vascular patency.
In order to understand the in vivo functions of PAI-1 in haemostasis, phenotypic characterization of complete PAI-1 deficiency is meaningful. A few reports exist on this deficiency (Mehta & Shapiro, 2008); however, only two such cases were genetically identified and defined (Fay et al, 1992; Iwaki et al, 2011a). In the first genetically identified case, two nucleotides were inserted into exon 4 of the Serpine1 gene. This insertion caused a frameshift with a premature stop codon, resulting in a truncated PAI-1 that was not secreted from the cells (Fay et al, 1992). In the second genetically identified case, one nucleotide was inserted into exon 3 of the Serpine1 gene. As in the previous case, this insertion also caused a frameshift and premature stop, which also resulted in the formation of a truncated PAI-1 that was never secreted (Iwaki et al, 2011a). The phenotypes manifested by these SERPINE1 mutations included severe, abnormal bleeding following surgery and during menstruation as well as impaired wound healing. The bleeding tendency in these cases was the most critical and life-threatening symptom but was typically delayed and therefore not similar to that observed in afibrinogenaemia (Acharya & Dimichele, 2008).
We recently reported that spontaneous miscarriage was another remarkable phenotype related to SERPINE1 deficiency. Pregnancies in patients with both afibrinogenaemia and congenital FXIII deficiency result in spontaneous miscarriage at approximately 7–8 weeks of gestation (Iwaki & Castellino, 2005). The PAI-1-deficient patient experienced a spontaneous miscarriage due to massive genital bleeding during the 19th week of the first pregnancy. Genital bleeding began during the 16th week of gestation, in the absence of supplemental therapies. On comparing PAI-1 deficiency to congenital afibrinogenaemia or Factor XIII deficiency, it is clear that genital bleeding begins much later in complete PAI-1 deficiency (Iwaki et al, 2011b). PAI-2 is synthesized by trophoblasts, which is why it is called ‘placental plasminogen activator inhibitor’. It was hypothesized that PAI-2 could compensate for the lack of functional PAI-1 in the placenta; however, the phenotypes observed in PAI-1 deficiency seem to refute that hypothesis.
To further characterize PAI-1 functions, Serpine1-deficient mice were generated (Carmeliet et al, 1993). After initial characterization of the deficient mice, it was found that the mice were not only viable, but also able to reproduce. Moreover, they did not show any tendency towards abnormal bleeding. These results were very different from those observed in human total PAI-1 deficiency. Although the reasons for the phenotypic discrepancies observed in complete PAI-1 deficiency between humans and mice are unclear, one possible reason may be because the clot lysis in rodents is 10 times slower than that in humans (Korninger & Collen, 1981). Based on the less severe symptoms in murine complete PAI-1 deficiency, combined with the slower clot-lysis time in rodents, we propose three potential hypotheses:
- The other known protease inhibitors containing PAI and/or plasmin inhibitory functions can compensate for PAI-1 deficiency.
- There are unknown PAIs in rodents that can compensate for PAI-1 deficiency.
- Murine fibrin is more resistant to plasmin and/or coagulation activities are more dominant in mice.
It is not known which of these hypotheses explains the different observations between mice and humans. These differences indicate that the extrapolation of mouse studies to humans in this field must be done with great care.
PAI-1 in inflammation
Lipopolysaccharide (LPS), derived from gram-negative bacteria, binds to the plasma LPS-binding protein (LBP; Tobias et al, 1989), resulting in the formation of a LBP-LPS complex that can bind to a cell-surface protein, CD14, on macrophages (Wright et al, 1990). The LBP-LPS-CD14 complex transfers LPS to the toll-like protein 4 (TRL4) and its accessory protein, MD2, resulting in the activation of TRL4. Activated TLR4 stimulates the nuclear factor (NF) κB pathway and three mitogen-activated protein kinase (MAPK) pathways – extracellular signal-regulated kinases (ERK) 1 and 2, c-Jun N-terminal kinase (JNK) and p38 in macrophages (Guha & Mackman, 2001). Following these cellular events, a variety of chemokines are released, including interleukin (IL)-8 (CXCL8), macrophage inflammatory protein (MIP)-1α (CCL3), MIP-1β (CCL4), and regulated on activation, normal T-cell expressed and secreted (RANTES, renamed as CCL5; Luster, 2002). In addition, the pro-inflammatory cytokines IL-1β (Eden & Turino, 1986) and tumour necrosis factor (TNF)-α (Chen et al, 1985) are also released. Following TNFα and IL-1β stimulation, an acute inflammatory reaction protein, IL-6, is produced (Scales et al, 1992) and induces the synthesis of acute-phase proteins, such as C-reactive protein (CRP) and fibrinogen.
There are a number of studies that indicate the relationship between coagulopathy and systemic inflammation (Esmon et al, 1999). These studies indicate that TNFα and IL-1β upregulate PAI-1 synthesis in endothelial cells (Emeis & Kooistra, 1986). Moreover, IL-6 possesses prothrombotic effects, such as elevation of D-dimer levels and decreasing the levels of anti-thrombin III (Mestries et al, 1994), both of which promote platelet aggregation (Burstein, 1994) and stimulate production of both tPA and PAI-1 by endothelial cells (Mestries et al, 1994). Hence, PAI-1 is a recognized acute phase protein, like CRP and fibrinogen. PAI-1 has also been recognized as an exacerbating factor in systemic inflammation, especially sepsis-induced disseminated intravascular coagulation (DIC) (Pralong et al, 1989). Indeed, plasma PAI-1 levels are markedly elevated during systemic inflammation, especially when triggered by gram-negative bacteraemia (Brandtzaeg et al, 1990), and high plasma PAI-1 levels also correlate to elevated mortality rates and frequencies of multiple-organ failure (Madoiwa et al, 2006).
In order to dissect the pivotal PAI-1functions in systemic inflammation, several in vivo experimental models have been employed. LPS administration to wild-type (WT) mice has been found to increase Serpine1 expression in several tissues (liver, kidney, lung and adrenal), but was greatly attenuated in C3H/HeJ mice in which TNFα is not released in response to LPS (Sawdey & Loskutoff, 1991). Thus, administration of LPS also induces Serpine1 expressions via TNFα release in mice and contributes to coagulopathy during the LPS challenge (Samad et al, 1996). Conversely, the use of a PAI-1 inhibitor during LPS administration to mice results in attenuated hypercoagulation and thus reduced mortality (Murakami et al, 1997).
Serpine1 deficiency resulted in decreased levels of neutrophil infiltration into the lungs of mice with Klebsiella pneumonia-induced pneumonia; the same infection induced Serpine1 expression in WT mice. As a result, Serpine1-deficient mice demonstrated an impaired host defence, as reflected by enhanced lethality and increased bacterial growth and dissemination. These effects could be largely prevented by transgenic overexpression of Serpine1 in the lung, by using a replication-defective adenoviral vector (Renckens et al, 2007). However, Serpine1 deficiency did not influence the outcome of pneumonia initiated by gram-positive Streptococcus pneumoniae (Rijneveld et al, 2003). These results indicate that for in vivo evaluation of PAI-1 function during systemic inflammation, especially during sepsis, the types of bacteraemia and coagulopathy induced by bacteraemia should be considered separately.
PAI-1 in atherosclerosis
Atherosclerosis is a self-sustaining, inflammatory, fibroproliferative disease that progresses in discrete stages and involves a number of cell types and effector molecules (Ross, 1993). The coagulation, anticoagulation and fibrinolytic systems are potentially important in atherosclerosis because fibrin deposits and fibrin degradation products are resident in the atherosclerotic plaque (Bini et al, 1989). Several investigators have also reported that high PAI-1 levels are associated with the presence of coronary artery disease. For example, patients with coronary atherosclerosis presenting with high levels of PAI-1 and/or tPA activity have atherosclerosis-associated carotid wall thickening (Lijnen & Collen, 1996). Survivors of myocardial infarction usually have higher PAI-1 levels than the general population, and these levels are correlated with the recurrence of myocardial infarction (Zorio et al, 2008). Several reports have also demonstrated the upregulation of Serpine1 expression in human atherosclerotic plaques (Lupu et al, 1993; Schneiderman et al, 1992). These results have led to the speculation that PAI-1 is an exacerbating factor in atherosclerosis; thus, animal models, especially Serpine1-deficient mice models, have been developed to validate this hypothesis.
Because of the differences in lipid metabolism between mice and humans, rapid clearance of low-density lipoprotein (LDL) and its precursors, very-low-density lipoprotein (VLDL) and intermediate-density lipoprotein (IDL) occurs in mice. Thus, the cholesterol present in murine plasma is mostly the artheroprotective high-density lipoprotein (HDL) fraction. As a result, WT mice are more resistant than humans to diet-induced elevation of LDL. The application of gene-targeting technology has facilitated the generation of mouse strains that possess many of the characteristics of lipid metabolism found in humans (Iwaki et al, 2008). The most widely studied murine models of familial hypercholesterolaemia (FH)-mediated human atherosclerosis are the LDL receptor-deficient (Ldlr−/−) (Ishibashi et al, 1993) and apoE-deficient (Apoe−/−) (Piedrahita et al, 1992; Plump et al, 1992) mouse lines.
Apoe−/−/Serpine1−/− mouse lines have been generated and characterized by several groups (Eitzman et al, 2000; Luttun et al, 2002; Sjoland et al, 2000; Zhu et al, 2001). Zhu et al (2001) and Sjoland et al (2000) measured the total cholesterol (TC) levels in Apoe−/−/Serpine1−/− and Apoe−/− mice on a high-fat diet. Zhu et al (2001) indicated that the TC levels in Apoe−/−/Serpine1−/− mice (38.4 ± 4.6 mmol/l) were significantly higher than those in Apoe−/− mice (28.3 ± 1.3 mmol/l). However, this difference was not observed by Sjoland et al (2000) (Apoe−/−/Serpine1−/− mice: 44.4 ± 9.9 mmol/l; Apoe−/− mice: 50.2 ± 4.3 mmol/l). More extensive investigations have not been conducted; thus, the extent of lipid metabolism alteration in Apoe−/− mice with a PAI-1 deficiency remains unresolved.
Sjoland et al (2000) maintained Apoe−/− and Apoe−/−/Serpine1−/− mice on a high-fat diet for 6, 15 and 30 weeks and compared the sizes of the plaques in the aortic sinuses. They concluded that there were no significant differences in plaque sizes between animals with either genotype. Eitzman et al (2000) maintained Apoe−/− and Apoe−/−/Serpine1−/− mice on a normal chow diet for 18, 30 and 52 weeks and found that there were significantly smaller plaques in the aortic arch and carotid artery in Apoe−/−/Serpine1−/− mice relative to Apoe−/− mice. On the other hand, Luttun et al (2002) fed mice either a high-fat or normal diet for 25 weeks and analysed various aortic segments, including the aortic valves, aortic arch, thoracic and abdominal aorta, carotid artery and brachiocephalic artery. They reported that more advanced plaque formation occurred in Apoe−/−/Serpine1−/− mice than in Apoe−/− mice. Thus, there is still considerable controversy involving the role of PAI-1 in animal models of atherosclerosis; these discrepancies are often referred as the ‘PAI-1 paradox’. PAI-1 is a multifunctional protein, and its vascular functions depend on the vascular beds, type of lesion and the experimental conditions (Diebold et al, 2008).
PAI-1 functions in atherosclerosis should be studied in models with or without fibrin formation (Fay et al, 2007). In the presence of fibrin, PAI-1 demonstrates atherogenic properties by stabilizing fibrin, which facilitates migration of vascular smooth muscle cells. In the absence of fibrin, this migration is restricted through inhibition of uPA and blocking of interactions between the smooth muscle cells and vitronectin.
Before the introduction of Serpine1 deficiencies into murine spontaneous atherosclerosis models like Apoe−/− and Ldlr−/− mice, the overall in vivo impacts of Serpine1 deficiency in atherosclerosis were thought to be examinable. However, the evaluation has proven to be more difficult than expected because atherosclerotic plaque formation is highly controlled by the balance of these PAI-1 functions, i.e. it is necessary to use multiple models to dissect the multiple functions of PAI-1 in atherosclerosis.
PAI-1 in metabolic syndrome
Metabolic syndrome is characterized by a hypercoagulable state caused by increased plasma levels of coagulation factors (tissue factor, factor VII and fibrinogen) as well as inhibition of the fibrinolytic pathway (increased PAI-1 and decreased tPA activity) (Nieuwdorp et al, 2005). It is well known that increased plasma PAI-1 levels are correlated to body-mass index, plasma triglyceride and insulin levels and systolic blood pressure (Juhan-Vague et al, 1991). Recent studies have shown that relatively large amounts of PAI-1 are synthesized in both human and murine adipose tissues (Samad et al, 1996; Shimomura et al, 1996) and that PAI-1 expression is similarly correlated to body-mass index (Alessi et al, 2000). Furthermore, human visceral adipose tissue synthesizes more PAI-1 than subcutaneous abdominal adipose tissue (Alessi et al, 1997). It is also well known that increased visceral adipose tissue mass is one of the causes of metabolic syndrome. On the other hand, weight loss by either dietary restriction or comprehensive lifestyle modification is effective in lowering plasma PAI-1 levels (Skurk & Hauner, 2004). From these facts, alteration of PAI-1 levels appears to be a very promising mechanism for treating metabolic syndrome. However, further work needs to be done to clarify whether or not an increased plasma PAI-1 level is a cause or a result of metabolic syndrome.
For clarification of this issue, investigations using Serpine1-deficient mice have been helpful and instructive. Surprisingly, Serpine1-deficient mice provided with a high-fat diet developed adipose tissue more rapidly than WT mice (Morange et al, 2000). Conversely, transgenic mice overexpressing murine Serpine1 had lower body weights and lower amounts of adipose tissue (Lijnen et al, 2003). In contrast, Serpine1 deficiencies reduced the adiposity and improved the metabolic profile of genetically obese and diabetic ob/ob mice (Schafer et al, 2001). Serpine1 deficiency was also found to prevent obesity and the development of insulin resistance in these obese and diabetic mice (Ma et al, 2004). Such discrepancies can be explained by the different genetic backgrounds of mice used in the studies. However, Serpine1 deficiency itself may not affect any metabolic syndrome in mice. Although it was expected that Serpine1-deficient mice could be used to answer several important questions about the role of PAI-1 in metabolic syndromes, results from the models have complicated these questions.
PAI-1 in fibrosis
Fibrosis in multiple tissues has been associated with increased expression of Serpine1. The pathology of chronic asthma is closely related to the expression of subepithelial extracellular matrix (ECM) and airway remodelling (Kucharewicz et al, 2003). Massive infiltrations of PAI-1-producing mast cells into the airways of asthmatic patients has been observed (Cho et al, 2004), and this infiltration has been associated with elevated PAI-1 levels and diminished lung function (Cho et al, 2011). Plasmin degrades the ECM, either by directly removing glycoproteins from the ECM or by activating matrix metalloproteinases (MMPs). PAI-1 prevents ECM degradation by blocking MMP activity and promotes fibrin accumulation and fibrosis via inhibition of fibrinolysis. Experiments with Serpine1-deficient mice demonstrated that these animals were protected against the accumulation of ECM and fibrosis in their lungs after bleomycin challenge (Cho et al, 2004). In a chronic LPS challenge model, subepithelial fibrin deposition diminished in Serpine1-deficient mice, whereas airway hyperreactivity and the expansion of the subepithelial area persisted in WT mice (Savov et al, 2003). In addition, inhalation of uPA has been shown to reduce airway remodelling in a murine asthma model (Matsuo et al, 2005). A recent study has also indicated that intra-airway administration of interfering RNA targeting PAI-1 was effective in attenuating allergic asthma in mice (Miyamoto et al, 2011).
The normal human kidney does not express Serpine1. However, it is overexpressed in pathologicalal conditions associated with renal fibrosis (Xu et al, 1996). In response to acute and chronic injury, Serpine1 expression can be induced in intrinsic glomerular cells, tubular epithelial cells, macrophages and fibroblasts (Eddy, 2002). In the unilateral ureteral obstruction (UUO) model, a model of induced renal fibrosis, total kidney collagen levels were more than threefold higher in WT mice than in Serpine1-deficient mice (Oda et al, 2001). Similarly, more severe fibrosis was observed in Serpine1-overexpressing mice in the UUO model (Matsuo et al, 2005).
Observations, such as those mentioned above, raise questions regarding the mechanism of how PAI-1 promotes renal fibrosis. Previously, it was hypothesized that PAI-1, as a SERPIN, would play a critical role in promoting renal fibrosis. This was expected to occur through the inhibition of plasmin generation by uPA and tPA and the restriction of MMP activation by plasmin, leading to the accumulation of ECM in the kidney. However, the severity of fibrosis in plasminogen-deficient mice was actually attenuated in the UUO model. In such experiments, attenuation of fibrosis severity was associated with significantly lower levels of phosphorylated extracellular signal-regulated kinase (ERK) and active transforming growth factor (TGF)-β (Zhang et al, 2007). When the UUO model was also applied to Plau-deficient mice, the severity of renal fibrosis was comparable to that observed in WT mice (Yamaguchi et al, 2007). Furthermore, less severe fibrosis developed in Plat-deficient mice in the UUO model (Yang et al, 2002). Therefore, the SERPIN activity of PAI-1 does not appear to play a critical role in renal fibrosis.
As stated previously, high plasma levels of PAI-1 are well correlated to coronary artery disease and are related to the onset of myocardial infarction (Juhan-Vague et al, 1996). Similarly, a sudden increase in plasma PAI-1 levels in patients with ST-elevated myocardial infarction results in a poor survival prognosis (Collet et al, 2003). In mice, altered Serpine1 expression in the heart has also been implicated in cardiac function changes associated with age (Sobel et al, 2006a) and insulin resistance after coronary occlusions (Sobel et al, 2006b). Furthermore, plasma PAI-1 levels were markedly elevated in patients with type 2 diabetes (McGill et al, 1994), and diabetic patients have been observed to be prone to cardiac fibrosis. Therefore, it has been speculated that elevated PAI-1 augments cardiac fibrosis.
However, studies in Serpine1-deficient mice have generated unanticipated results. In one study (Askari et al, 2003), a lack of PAI-1 resulted in decreased mouse survival and an increased incidence of cardiac rupture. Therefore, in mice, PAI-1 played a central role in ventricular remodelling after myocardial infarction. Another study showed that the absence of PAI-1 in an acute myocardial infarction model exacerbated cardiac functions and infarction size with increased inflammation and haemorrhage with reduced cardiac fibrosis (Zaman et al, 2007). Recently, two groups reported that spontaneous cardiac fibrosis with chronic heart failure was observed in aged Serpine1-deficient mice (Xu et al, 2010). Cardiac fibrosis was also observed in Plau-overexpressing mice (Moriwaki et al, 2004). Cardiac fibrosis, as with other fibroses, consists of both fibrin (as a result of coagulation) and other ECMs synthesized from mesenchymal cells, such as smooth muscle cells or fibroblasts. The lack of PAI-1 facilitates fibrin degradation and also promotes cell migration and synthesis of ECM. These two competing functions of PAI-1 cloud the relationship between PAI-1 and fibrosis, making organ-specific reactions an important consideration in fibrosis.
PAI-1 in cancer
High levels of PAI-1 have been detected in several human cancers and have been correlated to poor prognoses (Andreasen, 2007). As a SERPIN, PAI-1 not only restricts proteolysis but also regulates cell migration and proliferation as a result of the high-affinity binding of PAI-1 to vitronectin (Declerck et al, 1988). This interaction prevents the association of integrins with vitronectin and downregulates cell adhesion and migration. Because of these functions, the role of PAI-1 in tumourigenesis is very controversial.
In order to explore these PAI-1 functions in vivo, Serpine1-deficient and overexpressing mice have been used. The growth and metastasis of B16 mouse melanoma tumours was not affected by PAI-1 levels (Eitzman et al, 1996), while the development of primary breast tumours was not affected by PAI-1 levels (Almholt et al, 2003). In the former case, Serpine1 expression by the tumours may have affected the results; however, in the latter case, Serpine1 expression by the tumours was completely irrelevant to tumour development. Recently, an apparent impact of PAI-1 levels on tumourigenesis was not observed in a transgenic model of multi-stage epithelial carcinogenesis (Masset et al, 2011). From these studies, it seems that PAI-1 has very little impact on either tumourigenesis or metastasis.
Conversely, several studies have shown a positive relationship between increased PAI-1 and tumour angiogenesis. The growth of transplanted T241 fibrosarcomas was significantly suppressed in Serpine1-deficient mice and was accompanied by decreased cellular proliferation and increased apoptosis (Gutierrez et al, 2000). Similar results were obtained using transplanted malignant keratinocytes (Bajou et al, 1998) and neuroblastoma cells (Bajou et al, 2008). One of the reasons for these discrepancies observed in these studies, as compared to those previously mentioned, was that the PAI-1 concentrations in the latter models were different and the level of PAI-1 is a critical factor in angiogenesis. Physiological concentrations of PAI-1 promote in vivo tumour invasion and angiogenesis, whereas super-physiological concentrations of PAI-1 inhibit tumour angiogenesis (Bajou et al, 2004). Therefore, the role of PAI-1 in cancer might vary with each experimental condition, especially the PAI-1 concentration around the tumour.
A number of studies have been performed using Serpine1−/− mice to help in understanding PAI-1 functions. On the basis of these studies, several PAI-1 inhibitors have been synthesized for use in clinical trials, to treat thrombotic events such as strokes and coronary ischemia. Other disorders, in which elevated PAI-1 plays a relevant role, as discussed in this review, are also targeted by these inhibitors. However, most of them have not been validated in humans. From the results obtained from Serpine1−/− mice, it is believed that attenuation of PAI-1 activity does not induce severe bleeding. Life-threatening bleeding observed in a PAI-1 deficient patient suggests, however, that PAI-1 inhibitors should be used with caution, as is the case with other anti-coagulants, especially in menstruating women. To prevent and/or minimize such side effects, reliable assays for PAI-1 activities to control their dosage need to be established. Additionally, the inhibitors do not react equally well with both human and murine PAI-1, even though their structures are quite similar. Furthermore, plasma PAI-1 concentrations in mice are fivefold lower than those in humans, and the levels of PAI-1 in murine platelets are 500-fold lower than those in humans. Considering these challenges, careful attention must be paid when attempting to interpolate research from mouse models to clinical understanding of humans in this field.
T.I., T.U. and K.U. wrote the paper. This work was supported in part by Japan Society for the Promotion of Science (JSPS) KAKENHI 20890093 and 22790247 (to T.I.), Japan Science and Technology Agency (JST) AS232Z01751F (to T.I.), and the Uehara Memorial Foundation (to T.I.).