Link between plasminogen activator inhibitor-1 and cardiovascular risk in chronic hepatitis C after viral clearance

The pathophysiological implications of plasminogen activator inhibitor-1 (PAI-1) in HCV infection remain obscure. This prospective study evaluated 669 HCV patients, of whom 536 had completed a course of anti-HCV therapy and had pre-, peri- and post-therapy measurements of various profiles, including PAI-1 levels. Multivariate analysis demonstrated, before anti-HCV-therapy, platelet count and PAI-1-rs1799889 genotype were associated with PAI-1 levels. Among patients with a sustained virological response (SVR, n = 445), platelet count was associated with PAI-1 level at 24 weeks post-therapy. GEE analysis showed that PAI-1-rs-1799889 and interferon-λ3-rs12979860 genotypes affected PAI-1 levels early and late in therapy, respectively. At 24 weeks post-therapy, higher lipid, brain natriuretic peptide, homocysteine and PAI-1 levels and PAI-1 activity were noted only in SVR patients compared with pre-therapy levels. Within 24 weeks post-therapy, 2.2% of the SVR (mean age: 57.8 yr; 8 smoking males; the 2 females had pre-therapy hypercholesteremia or cardiovascular family history of disease) and 0% of the non-SVR patients experienced a new cardiovascular event. Platelet counts consistently correlated with PAI-1 levels regardless of HCV infection. PAI-1-rs-1799889 and interferon-λ3-rs12979860 genotypes mainly affected PAI-1 levels longitudinally. Within 24 weeks post-anti-HCV therapy, the SVR patients showed increasing PAI-1 levels with accelerating cardiovascular risk, especially the vulnerable cases.

The secretion of PAI-1 is stimulated by insulin, free fatty acids, atherogenic lipoproteins and chronic inflammation 6 . Consequently, PAI-1 levels are elevated during thrombotic, fibrotic, and cardiovascular events, as well as in the presence of hyperinsulinemia, hyperlipidemia, hypertension, non-alcoholic fatty liver disease and malignancy, especially those cancers with metastatic behavior and poor prognosis [4][5][6][7] . In contrast, PAI-1 secretion decreases with weight loss 8 . Additionally, PAI-1 works as an early response protein to modulate hepatocyte growth and differentiation 9 and as a marker of senescence 10 . Previous studies have suggested that a single guanosine insertion/deletion (4 G/5 G) polymorphism in the promoter region and another nearby single nucleotide polymorphism (SNP) in the PAI-1 gene [SERPINE1 (7q22.1)], as well as several SNPs in chromosomes other than chromosome 7, are associated with circulating levels of PAI-1 11 . These polymorphisms potentially increase the risk of IR and thrombosis formation 12,13 . With regard to the 4 G/5 G polymorphism, the 4 G allele is associated with moderately high PAI-1 levels. Furthermore, possible gene-environment interactions complicate this association, as differences in PAI-1 levels between individuals with the 4 G vs. the 5 G allele are more apparent in the presence of diseases that stimulate PAI-1 expression 14 . The dynamics of PAI-1 are particularly unpredictable in patients with chronic hepatitis C (CHC), as hyperfibrinolysis may occur due to elevated tissue-type plasminogen activator (tPA) levels 5 . In addition, HCV genotype-specific sustained virological response (SVR)-associated lipid alterations 2,15 might complicate the transcriptional regulation of PAI-1 1,2 . Given its complex gene/environmental coordination, the role of PAI-1 in cardiovascular risk for CHC patients remains elusive. Therefore, we sought to address the aforementioned enigmas by conducting a prospective study of PAI-1 levels in patients infected with various genotypes of HCV who had completed anti-HCV therapy. The study was adjusted for crucial confounders, including demographic characteristics as well as metabolic, liver, viral and genetic profiles.

Results
PAI-1-rs1799889 genotype correlated with pre-therapy PAI-1 levels in patients with CHC. The baseline (pre-therapy) characteristics of the patients with CHC are listed in Table 1. Patients with an SVR had lower levels of HCV RNA, homeostatic model assessment of insulin resistance (HOMA-IR) and brain natriuretic peptide (BNP) and a lower prevalence of cirrhosis and genotype 1 (G1) HCV infection. However, they had higher prevalences of G2 HCV infection and the interferon-λ 3 (IFNL3) CC genotype and higher platelet counts than the non-SVR patients. No notable differences were found in PAI-1 levels between the patients with and without an SVR.

Independent factors associated with pre-therapy PAI-1 levels in patients with CHC. Before
anti-HCV therapy, as listed in Table 2, univariate and multivariate regression analyses showed that age, body mass index (BMI), platelet count and triglyceride (TG) levels as well as the PAI-1-rs1799889 4 G/4 G genotype were associated with PAI-1 levels. HCV genotype and white blood cell (WBC) count were associated with HCV RNA expression. HCV genotype, BMI, liver cirrhosis and IFNL3 genotype were associated with SVR. Independent factors associated with the longitudinal trend in PAI-1 levels in patients with CHC. Using the GEE method, the factors longitudinally affecting PAI-1 levels were identified and are listed in Table 3. We found that sex, BMI, treatment, HOMA-IR, aspartate aminotransferase (AST) to platelet ratio index (APRI), platelet count, estimated glomerular filtration rate (eGFR), total cholesterol (TC), TG levels and PAI-1-rs1799889 and IFNL3-rs12979860 genotypes were independently associated with PAI-1 levels. Among these factors, the impacts of the categorical variables sex and PAI-1-rs1799889 and IFNL3-rs12979860 genotypes on PAI-1 levels were further elucidated and are presented in Fig. 1. The following observations were made: (1) Male patients consistently had higher PAI-1 levels than female patients throughout the therapy (Fig. 1A). (2) The positive impact of the PAI-1-rs 1799889 4 G/4 G genotype on PAI-1 level was most evident at the beginning of the therapeutic course and diminished as the course proceeded (Fig. 1B). (3) Patients with the IFNL3-rs12979860 CC genotype had higher PAI-1 levels than those with the non-CC genotype in the later part of the therapeutic course compared to the beginning (Fig. 1C).
Only patients with an SVR had increased PAI-1, BNP and homocysteine levels 24 weeks post-therapy. As shown in Table 4, paired t-tests demonstrated that the patients with an SVR had decreased ALT and APRI levels as well as increased TC, TG, PAI-1, BNP and homocysteine levels at 24 weeks post-therapy compared with pre-therapy levels. In contrast, the patients with no SVR did not exhibit these changes. Moreover, although the patients with an SVR had lower pre-therapy BNP levels than those without (Table 1)  The associations identified between independent factors and PAI-1 levels are shown in Fig. 2.
New cardiovascular events were noted only in patients with an SVR within 24 weeks post-therapy. Before anti-HCV therapy, metabolic syndrome was noted in 29.1% of the patients. At 24 weeks after anti-HCV therapy, 27% and 0% of the patients with no SVR had metabolic syndrome and cardiovascular events, respectively. In contrast, metabolic syndrome and new advent of cardiovascular events were noted in 27.8% and 2.2% (n = 10) of the patients with an SVR, respectively [cardiovascular events, non-SVR patients (0%) vs. SVR (2.2%), p = 0.025]. Among the 10 patients with an SVR who had cardiovascular events, eight were males who smoked. Of those, 3 suffered from acute myocardial infarction (two were rescued by angioplasty with stent implantation, and one was saved by coronary artery bypass grafting), 5 had acute cerebrovascular events (two were cerebral infarctions, one was an intracerebral hemorrhage, and two were transient ischemic attacks). Of the affected female patients, one had a strong family history of cardiovascular events and suffered from acute cerebral infarction. The other had pre-therapy hypercholesteremia and suffered from acute intestinal ischemia due to mesenteric vascular thrombosis. The mean and median age of the 10 patients were 57.8 and 55.0 years old, respectively. Most of the new cardiovascular events occurred 3-6 months after completing the anti-HCV therapy. Of the conventional risks for cardiovascular events, including sex, age, BMI, history of diabetes mellitus, hypertension, dyslipidemia, and smoking 16 , smoking [odds ratio (OR): 0.617, 95% confidence interval (CI) of OR:1.436-26.53] and dyslipidemia (OR: 5.25, 95% CI of OR: 1.27-21.68) were the independent risk factors associated with the advent of new cardiovascular events. There were no differences in the prevalences of these conventional risk factors between those with and without an SVR (Supplementary Table 2).
Negligible impacts from fibrosis and steatosis on hepatic PAI-1 expression in patients with CHC. The impact of hepatic fibrosis and steatosis on hepatic PAI-1 expression in CHC patients before   anti-HCV therapy were negligible, as only some biliary cells and endothelial cells ( Fig. 3A-D, black arrows), but no inflammatory cells or hepatocytes, expressed PAI-1, and the expression was very weak. Moreover, compared with the normal controls (Fig. 3E,F), the CHC patients did not exhibit differences in hepatic PAI-1 expression (0.15+ /− 0.07% vs. 0.11+ /−0.21%, p = 0.326). A hepatocellular carcinoma (HCC) sample showed that many malignant hepatocytes were strongly positive for PAI-1 expression (Fig. 3H, red arrows), while the biliary cells weakly expressed PAI (Fig. 3H, arrows). In contrast, normal hepatocytes in the HCC sample, even those close to areas with inflammation, fibrosis or steatosis, did not express PAI-I (Fig. 3G,H).

Discussion
To the best of our knowledge, this is the first prospective study to elucidate the relationships that exist among HCV infection, alterations in PAI-1 levels and cardiovascular risk factors. The most compelling results are as follows: (1) Patients with an SVR had similar PAI-1 levels but lower pre-therapy BNP levels than patients with no SVR. In the current study, all of the differences between the patients with and without an SVR in the measured pre-therapy variables, including viral load and HOMA-IR levels, have been previously clarified 2,15,17,18 , with the exception of BNP levels, which are well-established positive markers of cardiovascular events 19,20 . BNP levels were significantly lower in the patients with an SVR than in those without. This phenomenon may be related to the different viral loads rather than the different HOMA-IR levels between these groups because the negative association between BNP and HOMA-IR levels is blunted at pathological BNP levels 21 . For PAI-1, PAI-1 mRNA-binding protein may confer cAMP-mediated regulation of PAI-1 mRNA stability, allowing binding of PAI-1 mRNA to the HCV internal ribosomal entry site (IRES) 22 . This discovery suggests a potential link between Variants PAI-1 (ng/ml) HCV IRES biology and PAI-1 in associated cellular functions. Moreover, PAI-1 levels were reported to be predictive of interferon-based therapeutic responses in a cohort of 190 G1 CHC patients 23 . However, in the current study (669 patients with CHC, mainly G1 and G2), no association was noted between pre-therapy PAI-1 and HCV RNA levels, and no predictive role of PAI-1 level in anti-HCV therapeutic response was found. Although inflammation/fibrosis and hepatic steatosis are believed to increase PAI-1 levels 4-7 , our immunohistochemical (IHC) studies showed that the influence of hepatic fibrosis and steatosis on hepatic PAI-1 expression of CHC patients was negligible. Moreover, APRI, a positive marker of hepatic fibrosis in CHC 24 , was not associated with serum PAI-1 levels in pre-or post-therapy multivariate analyses. Additionally, no differences in hepatic PAI-1 expression were noted between the normal controls and the CHC patients before anti-HCV therapy. Collectively, the impact of HCV infection on PAI-1 level, if any, is likely indirect and of a non-hepatic origin. Of note, among the 5 PAI-1-associated SNPs tested, only PAI-1-rs1799889 was associated with PAI-1 levels. The CHC patients who were homozygous for the 4 G allele (4 G/4 G) of PAI-1-rs1799889 presented the highest PAI-1 levels. This is consistent with the finding that the protein encoded by the 4 G allele has higher activity The included patients with chronic hepatitis C were stratified according to (A) sex (male: 1; female: 0), (B) PAI-1-rs-1799889 genotype (4 G/4 G genotype: 1; non-4G/4 G genotype: 0) and (C) IFNL3-rs-12979860 genotype (CC genotype: 1; non-CC genotype: 0). The blood-drawing time points were as follows: 1, 2 weeks before therapy; 2, after 4 weeks of therapy; 3, after 12 weeks of therapy; 4, after 24 weeks of therapy; 5, after 36 weeks of therapy; 6, after 48 weeks of therapy; 7, after 60 weeks of therapy; and 8, after 72 weeks of therapy. than that encoded by the 5 G allele 25 . Because PAI-1 is considered a strong acute-phase reactant and the PAI-1-rs1799889 genotype has been described as a response polymorphism 14 , the impact of the PAI-1-rs1799889 4 G/4 G genotype was most evident in the pre-therapy PAI-1 levels and during the initial stages of the longitudinal trend for PAI-1. This impact vanished after the development of an SVR, when virus-associated inflammation disappeared. In contrast, platelet counts were consistently positively associated with PAI-1 levels, regardless of HCV infection status. These findings highlight the importance of blood platelets as a main source of serum PAI-1 25 . The impact of IFNL3-rs12979860 (which is highly associated with SVR) 1,15 on PAI-1 levels was evident during the later course of therapy, demonstrating that SVR did affect the longitudinal trend of PAI-1 levels. PAI-1 levels    increased after SVR in a platelet-, sex-, age-and HOMA-IR-dependent manner, but all the dependent variables remained unchanged. These findings indicate that sources other than platelets, such as visceral adipose tissue 26 , or transcriptional up-regulation subsequent to increased lipid profile 2,6,15 might account for the increased PAI-1 levels observed after SVR. Interestingly, BMI decreased but lipid profiles increased (Table 4) after SVR, showing a reciprocal trend in regulating PAI-1 levels [4][5][6][7] . Increased lipid profiles that counterbalanced the down-regulated effect of decreased BMI on PAI-1 levels might therefore be the main driving force for increased PAI-1 levels in CHC patients after SVR. Although the SVR-associated lipid alterations were different between G1 and G2 HCV infection 15 , no genotype-specific SVR-associated PAI-1 alteration was noted. As PAI-1 is necessary for modulating hepatocyte growth and differentiation 9 , the increased PAI-1 levels after SVR might also be associated with hepatocyte regeneration in response to viral clearance. However, increased PAI-1 levels have been strongly linked to increased cardiovascular risk 27,28 . One unit of PAI-1 activity is defined as the quantity of PAI-1 that can neutralize one unit of single-chain tPA in 10 minutes 29 . After SVR, PAI-1 activity increased while levels of tPA-PAI-1 complex and tPA decreased. All these alterations led to a hypercoagulable state. Moreover, BNP and homocysteine levels, which are both highly associated with cardiovascular events 19,20 , became elevated after the development of an SVR. In particular, pre-therapy levels of BNP were lower in the patients with an SVR than in those without and were supposed to decrease rather than increase after viral clearance. BNP has a role in cardiac protection arising from its capacity to regulate water and electrolyte levels 19,20 . Furthermore, BNP has antifibrotic and cytoprotective properties that inhibit adverse myocardial remodeling by suppressing PAI-1 mRNA and protein expression 19,20 . However, this inhibition was not effective in the patients with an SVR, as PAI-1 expression increased after the development of an SVR. Within 24 weeks post-therapy, ten (2.2%) of the patients with an SVR experienced a new-advent cardiovascular event. Most of the patients were old males who smoked or females with risk factors for cardiovascular events. In contrast, none of the patients with no SVR experienced any cardiovascular events. The higher incidence of cardiovascular events in the SVR vs. non-SVR group was confirmed by analyses that adjusted for the associated conventional risks. Paradoxically, several cohort studies have shown that anti-HCV therapy is associated with improved cardiovascular outcomes in CHC patients 2 , and a recent comprehensive meta-analysis also showed that these patients are at increased risk for cardiovascular event-related morbidity and mortality 30 . Although viral clearance eliminates hepatic injury, we propose that host metabolic homeostasis is perturbed, at least within 24 weeks post-therapy with interferon-based therapy. Therefore, special caution may be needed for populations that are vulnerable to cardiovascular events 31 . Conversely, for subjects who are free from conventional cardiovascular risks, the benefit of viral clearance would compensate for the temporary perturbation of host homeostasis in eliminating cardiovascular risk. Finally, because PAI-1 levels are predictive of many diseases in addition to cardiovascular events 7,8 , the data from the current study provide a reference for the evaluation of other co-morbidities. Given that adipose tissue is the major source for adipocytokines including PAI-1 4 , one of the major limitations of the current study was the lack of a pathological study of adipose tissue. Further studies of PAI-1 levels in CHC patients should include comprehensive surveys of adipose tissue pathology with transcriptional assays considering the influence of lipid profiles to confirm our findings and elucidate the associated molecular mechanisms. Furthermore, since only 2.2% of the patients with an SVR developed a new cardiovascular event, the difference in cardiovascular events between the patients with and without an SVR, although statistically significant, requires future studies with larger cohorts that can provide higher confirmatory power than the current study.