Potential conflict of interest: Nothing to report.
The gene encoding the human bile salt export pump (BSEP), ABCB11, is mutated in several forms of intrahepatic cholestasis. Here we classified the majority (63) of known ABCB11 missense mutations and 21 single-nucleotide polymorphisms (SNPs) to determine whether they caused abnormal ABCB11 pre-messenger RNA splicing, abnormal processing of BSEP protein, or alterations in BSEP protein function. Using an in vitro minigene system to analyze splicing events, we found reduced wild-type splicing for 20 mutations/SNPs, with normal mRNA levels reduced to 5% or less in eight cases. The common ABCB11 missense mutation encoding D482G enhanced aberrant splicing, whereas the common SNP A1028A promoted exon skipping. Addition of exogenous splicing factors modulated several splicing defects. Of the mutants expressed in vitro in CHO-K1 cells, most appeared to be retained in the endoplasmic reticulum and degraded. A minority had BSEP levels similar to wild-type. The SNP variant A444 had reduced levels of protein compared with V444. Treatment with glycerol and incubation at reduced temperature overcame processing defects for several mutants, including E297G. Taurocholate transport by two assessed mutants, N490D and A570T, was reduced compared with wild-type. Conclusion: This work is a comprehensive analysis of 80% of ABCB11 missense mutations and single-nucleotide polymorphisms at pre-mRNA splicing and protein processing/functional levels. We show that aberrant pre-mRNA splicing occurs in a considerable number of cases, leading to reduced levels of normal mRNA. Thus, primary defects at either the protein or the mRNA level (or both) contribute significantly to BSEP deficiency. These results will help to develop mutation-specific therapies for children and adults suffering from intrahepatic cholestasis due to BSEP deficiency. (HEPATOLOGY 2008.)
The human bile salt export pump (BSEP) is situated exclusively in the canalicular membrane of hepatocytes and is responsible for the translocation of bile salts from these cells into the canalicular space.1 BSEP is a 12-transmembrane span integral membrane protein belonging to the adenosine triphosphate–binding cassette (ABC) transporter superfamily.2 Previous characterization of recombinant human BSEP demonstrated a high affinity for taurocholate, glycochenodeoxycholate, and glycocholate, the main bile salts of human bile.3, 4 These affinities differ from those demonstrated for the BSEP orthologs of rats and mice and reflect the bile salt pool differences among these mammals.5, 6
Mutations in ABCB11 are associated with a phenotypical spectrum of autosomal recessive cholestatic liver diseases. Severe BSEP deficiency is a form of progressive familial intrahepatic cholestasis (PFIC) that presents in early childhood with neonatal hepatitis, leading to progressive cholestasis. End-stage liver disease is reached by the second decade, and liver transplantation is required.7 Currently, this is the only cure. Benign recurrent intrahepatic cholestasis (BRIC) patients typically have periods of cholestasis interspersed with symptom-free periods, and several ABCB11 mutations are uniquely associated with BRIC.8 Intrahepatic cholestasis of pregnancy (ICP) presents in the third trimester, can result in death of the unborn child, and may not resolve after the birth.9 The onset of acquired drug-induced cholestasis (DC) and oral contraceptive-induced cholestasis (CC) is unpredictable but they can resolve on removal of the causative drug or hormone.10
In studies of how clinically severe BSEP deficiency is mediated, several predicted ABCB11 missense mutations associated with this disorder have previously been introduced into equivalent positions in rat Abcb11, and their effects on Bsep were assessed in vitro in both Madin-Darby canine kidney (MDCK) and insect cells.11 Most of the seven mutations studied promoted retention of Bsep inside MDCK cells whereas taurocholate transport assays using insect cell membranes showed reduced functional activity in six cases. Species differences in BSEP, however, could result in differences in mutant behavior. Indeed, subsequent studies have shown differing results with respect to membrane localization and taurocholate transport for several mutations, depending on whether rat or human proteins were expressed.11–15
It cannot necessarily be assumed that a missense mutation causes disease by exchanging an important amino acid in a protein; the primary defect may actually be at the level of the production or stability of the gene's messenger RNA (mRNA).16, 17 Evidence has emerged that missense mutations located in the middle of exons, and therefore well away from the classical intron-exon boundaries, can cause altered splicing events.18 The same could be true for a synonymous or a nonsynonymous single-nucleotide polymorphism (SNP). In particular, any exonic nucleotide changes may abrogate exonic splicing enhancers (ESE) that are necessary for efficient splicing of the exon in which they reside, thus inducing exon skipping.19 Alternatively, mutations may create an exonic splicing silencer (ESS) or alter pre-mRNA secondary structures that are involved in exon definition, thus affecting the splice site that is used.20, 21 To define the molecular nature of exonic substitutions correctly, pre-mRNA splicing abnormalities and mRNA instability must therefore be investigated. Splicing events are overlooked by in vitro protein expression experiments, because complementary DNA (cDNA) constructs are ‘artificial’ templates in which splicing does not occur. However, in vitro minigene systems can be used to characterize such events.22
It has been documented in numerous human diseases that abnormal intracellular trafficking of mutant proteins is common.23 In most cases, mutant proteins are retained in the endoplasmic reticulum (ER) and degraded, such as for example, cystic fibrosis transmembrane regulator (CFTR) containing the ΔF508 mutation.24 It is of great clinical importance to determine whether any missense mutations in ABCB11 are associated with abnormal BSEP trafficking and then, in turn, whether the mutant proteins can be encouraged to reach the canalicular membrane and be functional there. The function can be determined using a baculovirus expression system previously used to assess wild-type BSEP function.3 Additionally, mutations causing mRNA destabilization or altered splicing may potentially be overcome by the development and administration of stabilizing agents.18
The experiments described in this study aim to provide a greater understanding of the mechanisms by which ABCB11 missense mutations cause human disease. Most known ABCB11 missense mutations and 21 common and disease-associated SNPs were analyzed for their effect on pre-mRNA splicing using a minigene system, and 37 clinically relevant mutant and five polymorphic proteins were expressed in vitro in Chinese Hamster Ovary (CHO)-K1 cells. Finally, bile salt transport studies in insect cells were used to assess any defects in BSEP function.
HepG2 cells (European Collection of Cell Cultures) were grown in Eagle's minimal essential medium (Cambrex, East Rutherford, NJ) supplemented with 1% (vol/vol) minimal amino acids, 1 × Glutamax, and 10% (vol/vol) fetal bovine serum (all Invitrogen Corporation, Paisley, UK) at 37°C in 5% CO2. CHO-K1 cells (European Collection of Cell Cultures) were grown in Dulbecco's modified Eagle's medium:Ham's F12 50% (vol/vol) containing 5% (vol/vol) fetal bovine serum (both Invitrogen Corporation) at 37°C in 5% CO2.
Minigene Construct Preparation.
ABCB11 exons and 500 bp of intronic sequence were amplified by polymerase chain reaction (PCR) using Pwo Superyield (Roche, Lewes, UK) and cloned into pCR-TOPO-II-blunt (Invitrogen Corporation). Clones were sequenced using Big Dye Version 3.1 (Applied Biosystems, Warrington, UK) on an ABI 3100-Avant automated sequencer (Applied Biosystems). Mutagenesis primers were designed to introduce the nucleotide sequence of 62 predicted missense mutations and 21 SNPs in human ABCB11 (Table 1; primer sequences are available from the authors on request) and site-directed mutagenesis was performed in the appropriate exon using Quick Change II Site-directed Mutagenesis Kit (Stratagene/Agilent Technologies, La Jolla, CA). After sequencing, the wild-type and variant exons were subcloned into the fibronectin intron present in the minigene vector pTB-Nde-minus.22, 25
Table 1. ABCB11 Missense Mutations and SNPs Functionally Analyzed in This Study
Four × 105 cells were plated in 60-mm dishes and 24 hours later were transfected with 3 μg of wild-type or variant ABCB11 minigene construct, or with the empty native minigene, using FuGene 6 (Roche), with a lipid:DNA ratio of 3:1. Cells were incubated for 48 hours before the isolation of total RNA using RNA-Bee, following the manufacturer's instructions (Biogenesis, Poole, UK). Precipitated RNA was deoxyribonuclease-treated (Promega, Southampton, UK), quantitated using a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Willington, DE), and 2.7 μg was reverse transcribed using Transcriptor Reverse Transcriptase (Roche) with 1 μg of random hexamer primers (Invitrogen Corporation).
For modulation of splicing experiments, 500 ng of plasmid that was empty or that contained the cDNA of the splicing factors SC35, heterogeneous nuclear riboprotein (hnRNP) A2 and hnRNPH25 was co-transfected with 3 μg of the appropriate ABCB11 minigenes. The lipid:DNA ratio was maintained at 3:1.
Two microliters of cDNA prepared from each minigene transfection was PCR amplified using VentR polymerase (New England Biolabs, Hitchin, UK) using the primers mFN-forward and mFN-reverse.25 Products were run on 2% agarose gels to analyze the resultant sizes and also cloned for sequencing purposes.
Taqman major groove binder probes labeled with FAM and with a nonfluorescent quencher (Applied Biosystems) were designed to the appropriate minigene-ABCB11 or ABCB11-minigene splice form and the native minigene using Primer Express software (Applied Biosystems; probe and primer sequences are available from the authors on request). Real-time PCR was performed as previously described,26 except that data were normalized to HPRT1 expression [delta (d) Ct] and plotted as fold difference of the mean Ct values (dd Ct) compared with the wild-type exon.
ABCB11 cDNA Construct Preparation.
A hemagglutinin tag was introduced to the 3′ end of the ABCB11 cDNA by PCR, replacing the histidine tag previously added.3 The cDNA was cloned into the transient mammalian expression vector pCI (Promega) and the baculovirus transfer vector pBlueBac4.5 (Invitrogen Corporation). Because of the instability of the human cDNA in Escherichia coli,3 all ligations and plasmid propagations were performed at 30°C in Stb12 MAX Efficiency E. coli (Invitrogen Corporation). Mutagenesis primers were designed to the nucleotide sequences predicted to result in 37 missense mutations and five SNPs in human ABCB11 (Fig. 5 and Table 1; primer sequences from the authors on request). Site-directed mutagenesis was performed, as described, in the ABCB11 cDNA, before the creation of full-length mutant ABCB11 cDNA constructs in the appropriate vector.
Transient Transfection of CHO-K1 Cells.
Eight × 105 cells were seeded in T25 cm2 flasks and 24 hours later transfected with 3 μg of each ABCB11 cDNA construct in pCI, using FuGene 6 (Roche), with a lipid:DNA ratio of 3:1. Cells were incubated for 48 hours before harvest. For glycerol experiments, 24 hours post-transfection either 500 μL of medium or 500 μL of glycerol (Sigma Aldrich, Gillingham, UK) were added to cells, which then were moved to 26°C for a further 24 hours before harvest. Membrane fractions were prepared essentially as previously described.26, 27 The blots shown are representative of two separate transfections for each mutant. For each set of transfections and subsequent membrane preparations, there was always a control experiment with wild-type BSEP, and so the mutants were always standardized to the 100% expression level of wild-type processed at the same time.
Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis and Western Blotting.
Samples of membrane fractions (1.5 μg each) were reduced and subjected to 7.5% sodium dodecyl sulfate polyacrylamide gel electrophoresis in duplicate gels. One gel was stained with Coomassie blue to confirm uniformity of protein loading among samples. The contents of the other gel were western blotted as previously described,26 except using a 1:300 dilution of an anti-hemagglutinin (Y-11) polyclonal primary antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and a 1:1,000 dilution of horseradish peroxidase–conjugated goat anti-rabbit secondary antibody (DAKO, Ely, UK).
Preparation of Recombinant Baculoviruses.
Sf9 and High 5 insect cells were grown in Sf900-II SFM and Express Five SFM, respectively, at 27°C. Recombinant ABCB11 baculoviruses were created by the recombination of the Bac-N-Blue Baculovirus genome (Invitrogen Corporation) with the appropriate ABCB11 cDNA in pBlueBac4.5.3 Membrane vesicles were prepared as described3 and were subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis and western blotting, as above.
The velocity of taurocholate transport was determined for wild-type and mutant BSEP using 6 μM 3H-taurocholate (Perkin Elmer LAS, Beaconsfield, UK) in transport assays, performed as described.3
In Vitro Analysis of Splicing Defects Associated with Mutant ABCB11 Minigenes.
Table 1 lists the ABCB11 nucleotide substitutions that were analyzed for their effect on pre-mRNA splicing: they are predicted to cause missense mutations or SNPs. The mutations are associated with PFIC,7, 28–36 BRIC,8, 12, 37, 38 or ICP, CC, and DC.39–42 The SNPs have been documented in the normal population in the studies mentioned previously, or in patients with primary biliary cirrhosis or primary sclerosing cholangitis.43–45
In vitro splicing of human ABCB11 was assessed in HepG2 cells, shown to be capable of expressing ABCB11,46 using a minigene system (illustrated in Fig. 1) previously used to characterize the splicing of ABCC7 (encoding CFTR).22, 25 The inclusion or exclusion of ABCB11 exons, whether wild-type, mutant, or polymorphic, within the fibronectin exon-bordering domain (EBD) exon was determined by reverse transcription polymerase chain reaction (RT-PCR). Sizing and sequencing of products could also detect aberrantly spliced variants. This latter information allowed the design of real-time PCR probes specific to splice variants and thus permitted quantitation of the various species (the results of which are summarized in Supporting Table 1). Fig. 2A-F shows the resultant PCR products from minigene transfections. The appropriate wild-type ABCB11 exon results are shown to the left of all mutant versions of that exon. Sequencing of a faint PCR product of 375 bp observed in some lanes revealed that this band was a nonspecific product (data not shown).
Our results show that several mutations increased exon skipping and thereby reduced the amount of normally spliced product, shown by an accumulation of the product generated by the native minigene (N). Notably, exon 7 nucleotide sequences encoding the predicted mutations A167V (c.500C>T) and E186G (c.557A>G) resulted, respectively, in only 80% and 60% wild-type splicing (Fig. 2A). To varying degrees, exon skipping was also seen for the nucleotide changes associated with other missense mutations: T463I (c.1388C>T; 90% wild-type splicing; Fig. 2C), A926P (c.2776G>C; 90%; Fig. 2E), A1110E (c.3329C>A; 90%), G1116R (c.3346G>C; 80%), R1128C (c.3382C>T; 85%), R1128H (c.3383G>A; 90%), and R1231W (c.3691C>T; 5%; Fig. 2F). A small amount of exon skipping was also seen for the SNPs G319G (c.957A>G; 95%; Fig. 2B) and E1186K (c.3556G>A; 90%; Fig. 2F). Additionally, the SNP associated with primary biliary cirrhosis, A1028A (c.3084A>G), exhibited the predominant exclusion of ABCB11 exon 24 (5%; Fig. 2E). The ESE prediction program RESCUE ESE47 indicates that an ESE is abolished in the presence of the variant G allele, whereas ESE Finder48, 49 shows that this allele results in an enhanced SRp40 binding site (score 2.86 versus 4.30).
Differential splice products were seen for six predicted ABCB11 missense mutations, leading to very low levels of wild-type splicing (indicated in parentheses): E297K (c.889G>A; 50%; Fig. 2B), D482G (c.1445A>G; 5%; Fig. 2C), R832C (c.2494C>T; 50%; Fig. 2E), and S1144R (c.3432C>A; 3%), R1153H (c.3458G>A; 3%), and S1154P (c.3460T>C; 3%; Fig. 2F). Figs. 3A through D illustrate the intraexonic alternative splice sites that were used in the presence of these nucleotide changes, determined from sequencing of the generated PCR products. No spliced ABCB11 PCR product was generated for the nucleotide changes associated with T586I (c.1757C>T; 0%; Fig. 2D) and R1231Q (c.3692G>A; 0%; Fig. 2F).
The Splice Site Prediction by Neural Network program50, 51 shows that the normal donor splice site at the exon 9–intron 9 boundary has a probability score of 0.62. A cryptic donor site adjacent to the nucleotide sequence of codon E297 (Fig. 3A) has a score of 0.68. On changing the wild-type exon 9 nucleotide sequence to the sequence of E297G (c.890A>G), the cryptic donor splice site disappears. However, for the nucleotide change that results in E297K (c.889G>A), the score for the cryptic donor splice site rises to 0.99, with the normal donor splice site being maintained at 0.62. Indeed, it can be seen that the normal and the cryptic splice sites are equally used by the E297K-containing exon 9 (Fig. 2B). Additional analysis of the exon 9 variant sequences to identify changes to ESE or ESS motifs was performed using RESCUE-ESE and ESE FINDER 2.0 or the FAS-ESS programme, respectively.47–49, 52, 53 An SC35 site was identified, which was maintained irrespective of which nucleotide form was analyzed, but the presence of sequence coding for either E297K or E297G resulted in the introduction of an ESS hexamer.
Scanning of exon 14 with the ESS prediction programme FAS-ESS52, 53 indicates that the mutant D482G (c.1445A>G) nucleotide sequence causes the introduction of two additional ESS hexamers, whereas the adjacent V481E (c.1442T>A) does not. This agrees with the observed degree of use of the cryptic splice site in exon 14 of these two mutations (Fig. 2B).
The cryptic acceptor splice site in exon 26 has a score of 0.23, according to the Splice Site Prediction by Neural Network program.50, 51 The mutations S1144R (c.3432C>A), R1153H (c.3458G>A), and S1154P (c.3460T>C) do not alter the predicted score for use of this site. Additionally, the mutations cause no alteration in the presence of ESS hexamers in this exonic location.52, 53
Modulation of Aberrant ABCB11 Splicing by Cellular Splicing Factors.
Positive-acting and negative-acting splicing factors affect the efficiency of splicing. The potential to modulate a selection of aberrant ABCB11 pre-mRNA splicing was assessed through the addition of the splicing factors SC35, hnRNPA2, and hnRNPH.25 The effect on the nucleotide change predicted to cause E1186K (c.3556G>A) was also assessed, as the role of this SNP in the development of disease is unclear. Results were confirmed by quantitative RT-PCR (data not shown).
Figure 4A illustrates that the addition of splicing factor SC35 increased the inclusion of ABCB11 exon 7 containing nucleotide sequence for E186G (c.557A>G) to the same level as wild-type exon 7, whereas that of the unspliced product remained constant. Additionally, the differential splicing associated with E297K (c.889G>A) was abolished on the addition of SC35, but on quantitative analysis was slightly enhanced with hnRNPA2 (2.5-fold) and hnRNPH (eightfold).
No added splicing factor substantially altered the cryptic splicing pattern associated with the D482G (c.1445A>G) nucleotide substitution (Fig. 4A). SC35 resulted in a twofold reduction of T463I (c.1388C>T) splicing (Fig. 4B). Splicing of exon 25 containing the nucleotide sequence encoding G1116R (c.3346G>C) was improved on the addition of SC35, with a reciprocal reduction in unspliced product, but no change was associated with hnRNPA2 and hnRNPH (Fig. 4B). The wild-type splicing of R1128C (c.3382C>T) was enhanced by these latter factors, however (Fig. 4B). None of the factors altered the splicing of A167V (c.500C>T; Fig. 4A), or R1231W (c.3691C>T; Fig. 4C). Finally, the splicing of the SNP E1186K (c.3556G>A) was maintained at 90% of wild-type levels on the addition of hnRNPA2 and hnRNPH, but was inhibited by SC35 (Fig. 4C).
Expression of BSEP Mutants to Assess Protein Processing Defects.
Wild-type, mutant, and polymorphic forms of human BSEP were transiently expressed in CHO-K1 cells to assess their effect on BSEP expression. The variants shown in Fig. 5 (details in Table 1) can be defined as (1) predicted missense mutations present in more than one family, (2) missense mutations associated with mutation-prone CpG dinucleotides, (3) SNPs with disease associations, and (4) missense mutations in which, depending on the nucleotide change in the codon, the wild-type amino acid has been substituted with different amino acids.
Mature wild-type human BSEP, in vivo or expressed in mammalian cells, has an apparent molecular weight of 160 kDa (Fig. 5A), 20 kDa larger than the protein expressed in insect cells. This is attributable to known differential glycosylation between mammalian and insect cells.3 Observed lower molecular weight forms of the wild-type protein correspond to newly made, partially processed BSEP within the biosynthetic pathway. Mutant proteins, which are not appropriately folded, are retained in the ER, and processing of N-linked glycans does not occur. Lower molecular weight forms are therefore diagnostic for ER retention of BSEP, excepting protein C-terminus deletions, which would also lead to a lower molecular weight. Certainly, a previous study confirmed, through N-glycoside F (PNGaseF) treatment, that human BSEP has an ER-associated form.13
Variant BSEP forms found in this study to have the same molecular weight and about the same abundance as the wild-type protein are I498T (c.1493T>C; Fig. 5A), N1211D (c.3631A>G), L1242I (c.3724C>A), E709K (c.2125G>A; Fig. 5C), T586I (c.1757C>T; Fig. 5E), as well as the SNP N591S (c.1772A>G; Fig. 5F). Mature BSEP, but at somewhat reduced levels, was detected for E135K (c.403G>A; Fig. 5B), T655I (c.1964C>T; Fig. 5C), R432T (c.1295G>C; Fig. 5.e), the PFIC-associated SNP Y818F (c.2453A>T; Fig. 5E), and the SNP A535A (c.1605C>T; Fig. 5F). The mutations N490D (c.1468A>G; Fig. 5A), R1128H (c.3383G>A; Fig. 5B), E297G (c. 890A>G), and A570T (c.1708G>A; Fig. 5D), and E186G (c.557A>G; Fig. 5E) resulted in significantly reduced levels of mature protein, and the L198P (c.593T>C) variant was barely detectable (Fig. 5C). In most of these cases, immature BSEP was also visible.
No BSEP protein, mature or immature, was observed for the mutations V284L (c.850G>C) and A588V (c.1763C>T; Fig. 5A) and I541L (c.1621A>C; Fig. 5B), indicating that these variants were rapidly degraded. For the remaining ABCB11 missense mutations little mature 160 kDa protein was seen, but instead low-molecular-weight bands indicative for immature protein. These bands were usually much less abundant than wild-type protein, suggesting that these mutants were largely retained in the ER and degraded.
The SNP V284A (c.851T>C) led to increased amounts of protein compared with wild-type BSEP (Fig. 5A). Similarly, the SNP A444V (c.1331C>T) resulted in enhanced levels of mature BSEP (Fig. 5F). This SNP is prevalent in various populations at a frequency of between 40% and 50%; the A444 form is associated with an increased incidence of several forms of intrahepatic cholestasis.37, 39–41
Modulation of Mutant Protein Processing by Glycerol and Low Temperature.
The addition of glycerol and incubation at a reduced temperature are conditions that promote, in cultured cells, the correct processing of membrane proteins trapped in the ER or Golgi.54–56 In an initial experiment, the effects of glycerol alone, incubation at 28°C alone, or glycerol and 28°C in combination on wild-type BSEP, R948C (c.2842C>T), and E297G (c. 890A>G) were assessed. Fig. 6A illustrates that, for the wild-type protein, the addition of 10% glycerol with a reduced incubation temperature enhanced the levels of mature BSEP. There was no such effect by any treatment for R948C, but for E297G, incubation at 28°C alone or in combination with 10% glycerol gave the same levels of 160 kDa BSEP as the wild-type protein.
Additional mutations associated with a predominance of immature protein (Fig. 5) were then incubated at 28°C with 10% glycerol (Fig. 6B-D). The following mutations showed an enrichment of mature BSEP: R1153H (c.3458G>A; Fig. 6B), R1128C (c.3382C>T), R1128H (c.3383G>A), R1231Q (c.3692G>A), R1050C (c.3148C>T; Fig. 6C) and R1268Q (c.3892G>A), A570T (c.1708G>A), and E297K (c.889G>A; Fig. 6D). D482G (c.1445A>G) showed an improvement on treatment, but mature protein levels were lower than those for wild-type BSEP (Fig. 6D). For the remaining mutations, this treatment led to an enrichment of yield of the lower-molecular-weight, immature BSEP forms, but had no effect on the mature protein levels.
Analysis of the Functional Properties of BSEP Mutants.
It is important to know whether disease mutations associated with normal BSEP protein levels, or those potentially susceptible to modulation of protein folding resulting in an increase in mature protein levels, have normal BSEP function. Wild-type and variant ABCB11 baculoviruses with the mutations N490D (c.1468A>G) and A570T (c.1708G>A) were created, and, after infection of Hi5 insect cells, inside-out membrane vesicles were prepared to perform taurocholate transport assays.3 The mutant N490D was chosen because it generated a mature form of BSEP, although in reduced abundance (Fig. 5A), and because the associated disease was severe,30 whereas A570T was amenable to glycerol and reduced temperature: these enhanced the abundance of a mature form of BSEP (Fig. 6D). Fig. 7 demonstrates that each BSEP variant was expressed to the same level as wild-type BSEP in Hi5 cells, with an apparent molecular weight of 140 kDa. A higher-molecular-weight band was also visible, presumably corresponding to aggregates.3 Adenosine triphosphate–dependent taurocholate transport activity of wild-type and mutant BSEP proteins was determined as previously described.3 Hi5 cells infected with a mock baculovirus had 5% of the activity of wild-type BSEP. Both mutant proteins had significantly reduced taurocholate transport activity compared with wild-type BSEP. The N490D mutation was associated with 27% and the A570T mutation with 60% of normal activity.
ABCB11, which encodes the bile salt export pump, BSEP, is a 28-exon gene that is mutated in severe PFIC and milder BRIC, ICP, CC, and DC.7, 8, 12, 28–45, 57–59 This study showed that some ABCB11 missense mutations were associated with splicing defects, most caused protein processing defects, and the two functionally analyzed showed a reduction in taurocholate transport. Additionally, the ability to modulate splicing and protein processing defects was examined; some mutations were amenable to such treatments.
Of the ABCB11 nucleotide changes assayed for pre-mRNA splicing defects, the introduction into exon 14 of the one predicted to generate D482G (c.1445A>G) resulted in the predominant use of a cryptic splice site within this exon (Figs. 2C, 3B). D482G is one of the most common PFIC missense mutations in European patients, and in vitro studies have differed in their conclusions on its effect on protein trafficking and protein function.11, 13, 14, 60 The altered pre-mRNA splicing of this mutant allele observed in this study (Fig. 3B) generates an mRNA template that on translation would result in the introduction of 14 novel amino acids followed by a stop codon. Therefore, if such splicing predominates in vivo, D482G-containing BSEP, per se, may only exist at very low levels in hepatocytes. Indeed, immunohistochemical analyses of liver from patients with this mutation show that detectable BSEP is lacking or greatly reduced.30, 31 The development of hepatocellular carcinoma and cholangiocarcinoma in BSEP-deficient patients is highly associated with splice site changes, deletions, insertions, and nonsense mutations that are predicted to result in a total absence of functional BSEP protein.30, 31, 35 In fact, of the nine missense mutations present in patients with hepatobiliary malignancy, five of these (56%) have been shown in this study to be associated with splicing defects.30, 31 It is envisaged that this is attributable to the more severe consequences of these types of mutations, whereby no BSEP is available to remove toxic bile salts from the liver. Of BSEP deficiency patients who developed malignancy, 16% have the D482G mutation,30, 31, 35 putatively making it a more severe disease-causing mutation than previously recognized. The aberrant splicing identified in this study could now provide an explanation for this occurrence, because it would result in a truncated, non-immunodetectable protein. Additionally, D482G-containing mRNA in vivo appears to be unstable, because in a previous small study from our group, liver from a patient heterozygous for D482G yielded no PCR-amplifiable ABCB11 exon 13 to exon 15 (indicating that the second mutation also resulted in deficiency of ABCB11 mRNA).61 The fact that the D482G-associated cryptic splicing could not be modulated by the addition of exogenous SC35 splicing factor (Fig. 4A) may make it more difficult to develop novel treatments to overcome this particular defect.
By contrast, the aberrant splice product associated with E297K (c.889G>A) was completely abolished on addition of the splicing factor SC35, whereas higher levels of hnRNPA2 and hnRNPH had a negative effect on the normal splicing of E297K (Fig. 4A). Therefore, the actual levels of positive and negative splicing factors in patients could potentially affect the levels of aberrant splicing associated with E297K in affected hepatocytes. This in turn could result in variation in clinical phenotype among patients with this mutation. Differences may be further exacerbated by the fact that the maturation of E297K protein can be modulated by conditions that facilitate protein folding by chaperones and thus increase protein release from the ER (Fig. 6D).
The nucleotide change resulting in the synonymous SNP A1028A (c.3084A>G) appears to cause the skipping of exon 24 (Fig. 2E). This SNP was documented previously in patients with primary biliary cirrhosis and was found in the heterozygous state in a patient with ICP.40, 43 c.3084A and c.3084G are present in near-equal proportions in most populations.45 The implications of this synonymous SNP for the development of cholestasis require further investigation. If a patient's allele contained the variant G as well as the nucleotide change resulting in E297K, for example, a double-negative effect on ABCB11 pre-mRNA splicing would be more severe. Conversely, a second nucleotide change in ABCB11 might abolish the negative effects of A1028A. This was recently shown for the gene MCAD, in which an exonic SNP halted exon skipping associated with the presence of a missense mutation in an ESE.62 Furthermore, the effects of intronic mutations and SNPs on ABCB11 exonic splicing are poorly understood and warrant further investigation.
E186G (c.557A>G) was associated with the moderate skipping of exon 7 when the corresponding nucleotide substitution was introduced (Fig. 2A). This mutation was previously identified in the heterozygous state in two patients with a BRIC phenotype.8, 37ABCB11 mRNA levels were the same as those in a normal adult.37 Measuring absolute levels of mRNA does not take into account increased transcription to compensate for instability of aberrantly spliced mRNA species, for example. The fact that the addition of the splicing factor SC35 could modulate the exon 7 skipping associated with E186G may mean that splicing of pre-mRNA containing the nucleotide change is a fluid situation in cells. Such splicing would depend on the activity of these splicing factors in the local environment of the mutant E186G ABCB11 pre-mRNA. Indeed, there is much evidence for pre-mRNAs associating with specific but numerous RNA-binding proteins that escort them to the cytoplasm.63 One patient was noted to have reduced canalicular membrane BSEP staining.37 Expression of E186G did indeed result in very low levels of mature BSEP (Fig. 5E), and thus a combination of reduced mRNA and protein levels could contribute to the clinical phenotype.
Interestingly, the patient in question was also heterozygous for the SNP V444A (c.1331T>C). A444 (c.1331C) and V444 (c.1331T) are present at 57% and 43%, respectively, in other populations; c.1331C is more frequent in patients with ICP, DC, and CC than is c.1331T.37, 39–45 Indeed, quantitation of BSEP levels in normal resected liver showed that those individuals with low or very low levels of BSEP had a higher frequency of the C allele than did those with normal BSEP levels.44 The A444 amino acid form has reduced levels of mature BSEP compared with V444 (Fig. 5F). In vitro expression and functional characterization of A444 in insect cells demonstrated that it and V444 had similar bile salt transport activity.39 Thus, it would appear that the lower abundance of A444 BSEP in the canalicular membrane predisposes patients to intermittent cholestasis. When on the same allele as a compromised nucleotide change, A444 could result in a more severe phenotype. Creating double mutants would allow further investigation of this scenario, as well as combining patient allelic information with immunohistochemical data.
Normal levels of mature BSEP were seen for the predicted missense mutations I498T (c.1493T>C; Fig. 5A), L1242I (c.3724C>A; Fig. 5C), and N1211D (c.3631A>G; Fig. 5C). Mutation I498 is located within the first nucleotide binding fold (NBF) between the Walker A and ABC signature motifs, which are domains highly conserved among ABC transporter superfamily members.2 The mutation I498T is associated with the less severe, intermittent BRIC clinical phenotype.38 Intermittent episodes of cholestasis may arise when any functional activity of I498T is compromised within a patient's liver. The second NBF harbors N1211D and L1242I: the former is located eight amino acids upstream of the ABC signature motif and is only conserved among BSEP orthologs, whereas the latter is within the Walker B motif and is totally conserved across BSEP, multidrug resistance (MDR) 1 (MDR1), MDR3, MDR-associated protein 2, and CFTR. Both mutations have been found in patients with PFIC, but N1211D may be a rare SNP (S Strautnieks, unpublished data). The lack of conservation among ABC transporters calls into question the importance of N1211, but functional studies are needed to confirm the effect of N1211D. With respect to L1242I, disruption of a Walker B motif is predicted to cause severe functional consequences.64
No detectable protein was observed for the predicted missense mutations V284L (c.850G>C) and A588V (c.1763C>T; Fig. 5A). V284A enhances, whereas V284L abolishes, BSEP protein levels. Valine, alanine, and leucine are aliphatic amino acids and are the essential “blocks” on which protein structures form.65 The introduction of leucine, which has one more methyl group than valine, may be detrimental to the stability of the intracellular loop between transmembrane domains 4 and 5, where V284 is located. The replacement of A588 with valine, within the first NBF of BSEP, may have a similar consequence on protein stability. A444V is a third example of a conservative aliphatic amino acid substitution, but with the maintenance of mature protein for both amino acid forms albeit with a clear difference in abundance (as discussed previously).
Most of the missense mutations expressed in vitro resulted in a considerable reduction of mature BSEP levels; any protein detected was instead the lower-molecular-weight BSEP forms (Fig. 5A-F). Molecular chaperones can promote such mutations to undergo correct processing, with the defective trafficking associated with the ΔF508 mutation in CFTR being successfully manipulated in vitro and in vivo.56 Treatment with glycerol and incubation at a reduced temperature resulted in increased mature protein levels for 61% of BSEP mutations, which presumably cause ER retention and degradation of the protein (Fig. 6A-D). These include the PFIC-associated mutations E297G (c.890A>G; Fig. 6A), R1128C (c.3382C>T) and R1231Q (c.3692G>A; Fig. 6C), and R1268Q (c.3892G>A; Fig. 6.d), the BRIC-associated mutations R1128H (c.3383G>A) and R1050C (c.3148C>T; Fig. 6C), and E297K (c.889G>A; Fig. 6D), as well as A570T (c.1708G>A), which can be associated with either form of disease. A recent study introduced E297G and D482G into human BSEP and assessed their trafficking in MDCK cells.66 The addition of sodium 4-phenylbutyrate prolonged the half-life of both mutant BSEP forms and resulted in increased functional expression of the proteins. This concurs with the data presented here, which show reduced levels of mature BSEP when E297G and D482G were expressed in vitro (Fig. 5D). Furthermore, the addition of 10% glycerol and incubation at 28°C resulted in a substantial and a marginal increase in mature E297G and D482G protein, respectively (Fig. 6A, D). For D482G, we found significant splicing defects (see previous discussion). Neither this treatment nor sodium 4-phenylbutyrate could have any clinical effect if normal D482G mRNA processing were almost completely disrupted by aberrant splicing in patients, as suggested by our results in vitro. However, if there were some wild-type splicing present, these agents could allow a partial rescue of phenotype, taking into account the slight defect in protein function previously determined for this mutant protein.15 Because there are no observed splicing defects associated with the nucleotide change associated with E297G, such therapies are possible in the future for this mutation.
Two mutations, N490D (c.1468A>G) and A570T (c.1708G>A), were expressed in insect cells and then assessed for their ability to transport taurocholate (Fig. 7A, B). The latter was chosen because levels of mature BSEP were enhanced on treatment with reduced temperature and glycerol (Fig. 6D), raising the question of whether the protein was actually functional. Despite equal expression of both mutants in these cells, N490D had only 27% and A570T had 60% of the activity of the wild-type protein, respectively. The phenotype of the patient with N490D was very severe, with death before the age of 1 year.30 A second mutation in ABCB11 has not been found but heterozygosity for V284A is present (c.851T>C; M. Jirsa, unpublished data). A570T was identified on both alleles in a patient with BRIC who required a partial external biliary diversion to treat progressive cholestasis at the age of 18 years.8 Both N490D and A570T are situated in the first NBF of BSEP; N490D is conserved across the MDR family, whereas A570T is also conserved in MDR-associated protein 2 and CFTR. However, the substitution N490D is more functionally detrimental and apparently more clinically severe than A570T. The recent expression of rat Bsep with A570T in MDCK cells gave comparable results to our study with respect to levels of mature protein.14 These authors attributed an approximately 50% reduction of taurocholate transport solely to reduced BSEP protein expression levels, whereas our data suggest an additional defect in transport activity. Whether this discrepancy is attributable to BSEP species variation or to differing experimental systems is unclear.
In this study, it can be seen that BRIC-associated mutations have a higher proportion of splicing or protein defects that can be modulated by exogenous factors. Additionally, BRIC mutations are less often associated with complete absence of mature BSEP protein, although levels may be very low (Fig. 5). These findings correlate well with the phenotype of these patients, less severe than that of PFIC patients; the mutations found in PFIC patients usually yield little mature protein and often generate substantial amounts of immature BSEP protein. A smaller-scale in vitro study of PFIC versus BRIC mutations reached similar conclusions.14
ESE and ESS prediction programs can be used to identify nucleotide substitutions that could alter splicing. For ABCB11, several missense mutations' nucleotide context was predicted to cause aberrant splicing, such as A588V (c.1763C>T; data not shown). However, this was not observed in the in vitro minigene splicing system. Similarly, nucleotide changes not predicted to cause any abnormalities, such as D482G (c.1445A>G), did disrupt splicing. Using more than one of these programs can enhance the probability of obtaining correct predictions, but assessing splicing directly is the only way to determine whether any defects actually occur. The minigene system provides an excellent tool to assess splicing changes on a large scale. Effects on protein folding/degradation and function are harder to predict. BSEP is an ABC transporter within the same subfamily as MDR1 and MDR3, and sharing many conserved domains with CFTR. The fact that most of the ABC transporter mutations result in ER retention is likely attributable to their multimembrane-spanning structure. This differs from genes encoding less complex proteins, such as neurofibromin 1 (NF1) and breast cancer 1, early onset (BRCA1), in which more than 50% of missense mutations result in aberrant splicing.16, 21 The complex structure of ABC transporters, requiring many membrane-spanning domains to be maintained during synthesis in the ER, gives an indication that trafficking problems might be predominant.
In this study, more than 80% of a gene's predicted missense mutations and SNPs have been analyzed for pre-mRNA splicing defects. In combination with information gained from in vitro protein expression studies, such analysis of disease-causing mutations in ABCB11/BSEP permits assignment of mutations into categories, which is an essential prerequisite to design better therapeutic regimens. Therefore, from this study, treatments could now be tailored toward pre-mRNA splicing or protein processing correction. BSEP deficiency is ideally suited to novel forms of therapy, such as gene or protein manipulation, because relatively low levels of correction could result in a major improvement in phenotype.67
The authors thank Milan Jirsa, Dita Cebecauerová, Leo Klomp, Catherine Williamson, Antal Németh, and Étienne Sokal for sharing unpublished patient data.