Impact of ABCB1 1236C > T-2677G > T-3435C > T polymorphisms on the anti-proliferative activity of imatinib, nilotinib, dasatinib and ponatinib

Overexpression of ABCB1 (also called P-glycoprotein) confers resistance to multiple anticancer drugs, including tyrosine kinase inhibitors (TKIs). Several ABCB1 single nucleotide polymorphisms affect the transporter activity. The most common ABCB1 variants are 1236C > T, 2677G > T, 3435C > T and have been associated with clinical response to imatinib in chronic myelogenous leukaemia (CML) in some studies. We evaluated the impact of these polymorphisms on the anti-proliferative effect and the intracellular accumulation of TKIs (imatinib, nilotinib, dasatinib and ponatinib) in transfected HEK293 and K562 cells. ABCB1 overexpression increased the resistance of cells to doxorubicin, vinblastine and TKIs. Imatinib anti-proliferative effect and accumulation were decreased to a larger extent in cells expressing the ABCB1 wild-type protein compared with the 1236T-2677T-3435T variant relatively to control cells. By contrast, ABCB1 polymorphisms influenced the activity of nilotinib, dasatinib and ponatinib to a much lesser extent. In conclusion, our data suggest that wild-type ABCB1 exports imatinib more efficiently than the 1236T-2677T-3435T variant protein, providing a molecular basis for the reported association between ABCB1 polymorphisms and the response to imatinib in CML. Our results also point to a weaker impact of ABCB1 polymorphisms on the activity of nilotinib, dasatinib and ponatinib.

vinblastine) and anthracyclines (e.g. doxorubicin) 12 . They were also suggested to play a role in the development of resistance against TKIs (i.e. imatinib, nilotinib and dasatinib) [13][14][15] . Beside the expression level of ABCB1, polymorphisms may also modulate ABCB1 activity and by consequence drug efficiency. More than 60 coding single nucleotide polymorphisms (SNPs) have been reported in the ABCB1 gene (www.pharmgkb.org) [16][17][18][19] . The three most common variants in the ABCB1 coding region are rs1128503 (1236C > T, Gly412Gly), rs2032582 (2677G > T/A, Ala893Ser/Thr) and rs1045642 (3435C > T, Ile1145Ile). They present a Minor Allele Frequency (MAF) of approximatively 50% in the Caucasian population and are in linkage disequilibrium. Several clinical trials have studied the impact of these three SNPs on the clinical response to imatinib. In one study, it was shown that the 1236C-2677G-3435C wild-type haplotype is associated with a decreased rate of major molecular response to imatinib (decreased frequency from 70% to 44.6%). In the same report, homozygous patients for the allele 1236T presented the best molecular response and the highest imatinib plasma concentrations 20 . Another study also showed that the 1236C-2677G-3435C haplotype was associated with higher resistance to imatinib 21 . However, other reports, including two meta-analyses, failed to confirm the impact of this haplotype either on the molecular response or on drug resistance in patients treated with imatinib 22,23 . Consequently, the effect of these SNPs towards imatinib remains controversial. However, results based on population studies are sometimes indecisive mainly because of the presence of numerous uncontrolled confounding factors. As a complement to population studies, recombinant cell lines are very useful to test the functional impact of genetic variants. Along this idea, two studies have analysed the TKIs transport activity of ABCB1 in transfected cultured cells. However, they did not confirm the involvement of the 1236T-2677T-3435T variant in imatinib transport or anti-proliferative effect 24,25 .
To further analyse the impact of ABCB1 polymorphisms, we set up two different cell models using HEK293 (human embryonic kidney) and K562 (human erythroleukemic) cell lines, as previously described 26,27 . HEK293 is a commonly used model to test ABC transporter variants whereas K562 is derived from a human myeloid leukaemia carrying the BCR-ABL1 fusion and is therefore particularly suitable for testing the activity of TKIs. In this report, we have evaluated the influence of these SNPs on ABCB1 activity towards imatinib. We have also tested other TKIs, nilotinib, dasatinib and ponatinib, which are reported ABCB1 substrates [28][29][30] . Little information is available regarding the impact of ABCB1 polymorphisms on these three drugs.

Generation of ABCB1 1236C > T-2677G > T-3435C > T recombinant cell lines. After transfection
of HEK293 and K562 cells with pcDNA3.1 and pEF-myc-cyto vectors (ABCB1 C-G-C , ABCB1 C-G-T , ABCB1 C-T-T or ABCB1 T-T-T ) respectively, recombinant cell lines expressing ABCB1 (thereafter called HEK C-G-C , HEK C-G-T , HEK C-T-T , HEK T-T-T or K562 C-G-C , K562 C-G-T , K562 C-T-T , K562 T-T-T ) or cell lines transfected with the empty vector (called HEK pcDNA3.1 or K562 pEF ) were selected in the presence of G418. Similar ABCB1 surface expression was ensured by sorting recombinant cells by fluorescence activated cell sorting (FACS) with fluorescence parameters gated on the same level of intensity. As depicted in Fig. 1a,b, comparable surface protein expression levels were demonstrated by analytic flow cytometry in recombinant models. No fluorescence signal was detected in HEK pcDNA3.1 or K562 pEF cell lines, suggesting negligible endogenous expression.
The subcellular localization of wild-type and variant ABCB1 proteins was evaluated by immunofluorescence staining of HEK293 cell lines (Fig. 2). A circular fluorescent staining was observed in recombinant models ( Fig. 2b-e) and indicated a membrane localization of ABCB1.

Impact of ABCB1 1236C > T-2677G
> T-3435C > T polymorphisms on the intracellular accumulation of rhodamine 123. We evaluated the ability of 1236C > T-2677G > T-3435C > T variants to export rhodamine (Rh) 123, a well characterized fluorescent substrate of ABCB1 26 . After incubation in the presence of Rh123, fluorescence levels were lower in all recombinant HEK293 cell lines compared to controls (Fig. 3a, p < 0.001), indicating a higher Rh123 efflux in ABCB1 transfected cells. LY335979, a specific ABCB1 inhibitor, restored Rh123 intracellular fluorescence in transfected cell lines (Fig. 3a, p < 0.001), indicating that the differences in fluorescence intensity can be ascribed to ABCB1 expression in HEK C-G-C , HEK C-G-T , HEK C-T-T and HEK T-T-T . There was no significant difference between the variants.
The same results were obtained in K562 cells. Indeed, we observed a lower fluorescence level in the ABCB1 transfected cell lines compared to control cell lines (Fig. 3b, p < 0.001) and these differences were abolished when ABCB1-mediated efflux was inhibited by LY335979 (Fig. 3b, p < 0.001).

Impact of ABCB1 1236C > T-2677G
> T-3435C > T polymorphisms on the cytotoxicity of doxorubicin and vinblastine. Since ABCB1 has been reported to transport doxorubicin and vinblastine, we assessed the influence of ABCB1 variant expression on K562 cell proliferation in the presence of these drugs to further characterize our model. We demonstrated that recombinant cells were more resistant to these anticancer drugs, compared to control cell lines (Fig. 4a [10 to 270 nM], p < 0.001; Fig. 4b, p < 0.001). The effect was particularly spectacular for low doses of doxorubicin. Furthermore, we observed no consistent difference in resistance between K562 C-G-T , K562 C-T-T and K562 T-T-T compared to K562 C-G-C (Fig. 4, p > 0.05). These observations suggest that variants do not alter the ABCB1 efflux activity towards these two specific substrates.

Impact of ABCB1 1236C > T-2677G
> T-3435C > T polymorphisms on anti-proliferative effects of tyrosine kinase inhibitors. Since ABCB1 has been reported to transport TKIs, we investigated the impact of ABCB1 variant expression on K562 cell proliferation in the presence of TKIs that target BCR-ABL namely imatinib, nilotinib, dasatinib and ponatinib. We observed that recombinant cells were more resistant to imatinib, compared to control cell lines (Fig. 5a), confirming that imatinib is a substrate of ABCB1. We next compared the four ABCB1 variants. The most striking effect was the resistance of K562 C-G-C to imatinib compared with K562 C-G-T , K562 C-T-T and K562 T-T-T (Fig. 5a). These results were confirmed by calculating half maximal Scientific RepoRts | 6:29559 | DOI: 10.1038/srep29559 inhibitory concentration (IC 50 ) for each cell line (Table 1). This effect was observed at concentrations that are clinically relevant. Our data suggest an increased activity of the wild-type protein (encoded by the CGC haplotype) towards imatinib compared with variant proteins.
ABCB1 expression also increased cell proliferation in the presence of nilotinib, dasatinib and ponatinib (Fig. 5). This increase was consistently observed in three independent experiments, although it was statistically significant only for dasatinib (Table 1) Table 1). These observations suggest that these polymorphisms have a much weaker impact on the transport of nilotinib, dasatinib and ponatinib by ABCB1, compared to imatinib.

Impact of ABCB1 1236C > T-2677G
> T-3435C > T polymorphisms on the intracellular accumulation of imatinib and nilotinib. We next sought to determine whether these polymorphisms affect the (e) ABCB1 C-G-C cells were stained with anti-ABCB1 antibody (green fluorescence). DAPI was used to stain nuclei (blue). intracellular accumulation of TKIs in transfected cell lines. We obtained radiolabelled imatinib and nilotinib (dasatinib and ponatinb were not available). As depicted in Fig. 6a,b, imatinib or nilotinib intracellular concentrations were strongly decreased in HEK293 transfected cell lines expressing the ABCB1 protein when compared to control cells (Fig. 6a,

Discussion
In this study, we show that ABCB1 1236C > T-2677G > T-3435C > T polymorphisms affect the sensitivity of human leukemic cells expressing BCR-ABL to TKIs, providing a molecular basis for the previously reported associations between this ABCB1 haplotype and the patient response to imatinib 20,21 .
First, HEK293 and K562 recombinant cell lines expressing the ABCB1 C-G-C , ABCB1 C-G-T , ABCB1 C-T-T or ABCB1 T-T-T haplotype were generated and carefully validated. As a control, we showed that the four variants decreased Rh123 concentration in a similar manner, in agreement with published data 31 . We also showed that all variants conferred a similar resistance to doxorubicin and vinblastine in concordance to a previous report showing that the efflux of vinblastine is not affected by these ABCB1 polymorphisms 31 .
We next used these validated models to assess the impact of ABCB1 variation on TKIs anti-proliferative activity. Our results confirmed that imatinib is a good ABCB1 substrate, as previously described 13,[32][33][34][35] . Indeed, ABCB1 expression decreased K562 cell sensitivity to imatinib as well as the intracellular accumulation of radiolabelled imatinib in HEK293. Furthermore, we showed that the wild-type protein (ABCB1 C-G-C ) conferred higher imatinib resistance compared to the variant protein (ABCB1 T-T-T ). Consistently, imatinib intracellular concentrations in cells expressing this variant protein were also significantly higher than in cells expressing the wild-type. These observations suggest that the variant haplotype decreases imatinib transport by ABCB1 and provide an explanation for previous in vivo studies that associated the wild-type haplotype (CGC) to imatinib resistance 20-23 .  Nilotinib, dasatinib and ponatinib are also reported as ABCB1 substrates but have been less studied than imatinib in this context 13,[28][29][30]34 . In our assays, nilotinib intracellular accumulation was much reduced upon ABCB1 expression. The cytotoxicity induced by nilotinib, dasatinib and ponatinib was also decreased, albeit  Table 1. to a variable extent, in cells expressing ABCB1. Moreover, polymorphisms did not affect nilotinib intracellular concentrations and had limited influence on the anti-proliferative effect of these three drugs, in contrast with imatinib. This observation suggests that the studied ABCB1 polymorphisms significantly affect the cell response to imatinib but not, or to a much lesser extent, to nilotinib, dasatinib, ponatinib, doxorubicin and vinblastine. This was reminiscent of our previous study on ABCB1 1199G > A SNP, which had also demonstrated a differential effect of ABCB1 for various substrates 26 . Among the three investigated coding SNPs, rs1128503 (1236C > T, Gly412Gly), rs2032582 (2677G > T/A, Ala893Ser/Thr), and rs1045642 (3435C > T, Ile1145Ile), only the 2677G > T SNP is associated to an amino acid substitution. Interestingly, in our study, we observed an impact of the 3435C > T synonymous SNP on the anti-proliferative activity of imatinib (see above) but not of the 2677G > T non-synonymous SNP. The functional impact of ABCB1 3435C > T synonymous SNP has been clarified in several studies [36][37][38] . Indeed, despite the fact that this is a synonymous SNP, this variant was shown to alter the kinetics of translation of the protein via the insertion of a rare codon. More precisely, this SNP was suggested to create a translational pause, which could slowdown translation by the ribosome and modify the kinetics of protein folding by chaperones 36 . We could however not confirm the importance of this particular polymorphism in the imatinib intracellular accumulation assay, possibly because this test is less sensitive and more prone to experimental variations.

ITK
Although the resistance to TKIs conferred by ABCB1 expression was not as striking as for doxorubicin, adapting imatinib dose according to ABCB1 genotype might be interesting for CML patients. Moreover, we showed that 2 nd generation TKIs are less dependent on the studied ABCB1 genetic polymorphisms, which are very frequent in the Caucasian population. Future studies should establish whether ABCB1 genotype affects the inter-patient variability of the response to TKI.
In summary, our in vitro results show that the ABCB1 1236C > T-2677G > T-3435C > T polymorphisms affect the anti-proliferative activity and the intracellular accumulation of imatinib.

Material and Methods
Material. Imatinib, nilotinib and dasatinib were purchased from LC Laboratories (Woburn, United States).
Generation of stable recombinant cell lines. HEK293 and K562 cell lines were obtained from ATCC and grown as previously described 26 . HEK293 and K562 cell lines were transfected with pcDNA3.1 and pEF-myc-cyto vectors, respectively and further selected in the presence of G418 (pcDNA3.1, 1 mg/ml and pEF-myc-cyto, 1.5 mg/ml) according to our previously published method 26,39 . Cells were then sorted as described below.
Characterization of ABCB1 expression. Flow cytometry. This experiment was performed as previously described 26  Immunofluorescence. This assay was performed as previously reported 26 , with slight modifications. One day before the experiment, HEK293 cells were plated at a density of 5 × 10 4 cells/well in complete medium. where y is the percentage of inhibition of proliferation compared to untreated cells, IC 50 is the concentration that is supposed to produce 50% inhibition, x is the drug concentration in logarithm and nH is the Hill coefficient. The coefficients where estimated by nonlinear adjustment with the use of JMP Pro version12 statistical Software based on the maximum likelihood estimation.
TKI accumulation. One day before the experiment, 3.5 × 10 5 transfected HEK293 cells were seeded in poly-L-lysine-coated 24-well plates in complete medium. 14 C-imatinib or 14 C-nilotinib were diluted with an equal amount of cold compound and added at five different total concentrations (from 0.3 to 5 μ M and from 0.0625 to 1 μ M, respectively) and cells were incubated for 120 min at 37 °C, 5% of CO 2 . After incubation with imatinib or nilotinib, the cells were washed two times in cold PBS. After centrifugation, the supernatant was discarded and cells were detached with ice-cold lysis buffer (0.1% triton X100 and 0.1% sodium deoxycholate). Radioactive imatinib or nilotinib were quantified in cell lysis using a Tri-Carb liquid scintillation β counter (Perkin Elmer), after addition of 4 ml of Ultima Gold liquid scintillation cocktail. The absolute amount of drug present in cell extracts was normalized to the amount of protein as quantified using the BCA kit (Thermo Scientific).
Statistical analysis. Experiment results are presented as mean with standard deviation. GraphPad InStat (Version 3.05) was used for statistical analyses (Fig. 3 and Table 1). Analyses of variance were used under the null hypothesis that the means of the compared groups were equal. Student-Newman-Keuls tests were performed when the differences between means were significant. JMP Pro (Version 11) was used for statistical analyses (Figs 4-6). Drug concentrations (in DPM per mg protein) and relative cell proliferation (%) of different cell lines were compared with a mixed-model analysis. The model was built on the maximum likelihood ratio, with cell line as the fixed factor, the replicate as the subject analysed and tested drug concentration as the repeated measurement. No particular structure was imposed on the variances and covariances between and within the tested concentrations of the repeated measurements. To test the overall effect of the haplotype, Dunnett post-Hoc Scientific RepoRts | 6:29559 | DOI: 10.1038/srep29559 comparison was used with empty vector transfected control sets as the reference. When indicated (see results), we restricted the statistical analyses over a range of tested concentrations (nM or μ M) for which ABCB1 effect was noticeable and not saturated.