Functional Characterization of Common Protein Variants in the Efﬂux Transporter ABCC11 and Identiﬁcation of T546M As Functionally Damaging Variant

ORIGINAL ARTICLE Functional characterization of common protein variants in the efﬂux transporter ABCC11 and identiﬁcation of T546M as functionally damaging variant

R Arlanov1, T Lang1, G Jedlitschky2, E Schaeffeler1, T Ishikawa3, M Schwab1,4 and AT Nies1

Multidrug resistance protein 8 (ABCC11)isanefflux transporter for anionic lipophilic compounds, conferring resistance to antiviral and anticancer agents like 5-fluorouracil (5-FU). ABCC11 missense variants may contribute to variability in drug response but functional consequences, except for the ‘earwax variant’ c.538G4A, are unknown. Using the ‘Screen and Insert’ technology, we generated human embryonic kidney 293 cells stably expressing ABCC11 missense variants frequently occurring in different ethnic populations: c.57G4A, c.538G4A, c.950C4A, c.1637C4T, c.1942G4A, c.4032A4G. A series of in silico prediction analyses and in vitro plasma membrane vesicle uptake, immunoblotting and immunolocalization experiments were undertaken to investigate functional consequences. We identified c.1637C4T (T546M), previously associated with 5-FU-related toxicity, as a novel functionally damaging ABCC11 variant exhibiting markedly reduced transport function of 5-FdUMP, the active cytotoxic metabolite of 5-FU. Detailed analysis of 14 subpopulations revealed highest allele frequencies of c.1637C4T in Europeans and Americans (up to 11%) compared with Africans and Asians (up to 3%).

The Pharmacogenomics Journal (2016) 16, 193–201; doi:10.1038/tpj.2015.27; published online 21 April 2015

INTRODUCTION More than 100 naturally occurring ABCC11 missense variants are Interindividual variability in response to antiviral and anticancer listed in the dbSNP database (National Centre for Biotechnology drug therapy is a major obstacle in clinical practice.1,2 Genetic Information: http://www.ncbi.nlm.nih.gov/SNP) with substantially variants affecting proteins involved in the absorption, disposition, different frequency distributions between ethnic groups. However, with the exception of the functionally damaging variant c.538G4A metabolism and elimination of these drugs may majorly contri- 17 bute to the variability in drug efficacy and/or development of (rs17822931, G180R), none of these variants has been character- adverse drug reactions. In addition to drug-metabolizing enzymes, ized in detail on the functional or cell biological level. The variant ‘ ’ drug transporters are important determinants of drug response G180R, also termed the earwax SNP , has gained much attention fl 3,4 because it determines the Mendelian trait of dry or wet earwax in because they mediate cellular drug uptake and ef ux. Members 18 of the ATP-binding cassette (ABC) transporter C subfamily mem- humans. Moreover, there appears to be a positive association among the ‘wet earwax type’, increased axillary osmidrosis17 and a bers are of particular interest since they have been identified as 19 – higher risk for breast cancer among Japanese women. Variant efflux transporters for antiviral and anticancer drugs.5 7 Here, G180R has also been suggested as a clinical biomarker predicting human ABCC11, which encodes multidrug resistance protein 8, has efficacy of nucleoside-based chemotherapy.19 In a pharmaco- emerged as a novel candidate gene for detailed investigation. genomic study, we recently identified an additional ABCC11 Among others, ABCC11 confers resistance to antiviral and anti- 4 fl missense variant, that is, T546M (c.1637C T, rs17822471), as a cancer nucleoside-based agents, such as 5- uorouracil (5-FU), further potential clinical biomarker.20 We showed that leucopenia, 9′-(2′-phosphonyl methoxyethyl) adenine (adefovir) and cytara- 8,9 a major toxicity type after 5-FU monotherapy of patients, is bine, as well as to methotrexate. ABCC11 also mediates the significantly associated with variant T546M (odds ratio: 3.31), cellular efflux of different endogenous anionic lipophilic com- 20 8 whereas variant G180R showed no association. Thus, functional pounds, including the cyclic nucleotides cGMP and cAMP, the analyses of ABCC11 variants are of clinical importance. 10 bile acids glycocholate and taurocholate, sulfated and glucur- Owing to the lack of an orthologous murine gene correspond- 10,11 onidated steroids as well as glutathione-conjugated sulfanyl- ing to human ABCC11 (ref. 21) an Abcc11 knockout mouse model alkalones.12 Because of the wide expression of ABCC11 messenger is not available that could help to identify the functional RNA in various tissues and cell types with particular high levels in consequences of a deficient ABCC11 transporter in humans. – breast and testis13 16 ABCC11 may be an important determinant Hence, systematic in vitro analyses of ABCC11 missense variants of the intracellular concentrations of endogenous compounds with respect to their expression, subcellular localization and and drugs. function are required.

1Dr Margarete Fischer-Bosch Institute of Clinical Pharmacology, Stuttgart and University of Tübingen, Stuttgart, Germany; 2Department of Pharmacology, Center of Drug Absorption and Transport (C_DAT), University Medicine of Greifswald, Greifswald, Germany; 3RIKEN Center for Life Science Technologies, Yokohama, Japan and 4Department of Clinical Pharmacology, Institute of Experimental and Clinical Pharmacology and Toxicology, University Hospital of Tübingen, Tübingen, Germany. Correspondence: Professor M Schwab, Dr Margarete Fischer-Bosch Institute of Clinical Pharmacology, Auerbachstr. 112, Stuttgart 70376, Germany. E-mail: [email protected] Received 2 September 2014; revised 23 February 2015; accepted 2 March 2015; published online 21 April 2015 Functional consequences of common ABCC11 variants R Arlanov et al 194

Table 1. ABCC11 missense variants selected for analysis of expression, subcellular localization and function

rs# NCBI Genetic variationa Amino acid Ethnicity Allele frequency (%) Nb Data sourcec

rs16945988 c.57G4A R19H CA 10.6 226 HapMap-CEU AF 25.9 224 HapMap-YRI AS 2.4 168 HapMap-JPT rs17822931 c.538G4A G180R CA 12.5 224 HapMap-CEU AF 0 120 HapMap-YRI AS 89.0 172 HapMap-JPT rs11863236 c.950C4A A317E CA 9.8 224 HapMap-CEU AF 29.6 226 HapMap-YRI AS 0.6 168 HapMap-JPT rs17822471 c.1637C4T T546M CA 7.5 120 HapMap-CEU AF 1.7 120 HapMap-YRI AS 0 90 HapMap-JPT rs16945930 c.1942G4A V648I CA 0.9 224 HapMap-CEU AF 0 120 HapMap-YRI AS 23.3 172 HapMap-JPT rs16945916 c.4032A4G H1344R CA 18.1 226 HapMap-CEU AF 41.5 86 HapMap-YRI AS 0 224 HapMap-JPT Abbreviations: AF, African; AS, Asian; CA, Caucasian. aNucleotide numbering reﬂects cDNA numbering with +1 corresponding to the A of the ATG translation initiation codon in the reference ABCC11 sequence NM_032583.3. bNumber of chromosomes analyzed. cNCBI dbSNP; http://www.ncbi.nlm.nih.gov/snp/.

In this study, we comprehensively analyzed six frequently described11 and transiently transfected with ABCC11 (WT or variants) in occurring ABCC11 missense variants: R19H (rs16945988), G180R pcDNA3.1_Hygro(-). (rs17822931), A317E (rs11863236), T546M (rs17822471), V648I (rs16945930) and H1344R (rs16945916). For functional analysis, we Preparation of crude membrane fractions stably expressed these variants in human embryonic kidney (HEK) Crude membrane fractions from stably transfected Rhek-IV cells were 293 cells using the ‘Screen and Insert’ technology, which we had prepared as previously described.22 In brief, cells were sonicated in previously established to elucidate functional consequences of hypotonic buffer, followed by centrifugation of the membrane fractions ABCC2 variants.22 Further, in silico phenotype predictions with four and storage of the collected fractions in hypotonic buffer at − 80 °C. different prediction programs were undertaken. Our study identiﬁed variant T546M, which we identiﬁed to be associated Preparation of plasma membrane vesicles 20 with a higher incidence of 5-FU-induced leukopenia, as a novel Plasma membrane vesicles from Rhek-IV cells stably expressing ABCC11 functionally damaging ABCC11 variant by using sulfated steroids WT or variants and control Rhek-IV cells were prepared from hypotonically as prototypic substrates and 5-FdUMP, the active cytotoxic lysed cells as described previously.23 Details are given in the Supple- metabolite of 5-FU. mentary Information.

MATERIALS AND METHODS Immunoblotting Immunoblotting of crude membrane fractions was performed as Chemicals described22 using the previously characterized KEA antibody.11 Protein Unless otherwise specified, all chemicals were obtained from Sigma-Aldrich expression levels were analyzed semiquantitatively by densitometry with GmbH (Taufkirchen, Germany). Radiolabeled [1,2,6,7-3H(N)]dehydro- AIDA v4.50 software (Raytest, Straubenhardt, Germany). Deglycosylation − epiandrosterone 3-sulfate (70.5 Ci mmol 1), [6,7-3H]estrone 3-sulfate was performed as described.17 See Supplementary Information for further − (45.6 Ci mmol 1) and [6-3H]5-fluoro-2’-deoxyuridine 5’-monophosphate details. (16.6 Ci mmol − 1) were purchased from American Radiolabeled Chemicals (Saint Louis, MO, USA), Perkin Elmer (Boston, MA, USA) and Hartmann fl Analytic (Braunschweig, Germany), respectively. Confocal uorescence microscopy Rhek and HepG2 cells were grown in eight-well chamber slides (Flexiperm, Sarstedt, Nümbrecht, Germany) at a density of 4 × 104 cells per well. Cells In vitro site-directed mutagenesis of ABCC11 were fixed with methanol at − 20 °C for 10 min and rehydrated in Human ABCC11 complementary DNA (cDNA) (referred to as ‘wild-type’, WT, phosphate-buffered saline. Cells were incubated with affinity-purified throughout the manuscript)17 was cloned with lox sites into the ABCC11 KEA antibody (1:500 dilution)11 for 1 h at room temperature and mammalian expression vector pcDNA3.1_Hygro(-) as described by us.20 washed 3 × in phosphate-buffered saline. Apical membranes of polarized This insertion vector and ABCC11 WT c DNA in pcDNA3.1_Hygro(-)17 transiently transfected HepG2 cells were identified by co-incubation of the served as templates for site-directed mutagenesis using the QuikChange ABCC11 antibody with anti-CD26 antibody (1:400 dilution; Alexis, Lausen, mutagenesis kit (Stratagene, Heidelberg, Germany). Mutagenesis primers Switzerland).24 Cells were then incubated with AlexaFluor 488-conjugated are given in Supplementary Information Table 1. DNA sequences were goat anti-rabbit immunoglobulin G and AlexaFluor 568-conjugated goat confirmed by sequencing. anti-mouse immunoglobulin G (1:300 dilution) as appropriate for 1 h at room temperature, and then washed again 3 × in phosphate-buffered Cell culture of HEK and HepG2 cells and transfection with ABCC11 saline. The coverslips were mounted onto glass slides using Mowiol. variants Images were taken with a confocal laser scanning microscope (TCS SP8, Leica Microsystems, Wetzlar, Germany). Generation of HEK293 cells with tetracycline-regulated d2EGFP expression (Rhek-IV), and Rhek-IV cells stably expressing ABCC11 WT or variant T546M were previously described by us.20 Rhek-IV cells expressing further ABCC11 Vesicle transport studies variants were generated following the same procedure.20 Details are given ATP-dependent transport of sulfated steroids and 5-FdUMP into plasma in the Supplementary Material. Human HepG2 cells were cultured as membrane vesicles was measured with the rapid filtration technique using

The Pharmacogenomics Journal (2016), 193 – 201 © 2016 Macmillan Publishers Limited Functional consequences of common ABCC11 variants R Arlanov et al 195 Mutation nomenclature The indicated cDNA numbers are relative to the ATG site starting with +1 corresponding to the A of the ATG translation initiation codon and based on the cDNA sequence from GenBank accession number NM_032583.3.

Analysis of allele frequencies and linkage disequilibrium of ABCC11 variants The publicly available databases ‘1000 Genomes’ (http://browser.1000gen omes.org/ index.html) and ‘HapMap’ (http://hapmap.ncbi.nlm.nih.gov/) were interrogated to assess allele frequencies and linkage disequilibrium (LD) of ABCC11 variants in different populations. HapMap and 1000 Genomes project data comprise a total of 1301 and 1092 individuals from 11 and 14 different populations, respectively. LD between pairs of ABCC11 variants was quantiﬁed by D´ and r2 statistics using the online tool LD TAG SNP Selection (http://snpinfo.niehs.nih.gov/snpinfo/snptag.htm) or the ‘variations tool’ within the 1000 Genomes browser using the link: http:// browser.1000genomes.org/Homo_sapiens/Variation/ HighLD?db = core; r = 16:48241879-48242879; v = rs17822471; vdb = variation; vf = 10030053.

In silico tools for predicting the functional consequences of ABCC11 variants Functional consequences of missense ABCC11 variants were predicted in silico with four commonly used and publicly available online tools, which had been applied before for predicting functional consequences of variants of membrane drug transporters:28,29 SIFT (http://sift.jcvi.org/), PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/), PMut (http://mmb2. pcb.ub.es:8080/PMut/) and SNPs3D (http://www.snps3d.org/). See Supple- mentary Information for further details.

Statistical analysis Data are expressed as means ± s.e.m. or s.d. Data comparisons among groups were performed by unpaired t-test or one-way analysis of variance followed by Dunnett’s post test to compare versus the control group or by Tukey’s post test to compare all groups. Analyses were performed with GraphPad Prism 5.04 (GraphPad Software, San Diego, CA, USA). P ⩽ 0.05 was considered statistically signiﬁcant.

RESULTS Figure 1. Expression of ABCC11 in Rhek cells. (a) A membrane Characterization of a Rhek cell line with tetracycline-regulated topology model of ABCC11 was generated using TOPO2 trans- ABCC11 expression membrane protein display software (www.sacs.ucsf.edu/TOPO2/). In extension of our initial characterization of a Rhek cell line with Transmembrane domain and loop regions were assigned based on tetracycline-regulated ABCC11 WT expression,20 we performed topology data from the TOPCONS database (http://topcons.cbr.su. fi se/). The antigen recognition site for the ABCC11 antibody KEA11 is additional analyses con rming that this cell system is valid for indicated by black boxes. (b) Immunoblot analysis of Rhek-IV studying the functional consequences of ABCC11 variants. In cells and Rhek cells stably expressing ABCC11 WT with (+) or immunoblot analyses, ABCC11 was predominantly detected at − without (− ) tetracycline (1 μgml 1). Deglycosylation of the ABCC11 ~ 180 kDa corresponding to the N-linked glycosylated ABCC11 protein was carried out using PNGase F. (c) Immunofluorescence form as previously described (Figure 1).17 The band at ~ 150 kDa is micrographs of ABCC11 in Rhek cells stably expressing ABCC11 consistent with the non-glycosylated ABCC11 form, as only this WT compared with untransfected Rhek-IV cells. Cells were fixed with methanol and staining was performed with ABCC11 antibody band remained after deglycosylation with PNGase F (Figure 1b). KEA. Bar, 20 μm. ABCC11 expression disappeared after 96 h exposure to tetracycline and it took up to 7 days to re-establish ABCC11 protein expression (Figure 1b). Moreover, Rhek-ABCC11 WT, but not Rhek- IV cells, showed clear ABCC11 immunofluorescence staining in the glass fiber filters23,25 and centrifugation of the vesicles through a Sephadex plasma membrane as detected by confocal laser scanning G-50 gel matrix in NICK spin columns,23 respectively. Details are given in the Supplementary Information. microscopy (Figure 1c). Finally, vesicular uptake studies with the two sulfated steroids and prototypic substrates DHEAS and E1 3S8,11 confirmed that ABCC11 is functionally active in the Rhek cells (Figure 2). The net ATP- and time-dependent uptake of ATPase activity [1,2,6,7-3H(N)]dehydroepiandrosterone 3-sulfate and [6,7-3H] ATPase activity of ABCC11 WT and T546M variant in plasma membrane estrone 3-sulfate was enhanced approximately threefold in vesicles was measured by determining the amount of inorganic phosphate liberated after 20 min incubation with a colorimetric assay kit (Innova membrane vesicles from ABCC11-expressing Rhek cells when Biosciences, Cambridge, UK) in the absence or presence of vanadate, compared with those from control membrane vesicles (Figures 2a which is a potent inhibitor of ABC function.26 The incubation buffer and d). Moreover, the potent ABCC11 inhibitor 4,4'-Diisothio- contained ouabain, sodium azide and ethylene glycol tetraacetic acid to cyano-2,2'-stilbenedisulfonic acid caused a 90% inhibition of the + + 2+ inhibit Na /K -ATPase, F1-F0-ATPase and Ca -dependent ATPases, ABCC11-mediated DHEAS and E13S transport at a concentration of 27 respectively. Details are given in the Supplementary Information. 50 μM (Figure 2e and f).

3 3 Figure 2. ATP-dependent transport of [ H]DHEAS (a, c, e) and [ H]E13S (b, d, f) into membrane vesicles from ABCC11 WT-transfected Rhek and control (Rhek-IV) cells. Membrane vesicles (20 μg) prepared from Rhek-ABCC11 cells and Rhek-IV cells were incubated for the indicated time 3 3 points at 37 °C in uptake medium containing 4 mM ATP or 4 mM AMP and 3 μM [ H]DHEAS (a)or1μM [ H]E13S (b), and (e and f) in the presence or absence of the ABCC11 inhibitor DIDS (50 μM). (c and d) ATP-dependent uptake was calculated by subtracting the values obtained in transport medium containing AMP from the values obtained in medium containing ATP. Values are means ± s.e.m. of at least two independent experiments measured in triplicates.

Selection of ABCC11 missense variants for functional complete genetic linkage in Caucasians (D’ =1, r2 = 0.95) (Supple- characterization mentary Figure 6). We therefore decided to construct cells In addition to the ‘earwax SNP’ G180R, ﬁve other ABCC11 missense expressing double-variant R19H/A317E. variants were prioritized for functional analysis based on the following criteria: (i) allele frequency ⩾ 5%, (ii) structural features and (iii) evolutionary conservation (Table 1, Supplementary Figure Protein expression of ABCC11 variants 5). According to this strategy we selected the missense variants To determine whether the transfected Rhek cells expressed R19H, G180R, A317E, T546M, V648I and H1344R for further the selected variants on protein level, membrane lysates analysis. Next, we performed a LD haplotype analysis of these were investigated by immunoblotting (Figure 3a). Similar to WT, variants revealing that variants R19H and A317E are in almost variants T546M, V648I, H1344R and R19H/A317E were

The Pharmacogenomics Journal (2016), 193 – 201 © 2016 Macmillan Publishers Limited Functional consequences of common ABCC11 variants R Arlanov et al 197 ABCC11 WT (Figure 5a). Remarkably, T546M resulted in a largely reduced ATP-dependent 5-FdUMP transport (Figure 5b).

ATPase activity of T546M variant ABC transporters utilize the energy derived from ATP hydrolysis to transport substrates across cell membranes.26 Because T546M is located within the highly conserved Walker A motif of the first nucleotide-binding domain (Supplementary Figure 5) ATP hydrolysis might be affected. We therefore measured the amount of inorganic phosphate liberated by membrane vesicles from ABCC11 WT, T546M variant and Rhek-IV control cells (Table 2). Vanadate-sensitive ATPase activity was already detectable in Rhek- IV control vesicles, which is probably due to ABC transporters endogenously expressed in HEK cells, such as ABCC4.31 ABCC11 WT vesicles had a significantly higher ATPase activity than control vesicles, whereas T546M vesicles had a somewhat higher activity than control vesicles, which was significantly lower than WT.

Subcellular localization of ABCC11 variants To investigate whether ABCC11 variants affected plasma membrane localization, immunofluorescence studies were performed (Figures 6 and 7). Similar to the localization of ABCC11 WT, variants T546M, V648I, H1344R and R19H/A317E were predominantly localized in the plasma membrane of Rhek cells. In agreement with previous studies,17 expression of ‘earwax SNP’ G180R at the plasma membrane was very low and variant protein was mainly detected intracellularly. Because Rhek cells are nonpolarized and we aimed to analyze localization of ABCC11 variants in a polarized cell system as well, we transiently transfected human HepG2 cells (Figure 7). These cells achieve hepatic polarity and form apical vacuoles between adjacent cells.24 As we showed previously,11 ABCC11 WT localized in the apical membrane of polarized HepG2 cells (Figure 7). Similarly, all ABCC11 variants were also detectable in the apical membrane (Figure 7). Figure 3. (a) Immunoblot analysis of Rhek-IV clones stably expressing ABCC11 WT or variants. (b) Protein expression was analyzed semiquantitatively by densitometry of the 180 kDa band corre- Functional prediction of the investigated ABCC11 missense sponding to the mature glycoprotein. For standardization, ABCC11 variants protein levels were normalized to β-tubulin expression. ABCC11 WT ± Four different bioinformatic phenotype prediction tools (SIFT, expression was set as 100%. Data are means s.e.m. of four PolyPhen-2, PMut, SNPs3D) were used to assess the functional experiments. *** indicates a P-value o0.001 compared with the expression level from ABCC11 WT. consequences of all six ABCC11 missense variants (Supplementary Table 2). All tools predicted the ‘earwax variant’ G180R correctly as damaging substitution. Moreover, variant T546M was also correctly predicted to be damaging by three (SIFT, PolyPhen-2, SNPs3D) out of the four tools. The results for all other variants predominantly expressed as mature glycoprotein of ~ 180 kDa. In ‘ ’ (R19H, A317E, V648I, H1344R) were discordant thereby preventing contrast, the earwax variant G180R was only weakly expressed reliable functional prediction. and only the unglycosylated form was detectable, consistent with a previous study.17 Semiquantification of the glycosylated ABCC11 form relative to β-tubulin confirmed that variants T546M, V648I, Analysis of allele frequencies and LD patterns of ABCC11 T546M fi H1344R and R19H/A317E were expressed comparable to WT, Our studies identi ed T546M as a novel functionally damaging whereas expression of variant G180R was very low (Figure 3b). ABCC11 variant. Because this may be of clinical relevance, we performed a detailed analysis of its allele frequencies and LD patterns. By the use of the 1000 Genomes project data different Transport activity of ABCC11 variants allele frequencies of ABCC11 variant T546M were found for the 14 Next, we determined whether the selected variants were subpopulations from Europeans, Americans, Africans and Asians (Supplementary Table 3). The c.1637 T (546M) allele occurs more functionally active by using the prototypic substrates DHEAS frequently in European and American subpopulations with and E 3S (Figure 4). The ‘earwax variant’ G180R showed virtually 1 minor allele frequencies between 4 and 11% compared with no DHEAS transport and largely reduced E13S transport. Of note, African and Asian subpopulations with allele frequencies between fi the T546M variant also resulted in a signi cantly decreased 0 and3%. In addition, we performed a pairwise LD analysis for transport activity of DHEAS and E13S compared with WT. In variants in different subpopulations, which are captured by T546M contrast, transport activities of the other investigated variants (rs17822471) on chromosome 16 (Supplementary Table 4). LD (V648I, H1344R, R19H/A317E) were comparable to that of WT. analysis showed that rs17822471 is in perfect LD (D´= 1, r2 =1) Moreover, we investigated whether the T546M variant also with six and three intronic variants of ABCC11 in Mexicans and showed an impaired transport of 5-FdUMP, which is the active Puerto Ricans, respectively. Moreover, T546M (rs17822471) is metabolite of 5-FU30 and had been previously identified as linked in 100% with the ABCC11 A666T variant (rs112416246) in ABCC11 substrate.8 5-FdUMP was confirmed as substrate of the Japanese (JPT: Japanese in Tokyo, Japan) and Kenyans (LWK:

Figure 4. ATP-dependent transport of DHEAS and E13S into membrane vesicles from ABCC11 WT, ABCC11 variants and control Rhek-IV cells. 3 3 Membrane vesicles (20 μg) were incubated with (a and c)3μM [ H]DHEAS or (b and d)1μM[ H]E13S in the presence of 4 mM ATP or 4 mM AMP for 10 min at 37 °C as described in Materials and Methods. Vesicle-associated radioactivity was determined by rapid ﬁltration. Uptake of 3 3 [ H]DHEAS and [ H]E13S was calculated by subtracting the values obtained in transport medium containing AMP from the values obtained in medium containing ATP and minus uptake of values measured for membrane vesicles from control Rhek-IV cells (shown as % of WT). Values shown are means ± s.e.m. of at least two independent experiments measured in triplicates and performed with two independently isolated membrane vesicle preparations of each variant. *** indicates a P-valueo0.001 compared with the uptake of DHEAS and E13S from ABCC11 WT. (c and d) Uptake values relative to ABCC11 protein levels, which were determined semiquantitatively by immunoblotting.

Luhya in Webuye, Kenya) with calculated allele frequencies of the the plasma membrane may be different in polarized cells, we also double-variant T546M/A666T of 1.7% and 2.6%, respectively analyzed localization of ABCC11 WT and variants in HepG2 cells, (Supplementary Tables 3 and 4). which achieve polarization as previously described.11 All ABCC11 variants and WT were able to reach the apical membrane in this cell system. Of note, also ‘earwax variant’ G180R was detected in DISCUSSION the apical membrane. This is consistent with the observation that In the present study, we comprehensively analyzed the con- variant G180R does not affect the luminal plasma membrane sequences of six frequently occurring ABCC11 missense variants localization of ABCC11 in secretory cells of ceruminal glands.17 on protein expression, localization and function. ABCC11 variants We used the two sulfated steroids DHEAS and E13S, which are are of clinical interest, because ABCC11 is relevant for drug established prototypic substrates of ABCC11,10,11 for functional disposition with potential consequences for drug response.19 assessment of the six investigated variants. The initial character- Recently, we described the ‘Screen and Insert’ method and ization of G180R showed that this variant resulted in a significantly successfully used this technique for the functional characterization reduced cellular cGMP efflux.18 Our results now extend this of ABCC2 protein variants, another member of the ABC-trans- characterization to further prototypic substrates, DHEAS and E13S, porter subfamily C.22 Using this method, we have now generated and confirm that G180R severely compromises function. Transport HEK293 cell lines that stably express and stringently regulate the of both sulfated steroids was also largely impaired by variant expression of ABCC11 WT and missense variants. T546M, but not to the same extent as observed for the ‘earwax Consistent with previously published data,17 the immunoblot variant’ G180R. As expected, expression levels of G180R are very analyses showed that, with the exception of the ‘earwax variant’ low due to its rapid proteasomal degradation,17 explaining the G180R, protein levels of all ABCC11 variants were comparable to almost absent transport function. In contrast, for T546M, expres- that of WT ABCC11. Moreover, our immunolocalization analyses sion levels of the glycosylated form and membrane localization confirmed that expression of ‘earwax variant’ G180R at the plasma are comparable to WT ABCC11 suggesting that T546M compro- membrane of HEK cells was very low and mainly detected mises intrinsic transport function. We hypothesize that this is due intracellularly. In contrast, as already evident from the immuno- to the fact that T546M is located within the highly conserved blots, WT and the other variants (T546M, V648I, H1344R, R19H/ Walker A motif of the first nucleotide-binding domain (Supple- A317E) were expressed at considerable levels and readily mentary Figure 5) and that amino-acid alterations in this region detectable in the plasma membrane of Rhek transfectants. severely compromise function. This is supported by our finding Because Rhek cells are nonpolarized and because targeting to that T546M has a reduced ATPase activity. Moreover, in a study on

The Pharmacogenomics Journal (2016), 193 – 201 © 2016 Macmillan Publishers Limited Functional consequences of common ABCC11 variants R Arlanov et al 199 the ABCC1 protein, which belongs to the same subfamily as ABCC11, the substitution of lysine684 with methionine within the Walker A motif of the ﬁrst nucleotide-binding domain signiﬁcantly reduced ATP-dependent transport function, whereas expression levels of WT ABCC1 and mutated ABCC1 were similar.32

Table 2. ATPase activity of membrane vesicles from ABCC11 WT, variant T546M and control Rhek-IV cells

Cell line Vanadate-sensitive Pi release nmol mg − 1 protein min − 1

Rhek-IV 56.0 ± 17.6 control ABCC11 95.2 ± 27.4 Po0.01 vs WT Rhek-IV ABCC11 64.8 ± 12.2 Po0.05 vs T546M ABCC11 WT Data are means ± s.d. of nine determinations. P-values were calculated by ANOVA followed by Tukey’s post test.

Figure 5. Time course of ATP-dependent transport of FdUMP into membrane vesicles from ABCC11 WT (a), ABCC11 T546M variant (b) and control Rhek-IV (c) cells. Membrane vesicles were incubated 3 with 1 μM [ H]FdUMP in the presence of 4 mM ATP or 4 mM AMP at 37 °C for the indicated time points. Vesicle-associated radioactivity was determined by centrifugation of the vesicles through Sephadex gel matrix. Values are means ± s.d. of five determinations. Figure 6. Localization of ABCC11 variants in stably transfected Rhek cells by immunofluorescence. ABCC11 protein variants were probed with an ABCC11 antibody11 and then with anti-rabbit IgG labeled with green fluorescent AlexaFluor 488 dye. Subcellular localization was analyzed by confocal laser scanning microscopy. Bar, 20 μm; same magnification for all pictures.

Table 3. Conservation of the ABCC11 missense variants among other ABCC orthologs and homologs

Protein Speciesa R19H G180R A317E T546M V648I H1344R

ABCC11 Human R G A T V H Gorilla R G E T V R Macaque L G E T V R Dog L G E T V R Rabbit L G E T V R Bovine L G Q T V R ABCC1 Human I G N V V R ABCC2 Human C S N V T K ABCC3 Human T N N V V R ABCC4 Human ─ QQV VK ABCC5 Human R G M V L R ABCC6 Human S G N V V R ABCC10 Human C G N V V R ABCC12 Human ─ GQVVH Multiple sequence alignment was performed using ClustalW (www.ebi.ac. uk/clustalw). Bold text indicates the ABCC11 reference amino-acid sequences for human, gorilla, macaque, dog, rabbit and bovine. See Supplementary Material for accession numbers.

between functional in vitro data and eight prediction tools for novel identified cytochrome P450 2B6 variants in Rwandeses was shown.33 SIFT was the best prediction tool, which is of interest because this tool is based only on multiple sequence alignments, similar to those shown in Table 3. The limitation of in silico prediction tools, however, is that only effects of single missense variants can be assessed. In case of the ABCC11 variants R19H and A317E, which are in complete linkage in Caucasians, the SIFT prediction for both single variants were consistent with the functional consequences of the double-variant R19H/A317E in our in vitro experiments. However, previous data regarding the double-variant V1188E/C1515Y in another ATP-dependent efflux transporter (ABCC2)22 clearly indicated that the information of in silico tools for both single variants did not correctly predict functionality of the double-variant. Thus, experimental studies are definitively required to assess reliably functional consequences of double-variants. Because the T546M variant showed a largely impaired transport Figure 7. Localization of ABCC11 variants in transiently transfected activity considerable alteration of the pharmacokinetics of ABCC11 polarized HepG2 cells by immunofluorescence. ABCC11 protein drugs may be expected resulting in potential clinical conse- variants were detected with an ABCC11 antibody11 (green) and the quences, such as adverse drug reactions. ABCC11 is directly apical membranes with an antibody against CD26 (red). Subcellular involved in 5-FU resistance8,34 by the efflux transport of the active localization was analyzed by confocal laser scanning microscopy on metabolite 5-FdUMP.8 Recently, we identified a significant a TCS SP8 (Leica). Arrows point to ABCC11 protein localized in the association of T546M with severe leucopenia in European patients apical membrane of polarized cells. Bar, 10 μm; same magnification 20 for all pictures. treated with 5-FU monotherapy in vivo, which is mechanistically now corroborated by our in vitro functional data of a markedly impaired transport activity of 5-FdUMP by the T546M variant. A decreased transport activity in 5-FU-treated cancer patients 4 Because there are 100 yet uncharacterized ABCC11 missense carrying T546M might result in higher intracellular levels of 5- variants and functional characterization of missense variants is FU, thereby contributing subsequently to 5-FU-related toxicity. ‘ ’ rather low-throughput , reliable in silico prediction tools would be Additional clinical studies are warranted to further elucidate the helpful. We therefore compared the functional data of the six impact of T546M variant on 5-FU toxicity. investigated ABCC11 missense variants with the predicted out- In summary, we have identified T546M as a novel functionally come of the four different in silico tools SIFT, PolyPhen-2, PMut damaging ABCC11 variant frequently occurring in Caucasians. The and SNPs3D, which had been used to predict functional in silico tools SIFT and PolyPhen-2 may be useful for functional 28,29 consequences of variants of membrane drug transporters. prediction of the large number of rare and yet uncharacterized The variants G180R and T546M, which we identified as function- ABCC11 variants. However, in silico prediction is not satisfactory in ally damaging in our in vitro experiments, were correctly predicted all cases and does not take into account the functional by 100% and 75%, respectively, confirming that in silico analysis consequences of double-variants. Therefore, HEK293 cells stably may be used to identify additional functionally damaging variants transfected using the ‘Screen and Insert’ method as described in in the future. This observation is corroborated by investigations this study provide a suitable cell model for in vitro studies of related to drug-metabolizing enzymes where a high concordance missense variants in the membrane transporter ABCC11.

The Pharmacogenomics Journal (2016), 193 – 201 © 2016 Macmillan Publishers Limited Functional consequences of common ABCC11 variants R Arlanov et al 201 CONFLICT OF INTEREST 16 Matsson P, Yee SW, Markova S, Morrissey K, Jenkins G, Xuan J et al. Discovery of The authors declare no conflict of interest. regulatory elements in human ATP-binding cassette transporters through expression quantitative trait mapping. Pharmacogenomics J 2012; 12:214–226. 17 Toyoda Y, Sakurai A, Mitani Y, Nakashima M, Yoshiura K, Nakagawa H et al. Earwax, ACKNOWLEDGMENTS osmidrosis, and breast cancer: why does one SNP (538G4A) in the human ABC transporter ABCC11 gene determine earwax type? FASEB J 2009; 23: 2001–2013. We gratefully acknowledge Sabine Rekersbrink and Silvia Hübner for expert technical 18 Yoshiura K, Kinoshita A, Ishida T, Ninokata A, Ishikawa T, Kaname T et al. A SNP in assistance. RA, TL, ES, MS and ATN were supported by the Robert Bosch Stiftung, the ABCC11 gene is the determinant of human earwax type. Nat Genet 2006; 38: Stuttgart, Germany, RA and TL were supported by the German Research Foundation 324–330. DFG (LA 2406/2-1). GJ was supported by the German Research Foundation DFG (JE 19 Ishikawa T, Toyoda Y, Yoshiura K, Niikawa N. Pharmacogenetics of human ABC 234/4-1). ATN and MS were supported by the 7FP EU Initial Training Network transporter ABCC11: new insights into apocrine gland growth and metabolite program‚ FightingDrugFailure’ (GA-2009–238132). MS and ES were supported in part 3 by the Federal Ministry for Education and Research (BMBF, Berlin, Germany; grant 03 secretion. Front Genet 2013; : 306. IS 2061C and 0315755). 20 Magdy T, Arlanov R, Winter S, Lang T, Klein K, Toyoda Y et al. ABCC11/MRP8 polymorphisms affect 5-fluorouracil-induced severe toxicity and hepatic expression. Pharmacogenomics 2013; 14:1433–1448. REFERENCES 21 Shimizu H, Taniguchi H, Hippo Y, Hayashizaki Y, Aburatani H, Ishikawa T. Char- acterization of the mouse Abcc12 gene and its transcript encoding an ATP- 1 Meyer UA, Zanger UM, Schwab M. Omics and Drug Response. Annu Rev Phar- binding cassette transporter, an orthologue of human ABCC12. Gene 2003; 310: macol Toxicol 2013; 53: 475–502. 17–28. 2 Wang L, McLeod HL, Weinshilboum RM. Genomics and drug response. N Engl J 22 Arlanov R, Porter A, Strand D, Brough R, Karpova D, Kerb R et al. Functional Med 2011; 364: 1144–1153. characterization of protein variants of the human multidrug transporter ABCC2 by 3 Giacomini KM, Huang SM, Tweedie DJ, Benet LZ, Brouwer KL, Chu X et al. Mem- a novel targeted expression system in fibrosarcoma cells. Hum Mutat 2012; 33: brane transporters in drug development. Nat Rev Drug Discov 2010; 9:215–236. 750–762. 4 Schuetz JD, Swaan PW, Tweedie DJ. The role of transporters in toxicity and dis- 23 Keppler D, Jedlitschky G, Leier I. Transport function and substrate specificity of ease. Drug Metab Dispos 2014; 42:541–545. multidrug resistance protein. Methods Enzymol 1998; 292:607–616. 5 Errasti-Murugarren E, Pastor-Anglada M. Drug transporter pharmacogenetics in 24 Keitel V, Kartenbeck J, Nies AT, Spring H, Brom M, Keppler D. Impaired protein nucleoside-based therapies. Pharmacogenomics 2010; 11:809–841. maturation of the conjugate export pump multidrug resistance protein 2 as a 6 Slot AJ, Molinski SV, Cole SP. Mammalian multidrug-resistance proteins (MRPs). consequence of a deletion mutation in Dubin-Johnson syndrome. Hepatology Essays Biochem 2011; 50:179–207. 2000; 32:1317–1328. 7 Nies AT, Lang T. Multidrug resistance proteins of the ABCC subfamily. In: You G, 25 Rädisch S, Dickens D, Lang T, Bonnett L, Arlanov R, Johnson MR et al. A com- Morris ME (eds). Drug Transporters: Molecular Characterization and Role in Drug prehensive functional and clinical analysis of ABCC2 and its impact on treatment Disposition, 2 edn. John Wiley & Sons: Hoboken, 2014, pp 161–185. response to carbamazepine. Pharmacogenomics J 2014; 14:481–487. 8 Guo Y, Kotova E, Chen ZS, Lee K, Hopper-Borge E, Belinsky MG et al. MRP8, ATP- 26 Linton KJ, Higgins CF. Structure and function of ABC transporters: the ATP switch binding cassette C11 (ABCC11), is a cyclic nucleotide efflux pump and a resistance fl fl 453 – factor for fluoropyrimidines 2',3'-dideoxycytidine and 9'-(2'-phosphonylmethox- provides exible control. P ugers Arch 2007; :555 567. yethyl)adenine. J Biol Chem 2003; 278: 29509–29514. 27 Sarkadi B, Price EM, Boucher RC, Germann UA, Scarborough GA. Expression of the 9 Toyoda Y, Ishikawa T. Pharmacogenomics of human ABC transporter ABCC11 human multidrug resistance cDNA in insect cells generates a high activity drug- 267 – (MRP8): potential risk of breast cancer and chemotherapy failure. Anticancer stimulated membrane ATPase. J Biol Chem 1992; :4854 4858. Agents Med Chem 2010; 10:617–624. 28 Ramsey LB, Bruun GH, Yang W, Trevino LR, Vattathil S, Scheet P et al. Rare versus 10 Chen ZS, Guo Y, Belinsky MG, Kotova E, Kruh GD. Transport of bile acids, sulfated common variants in pharmacogenetics: SLCO1B1 variation and methotrexate 22 – steroids, estradiol 17-beta-D-glucuronide, and leukotriene C4 by human multi- disposition. Genome Res 2012; :1 8. drug resistance protein 8 (ABCC11). Mol Pharmacol 2005; 67:545–557. 29 Nies AT, Niemi M, Burk O, Winter S, Zanger UM, Stieger B et al. Genetics is a major 11 Bortfeld M, Rius M, König J, Herold-Mende C, Nies AT, Keppler D. Human multi- determinant of expression of the human hepatic uptake transporter OATP1B1, 5 drug resistance protein ABCC11 (MRP8), an apical efflux pump for steroid sulfates, but not of OATP1B3 and OATP2B1. Genome Med 2013; :1. is an axonal protein of the central and peripheral nervous system. Neuroscience 30 Nies AT, Magdy T, Schwab M, Zanger UM. Role of ABC transporters in 2006; 137: 1247–1257. fluoropyrimidine-based chemotherapy response. Adv Cancer Res 2015; 125: 12 Baumann T, Bergmann S, Schmidt-Rose T, Max H, Martin A, Enthaler B et al. 217–243. Glutathione-conjugated sulfanylalkanols are substrates for ABCC11 and gamma- 31 Reid G, Wielinga P, Zelcer N, van der Heijden I, Kuil A, de Haas M et al. The human glutamyl transferase 1: a potential new pathway for the formation of odorant multidrug resistance protein MRP4 functions as a prostaglandin efflux transporter precursors in the apocrine sweat gland. Exp Dermatol 2014; 23:247–252. and is inhibited by nonsteroidal antiinflammatory drugs. Proc Natl Acad Sci USA 13 Bera TK, Lee S, Salvatore G, Lee B, Pastan I. MRP8, a new member of ABC trans- 2003; 100: 9244–9249. porter superfamily, identified by EST database mining and gene prediction pro- 32 Gao M, Cui HR, Loe DW, Grant CE, Almquist KC, Cole SP et al. Comparison of the gram, is highly expressed in breast cancer. Mol Med 2001; 7:509–516. functional characteristics of the nucleotide binding domains of multidrug resis- 14 Tammur J, Prades C, Arnould I, Rzhetsky A, Hutchinson A, Adachi M et al. Two new tance protein 1. J Biol Chem 2000; 275: 13098–13108. genes from the human ATP-binding cassette transporter superfamily, ABCC11 and 33 Radloff R, Gras A, Zanger UM, Masquelier C, Arumugam K, Karasi JC et al. Novel ABCC12, tandemly duplicated on chromosome 16q12. Gene 2001; 273:89–96. CYP2B6 enzyme variants in a Rwandese population: functional characterization 15 Yabuuchi H, Shimizu H, Takayanagi S, Ishikawa T. Multiple splicing variants of two and assessment of in silico prediction tools. Hum Mutat 2013; 34:725–734. new human ATP-binding cassette transporters, ABCC11 and ABCC12. Biochem 34 Oguri T, Bessho Y, Achiwa H, Ozasa H, Maeno K, Maeda H et al. MRP8/ABCC11 Biophys Res Commun 2001; 288:933–939. directly confers resistance to 5-fluorouracil. Mol Cancer Ther 2007; 6:122–127.

Supplementary Information accompanies the paper on the The Pharmacogenomics Journal website (http://www.nature.com/tpj)