ABCC5 Gene Variant c.1146A>G Reduces MRP5 Expression and Can be a Potential Marker to Manage Cilostazol Induced Headache

Background: Patients taking cilostazol, a representative phosphodiesterase type III inhibitor used for vasodilation, occasionally complain of headaches. Cerebral arteriolar relaxation, due to an increase in intracellular cAMP or cGMP, is believed to be associated with a severe form of cilostazol-induced headaches. Multidrug resistance protein 5 (MRP5) is an important regulator of cAMP and cGMP could be a regulatory protein of this cilostazol induced headache. Methods: The response to cilostazol on the basis of MRP5 genetic variations was studied in phase-I clinical trial including 101 healthy Korean individuals. Quantitative real time PCR, Western blot, confocal analysis and drug transporter assay was performed to detect and evaluate the activity of MRP5. Statistical analysis was performed by using SPSS 18.0 and GraphPad Prism 4.0 software. Results: Population genetic and pharmacokinetic analyses indicated that a group of MRP5 genetic variations in linkage disequilibrium with c.1146A>G (rs7636910) is associated with the no or mild forms of cilostazol-induced headaches, but did not affect the systemic distribution of cilostazol or its metabolites. Quantitative real-time PCR and pyrosequencing assays of blood cells revealed that individuals with the c.1146G allele, a protective allele against cilostazol-induced headaches, had a 0.68-fold lower mRNA expression of MRP5 than that of individuals with c.1146A allele. In addition, in a gene transfection experiment using MDCKII cells, the G variant was found to be reduced the mRNA and protein expression of MRP5. Conclusion: These results suggest that MRP5 has an important role in the regulation of cilostazol-induced chronic headache and the c.1146A>G variation could be a potential marker to manage/treat cilostazol-induced headaches. Trial


Background
Cilostazol is a representative phosphodiesterase (PDE) type III inhibitor that is used widely to mitigate ischemic symptoms in patients with intermittent claudication and as an antiplatelet agent for stroke prevention [1,2]. The post-marketing surveillance has identified several adverse drug reactions (ADRs) of cilostazol that occur in the cardiovascular, digestive, nervous, respiratory, urinary systems, and in skin. Additionally, headaches were the most frequently reported reason for cilostazol withdrawal [1,3].
Several studies reported that cilostazol induced a migraine-like headache. Those reports suggested the mechanism of headache was due to accumulation of cyclic adenosine monophosphate (cAMP) in brain cells that induced by cilastazol [4][5][6]. The PDE III enzymes catalyzes the degradation of both cyclic adenosine monophosphate (cAMP) and cyclic guanosine monophosphate (cGMP). Therefore, cilostazol, an inhibitor of PDE III, can increase the intracellular levels of cAMP and cGMP. Both of cGMP and cAMP, which are synthesized in the nitric oxide (NO)-guanylate cyclase pathway and adenylate cyclase pathway, respectively, can cause headaches resulting from cerebral vascular dilation through regulating the intracellular status of cyclic nucleotides in the vascular smooth muscle [7][8][9].
The ATP-binding cassette (ABC) transporter superfamily is a group of transport proteins located in the cell membrane that mediate the extracellular transport of various endogenous and exogenous substances. Thus, ABC transporters are considered as one of the main causes of drug resistance mechanism as they alter the intracellular pharmacokinetics of substrate drugs [10]. Multidrug resistance proteins (MRPs) are members of the ABC transporters family. Of the transporters, MRP4 and 5 have been shown to mediate ATP-dependent cGMP efflux in the cerebral arteries and peripheral tissues [8,11]. In fact, it was reported that overexpression of MRP4 and 5 caused resistance to nucleoside analog drugs in cancer chemotherapy and antiviral treatment [12,13]. If this is the case, it can be hypothesized that the MRP4 and 5 proteins may also be involved in headaches associated with the nucleoside analog-mediated mechanism.
Although cilostazol is not a nucleoside analog, metabolism and disposition of this drug and its metabolites are partly regulated by MRP4 and thus also can be by MRP5 [14]. Therefore, the function or expression of MRP4 and 5 is likely to affect the occurrence or extent of cilostazol-induced headaches. However, the genetic polymorphisms in genes encoding nucleoside analog drug transporters are a known cause of individual variation associated with the development of headaches, there is no report yet regarding these proteins related occurrence of headaches [15,16]. In the present study, the association of MRP5 genetic variations with the occurrence of headache induced by cilostazol treatment is evaluated in 101 healthy Korean participants. The study was also investigated a predictable marker causing cilostazol induced headaches, which may help the physician to take decision about treatment strategy.

Methods
The study plan was reviewed and approved by the institutional review board of Severance Hospital in the Yonsei University Health System, Seoul, Korea and the trial was registered at www.clinicaltrials.gov (the clinical trial registry number is NCT01455558).

Pharmacokinetics
Six separate clinical trials comparing the PK profiles between newly developed sustained release (SR) formulations and an immediate release (IR) formulation of cilostazol in healthy Korean participants were completed at the Yonsei University Severance Hospital (Seoul, Korea) from August 2008 to August 2009, however the data only from IR formulation were used in this study [17]. In brief, each participant received two IR (100 mg × two tablets, BID) formulations of cilostazol (PLETAAL), with one tablet administered every 12 h. Venous blood samples were collected into heparin tubes at 0, 1, 2, 3, 4, 6, 8, 10,12,13,14,15,16,18,20,22,24,36,48, and 72 h after drug administration. The plasma concentrations of cilostazol and its metabolites (OPC-13015 and OPC-13213) were measured by using liquid chromatography tandem mass spectrometry (LC-MS/MS), and PK profiles, including the parameters C m ax , AUC inf , and t 1/2 , were computed by using Phoenix WinNonlin version 5.3 (Pharsight Corporation, Mountain View, CA).

Participants
The volunteers who had participated in this clinical trials were healthy Korean males or females between 19 and 55 years of age, within 20% of their ideal body weight, and without congenital abnormalities or chronic diseases (17). Informed consent for the analysis of their genetic information was obtained from all 101 participants. In total, 33 volunteers that gave consent for MRP5 mRNA analysis were newly recruited, in accordance with the same criteria as those of clinical trials, because this study using clinical trial data was planned retrospectively.

Adverse Drug Reactions
Basic vital signs (blood pressure, body temperature, and pulse rate), a personal symptom interview, electrocardiograms (ECGs), pregnancy tests (human chorionic gonadotropin in blood) for females, and laboratory tests (hematology, blood chemistry, and urinalysis) were performed at appropriate times. An ADR was defined as any unfavorable result in the checked information and was recorded on the case-report forms [17]. All participants who had enrolled in the clinical trials were questioned about headaches by using the 10-point VRS during a personal interview. The headache intensity was categorized into four groups, no headache (0), mild headache (1-3), moderate headache (4-6), and severe headache (7)(8)(9)(10) according to the numeric rating scale (NRS) used in an earlier report (18). For statistical analysis, the four groups were re-categorized into two groups: no or mild headache (0-3), and moderate or severe headache (4-10).

Immunoblotting
HEK293 cells (1×10 6 cells) were seeded in 60 mm plates at the day before transfection and then transfected for 36 h by using Lipofectamine 2000. The cells were washed with ice-cold phosphate-buffered saline (PBS), lysed in radioimmunoprecipitation assay (RIPA) lysis buffer containing 0.5% sodium dodecyl sulfate (SDS), 1% Nonidet P-40, 1% sodium deoxycholate, 150 mM Tris-hydrochloride (HCl, pH 7.4), and protease inhibitors. The protein concentration was determined by using Bio-Rad protein assay dye reagent. The proteins were separated by using SDS-polyacrylamide gel electrophoresis (PAGE) in a 4%-12% gradient gel (Invitrogen) and transferred to PVDF membranes (Amersham Biosciences). Non-specific binding of protein to the membranes was blocked by 5% nonfat skim milk for 1 h incubation at 37°C. The membranes were washed three times (each of 10 min) with tris-buffered-saline (TBS) containing tween 20 (0.1 %) and probed with anti-MRP5 (1:500, sc-20769, Santa Cruz Biotechnology), anti-beta actin antibodies (1:10,000, A5441, Sigma) for 30 min at 37°C. The membranes were washed three times again as mentioned before and incubated with horseradish-peroxidase-conjugated anti-mouse IgG secondary antibody (1:2,000, Santa Cruz) for 2 h at room temperature. The membranes were then exposed to a medical x-ray film using the enhanced chemiluminescence (ECL) system (Santa Cruz).

Immunocytochemistry
HEK293 cells (1×10 6 cells) were seeded in 60 mm plates containing two microscope cover glasses at the day before transfection. The cells were then transfected with cDNA Diego, CA) software packages were used for statistical analyses. All genotyped SNPs underwent Hardy-Weinberg equilibrium testing before case-control analysis. Categorical data were evaluated by using the chi-square and Fisher's exact tests, whereas ANOVA, ttest, and Kruskal-Wallis test were applied to continuous data. A significant effect of a certain genotype on the clinical phenotype was estimated as an OR with a 95% CI by using a binary logistic regression method. The fundamental standard P value of significance was based on two-sided comparisons and was selected as 0.05. However, the P value was adjusted with Bonferroni's correction when multiple statistical analyses were performed at once.

Demographic and clinical characteristics of healthy volunteers
A 10-point verbal rating scale (VRS) was used to evaluate the severity of headaches. In total, 101 healthy human volunteers who had received a standard preparation of cilostazol (PLETAAL) as participants of bioequivalence trials of new cilostazol regimens were divided into two groups: 1) those who experienced no or mild headaches (n = 68, VRS score 0-3); and 2) those who experienced moderate to severe headaches (n = 33, VRS score 4-10).
The baseline demographic characteristics were not statistically different between the two groups ( Table 1). The mean age of the volunteers was 24.8 years, and 84 (83.2%) were male. The mean height and weight were 173.2 cm and 67.3 kg, respectively. The number of participants those regularly smoked and consumed alcohol were 27 (26.7%) and 71 (70.3%), respectively.

MRP5 variations associated with cilostazol-induced headaches
Although almost 1360 single nucleotide polymorphisms (SNPs) have been identified in the human ABCC5 gene (the gene that codes for the MRP5 protein), the variant selected in this study was yet to find any association with cilostazol induced headache (www.genecards.org). Thirteen genetic variations in the promoter and coding regions of MRP5 were found in our 101 clinical trial participants by gene scanning (Fig. 1). Of these 13, three SNPs (c.222G>A, c.2714T>G, c.2847C>T) were excluded from the study because of their low frequency (minor allele frequency <5%). The c.4896G>A and c.5557A>G SNPs located in the 3′ untranslated region had perfect linkage disequilibrium (LD) in the Korean population (D′ and r 2 = 1); c.4896G>A was used as a tag SNP. The genotype frequencies of each polymorphism are shown in Table 2

Logistic regression analysis with clinical variables
The G allele of c.1146A>G was associated with a lower chance of headaches than the A allele. The relative effect of MRP5 genetic variation on cilostazol-induced headaches was quantified by using logistic regression analysis that integrated all available clinical variables ( Table 3) immunoblotting, and the expression and location of MRP5 protein was confirmed through immunocytochemistry. In the control vector, low MRP5 mRNA and protein expression was found, with high expression in the wild-type (A allele), and lower expression with the G allele compared with the wild-type ( Fig. 4A and B). In immunocytochemistry, MRP5 protein expression was found along with the cell membrane (in the same position as the membrane marker P-cadherin) of wild type (A allele) MRP5 gene expressing HEK cells.
Whereas, the MRP5 protein expression was found to be very low in mutant (G allele) type of HEK cells (Fig. 4C).

Discussion
Cilostazol was administered as a reference drug to the study participants in six separate trials. The data from the volunteers who reported moderate or severe headaches after cilostazol administration were combined for analysis. Before the clinical trial started, the medical history of the participants was confirmed from a personal interview. The headache severity was evaluated by using a 10-point VRS and the participants were divided into two groups: those with no or mild headaches were assigned to the ADR-No group, and those with moderate or severe headaches were assigned to the ADR-Yes group. Participants with mild headaches were classified into the ADR-No group was because they may have felt that they had a mild headache when questioned about headaches during the personal interview. Moreover, people who are not taking any medications and with no known disease can experience mild headaches due to environmental factors such as stress, alcohol and caffeine withdrawal, neck stiffness, and light sensitivity [19]. Therefore, participants with mild headaches were separated from those with moderate or severe headaches to ensure that the headaches were due to cilostazol intake only.
In the genetic association test, cilostazol-induced headaches were found to be significantly associated with three SNPs after conservative Bonferroni correction for multiple testing (Table 2). However, only the c.1146A>G SNP was found to be positively associated with MRP5 mRNA expression in vivo (Fig. 2). Therefore, the relationship between these SNPs was examined to interpret the results. D′ and R-square were used most often to measure LD and were computed by using the Haploview program [20, 21].
As shown in Figure 1, three SNPs were considerably connected to each other, although not in perfect linkage. Indeed, the D′ value for all three SNPs was 1.0. The r 2 values were 0.28 between c.1146A>G (rs7636910) and c.1147+94G>A (rs55695073), 0.44 between c.1147+94G>A (rs55695073) and c.3624C>T (rs3749442), and 0.65 between c.1146A>G (rs7636910) and c.3624C>T (rs3749442). In the haplotype analysis of the three loci, the frequency of the GGC haplotype was 41.1%, showing an association with the no or mild headaches group, whereas the frequency of the AAT haplotype was 29.0%, and showed an association with the moderate or severe headaches group (Supplementary Table 1).
However, the frequencies of the AGC and AGT haplotypes were 20.3% and 9.7%, respectively. These two haplotypes showed no association with cilostazol-associated headaches, indicating that crossover would have had to occur in a region between c.1146A>G (rs7636910) and c.1147+94G>A (rs55695073). According to the results, the synonymous c.1146A>G SNP might have a practical function of the regulating MRP5 mRNA expression, and other SNPs might be connected to c.1146A>G but did not have this function.
With regard to the molecular mechanism of the c.1146A>G SNP, this variation is a synonymous SNP at position 382. However, c.1146A>G is located at the second to last nucleotide of the eighth exon within the U1 small nuclear RNA (snRNA) binding site [22]. Therefore, one possibility is that if the binding site of the U1 snRNA is changed by the mutation, the splice variant of MRP5 will be produced during the process of converting pre-mRNA to mature RNA. To confirm this, RT-PCR was performed by using cDNA samples synthesized from the mRNA of fresh human blood. However, the length of the splicing product was identical from the AA, AG, and GG genotypes, and a new splice variant of MRP5 was not identified (Supplementary Fig. 1). The other possibility was that synonymous variations in the coding regions can alter mRNA expression. Therefore, we performed a pyrosequencing assay of samples from heterozygous individuals. This is an allele-specific expression analysis method that is much more sensitive to relatively small changes in expression, whereas across-sample comparisons include much more noise (biological variation between samples and technical noise from normalizing to a housekeeping gene) and can only detect larger differences in expression. Thus, for the allelic expression differences in MRP5 an evaluation of whether the c.1146A>G SNP was associated with allelic expression differences was performed. It was found that c.1146G resulted in significantly lower MRP5 mRNA expression than c.1146A (Fig. 3). The differences in the two allele-specific transcripts were as large as ~1.5-fold.
It was necessary to confirm MRP5 mediated transportation of cilostazol to conclude whether the pharmacokinetics of cilostazol affected headaches or not. Studies were performed to examine whether cilostazol could be transported via MRP5 in MDCKII-MRP5 cells. The protein expression of MRP5 in MDCKII-MRP5 cells was found to be increased depending on the total protein loaded, and MRP5 was found to be expressed in along with the cell membrane (Supplementary Fig. 2A and B). These results showed that in vitro test system was capable of detecting substrates of MRP5. To confirm the behavior of transient MDCKII-MRP5 cell systems, the efflux ratio (P app basal to apical/P app A to B) of carboxy dichlorofluorescein (1 μM CDF, probe substrate for MRP5) in MDR5-expressing cells was determined and was found to be 2.0 fold higher than that of apical to basal ratio ( Supplementary Fig. 2C). The presence of 100 μM MK-571 (a reference inhibitor for MRP5) significantly decreased the B to A transport of CDF from 264.5 × 10 -6 to 92.2 × 10 -6 cm/s (p < 0.01). Both A to B and B to A direction transport rates of cilostazol across MDCKII-MRP5 cells were increased in a concentration-dependent manner. As shown in Supplementary Figure 2D and Supplementary Table 2, the efflux ratio (P app B to A/P app A to B) for 10, 30, 60, and 100 μM cilostazol was 0.8, 1.1, 0.9, and 1.4, respectively.
According to the US FDA guidance for industry, to be a substrate of efflux transporter protein, the efflux ratio cut off value (P app B to A/P app A to B) should be above 2 [23]. So, the results in this experiment indicated that the cilostazol is not a substrate of MRP5 transporter.
Accordingly, the concentration of cilostazol and its active metabolites in blood were compared for each genotype of c.1146G>A (Supplementary Fig. 3). Various PK parameters of cilostazol and its active metabolites, including the maximum observed plasma concentration (C m ax ), the area under the plasma concentration-time curve from time zero to infinity (AUC inf ), and the terminal elimination half-life (t 1/2 ), were not so different, suggesting that the MRP5 polymorphism was not related to the systemic distribution of cilostazol or its metabolites (Supplementary Table 3).
At the beginning of this study, the correlation between MRP4 as well as MRP5 genetic variants and cilostazol-induced headaches was identified. The MRP4 gene was analyzed for 17 genetic variants present in Koreans. However, no genetic variation was found to be significantly related with headache (Supplementary Table 4).
NO is a key molecule in primary headaches, and the vasodilation effect of NO is the main cause of headaches [7,24]. The role of NO in the regulation of vascular tone can be categorized into two major mechanisms [25]. Several studies have suggested that NO regulates vascular tone without affecting the NO/cGMP pathway [26][27][28][29]. However, other research suggested that NO/cGMP pathway can regulates the mechanism of headache, and our study focused on a cGMP transporting protein as the factor controlling headaches [30,31]. Xu et al. reported that vessel diameters were increased to a smaller extent in MRP5knockdown rats than in control rats when using the NO donor S-nitroso-Nacetylpenicillamine (SNAP) in native conditions [32]. An increase in cAMP reduced SNAPinduced vasodilation in MRP5 knockdown rats [8]. These data are consistent with our results showing that participants receiving cilostazol with the c.1146A>G mutant had fewer headaches and lower MRP5 mRNA expression than participants with the wild-type allele. In MRP5-knockdown rats, the reduction in cGMP-induced vasodilatation may be due to increased PDE5 activity and increased cAMP levels, although the exact mechanisms require further investigation. A similar analogy may explain the protective effect of the c.1146G allele on cilostazol-induced headaches. Cilostazol is a specific PDE III inhibitor.
PDE III is referred to as a "cAMP-preferring" enzyme and is most highly expressed in vascular smooth muscle cells [33]. A role for cAMP in the promotion of NO/cGMPdependent vasodilation was proposed, in which cAMP mainly inhibited cGMP transportation to the outside of cells through MRP5 and also enhanced the PDE5-mediated cGMP breakdown to 5′-GMP [8]. Therefore, participants with the c.

Conclusion
In total, 101 volunteers who had completed clinical trials of cilostazol were analyzed to determine whether headaches had occurred or not. Statistical significance of the c.1146G

Funding
The study had no funding. a χ 2 or t-test used as appropriate.     The MRP5 protein expression pattern was compared with the MRP5 wild-type allele and c.1146A>G by using immunocytochemistry. Pan-cadherin was used as a membrane marker. Scale bar = 10 μm.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.