Dulbecco’s Modified Eagle Medium (DMEM), Hank’s Buffered Salt Solution (HBSS), and 4-(4-(dimethylamino)styryl)-N-methylpyridinium (ASP⁺) were obtained from Life Technologies (Darmstadt, Germany). Poly-D-lysine (1–5 kDa), 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid (HEPES), bicinchoninic acid, and copper sulfate pentahydrate were obtained from Sigma-Aldrich (Taufkirchen, Germany). Dulbecco’s phosphate-buffered saline (DPBS) and additives for cell culturing were obtained from PAN-Biotech (Aidenbach, Germany). Twelve-well plates were obtained from CytoOne (Langenselbold, Germany), 6-well plates from Corning GmbH (Kaiserslautern, Germany), and tissue culture flasks from Sarstedt (Nümbrecht, Germany). All chemicals used in this study were purchased from commercial sources and had purities ≥95%.

HEK293 (Thermo Fisher Scientific, Darmstadt, Germany), Huh7, and HepG2 (ATCC, Manassas, United States) cells were cultured in DMEM supplemented with 10% fetal calf serum (FCS), 100 U/mL penicillin and 100 μg/ml streptomycin at 37°C and 5% CO₂.

The 8 bp insertion in exon 7 that results from alternative splicing of rs35854239 was introduced into an OCT1 encoding pcDNA5/FRT vector by site-directed mutagenesis as described previously (Seitz et al., 2015). The used primer pair 1 is listed in Supplementary table S1. The sequence was validated by capillary sequencing prior to transient transfection into HEK293 cells.

HEK293 cells were seeded at a density of 5 × 10⁵ cells per well in a 12-well plate precoated with poly-D-lysine. Twenty-four hours after seeding, cells were transfected with 2 µg of the alternatively spliced OCT1 vector DNA using Lipofectamine™ 2000 (Invitrogen, Darmstadt, Germany) according to the manufacturer’s instructions. Transfection efficiency was evaluated by co-transfection with 0.5 µg of the green fluorescent protein coding vector pGFP-tpz (Thermo Fisher Scientific). The next day, uptake experiments were performed at 37°C and pH 7.4 using HBSS+ (HBSS supplemented with 10 mM HEPES buffer). Cells were washed once with pre-warmed (37°C) HBSS+. Uptake was initiated by adding 20 µM ASP⁺ diluted in HBSS+ and stopped after two minutes by adding ice-cold HBSS+. Cells were washed twice with ice-cold HBSS+ and lyzed with RIPA buffer. Fluorescence of ASP⁺ in lysates was measured with an excitation of 485 nm and emission of 612 nm using the Tecan infinite M200 Microplate Reader (Tecan Group Ltd., Männedorf, Switzerland). ASP⁺ fluorescence intensities were normalized to the total protein amount in the samples as measured using the bicinchoninic acid assay (Smith et al., 1985).

Splicing of exon 7-intron 7 was analyzed in the splicing vector pSPL3b, further referred to as minigene. Exon 7-intron 7 of OCT1 for both rs35854239 genotypes was amplified with primer pair 2, listed in Supplementary table S1. The PCR product was cloned into the pSPL3b vector after restriction of the PCR product and the vector with PstI and EcoRV. The Met408 and Val408 were introduced by site-directed mutagenesis using primer pairs 3 and 4, respectively, listed in Supplementary table S1. The minigene constructs were validated by capillary sequencing and then used for transient transfection into Huh7 and HepG2 cells.

Huh7 and HepG2 cells were seeded in six-well plates at a density of 4 × 10⁵ and 1.7 × 10⁶ cells per well, respectively. After 24 hours, cells were transfected with 2 µg minigene vector DNA using Lipofectamine™ 2000 as described above. Transfection efficiency was evaluated by co-transfection with pGFP-tpz as described above. Forty-eight hours after transfection, cells were lysed and RNA was isolated using the RNeasy Plus Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions.

After RNA isolation from transfected cells, complementary DNA (cDNA) was synthesized using the MultiScribe™ Reverse Transcriptase Kit (Applied Biosystems). Spliced exon 7 was amplified with primer pair 5 listed in Supplementary table S1. PCR products were separated by gel electrophoresis, bands were visualized under UV light and band intensities were quantified using the Fiji software (ImageJ version 1.52p, National Institutes of Health, Bethesda, United States).

Human liver samples were obtained from normal liver tissue that had to be removed for technical reasons during liver surgery or from organ donors. Patients gave their informed consent for research use of the removed liver tissue, and the procedures were approved by the ethics committee of the University Medicine Göttingen, Georg-August-Universität Göttingen (application number 26/01/17) and the ethics committee of the Pomeranian Medical University (application number KB-0012/64/12). Deep-frozen human liver samples were homogenized using the Mikro dismembrator S (B. Braun, Melsungen, Germany) at 2500 rpm for 1 min. DNA was isolated using the DNeasy Blood & Tissue kit (QIAGEN) according to the manufacturer’s instructions. For the genotyping of functionally relevant polymorphisms in the SLC22A1 gene, the single base primer extension method was used as described previously (Seitz et al., 2015) using primer 6 listed in Supplementary table S1. RNA from human liver samples was isolated from homogenates using the RNeasy Plus Mini Kit and cDNA was synthesized as described above.

The ratio of correctly vs. alternatively spliced exon 7 in the transfected cell lines with hepatic origin and human liver samples was analyzed using pyrosequencing. Spliced exon 7 from minigene experiments and from cDNA from human liver samples was amplified using primer pair 7 (minigene) and primer pair 8 (liver samples) listed in Supplementary table S1. Samples were prepared using PyroMark™ Binding and Annealing Buffer (QIAGEN) and the PyroMark™ Vacuum Prep Station (Biotage, Uppsala, Sweden). Pyrosequencing was carried out on the PyroMark™ Q96 ID (Pyrosequencing AB, Uppsala, Sweden) with PyroMark™ Gold Q96 reagents (QIAGEN) using the primer 9 (minigene) and primer 10 (liver samples).

Next-generation DNA and RNA sequencing was performed with cDNA from minigene experiments and DNA and cDNA from human liver samples. Exon 7 and its 3′ flanking region were first amplified using the primer pairs 11 to 14 listed in Supplementary table S1. The PCR products were purified by magnetic separation using Agencourt^® AMPure^® XP reagent (Beckman Coulter GmbH, Krefeld, Germany). Unique indices were attached to the purified amplicons by PCR using the Nextera^® XT Index Kit v2 (Illumina Inc., San Diego, United States). The samples were again cleaned up with Agencourt^® AMPure^® XP reagent. All samples were pooled in appropriate ratios. The pooled library was quantified using the Qubit^® 2.0 fluorometer and the Qubit^® dsDNA BR assay kit (Thermo Fisher Scientific) and diluted to DNA concentration of 2 nM. DNA was denatured and diluted according to the manufacturer’s instructions. As internal control and to increase variability within the sequencing run, 30% PhiX control (Illumina) was spiked in prior to denaturation. The sequencing run was performed using the MiSeq^® Reagent Kit v3 (600 cycles) and paired-end 221 reads on the Illumina MiSeq™ (Illumina). The sequencing run was analyzed using the IGV v.2.6.3 software (Broad Institute, Cambridge, United States).

The paired-end sequence reads were merged using PEAR (release 0.9.11; Zhang et al., 2014). The mapping to a reference sequence was performed with Bowtie2 (Langmead et al., 2012) version 2.3.4.1. DNA was mapped against the human genome assembly (hg19) downloaded from the UCSC Genome Browser. The cDNA from human liver samples was mapped against the “Homo sapiens solute carrier family 22 member 1 (SLC22A1), transcript variant 1, mRNA” (NM_003057.3) with artificially introduced duplication in it in order to better visualize the possible insertion using local mapping mode. The cDNA from minigene was mapped against the minigene itself. For both experiments with human liver samples we calculated subsequently the sequencing depth for each allelic combination of rs35854239 variant and rs628031 (Met408Val) variant with our own script.

Five x 10⁵ HEK293 cells were seeded on poly-D-lysine coated cover slips and transfected as described above. One day after transfection, cells were washed twice with PBS and were fixed with 100% ethanol for 20 min at –20°C. After washing three times with PBS, cell membranes were permeabilized with PBS containing 0.4% Tween 20. Cells were washed three times with PBS and subsequently blocked for 3 hours with blocking buffer (5% FCS in PBS). OCT1 was stained using the NBP1-51684 (2C5) antibody (Novus Biologicals, Abingdon, United Kingdom). Cells were co-stained with the EP 1845Y antibody (Abcam, Cambridge, United Kingdom) against the membrane marker Na⁺/K⁺-ATPase. The primary antibodies against OCT1 and Na⁺/K⁺-ATPase were diluted in blocking buffer in a dilution of 1:400 and 1:200, respectively. Per cover slip, 50 µL antibody solution was added, cells were covered with parafilm and incubated in a humid chamber overnight. The next day, after washing three times with PBS, fluorescently-labeled secondary antibodies (Alexa Fluor^® 546 goat anti-rabbit IgG (H + L), polyclonal and Alexa Fluor^® 488 goat anti-mouse IgG (H + L), polyclonal; Thermo Fisher Scientific) were diluted 1:400 in PBS, added and incubated for 2 hours in the dark. Cells were washed three times with PBS and cover slips were mounted with Roti^® Mount Fluor Care DAPI (Carl Roth, Karlsruhe, Germany). The staining was analyzed using the laser scanning microscope LSM780 (Carl Zeiss Microscopy GmbH, Oberkochen, Germany). Images were processed using the Fiji software.

OCT1 mRNA and OCT1 protein expression data were extracted from a previous study describing expression of OCT1 and OCT3 in human liver samples (Nies et al., 2009). Analysis was performed on the subset of samples (n = 90) that were from individuals who were non-cholestatic and had no hepatocellular, cholangiocellular or gallbladder carcinoma (Schaeffeler et al., 2011; Nies et al., 2013). The study was approved by the ethics committees of the Charité, Humboldt University (Berlin, Germany) and the University of Tübingen (Tübingen, Germany) in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from each patient.

The clinical trials on the effects of SLC22A1 genetic variants on fenoterol and sumatriptan pharmacokinetics have been described in details before (Matthaei et al., 2016; Tzvetkov et al., 2018). The rs35854239 variant was genotyped using the available DNA from those studies and the single base primer extension method as described previously (Seitz et al., 2015) with the SNaPshot primers listed in Supplementary table S1.

Differences in OCT1 mRNA and protein expression, or drug plasma concentration between homozygous wild-types and homozygous duplication allele carriers were performed using the Mann-Whitney-U test. Differences between DNA and RNA allele frequencies in allelic expression imbalance analyses were performed using the paired sample t-test. All analyses were performed using SPSS Statistics version 25 (SPSS INC., IBM, Chicago, IL) Statistical significance was defined as p < 0.05. Post-hoc power calculations of the clinical studies were performed with the G*Power software version 3.1.9.4 (Faul et al., 2007).

We used minigene assays to quantify the percentage of alternatively spliced transcripts in the 8 bp duplication allele of the rs35854239 variant. For this purpose, exon 7, including 306 bp upstream and 310 bp downstream of the flanking intronic regions, was cloned in the minigene vector pSPL3b between the exons 1 and 2 of the rabbit β-globulin gene. Next to the construct carrying the 8 bp duplication allele, the wild-type allele was also cloned and used as a control in the analyses (Figure 1B). Two independent minigene clones containing the duplication allele were analyzed to account for potential artifacts from the quality of the clones and the DNA preparation. The minigene constructs were transiently transfected into HepG2 and Huh7 cells, and the resulting correctly and alternatively spliced transcripts were quantified 48 hours later using three independent quantification techniques: semi-quantitative PCR, pyrosequencing and deep RNA-sequencing. In all cases, first, total RNA was reverse transcribed into cDNA. For the semi-quantitative PCR, the spliced SLC22A1 exon 7 was amplified by PCR using primers within the flanking exons of the rabbit β-globulin gene. The PCR products were separated by gel electrophoresis to enable the selective identification of the correctly and the alternatively spliced transcripts, and the band intensities were quantified (Figure 1C). As expected, the wild-type allele was spliced 100% correctly. However, the duplication allele was only spliced 47% correctly (range 42–51%) in transiently transfected Huh7 cells and 52% correctly (range 46–58%) in HepG2 cells (data not shown).

Next, we used pyrosequencing to quantify more precisely the ratio of the alternatively spliced transcripts. The pyrosequencing quantification method was validated by calibration series of vectors encoding the correctly or alternatively spliced minigene (Supplementary Figure S1). The pyrosequencing-based quantification showed that in Huh7 cells, the duplication allele was correctly spliced in 49% (range 46–52%) of all transcripts (Figure 1D). This was highly comparable with the previously semi-quantitatively determined ratios. In HepG2 cells, the duplication allele was correctly spliced in 57% (range 55–60%) of all transcripts.

Finally, the minigene insertion allele clone 1 spliced in Huh7 or HepG2 cells at 48 h was reanalyzed using deep massive-parallel sequencing (Figure 1E). The average depth of targeted RNA sequencing was 59,135 reads (range 32,902–84,691). The quantification of reads carrying the 8 bp insertion as a result of alternative splicing showed a percentage of 62% correctly spliced minigene in both cell lines. These results confirm the alternative splicing of rs35854239. More importantly, these results suggest that the 8 bp duplication causes erroneous splicing in only a part of the transcripts. Estimated by the data of all experiments maximally 52% of the transcripts are erroneously spliced. Thus, our in vitro data suggest that even in homozygous carriers of the duplication about the half of OCT1 transcripts will be correctly spliced.

Within the SLC22A1 gene, exon 7 harbors the highest density of coding functionally relevant polymorphisms. Thereof, the Met408Val substitution is almost completely linked to the rs35854239 duplication (Figure 2A). Under native conditions, it could not be excluded that the coding variant substantially contributes to the effects of splicing. Here were took advantage of the minigene technique and addressed separately the effects of the two variants on splicing. The Val408Met substitution alone did not significantly affect splicing, neither on duplication nor on wild-type rs35854239 background (Figure 2B). Therefore, it could be concluded that the effects on splicing are completely caused by the rs35854239 duplication and there is no contribution of the highly linked Met408Val.

The alternative splicing leads to an 8 bp longer exon 7, entailing a frame shift. This results in an altered amino acid sequence after codon 425 followed by a premature stop after seven amino acids. The resulting truncated OCT1 protein p. Asp426fs consists of the first nine TMHs only (Figure 3A).

To analyze whether the truncated OCT1 protein p. Asp426fs is able to function as an uptake transporter, the 8 bp insertion sequence was introduced between exon 7 and exon 8 of the OCT1 carrying pcDNA5 vector using site-directed mutagenesis. The resulting vector was transiently transfected into HEK293 cells and the uptake of the model OCT1 substrate ASP⁺ was compared to the uptake of the wild-type OCT1. Three independent clones of p.Asp426fs were analyzed. The alternatively spliced OCT1 protein showed no transport activity (Figure 3B). ASP⁺ uptake of all alternatively spliced OCT1 clones was at same levels as the empty pcDNA5 vector, indicating no OCT1-mediated substrate uptake.

Immunofluorescence staining revealed aberrant membrane localization of the alternatively spliced OCT1 (Figure 3C). This demonstrates that the truncated OCT1 protein, which results from alternative splicing of exon 7, is completely inactive and lacks correct membrane localization.

Minigene analyses in cell lines with hepatic origin showed that the 8 bp duplication leads to maximally 52% of alternatively spliced transcripts that encode a non-functional protein. In order to validate these results in vivo, we quantified the effects of the 8 bp duplication on OCT1 mRNA splicing in human liver. To this end, DNA from 24 liver samples was genotyped for rs35854239 and the correct splicing of the exon 7-intron 7 junction was quantified using pyrosequencing and deep RNA-sequencing (Figure 4).

Using pyrosequencing, we observed 99% correctly spliced SLC22A1 mRNA irrespective of the genotype of the liver donors. Alternative splicing of OCT1 in liver samples from homozygous or heterozygous 8 bp duplication allele carriers appeared with a maximum of 2.1% (Figure 4A). This percentage is far below the observed results in minigene experiments (Figure 1) but is still substantially higher than the observed 0.4% in the SLC22A1 mRNA from donors with wild-type genotype, which can only be spliced correctly.

Using deep RNA-sequencing, we detected very low levels of alternatively spliced transcripts that were, however, dependent on the rs35854239 genotype (Figure 4B). The average depth of sequencing was 74,326 reads per RNA sample (range 41,116–133,715). The liver samples from homozygous duplication allele carriers showed mean values of 0.58% alternatively spliced transcripts (range 0.19–1.14%). In heterozygous genotypes, alternative splicing was detected with a mean of 0.36% (range 0.08–0.83%). The samples of the homozygous wild-type allele carriers showed a mean of 0.02% alternatively spliced transcripts indicating very low levels of possible contamination with this highly sensitive method. In conclusion, both techniques demonstrated that despite the close to 50% probability of alternative splicing of the rs35854239 duplication allele estimated by the minigene assays, alternatively spliced OCT1 could only barely be detected in human liver samples. These results suggest that the alternatively spliced transcripts may be recognized and rapidly degraded under native conditions.

To verify this, we performed an allelic expression imbalance analysis in the human liver samples. We took advantage of the strong genetic linkage between the duplication allele of rs35854239 and the A-allele of the coding variant Met408Val (rs628031, 1222A>G, r ² = 0.95; Figure 5A). Based on the strong linkage, in heterozygous carriers of the rs35854239 duplication and Met408Val A-allele haplotype, a degradation of alternatively spliced OCT1 could be detected as a lower abundance of the Met408 A-allele in the RNA transcripts compared to the expected 50% of the DNA reads (Figure 5B). We used all nine liver samples from which both DNA and RNA was available and applied deep sequencing for quantification. While the A-allele was detected in 50% of the DNA reads (range 49–53%), the abundance in RNA was significantly decreased to 42% (range 40–44%, p = 2.77 × 10⁻⁷, paired t-test; Figure 5C). This result supports the degradation of the alternatively spliced transcripts and suggests that the presence of the rs35854239 duplication will result in the reduction of OCT1 mRNA levels, and as a consequence OCT1 protein in general.

In addition, more precise analyses of the sequencing reads suggest that the correct “canonical” splicing may be preferred under native conditions. Indeed, reads of alternatively spliced RNA carrying the A-allele of Met408Val were almost not detectable (Figure 5D). However, the reads of correctly spliced RNA carrying A-allele Met408Val were 71.3% of the G-allele reads (range from 65.8 to 76.7%) instead of the expected 50% from the minigene analyses. This suggests that in parallel to degradation of the alternatively spliced transcripts also a preference for correct splicing of rs35854239 duplication allele under native conditions may exist.

To address this, we analyzed total OCT1 mRNA and protein expression in 73 human liver samples. The liver samples had been characterized for their OCT1 expression before (Nies et al., 2009). We included only those samples lacking the Arg61Cys substitution, which is known to significantly affect OCT1 protein levels in the liver (Supplementary table S2 and (Nies et al., 2009) by affecting the correct membrane localization (Seitz et al., 2015). The OCT1 mRNA expression was on median 47% lower in homozygous duplication than in homozygous wild-type rs35854239 allele carriers (median of 0.014 and 0.026 transcripts per beta-actin transcript, respectively; p = 0.007; Figure 6A). The OCT1 protein levels were on median 35% lower in homozygous duplication than in homozygous wild-type rs35854239 allele carriers (median of 3.91 and 6.04, respectively; p = 0.045; Figure 6B). However, although statistically significant, the effects of rs35854239 genotypes could explain only 9% of the variability of mRNA and protein expression in this sample set.

Finally, we analyzed to what extend the decrease of OCT1 expression in carriers of the rs35854239 duplication allele leads to changes in sumatriptan and fenoterol pharmacokinetics in humans. We took advantage of the existing studies on the effects of OCT1 genotypes on the pharmacokinetics of both drugs (Matthaei et al., 2016; Tzvetkov et al., 2018) and analyzed them in the context of the rs35854239 genotype. The AUC of sumatriptan was slightly increased in homozygous rs35854239 duplication allele carriers compared to the wild-type (means of 7187 vs. 6277 min × ng/ ml, respectively, Figure 7A). However, this increase was not significant and was on average by 14% compared to the observed 127% increase in poor OCT1 transporters (homozygous or compound heterozygous carriers of the coding variants Arg61Cys, Gly401Ser, Gly465Arg) observed in the same study.

Even more, the AUC of fenoterol was not higher in homozygous carriers of the rs35854239 duplication allele compared to the wild-type (means of 84.25 vs. 86.84 min × ng/ml, respectively; Figure 7B). In comparison, poor OCT1 transporters showed 1.89-fold higher AUCs for fenoterol. This data suggests that compared to the well-known loss-of-function coding variants, the 8 bp duplication shows only limited effects on drugs pharmacokinetics.