Hepatic PPARα is critical in the metabolic adaptation to sepsis

Abstract

Background and aims

Although the role of inflammation to combat infection is known, the contribution of metabolic changes in response to sepsis is poorly understood. Sepsis induces the release of lipid mediators, many of which activate nuclear receptors such as the peroxisome proliferator-activated receptor (PPAR)α, which controls both lipid metabolism and inflammation. However, the role of hepatic PPARα in the response to sepsis is unknown.

Methods

Sepsis was induced by intraperitoneal injection of Escherichia coli in different models of cell-specific Pparα -deficiency and their controls. The systemic and hepatic metabolic response was analysed using biochemical, transcriptomic and functional assays. PPARα expression was analysed in livers from elective surgery and critically ill patients and correlated with hepatic gene expression and blood parameters

Results

Both whole body and non-hematopoietic Pparα -deficiency in mice decreased survival upon bacterial infection. Livers of septic Pparα -deficient mice displayed an impaired metabolic shift from glucose to lipid utilization resulting in more severe hypoglycemia, impaired induction of hyperketonemia and increased steatosis due to lower expression of genes involved in fatty acid catabolism and ketogenesis. Hepatocyte-specific deletion of PPARα impaired the metabolic response to sepsis and was sufficient to decrease survival upon bacterial infection. Hepatic PPARA expression was lower in critically ill patients and correlated positively with expression of lipid metabolism genes, but not with systemic inflammatory markers.

Conclusion

Metabolic control by PPARα in hepatocytes plays a key role in the host defense to infection.

Introduction

Sepsis, the systemic inflammatory response to poorly controlled infection, causes significant morbidity/mortality [1]. Sepsis is often complicated by multiple organ failure, requiring intensive care. Recently, mortality in sepsis has decreased largely due to improved supportive strategies for critically ill patients, such as mechanical ventilation, renal replacement therapy and antibiotics. While current therapeutic strategies targeting the inflammatory response have been disappointing [1,2], metabolic interventions, such as intensive insulin therapy [3] and controlled caloric deficit through delayed administration of parenteral nutrition [4], have shown some promise, suggesting that appropriate adaptation of energy metabolism contributes to proper defense against pathogens [5].

The early pro-inflammatory response to infection requires glycolysis and non-insulin-mediated glucose uptake to rapidly meet the high energy demand of innate immune cells [6]. In this phase, hepatic gluconeogenesis increases to maintain plasma glucose concentrations [7]. As sepsis sets in, plasma free fatty acid (FFA) and glycerol levels rise due to enhanced adipose tissue (AT) lipolysis [7]. In response, organs such as the liver, muscle and heart, shift from glucose to FA utilization [8] and enhance mitochondrial activity [9].

PPARα is a nuclear receptor activated by fatty acids and derivatives regulating both metabolism and inflammation [10]. PPARα is highly expressed in metabolic tissues, such as liver, heart, kidney and muscle, the vasculature (endothelial cells, smooth muscle cells) as well as in the immune system (monocytes/macrophages, neutrophils, etc.) [11]. During fed-to-fasted transition, hepatic PPARα expression increases [12] and is activated by the influx of AT-released FFA, orchestrating a shift from glucose to FA utilization driving ketone body and glucose production by the liver [12,13].

PPARα also exerts anti-inflammatory activities by inhibiting NFκB and AP1 signaling [10]. Consequently, Pparα -deficient mice display a prolonged inflammatory response upon sterile inflammation [14]. Conversely, upon polymicrobial infection, Pparα -deficient mice display decreased survival associated with a reduced proinflammatory immune response [15], through mechanisms involving nonhematopoietic PPARα [16]. Although cardiac PPARα contributes to sepsis survival by increasing cardiac performance and FA oxidation [17], the specific contribution of PPARα in the liver has not yet been addressed.

Materials and methods

Mice

Whole body Pparα knockout (KO) and wild type (WT) littermate C57BL/6J mice (gift of F.Gonzalez [18]) were bred at the Institut Pasteur de Lille (IPL) transgenic rodent facility (SPF status). Hepatocyte-specific Pparα -deficient (Ppara fl/fl, Albumin-Cre⁺, Pparα ^hepKO) and corresponding littermate controls (Ppara fl/fl, Albumin-Cre^-) mice were generated and bred at INRA’s rodent facility (conventional health status, Toxalim, Toulouse, France), as described [19]. In survival experiments comparing Pparα ^hepKO to whole body Pparα KO mice, all mice were bred at the INRA transgenic rodent facility. Infection experiments were conducted on female (8-12 weeks) mice fed a standard rodent diet (Safe 04 U8220G10R) at IPL facility. All animals were housed under temperature-controlled (at 22-24°C), 12-hour light/dark cycle conditions. Experimental procedures were approved by the Nord Pas-de-Calais Ethics committee (CEEA 75, APAFIS#7738-2015121713177853 v9).

Bacterial cultures and infection

DH5α E.coli were grown in LB Broth at 37°C to an OD₆₀₀ of 0.6, equivalent to 4-5x10⁸ CFU/mL, collected by centrifugation, washed once with sterile PBS, and resuspended in cold PBS at 4-7x10⁸ CFU/mL. Concentration and viability were confirmed by plaque assay colony counting. Mice (10-15 mice/group) were injected intraperitoneally (i.p.) with 4-7x10⁸ CFU/mouse in 1 mL PBS and survival rates were monitored every 6hrs for a week. For biochemical characterization, mice were killed by cervical dislocation 16hrs post infection and serum and livers collected.

Human study

Post-mortem liver biopsies were taken from ICU-patients (n=46), enrolled in a randomized controlled trial [3]. All deaths occurred after multidisciplinary decision to restrict therapy when further treatment was judged to be futile. During this trial, for postmortem tissue sampling for academic purposes, each patient or his/her legal representative consented upon admission, via a hospital-wide information and consent procedure that required active opting-out when not consenting. Opting-out remained possible until time of death. This strategy was approved by the Institutional Ethical Review Board. Liver samples were harvested within minutes after death. Control liver biopsies from 20 demographically matched patients (written informed consent obtained prior to the procedure) undergoing an elective restorative rectal resection were obtained. All protocol and consent forms were approved by the Institutional Ethical Review Board of the KU Leuven (ML1094, ML2707). Baseline and outcome variables are indicated in Supplementary Table 1. Liver biopsies were taken from liver segment IVb, snap-frozen in liquid nitrogen, and stored at -80°C until analysis (see Supplementary Material & Methods (Suppl.M&M)).

Biochemical analysis

Plasma aspartate aminotransferase (AST) and alanine aminotransferase (ALT) (Biolabo), free fatty acids (FFA) (Diasys), β-hydroxybutyrate (ketone bodies) (Thermo Fisher) were determined by colorimetic assays. Plasma Tnfa, Kc and IL-6 protein levels were measured by ELISA (R&D Systems) and MPO activity as described [20]. Cytokine levels were quantified in serum of critically ill patients collected on the day of biopsy (last day alive in the ICU) [21].

Liver transcriptomic analysis

RNA extraction and analysis are detailed in Suppl.M&M. RNA microarray analysis was performed using Mouse Gene 2.0ST arrays. Array data processing was performed using Bioconductor in the R-environment (r-project.org). Gene expression was calculated after normalizing signal using robust multichip averaging (RMA) in the oligo package [22]. Differential gene expression between groups (Pparα WT uninfected, Pparα WT infected, Pparα KO uninfected and Pparα KO infected) was assessed using limma package [23] with a threshold of 5% false discovery rate (FDR). Differentially expressed genes were clustered using the hopach package [24] with the cosine distance metric. Gene Ontology (GO) terms enrichment of selected clusters was performed using the clusterProfiler package [25,26]. For KEGG pathway analysis, data were analyzed using Partek software.

Mitochondrial respiration

Liver samples (125mg) were minced and dounce homogenized by 8-10 strokes in ice-cold MIR05 respiratory buffer (20mM HEPES, 10mM KH₂PO₄, 110mM sucrose, 20mM taurine, 60mM K-lactobionate, 0.5mM EGTA, 3mM MgCl₂·6H₂O, 1g/L BSA (fatty acid free)). Liver homogenates (50μl) were introduced into O2K oxygraph chambers (Oroboros Instruments, Innsbruck, Austria) to assess oxygen consumption in presence of pyruvate (5mM) and malate (2mM) (state 2 respiration), followed by ADP (0.5mM) (state 3 respiration). To measure β-oxidation, octanoylcarnitine (25μM) and malate (2mM) were added, followed by ADP (0.5mM). The respiratory control ratio (RCR) was calculated as the state 3:state 2 ratio. Finally, cytochrome c (10μM) was added to measure mitochondrial integrity.

Histological analysis

Frozen liver samples were embedded in Frozen Section Medium (NEG-50, Richard-Allan Scientific), stained with anti-Moma2 (ab33451, abcam) or anti-Ly6G (1A8, BD Pharmingen) antibodies and counterstained with hematoxylin. Ly6G and Moma2 staining areas were determined by color detection using a Nikon Eclipse Ti microscope and a color video camera coupled to the NIS Elements software (Nikon).

Hepatic triglyceride measurement

see Suppl.M&M.

Statistical analysis

Groups were compared using the Log-rank (Mantel-Cox) Test (survival test), 2-way ANOVA, 2-tailed non-paired t-tests or nonparametric Wilcoxon tests (mouse and human studies) and expressed as means ± SEM using the GraphPad Prism software. Significance of correlations between parameters was assessed by calculation of the Pearson (r) correlation coefficient using GraphPad Prism software.

Results

Whole body and non-hematopoietic Pparα-deficiency aggravate mortality upon bacterial infection

To evaluate the role of Pparα in sepsis, whole body Pparα -deficient (KO) and wild type (WT) mice were inoculated with Gram-negative Escherichia coli (E. coli). Three days after infection, mortality was 0% in Pparα WT compared to 75% in Pparα KO mice (Fig.1A). As PPARα is expressed in immune cells and exerts anti-inflammatory actions, we assessed whether restoration of PPARα in the hematopoietic compartment conferred protection to infection. Chimeric mice were generated by transplanting lethally irradiated whole body Pparα KO and Pparα WT mice with Pparα WT bone marrow (WTbm->KO and WTbm->WT, respectively, see Suppl.M&M). Surprisingly, Pparα -deficiency in non-hematopoietic cells was still associated with increased mortality in response to infection (20% Pparα WTbm->WT compared to 70% Pparα WTbm->KO mortality 3 days after infection; Suppl.Fig.1A), revealing a protective role for non-hematopoietic Pparα in response to sepsis.

Figure 1. Whole body Pparα -deficiency enhances mortality upon bacterial sepsis and impairs the metabolic and inflammatory response to bacterial infection.

Pparα WT and KO mice were injected (ip) with vehicle (PBS) (-) or E.coli (4x10⁸ living bacteria) (Inf) (+). (A). Survival was followed for 8 days after injection (n=8-10 mice/group). Plasma was collected 6 (G,H) or 16hrs (B-F, I-J) after injection and (B) free fatty acids (FFA), (C) aspartate aminotransferase (AST), (D) alanine aminotransferase (ALT), (E) ketone bodies, (F) blood glucose, (G) Tnfa, (H) Kc/Cxcl1 concentrations and myeloperoxidase (MPO) activity (I) were measured as described in methods. (J) Bacterial levels in peritoneal fluid and blood were measured by retro-culture (n=3 mice/group). Statistical differences are indicated (Survival test: Log-rank (Mantel-Cox) Test: ** p<0.01. 2-way ANOVA: *** p<0.001, ** p<0.01 and * p <0.05 for effect of infection; §§§ p<0.001; §§ p<0.01; § p<0.05 for genotype effect; ns: non-significant)

Whole body Pparα-deficiency results in impaired metabolic and inflammatory responses to bacterial sepsis

To understand the mechanisms of increased mortality of Pparα KO mice in response to sepsis, the systemic metabolic and inflammatory response to bacterial infection was characterized. Plasma FFA levels significantly increased during sepsis to a similar extent in Pparα WT and KO mice at 16hrs post-infection (Fig.1B), when maximal changes in metabolic parameters are observed (Suppl.Fig.2). Plasma AST and ALT levels were similar in Pparα WT and KO mice, suggesting comparable tissue damage (Fig.1C,D). Conversely, plasma ketone body levels increased strongly upon infection in WT, but not in Pparα KO mice (Fig.1E). In addition, infection-induced hypoglycemia was more pronounced in Pparα KO mice, suggesting defective glucose homeostasis (Fig.1F). Interestingly, infected chimeric Pparα WTbm->KO mice also displayed lower plasma ketone body levels and more pronounced hypoglycemia compared to WTbm->WT controls (Suppl.Fig.1B,C). These results demonstrate that the metabolic response to sepsis depends on PPARα expression in non-hematopoietic derived cells. In line with previous observations [15], Pparα -deficiency resulted in decreased, rather than increased, inflammatory responses as illustrated by lower plasma tumor necrosis factor alpha (Tnfa) and chemokine Kc/Cxcl1 levels 6hrs after infection (Fig.1G,H). Since hepatic PPARα contributes to systemic inflammation [27], histological analysis of livers from infected Pparα WT and KO mice was performed. Surprisingly, whole body Pparα -deficiency did not affect infection-induced neutrophil or monocyte/macrophage recruitment in livers as assessed by Ly6G and Moma2 stainings (Suppl.Fig.3). Plasma myeloperoxidase activity (Fig.1I) and bacterial dissemination into peritoneum and blood (Fig.1J) were also unaffected by whole body Pparα -deficiency.

Whole body Pparα-deficiency modulates the hepatic metabolic and inflammatory gene expression responses to bacterial sepsis

To identify molecular pathways regulated by PPARα in response to sepsis, microarray analysis was performed on livers from whole body Pparα WT and KO mice 16hrs post-infection and from uninfected controls. In addition to inflammation-related pathways, the Gene Ontology (GO) term “Oxidation-Reduction Process,” was altered by infection in Pparα WT mice, suggesting altered mitochondrial function (Fig.2A). Unbiased clustering analysis of genes affected by sepsis in either whole body Pparα WT or KO mice revealed 3 gene clusters (Fig.2B,C and Suppl.Fig.4). While cluster 1 (black) and cluster 3 (purple) contained genes whose response to infection (induced or suppressed, respectively) was largely maintained in whole body Pparα KO mice, genes in cluster 2 (green) were either unresponsive or responded in the opposite direction in Pparα KO mice despite strong regulation in Pparα WT mice upon infection. Cluster 1 was enriched in genes associated with inflammatory response and ROS metabolism, and cluster 3 was enriched in genes involved in cellular glucuronidation and response to xenobiotics (Suppl.Fig.4). Interestingly, Cluster 2 contained many genes related to FA metabolism and circadian rhythm, many of which are bona fide PPARα targets. Indeed, carnitine palmitoyl transferase (Cpt)1a, acyl-Coenzyme A oxidase (Acox)1 and pyruvate dehydrogenase kinase (Pdk)4, were induced by sepsis in Pparα WT mice, but either reduced or unchanged by infection in Pparα KO mice (Fig.2D). Together, these results demonstrate that the most profound transcriptional differences between whole body Pparα WT and KO mice in response to sepsis are related to regulation of PPARα’s metabolic targets, rather than to effects on inflammation or other pathways.

Figure 2. Whole body Pparα -deficiency modulates hepatic metabolic and inflammatory transcriptional responses to infection.

Livers from Pparα WT and KO mice injected (ip) with vehicle (PBS) (Control) or E.coli (4x10⁸ live bacteria) (Infected) were collected after 16hrs and transcriptomic analysis was performed (n=6 mice/group). (A) Top enriched GO terms for genes differentially expressed comparing infection vs control in Pparα WT mice. (B) Hierarchical clustering and (C) dot plot of genes affected by infection in Pparα WT or KO mice. (D) Selected genes from the “fatty acid oxidation pathway" GO term of Cluster 2. (E) Mitochondrial respiration measured with (pyruvate/malate (PYR) or octanoylcarnitine/malate (OCTA) as described in methods (n=7-8 mice/group). Statistical differences are indicated (2way ANOVA: *** p<0.001, ** p<0.01 and * p<0.05 for effect of infection; §§§ p<0.001; §§ p<0.01; § p<0.05 for genotype effect; ns: non-significant).

Whole body Ppar α-deficiency impairs the hepatic metabolic shift from glucose to lipid utilization during bacterial sepsis

To further characterize the metabolic contribution of PPARα upon sepsis, glucose and lipid metabolism gene expression was measured in livers of infected Pparα WT and KO mice (Supp.Fig.5). Sepsis increased expression of genes involved in gluconeogenesis, such as phosphoenolpyruvate carboxykinase (Pck)1 and fructose-1,6-biphosphatase (Fbp)1, although the magnitude of response was lower in Pparα KO mice (Pparα WT vs KO: Pck1: 3.00vs1.67; Fbp1: 2.81vs1.44-fold induction) (Supp.Fig.5A). However, sepsis raised expression of Pdk4, which inhibits the final step of glycolysis,>100-fold in a PPARα-dependent manner (Supp.Fig.5B). Similarly, sepsis strongly increased (>12-fold) hepatic adipose triglyceride lipase (Atgl/Pnpla2) expression in Pparα WT mice only (Supp.Fig.5C). Moreover, sepsis increased several PPARα target genes involved in hepatic lipid metabolism such as FA uptake (e.g. Fatp1, Cd36), FA activation (e.g. Acsl1), peroxisomal FA β-oxidation (Acox1) and mitochondrial FA transport and β-oxidation (e.g. Cpt1a, Lcad) (Supp.Fig.5D). Interestingly, except for Cpt1a, sepsis-mediated induction of these genes was PPARα-dependent. Moreover, expression of 3-hydroxy-3-methylglutaryl-Coenzyme A synthase (Hmgcs)2, the rate-limiting enzyme for β-hydroxybutyrate production, was lower in Pparα -deficient mice (Supp.Fig.5E). Furthermore, infected chimeric Pparα WTbm->KO mice, expressing Pparα only in hematopoietic cells, also displayed lower hepatic expression of Pdk4, Cd36, Acox1, Lcad, Hmgcs2 and Atgl compared to WTbm->WT mice (Supp.Fig.5F,G), demonstrating that the metabolic transcriptional response to sepsis in liver depends on non-hematopoietic PPARα.

Functional analysis of liver mitochondria also suggested impaired FA utilization in whole body Pparα KO mice. Unaltered citrate synthase activity indicated that Pparα -deficiency did not impact mitochondrial quantity (Supp.Fig.5H). Uninfected whole body Pparα KO livers displayed lower RCR values upon incubation with pyruvate and malate (PYR) and, to a lesser extent, with octanoylcarnitine and malate (OCTA) compared to uninfected Pparα WT (Fig.2E). However, bacterial infection shifted the respiration rate from pyruvate/malate to octanoylcarnitine/malate in Pparα WT, whereas this shift was less apparent in Pparα KO livers (Fig.2E). Together, these data indicate that sepsis shifts the transcriptional and metabolic program from glucose to lipid utilization and this is impaired in whole body Pparα KO livers.

Hepatocyte-specific PPARα expression is required for proper metabolic response and survival upon bacterial infection.

Because the results of both transcriptomic and metabolic analysis suggested that hepatic PPARα is activated upon sepsis, we postulated that hepatic PPARα may be critical to the organism’s response to infection. Therefore, hepatocyte-specific Pparα^hepKO were subjected to bacterial infection and their response compared to Pparα ^hepWT and whole body Pparα KO mice. Similar to whole body Pparα KO mice, Pparα ^hepKO showed increased mortality compared to Pparα ^hepWT mice (Fig.1A&3A). Infected Pparα ^hepKO mice also displayed lower ketone body levels compared to Pparα ^hepWT, despite similar levels of plasma FFA, ALT and AST (Fig.3B-D and Supp.Fig.6). Liver triglyceride (TG) content increased more markedly upon bacterial infection in Pparα ^hepKO vs Pparα ^hepWT mice (3.1 vs 2.2-fold, Fig.3F), suggesting defective hepatic lipid utilization. Surprisingly, infected Pparα ^hepKO mice displayed no significant differences in glycaemia (Fig.3E), nor plasma levels of inflammatory cytokines (Tnfa, Kc and IL-6) at both 5hrs and 16hrs post-infection (Fig.3G,H & Supp.Fig.7) compared to Pparα ^hepWT. In addition, bacterial dissemination in peritoneum and blood did not differ between Pparα ^hepWT and Pparα ^hepKO mice (Fig.3I). These data suggest that the increased mortality observed in whole body Pparα KO is unlikely caused by the different glycemic and systemic inflammatory responses.

Figure 3. Hepatocyte-specific Pparα -deficiency results in metabolic perturbations and aggravates mortality during bacterial infection.

(A) Pparα ^hepWT, Pparα ^hepKO, and Pparα KO mice were injected (ip) with E.coli (7x10⁸ live bacteria). Survival was followed for 8 days after bacterial infection (n=12-18 mice/group). Plasma was collected 5 (G,H) or 16hrs (B-E) 16hrs after bacterial infection and (B) free fatty acids (FFA), (C) alanine amino transferase (ALT), (D) ketone bodies, (E) blood glucose, (G) Tnfa and (H) Kc/Cxcl1 concentrations were measured as described in methods. Livers, peritoneal fluid and blood were collected 16hrs after injection and (F) TG content and (I) bacterial levels were determined as described in methods (n=7-8 mice/group). Statistical differences are indicated (Survival test: Log-rank (Mantel-Cox) Test. * p<0.05, *** p<0.001 compare to survival of PPARα ^hepWT; 2-way ANOVA: §§§ p<0.001; §§ p<0.01; § p<0.05 for genotype effect; ns: non-significant).

Similar to whole body Pparα KO mice, Pparα^hepKO mice displayed major defects in sepsis-modulated regulation of several genes involved in hepatic glucose (i.e. Pdk4, Fig.4A) and lipid metabolism (e.g. Cd36, Acox1, Vlcad and Lcad, Fig.4B), as well as Pparα itself (Fig.4C). Interestingly, the induction of hepatic Atgl and Cpt1a expression upon sepsis was independent of hepatocyte PPARα (Fig.4B). Conversely, hepatic Hmgcs2 expression was virtually undetectable upon sepsis in Pparα^hepKO mice (Fig.4B). Bacterial infection increased Atgl, but not Hmgcs2 or Dgat1 expression at the protein level (Supp.Fig.8). Moreover, Hmgcs2 protein levels were lower in Pparα^hepKO mice both in uninfected and infected conditions. Whereas Atgl protein induction appeared less pronounced upon sepsis in Pparα^hepKO mice, Dgat1 protein expression only increased in Pparα^hepKO mice, both consistent with increased hepatic TG content. Together, these data indicate that hepatocyte-specific PPARα-deficiency profoundly affects the hepatic metabolic response to infection.

Figure 4. Hepatocyte-specific Pparα -deficiency impairs the response of lipid metabolism genes to bacterial infection.

Pparα ^hepWT, Pparα ^hepKO mice were injected (ip) with vehicle (PBS) (-) or with E.coli (6x10⁸ live bacteria) (inf) (+). Livers were collected 16hrs after infection and hepatic mRNA expression of (A) Pdk4, (B) genes involved in lipid metabolism and (C) Ppara was measured. Statistical differences are indicated (2way ANOVA: *** p<0.001, ** p<0.01 and * p<0.05 for effect of infection; §§§ p<0.001; §§ p<0.01; § p<0.05 for effect of genotype; ns: non-significant).

Pparα-deficiency regulates the inflammatory response, but does not promote innate immune cell recruitment or inhibit autophagy in the liver upon bacterial sepsis

To determine the contribution of non-hematopoietic PPARα to systemic inflammation in sepsis, expression of inflammatory genes was measured in livers and spleens from whole body Pparα KO, chimeric Pparα WTbm->KO mice and their respective controls. The induction of Tnfa, Mcp1, Il6 and Ifng upon sepsis was markedly attenuated in livers (Fig.5A), and, to a lesser extent, in spleens (Suppl.Fig.9) of whole body Pparα KO mice compared to controls. Likewise, induction of the vascular inflammation markers Vcam1 and Icam1 (Fig.5B) and the mitochondrial anti-oxidant enzyme Sod2 (Fig.5C) was lower in livers of whole body Pparα KO mice than in their WT counterparts. By contrast, the hepatic and vascular inflammatory and anti-oxidant responses (Fig.5D&E), as well as MPO activity (Fig.5F) and immune cell recruitment (Suppl.Fig.10) were similar in chimeric Pparα WTbm->KO and WTbm->WT mice, despite differing survival outcomes (Fig.1A, Supp.Fig.1A). Interestingly, inPparα ^hepKO mice, the induction of inflammatory genes (Fig.6A) upon sepsis was either higher (Tnfa) or unchanged (Mcp1, Il6), whereas neutrophil and monocyte/macrophage recruitment was again similar (Ly6G and Moma2 stainings, Fig.6C-E) upon infection. Altogether, these data indicate that the attenuated inflammatory response observed in livers of whole body Pparα KO mice depends on hematopoietic, but not hepatic Pparα expression.

Figure 5. The sepsis-induced inflammatory response occurs through hematopoietic PPARα.

Whole body Pparα WT and KO mice (A-C) or chimeric Pparα WTbm->WT and WTbm->KO mice (D-F) were injected (ip) with vehicle (PBS) (-) or E.coli (4x10⁸ live bacteria) (Inf) (+). Livers were collected 16hrs after infection and mRNA expression of genes involved in inflammation (A, D), endothelial activation (B, E), and oxidative stress (C, E), was analyzed using RT-Q-PCR. Plasma myeloperoxidase (MPO) activity (F) was measured as described in methods (n=8 mice/group). Statistical differences are indicated (2way ANOVA: *** p<0.001, ** p<0.01 and * p<0.05 for effect of infection; §§§ p<0.001; §§ p<0.01; § p<0.05 for genotype effect; ns: non-significant).

Figure 6. Hepatocyte-specific Pparα -deficiency modulates the inflammatory response in the liver without affecting innate immune cell recruitment upon bacterial infection.

Pparα ^hepWT, Pparα ^hepKO mice were injected (ip) with vehicle (PBS) (-) or E.coli (6x10⁸ live bacteria) (inf) (+). Livers were collected 16hrs after infection and mRNA expression of genes involved in (A) inflammation, (B) endothelial activation and oxidative stress (B) was analysed using RT-Q-PCR. Liver sections stained for Ly6G (E, top panel) and Moma2 (E, bottom panel) and quantified (C, D) respectively using NIS Element software (n=7-8 mice/group) (Bar = 100μm). Statistical differences are indicated (2way ANOVA: *** p<0.001, ** p<0.01and * p<0.05 for effect of infection; §§§ p<0.001; §§ p<0.01; § p<0.05 for genotype effect; ns: non-significant).

Because autophagy may play a protective role during sepsis [28] and PPARα mediates fasting-induced autophagy [29], markers of autophagy were assessed in whole body and hepatocyte-specific Pparα -deficient mice. Whereas sepsis increased expression of certain autophagy genes, their regulation was not different in whole body Pparα KO mice (Supp.Fig.11A). Moreover, hepatocyte-specific Pparα -deficiency (Supp.Fig.11B-C) rather resulted in more pronounced induction of Ulk1, Atg5, Bnip3 and Becn1 gene expression and Lc3b-II/I protein ratio, suggestive of a compensatory induction of autophagy to combat the deleterious response to sepsis in Pparα^hepKO mice. Altogether, these data indicate that hepatocyte PPARα expression contributes to protection against sepsis by controlling the systemic metabolic response.

PPARα expression and activity is lower in livers of critically ill patients

To determine whether hepatic PPARα expression is altered in critically ill human, livers from non-surviving critically ill patients and healthy controls were analysed for PPARα and target gene expression (Fig.7A). Interestingly, PPARA expression was lower in livers of critically ill patients. Moreover, expression of genes involved in TG lipolysis (ATGL), glucose oxidation (PDK4), FA uptake and β-oxidation (CD36, LCAD) and ketogenesis (HMGCS2) were also lower and correlated with PPARA expression (Fig.7A,B). Surprisingly, PPARA mRNA levels did not correlate with serum cytokine levels in these patients (Fig.7B), suggesting a critical role for hepatic PPARα in the metabolic, but not in the inflammatory response to sepsis in critically ill human patients.

Figure 7. PPARα gene expression in liver biopsies of critically ill patients correlates with decreased expression of FA utilization genes.

(A) Liver biopsies from elective surgery (n=20) and critically ill patients (n=46) were collected and mRNA was analysed using RT-Q-PCR. Statistical differences are indicated (Wilcoxon test: *** p<0.001; ** p<0.01; compared to healthy). Serum cytokines were quantified as described in methods. (B) Correlations of hepatic PPARA mRNA expression with metabolic gene expression or serum cytokine levels from critically ill patients were calculated. Statistical differences are indicated (Pearson (r): *** p<0.001; ** p<0.01).

Discussion

Our results demonstrate that PPARα protects against sepsis primarily by controlling the metabolic response in the hepatocyte, by shifting its energy utilization from glucose to FA and by increasing ketogenesis.

The host defense toward bacterial infection is a complex response involving resistance (to limit microbial burden) and tolerance (to limit tissue injury and organ dysfunction) mechanisms. These processes require metabolic reprogramming in immune and non-immune cells [30]. Resistance is characterized by a balance between local activation of pro-inflammatory pathways to restrain and eliminate invading pathogens and anti-inflammatory pathways required to prevent exaggerated systemic inflammation [31,32]. Our data show that whole body Pparα -deficiency attenuates organ and systemic inflammatory responses upon infection. This contrasts with previous observations in models of sterile chronic and acute inflammation in which Pparα -deficiency results in exacerbated inflammatory responses to endotoxemia in vascular, splenic and liver cells [33]. Accordingly, whole body Pparα -deficiency also resulted in a decreased pro-inflammatory response and survival in a cecal ligation sepsis model [15].

Interestingly, our data and others’ indicate that non-hematopoietic PPARα action is an important determinant of survival. Studies by Standage et al. suggest that heart PPARα expression contributes to survival during sepsis by increasing cardiac performance and FA oxidation [16,17]. In the present study, we demonstrate that PPARα expression and activation in hepatocytes, but not immune cells, contributes to protection against sepsis by promoting an appropriate metabolic response, hence improving survival. Sepsis activates hepatic PPARα, which results in activation of FA metabolism and ketogenesis-related target genes, and elevates plasma ketone body levels. Indeed, Pparα ^hepKO mice displayed higher hepatic TG accumulation, reduced plasma ketonemia, and lower hepatic expression of FA metabolism and ketogenesis genes. Decreased ketogenesis in septic Pparα-deficient mice was unlikely due to defective AT lipolysis, since plasma FFA levels were unaffected by whole body nor hepatocyte-specific Pparα -deficiency.

Defective FA oxidation and ketogenesis in septic Pparα -deficient mice may indirectly contribute to the aggravation of hypoglycaemia and mortality. Mouse models displaying FA oxidation defects are often hypoglycemic upon LPS administration, e.g. Mcad-deficient mice [34,35] and Tbp2-deficient mice [36]. Moreover, several lines of evidence indicate that PPARα plays an important role in glucose homeostasis [37]. Under septic conditions, the hypoglycemia in whole body Pparα -deficient mice may involve increased hepatic glucose utilization. Indeed, Pparα -deficiency impairs the induction of Pdk4 gene expression upon sepsis, whereas the expression of gluconeogenic genes was not affected. Still, assessment of mitochondrial respiration in Pparα -deficient livers suggests that infection does not increase pyruvate oxidation. Thus, the hypoglycemia might also result from increased peripheral glucose uptake by metabolic organs, such as the heart, muscle and brain, to compensate for both the inability to catabolize FA [38,39] and reduced ketone body availability. In line, septic Pparα ^hepKO mice display less pronounced hypoglycaemia than whole body Pparα –deficient mice. However, it has been shown that decreased FA oxidation and ketogenesis as a result of Pparα or Fgf21-deficiency can lead to an inability to maintain tissue tolerance to bacterial sepsis leading to neuronal dysfunction and death [40]. Moreover, ketone body therapy protects against lipotoxicity and acute liver failure in Pparα -deficient mice [41]. Altogether, these and our data suggest that the increased mortality during sepsis may be caused by a deficiency in beneficial energetic substrates produced by FA oxidation in hepatocytes, such as ketone bodies, to maintain tissue protection.

Interestingly, in livers of non-surviving critically ill patients, PPARA mRNA levels are lower and correlate with the lower expression of genes involved in lipid and glucose metabolism, but not with plasma markers of inflammation. These data corroborate findings from a clinical metabolomic study showing that lactate, pyruvate, acetyl-carnitine and several citric acid cycle metabolites, were higher in sepsis non-survivors compared to survivors, suggesting that a profound defect in FA β-oxidation, possibly as a result of mitochondrial dysfunction, is associated with the incidence of death in critical ill patients [42,43].

In conclusion, we have shown that Pparα -deficiency in hepatocytes during sepsis is deleterious as it impairs the adaptive metabolic shift from glucose to FA utilization. While most current approaches to treat sepsis aim to harness the inflammatory response, our results might pave the way for strategies based on adaptive energy homeostasis.

Lay summary.

As the main cause of death of critically ill patients, sepsis remains a major health issue lacking efficacious therapies. While current clinical literature suggests an important role for inflammation, metabolic aspects of sepsis have been mostly overlooked. Here, we show that mice with an impaired metabolic response, due to deficiency of the nuclear receptor PPARα in the liver, exhibit enhanced mortality upon bacterial infection despite a similar inflammatory response, suggesting that metabolic interventions may be a viable strategy for improving sepsis outcomes.

Highlights.

• -Sepsis activates hepatic PPARα
• -PPARα plays a protective role in sepsis
• -Pparα -deficiency impairs FA utilization in the liver during sepsis
• -Hepatocyte Pparα -deficiency worsens the outcome of bacterial infection
• -PPARα activity is lower in livers of non-surviving critically ill patients