Levosimendan pretreatment improves survival of septic rats after partial hepatectomy and suppresses iNOS induction in cytokine-stimulated hepatocytes

We evaluated the survival effects and biochemical profiles of levosimendan in septic rats after partial hepatectomy and investigated its effects in cultured hepatocytes. Thirty-two rats underwent 70% hepatectomy and were randomised equally into four groups, followed by lipopolysaccharide (LPS) injection (250 µg/kg, i.v.) after 48 h. Levosimendan was given (i.p.) 1 h before LPS injection [group (A) levosimendan 2 mg/kg; (B) 1; (C) 0.5; (D) vehicle]. Survival at 7 days was increased significantly in group A compared with that in group D [A: 63%; B: 38%; C: 13%; D: 0%]. In serum, levosimendan decreased the level of tumour necrosis factor-α, interleukin (IL)-1β, IL-6 and nitric oxide (NO). In remnant livers, levosimendan inhibited inducible nitric oxide synthase (iNOS) gene expression. In primary cultured rat hepatocytes stimulated by IL-1β, levosimendan suppressed NO production by inhibiting iNOS promoter activity and stability of its mRNA.

Effect of levosimendan on histopathological changes. Histopathology revealed the change in regeneration of rat livers 48 h after 70% hepatectomy: ballooning hepatocytes and spreading of lipid droplets (Fig. 5a). After 4 h of LPS injection, focal necrotic hepatocytes were prominent at the centrilobular zone and midzone in both groups of PH/LPS with vehicle and levosimendan (Fig. 5b,c). Few myeloperoxidase (MPO)-positive cells were infiltrated in livers 48 h after 70% hepatectomy: (0.3 cells/mm 2 , Fig. 5d). Severe infiltration of MPO-positive cells was recognized in specimens of rat livers after 4 h of LPS injection with vehicle (Fig. 5e).   Levosimendan pretreatment did not inhibit the infiltration of MPO-positive cells significantly in remnant livers (P = 0.7) (Fig. 5f,g). Apoptotic bodies were evaluated by terminal deoxynucleotidyl transferase-mediated dUTP-digoxigenin nick-end labelling (TUNEL) staining, and few positive nuclei were detected in rats 48 h after 70% hepatectomy (4 per 1,000 nuclei, Fig. 5h). The difference in the percentage of TUNEL-positive cells in all nuclei was not significant in the absence (Fig. 5i) or presence of levosimendan pretreatment (P = 0.9) (Fig. 5j,k).
Effect of levosimendan on induction of expression of NO, iNOS protein and iNOS mRNA in IL-1β-stimulated cultured hepatocytes. In the culture medium, simultaneous administration of levosimendan (20 µM) with IL-1β (1 nM) reduced the level of nitrite (NO metabolite) time-dependently, which was increased by single administration of IL-1β (Fig. 6a). Levosimendan reduced the production of NO and iNOS protein dose-dependently, and decreased production to a near-basal level at a concentration of 20 µM (Fig. 6b, upper and middle). The level of lactate dehydrogenase (LDH) in the culture medium was not increased by ≤20 µM of levosimendan (Fig. 6b, lower). A dose of 20 µM was used in subsequent in vitro experiments. Reverse transcription-polymerase chain reaction (RT-PCR) revealed that levosimendan reduced expression of iNOS mRNA in each hour (Fig. 6c).

Effect of levosimendan on the activity of iNOS promoters, iNOS antisense transcription and
intranuclear level of nf-κB in IL-1β-stimulated hepatocytes. The scheme of the constructs containing firefly luciferase controlled by the iNOS promoter (pRiNOS-Luc-SVpA and pRiNOS-Luc-3′UTR) is shown in Fig. 7a. Levosimendan inhibited relative luciferase activities on both constructs, which were increased by IL-1β single administration (Fig. 7b). RT-PCR revealed that levosimendan inhibited expression of the iNOS antisense transcript at 3 h and 6 h (Fig. 7c). EMSAs with nuclear extracts did not show an inhibitory effect of levosimendan on NF-κB activation (Fig. 7d). Further, we could not detect significant influences of levosimendan on NF-κB nuclear translocation, IκB degradation and phosphorylation of NF-κB p65 (Ser 536 ) (Supplementary information file; M, K, and L).

Discussion
Two experimental models of sepsis with acute liver injury can be employed: (i) simultaneous administration of D-galactosamine/LPS 29-31 ; (ii) PH/LPS. The lethal activity of endotoxins is enhanced considerably under both models, but the PH/LPS-model exhibits more severe and refractory symptoms 32 , and is closer to a specific clinical situation.
A pilot study revealed that a 50% lethal dose of LPS for this model was ≈100 µg/kg, and that >90% of rats died at a LPS dose of 250 µg/kg (i.v.) (T. O., unpublished observation). We chose the doses of levosimendan by reference to a similar study of ischemia-reperfusion injury in rat mesenteries 33 . In our preliminary study, administration of levosimendan (2 mg/kg) 1 h after LPS injection showed no effect on survival (data not shown). In contrast, pretreatment of levosimendan increased the survival of PH/LPS-model rats in a dose-dependent fashion, though a significant difference was only found between group A (doses of 2 mg/kg) and group D (vehicle) by post hoc analysis. Levosimendan pretreatment prevented an increase in expression of proinflammatory cytokines in serum and their mRNAs in remnant livers. Expression of iNOS in remnant livers and NO in serum (which are proinflammatory mediators) was also inhibited by levosimendan pretreatment. Those effects would probably involve inhibition of NF-κB activation, because NF-κB has an important role as a transcriptional factor of iNOS gene 28 . However, levosimendan did not inhibit NF-κB activation significantly shown in EMSA experiments in remnant livers. We should mention of the limited number of experimental animals we used and there probably existed the influence of other transcriptional factors such as hypoxia-inducible factor-1α 34 or nuclear respiratory factor 2 35 . According to a reported study of septic mice, Wang et al. concluded that levosimendan did not inhibit the LPS-induced activation of NF-κB significantly, which is a similar result to our study 8 . Levosimendan demonstrated a hepatoprotective effect in that levels of transaminases in serum decreased significantly in the levosimendan group 4 h after LPS injection. However, histopathology revealed that levosimendan did not inhibit both the infiltration of MPO-positive cells (i.e., necrotic change) and TUNEL-positive cells (i.e., apoptotic change). The results of histopathology will cause controversy whether a hepatoprotective effect of levosimendan determined the survival benefit in our study. We assume that D-galactosamine/LPS-model would be essential for examining a heaptoprotective effect of levosimendan against LPS-induced acute liver injury 36 , but this model would not surely represent for septic shock 37 . As a limitation, we could not adopt a blinded maneuver of each group for a practical reason when we injected LPS and/or levosimendan. However, two researchers (T. S. and T. O.) assured the quality of experiments of PH/LPS.
In IL-1β-stimulated primary cultured hepatocytes, levosimendan suppressed NO production in a time-and dose-dependent fashion through inhibition of iNOS gene expression. We set the concentration of levosimendan www.nature.com/scientificreports www.nature.com/scientificreports/ at 20 μM in the experiments, because the levels of LDH in culture medium were slightly elevated at the concentration of 100 μM of levosimendan (data not shown), which implied cytotoxicity caused by the overdose of levosimendan, but levosimendan had no such effects at 1-20 μM. The experiments with iNOS promoter constructs demonstrated that levosimendan inhibited iNOS expression during the synthesis and stabilisation of mRNA. iNOS promoter activity measured with the constructs represented the intensity of NF-κB-dependent transcription because both constructs have two NF-κB binding sites (κB) in each promoter area. However, EMSAs revealed that the binding activity of nuclear extracts to the NF-κB consensus oligonucleotide was not inhibited by levosimendan. We conducted the additional experiments to investigate the NF-κB nuclear translocation, IκB degradation and phosphorylation of NF-κB p65 (Ser 536 ), which are the important signalling steps to stimulate NF-κB activation. However, we could not detect significant influences of levosimendan on these steps (Supplementary file). From the results above, we concluded that levosimendan did not inhibit the activating steps of NF-κB in cultured hepatocytes. This result suggests that levosimendan might affect the synthesis of iNOS mRNA through signalling pathways and transcription factors other than NF-κB. We found that the iNOS antisense-transcript had a key role in stabilising iNOS mRNA by interacting with the 3′-ultratranslated region (UTR) and adenylate-uridylate-rich sequence elements-binding proteins 37 . Levosimendan demonstrated an inhibitory effect on expression of iNOS antisense transcripts. An anti-inflammatory profile of levosimendan was also shown in hepatocytes because of inhibition of the mRNA expression of TNF-α, CINC-1 and IL-1RI. Note that our in vitro study was not a complete reproduction of PH/LPS-model in two points that we did not use the direct cultured hepatocytes from all groups in PH/LPS-model, and we used a single cytokine (IL-1β) to stimulate the hepatocytes. The results from our in vitro study should be considered as reference to understand the anti-inflammatory mechanism of levosimendan.
Some in vitro studies have shown that levosimendan can down-regulate iNOS induction and NO production in response to inflammatory stimuli in macrophages 12 , cardiac fibroblasts 38 and hepatocytes 39 . Differences in the signalling events leading to activation of iNOS transcription between cell types might exist. Sareila et al. reported that levosimendan did not affect the activation, nuclear translocation or DNA binding of NF-κB in J774 macrophages, but inhibited NF-κB-dependent transcription in L929 fibroblasts 12   www.nature.com/scientificreports www.nature.com/scientificreports/ As an in vivo model of sepsis, CLP-model 7,8 has previously been used to show a survival benefit of levosimendan. Authors selected continuous infusion via a catheter in the jugular vein 7 or an intraperitoneal osmotic pump 8 of levosimendan, whereas we used intraperitoneal bolus administration. One may argue that intraperitoneal bolus administration of levosimendan does not represent the clinical situation accurately. However, the sepsis model caused by LPS injection does not fully represent human sepsis because LPS causes a much earlier peak of expression of pro-inflammatory cytokines compared with that seen in human sepsis. A survival curve of PH/ LPS model is more precipitous that the majority of positive control rats died at 6 h after LPS injection compared with CLP-model that two-thirds of controlled rats survived at 9 h after operation 7 . Levosimendan has a half-life of ≈1 h but its active metabolite, OR-1896, has a half-life of 80 h 40 , which could cover the duration of effect of a LPS bolus administration. Continuous infusion of levosimendan in the PH/LPS-model may merit further study.
As in vivo model of liver injury, Grossini et al. reported that levosimendan protected against ischemiareperfusion injury through mechanisms related to NO production and mitochondrial ATP-dependent potassium-channel function 11 . Taken together, levosimendan would have a beneficial effect in liver surgery/transplantation. The results of our study lead us to recommend levosimendan pretreatment for sepsis management after acute liver injury.

Animals.
All animal experiments were undertaken in accordance with the Guidelines for the care and use of laboratory animals (National Institutes of Health, Bethesda, MD, USA). The study protocol was approved by the Animal Care Committee of Kansai Medical University (Permission numbers: 17-023(01) and 18-027(01)).
Rats (specific pathogen-free) were purchased from Charles River Laboratories Japan (Yokohama, Japan) and maintained in a room at 22 °C under a 12-h light-dark cycle with a diet of γ-irradiated CRF-1 (Oriental Bioservice, Kyoto, Japan) and water ad libitum.
Drugs. Levosimendan was purchased from Wako Pure Chemical Industries (Osaka, Japan). Levosimendan was resolved in dimethyl sulfoxide (DMSO) and stored at −80 °C. For PH/LPS experiments, resolved levosimendan was diluted by 1 ml of normal saline for each rat so that the DMSO concentration was 2%. Isoflurane, pentobarbital sodium, collagenase, Transaminase CII-test kit, LDH-Cytotoxicity Assay kit and PicaGene Luminescence kit were from Wako Pure Chemical Industries. LPS (Escherichia coli O111:B4) and mouse anti-β-tubulin were from Sigma-Aldrich Japan (Tokyo, Japan). Creation of the PH/LPS model. The procedure for 70% partial hepatectomy is based on experiments described elsewhere 41 . Briefly, male Sprague-Dawley rats (8 weeks; 310 ± 10 g) were anaesthetised using pentobarbital sodium (40 mg/kg, i.p.) and isoflurane (0-2%). A laparotomy was done with a midline incision (≈3 cm). The left lateral and left median lobe of the liver were removed after ligation, followed by wound closure. Operated rats were randomised immediately and equally into four groups: (A) levosimendan 2 mg/kg; (B) levosimendan 1 mg/kg; (C) levosimendan 0.5 mg/kg; (D) vehicle (normal saline). Forty-eight hours after surgery, 250 µg/kg body weight of LPS in saline was injected into the penile vein. Levosimendan (i.p.) was given 1 h before LPS injection. Survival was evaluated during 7 days after LPS injection, and then rats were killed by isoflurane. As an exploratory experiment, samples of blood and remnant liver were taken from Sprague-Dawley rats 0 h, 1 h and 4 h after LPS injection with or without levosimendan pretreatment (n = 3-5 in each group). A scheme of the experimental protocol is shown in Fig. 1. Isolation and culture of primary hepatocytes. The isolation and culture of rat hepatocytes is based on experiments described elsewhere 42,43 . Hepatocytes were isolated from livers of Wister rats (200-220 g) by collagenase perfusion via the portal vein, followed by centrifugation (50 × g, 70 sec, 4 °C; four times). Isolated hepatocytes were suspended at 6 × 10 5 cells/mL in Williams' E (WE) culture medium, supplemented with 10% newborn calf serum, Hepes (5 mM), penicillin (100 U/mL), streptomycin (100 μg/mL), fungisone (0.25 μg/mL), aprotinin (0.1 μg/mL), dexamethasone (10 nM) and insulin (10 nM). The cells were seeded into 35-or 100-mm plastic dishes (2 or 10 mL/dish; Falcon Plastic, Oxnard, CA, USA) and cultured at 37°C in a CO 2 incubator under a humidified atmosphere of 5% CO 2 in air for 2 h. The medium (1.5 mL/35-mm dish) was replaced with fresh serum-free and hormone-containing WE medium (first medium change), then with fresh serum-and hormone-free WE medium at 5 h (second medium change), and the cells were cultured overnight. As cells were cultured two days or more before use in experiments, fresh serum-free and hormone-containing WE medium was used in the second medium change, with this medium subsequently changed every day. Then, cells were treated with recombinant human IL-1β (1 nM) in the presence or absence of levosimendan.
Biochemical analyses. Serum levels of TNF-α, IL-1β and IL-6 were measured using commercial ELISA kits. The sum of nitrite and nitrate (stable metabolites of NO) in the serum, or nitrite in the culture medium, was measured using the Griess reagent method 44 . Serum levels of AST and ALT were determined using commercial kits. LDH activity in the culture medium was measured using a commercial kit according to manufacturer instructions.
Western blotting. Protein extracts of liver sections and hepatocytes were prepared for western blotting, as described previously 29 . They were subjected to a 7.5% gel, and electroblotted. Immunostaining was done using primary antibodies against iNOS and β-tubulin (internal control), followed by visualisation with ECL Western Blotting Detection Reagents for iNOS and Luminate Forte Western HRP for β-tubulin. The bands corresponding to each protein were quantified by densitometry using ImageJ (San Diego, CA, USA) 45 .
Real-time PCR was done using SYBR Green and primers for each gene. Primer sequences were synthesised by Eurofins Genomics (Tokyo, Japan) ( Table 1). The conditions of thermal cycling using Rotor-Gene Q (Qiagen, Stanford, VA, USA) were 95 °C for 5 min followed by 40 cycles of 95 °C for 5 s and 60 °C for 10 s. Collection and analyses of data were done using the software included with the system. mRNA levels of each gene were measured as CT threshold levels and normalised to those of eukaryotic elongation factor-1α.