Critical illness evokes elevated circulating bile acids related to altered hepatic transporter and nuclear receptor expression


  • Potential conflict of interest: Nothing to report.

  • Supported by grants from the Research Foundation Flanders (G.0278.03), the Research Fund of the Katholieke Universiteit Leuven (KULeuven GOA 2007/14), the Methusalem grant of the Flemisch Government (to G.VdB.) and National Health and Medical Research Council of Australia Project Grant 512354 (to C.L.). D.M. holds a postdoctoral Clinical Research Fund fellowship from the University Hospitals KULeuven. LL holds a post-doctoral Research Fund fellowship from the Research Foundation Flanders.


Hyperbilirubinemia is common during critical illness and is associated with adverse outcome. Whether hyperbilirubinemia reflects intensive care unit (ICU) cholestasis is unclear. Therefore, the aim of this study was to analyze hyperbilirubinemia in conjunction with serum bile acids (BAs) and the key steps in BA synthesis, transport, and regulation by nuclear receptors (NRs). Serum BA and bilirubin levels were determined in 130 ICU and 20 control patients. In liver biopsies messenger RNA (mRNA) expression of BA synthesis enzymes, BA transporters, and NRs was assessed. In a subset (40 ICU / 10 controls) immunohistochemical staining of the transporters and receptors together with a histological evaluation of cholestasis was performed. BA levels were much more elevated than bilirubin in ICU patients. Conjugated cholic acid (CA) and chenodeoxycholic acid (CDCA) were elevated, with an increased CA/CDCA ratio. Unconjugated BA did not differ between controls and patients. Despite elevated serum BA levels, CYP7A1 protein, the rate-limiting enzyme in BA synthesis, was not lowered in ICU patients. Also, protein expression of the apical bile salt export pump (BSEP) was decreased, whereas multidrug resistance-associated protein (MRP) 3 was strongly increased at the basolateral side. This reversal of BA transport toward the sinusoidal blood compartment is in line with the increased serum conjugated BA levels. Immunostaining showed marked down-regulation of nuclear farnesoid X receptor, retinoid X receptor alpha, constitutive androstane receptor, and pregnane X receptor nuclear protein levels. Conclusion: Failure to inhibit BA synthesis, up-regulate canalicular BA export, and localize pivotal NR in the hepatocytic nuclei may indicate dysfunctional feedback regulation by increased BA levels. Alternatively, critical illness may result in maintained BA synthesis (CYP7A1), reversal of normal BA transport (BSEP/MRP3), and inhibition of the BA sensor (FXR/RXRα) to increase serum BA levels. (HEPATOLOGY 2011;)

Almost 20% of the intensive care unit (ICU) patients develop ICU jaundice or cholestasis, which has been linked to an increased risk of mortality and length of stay.1, 2 Currently there is no consensus on the definition of cholestasis during critical illness. Most commonly, routine laboratory measurements of bilirubin and alkaline phosphatase (ALP)/gamma-glutamyl transpeptidase (GGT) with different cutoffs are used.2-4 Therefore, in clinical practice ICU cholestasis is the equivalent of conjugated hyperbilirubinemia. Because a causal link between hyperbilirubinemia and worse outcome is missing, it may even be a biochemical epiphenomenon. Additionally, the reliability of hyperbilirubinemia as a marker of cholestasis in critically ill patients may be questionable, because there are many factors that can influence the levels of bilirubin. The weakness of bilirubin as a marker of cholestasis during critical illness and the absent mechanistic underpinning of ICU cholestasis were the main drivers for this study.

To date, the behavior and impact of bile acids (BAs) during ICU jaundice has been neglected, despite their crucial role in bile formation,5 lipid/cholesterol metabolism, and energy and glucose homeostasis.6 Also, studies of the BA transporters and their regulatory network of nuclear receptors (NRs) has so far been focused on either chronic cholestatic liver disorders, such as primary biliary cirrhosis, or familial intrahepatic cholestasis,7 or on acute animal models of sepsis.8 Endotoxin-induced proinflammatory cytokines lead to reduced Na+/taurocholate cotransporting polypeptide (NTCP) and organic anion transporting polypeptide (OATP) expression.8 Expression of the canalicular efflux pumps, bile salt export pump (BSEP), and multidrug resistance-associated protein (MRP) 2 is reduced during rat endotoxemia, whereas multidrug resistance protein (MDR) 1 expression is increased.8 MRP3 and MRP4, inducible basolateral efflux pumps, are strongly up-regulated and may serve as an alternative escape route for cytotoxic compounds from hepatocytes into sinusoidal blood.7

BA metabolism and transporter function is regulated by a complex network of NRs, together with their coactivators and corepressors.9 In physiological conditions farnesoid X receptor (FXR) is the “bile acid sensor”. FXR activation is known to lead to repression of basolateral BA uptake (NTCP, OATP1B1) and BA synthesis. FXR activation at the same time induces canalicular (BSEP, MRP2, MDR3) and basolateral efflux systems (organic solute transporter alpha/beta). Recently, it has become clear that the NRs vitamin D receptor (VDR), pregnane X receptor (PXR), and constitutive androstane receptor (CAR) have significant regulatory roles in BA metabolism and/or transport.9

The aim of this study was to examine a large cohort of critically ill patients to gain mechanistic insights into ICU jaundice, with a focus on BAs, hepatocytic transporters involved in bile production, as well as their regulating NRs. An understanding of these mechanisms has the potential not only to expand our knowledge of hepatic metabolic dysfunction in the critically ill, but may also convey hints whether hyperbilirubinemia or increased serum BAs are a biochemical epiphenomenon of a failing hepatobiliary system or a desired compensatory reaction during critical illness.


ALT, alanine aminotransferase; ALP, alkaline phosphatase; AST, aspartate aminotransferase; BSEP, bile salt export pump; CA, cholic acid; CAR, constitutive androstane receptor; CDCA, chenodeoxycholic acid; CK7, cytokeratin 7; CYP, cytochrome P450; DCA, deoxycholic acid; FXR, farnesoid X receptor; G-CA, glycocholic acid; G-CDCA, glycochenodeoxycholic acid; GGT, gamma-glutamyl transpeptidase; HPRT, hypoxanthine phosphoribosyltransferase; ICU, intensive care unit; IQR, interquartile range; LCA, lithocholic acid; MDR, multidrug resistance protein; MRP, multidrug resistance-associated protein; NTCP, Na+/taurocholate cotransporting polypeptide; OATP, organic anion transporting polypeptide; PXR, pregnane X receptor; RXRα, retinoid X receptor alpha; SHP, short heterodimeric partner; SEM, standard error of the mean; T-CA, taurocholic acid; T-CDCA, taurochenodeoxycholic acid; VDR, vitamin D receptor.

Material and Methods

Patients and Serum Analysis.

Postmortem liver biopsies were taken from ICU patients (n = 130), enrolled in two large randomized controlled trials studying the effects of intensive insulin therapy in critically ill patients.10, 11 All deaths occurred after a multidisciplinary decision to restrict therapy when further treatment was judged to be futile. All liver samples were harvested within minutes after death. For comparison, liver biopsies from 20 demographically matched patients (controls) undergoing an elective restorative rectal resection were obtained. All protocol and consent forms were approved by the Institutional Review Board of the Katholieke Universiteit Leuven. Written informed consent was obtained from all patients, or, when the patient was unable to give consent, from the closest family member. All liver biopsies were taken from liver segment IVb, snap-frozen in liquid nitrogen, and stored at −80°C until analysis.

For all ICU patients, blood samples were taken on admission and the last day of ICU stay. Liver biochemistry was analyzed by routine automated laboratory assays: total bilirubin, alanine aminotransferase (ALT); aspartate aminotransferase (AST), GGT, and ALP. Blood samples from controls were taken preoperatively. Sepsis was defined according to Bone criteria as suspected or documented infection on the day of admission to the ICU and fulfillment of at least two of the three criteria for the systemic inflammatory response syndrome (receiving ventilatory support, white-cell count ≤4,000 or ≥12,000 per cubic millimeter, and body temperature ≤36°C or ≥38°C).12 Serum concentrations of cytokines were quantified by a multiplexed microbead suspension enzyme-linked immunosorbent assay (Biosource, Carlsbad, CA) using the Luminex 100 system (Austin, TX) as published.13

Individual serum BAs were quantified by high-performance liquid chromatography / mass spectrometry using authentic BA standards and deuterated internal standards.14

Gene and Protein Expression on Liver Biopsies.

Total RNA was isolated and quantified as described.15 Commercial gene expression assays from Applied Biosystems were used and are listed in Supporting Data Table 4. Data are expressed as fold increase relative to the mean of the control patients. Immunoblot analysis of CYP7A1 was performed as described in the online supplement.

Histological and Immunohistochemical Analysis.

For histological and immunohistochemical analysis, liver sections from a randomly chosen subset of study patients (40 ICU and 10 controls) were used. Four-μm-thick sections were cut from frozen samples and stained with hematoxylin and eosin for a general histological assessment. For evaluation of bilirubinostasis and ductular reaction, sections were stained by Hall's method and for cytokeratin 7 (CK7) (Dako, Glostrup, Denmark). For immunohistochemistry, 5-μm-thick frozen sections were dried overnight at room temperature, fixed in acetone for 10 minutes, and washed in phosphate-buffered saline (PBS) immediately prior to use. Sections were incubated with primary antibodies for 30 minutes at room temperature. The primary antibodies used are listed in Supporting Data Table 5. For the staining of CK7, OATP2/8, MRP3, MRP2, MDR1, and MDR3 the second and third step consisted of peroxidase-labeled rabbit anti-mouse and peroxidase-labeled swine anti-rabbit immunoglobulins (both Dako). Secondary and tertiary anti-bodies were diluted (1:50 and 1:100, respectively) in PBS (pH 7.2) containing 10% normal human serum. For the staining of BSEP, the slides were incubated with an anti-rabbit peroxidase-conjugated Envision antibody (Dako) and subsequently incubated with a goat peroxidase anti-peroxidase complex (goat PAP complex; Dako). For NTCP staining a protein block was performed prior to the application of the primary antibody to counteract the strong sinusoidal staining and the secondary step consisted of peroxidase-labeled swine antirabbit IgG (dilution 1:100; Dako), followed by peroxidase-labeled rabbit anti-swine IgG (dilution 1:100; Dako).

For the staining of the NR, sections were incubated with the primary antibody for 30 minutes at room temperature and subsequently incubated for 30 minutes at room temperature with antimouse peroxidase-conjugated Envision antibody (Dako). All incubations were performed for 30 minutes at room temperature and followed by a wash in three changes of PBS for 5 minutes.

For all immunohistochemical stainings, 3-amino-9-ethylcarbazol (AEC) in 0.01% H2O2 was used as substrate-chromogen. The sections were counterstained with hematoxylin. Negative controls consisted of omission of the primary antibody and were consistently negative. To ensure uniform handling of samples, all sections were processed simultaneously. All immunohistochemically stained slides were evaluated for staining patterns and intensities by four observers (T.R., Y.V., J.W., and L.L.). Histological changes, i.e., portal inflammation; hepatocellular, canalicular, and ductular bilirubinostasis; ductular reaction; steatosis and centrolobular necrosis were graded using a semiquantitative scoring system. BA transporter expression was semiquantitatively graded as compared with what was deemed normal by the pathologist. For the assessment of NRs, intensity of nuclear localized staining was scored.

Statistical Analysis.

Statistical analysis was performed using Statview 5.0.1 (SAS Institute, Cary, NC). All quantitative values were assessed for normality. Values with normal distribution, and those that were normalized after logarithmic transformation, are represented as mean ± standard error of the mean (SEM) and were compared using the unpaired Student's t test. The nonnormally distributed data were represented as medians and interquartile range (IQR) (1st-3rd) and compared by the nonparametric Mann-Whitney U test. Nominal and ordinal variables (expressed as numbers and percentages) were compared by Fisher's exact test. Correlations between variables were calculated using either Pearson's or Spearman's rank correlation test. For all comparisons P < 0.05 was deemed significant.


Biochemistry in ICU and Control Patients.

Baseline characteristics of ICU (n = 130) and control (n = 20) patients are described in Table 1. The total ICU population, as well as the subset used for immunohistochemical analysis, was matched with control patients for gender, age, and body mass index (Supporting Data Table 1).

Table 1. Baseline Characteristics of Control and ICU Patients
 Control (n=20)ICU (n=130)P-value
  1. Baseline characteristics for all control and ICU patients. Represented P- values are calculated for the comparison between ICU and control patients. All data are represented as mean ± SEM or median with IQR (25th-75th percentiles) as appropriate.

  2. Abbreviations: ICU, intensive care unit; BMI, body mass index; LOS, length of stay; APACHE, acute physiology and chronic health assessment evaluation; CRP, C-reactive protein; ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma-glutamyltranspeptidase; ALP, alkaline phosphatase; TNFα, tumor necrosis factor alpha; IL, interleukin.

Gender (% male)65640.9
Age (years)68 ± 368 ± 10.7
BMI (kg/m2)25.1 ± 0.624.9 ± 0.40.4
LOS ICU (days) 10 (6-21) 
APACHE II (score) 19 (12-27) 
Diagnostic Group (n,%)   
 Cardiovascular disease/high-risk cardiac or complicated vascular surgery 33 (25) 
 Respiratory disease/complicated pulmonary or esophageal surgery 42 (32) 
 Gastrointestinal or hepatic disease/complicated abdominal surgery 16 (12) 
 Neurology/neurosurgery 14 (11) 
 Hematology/oncology 9 (7) 
 Solid organ transplant 1 (<1) 
 Polytrauma 3 (2) 
 Renal/metabolic 2 (2) 
 Other 10 (8) 
Sepsis (n, %) 65 (50) 
Serum markers on admission   
 CRP (mg/L) 131 (61-192) 
 ALT (IU/L) 29 (15-59) 
 AST (IU/L) 48 (27-95) 
 Total Bilirubin (mg/dL) 1.18 (0.63-2.41) 
 GGT (IU/L) 40 (25-78) 
 ALP (IU/L) 175 (125-258) 
Serum markers on day of biopsy  
 CRP (mg/L)7 (4-28)150 (88-219)<0.0001
  TNFα (pg/mL) 5160 (8-23825) 
  IL-1β (pg/mL) 6750 (18-19963) 
  IL-6 (pg/mL) 92910 (535-269918) 
 ALT (IU/L)15 (14-18)46 (26-125)0.0002
 AST (IU/L)18 (15-21)62 (36-157)<0.0001
 Total Bilirubin (mg/dL)0.37 (0.29-0.46)2.89 (1.08-8.87)<0.0001
 GGT (IU/L)25 (20-46)75 (41-204)0.004
 ALP (IU/L)220 (180-242)350 (209-689)0.02
 Total bile acids (μM)0.62 (0.42-0.92)6.88 (3.12-17.71)<0.0001

Serum total bilirubin on the last day of ICU stay was 8-fold higher in ICU patients than in controls (Table 1) and the hyperbilirubinemia was predominantly conjugated. Compared with controls, serum ALP and GGT levels in ICU patients were 1.6- and 3-fold higher, respectively (Table 1). In parallel, serum total BAs were 11-fold higher (P < 0.0001) in ICU patients (Table 1), this increase being mainly attributable to conjugated BAs (Table 2). There was no effect of tight glycemic control on circulating bilirubin or BA levels. There was an increase in conjugation percentage for the primary BA cholic acid (CA) (98.3% in patients versus 55.6% in controls) and chenodeoxycholic acid (CDCA) (95.9% in patients versus 37.5% in controls) (P < 0.0001). Serum levels of glycocholic acid (G-CA) were on average 83-fold increased in ICU versus control patients, whereas glycochenodeoxycholic acid (G-CDCA) was 34-fold higher in the critically ill population. Taurocholic acid (T-CA) was 22-fold and taurochenodeoxycholic acid (T-CDCA) was 39-fold increased in critical illness. Serum levels of the unconjugated BAs CA, CDCA, and deoxycholic acid (DCA) did not differ between the two populations. The ratio of unconjugated CA/CDCA (0.5 in patients versus 0.3 in controls, P = 0.003) as well glycoconjugated CA/CDCA (1.1 in patients versus 0.4 in controls, P < 0.0001) was higher in critically ill patients. After logarithmic transformation, serum levels of total bilirubin correlated strongly with G-CA, G-CDCA, T-CA, and T-CDCA on the day of biopsy, as shown in Fig. 1. Changes in serum markers of cholestasis and bilirubinostasis in the subset of ICU patients used for immunohistochemical analysis were similar to those seen in the entire ICU population used for messenger RNA (mRNA) analysis (data not shown).

Table 2. Serum Levels of Bile Acids for Control and ICU Patients
Bile Acid (μM)Control (n=20)ICU (n=130)P-value
  • Serum levels of bile acids for control and ICU patients on day of biopsy. Levels are expressed in μM and are represented as median with IQR (25th-75th percentiles).

  • *

    More than 60% of the samples were below the detection limit.

  • Abbreviations: CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; LCA, lithocholic acid; UDCA, ursodeoxycholic acid; G-CA, glycocholic acid; G-CDCA, glycochenodeoxycholic acid; G-DCA, glycodeoxycholic acid; G-UDCA, glycoursodeoxycholic acid; T-CA, taurocholic acid; T-CDCA, taurochenodeoxycholic acid; T-DCA, taurodeoxycholic acid; T-UDCA, tauroursodeoxycholic acid; T-LCA, taurolithocholic acid.

CA0.04 (0.01-0.05)0.05 (0.02-0.15)0.06
CDCA0.10 (0.09-0.15)0.09 (0.05-0.18)0.5
DCA0.05 (0.05-0.08)0.05 (0.05-0.13)0.4
LCA*0.04 (0.04-0.04)0.04 (0.04-0.04)0.2
UDCA*0.01 (0.01-0.07)0.01 (0.01-0.01)0.2
G-CA0.03 (0.03-0.09)2.49 (0.68-7.25)<0.0001
G-CDCA0.05 (0.05-0.34)1.70 (0.82-5.82)<0.0001
G-DCA*0.01 (0.01-0.01)0.01 (0.01-0.14)0.05
G-UDCA*0.02 (0.02-0.02)0.02 (0.02-0.13)0.1
T-CA0.02 (0.02-0.02)0.43 (0.13-1.31)<0.0001
T-CDCA0.01 (0.01-0.01)0.39 (0.01-1.16<0.0001
T-DCA*0.03 (0.03-0.03)0.03 (0.03-0.03)0.1
T-UDCA*0.01 (0.01-0.01)0.01 (0.01-0.01)0.7
T-LCA*0.01 (0.01-0.01)0.01 (0.01-0.01)0.9
Figure 1.

Correlation between serum levels of BAs and total bilirubin. Correlation between serum levels of conjugated primary BAs G-CA, G-CDCA, T-CA, and T-CDCA and total bilirubin for ICU and control patients on day of biopsy. Open dots represent values below detection limit. Abbreviations: G-CA, glycocholic acid; G-CDCA, glycochenodeoxycholic acid; T-CA, taurocholic acid; T-CDCA, taurochenodeoxycholic acid.

Serum levels for tumor necrosis factor alpha (TNFα), interleukin (IL)-1β, IL-6 are shown in Table 1. In over 80% of the ICU patients levels of IFNγ, IL-2, IL-4, and IL-5 were undetectable, whereas in the control patients all measured cytokines were below the assay detection limits.

Histological Characteristics of Cholestasis in ICU and Control Patients.

Liver histology and immunohistochemical staining were performed in a random subset of 40 ICU patients and 10 controls (Table 3). The majority of the ICU biopsies exhibited typical histological features of intrahepatic cholestasis (Fig. 2). In 82% of the liver biopsies from the ICU patients hepatocellular and canalicular bilirubinostasis was present, whereas in 34% ductular bilirubinostasis also was present. This was absent in the control biopsies. A mild ductular reaction was seen in 20% of controls compared with ICU biopsies that showed a mild (37%) to severe (47%) ductular reaction. In 42% of ICU patients signs of cholangiolitis were observed. In contrast, the presence of portal inflammation did not differ between ICU patients and controls.

Table 3. General Histology, Immunohistochemistry of Hepatobiliary Transporters, and Nuclear Receptors of Liver Sections of Control and ICU Patients
IHC ScoreControl (n=10)ICU (n=40)P-value
  1. Histological assessment (for bilirubinostasis, ductular reaction, and portal inflammation), immunohistochemical assessment of the expression of hepatobiliary transporters and nuclear receptors of liver sections of 40 ICU and 10 control patients. Data are represented as numbers and percentages (in parentheses). Represented P-values are calculated for the comparison ICU versus control patients.

  2. Abbreviations: IHC, immunohistochemistry; NTCP, Na+/taurocholate cotransporting polypeptide; OATP, organic anion transporting polypeptide; MRP, multidrug resistance-associated protein; BSEP, bile salt export pump; MDR, multidrug resistance protein; CAR, constitutive androstane receptor; FXR, farnesoid X receptor; VDR, vitamin D receptor; RXRα, retinoid X receptor alpha; PXR, pregnane X receptor.

GeneralBilirubinostasis  <0.0001
 010 (100)7 (18) 
 10 (0)17 (45) 
 20 (0)14 (37) 
Ductular reaction  0.0009
 08 (80)6 (16) 
 12 (20)14 (37) 
 20 (0)11 (29) 
 30 (0)7 (18) 
Portal inflammation  0.6
 04 (40)11 (29) 
 16 (60)24 (63) 
 20 (0)3 (8) 
Hepatobiliary transportersNTCP  0.7
 -20 (0)4(11) 
 -13 (33)11 (29) 
 03 (33)8 (22) 
 13 (33)11 (30) 
 20 (0)3 (8) 
OATP2/8  0.07
 -10 (0)5 (15) 
 07 (78)11 (32) 
 12 (22)11 (32) 
 20 (0)7 (21) 
MRP3  <0.0001
 06 (67)3 (8) 
 13 (33)9 (24) 
 20 (0)26 (68) 
BSEP  0.02
 -20 (0)11 (28) 
 -12 (22)16 (42) 
 07 (78)11 (28) 
MRP2  0.02
 08 (80)13 (34) 
 12 (20)10 (26) 
 20 (0)15 (40) 
MDR1  0.05
 09 (90)18 (47) 
 11 (10)13 (34) 
 20 (0)7 (18) 
MDR3  0.0002
 09 (90)8 (21) 
 11 (10)14 (37) 
 20 (0)16 (42) 
Nuclear receptorsCAR  <0.0001
 00 (0)3 (8) 
 10 (0)18 (47) 
 21 (11)12 (32) 
 38 (89)5 (13) 
VDR  0.2
 01 (11)13 (33) 
 16 (67)13 (33) 
 22 (22)14 (35) 
FXR  0.04
 01 (10)7 (18) 
 12 (20)12 (31) 
 22 (20)16 (41) 
 35 (50)4 (10) 
RXRα  0.0004
 01 (10)3 (8) 
 10 (0)14 (36) 
 20 (0)14 (36) 
 39 (90)8 (21) 
PXR  0.01
 01 (10)14 (35) 
 10 (0)12 (30) 
 24 (40)9 (23) 
 35 (50)5 (13) 
Figure 2.

Representative liver sections for bilirubinostasis and ductular reaction. Left panel: Control patients. Right panel: ICU patients. (A) Control patient with normal liver tissue with normal portal tract (PT) and no signs of cholestasis (left panel). Extensive cholestasis with hepatocellular (arrowhead), canalicular (short arrow), and ductular (long arrow) bilirubinostasis. PT with a dilated ductulus filled with a bile plug (right panel + small frame in the left lower corner). (B) Normal CK7 staining of intralobular bile duct (left panel). Increased CK7 staining with ductular proliferation at the interface of the portal tract and the liver parenchyma (right panel). Abbreviations: H&E, hematoxylin and eosin; CK, cytokeratin; ICU, intensive care unit; PT, portal tract; CV, centrolobular vene.

Morphological signs of cholestasis were linked with biochemical markers of cholestasis measured on the day of the biopsy. The degree of bilirubinostasis correlated with serum levels of total bilirubin (ρ = 0.816, P < 0.0001), ALP (ρ = 0.472, P = 0.008), GGT (ρ = 0.495, P = 0.008), G-CA (ρ = 0.775, P < 0.0001), G-CDCA (ρ = 0.726, P < 0.0001), T-CA (ρ = 0.739, P < 0.0001), and T-CDCA (ρ = 0.566, P = 0.0001). The presence of ductular reaction also correlated with the serum levels of total bilirubin (ρ = 0.709, P < 0.0001), ALP (ρ = 0.539, P = 0.002), GGT (ρ = 0.483, P = 0.009), G-CA (ρ = 0.591, P < 0.0001), G-CDCA (ρ = 0.598, P < 0.0001), T-CA (ρ = 0.696, P < 0.0001), and T-CDCA (ρ = 0.658, P < 0.0001).

BA Synthesis Enzymes in Prolonged Critically Ill Patients.

Hepatic mRNA expression of CYP7A1, the rate-limiting step in BA synthesis, was decreased by 94% in ICU patients compared with controls (P < 0.0001), but CYP7A1 protein expression did not differ between the two groups. However, within the ICU group an inverse correlation between CYP7A1 protein and the serum levels of total BAs was observed. (ρ = −0.347, P = 0.0001). In contrast, mRNA expression of CYP8B1, an enzyme involved in the synthesis of CA, was increased by 240% (P < 0.0001).

Bile Salt Transporters in Prolonged Critically Ill Patients.

In ICU patients, mRNA expression of the basolateral uptake transporters NTCP, OATP2, and OATP8 was down-regulated compared with controls (Fig. 3), but NTCP immunohistochemical staining did not differ between groups (Table 3). OATP2/8 staining had a clear intensity gradient from centrolobular to periportal regions in control patients. In 11/34 ICU patients a more uniform and extended staining with gradient fading was observed (Table 3).

Figure 3.

mRNA levels of hepatobiliary transporters and nuclear receptors in control and ICU patients. Upper panel: mRNA levels of hepatic basolateral influx pumps (NTCP, OATP2, OATP8), basolateral efflux transporters (MRP3, MRP4), and canalicular efflux pumps (BSEP, MRP2, MDR1, MDR3) of 130 ICU patients. Lower panel: mRNA levels of hepatic nuclear receptors (FXR, VDR, CAR, PXR, RXRA, and SHP) of 130 ICU patients. mRNA levels are expressed relative to the mRNA expression of the housekeeping gene HPRT and relative to 20 control patients. Data are represented as median with IQR (25th-75th percentiles). Abbreviations: NTCP, Na+/taurocholate cotransporting polypeptide; OATP, organic anion transporting polypeptide; MRP, multidrug resistance-associated protein; BSEP, bile salt export pump; MDR, multidrug resistance protein; HPRT, hypoxanthine phosphoribosyltransferase; FXR, farnesoid X receptor; VDR, vitamin D receptor; CAR, constitutive androstane receptor; PXR, pregnane X receptor; RXRA, retinoid X receptor alpha; SHP, short heterodimeric partner.

In contrast, mRNA levels of MRP3 and MRP4, the basolateral efflux transporters, were strongly up-regulated. Immunohistochemistry confirmed a marked up-regulation of MRP3 staining in ICU patients compared with control subjects (P < 0.0001) (Table 3). Moreover, whereas controls only exhibited basolateral MRP3 staining in the centrolobular zone of the liver lobule, ICU patients showed a strong panlobular honeycomb staining pattern (Fig. 4). For MRP3, mRNA and protein levels were in agreement (ρ = 0.432, P = 0.004). Moreover, MRP3 expression correlated positively with the degree of bilirubinostasis both at the mRNA level (ρ = 0.529, P = 0.0003) and protein level (ρ = 0.591, P < 0.0001). There was also a strong correlation between the MRP3 protein levels and biochemical markers of cholestatic liver dysfunction, i.e., the serum levels of total bilirubin (ρ = 0.625, P = 0.0003), GGT (ρ = 0.519, P = 0.005), ALP (ρ = 0.551, P = 0.002), G-CA (ρ = 0.494, P = 0.0008), and G-CDCA (ρ = 0.484, P = 0.001). Due to technical limitations, we were not able to stain for MRP4.

Figure 4.

Representative liver sections for MRP3, MDR3. Left panel: Control patients. Right panel: ICU patients. (A) Normal basolateral MRP3 staining showing clear centrolobular and midzonal activity (left panel). Markedly up-regulated panlobular honeycomb MRP3 staining pattern (right panel). (B) Normal pattern of a fine canalicular linear MDR3 staining (left panel). Strong double-stranded pattern of MDR3 staining around multiple dilated canaliculi (right panel). Abbreviations: MRP, multidrug resistance-associated protein; MDR, multidrug resistance protein; CV, centrolobular vene; PT, portal tract.

mRNA expression of the canalicular efflux pumps BSEP, MRP2, MDR1, and MDR3 was significantly higher in ICU patients compared with control patients (Fig. 3). In contrast to the increased mRNA expression, protein expression of BSEP was down-regulated (Table 3). The normal regular BSEP immunohistochemical staining pattern became irregular and discontinuity was observed in cholestatic zones. Severely cholestatic areas had no discernable immunostaining. In concert with the mRNA expression, MRP2 immunostaining was up-regulated in ICU patients in comparison with controls (P = 0.02) and correlated well with the degree of bilirubinostasis (ρ = 0.512, P = 0.0004), ductular reaction (ρ = 0.433, P = 0.003), and the serum levels of total bilirubin (ρ = 0.502, P = 0.003). MDR1 and MDR3 staining was also up-regulated. At the canalicular domain of the hepatocytes, a fine linear MDR3 pattern, seen in control subjects, evolved towards a strong double strand pattern of staining around multiple dilated canaliculi in ICU patients (Fig. 4). This is indicative of a very strong up-regulation of MDR3 protein. Similar to MRP3, MDR3 protein levels correlated with the degree of bilirubinostasis seen on the liver sections (ρ = 0.569, P < 0.0001) and serum levels of total bilirubin (ρ = 0.745, P < 0.0001), GGT (ρ = 0.402, P = 0.03), ALP (ρ = 0.437, P = 0.01), G-CA (ρ = 0.639, P < 0.0001), and G-CDCA (ρ = 0.548, P = 0.0002) MDR1 protein staining showed a similar up-regulation as MDR3 staining (P = 0.05).

Nuclear Receptors in Prolonged Critically Ill Patients.

In ICU patients, mRNA expression of the NRs FXR, VDR, PXR, and RXRα was up-regulated in comparison with control subjects. mRNA expression of CAR and SHP did not differ between groups. (Fig. 3). In contrast to the increased mRNA expression, FXR, PXR, and RXRα immunostaining in the nuclei was effectively absent in ICU patients but clearly visible in controls (Table 3, Fig. 5). VDR protein expression did not differ between ICU and control patients.

Figure 5.

Representative liver sections for FXR, CAR, and RXRα. Left panel: Control patients. Right panel: ICU patients. Normal liver with nuclear FXR (A), CAR (B), and RXRα (C) immunostaining clearly present in controls (left panel). Strongly decreased FXR (A), CAR (B), and RXRα (C) nuclear immunostaining in ICU patients (right panel). Abbreviations: FXR, farnesoid X receptor; CAR, constitutive androstane receptor; RXRα, retinoid X receptor alpha; PT, portal tract.

Nuclear CAR staining was clearly decreased in ICU patients. Control subjects showed both cytoplasmic and intense nuclear staining, with a clear intensity gradient from periportal to centrolobular regions, whereas ICU patients only showed discrete positive cytoplasmic staining and a marked reduction in nuclear staining (Fig. 5).

Overall there was no correlation between mRNA and protein levels for all NRs. In contrast, nuclear staining correlated inversely with histological and biochemical cholestatic parameters. For example, patients with the lowest levels of nuclear CAR and RXRα staining demonstrated the most severe bilirubinostasis. Serum levels of total bilirubin on the day of biopsy inversely correlated with the nuclear immunolocalization of CAR (ρ = −0.589, P < 0.0006), FXR (ρ = −0.416, P < 0.01), and RXRα (ρ = −0.553, P < 0.001). RXRα staining also correlated well with BSEP apical protein visualization (ρ = 0.581, P < 0.0001).


This study of postmortem liver biopsies in conjunction with pre-agonal serum analyses found that BA levels are much more increased during critical illness than the bilirubin concentrations. Critical illness was also associated with maintained CYP7A1 levels, decreased apical BSEP protein, increased basolateral MRP3 protein expression. Nuclear localization of FXR and its heterodimeric partner RXRα was diminished in critically ill patients.

Although bilirubin levels increased 8-fold during critical illness, the larger increase in circulating total BAs mainly consisted of glycine and taurine conjugates of CA and CDCA. Unconjugated CA and CDCA did not differ from controls. This indicates that the hepatocytes are able to conjugate potentially toxic BAs, either de novo synthesized or enterohepatically recirculated. It also suggests that the transport of the conjugated BA toward the apical bile canaliculi is strongly shifted to the blood. The ratio of CA to CDCA was also increased in critically ill patients, consistent with the increased expression of hepatic CYP8B1 mRNA. This shift may represent a reduction in FXR-mediated FGF19 production by small bowel enterocytes due to reduced BAs being excreted into the gut, as FGF19 has recently been recognized to repress CYP8B1 in mice.16

Despite the strongly elevated serum BA levels during critical illness, CYP7A1, the rate-limiting step in de novo BA synthesis, was only repressed at the mRNA level but not at the protein level. This is in line with the absence of increased SHP mRNA expression in ICU patients, which mediates BA repression of CYP7A1.17 Furthermore, FXR and its heterodimeric partner RXRα, which act in concert with SHP to suppress BA synthesis enzymes, were absent from the hepatocytic nucleus, where they exert transcriptional activity through direct binding to DNA. This may imply an at least partial loss of the sensing of BA and its feedback regulation of de novo BA production, in light of the increased circulating BAs in ICU patients. Alternatively, critical illness may induce elevated BA levels by suppressing the BA sensor FXR and maintaining (CYP7A1) and/or shifting (CYP8B1) BA synthesis. Cytoplasmic retention of RXR has also been found in models of acute liver inflammation18, 19 and advanced extrahepatic cancer.20 In the present study other NRs relevant to BA regulation, namely, PXR and CAR, also did not localize to the nucleus. The lower nuclear levels of PXR and CAR may not only affect bile formation, but also metabolic processes in the liver, such as energy homeostasis.21

BAs, and bilirubin, are transported by the hepatocyte by way of the hepatobiliary transporters. In this study the most prominent changes in the expression profile of the hepatic BA transporters during prolonged critical illness were observed in the basolateral efflux transporters MRP3 and MRP4. Normally, MRP3 and MRP4 are expressed at very low levels in hepatocytes, but they become up-regulated by inflammation and during long-standing cholestasis, presumably shifting transport of BAs back into sinusoidal blood for elimination by the kidneys.7 Immunohistochemical expression of BSEP in the hepatocyte canalicular domain was dramatically reduced in ICU patients, especially in regions of bilirubinostasis, despite an increase in BSEP mRNA expression. Decreased expression of BSEP is a major contributor22 to the cholestatic phenotype of the prolonged critically ill patient, as BAs will accumulate within the hepatocytes. In contrast to findings from chronic cholestatic disorders7 and animal models of cholestasis23 and sepsis,24 MRP2, the main canalicular bilirubin transporter, was up-regulated during critical illness. This seems difficult to reconcile with the elevated serum bilirubin levels. Nevertheless, it may fit with the rather moderate increase in serum bilirubin, compared with the changes in serum BA concentrations. Besides, bile formation is a secretory process that depends on osmotically active solutes, mainly BAs. If the bile flow is hampered as a consequence of retained BAs, bilirubin will also be retained, essentially as a biochemical epiphenomenon.

The data on changes of BA synthesis and disposition and their regulation by the NR do not allow us to state whether they are beneficial or a failing compensatory response. Data from hepatocyte-RXRα-null mice indicate that these mice are protected against WY-14,643-induced liver injury by the up-regulation of Mrp3 expression and increased efflux of BAs into blood for renal excretion.25 FXR knockout mice have a lower mortality rate and less liver injury during bile duct ligation. These FXR knockout mice strongly increased Mrp4 and reduced Bsep expression.26 However, FXR knockout mice exhibit more hepatotoxicity when challenged with a CA-enriched diet.27 Also, FXR agonists could be beneficial for patients with cholestatic liver diseases.28 CAR knockout mice show lower levels of serum and liver primary BAs than wildtype mice during bile duct ligation.29 Moreover, these CAR knockout mice are resistant to acetaminophen liver toxicity.30 Similarly, an increased bilirubin clearance has been demonstrated in PXR knockout mice.31

Histopathology of ICU patient liver biopsies revealed classic changes of cholestasis, namely, bilirubinostasis, ductular proliferation, and variable inflammation. Increased levels of serum bilirubin and conjugated BAs correlated strongly with the microscopic signs of bilirubinostasis and ductular reaction. Ductular proliferation and ductular differentiation of the hepatocytes are considered part of an adaptive, protective response to cholestasis. Canalicular MDR3 was also up-regulated in ICU patients. Given the key role of biliary phospholipids in protecting bile duct epithelium from the potentially toxic biliary content, up-regulation of MDR3 might also exert a compensatory action, protecting the canalicular membrane and biliary epithelium. Because MRP3 correlated well with histological bilirubinostasis and serum bilirubin and conjugated BAs levels, MRP3 up-regulation is a likely compensatory reaction to cholestasis, as has been observed in animal bile duct ligation models of cholestasis.32 The up-regulation of MRP3 (and MRP4) provides a mechanism to limit hepatocellular retention of hydrophobic BAs and other potentially toxic compounds that would normally be destined for biliary excretion. This is in keeping with the selective increase in serum taurine and glycine conjugated BAs, which have been conjugated by hepatocytes and transported back into the circulation. MRP3 up-regulation has also been shown in acute sepsis models without longer-lasting cholestasis.33 Unexpectedly, there was a lack of association between ICU cholestasis and markers of inflammation, suggesting that inflammation is not the main contributor to cholestasis in prolonged critical ill patients, as it does in acute sepsis or septic shock.8

A limitation of this study is reliance on liver biopsy samples taken immediately postmortem, which is an inherent confounder. However, ethically it was not possible to obtain study-programmed liver samples in unselected critically ill patients. Therefore, findings at the tissue level were always interpreted in the context of serum marker changes from the pre-agonal phase.

In summary, critical illness is associated with a strong increase in serum BA levels. Maintenance of BA synthesis, suppression of FXR/RXRα, with lowering of apical BSEP and elevated basolateral MRP3 expression may either be a desired response during critical illness to raise serum BA concentrations or it may be a failing feedback regulation on BA formation and disposition, caused by cholestasis, i.e., increased serum bilirubin and BA.