Liver dysfunction and phosphatidylinositol-3-kinase signalling in early sepsis: experimental studies in rodent models of peritonitis

Abstract

Background: Hepatic dysfunction and jaundice are traditionally viewed as late features of sepsis and portend poor outcomes. We hypothesized that changes in liver function occur early in the onset of sepsis, yet pass undetected by standard laboratory tests.

Methods and findings: In a long-term rat model of faecal peritonitis, biotransformation and hepatobiliary transport were impaired, depending on subsequent disease severity, as early as 6 h after peritoneal contamination. Phosphatidylinositol-3-kinase (PI3K) signalling was simultaneously induced at this time point. At 15 h there was hepatocellular accumulation of bilirubin, bile acids, and xenobiotics, with disturbed bile acid conjugation and drug metabolism. Cholestasis was preceded by disruption of the bile acid and organic anion transport machinery at the canalicular pole. Inhibitors of PI3K partially prevented cytokine-induced loss of villi in cultured HepG2 cells. Notably, mice lacking the PI3Kγ gene were protected against cholestasis and impaired bile acid conjugation. This was partially confirmed by an increase in plasma bile acids (e.g., chenodeoxycholic acid [CDCA] and taurodeoxycholic acid [TDCA]) observed in 48 patients on the day severe sepsis was diagnosed; unlike bilirubin (area under the receiver-operating curve: 0.59), these bile acids predicted 28-d mortality with high sensitivity and specificity (area under the receiver-operating curve: CDCA: 0.77; TDCA: 0.72; CDCA+TDCA: 0.87).

Conclusions: Liver dysfunction is an early and commonplace event in the rat model of sepsis studied here; PI3K signalling seems to play a crucial role. All aspects of hepatic biotransformation are affected, with severity relating to subsequent prognosis. Detected changes significantly precede conventional markers and are reflected by early alterations in plasma bile acids. These observations carry important implications for the diagnosis of liver dysfunction and pharmacotherapy in the critically ill. Further clinical work is necessary to extend these concepts into clinical practice. Please see later in the article for the Editors' Summary.

Figure 1. Rat model allowing assessment of early changes in gene expression, and relationships to outcome.

(A) Protocol for instrumentation, induction of sepsis, echo-guided prognosis stratification, and tissue harvesting. (B) The heatmap displays expression profiles of 1,373 genes with significantly different expression levels among the naïve, sham, predicted sepsis survivor, and predicted non-survivor groups (n = 4 per group). k-Means clustering revealed four gene clusters with comparable behaviour. Associated colours represent variance-normalised expression for individual genes: red colour intensity indicates a relative expression greater than the mean, blue indicates the opposite. (C) The transcripts identified in (B) were analysed by Ingenuity Pathway Analysis to identify canonical pathways found to be up- or down-regulated in animals predicted to die. Presented is a bar chart of the most statistically significant (corresponding to p-value) down-regulated (blue) and up-regulated (red) pathways associated with poor prognosis. Numbers in parentheses represent the number of differentially regulated transcripts involved in a defined pathway.

Figure 2. Sepsis-induced excretory dysfunction, as visualised by accumulation of the xenobiotic indocyanine green and bilirubin.

(A) Non-invasive whole body near-infrared fluorescence imaging of ICG at 15 h after sepsis induction. Shown are representative images from sham-treated animals (left panels) and animals subjected to lethal sepsis (right panels) for five biological replicates each. While ICG is eliminated via hepatobiliary excretion into the duodenum within the 300-min observation period in sham animals, the dye accumulates in the livers of septic animals, with an almost absence of fluorescence signal in the gut, as confirmed after laparotomy (lower panels). The dotted line divides upper and lower abdominal quadrants for orientation. (B) Subsequent epifluorescence microscopic examination of liver surfaces at 300 min after ICG administration in sham and septic animals. Stars indicate central veins (magnification 50×, pseudo-coloured). (C) ICG fluorescence intensities around central veins were significantly higher in septic compared to sham-operated animals (n = 4 animals/group, five perivenous and five periportal areas per animal, *p = 0.031 compared to sham). (D) Representative micro-Raman images of tissue sections from liver sections obtained from sham-operated and septic rats 15 h post-insult. In the livers of the septic rats, relative intensities of the bilirubin component are elevated in the perivenous region (stars). In contrast, only minor local spots of accumulated bilirubin could be verified in livers of control animals. Scale bars: 50 µm. (E) Raman spectrum of the bilirubin component (red trace) with crystalline bilirubin for comparison (black trace). Main spectral contributions of bilirubin are resolved near 971, 1,229, and 1,606 cm⁻¹ (arrows at 971 and 1,229 cm⁻¹; high peak at 1,606 cm⁻¹).

Figure 3. Altered expression of canalicular bile acid and organic anion transporters in sepsis.

(A) Distribution and localisation pattern of Bsep and Mrp2 in livers obtained from control and septic rats using confocal microscopy. Na,K-ATPase was used to stain the basolateral plasma membrane. Staining for Bsep revealed a substantial decrease in protein expression after sepsis induction. In sham-treated animals immunoreactive Mrp2 delineates bile canaliculi. The Mrp2 staining pattern in septic rats is irregular, disrupted, and accompanied by recovery of Mrp2 inside the cell, as shown by punctuate staining in the pericanalicular region. Lowermost panels represent merged pictures of Na,K-ATPase and Mrp2. Scale bar: 20 µm. (B) Electron micrographs of freeze-fracture-immunolabelled Mrp2 protein in the plasma membrane of isolated hepatocytes at 15 h post-infection to further study irregular staining observed in sepsis. The cell surface of sham-treated animals is densely covered with microvilli, visible as cross-fractured stubs (stars). A lengthwise fracture of the microvilli membrane exposes a dense labelling with Mrp2 protein arranged in small clusters along microvilli (white arrowheads). By contrast, hepatocytes isolated from septic animals exhibited a loss of microvilli, while Mrp2 was predominantly found in vesicles located closely beneath the canalicular membrane (black arrowheads). The inset shows a top view of a bile canaliculus formed by two adjacent hepatocytes (scale bar: 1 µm). Scale bars in upper and lower panels indicate 200 nm and 100 nm, respectively. EF, exoplasmic fracture face; PF, protoplasmic fracture face.

Figure 4. Polymicrobial sepsis causes deranged bile acid conjugation and transport.

At 15 h after sepsis induction, plasma, liver tissue, and bile were subjected to targeted metabolomics. Expression of BAAT, facilitating conjugation to taurine and glycine, was quantified by immunoblotting. (A) The plot depicts median log₂ fold changes of unconjugated as well as glycine- and taurine-conjugated bile acid in plasma, liver, and bile, comparing septic to sham-operated rats (n = 12 per group, *p<0.05 or **p<0.01 compared to sham). (B) Conjugation index as a surrogate for the observed conjugation defect reflected by the ratio of unconjugated bile acids CA and CDCA to the corresponding taurine (TCA and taurochenodeoxycholic acid) and glycine (GCA and glycochenodeoxycholic acid) conjugates in plasma, liver and bile (ratio given as log₂ fold change, n = 12 per group). (C and D) Representative immunoblots of BAAT 15 h after sepsis induction in cytosolic (c) as well as peroxisomal (p) fractions, with corresponding densitometric analysis (n = 5 for sham, n = 8 for sepsis; BAAT (c): *p = 0.002; BAAT (p): *p = 0.006 compared to sham). Densitomentric values are normalised to β-actin. DCA, deoxycholic acid; GDCA, glycodeoxycholic acid; GCDCA, glycochenodeoxycholic acid; GLCA, glycolithocholic acid; GLCAS, glycolithocholic acid sulphate; GUDCA, glycoursodeoxycholic acid; TCDCA, taurochenodeoxycholic acid; TLCA, taurolithocholic acid; TLCAS, taurolithocholic acid sulphate; TUDCA, tauroursodeoxycholic acid; UDCA, ursodeoxycholic acid.

Figure 5. Plasma bilirubin and (un)conjugated bile acid levels in patients on the day of diagnosis of severe sepsis.

(A) The plot depicts median log₂ fold changes of bilirubin (Bili), and unconjugated as well as glycine- and taurine-conjugated bile acid quantities in plasma of severely septic patients (n = 48) fulfilling American College of Chest Physicians/Society of Critical Care Medicine consensus criteria compared to non-septic controls (n = 20) (*p<0.05 compared to controls). (B) Receiver operating characteristics of bilirubin, CDCA, TDCA, or the combined performance of CDCA+TDCA in predicting 28-d mortality. DCA, deoxycholic acid; GDCA, glycodeoxycholic acid; GCDCA, glycochenodeoxycholic acid; GLCA, glycolithocholic acid; GLCAS, glycolithocholic acid sulphate; GUDCA, glycoursodeoxycholic acid; TCDCA, taurochenodeoxycholic acid; TLCA, taurolithocholic acid; TLCAS, taurolithocholic acid sulphate; TUDCA, tauroursodeoxycholic acid; UDCA, ursodeoxycholic acid.

Figure 6. Polymicrobial sepsis severely impairs the activity of enzymes responsible for phase I and II biotransformation.

(A) Activities of CYP1A, CYP2A, CYP2B, CYP2C, and CYP2E were assessed by ethoxycoumarin O-deethylation, while (B) CYP3A activity was quantified using the ethylmorphine N-demethylation model reaction for phase I biotransformation. Glutathione-S-transferase activity (C) and bilirubin glucuronidation (D), representing typical phase II conjugation reactions, were assessed by the model reaction 1-chloro-2,4-dinitrobenzene conjugation, resulting in the formation of dinitrobenzene-glutathione conjugate (GS-DNB), or the Burchell method, respectively (n = 5 for sham, n = 8 for sepsis, *p = 0.004 compared to sham).

Figure 7. Stimulation of HepG2 cells with prototypical sepsis-associated mediators triggers internalisation of pseudovilli via PI3K signalling.

(A) Representative electron microscopic images of HepG2 cells subjected to non-specific (Wortmannin) or specific (AS605240) inhibition of PI3Kγ prior to stimulation with a mix of TNF-α (50 ng/ml), IL-1β (10 ng/ml), IFN-γ (10 ng/ml), and LPS (100 ng/ml) for 6 h. Untreated and cytokine-mix-stimulated cells served as controls. Arrowheads and the inset indicate pseudovilli, representing the equivalent for canalicular microvilli formed by this cell line. Scale bars: 10 µm, inset scale bar: 5 µm. (B) For quantification, the ratio of cells carrying pseudovilli to the total number of cells within the observation field was analysed (p<0.001, post-hoc test: *p<0.05 compared to control, ^# p<0.05 compared to cytokine mix, n = 9–13 fields of interest per treatment). (C) Representative electron micrographs of freeze-fracture-immunolabelled Mrp2 transporters in the plasma membrane of HepG2 cells. Cells were treated as described in (A). Inset indicates a cluster of membrane-incorporated immunolabelled Mrp2 protein. Scale bars: 200 nm, inset scale bar: 100 nm. (D) For quantification, Mrp2 particles were assessed and normalised to the protoplasmic fracture face area. (p<0.001, post-hoc test: *p<0.05 compared to control, ^# p<0.05 compared to cytokine mix, n = 23–25 fields of interest per treatment).

Figure 8. PI3Kγ^−/− mice are protected against hepatic neutrophil infiltration and cholestasis despite increased inflammatory mediators.

(A) Representative electron micrographs for morphologic assessment of bile canaliculi in liver tissue obtained from wild-type and PI3Kγ^−/− mice 6 h after sepsis induction. Stars indicate widened bile canaliculi with disrupted brush borders in wild-type mice. Small vesicles located closely beneath the canalicular plasma membrane were observed in wildtype mice, indicative of subapical vesicles (arrowheads). By contrast, brush borders of PI3Kγ knock-out mice were well maintained. Scale bars: 1 µm. (B) Levels of total bilirubin in sham-operated and septic wild-type and PI3Kγ^−/− animals (p<0.001, post-hoc test: *p<0.001 wild-type sham versus wild-type sepsis, ^# p = 0.008 wild-type sepsis versus knock-out sepsis; n = 7–9 per group). (C) Assessment by intravital microscopy of infiltrating esterase-positive leucocytes per square millimetre of liver surface in wild-type and PI3Kγ^−/− mice at 6 h under sham or septic conditions (p<0.001, post-hoc test: *^,# p<0.05 wild-type sham versus wild-type sepsis, as well as wild-type sepsis versus knock-out sepsis; n = 3 animals per group with five regions of interest per animal). (D–F) Mean levels of TNF-α, IL-6, and MCP-1 from wild-type and PI3Kγ-deficient mice at 6 h under sham and septic conditions (p<0.001, post-hoc test: ^# p<0.001 for wild-type sham versus wild-type sepsis, knock-out sham versus knock-out sepsis and wild-type sepsis versus knock-out sepsis; n = 7–9 per group). (G and H) Representative immunoblots of BAAT 6 h after sepsis induction in cytosolic (c) and peroxisomal (p) fractions, with corresponding densitometric analysis (n = 4 each for wild-type and PI3Kγ^−/− samples). Densitomentric values are normalised to β-actin. (I) Conjugation index reflected by the ratio of unconjugated bile acids CA and CDCA to the corresponding taurine (TCA and taurochenodeoxycholic acid [TCDCA]) and glycine (GCA and glycochenodeoxycholic acid [GCDCA]) conjugates in plasma of wild-type and PI3Kγ^−/− mice at 6 and 15 h (ratio given as log₂ fold change, n = 5–7 animals/group).