The transporter “variome”: The missing link between gene variants and bile salt transporter function

Abstract

The human bile salt export pump, encoded by the ABCB11 gene, is a member of the ATP-binding cassette (ABC)-containing family of genes, and the central molecule in the transport of bile acid from hepatocytes into the bile canaliculus. Located exclusively at the canalicular membrane of the hepatocyte, defects cause a spectrum of cholestatic disorders, depending on their impact on protein function: from severe progressive familial intrahepatic cholestasis type 2 (PFIC type 21) via benign recurrent cholestasis2 to intrahepatic cholestasis of pregnancy,3 or cholestasis induced by contraceptives or drugs.4 ABC, ATP-binding cassette; mRNA, messenger RNA; PFIC, progressive familial intrahepatic cholestasis; SNP, single-nucleotide polymorphism. Studies in PFIC families have led to the identification of a variety of genetic aberrations, some of which could be shown to decrease taurocholate transport activity.5 Other protein alterations cause the absence of ABCB11 protein by introducing stop codons, or, as described for numerous other diseases,6 by abnormal trafficking of mutant protein, leading to retention and degradation of the immature protein in the endoplasmatic reticulum. A potential treatment for patients carrying trafficking mutants would be to use a transporter ligand that acts as a pharmacological chaperone to correct the folding defect.7 A unique expression pattern and the clearly defined function with a low, or even nonexistent, functional redundancy make ABCB11 a highly interesting target to study genotype-phenotype correlation in hepatobiliary diseases. In the article presented in this issue of HEPATOLOGY, Byrne et al.8 took upon themselves the formidable task of characterizing the majority of sequence variants known to be associated with ABCB11 deficiency. Following a previous study on substrate specificity and identification of inhibitors,9 they now extended their analyses of disease-causing mechanisms from protein function tests to in vitro analysis of pre-mRNA splicing events. In a highly impressive effort, they analyzed 62 known ABCB11 missense mutations and 21 single-nucleotide polymorphisms (SNPs) with regard to their impact on pre-mRNA splicing, abnormal processing of ABCB11 protein, or alterations in ABCB11 function. Alterations in the nucleic acid sequence can lead to protein abnormalities via different mechanisms: Traditionally the best known mechanism is the translation of the variant nucleic acid into a different amino acid. The effect that such a replacement has on protein function depends on the position and the biochemical nature of the replaced amino acid. If the alteration leads to a premature stop codon, the result is a nonfunctional product. The complete abolishment of functional protein leads to total absence of protein in immunohistochemical detection10 and usually the most severe phenotypes. However, the genotype-phenotype correlation of sequence variants is not always straightforward, and some seemingly mild alterations with regards to amino acid changes can lead to clinically relevant phenotypes in vivo, which has been a conundrum for some time.11 Only recently, it has become obvious that even subtle and seemingly innocuous genetic variants can lead to severe consequences via an alternative route not involving direct changes in the amino acid sequence: Nucleotide exchanges within or outside the coding region may cause disease by changing the structure and stability of the gene's messenger RNA (mRNA). Missense mutations well away from the known splice junctions may disrupt mRNA stability or lead to the abolishment or generation of splice sites. Missplicing can introduce premature stop codons and cause a phenotype of similar severity as nonsense mutations, whereas alterations in RNA secondary structure can impinge on the splicing process and the stability of pre-mRNA, thereby decreasing the amount of protein available. Byrne et al.8 expressed mutant protein for 42 variants (37 missense mutations and five SNPs) associated with amino acid alterations and analyzed abundance of protein in the membrane fraction. Bile acid transport experiments were used to assay defects in bile salt export pump function in proteins of seemingly normal abundance and localization. To assess potential changes in pre-mRNA splicing, 62 predicted missense mutations and 21 SNPs were introduced by in vitro mutagenesis and mutant sequences were cloned into a minigene vector containing flanking sequence from the fibronectin extra domain B (EBD) and 3 α-globin exons. Differentially spliced products were amplified with primers from inside the flanking vector, sequenced, and quantified by differential real-time polymerase chain reaction. Using splice site prediction programs, as demonstrated in this study, is a possible start, but due to the level of false positive and false negative results, not a viable option to the costly and labor-intensive approach of in vitro mutagenesis, minigene-cloning, and splicing analysis. Using all these methods, the study reveals a highly complex and variable picture of how genetic variants cause functional defects by several mechanisms. The results of the present study are consistent with data from recent high-throughput sequencing of transcripts, demonstrating that exon skipping is the most prevalent form of alternative splicing.12 This type of analysis also explains why some transporter variants that may appear to have rather low impact on the protein level, such as p.D482G,13 one of the most common PFIC missense mutations in Europe, give rise to a comparatively severe phenotype: The authors describe how this variant leads to altered pre-mRNA splicing, resulting in an mRNA which, when translated, introduces 14 novel amino acids, followed by a stop codon. This represents a much more serious alteration than the replacement of one amino acid, and the severity of the splice-induced abnormality explains the almost complete lack of detectable BSEP protein in immunohistochemical analysis of livers from patients with the p.D482G mutation.13, 14 Of note, a patient with obstetric cholestasis carrying this variant also displayed a severe phenotype resolving only in part after delivery.3 Additional weight is added to the results from the splice site analyses by the fact that the development of hepatocellular carcinoma and cholangiocarcinoma is highly associated with splice site changes, deletions, insertions, and nonsense mutations predicted to result in total absence of BSEP protein. Of note, five of nine mutations present in patients with hepatobiliary malignancy have been shown to be associated with splicing defects.13, 14 As seen in all good studies, this one raises questions and opens new discussions: How accurately do splicing events within the minigene reflect the in vivo situation in a distinct genomic context? Which distant SNPs should be interrogated for effects on splicing to generate an accurate inventory of variants? Should even synonymous exonic SNPs be investigated, which might have an impact on ABC transporter substrate specificity due to altered folding and membrane insertion?15 These data add substantial information to the development of a comprehensive genotype-phenotype association matrix for the ABCB11 gene. Not just do they describe the association of a genetic variant with disease in a family, but furthermore they describe in detail the mechanism by which this variant leads to such profound changes in protein expression and function. Furthermore, the authors present critical information on how to relate genotypes and phenotypes and provide the experimental framework for developing risk assessment tools for hepatobiliary diseases based on either total genome sequence of the individual or array analysis of SNPs known to convey certain levels of risk. Describing in an equally detailed and comprehensive way the mechanisms to change protein sequence and functionality reveals a complex picture of how multiple layers of modulation can be added by sequence alterations in the most inconspicuous places. It also makes us realize once more that gene expression is anything but straightforward, and that any DNA variant has to pass multiple checkpoints in transcription and translation before being turned into a protein (Fig. 1). Exemplary ways how gene variants can lead to aberrant protein expression. A = exonic, nonconservative variation causing an amino acid replacement. The resulting changes in protein structure can cause: (1) Functional impairment at the respective site of function (i.e., at the canalicular membrane) without detectable reduction (as seen in ABCB11 p.I498T). (2) Retention and degradation of the protein in the endoplasmatic reticulum (trafficking mutant, for example ABCB11 variants p.V284L and p.A588V). (3) Generation of a complete but instable mRNA, which results in low transcript abundance, adding low protein expression to the impact of the structural abnormality. (4) Generation of a novel splice donor or acceptor site by an exonic variant may create exon skipping or missplicing, thereby creating a premature stop codon (as seen in ABCB11 p.D482G) or generating instable mRNA species. B = Intronic variants may lead to: (5) Aberrant splicing by disrupting splice donor and acceptor sites, resulting in abnormal transcripts and nonfunctional protein. To interfere with splicing or protein function, they do not necessarily have to be positioned directly at the splice site,16 but further experiments are required to assess the exact impact of intronic mutations and SNPs. The rapid progress in individualized sequencing will enable clinicians to obtain information on genetic variation for various genes of interest at low cost. Hence, the problem arises how to interpret these data. This is exemplified by yet another ATP-dependent transporter, ATP7B, which encodes the Wilson disease gene.16 Although this was one of the first genes identified by positional cloning more than 15 years ago, the large number of ATP7B variants identified since then is still undergoing functional validation to prove pathogenic effects of individual variants.17 Studies like the one by Byrne et al.8 in this issue are there to fill the gap between sequence variation (“variome”) and altered transport (“phenome”).18 Similar assays are needed for correct classification of the risk conveyed by other transporter variants, a prerequisite for reliable molecular diagnoses. The “beauty” of genes like ATP7B and ABCB11 as study objects is the fact that due to their unique roles, nucleotide exchanges leading to functional impairment will result in detectable phenotypes in the majority of cases. For many other genes, sequence variants outside the coding regions are harder to associate with disease because there is functional redundancy, hence lower protein abundance will not necessarily result in detectable pathological changes. The report by Byrne et al.8 in this month's issue comprehensively describes mechanisms of disease and suggests potential therapeutic intervention strategies based on these mechanisms for a large number of transporter variants. In addition, it provides ample evidence for the association of noncoding gene variants with disease progression and susceptibility. After reading the article, everyone should look at SNPs with different eyes.