Potential conflict of interest: Nothing to report.
Hepatocellular carcinoma (HCC) is rare in young children. We attempted to see if immunohistochemical and mutational-analysis studies could demonstrate that deficiency of the canalicular bile acid transporter bile salt export pump (BSEP) and mutation in ABCB11, encoding BSEP, underlay progressive familial intrahepatic cholestasis (PFIC)—or “neonatal hepatitis” suggesting PFIC—that was associated with HCC in young children. We studied 11 cases of pediatric HCC in the setting of PFIC or “neonatal hepatitis” suggesting PFIC. Archival liver were retrieved and immunostained for BSEP. Mutational analysis of ABCB11 was performed in leukocyte DNA from available patients and parents. Among the 11 nonrelated children studied aged 13-52 months at diagnosis of HCC, 9 (and a full sibling, with neonatal hepatitis suggesting PFIC, of a tenth from whom liver was not available) had immunohistochemical evidence of BSEP deficiency; the eleventh child did not. Mutations in ABCB11 were demonstrated in all patients with BSEP deficiency in whom leukocyte DNA could be studied (n = 7). These mutations were confirmed in the parents (n = 14). With respect to the other 3 children with BSEP deficiency, mutations in ABCB11 were demonstrated in all 5 parents in whom leukocyte DNA could be studied. Thirteen different mutations were found. In conclusion, PFIC associated with BSEP deficiency represents a previously unrecognized risk for HCC in young children. Immunohistochemical evidence of BSEP deficiency correlates well with demonstrable mutation in ABCB11. (HEPATOLOGY 2006;44:478–486.)
Progressive familial intrahepatic cholestasis (PFIC) with normal or only slightly elevated serum concentrations of γ-glutamyltranspeptidase (GGT) activity, or low-GGT PFIC, once known as “Byler disease,” encompasses several disorders.1 These are defined by persistent nonremitting conjugated hyperbilirubinemia with elevated serum concentrations of bile acids (hypercholanemia), lack of abnormal bile acid species in serum and urine, patency of a normally formed biliary tract, and, unless partial external biliary diversion (PEBD) or ileal exclusion is undertaken,2–4 evolution into end-stage liver disease. Two principal categories are recognized within low-GGT PFIC: familial intrahepatic cholestasis 1 (FIC1) deficiency and bile salt export pump (BSEP) deficiency. The former is caused by mutation in ATP8B1, which encodes FIC1.5, 6 FIC1 is a putative aminophospholipid flippase.7 The latter is caused by mutation in ABCB11,8 which encodes BSEP.8, 9 BSEP is the principal conveyor of bile acids from hepatocyte cytoplasm into bile canaliculus.10, 11 At biopsy on presentation in infancy, patients with PFIC due to FIC1 deficiency generally have bland canalicular cholestasis, whereas patients with PFIC due to BSEP deficiency generally have “neonatal hepatitis.”1, 12
Hepatocellular carcinoma (HCC) is rare in children.13 We identified 11 unrelated children with clinically diagnosed PFIC in whom HCC was diagnosed between the ages of 13 and 52 months. In 10 of these children, BSEP deficiency was demonstrated immunohistochemically and mutation in ABCB11 was demonstrated via molecular analysis using materials from the children or their family members. We describe clinical and histopathological findings, attempt correlation with results of mutational analysis and processes of carcinogenesis, and suggest implications for management.
HCC was found in the explanted liver of a boy with intrahepatic cholestasis (patient A; Table 1) and BSEP deficiency. The boy's disorder, which manifested at 3 weeks of age, was characterized by failure to thrive and icterus. Clinical and laboratory studies revealed hypercholanemia, conjugated hyperbilirubinemia, and low GGT. Urine screening did not identify abnormal bile acid species. Imaging studies did not suggest bile duct obstruction or malformation. Microscopy of a liver biopsy specimen obtained at 35 days of age identified “neonatal hepatitis” with intralobular cholestasis and anisocytosis, rosetting, edema, multinucleation, and individual cell necrosis of hepatocytes; portal-tract cholestasis was not found (Fig. 1A). Immunohistochemical studies demonstrated normal expression of multidrug resistance-associated protein 2 (MRP2) (Fig 1B), like BSEP an adenosine triphosphate (ATP)-binding cassette protein involved in canalicular transport; MRP2 here served as a functional control useful in assessing tissue preservation.9 Marking for BSEP (see Methods section) was entirely absent (Fig. 1C). Allogeneic hepatocytes were infused14 in hopes of averting orthotopic liver transplantation (LT), and tacrolimus was given. No clinical effect was apparent, although GGT rose slightly, consonant with greater access of bile acids to the canalicular lumen.1, 15 The child was listed for LT. The serum concentration of α-fetoprotein (AFP) had been 2 IU/L as a neonate (nl < 7). This level had risen to 34 IU/L before hepatocyte transplantation and reached 199 IU/L by LT (21 months), after which it fell immediately. The deep green-brown explanted liver contained a pink-white mass 0.5 cm in diameter. Microscopy of the mass revealed HCC. Abnormalities in nontumoral liver included cirrhosis, mild inflammation, and hepatocellular and canalicular cholestasis, with hepatocellular edema (Fig. 1D). The tumor exhibited clear-cell change, with intracellular inclusions consistent with glycoprotein (Fig. 1E). On immunostaining, tumor-cell cytoplasm and inclusions marked for AFP and tumor nuclei marked for p53 but not for β-catenin. Typing of microsatellite marker loci on 6 different chromosomes using DNA from lesion and from patient and donor leukocytes found the tumor to be of native liver rather than allogeneic-hepatocyte origin. Mutational analysis of ABCB11 using leukocyte DNA revealed compound heterozygosity for IVS16-8T>G, which induces direct splicing of exon 16 to exon 18, and 1939delA, which introduces a frameshift, with subsequent termination codon, in exon 16. The boy was healthy 32 months after LT.
Table 1. Patients, Clinical Courses, and ABCB11 / BSEP Mutations
Sibling(s) With PFIC
Age, PFIC Manifest
Age, HCC Diagnosed
Outcome to Date
A/Male; Northern European Caucasian
Cholestasis from 3 wk
Hepatocyte infusion, 16 mo
21 mo (incidental in explant; AFP 199, nl < 7), at LT
Healthy (2 y, 8 mo after LT)
1939delA/IVS16-8T>G (compound heterozygote)
647K then VFTSLX/splice site disruption
B/Female; Northern European Caucasian
Cholestasis from 2 wk, hospitalized for evaluation aged 12 wk
28 mo, at open biopsy; AFP not determined
Palliative care only; death aged 33 mo
IVS18+1G>A/74C>A (compound heterozygote)
Splice site disruption/S25X
C/Male; Northern European Caucasian
Cholestasis from birth
23 mo (AFP 30k, nl < 5; liver mass); histologic diagnosis at necropsy, 24 mo
Palliative care only; death aged 24 mo
1445A>G/3691C>T (compound heterozygote)
D/Male; Northern European Caucasian
Cholestasis from 3 wk
Partial external biliary diversion (PEBD), 11 mo; decreased pruritus temporarily, clinical-laboratory test results and growth no better
22 mo (AFP 158k, nl < 15); liver mass; lung and bone lesions; chemotherapy given; histologic diagnosis at LT, 25 mo
Death, 6 d after LT (sepsis; no HCC at necropsy)
E/Male; Northern European Caucasian
Growth failure from 6 mo; diagnosed 9.5 mo
PEBD, 10 mo; no response
29 mo (incidental in explant; AFP 6.4k, nl < 9), at LT
Healthy (5 y, 10 mo after LT)
IVS7+1G>A/890A>G (compound heterozygote)
Splice site disruption/E297G
F/Male; Northern European Caucasian
Cholestasis from 6 wk
16 mo (clinically unsuspected), at necropsy; AFP not determined
Death with metastatic HCC in lungs
Splice site disruption/not known
Cholestasis from 6 wk
15 mo (AFP 11k, nl < 7; liver mass); histologic diagnosis at LT, 16 mo
Healthy (1 y, 7 mo after LT)
H/Male; Northern European Caucasian
Evaluation at 6 mo for jaundice and growth failure
PEBD, 32 mo; excellent response (pruritus resolved, bile salts in serum normal range, growth resumed)
52 mo (marked increase in abdominal size; tumor metastasized at diagnosis; AFP 2x106, nl < 9), at open biopsy
Palliative care only; death 3 wk after diagnosis
890 A>G/IVS13del-13ˆ-8 (compound heterozygote)
E297G/splice site disruption
I/Male; Central Asian Caucasian
Cholestasis from birth
13 mo (incidental in explant; AFP 831, nl < 23), at LT
Healthy(1 y, 11 mo after LT)
Splice site disruption
J/Male; Central Asian Caucasian
Cholestasis from 1 wk
14 mo (AFP 4k, nl < 23; liver mass), at biopsy; confirmed at LT, 15 mo
Healthy (1 y, 2 mo after LT)
K/Male; Northern European Caucasian
Cholestasis from 3 mo
26 mo (marked increase in abdominal size; tumor metastasized at diagnosis; AFP not measured), at open biopsy
Death with metastatic HCC in lungs, aged 27 mo
This experience prompted a review of cases of HCC associated with PFIC at King's College Hospital and elsewhere. Ten additional young children with HCC associated with intrahepatic cholestasis were identified; all 11 children were unrelated (Table 1). Three of the children had been subjects of previous reports (patients B, F, and K).13, 16–20 Each patient exhibited conjugated hyperbilirubinemia without biliary tract obstruction in infancy. Liver biopsy in 9 patients revealed “neonatal hepatitis”; in 2 patients, each of whom had 3 affected siblings with biopsy-documented “neonatal hepatitis,” the parents refused biopsy. In the 9 children specifically assessed, GGT was low. These children also had no evidence of bile acid synthesis disorder on urine screening. Infection with hepatitis B virus or hepatitis C virus was not demonstrated in any of the patients. Tyrosinemia type I (TTI) and other metabolic disorders were excluded in all 11 patients. Serum concentrations of AFP were elevated in the 8 children specifically assessed. Five of the children died from HCC; the other 6 underwent LT. Of these 6, one died of sepsis, without tumor, shortly after LT, and 5 are well.
Paraffin-wax blocks containing tumor and nontumoral liver were retrieved; sections were stained and immunostained (see Methods section). Genomic DNA was obtained from patients' peripheral blood in 7 instances (patients A, D, E, G, H, I, and J) and from parents' peripheral blood in 10 instances (patients A-J). Mutational analysis of ABCB11 was undertaken (see Methods section). All studies were considered routine diagnostic assessment.
Tumor was available from all 11 patients with HCC. Nontumoral liver was available from 10 patients. In the exception (patient F),16, 17 whose sister also had intrahepatic cholestasis manifesting as “neonatal hepatitis”,18, 19 liver could not be retrieved from archives. The patient's lung, with metastatic tumor, and liver from his sister were used. Tissue sections cut at 4 μm were stained with hematoxylin-eosin and, after diastase digestion, with periodic acid — Schiff technique. Parallel sections were immunostained (DAKO Chem-Mate; DAKO, Ely, UK) with antibodies raised in rabbit against BSEP21 and raised in mouse against an ATP-binding cassette protein, MRP2 (Signet/Bioquote, York, UK). Sections containing tumor were similarly stained with antibodies against AFP (DAKO), p53 (DAKO), and β-catenin (Novocastra, Newcastle-upon-Tyne, UK), markers of stages in or routes toward malignant change in hepatocytes.22 The anti-BSEP antibody was raised, as described,11, 23 against an oligopeptide of the C-terminal 13 amino acids of BSEP (Neosystems, Strasbourg, France) and was affinity-purified (AminoLink Kit; Pierce Biotechnology, Boston, MA) against the same oligopeptide. Archival liver from 2 adults with Dubin-Johnson syndrome, in which MRP2 generally is lacking, and 30 children with cholestatic disease of known etiology (ATP8B1 mutation, n = 10; ABCB11 mutation, n = 5; ABCB4 mutation, n = 3; bile acid synthesis disorder, idiopathic acute liver failure, extrahepatic biliary atresia, α1-antitrypsin storage disorder, Alagille syndrome, and TTI, n = 2 each) also were evaluated for MRP2 and BSEP expression. (All commercial products were used according to the manufacturers' instructions.)
DNA was extracted from peripheral-blood leukocytes using the QIAamp blood DNA mini-kit (QIAGEN, Crawley, UK). ABCB11 was analyzed by sequencing of PCR products of all 28 coding and noncoding exons. Primer sequences for exonic amplification, available on request, included up to 50 bp of intronic flanking sequence and hence all sequences critical for mRNA splicing.
PCR amplification and product purification were performed using Roche FastTaq amplification and High Pure PCR purification systems (Roche Diagnostics Ltd, Lewes, UK). Bidirectional sequencing was performed using the 3.1 Dye Terminator Cycle Sequencing kit (Applied Biosystems, Warrington, UK), followed by ethanol precipitation and capillary gel electrophoresis on a 3100-Avant Genetic Analyzer (Applied Biosystems). Sequence analysis was performed using Sequencher software (Gene Codes, Ann Arbor, MI).
Microsatellite Typing (Patient A).
DNA was extracted from paraffin-embedded tumor using the QIAgen DNeasy Tissue kit (QIAGEN) as well as from leukocytes of both patient A and the hepatocyte donor. Fluorescently labeled primers were used to amplify informative microsatellite marker loci on 6 chromosomes. PCR products were separated on a 373 Automated DNA Sequencer and analyzed using Genescan and Genotyper software (all Applied Biosystems). Haplotypes for tumor and for patient and hepatocyte-donor leukocytes were constructed and compared.
In 6 patients with HCC, liver and tumor were sampled at LT (patients A, D, E, G, I, and J). In 3 patients, liver and tumor were sampled at necropsy (patients C, F, and K). In 2 patients with widespread disease, nontumoral liver was obtained at presentation and follow-up biopsies, whereas tumor was obtained at diagnostic laparotomy (patients B and H). All 11 patients had cirrhosis when tumor was diagnosed.
Tumors in the 6 explanted livers consisted of a single nodule in 4 patients (patients A, D, E, and I), 2 nodules in 1 patient (patient J), and 3 nodules in 1 patient (patient G). Tumor in 1 patient (patient A) lacked bile pigment and consisted of cells with cleared cytoplasm containing globules of eosinophilic material (Fig. 1E). Tumors were well-differentiated in 9 patients (Fig. 1F) and occasionally exhibited bile pigment at the centers of rosettes of cells resembling hepatocytes. In the patient with 3 tumor nodules (patient G), 2 of the nodules were well-differentiated, with bile production, and 1 was composed of clear cells. Neither mucin nor mucus was identified in any tumor. Although occasional hemopoietic cells were found within tumor, no lesion had the zonally biphasic smaller-cell (embryonal)/larger-cell (fetal) appearance of hepatoblastoma, and heterologous elements (osteoid, squamous epithelium, melanin) were found in none of the patients.
In 8 of the patients with HCC from whom nontumoral liver was available and in liver from the sister (who also had severe intrahepatic cholestasis) of the patient with HCC from whom nontumoral liver was not available (patient F), BSEP was not detected immunohistochemically at any site in nontumoral liver. In nontumoral liver from another patient with HCC (patient H), scant staining for BSEP was seen along occasional canaliculi. In yet another patient (patient K) and his brother, BSEP expression was intact. MRP2 was well-expressed at canalicular margins in all nontumoral liver. BSEP expression was not found in any tumor except in patient K.
All tumors, however, expressed MRP2 at margins of rosettes or, at cell borders, in linear patterns consistent with canalicular margins. Canaliculi in all samples from the panel of comparison materials but those with Dubin-Johnson syndrome expressed MRP2. Canaliculi in all samples from the panel but those with known ABCB11 mutation expressed BSEP. AFP was demonstrated in cytoplasm in all tumors. Nuclei in all tumors but 3 (patients B, J, and K) stained for p53; β-catenin accumulation in nuclei was not detected in any of the tumors.
Mutation in ABCB11 was found on both alleles in 9 families and on 1 allele in the family of patient F, from which only maternal leukocytes were available. In 7 patients, mutations were found on analysis of proband leukocyte DNA; these were confirmed in parental material for each. With respect to the other 4 patients, from whom no peripheral blood leukocyte DNA was available (patients B, C, F, and K), mutation in ABCB11 was found in material from the 5 parents who consented to genetic studies (both parents of patient B, both parents of patient C, and only one parent of patient F). The parents of patient K could not be traced, and mutational analysis was not attempted.
A total of 13 different mutations, 10 novel, was identified (Table 1). The common Polish mutation 1445A>G (D482G)8 was present in one parent of a patient from whom leukocytes were not available (patient C). The common European missense mutation 890A>G (E297G)8 was present in 3 patients and their parents (patients D, E, and H). Except for 890A>G (E297G), none of the changes found was present in 500 control individuals representative of all major populations (University of California San Francisco Pharmacogenetics Project24; http://pharmacogenetics.ucsf.edu).
Microsatellite Typing (Patient A).
At all marker loci studied, haplotypes constructed for tumor matched those for patient DNA rather than for hepatocyte-donor DNA.
Several descriptions exist of HCC in early childhood associated with “giant-cell hepatitis” or “neonatal hepatitis”.13, 16–20, 25, 26 Two such children had one sibling with fatal intrahepatic cholestasis,16–19, 26 and another had 3 siblings with intrahepatic cholestasis, one of whom died from HCC in adolescence.20 HCC also has been described in older children or children of unstated age, adolescents, or young adults with “neonatal hepatitis”, familial cholestatic cirrhosis of childhood, Byler disease, or PFIC.20, 27–32 What causes giant-cell hepatitis or “neonatal hepatitis” in these children, or what sort of PFIC affects them, is a matter of debate.
In a search of published instances of HCC in early childhood and in a review of materials at five pediatric hepatology centers, we identified 11 children, 3 of whom had been subjects of the case reports cited above,13, 16–20 in whom “neonatal hepatitis” and persistent cholestasis were associated with development of HCC at less than 52 months of age and from whom (and whose families) archival materials or blood samples were available for study. Archival materials for 2 other similar children25, 26 were sought but were unavailable (T. Higgins and S. Falkmer, personal communications). Attempts to gain access to materials from a child with PFIC and with HCC manifest at an unstated age32 were unsuccessful.
Liver disease in 9 of the 11 patients met the clinicopathological criteria set forth for the diagnosis of PFIC. In the other 2 patients (patients F and K),16–20 GGT was not measured, and primary disorders of bile acid synthesis were not excluded. In all 11, clinicopathological findings suggested BSEP deficiency. In 9, BSEP could not be demonstrated immunohistochemically at canalicular margins (8 patients) or was present but very scant (patient H). In patient F,16, 17 nontumoral liver was not available for assessment of BSEP expression. The liver of a sister with cholestatic liver disease fatal in childhood and initially manifest as “neonatal hepatitis”,18, 19 however, entirely lacked immunohistochemically demonstrable BSEP. Liver disease in this sibling pair thus was assessed as likely due to BSEP deficiency. In 10 of 11 instances of HCC in early childhood associated with clinically diagnosed PFIC, deficiency of BSEP expression was directly demonstrated or could be reasonably inferred. The exception was patient K20 (see below).
Lesions in ABCB11 predicted to disrupt synthesis of functional BSEP were identified in all 10 families studied by mutational analysis (Table 1). Mutation was found on both alleles of ABCB11 in 9; paternal DNA was not available for the exception (patient F), and only a maternal change could be identified. Two ABCB11 mutations were found in each of the 7 children with deficiency of BSEP expression from whom peripheral blood leukocytes were available. With respect to the other 3 children with proven or likely deficiency of BSEP expression, each of the 5 parents whose DNA could be studied proved to carry a single ABCB11 change. The lack of immunohistochemically detectable BSEP in liver tissue from children B and C and from the sibling of child F argues strongly that B, C, and F each carried 2 defective ABCB11 alleles. In all 10 of the patients whose liver disease was associated with BSEP deficiency, then, mutation in ABCB11 was directly demonstrated or could be reasonably inferred.
No deficiency of BSEP expression was found in liver from either patient K or his older brother, both of whom had clinically diagnosed PFIC and both of whom died of HCC.20 Liver disease in this sibling pair thus was assessed as not likely due to BSEP deficiency. The status of patient K with respect to ABCB11 mutation is unknown. Though lesions in ABCB11 impeding BSEP function but not BSEP expression may have been present, the liver disease clinically diagnosed as PFIC and manifest as “neonatal hepatitis” in that child and his siblings20 perhaps represented a disease different from BSEP deficiency. The definition of PFIC now generally in use1 did not yet exist 30 years ago, when patient K and his siblings were evaluated: GGT values and results of bile acid synthesis defect screening, in particular, were not used to lessen heterogeneity within a wide mix of cholestatic disorders.
The polyclonal, affinity-purified anti-BSEP antibody that we employed21 identified appropriately distributed canalicular BSEP in patients with a variety of liver diseases of known etiology other than mutation of ABCB11. As in a previous study using snap-frozen tissue,9 we assessed nonspecific preservation of canalicular antigens in terms of expression of MRP2, which, like BSEP, is an ATP-binding cassette protein and a canalicular transporter. MRP2 expression was present and unremarkable in all materials except, as expected, those from patients with Dubin-Johnson syndrome.
In all 10 children with HCC and deficiency in expression of BSEP, mutational analysis in ABCB11 found lesions predicted to abrogate synthesis of functional BSEP. Failure to express immunohistochemically identifiable BSEP, or severe deficiency in BSEP expression (patient H), thus was paired with mutation in ABCB11 in every instance studied. All patients with mutation in ABCB11 expressed immunohistochemically identifiable MRP2. We conclude that MRP2 served as an appropriate technical control. We also conclude that lack of expression of BSEP, when assessed immunohistochemically, correlated well with demonstration of mutation in ABCB11.
We could not associate progression to HCC with any particular kind of mutation in ABCB11, although splice-site mutations formed a plurality among the 13 mutations identified in the 10 families genetically studied. The various predicted effects (Table 1) include splicing defects (patients A, B, E, F, H, and I), a frameshift (patient A), missense changes at conserved amino acid residues (patients C, D, E, and H), and direct introduction of premature termination codons (patient B, exon 2; patient G, exon 13; patient J, exon 19). The frameshift, which was found in exon 16, generates 6 altered amino acid residues followed by premature termination. Among the missense changes is 1 in exon 27 (patient C); those in patients D, E, and H (890A>G [E297G], exon 9) and the other mutation in patient C (1445A>G [D482G], exon 14) have been described previously.8 The 890A>G (E297G) and 1445A>G (D482G) mutations reportedly affect BSEP transport activity,21, 33 with altered trafficking of BSEP to the canalicular membrane.33–35 The mutation in patient C, 3691C>T (R1231W), is predicted to alter an amino acid residue between the ATP-binding cassette signature motif and the Walker B motif of BSEP. This residue is conserved in all MDR subfamily members and in the related transporters MRP2 and CFTR.
One of the mutations predicted to alter splicing has been described. IVS18+1G>A, found in patient B, was (in combination with 3148C>T [R1050C]) associated with “benign” recurrent intrahepatic cholestasis (BRIC)—namely, intrahepatic cholestasis with intermittent clinical manifestations (published as IVS19+1G>A through typographical error [L. Klomp, personal communication]).36 IVS16-8T>G, found in patient A, has been associated with PFIC in several other patients. This mutation leads to skipping of exon 17 with an associated frameshift and introduction of 5 amino acid residues followed by protein truncation (unpublished data). IVS13del-13ˆ-8 removes 6 bases of the 3′ splice site and changes the location of the branch point sequence. IVS18+1G>A and the other 3 splice site changes (Table 1) affect the invariant GT 5′ donor splice site consensus sequences and are predicted to lead to exon skipping and associated frameshifts or to cryptic splice site activation.
Of interest is the difference in phenotype between patient B, with PFIC in association with the genotype IVS18+1G>A / 74C>A (S25X), and a patient with BRIC whose less severe disease was associated with the IVS18+1G>A / 3148C>T (R1050C) genotype.24 Factors modulating penetrance of ABCB11 mutation have yet to be defined. Also of interest is the presence among our patients of the 890A>G (E297G) mutation (patient D, homozygous; patients E and H, heterozygous) and the 1445A>G (D482G) mutation (patient C, heterozygous). The former can be associated with either PFIC or BRIC8, 24; preliminary observations suggest association of both with favorable response to PEBD.
Our patients D, E, and H, all with 890A>G (E297G) mutation, came to PEBD. Patients D and E had no immunohistochemically identifiable BSEP; patient H retained some expression, which was very scant. Patient H responded well to PEBD, but HCC developed nonetheless. Close monitoring appears in order for BSEP-deficient PFIC even when clinical response to PEBD is good. It may also be in order for BSEP-deficient BRIC.
Our patients' tumors were not morphologically uniform. No tumor met criteria for the diagnosis of hepatoblastoma or cholangiocarcinoma. (A hepatocellular malignancy interpreted as hepatoblastoma has been described in association with PFIC; we have demonstrated deficiency of BSEP expression in ambient liver, but consider the tumor more likely HCC.37) In 9 of the 10 patients with demonstrated or inferred deficiency of BSEP expression, the single lesion was a well-differentiated HCC with bile production. Clear-cell features were found in the other patient with a single lesion (patient A). In the patient who had 3 tumors (patient G), 1 had clear-cell features; and 2 were well-differentiated HCC.
Common characteristics of the tumors, however, were increased synthesis of AFP and nuclear accumulation of p53 protein. None of the tumors exhibited nuclear accumulation of β-catenin. Although these observations may implicate specific pathways toward malignancy,22 they identify no particular carcinogenic species or event. Of interest is a preliminary report of cholangiocarcinoma in 2 older children with PFIC and ABCB11 mutation.38 This may indicate mutagenesis affecting cells that can differentiate along either hepatocellular or cholangiocytic lines.
The mechanism of hepatocarcinogenesis in BSEP deficiency is unclear. Perhaps hepatocarcinogenesis in BSEP deficiency or malfunction is among nonspecific consequences of various mutations in ABCB11. One such consequence—increased intracellular concentrations of bile acids—may be mutagenic.39 This, however, fails to account for the dearth of HCC in other cholestatic and hypercholanemic disorders of infancy such as extrahepatic biliary atresia.13
It is also possible that TTI provides an analogy. In TTI, a wide range of mutations in fumarylacetoacetate hydrolase can result in inhibition of DNA ligase 1 by succinylacetone, leading to accumulation of genetic injury.40 Any specific agent predisposing to malignancy in BSEP deficiency has yet to be identified.
Of interest in BSEP deficiency, and distinct from TTI, is the speed with which HCC may develop. Although pharmacotherapy and intervention with LT have reduced the proportion of children with TTI who develop HCC to as low as 10%,41, 42 HCC historically occurred in 37% of children with TTI aged > 2 years.43 Among the cases reviewed, although dates of diagnosis are not cited, HCC led to death in no patient aged < 4 years.43 Instances of HCC manifesting before the age of 24 months in TTI appear rare.44, 45 By contrast, 7 of our 10 patients with BSEP deficiency and HCC were aged < 24 months at diagnosis of HCC.
PFIC—whether due to deficiency of FIC1, deficiency of BSEP, or other causes—is treated principally with supportive measures. It can be argued that to identify BSEP deficiency as underlying PFIC has few practical implications for care. Our findings imply that, on the contrary, identification of BSEP deficiency may be important. We suggest that monitoring for development of HCC is indicated in BSEP deficiency, perhaps with determinations of serum concentrations of AFP and with sonography. Whether such monitoring is indicated in FIC1 deficiency is an open question, though we know of no instance of HCC associated with that condition.
Our observations also call into question treatment with allograft hepatocyte infusions14 or even gene therapy in PFIC owing to BSEP deficiency. To establish clones of exogenous or modified hepatocytes that secrete bile acids may not only effectively treat cholestasis and pruritus but also reduce risk of malignancy liver-wide. However, because this approach may leave BSEP-deficient—and possibly premalignant—hepatobiliary cells in place, it is perhaps less desirable than LT.
In conclusion, mutation in ABCB11, with deficiency of immunohistochemically identifiable BSEP, increases risk of HCC in early life. Immunohistochemical deficiency of BSEP in PFIC is closely correlated with demonstrable mutation in ABCB11. These findings suggest approaches to more efficient diagnosis of PFIC associated with ABCB11 mutation and warrant particular care in the observation and management of patients with PFIC.
We thank Anne Rayner for excellent technical work.