Main

Although BA is the most common indication for pediatric liver transplantation and CFLD is the second most common specific cause of mortality in CF, there is a fundamental lack of understanding of the pathophysiology of both diseases. In spite of different etiologies, these diseases share the final common pathway of cirrhosis. The histologic hallmarks of BA are cholestasis, progressive inflammation, bile duct proliferation, and fibrosis(1). In CFLD, focal biliary cirrhosis is considered a pathognomonic lesion which may progress to multilobular biliary cirrhosis, and is thought to be initiated by obstruction of bile ductules with eosinophilic material(2). Because TGF-β1 is a major fibrogenic cytokine in experimental and adult liver diseases, we postulated that it may play a role in hepatic fibrosis in these two common fibrotic liver diseases of childhood.

TGF-β1 is a multifunctional polypeptide growth factor which has been implicated in fibrosis in a number of organs including skin, lung, kidney, and liver(3). TGF-β1 has been associated with hepatic fibrogenesis in experimental animals treated with a number of different hepatotoxins and infectious agents. For example, in rats with hepatic fibrogenesis secondary to either carbon tetrachloride, thioacetamide, or Schistosoma mansoni, hepatic expression of TGF-β1 mRNA and protein was increased(49). In the bile duct ligation model of fibrosis in rats, increased TGF-β1 was present in bile duct epithelium(10). Transgenic mice in which TGF-β1 was overexpressed developed hepatic fibrosis(11).

Hepatic expression of TGF-β1 mRNA and protein was increased in human adults with hepatic fibrosis, and TGF-β1 protein was localized to areas of extracellular matrix deposition(12,13). In adults with chronic liver disease, hepatic TGF-β1 mRNA correlated closely with procollagen mRNA and histologic activity scores(14). Shirai et al.(15) showed that plasma TGF-β1 protein was increased in adults with cirrhosis compared with adult control subjects (3.7 ± 2.1 ng/mL versus 1.4 ± 0.8 ng/mL, p < 0.05). However, no studies have examined the role of TGF-β1 in fibrotic liver diseases of childhood. We hypothesized that both the hepatic expression and plasma concentrations of TGF-β1 protein would be increased in children with hepatic fibrosis secondary to BA or CFLD. For comparison, studies were also performed in liver biopsy samples from patients with liver disease secondary to α1-antitrypsin deficiency, drug hepatitis, total parenteral nutrition cholestasis, and primary sclerosing cholangitis.

METHODS

Human studies: plasma analysis. Plasma was obtained from the following groups: infants and children with BA undergoing liver transplantation (n = 9, 2 ± 1 y of age) and healthy infants and children, with no history of liver disease, undergoing elective surgery(n = 9, 2 ± 0 y of age). Surgical procedures included hypospadias repair, circumcision, herniorrhaphy, or myringotomy tube placement. Healthy adolescents (n = 10, 13 ± 1 y of age) and healthy adults (n = 10, 30 ± 2 y of age) were also studied. Patients with CF without clinical or biochemical evidence of liver disease (serum ALT < 2 times normal) (n = 14, 12 ± 1 y of age) and with CFLD (ALT > 2 times normal, n = 11, 16± 4 y of age) were recruited from the Cystic Fibrosis Clinic at the Johns Hopkins Children's Center. Plasma TGF-β1 was also analyzed in one patient each with total parenteral nutrition cholestasis, drug hepatitis, hepatitis C, and primary sclerosing cholangitis. The study was approved by the Johns Hopkins Committee on Clinical Investigation, and informed consent was obtained from all patients and/or parents and/or legal guardians.

Plasma collection. In patients with BA, blood samples were obtained via an indwelling central venous catheter after induction of general anesthesia, before liver transplantation. In the healthy infants and children undergoing surgery, blood was collected via venipuncture after induction of general anesthesia, before surgery. In all other groups, blood was collected percutaneously. Care was taken to obtain plasma as atraumatically as possible to minimize artifactual TGF-β1 derived from endothelium and platelet stores. The first 2 mL of blood were discarded, and 5 mL were gently drawn into a separate syringe and immediately transferred to a chilled heparinized tube. Vacuum was released from the tube before adding the sample. Samples were centrifuged at 1000 × g for 30 min, within 1 h of collection. The plasma supernatant fraction was stored at -70°C. Interassay variation over 12 mo of storage at -70°C was 14%. Intraassay variation was 10%.

Plasma sample preparation and TGF-β1 ELISA protocol. Nunc Maxisorb ELISA microtiter 96-well plates (PGC Scientific, Gaithersburg, MD) were coated with 1 µg/µl monoclonal anti-TGF-β1,2,3(Genzyme, Cambridge, MA) in PBS overnight at room temperature(16). All subsequent incubations were done at room temperature. Plates were treated with blocking buffer (10 mg/mL crystalline BSA) the following day. Plasma samples were then added to plates for 1.5 h.

To activate the TGF-β1, plasma was treated with 2.5 N glacial acetic acid/10 M urea followed by addition of 2.7 N NaOH to obtain a pH of 7.0. Neutralizing buffer containing 0.1% Tween 20, 200 mM Tris-HCl, was used to dilute samples 1:4. Recombinant human TGF-β1 standards (R&D Systems, Minneapolis, MN) were assayed in duplicate with samples. After a PBS wash, plates were incubated 1 h with monoclonal chicken anti-TGF-β1 (R&D Systems). Next, alkaline phosphatase-conjugated rabbit anti-chicken IgG(Sigma Chemical Co. Immunochemicals, St. Louis, MO) was allowed to bind to the second antibody for 1 h. Color development was observed with addition of p-nitrophenyl phosphate in 1 M diethanolamine buffer. The TGF-β1 concentration was determined by absorbance at 450 nm.

PF4 assay. PF4 was measured as an estimate of the platelet-derived pool of TGF-β1 in each sample; ELISA kits were obtained from American Bioproducts, Parsippany, NJ, and all procedures were carried out at room temperature. Samples were assayed in duplicate and repeated simultaneously with PF4 standards. After addition of samples and standards to precoated wells, plates were incubated 1 h in the dark. Peroxidase-conjugated anti-PF4 antibody was added, and plates were again incubated for 1 h. Color development was observed over a 3-min period after addition of a reagent containing in 30% hydrogen peroxide; then 1 N HCl was added to stop the reaction. PF4 concentrations were determined by absorbance at 490 nm. Intraassay variation was <5%.

Selection of liver biopsy specimens. Archival liver biopsy specimens were obtained from patients with BA (n = 10, mean age 0.8 ± 0.3 y), CFLD (n = 10, 15 ± 2 y), and control subjects with histologically normal livers obtained during staging laparotomy for Hodgkin's disease (n = 10, 14 ± 3 y). In addition, histologically normal liver biopsies were obtained from two children undergoing staging laparotomy for neuroblastoma (ages 2 and 3 y) as were two liver biopsies from infants with hepatic fibrosis secondary toα1-antitrypsin deficiency (ages 3 and 10 mo), and one liver biopsy each from a patient with total parenteral nutrition cholestasis, drug hepatitis, or primary sclerosing cholangitis. All samples were obtained from the Pathology Departments of the Johns Hopkins or Columbus Children's Hospitals.

Preparation of liver biopsy specimens with antibodies to TGF-β1 protein. Five-micrometer thick tissue sections were analyzed for TGF-β1 protein according to the method of Flanders et al.(17). Paraffin-embedded samples were deparaffinized and treated with 0.6% hydrogen peroxide in methanol to minimize background from endogenous peroxidase in the tissue. After a rinse with Tris-buffered saline containing 0.1% BSA, slides were covered with bovine testicular hyaluronidase(1 mg/mL) and incubated at 37°C for 30 min in a humidified chamber. A blocking solution (Tris-buffered saline containing 1% BSA and 5% normal goat serum) was added and slides were incubated in a humidified chamber for 30 min. TGF-β1 LC antibody (gift from Dr. Michael Sporn), diluted in Tris-buffered saline/1% BSA, was maintained at 4°C overnight. The secondary antibody, biotinylated goat anti-rabbit IgG (Vectastin Elite ABC Kit, Vector Labs, Burlingame, CA) was then applied. Color was developed with a peroxidase substrate; slides were then counterstained with Mayer's hematoxylin.

The TGF-β1 was scored semiquantitatively by three investigators who were blinded to the origin of the sample and diagnosis. The following scoring schema was used: samples with trace or no fibrosis = 1, moderate fibrosis = 2, and severe fibrosis = 3. Samples prepared for TGF-β1 labeling were assessed similarly; samples with trace or no staining = 1, moderate staining = 2, marked staining = 3.

Statistics. The relationships among plasma TGF-β1 and group ages for BA and the age-appropriate control group were examined using linear regression and Mann-Whitney rank sum test. ANOVA was used to analyze plasma TGF-β1 and ages in the CFLD and age-appropriate control groups. Histologic scoring was analyzed by a log-linear model of a contingency table. The α level for statistical significance was 0.05. Continuous data results are reported as mean ± SEM.

RESULTS

In a pilot study of healthy subjects, there was an inverse correlation between the plasma concentration of TGF-β1 and age (Fig. 1A). A similar relationship was observed between age and the plasma PF4 concentration (Fig. 1B). TGF-β1 and PF4 demonstrated a strong positive association (Fig. 1C).

Figure 1
figure 1

(A) Plasma TGF-β1vs age in healthy control subjects. (B) Plasma PF4vs age in healthy control subjects. (C) Plasma TGF-β1 vs PF4 in healthy control subjects.

Clinical and laboratory data pertinent to patients in the plasma TGF-β1 study are shown in Table 1. In patients with BA, serum total and direct bilirubin, AST, ALT, and alkaline phosphatase were increased compared with control subjects, and to patients with CFLD. Hematologic indices were comparable in both disease groups, except that the infants with BA (who were undergoing liver transplantation) exhibited anemia consistent with chronic severe liver disease.

Table 1 Clinical and laboratory data

In patients with BA, the plasma concentration of TGF-β1 was decreased compared with values for age-appropriate controls (Fig. 2). The ages of the two groups were comparable: 2 ± 1 y of age for BA versus 2 ± 0 y of age for healthy infants and children. Plasma PF4 was also reduced in BA patients compared with control subjects: 83± 17 versus 143 ± 2 IU/mL, p < 0.005. When TGF-β1 values were analyzed by multivariate analysis using PF4 as an indirect index of platelet-derived TGF-β1, the reduction of TGF-β1 in BA was independent of the values for PF4.

Figure 2
figure 2

Plasma TGF-β1 in patients with BA or CFLD and control subjects. **p < 0.01 BA vs healthy "infants" (2 ± 0 y old), referred to in the text as "healthy infants and children"). *p < 0.05 CFLD vs CF well and vs healthy teens.

Decreased plasma TGF-β1 concentrations were also found in CFLD patients compared with age-appropriate CF patients without liver disease and non-CF, healthy age-appropriate subjects (Fig. 2). The respective ages of these three groups did not differ: 16 ± 4 y of age for CFLD, 12 ± 1 y of age for "well" CF patients, and 13 ± 1 y of age for healthy adolescents. Because two of the CFLD patients were adults(39 and 42 y old), and our data in healthy subjects had shown that plasma TGF-β1 decreased with age, the plasma TGF-β1 values for the CFLD patients were analyzed after deletion of the two values for those adults. Plasma TGF-β1 values for the 9 remaining CFLD patients (who were 11± 4 y of age) were 2.7 ± 3.3 ng/mL, almost identical to values when all 11 CFLD patients were analyzed (2.4 ± 3.0 ng/mL). PF4 values for CFLD were not significantly different from the CF controls; 45± 12 versus 74 ± 12 IU/mL (p = 0.074). When the TGF-β1 and PF4 data were analyzed by multivariate analysis, the difference in TGF-β1 between CFLD and the two control groups was not independent of the effect of PF4.

Liver biopsy samples from patients with BA demonstrated advanced biliary cirrhosis in each patient. Other findings included cholestasis and bile duct proliferation. Centrolobular necrosis was seen in one patient. Among patients with CFLD, five of nine specimens showed cirrhosis, whereas macrovesicular fat was the dominant finding in two, and focal fibrosis was the significant finding in two. All control patients had histologically normal livers.

Figure 3A illustrates the increased hepatic TGF-β1 protein and fibrosis in a representative patient with BA. TGF-β1 staining was identified in hepatocytes and Kupffer cells, in regions adjacent to fibrotic bands. Similar findings were noted in CFLD(Fig. 3B). In contrast, the liver biopsies of the children with hepatic fibrosis secondary to α1-antitrypsin deficiency exhibited markedly increased staining of TGF-β1 protein throughout the liver lobule (Fig. 3,C and D). This generalized increase in staining of TGF-β1 protein was also evident in the liver biopsy of the patient with total parenteral nutrition cholestasis(Fig. 3,E and F. In the histologically normal liver biopsy control specimens, the brown staining characteristic of TGF-β1 protein staining was minimal or absent.

Figure 3
figure 3

Immunohistochemistry for TGF-β1 protein in the liver, done with LC antibody. (A) This 2-y-old patient with BA shows brown cytoplasmic staining of hepatocytes (indicative of TGF-β1 protein) located immediately adjacent to the regions of fibrosis (arrow) (100×). (B) The liver of a 12-y-old patient with CF also shows brown cytoplasmic staining of the hepatocytes adjacent to the portal tracts (arrow) (200×).(C and D) This cirrhotic liver of a 10-mo-old withα1-antitrypsin deficiency shows strong brown cytoplasmic staining of hepatocytes (arrows). C, 100×;D, 200×. (E and F) Staining for TGF-β1 in the liver of a patient with total parenteral nutrition cholestasis. Note the generalized increase in staining of TGF-β1 protein throughout the liver lobule. E, 100×; F, 150×.

Table 2 shows the relationship between TGF-β1 scoring and fibrotic scoring for 30 liver biopsy samples studied; 10 each from patients with either BA, CFLD, and control subjects with Hodgkin's disease and histologically normal livers. There was a strong association between hepatic fibrosis and TGF-β1 protein scores. The likelihood ratio of this association was 23.38, p < 0.007.

Table 2 Two-way contingency table analysis of fibrosis and TGF-β1 protein in liver

Table 3 contains data for the five BA patients in whom it was possible to analyze both plasma and hepatic TGF-β1 protein, as well as three patients with liver disease of diverse etiologies and one patient in whom only plasma TGF-β1 protein values could be measured. In this small subset, no correlations between the variables were evident. The r value = 0.0294 for plasma TGF-β1 versus PF4,p = 0.963. Neither a negative correlation between plasma and hepatic TGF-β1 nor a positive correlation between plasma TGF-β1 and PF4 was observed.

Table 3 Liver and plasma TGF-β1, in childhood liver disease

DISCUSSION

The strong association between hepatic TGF-β1 protein and hepatic fibrosis in several common types of pediatric liver disease was a major finding of this study. The association between hepatic fibrosis and this fibrogenic cytokine was evident in parenchymal disease (secondary toα1-antitrypsin deficiency, total parenteral nutrition cholestasis, and drug hepatitis) as well as biliary tract disease. To our knowledge, this is the first report of this association in childhood liver disease. This association was evident both from assessment by a semi-quantitative scoring system as well as by subjective review of the typical patterns of TGF-β1 staining. In the liver biopsy samples with biliary fibrosis, the strongest staining was in the perifibrotic hepatocytes, whereas in the parenchymal liver diseases, staining was generalized throughout the liver lobule.

Our finding of increased hepatic TGF-β1 protein in association with hepatic fibrosis suggests that this cytokine may play a role in the fibrogenesis of several common childhood liver diseases. Animal and human adult data, discussed earlier, also suggest that TGF-β1 is a fibrogenic cytokine. On the other hand, scoring of the TGF-β1 labeling may have been biased by the presence of fibrosis in each specimen so that scoring may not have been done in a truly blinded fashion. However, bias of this sort was minimized by using three investigators to score the samples independently, without prior knowledge of the sample source. It seems unlikely that the perifibrotic TGF-β1 staining in liver biopsies from the young BA patients (mean age < 2 y) represents a physiologic growth-related increase, given the minimal TGF-β1 staining in the histologically normal livers of the two young subjects with neuroblastoma.

The second major finding of our study was the effect of age on PF4 and TGF-β1, demonstrating that conclusions regarding roles of these two factors in a given disease cannot be made unless efforts are made to control for age. The increased plasma TGF-β1 protein in young human subjects is consistent with the finding of increased plasma TGF-β1 protein in both healthy control mice and mice transgenic for mature TGF-β1 tested at 2 wk of age in comparison with values thereafter(11). In that study, the authors speculated that the increased TGF-β1 in the plasma of the 2-wk-old nontransgenic mice was in the biologically inactive latent form, because the control mice did not exhibit the increased hepatic expression of TGF-β1 protein, type I collagen, and hepatic fibrosis characteristic of the transgenic mice. They postulated that the plasma TGF-β1 in the transgenic mice was in the mature biologically active form. It should be noted that in our study as in the mouse study, the assay used in our study measures both mature and latent forms of the cytokine.

An unanticipated finding in our study was that plasma TGF-β1 was decreased in children with hepatic fibrosis secondary to either BA or CFLD compared with age-appropriate healthy controls. In BA, this effect was independent of PF4 values, whereas in CFLD the reduction in plasma TGF-β1 was found to be related to the reduction in plasma PF4. The association of plasma TGF-β1 and PF4 in patients with CFLD suggests that the decreased plasma TGF-β1 values in these patients could merely be an artifact secondary to the thrombocytopenia of chronic liver disease. This possibility seems unlikely given the normal average platelet counts in the patients with CFLD (Table 1), very similar to platelet counts for the infants with BA, in whom there was no association between plasma TGF-β1 and PF4.

Although in theory, children with fibrotic liver disease could exhibit decreased release of platelet-derived TGF-β1, adults with cirrhosis actually exhibit either normal or increased platelet activation(18). Another possible explanation for the decreased plasma TGF-β1 in the BA patients is that the method of obtaining blood samples (via a central catheter) may have resulted in less platelet activation and release of TGF-β1 than the venipuncture technique used in the age-appropriate control subjects. However, this explanation could not be used in the CFLD patients in whom the blood drawing technique was identical to those used in control subjects.

Other explanations for the decreased plasma TGF-β1 protein in patients with BA and CFLD include decreased production or release, increased clearance or catabolism, or factors in the plasma of these children that interfere with immunorecognition of TGF-β1 by the antibody used in our assay. Also, because the BA patients were chronically ill, undergoing liver transplantation, utilization of other chronically ill children as control subjects might have been instructive.

However, the data in Table 3 suggest that patients with liver disease do not always exhibit decreased plasma TGF-β1 and that there is little correlation between hepatic expression and plasma concentration of the cytokine in children with liver disease. Even the reports in adults with liver disease demonstrating increased plasma concentrations of TGF-β1 in patients with cirrhosis and hepatocellular carcinoma reported a wide range of values in liver disease state, with considerable overlap between values for healthy adults and those with proven cirrhosis(15).

In conclusion, these studies have established the association of hepatic TGF-β1 protein and hepatic fibrosis in several common liver diseases of childhood based on immunohistochemical identification of TGF-β1 protein in hepatic tissue. Although association does not imply causation, substantial data from experiments in animals and human adults support a pathogenic role for this cytokine in hepatic fibrosis of children. These findings suggest that an important area for future research will be in the determination of the factors in the fibrotic liver diseases of childhood which promote the hepatic expression of this fibrogenic cytokine. Our data also suggest that plasma TGF-β1 is not a useful marker of hepatic expression of this protein in children with liver disease.