Toll-like receptor 4–dependent Kupffer cell activation and liver injury in a novel mouse model of parenteral nutrition and intestinal injury


  • Karim C. El Kasmi,

    1. Section of Pediatric Gastroenterology, Hepatology, and Nutrition, University of Colorado School of Medicine, Aurora, CO
    2. Digestive Health InstituteAurora, CO
    Search for more papers by this author
  • Aimee L. Anderson,

    1. Section of Pediatric Gastroenterology, Hepatology, and Nutrition, University of Colorado School of Medicine, Aurora, CO
    2. Digestive Health InstituteAurora, CO
    Search for more papers by this author
  • Michael W. Devereaux,

    1. Section of Pediatric Gastroenterology, Hepatology, and Nutrition, University of Colorado School of Medicine, Aurora, CO
    2. Digestive Health InstituteAurora, CO
    Search for more papers by this author
  • Sophie A. Fillon,

    1. Section of Pediatric Gastroenterology, Hepatology, and Nutrition, University of Colorado School of Medicine, Aurora, CO
    2. Digestive Health InstituteAurora, CO
    Search for more papers by this author
  • J. Kirk Harris,

    1. Section of Pulmonary Medicine, Department of Pediatrics, University of Colorado School of Medicine, Aurora, CO
    2. Breathing Institute, Aurora, CO
    Search for more papers by this author
  • Mark A. Lovell,

    1. Department of Pathology, University of Colorado School of Medicine, Aurora, CO
    2. Children's Hospital Colorado, Aurora, CO
    Search for more papers by this author
  • Milton J. Finegold,

    1. Department of Pathology, Texas Children's Hospital and Baylor College of Medicine, Houston, TX
    Search for more papers by this author
  • Ronald J. Sokol

    Corresponding author
    1. Section of Pediatric Gastroenterology, Hepatology, and Nutrition, University of Colorado School of Medicine, Aurora, CO
    2. Digestive Health InstituteAurora, CO
    3. Children's Hospital Colorado, Aurora, CO
    • Colorado Clinical and Translational Sciences Institute, Children's Hospital Colorado, Box B290, 13123 E. 16th Ave., Aurora, CO 80045
    Search for more papers by this author
    • fax: 720-777-7277

  • Potential conflict of interest: Nothing to report.

  • The contents of this article are the authors' sole responsibility and do not necessarily represent the official views of the National Institutes of Health.


Infants with intestinal failure who are parenteral nutrition (PN)-dependent may develop cholestatic liver injury and cirrhosis (PN-associated liver injury: PNALI). The pathogenesis of PNALI remains incompletely understood. We hypothesized that intestinal injury with increased intestinal permeability combined with administration of PN promotes lipopolysaccharide (LPS)–Toll-like receptor 4 (TLR4) signaling dependent Kupffer cell (KC) activation as an early event in the pathogenesis of PNALI. We developed a mouse model in which intestinal injury and increased permeability were induced by oral treatment for 4 days with dextran sulphate sodium (DSS) followed by continuous infusion of soy lipid-based PN solution through a central venous catheter for 7 (PN7d/DSS) and 28 (PN28d/DSS) days. Purified KCs were probed for transcription of proinflammatory cytokines. PN7d/DSS mice showed increased intestinal permeability and elevated portal vein LPS levels, evidence of hepatocyte injury and cholestasis (serum aspartate aminotransferase, alanine aminotransferase, bile acids, total bilirubin), and increased KC expression of interleukin-6 (Il6), tumor necrosis factor α (Tnfα), and transforming growth factor β (Tgfβ). Markers of liver injury remained elevated in PN28d/DSS mice associated with lobular inflammation, hepatocyte apoptosis, peliosis, and KC hypertrophy and hyperplasia. PN infusion without DSS pretreatment or DSS pretreatment alone did not result in liver injury or KC activation, even though portal vein LPS levels were elevated. Suppression of the intestinal microbiota with broad spectrum antibiotics or ablation of TLR4 signaling in Tlr4 mutant mice resulted in significantly reduced KC activation and markedly attenuated liver injury in PN7d/DSS mice. Conclusion: These data suggest that intestinal-derived LPS activates KC through TLR4 signaling in early stages of PNALI. (HEPATOLOGY 2012)

Since its first clinical use in the 1960s, parenteral nutrition (PN) has significantly improved the survival of infants who are unable to tolerate enteral feedings, especially those with intestinal failure caused by necrotizing enterocolitis, short bowel syndrome, intestinal atresias, and other gastrointestinal malformations.1 However, a significant complication in PN-infused infants is the development of cholestatic PN-associated liver injury (PNALI) that can rapidly progress to cirrhosis.2 Multivisceral transplantation then becomes necessary for survival in many of these infants.3 To date, the pathogenesis of PNALI remains poorly understood, which has been a critical barrier for the design of more effective treatment and prevention methods. One of the major obstacles to advances in this field has been the lack of an animal model that allows investigation of the early events that trigger liver injury. The readily available spectrum of transgenic mice makes a mouse model of PNALI highly desirable in order to study cellular, molecular, and immunological mechanisms that may initiate liver injury.

The clinical observation that severity and chronicity of PNALI are increased in those PN-dependent infants with underlying intestinal inflammation4 may provide important clues to a better understanding of the pathogenesis of PNALI. For example, the lack of enteral feedings in PN-infused infants may significantly reduce intestinal motility, favoring bacterial overgrowth,4-6 alterations in mucosal immunity, and subsequent aggravation of underlying inflammation.6 In addition, the remaining intestine after surgical resection may have impaired ileal bile acid transport, lack the barrier function of the ileal-cecal valve, be inflamed by the underlying disease process (e.g., necrotizing enterocolitis), or may have nonspecific inflammation (e.g., jejunitis and short gut colitis7) or disturbed vascular perfusion. Together with the presumed a priori increased intestinal permeability of infants,8 this may result in a significant compromise of the intestinal barrier function.9 As a result, intestinal-derived Toll-like receptor (TLR) agonists such as bacterial proteins, lipids, or nucleic acids may be absorbed into the portal circulation in large amounts. Activation of TLR signaling in Kupffer cells (KCs) may subsequently initiate inflammatory pathways that promote hepatocyte injury, cholestasis, apoptosis, and necrosis, as well as activation of stellate cells.10 Intriguingly, KC hyperplasia and inflammation are characteristic of the liver histopathology in human infants and adults with PNALI.11 Recent work in mouse models of cholestatic,12, 13 alcoholic,13 and nonalcoholic14 liver injury has implied a dysfunctional intestinal barrier, increased absorption of gut-derived TLR agonists—specifically lipopolysaccharide (LPS)—and subsequent TLR-dependent activation of KCs as disease-initiating and -promoting mechanisms (referred to as the gut-liver axis).

Based on these observations, we hypothesized that intestinal injury and increased intestinal permeability combined with administration of PN promotes TLR4-dependent KC activation as an early event in the pathogenesis of PNALI. In order to test this hypothesis, we designed a mouse model that specifically replicates the initial events that occur in human infants at greatest risk for the development of PNALI, namely the combination of PN infusion and intestinal injury/increased intestinal permeability. Using this mouse model we investigated the role of KC activation by the intestinal microbiota through TLR4 signaling during the early stages of liver injury in PNALI.


ALT, alanine aminotransferase; ANOVA, analysis of variance; AST, aspartate aminotransferase; cDNA, complementary DNA; CVC, central venous catheter; DSS, dextran sulphate sodium; FITC, fluorescein isothiocyanate; H&E, hematoxylin and eosin; Il1β, interleukin-1β; Il6, interleukin-6; KC, Kupffer cell; LPS, lipopolysaccharide; MPO, myeloperoxidase; mRNA, messenger RNA; NS, normal saline; PCR, polymerase chain reaction; PN, parenteral nutrition; PNALI, PN-associated liver injury; qRT-PCR, quantitative reverse-transcription PCR; Tgfβ, transforming growth factor beta; TLR, Toll-like receptor; Tnfα, tumor necrosis factor alpha; TSBA, total serum bile acids.

Materials and Methods

Mouse Model.

C57BL/6 wild-type and Tlr4 mutant (B6.B10ScN-Tlr4lps-del/JthJ15) adult male mice (8-10 weeks old; The Jackson Laboratory, Bar Harbor, ME) were exposed ad lib to 2.5% dextran sulphate sodium (DSS) in drinking water for 4 days. Mice then received regular drinking water for 24 hours (referred to as DSS pretreatment) before placement of a central venous catheter (CVC) (Silastic tubing, 0.012 inches internal diameter; Dow Corning) into the right jugular vein. The proximal end of the CVC was tunneled subcutaneously, exited between the shoulder blades, and connected to an infusion pump (Harvard Apparatus, Holliston, MA). Mice were placed in a rubber harness (Instech Laboratories, Plymouth Meeting, PA) and recovered from surgery for 24 hours with intravenous normal saline (NS) infusion at a rate of 0.23 mL/hour and given ad lib access to chow and water. After 24 hours, mice were continuously infused for 7 or 28 days with PN (PN7d/DSS and PN28d/DSS, respectively) at a rate of 0.29 mL/hour providing a caloric intake of 8.4 kcal/24 hours (Table 1). All PN-infused mice had access to water ad lib but not to chow. Several control groups were also studied. DSS pretreated mice were infused with NS for 7 days instead of PN (NS/DSS). DSS pretreated mice without CVC placement were given free access to chow and water for either 1 day (DSS mice) or 8 days (DSS+8d chow). Other mice were not DSS pretreated but received PN or NS, respectively, in the same manner as PN/DSS and NS/DSS mice (PN and NS). Unmanipulated mice (chow mice) had free access to chow and water for a period of 12 days. All mice were individually housed in metabolic cages. DSS pretreatment did not cause weight loss; however, mice lost a mean of 1.45 g and 2.6 g during the 7-day and 28-day periods of PN/DSS administration, respectively (Table 2). Blood was collected from the retro-orbital plexus. Serum was analyzed by the University of Colorado Hospital Clinical Chemistry Laboratory for aspartate aminotransferase (AST), alanine aminotransferase (ALT), and total bilirubin levels. Total serum bile acids (TSBA) were analyzed using a total bile acid detection kit (Diazyme Laboratories, Poway, CA). All animals were treated humanely, and all animal procedures were approved by the Institutional Animal Care and Use Committee.

Table 1. Parenteral Nutrition Solution Components Used in Experiments per 100 mL
  • *

    Obtained at Abbott Laboratories (Abbott Park, IL).

  • MTE-5 contains zinc, copper manganese, chromium, and selenium.

Aminosyn 15%*26.7 mL; 4 g
Dextrose36 mL; 25.5 g
Lipid10 mL; 2 g
Sodium phosphate1.34 mmol
Potassium chloride1.6 mEq
Sodium chloride3.2 mEq
Potassuim acetate12 mEq
Magnesium sulphate0.8 mEq
Calcium gluconate1.32 mEq
Multitrace-5 concentrate (MTE-5)0.1 mL
Heparin500 U
Nitrogen content0.61 g
Nonprotein calories105.68 kcal
Protein calories16.02 kcal
Dextrose calories85.68 kcal
Lipid calories20.00 kcal
Total calories121.7 kcal
Multivitamins2 mL
Table 2. Weight Change According to Treatment
GroupDSS PretreatmentPN or NS Infusion
Day 0Day 4Δ (g)*PDay 5Day 12 or 33Δ (g)*P
  • Abbreviations: n.a., not applicable for groups of mice that were not DSS pretreated; n.s., not significant as determined by P > 0.05.

  • Mice were weighed and DSS pretreatment was initiated on day 0. DSS was removed from drinking water on day 4 and mice were weighed again. On day 5, mice were weighed again, underwent central venous catheter placement, and were infused with NS for 24 hours. On day 6, PN or NS infusion was initiated for 7 days through day 12 or for 28 days through day 33. P values compare day 0 versus day 4 and day 5 versus day 12 or 33 by t test. Values presented are the mean ± SEM of body weight in grams (g). Note that DSS pretreatment did not result in significant weight loss. Note that PN administration was associated with a loss of 1.34 (PN) to 1.43 (PN/DSS7d) and 2.60 (PN/DSS28d) g of body weight, respectively. Weight loss in PN/DSS-treated TLR4 mutant mice was similar to weight loss in PN/DSS-treated wild-type mice. Treatment of PN/DSS mice with oral antibiotics resulted in reduced but still significant weight loss (−0.27 g).

  • *

    Change in body weight.

  • #

    Data at day 33, after 28d of PN.

NS/DSS (n = 27)22.17 (± 0.34)22.67 (± 0.37)+0.50n.s.22.30 (± 0.29)23.30 (± 0.27)+1.00<0.05
PN/DSS7d (n = 30)22.87 (± 0.36)22.93 (± 0.30)+0.06n.s.22.09 (± 0.21)20.64 (± 0.30)−1.45<0.05
PN/DSS28d (n = 5)23.06 (± 0.40)23.80 (± 0.37)+0.74n.s.23.80 (± 0.37)21.20 (± 0.45)#−2.60<0.05
PN/DSS7d+Abx (n = 11)22.55 (± 0.45)22.73 (± 0.38)+0.18n.s.22.45 (± 0.37)22.18 (± 0.46)−0.27<0.05
PN/DSS7d tlr4mut (n = 7)22.00 (± 0.71)22.20 (± 0.86)+0.20n.s.21.60 (± 1.20)20.40 (± 0.87)−1.20<0.05
PN (n = 12)n.a.n.a.n.a.n.s.23.42 (± 0.65)22.08 (± 0.35)−1.34<0.05
NS (n = 12)n.a.n.a.n.a.n.s.21.33 (± 0.41)22.00 (± 0.40)+0.67<0.05
Chow-fed (n = 12)n.a.n.a.n.a.n.s.23.33 (± 0.33)24.67 (± 0.33)+1.34<0.05

Intestinal Permeability and Portal Vein LPS Measurements.

Intestinal permeability was examined by oral gavage with 200 μL fluorescein isothiocyanate (FITC)-dextran (4,000 kD at 80 mg/mL; Sigma-Aldrich) on the day of sacrifice. Four hours after gavage, mice were anesthetized with intraperitoneal pentobarbital and blood was drawn from the portal vein (400 μL). Serum was prepared and analyzed for fluorescence measured at excitation wavelength of 494 nm and emission wavelength of 518 nm. LPS was measured in serum obtained from the portal vein using the Limulus Amebocyte Lysis Endpoint Assay (Lonza; Williamsport, PA) according to the manufacturer's instructions.

Suppression of Intestinal Microbiota.

A group of wild-type PN7d/DSS mice was exposed ad lib to an oral (drinking water) cocktail of four broad-spectrum antibiotics during the course of the 7-day PN infusion. The antibiotics, mixed fresh every 24 hours, included vancomycin (1 g/L), streptomycin (2 g/L), ampicillin (2 g/L), and metronidazole (2 g/L) based on the protocol of Fagarasan et al.16

Bacterial Load Assay.

Fecal samples were obtained from the descending colon on the day of sacrifice using sterile technique and were snap-frozen and stored at −70°. Total DNA was extracted, and bacterial 16S ribosomal DNA was polymerase chain reaction (PCR)-amplified using specific primers as described.17 Total bacterial ribosomal DNA gene copy numbers were measured in triplicate using the assay developed by Nadkarni et al.18 and copies per nanogram template DNA were calculated.

KC and Splenic Macrophage Purification.

Pooled livers (4-6 mice) were minced in cold Hanks buffer (Gibco/Invitrogen, Carlsbad, CA), incubated in Liberase (Hoffman La Roche, Germany) and DNAse (Sigma-Aldrich, St. Louis, MO) for 30 minutes at 37°C followed by centrifugation at 25g and separation over a 16% Histodenz gradient at 1,500g for 20 minutes. CD11b+ cells (referred to as KCs in this study) were then purified using positive selection with magnetic CD11b beads (MACS, Miltenyi, Auburn, CA). FACS-Calibur analysis demonstrated >95% purity using a CD11b-PE antibody (Miltenyi, Auburn, CA). Average yield was 500,000 KCs per 4-6 pooled livers from which an average of 600 ng complementary DNA (cDNA) was synthesized. Quantitative reverse-transcription PCR (qRT-PCR) was performed with 100 ng of cDNA/well in triplicate assays. KC and cDNA yields from Tlr4 mutant mice were up to 50% lower. Spleens were pooled (n = 3 each) from NS/DSS and PN7d/DSS mice, minced in cold Hanks buffer, centrifuged at 200g for 5 minutes followed by red cell lysis, and underwent subsequent positive selection (referred to as splenic macrophages in this study) using CD11b beads as described above.

RNA Isolation and Quantitative Gene Expression Analysis.

RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA), DNAse-treated (Ambion, Austin, TX) and reverse-transcribed with iScript (BioRad, Hercules, CA). Gene expression was analyzed by qRT-PCR on an Applied Biosystems 7300 cycler using commercially available TaqMan gene expression assays (Applied Biosystems). Data are expressed as normalized gene expression relative to chow mice or NS/DSS treated mice using the Δ/Δ Ct method. RNA was extracted, prepared, and analyzed from whole liver tissue as above.

Histological Analysis.

Liver tissue, terminal ileum, and colon were removed at sacrifice, formalin-fixed, paraffin-embedded and prepared for histology. Sections were stained with hematoxylin and eosin (H&E), Masson's trichrome, and sirius red according to standard staining protocols. Slides were examined by two pediatric pathologists (M. F. and M. L.) who were blinded to the treatment groups. KCs and activated neutrophils were visualized using antibodies specific for F4/80 (clone BM8; BMA Biomedicals, Switzerland) and myeloperoxidase (MPO; Abcam, Cambridge, MA) using standard immunohistochemistry protocols.

Statistical Analysis.

One-way analysis of variance (ANOVA) and Bonferroni's correction were used to determine statistical significance when more than two groups of mice were compared. The t test was used for comparison between two groups. P < 0.05 was considered statistically significant.


DSS Pretreatment Induces Intestinal Inflammation and Increased Intestinal Permeability.

We first determined whether the DSS pretreatment protocol, which uses a relatively short DSS treatment time (4 days) and low DSS dose (2.5%) compared with other DSS protocols,19 would induce intestinal inflammation and increased permeability without concomitant liver injury and KC activation. DSS pretreatment uniformly led to small amounts of visible blood in the fecal pellets starting on day 4 of DSS, which persisted for an additional 2-3 days, providing clinical evidence of intestinal inflammation. H&E and immunostains of terminal ileum and colon obtained after DSS pretreatment revealed focal mixed inflammation and focal infiltrates of F4/80-positive macrophages and MPO-positive neutrophils in submucosa of the colon (Supporting Fig. 1). Intestinal permeability was assessed in DSS mice (n = 10) and PN7d/DSS mice (n = 6) and compared with chow mice for absorption of FITC-dextran into the portal blood. Untreated chow mice that were not gavaged with FITC-dextran (n = 8) had virtually no serum fluorescence. DSS mice gavaged with FITC-dextran had 8-fold higher (P < 0.05) fluorescence in portal blood serum compared with FITC-dextran–gavaged chow mice (n = 5). Fluorescence in portal blood from PN7d/DSS mice was also 8-fold increased compared with chow mice (P < 0.05) (Fig. 1A).

Figure 1.

Increased intestinal permeability combined with PN infusion in mice induces liver injury and cholestasis. (A) DSS-pretreated (DSS; n = 10), PN7d/DSS (n = 6), and untreated control mice (chow; n = 5) were gavaged with FITC-dextran to determine intestinal permeability. Untreated control mice not gavaged with FITC-dextran are shown as a baseline control (chow; n = 8). Four hours later, fluorescence was measured in portal serum specimens. Intestinal permeability was increased after DSS pretreatment and remained elevated after PN7d/DSS treatment. *P < 0.05 DSS and PN7d/DSS versus chow; determined by one-way ANOVA. (B) Increased absorption of LPS in mice with increased intestinal permeability. Portal serum LPS levels [EU/L] were measured in untreated chow mice (chow; n = 8), DSS-pretreated mice (DSS; n = 8), DSS+8d chow mice (n = 8), PN mice (n = 3), and PN7d/DSS mice (n = 6). DSS treatment increased LPS absorption. PN treatment in the absence of DSS pretreatment (PN) also increased LPS absorption. LPS absorption was further significantly increased in PN7d/DSS mice compared with the other groups. *P < 0.05 (one-way ANOVA). #P < 0.05 PN versus chow and P > 0.05 PN versus DSS and DSS+8d chow (one-way ANOVA). (C-F) The combination of PN infusion with increased intestinal permeability induced hepatocyte injury (C,D) and cholestasis (E,F) in mice after 7 and 28 days of PN. *P < 0.05 PN/DSS versus controls (one-way ANOVA). Mean values and SEM are shown.

We next determined whether increased intestinal permeability was associated with absorption of LPS into the portal venous circulation. LPS levels in portal blood were significantly increased (2-fold) in DSS mice (n = 8) compared with chow mice (n = 8) and further increased in DSS+8d chow mice (n = 8). PN treatment alone resulted in increased LPS levels that were similar to both DSS and DSS+8d chow mice. PN7d/DSS treatment (n = 6) was associated with even higher levels of portal blood LPS compared with PN, DSS, and DSS+8d chow mice (P < 0.05) (Fig. 1B). These data indicate that administration of PN combined with underlying intestinal injury promotes persistently increased intestinal permeability and significant absorption of LPS into the portal circulation. We next tested whether DSS pretreatment induced liver injury. DSS (n = 9) and DSS+8d chow mice (n = 6) did not show elevations of AST, ALT, TSBA, or bilirubin levels compared with chow mice (n = 20) (Fig. 1C-F). RNA expression analysis in whole liver homogenates (n = 5 each) and isolated KCs (pooled from 4-6 mice) showed no increased expression of the proinflammatory cytokines interleukin-6 (Il6), interleukin-1β (Il1β), and tumor necrosis factor α (Tnfα) in DSS, DSS+8d chow, and NS/DSS mice relative to chow mice (Figs. 3 and 4). Furthermore, histology of livers obtained from DSS mice, DSS+8d chow mice, and NS/DSS mice showed no inflammation, cholestasis, steatosis, fibrosis, or hepatocyte death, and immunostaining using F4/80 revealed normal appearance and numbers of KCs (data not shown).

PN Infusion Combined With Increased Intestinal Permeability Induces Liver Injury.

We next determined whether combining PN infusion with increased intestinal permeability would induce liver injury. We randomized DSS-pretreated mice into two groups that were infused with either PN (PN7d/DSS mice; n = 30) or NS (NS/DSS mice: n = 20) for 7 days. PN7d/DSS treatment resulted in significantly elevated AST (≈4-fold; P = 0.0006), ALT (≈4 fold; P = 0.0008), TSBA (>10-fold; P = 0.0037), and total bilirubin (≈2-fold; P = 0.0137) levels compared with NS/DSS, DSS (n = 9), DSS+8d chow (n = 6), and untreated chow mice (n = 20) (Fig. 1C-F). We next tested whether infusion with PN by itself would promote liver injury. PN infusion for 7 days by itself did not cause significant liver injury as measured by AST, ALT, and bilirubin levels in PN mice (n = 10) compared with NS (n = 12) or untreated chow mice (n = 20; P > 0.05) (Fig. 1C-E); mildly elevated TSBA levels were observed in PN mice compared with NS and untreated chow mice (P < 0.05) (Fig. 1F). After 7 days of infusion, PN mice did also not display intestinal inflammation (Supporting Fig. 2). Thus, only PN in combination with DSS-induced intestinal injury/permeability led to significant liver injury within 7 days. Histology of livers stained with H&E, Masson's trichrome, and sirius red did not show significant alterations in PN7d/DSS mice compared with NS/DSS and chow controls (data not shown). The relatively normal liver histology in PN7d/DSS mice is consistent with this model representing an early phase of liver injury.

We next determined the effect of longer PN infusion in DSS pretreated mice. Infusion of PN in DSS-pretreated mice was performed for 28 days (PN28d/DSS mice; n = 5). AST, TSBA, and bilirubin levels remained elevated and were comparable to those in PN7d/DSS mice (n = 30), while ALT levels were even further increased after 28 days of PN (Fig. 1C-F). Liver histology at 28 days showed focal mixed inflammation in the parenchyma, hepatocyte apoptosis, loss of hepatocytes (peliosis), and KC hyperplasia and hypertrophy, consistent with KC activation (Fig. 2). We also detected increased numbers of MPO-positive neutrophils in livers from PN28d/DSS mice (Fig. 2I). There was no evidence of steatosis, portal or canalicular bile retention, bile duct injury, cholangitis, or portal fibrosis.

Figure 2.

PN28d/DSS treatment induces histological liver lesions. (A-C) H&E-stained liver from PN28d/DSS mice shows focal areas of mixed hepatic inflammation (arrows in panel A), loss of hepatocytes (arrows in panel B), and focal hepatocellular apoptosis (arrows in panel C pointing to Councilman bodies). (D-F) Immunohistochemistry using pan macrophage marker F4/80 demonstrates KC hyperplasia and hypertrophy in PN28d/DSS mice (F) compared with PN7d/DSS mice (E) and chow mice (D). (G-I) Immunohistochemistry using specific antibodies for MPO demonstrates increased numbers of activated neutrophils (arrows) in PN28d/DSS mice (I) compared with PN7d/DSS mice (H) and chow mice (G).

PN/DSS Treatment Is Associated With Activation of KCs.

We next tested whether liver injury in PN7d/DSS mice was associated with activation of KCs. KCs were purified from pooled livers (4-6 mice each) derived from wild-type PN7d/DSS (PN/DSS7dTlr4wt), NS/DSS, and untreated chow control mice. Due to limited amounts of RNA obtained from purified KCs we focused our gene expression analysis on canonical cytokines downstream of TLR signaling (i.e., expression of Il6, Tnfα, and transforming growth factor [Tgfβ]). As a positive control for TLR-activated KC gene expression, KCs were purified from livers of mice that received intraperitoneal injection of 10 mg/kg LPS (from E. coli serotype O111:B4; Sigma-Aldrich) 16 hours prior to sacrifice. Transcription of Il6 (2.5-fold), Tnfα (2-fold), and Tgfβ (40-fold) was significantly (P < 0.05) up-regulated in KCs derived from PN7d/DSS Tlr4wt mice compared with cells from NS/DSS and chow control mice (P < 0.05), and was comparable to that detected in KCs from LPS-injected mice (Fig. 3A-C). KCs purified from pooled livers of PN or DSS mice had expression levels of Il6, Tnfα, and Tgfβ similar to KCs from NS and chow mice (Fig. 3A-C), indicating that neither PN nor DSS treatment alone promoted KC activation. Importantly, transcriptional analysis of purified CD11b-positive splenic macrophages isolated from PN7d/DSS mice (n = 3) did not reveal increased cytokine transcription for Il6, Tnfα, and Tgfβ compared with NS/DSS controls (n = 3) (Fig. 3D). Furthermore, transcript levels of Il6, Il1β, and Tnfα (Fig. 4A-C) were not elevated in RNA from whole liver homogenate from PN7d/DSS mice (n = 10) compared with NS/DSS (n = 5), DSS (n = 5), NS (n = 5), PN (n = 5), and chow mice (n = 5), but were elevated in PN28/DSS mice (n = 5). These data indicate that functional activation of KCs is an early event in PNALI.

Figure 3.

KC activation in PN7d/DSS mice. (A-C) qRT-PCR analysis depicting relative transcript levels of Il6 (A), Tnfα (B), and Tgfβ (C) mRNA in purified KCs. KCs derived from LPS-injected mice (LPS) served as a positive control for innate activation of KCs. KC activation is increased in PN7d/DSS wild-type mice (PN7d/DSSTlr4wt) compared with controls. Note that DSS treatment alone (DSS, DSS+8d chow, NS/DSS) or PN infusion in the absence of DSS pretreatment (PN) does not induce KC activation. KC activation is attenuated after suppression of bowel flora (PN7d/DSSTlr4wt+Abx) or ablation of TLR4 signaling (PN7d/DSSTlr4mut) compared with KCs from PN7d/DSS wild-type mice (PN7d/DSSTlr4wt). (D) qRT-PCR analysis depicting relative transcript levels of Il6, Tnfα, and Tgfβ in purified splenic macrophages from NS/DSS and PN7d/DSS mice (pooled from n = 3 mice per group). Note that cytokines are not increased in splenic macrophages derived from PN7d/DSS mice. Data were normalized to hypoxanthine-phosphoribosyl-transferase-1 (Hprt1) as an endogenous control gene. mRNA levels are expressed relative to results obtained from untreated chow mice (A-C) or NS/DSS mice (D). Mean values and SEM of two independent experiments that were analyzed in triplicate are shown. *P < 0.05 PN7d/DSSTlr4wt and LPS versus controls. #P < 0.05 PN7d/DSSTlr4wt+Abx and PN7d/DSSTlr4mut versus PN7d/DSSTlr4wt (one-way ANOVA).

Figure 4.

Inflammatory gene transcription is increased in livers from PN28d/DSS mice. (A-C) RT-PCR analysis depicting relative transcript levels of Il6 (A), Tnfα (B), and Il1β (C) mRNA in whole liver homogenates. Elevated transcription was present at 28 but not 7 days of PN/DSS. Data were normalized to hypoxanthine-phosphoribosyl-transferase-1 (Hprt1) as an endogenous control gene. mRNA levels are expressed relative to results obtained from untreated chow mice. Mean values and SEM from n = 5 samples tested in triplicate are shown. *P < 0.05 PN28d/DSS versus all other groups (one-way ANOVA).

KC Activation and Liver Injury Are Dependent on the Intestinal Microbiota.

The intestinal microbiota are a major source of intestinal-derived LPS. We therefore tested whether intestinal microbiota were involved in promoting KC activation and liver injury in the mouse model. Intestinal microbiota were suppressed by exposing wild-type PN7d/DSS mice ad lib to an oral cocktail of four broad-spectrum antibiotics in drinking water during the entire 7-day period of PN infusion (PN7d/DSS Tlr4wt+Abx mice). qRT-PCR analysis of copy numbers of bacterial 16S DNA in colonic feces demonstrated a >98% reduction (4.8*105 → 4.7*103, P < 0.05) of the intestinal microbiota in PN7d/DSS Tlr4wt+Abx–treated mice (n = 7) relative to PN7d/DSSTlr4wt (n = 13) mice (Fig. 5A). KC messenger RNA (mRNA) expression of Il6, Tnfα, and TGFβ was significantly lower (P < 0.05) in PN7d/DSSTlr4wt+Abx mice compared with PN7d/DSSTlr4wt mice, and similar to expression in KCs isolated from NS/DSS mice (P > 0.05) (Fig. 3A-C). Antibiotic treatment was also associated with a significant reduction in liver injury (AST, ALT) (Fig. 5B,C). Cholestasis was almost completely prevented in PN7d/DSSTlr4wt+Abx mice, with TSBA and total bilirubin values comparable to NS/DSS (n = 12) and chow mice (n = 20) (Fig. 5D,E). Thus, oral antibiotic treatment reduced colonic bacterial load and prevented both KC activation and liver injury.

Figure 5.

Liver injury is attenuated after suppression of bowel flora or ablation of TLR4 signaling. (A) qRT-PCR to determine copy number of bacterial 16S DNA genes in colonic stool samples after oral antibiotic treatment. Note the significant reduction in bacterial load in stool from PN7d/DSSTlr4wt+Abx mice. Note that ablation of TLR4 signaling does not reduce bacterial load in stool of PN7d/DSSTlr4mut mice compared with wild-type mice. *P < 0.05 PN7d/DSSTlr4wt+Abx versus other groups (one-way ANOVA). (B-E) Hepatocyte injury (B,C) and cholestasis (D,E) were attenuated in PN/DSS wild-type mice that were treated with oral antibiotics (PN7d/DSSTlr4wt+Abx; n = 11) or had nonfunctional TLR4 signaling (PN7d/DSSTlr4mut; n = 12) compared with PN/DSS wild-type mice (PN7d/DSSTlr4wt; n = 15). NS/DSS mice (n = 12) and untreated chow mice (n = 20) served as controls. Mean values and SEM are shown. *P < 0.05 PN7d/DSSTlr4wt versus controls and PN7d/DSSTlr4wt+Abx and PN7d/DSSTlr4mut versus PN7d/DSSTlr4wt. #P < 0.05 PN7d/DSSTlr4wt+Abx and PN7d/DSSTlr4mut versus controls (one-way ANOVA).

TLR4 Signaling Significantly Contributes to KC Activation and Liver Injury.

We next examined the role of TLR4 signaling in this model by using Tlr4 mutant mice with defective TLR4 signaling. Tlr4 mutant (PN7d/DSSTlr4mut; n = 11) and wild-type (PN7d/DSSTlr4wt; n = 10) mice underwent DSS pretreatment followed by PN infusion for 7 days. Il6 transcription (tested as a representative canonical cytokine, since RNA amounts were very limited from Tlr4 mutant mice due to reduced numbers of purified CD11b+ KCs) in KCs purified from PN7d/DSSTlr4mut mice was markedly reduced compared with PN7d/DSSTlr4wt mice and was similar to NS/DSS control mice (Fig. 3B). Importantly, hepatocyte injury and cholestasis were significantly attenuated in PN7d/DSStlr4mut mice, with significant reductions in AST and ALT levels (P < 0.05) and marked reductions in bilirubin and TSBA compared with PN7d/DSStlr4wt–treated mice (Fig. 5B-E). Attenuation of liver injury and KC activation was not associated with reduced KC numbers in Tlr4 mutant mice (Supporting Fig. 2).


Although the pathogenesis of PNALI is poorly understood, the severity and chronicity of PNALI are clearly increased in infants with underlying intestinal inflammation, presumed small bowel bacterial overgrowth, and increased intestinal permeability.4 These factors together may increase translocation of bacteria and/or microbial-derived TLR agonists into the portal circulation. Of note, KC hyperplasia and inflammation are characteristic of the liver histopathology in human infants with PNALI.11

The present study was designed to test the hypothesis that an early upstream event in the pathogenesis of PNALI is activation of KCs by intestinal-derived TLR agonists. We therefore designed a mouse model that mimics the early pathophysiology present in the human infant at greatest risk for PNALI by providing continuous PN infusion with intestinal injury and increased permeability. Three observations support this approach: (1) we have reported noninfectious colitis complicating short bowel syndrome, verified by sigmoidoscopy and on colonic biopsies7; (2) Kaufman et al.4 reported significant small intestinal inflammation on biopsies (associated with bacterial overgrowth) in TPN-dependent children with short bowel syndrome; and (3) intestines that are removed at the time of intestinal transplant in children with intestinal failure typically show chronic inflammation (unpublished data). We took advantage of the known effects of DSS on intestinal integrity, which allowed us to modulate the severity of intestinal inflammation so that the mice could tolerate PN without the usual weight loss observed in most DSS models.19 Our study has provided evidence that early liver injury in PNALI is induced by a mechanism that involves TLR4-dependent activation of KCs by LPS absorbed from the gut microbiota. Intestinal injury by itself or PN infusion alone was insufficient to induce KC activation or liver injury. KC activation in PN7d/DSS-treated mice was demonstrated by increased transcription of canonical TLR signaling-NFκB–dependent cytokines. Whole liver homogenates or splenic macrophages derived from PN7d/DSS mice did not show increased transcription of proinflammatory cytokines, providing evidence that isolated activation of the KCs was a key early event in this liver injury. The very early phases of this liver injury were not associated with visible histopathological alterations of the liver, similar to what is observed in 30%-40% of human infants who receive TPN for less than 2 weeks.20 However, after 28 days of PN infusion following DSS pretreatment, we observed marked KC hyperplasia and hypertrophy (consistent with KC activation), which was associated with recruitment of neutrophils, hepatic inflammatory infiltrates, hepatocellular apoptosis, and parenchymal peliosis. Importantly, these lesions at 28 days were associated with increased transcription of proinflammatory cytokines from liver homogenate, suggesting a progression and broadening of the inflammatory process. We acknowledge that adult rather than infant mice were used in these experiments because of the feasibility of placing central catheters. We speculate that if CVCs could have been placed in infant mice, we would have seen an exaggerated cholestatic injury.

The observation that KC activation and liver injury were observed only in PN mice with concomitant intestinal injury is consistent with the proposed pathogenesis of other chronic liver injuries, in which absorption of TLR agonists through a disrupted intestinal barrier results in TLR-dependent activation of KCs with subsequent generation of proinflammatory and profibrogenic mediators.12, 21 Under normal circumstances, a functional intestinal epithelial barrier ensures that only trace amounts of LPS and other TLR agonists enter into the portal/sinusoidal circulation in healthy humans and rodents. Our data support the notion that disruption of the intestinal barrier function in DSS mice led to increased absorption of LPS. Combining administration of PN with DSS treatment appeared to have an additive effect on LPS absorption with subsequent activation of KCs.

Using a genetic approach, we have demonstrated that interruption of LPS-TLR4 signaling in Tlr4 mutant mice attenuated KC transcription of Il6. Only a small number of CD11b+ KC could be isolated from Tlr4 mutant mice, which may be due to attenuated NFκB-mediated CD11b expression in Tlr4 mutant mice22 and, as our results suggest, not due to fewer numbers of KCs in Tlr4 mutant mice. This reduced the amount of recovered RNA so that the number of cytokines that could be examined by mRNA expression was limited; thus, we focused on Il6 as a canonical marker for TLR-dependent KC activation. Importantly, attenuated KC activation in Tlr4 mutant mice was associated with reduced liver injury and cholestasis. These findings are in keeping with previous studies using Tlr4 mutant mice in a bile duct–ligated model of cholestasis in which progression to fibrosis was attenuated by disrupting TLR4 signaling.12 Our data also demonstrate that liver injury, albeit significantly reduced, still developed to a limited extent in Tlr4 mutant mice. This finding implies that other TLR signaling pathways or other cell types may also be involved in the liver injury. In this regard, TLR9-dependent activation of murine KCs has been shown to be critically involved in progression from steatosis to steatohepatitis and fibrosis.23 Moreover, Lichtman et al.24 have demonstrated that intestinal-derived peptidoglycan, a TLR2 agonist, played a major role in liver injury in rats with bacterial overgrowth of the small intestine in a KC-dependent mechanism. Further studies using pharmacological and/or genetic ablation of individual TLR signaling components isolated to KCs are required for a more detailed definition of KC activation in this model. We also identified that CD11b+ splenic macrophages from PN/DSS mice were not activated, indicating that systemic activation of macrophages (as in bacteremia and sepsis) did not play a role. The most likely factor that could account for this hepatic macrophage specificity is intestinal absorption of TLR agonists into the portal circulation triggering TLR signaling in sinusoidal KCs. However, treatment of mice with DSS alone or PN alone did not initiate liver injury or KC activation, albeit increased LPS absorption was observed in these mice; thus, factors associated with the infusion of PN (e.g., lack of enteral feedings or constituents of the PN solution) must be involved as well.

Recent studies have implicated the intestinal microbiota in promoting liver injury and progression to fibrosis by a mechanism that involves TLR-dependent activation of KCs.12 To test the role of intestinal microbiota, we employed an oral cocktail of nonabsorbable broad spectrum antibiotics during the time of PN infusion, which significantly reduced intestinal bacterial flora, and both attenuated liver injury and prevented induction of proinflammatory cytokines in KCs. This antibiotic cocktail has been shown to not alter the number of KCs, TLR4 expression, or KC responsiveness to LPS.12 Thus, reduction of intestinal bacterial load appeared to have inhibited activation of KCs and the development of liver injury and cholestasis, which is further supported by unpublished metagenomic data from our group that demonstrate a significant reduction in gram-negative (LPS source) bacterial species in PN7d/DSS/Abx mice. A complete metagenomic characterization of the intestinal microbiome in this mouse model is underway and will be the subject of a future report.25

One intriguing aspect of this mouse model is that KC activation and liver injury were not observed in mice receiving either DSS or PN alone, but that the combination was required. Based on this study and the work of others, we propose that the early stages in the pathogenesis of PNALI may involve the following. The lack of enteral feedings during PN infusion and underlying intestinal dysfunction/dysmotility promote bacterial overgrowth, which enhances intestinal inflammation and compromises barrier function.4-6, 9 Together with preexisting intestinal injury and the presumed increased intestinal permeability of infants,8 absorption of TLR agonists is enhanced. KC activation is then induced through TLR signaling and potentially by proinflammatory eicosanoids as well as lipid peroxidation products derived from omega-6 fatty acids present in the standard soy lipid-based PN.26 Cholestasis and liver injury may be further amplified by the inhibitory effects on the expression of canalicular bile acid transporters by proinflammatory cytokines (e.g., Il6, Il1β) derived from activated KCs27 and from soy lipid-derived phytosterols in the PN solution.28 We propose that this mouse model will be useful in elucidating the pathogenesis of early initiating mechanisms of PNALI and in further characterization of the role of the gut-liver axis in liver disease.


We thank Daniel H. Teitelbaum, M.D., University of Michigan, for assistance with the development of the PNALI mouse model.