Role of PAI-1 in hepatic steatosis and dyslipidemia

Plasminogen activator inhibitor 1 (PAI-1) is a functional biomarker of the metabolic syndrome. Previous studies have demonstrated that PAI-1 is a mechanistic contributor to several elements of the syndrome, including obesity, hypertension and insulin resistance. Here we show that PAI-1 is also a critical regulator of hepatic lipid metabolism. RNA sequencing revealed that PAI-1 directly regulates the transcriptional expression of numerous genes involved in mammalian lipid homeostasis, including PCSK9 and FGF21. Pharmacologic or genetic reductions in plasma PAI-1 activity ameliorates hyperlipidemia in vivo. These experimental findings are complemented with the observation that genetic deficiency of PAI-1 is associated with reduced plasma PCSK9 levels in humans. Taken together, our findings identify PAI-1 as a novel contributor to mammalian lipid metabolism and provides a fundamental mechanistic insight into the pathogenesis of one of the most pervasive medical problems worldwide.


Results
PAI-1 inhibition reduces PCSK9 expression. Selective pharmacologic inhibition of PAI-1 prevents the development of NAFLD in mouse models of diet-induced obesity 16,17 . To investigate the mechanistic role of PAI-1 in hepatic lipid metabolism, RNA sequencing (RNA-seq) was performed on hepatic mRNA from mice treated with TM5614, a novel orally active small molecule inhibitor of PAI-1, and compared their gene expression patterns with those from hepatic mRNA collected from age-matched littermate control mice consuming a standard chow (SC) diet. Cluster analysis indicated homogeneity amongst samples in each group, with numerous changes induced by blocking PAI-1 in RNA expression across the genome (Fig. 1A). Hierarchical clustering and gene ontology analysis revealed that short-term administration of a PAI-1 inhibitor generated a coordinated, robust, and significant alteration in the expression of numerous genes involved in metabolism of lipids (160 genes) and fatty acid metabolism (156 genes) in response to treatment (Fig. 1C, S1A). Differential gene expression analysis indicated that Proprotein convertase subtilisin/kexin type 9 (Pcsk9) was the most highly downregulated gene transcript in mice treated with TM5614 (86% decrease, adjusted p = 4.34 × 10 −32 , Fig. 1B), whereas Fibroblast growth factor 21 (FGF21) was the most upregulated transcript (568% increase, adjusted p = 5.66 × 10 −28 , Fig. 1B). Consistent with these transcriptional changes, treatment with TM5614 led to a 40% decrease in plasma PCSK9 levels (Fig. 1D, p < 0.05) and this was accompanied by a 25% reduction in plasma cholesterol levels (Fig. 1E, p < 0.01). In humans, circulating PCSK9 reduces low density lipoprotein receptor (LDLR) levels by targeting the receptor for degradation, thus impairing LDL cholesterol (LDL-C) clearance from circulation and increasing plasma LDL-C levels 20 . Wild-type (WT) mice have minimal levels of LDL-C, as their plasma cholesterol is mostly carried by high density lipoproteins (HDL) 21 . In contrast to humans, both LDL and HDL particles in mice can be cleared by the LDLR 22 . As expected, the decrease in total cholesterol caused by PAI-1 inhibition was driven by a reduction in HDL cholesterol (HDL-C) level ( Figure S2A,B). Consistent with these findings in vivo, incubation with TM5614 in HepG2 cells resulted in a reduction in PCSK9 protein levels (48% intracellularly ( p < 0.05), 69% extracellularly ( p < 0.01) and a concomitant increase in LDLR protein levels (65%, p < 0.001), suggesting that PAI-1 regulation of PCSK9 levels has an impact on cholesterol clearance mediated by increased LDLR levels ( Figure S3). To exclude the influence of an off-target effect of the drug on the observed alterations in hepatic gene expression, a similar set of studies was performed using heterozygous PAI-1 deficient mice (Pai-1 +/− ) and wild-type (WT) littermate controls. RNA-seq was performed on hepatic mRNA from 20 week-old mice maintained on the SC diet. Cluster analysis indicated homogeneity amongst samples in each group, with multiple changes in RNA expression across the genome induced by lower PAI-1 levels ( Fig. 2A). Similar to the effects seen with the administration of the PAI-1 inhibitor, differential gene expression analysis identified Pcsk9 as one the most highly downregulated gene transcript identified in Pai-1 +/− mice (Fig. 2B). Furthermore, Pai-1 heterozygosity also resulted in a coordinated and robust alteration in the expression of genes involved in lipid metabolism (Figure S1B), and these results were corroborated by a gene ontology analysis (Fig. 2C). Plasma PCKS9 levels trended lower in Pai-1 +/− mice but did not reach statistical significance (Fig. 2D). Pharmacological inhibition of PAI-1 shared 771 differentially expressed genes with heterozygous Pai-1 deficiency ( Figure S1C). The observed 16% reduction in decrease in total cholesterol levels ( Fig. 2E; p = 0.059) suggest that the observed transcriptional changes generate the expected changes in lipid homeostasis.

PAI-1 inhibition in hypercholesterolemia lowers PCSK9 levels and corrects dyslipidemia in vivo.
To investigate further the effects of PAI-1 inhibition on cholesterol metabolism, WT C57BL/6 J mice were fed a high-fat, high-sugar (HFHS) diet to induce obesity and hyperlipidemia. Mice were placed on HFHS diet for 16 weeks, and then treatment with TM5614 was initiated in half of the experimental animal cohort for an additional 10 weeks, while the other half of the cohort remained, untreated, on the HFHS diet. The HFHS diet increased total cholesterol, LDL-C, and HDL-C, as expected. There were no differences between baseline or end-of-study weight between groups ( Figure S4A,B). Consistent with prior studies, inhibition with TM5614 was associated with preserved glucose tolerance and decreased hepatic steatosis ( Figure S4C-E). The HFHS diet produced a 16-fold increase in plasma levels of active PAI-1 ( p < 0.001), whereas TM5614 administration reduced plasma levels of active PAI-1 by 56% ( p < 0.05) (Fig. 3A). Inhibition of PAI-1 led to a lowering of plasma total cholesterol by 37% ( p < 0.05), LDL-C by 48% ( p = 0.06), and HDL-C by 33% ( p < 0.05) (Fig. 3B,C). Hepatic Pcsk9 mRNA expression was reduced by 81% ( p < 0.001) in mice fed a HFHS supplemented with TM5614. Drug treatment also reduced the expression of other genes involved in lipid metabolism (Fig. 3D), including Srebp1a (70%, p < 0.01), Srebp1c (81%, p < 0.001), Srebp2 (42%, p < 0.001), and HMG-CoA reductase (55%, p < 0.0001). The reduction in Pcsk9 mRNA expression was associated with a 55% reduction in the corresponding plasma levels of PCSK9 protein ( p < 0.01, Fig. 3E), while corresponding measures of Ldlr mRNA and protein were essentially unchanged (Figs. 3D,F).
We designed and performed similar studies using heterozygous null mice with partial PAI-1 deficiency (Pai-1 +/− ) fed a HFHS diet for 16 weeks. As expected, plasma PAI-1 levels in these mice are approximately 50% ( p < 0.05) of those of littermate controls (Fig. 4A). In these studies, partial PAI-1 deficiency was associated with a 20% lower plasma cholesterol level ( p < 0.05), driven by a decrease in LDL-C and HDL-C (Fig. 4B,C). Hepatic Pcsk9 transcript levels were reduced by 50% ( p < 0.05) (Fig. 4D). Plasma levels of PCSK9 were also reduced by 61% ( p < 0.001) ( Figure S5). A comparison of PCSK9 plasma levels in all of the experimental groups described in this paper is shown in Figure S5. Taken together, these data show that PAI-1 directly influences the expression of PCSK9.
To validate the relationship between plasma PAI-1 and PCSK9 levels in humans, we studied plasma samples from a second human cohort of adults (mean age 69.5, 66% female). This cohort included control patients and patients with heart failure and preserved ejection fraction (HFpEF). There was again a positive correlation between PAI-1 and PCSK9 levels in these individuals, similar to that seen in mice and in the Amish cohort ( Fig. 5E, p < 0.0001; R = 0.351). Furthermore, we analyzed PAI-1 levels in hypercholesterolemic patients treated with therapeutic monoclonal antibodies that block PCSK9 binding to the LDLR (evolocumab and alirocumab). In aggregate, treatment with PCSK9 inhibitors was followed by a substantial increase in total plasma PCSK9 levels ( Figure S6, p < 0.001) coupled with a significant reduction in plasma PAI-1 levels (Fig. 5F, p = 0.0029). These data suggest that modulation of PCSK9 activity has a novel and reciprocal effect on plasma PAI-1 levels.

Discussion
In this study, we demonstrate for the first time that PAI-1 is directly involved in lipid metabolism. Previous clinical studies have reported that plasma PAI-1 levels correlate with levels of VLDL 23 . This physiological relationship was explained by direct effects of VLDL in inducing PAI-1 expression. We now demonstrate the presence of a converse operational paradigm, that pharmacologic or genetic reductions in PAI-1 alone are sufficient to produce alterations in lipid homeostasis. The administration of a specific PAI-1 inhibitor generates an unexpected and multifaceted hepatic transcriptional response in numerous genes involved in lipid metabolism. The fundamental validity of this finding is supported by our uncovering of a similar pattern of transcriptional alterations in Pai-1 +/− mice. Furthermore, in mice fed a HFHS diet, modulation of PAI-1 activity yields significant effects on the transcription of numerous critical regulators of lipid biosynthesis, including Pcsk9, Srebp1a, Srebp1c, Srebp2, and HMG-CoA reductase. These alterations directly lead to a reduction in plasma cholesterol, LDL-C, and HDL-C levels. Taken together, these data demonstrate that under normal conditions, modest alterations in plasma PAI-1 activity have a pervasive and physiologically significant impact on hepatic lipid metabolism. Furthermore, this newly identified regulatory role persists under conditions of metabolic stress brought on by feeding a HFHS diet. Plasma PAI-1 levels are elevated in patients with NAFLD, and there is accumulating evidence that PAI-1 plays a causal role in the pathogenesis of this condition 16,17 . Recently, Lu et al.reported that a DNA methylation estimator of plasma PAI-1 levels (DNAm PAI-1) is a remarkably powerful predictor of hypertension, T2DM, NAFLD, and plasma lipid levels in humans 24 . While these and other studies have identified important statistical associations, a mechanism to explain the link between PAI-1 and NAFLD has not yet been previously reported. In our attempts to address this question in an unbiased manner, we systematically investigated the effects of TM5614 on hepatic gene expression, and in the process uncovered a novel function of PAI-1 in the development of dyslipidemia. To our knowledge, this is the first study that mechanistically associates PAI-1 with hyperlipidemia, and its remarkable juxtaposition in metabolic control regulating PCSK9.
The RNA-seq dataset presented here revealed a role for PAI-1 in the expression of numerous genes responsible for lipid metabolism. These exploratory findings were confirmed by selective qRT-PCR analysis indicating that PAI-1 inhibition reduces hepatic Pcsk9 expression, as well as numerous hepatocellular enzymatic components that regulate de novo cholesterol synthesis and fatty acid metabolism. PCSK9 is broadly recognized as an important regulator in lipid metabolism, largely through its effects in promoting the degradation of LDL receptors, thus impairing receptor-mediated LDL-C clearance 25 . In clinical practice, PCSK9 is a validated target in lipidlowering therapy. The administration of monoclonal antibodies blocking the binding of PCSK9 to the LDLR www.nature.com/scientificreports/ yields ~60% reduction in LDL-C levels of patients suffering from hypercholesterolemia 26 . The effects of PAI-1 inhibition on the hepatic expression of Pcsk9 translated to a decrease in plasma PCSK9 concentrations in mice and a resultant significant reduction in serum cholesterol. This same dataset revealed a novel and robust effect of PAI-1 inhibition in increasing the expression of Fgf21. While PCSK9 suppression was consistently observed irrespective of diet and method of reducing PAI-1 activity (pharmacologic or genetic), the effects of TM5614 on hepatic Fgf21 expression were only observed in healthy mice on a standard chow diet. While the metabolic effects and potential benefits of FGF21 are numerous, the beneficial effects of PAI-1 inhibition on lipid metabolism do not appear to involve FGF21. In order to gain some insight into the potential utility of PAI-1 inhibition in patients with the metabolic syndrome, we studied the effects of TM5614 in a diet-induced dyslipidemia model. Oral administration of TM5614 effectively reduced hepatic Pcsk9 mRNA by over 80% with a corresponding 55% decrease in plasma PCSK9 levels. These effects were accompanied by a significant reduction in plasma HDL-C and LDL-C. The observed decrease in both HDL-C and LDL-C reported here is consistent with prior studies in mice using monoclonal antibodies that block PCSK9 activity 27 . Importantly, other genes involved in lipid synthesis and metabolism were also downregulated in this experimental model of diet-induced dyslipidemia, including Srebp1/2 and HMG-CoA reductase. These changes are likely secondary to the increased uptake of intracellular cholesterol caused by the loss of PCSK9 28 . However, a direct role for PAI-1 on these core lipid synthesis genes cannot be excluded. In contrast, while PAI-1 inhibition had a significant effect in reducing LDL-C in this model, it was not accompanied by the anticipated augmentation of LDLR protein in hepatic lysates 29 . The apparent lack of change in LDLR levels may reflect numerous competing effects present in the model, including a partial reduction in circulating PCSK9 levels, cellular heterogeneity in responses, the restoration of receptor density over time, or artifacts of experimental design, including duration of therapy and the use of whole cell extracts as opposed to purified membranes.
The experimental evidence presented here clearly indicates that PAI-1 regulates lipid metabolism in mice. There is a paucity of information on the relationship between PAI-1 and PCSK9 in the clinical literature. Importantly, we did find that genetic deficiency of PAI-1 is associated with a reduction in hepatic Pcsk9 expression in mice and in human plasma PCSK9 levels. While this relationship is significant, it probably underestimates the effects of PAI-1 on PCSK9 levels. Multiple factors likely contribute to the relatively narrow range [19 ± 14 ng/ ml (mean ± SD)] of plasma PAI-1 levels in the unaffected Amish individuals reported in this study, including lifestyle, diet and fitness, and these may have blunted the effects on plasma PCSK9. We also uncovered the presence of a relatively robust correlation between plasma levels of PAI-1 and PCSK9 in patients that do not have genetic PAI-1 deficiency. Interestingly, PAI-1 levels also decreased dramatically in hypercholesterolemic patients who were treated with monoclonal antibodies that inhibit the function but raise the levels of plasma PCSK9. This newly recognized relationship indicates that pharmacologic inhibition of PCSK9 in hypercholesterolemic individuals not only reduces LDL-C but may also provide a secondary cardiovascular benefit by concomitantly reducing plasma PAI-1. In aggregate, these findings strongly suggest the presence of a reciprocal regulatory loop between PAI-1 and PCSK9.
While the effects of PAI-1 on hepatic lipid metabolism are extensive and likely important, the molecular mechanisms that explain these novel effects are incompletely understood. The aggregate effects of PAI-1 on hepatic gene expression likely involve a complex physiological calculus that includes: (1) the direct binding of PAI-1 binding to LRP and effects of Jak/STAT signaling 30,31 ; (2) PAI-1 inhibition of the uPA-dependent activation of proHGF and subsequent alterations in the binding of active HGF to hepatic cMET receptors 32 ; and (3) augmented transcription of FGF21 and increased hepatic signaling through the heterodimeric FGFR/β-Klotho receptor [33][34][35][36] . Other mechanisms may be related to the effects of PAI-1 on the furin-dependent activation of the insulin receptor 37 .
In conclusion, we have shown that pharmacological inhibition of PAI-1 prevents hepatic steatosis and reduces serum cholesterol levels through a mechanism that involves reduced PCSK9 synthesis (Fig. 6). Given that PAI-1 levels are consistently elevated in patients with the metabolic syndrome, our finding of a direct role for PAI-1 in the regulation of lipid metabolism provides further support to the central role of PAI-1 in the cardiovascular complications of the metabolic syndrome, and in the molecular pathogenesis of the metabolic syndrome itself. We anticipate that novel therapeutic agents that either reduce the synthesis or block the function of PAI-1 will be of value in the prevention and treatment of the metabolic syndrome and its vascular consequences.

Materials and methods
Materials. TM5614 (FW = 446.43 g/mol). A 20 mM stock solution was prepared in DMSO and stored at − 20 °C. In cell culture, TM5614 was added to a final concentration of 10 µM with DMSO < 0.5%. For animal feeding experiments, dry TM5614 was mixed into the powdered mouse food which were then formed into pellets given at a dose of 20 mg/kg/day. Human cohorts. Plasma samples from members of the Bern Amish Kindred who harbor a frameshift mutation in SERPINE1 (SERPINE1 +/+ (n = 17, 7 males) and SERPINE1 +/− (n = 16, 8 males) were assessed to quantify PCSK9 plasma levels. This study was previously approved by both the Northwestern University and the Indiana Hemophilia and Thrombosis Center Institutional Review Board as previously described 19 . All methods were performed in accordance with the relevant guidelines and regulations. Informed consent was obtained from all subjects.
Data was analyzed from a cohort of patients with heart failure with preserved ejection fraction (HFpEF). Patients were enrolled in this prospective observational study from the outpatient HFpEF clinic (for HFpEF patients) and from the outpatient general cardiology clinic or cardiac catheterization laboratory (elective cases) for the control patients. Control patients were deemed eligible if they had 1 or more risk factors for HFpEF (e.g., Scientific Reports | (2021) 11:430 | https://doi.org/10.1038/s41598-020-79948-x www.nature.com/scientificreports/ hypertension, obesity, diabetes, chronic kidney disease) and only if there was no history of heart failure (any type, including those with recovered LVEF). In this analysis, 248 unique plasma proteins (including PAI-1 and PCSK9) were quantified by a multiplex immunoassay (Olink, www.olink .com) using the commercially available panels cardiovascular disease II, IIII, and inflammation. This study was approved by the Northwestern University Institutional Review Board. Plasma samples from hypercholesterolemic patients before and while on a PCSK9 inhibitory therapy (Alirocumab, 75 mg every 2 weeks, or Evolocumab, 140 mg every 2 weeks) were obtained from the OHSU Center for Preventive Cardiology Registry and Biorepository, approved by OHSU Institutional Review Board (17,329).
Animal models. Wild-type C57BL/6 J male mice (Jackson Laboratories, Bar Harbor, ME) were fed a standard chow (SC) diet and underwent 14/10-h light/dark cycling with free access to food and water. Animal protocols were approved by the Northwestern University Institutional Animal Care and Use Committee (IACUC). A first cohort of 20-week-old mice was used to evaluate the acute effect of inhibiting PAI-1 with TM5614. Half of the mice were fed SC and the other half SC supplemented with TM5614 at a dose of 20 mg/kg/day for 10 days. A second cohort of 6-week-old mice was used to evaluate the long-term effects of inhibiting PAI-1 with TM5614 in mice fed a high-fat, high-cholesterol, high-sugar diet (HFHS). The HFHS diet is composed of 40% energy as fat (milk fat, 12% saturated) with 2% cholesterol (AIN-76 Western Diet, Test Diet) and drinking water supplemented with 42 g/l of 55% fructose/45% glucose by weight. Mice were fed HFHS for 4 months to induce obesity. After this period, half of the mice were treated with TM5614 (as described for the acute exposure) for a period of 2 months. A third cohort of mice Pai-1 +/− were generated by the breeding of Pai-1 −/− and WT mice (Jackson Laboratories, Bar Harbor, ME). These mice were maintained on SC.
At the end of these protocols, blood was collected by retro-orbital bleeding in citrated tubes and centrifuged to collect plasma and mice were euthanized by isoflurane followed by cervical dislocation. Mouse livers were flushed with ice-cold saline, a portion of the liver was put into RNA LATER (Cat# 76104, Qiagen) prior to RNA extraction. The remainder of the tissue was snap frozen in liquid nitrogen. All methods were performed in accordance with the relevant guidelines and regulations. www.nature.com/scientificreports/ RNA sequencing and analysis. RNA isolation and sequencing were performed as previously described 38 .
Briefly, total RNA was extracted from murine livers using an RNA Easy mini kit  Hepatic protein quantifications. Hepatic proteins were extracted from liver tissue homogenized with lysis buffer (RIPA Buffer (Sigma-Aldrich, Cat# R0278) supplemented with 1X proteinase inhibitor (Thermo Fisher Scientific, Cat# A32963) and total protein content were performed as described 39 . mLDLR (R&D Systems, Cat# MLDLR0) protein levels were quantified from the total extracts (20 µg) using ELISA methodology following the manufacturers recommendation.
In vitro cellular studies. Human hepatocellular carcinoma HepG2 (ATCC HB-8065) cell line was routinely grown as previously described 40 at 37 °C in a 5% CO 2 humidified atmosphere. 2.5 × 10 5 cells were plated. The following day (day 2) media was changed to serum-free media. On day three cells could be treated with 10 µM TM5614 (or vehicle) for 48 h. On day five, cell culture media was collected and stored at − 20 °C until analysis. Cells were washed twice with ice-cold phosphate-buffered saline and scraped in 100 µL lysis buffer (same buffer as described for tissue protein extraction). Lysates were incubated on ice for 1 h (with mixing every 15 min) and spun at 14,000 × g at 4 °C for 20 min, supernatants were collected and quantified by a Lowry assay (DC Protein Assay, Cat# Bio-Rad, 5000111) then stored at − 20 °C until used. 20 µL of cell culture medium or 40 µg of total extracts were resolved in SDS-PAGE (4-12% or 12% Bis-Tris precast acrylamide gels, Thermo Fischer Scientific) and later transferred to nitrocellulose membranes (GE Healthcare, Cat# 10600003) and blot for human PCSK9 (using a rabbit anti-PCSK9 (1:1000 dilution, MBL International, Cat# CY-P1037) as a primary antibody and a goat anti-rabbit (1:15,000 dilution, LI-COR Biosciences, Cat# 925-32211) as the secondary antibody), LDLR (using a goat anti-LDLR (1:1000 dilution, R&D, Cat# AF2148) as the primary antibody and a donkey anti-goat (1:15,000 dilution, LI-COR Biosciences, Cat# 925-32214) as the secondary antibody), actin (using a mouse antiactin (1:2000 dilution, Sigma-Aldrich, Cat# A5441) as the primary antibody and a goat anti-mouse (1:15,000 dilution, LI-COR Biosciences, Cat# 926-68070) as the secondary antibody).
Glucose tolerance testing. Intraperitoneal glucose tolerance tests (IPGTT) were performed on mice fasted for 16 h. Glucose was injected by IP at 2 g/kg of body weight. Blood glucose was obtained from tail veins at multiple time points (0-120 min) and measured using a NovaMax Plus glucometer (Novacares).
Hepatic triglyceride content. Hepatic triglyceride levels were measured using an Infinity spectrophotometric assay (Thermo Electron Corporation, Cat# TR22421) following lipid extraction using the Folch method 41 .