Uptake/Efflux Transport of Tramadol Enantiomers and O-Desmethyl-Tramadol: Focus on P-Glycoprotein

Abstract: The analgesic effect of tramadol (TMD) results from the monoaminergic effect of its two enantiomers, (+)-TMD and (−)-TMD as well as its opioid metabolite (+)-O-desmethyl-tramadol (M1). P-glycoprotein (P-gp) might be of importance in the analgesic and tolerability profile variability of TMD. Our study investigated the involvement of P-gp in the transepithelial transport of (+)-TMD, (−)-TMD and M1, using a Caco-2 cell monolayer model. The bidirectional transport of racemic TMD and M1 (1–100 µM) across the monolayers was investigated at two pH conditions (pH 6.8/7.4 and 7.4/7.4) in the presence and absence of P-gp inhibitor cyclosporine A (10 µM) and assessed with the more potent and specific P-gp inhibitor GF120918 (4 µM). Analytical quantification was performed by liquid chromatography coupled to the fluorescence detector. A net secretion of (+)-TMD, (−)-TMD and M1 was observed when a pH gradient was applied (TR: Papp(B − A)/Papp(A − B): 1.8–2.7; P < 0.05). However, the bidirectional transport of all compounds was equal in the non-gradient system. In the presence of P-gp inhibitors, a slight but significant increase of secretory flux was observed (up to 26%; P < 0.05) at both pH conditions. In conclusion, (+)-TMD, (−)-TMD and M1 are not P-gp substrates. However, proton-based efflux pumps may be involved in limiting the gastrointestinal absorption of TMD enantiomers as well as enhancing TMD enantiomers and M1 renal excretion. A possible involvement of uptake carriers in the transepithelial transport of TMD enantiomers and M1 is suggested.

Tramadol hydrochloride (TMD) is a centrally acting analgesic structurally related to codeine and morphine [1]. Experimental and clinical studies have demonstrated the analgesic efficacy and tolerability of TMD in acute and chronic nociceptive and neuropathic pain of both malignant and non-malignant origin [2][3][4]. TMD is mainly metabolised by O-and N-demethylation and by conjugation reactions, forming glucuronides and sulfates. The O-demethylation of TMD to its opioid active moiety O-desmethyl-tramadol (M1) is catalysed by cytochrome P450 (CYP) 2D6, whereas Ndemethylation to N-desmethyl-TMD (M2) is catalysed by CYP2B6 and CYP3A4 ( fig. 1) [5,6]. Both enantiomers of TMD, (+)-TMD and ())-TMD, as well as its major metabolite M1 contribute to its analgesic activity via different mechanisms. M1 acts mainly as a l opioid agonist (Ki: 0.0034 lmol/l) whereas (+)-TMD and ())-TMD inhibit serotonin and norepinephrine reuptake, respectively (Ki: 0.53 and 0.43 lmol/l), thus enhancing the inhibitory effect on pain transmission in the spinal cord [7,8]. The complementary and synergistic actions of the three compounds confer a unique pharmacodynamic profile to TMD [9].
However, wide variability in the pharmacokinetic/ pharmacodynamic properties of TMD has been shown.
P-glycoprotein (P-gp/MDR1/ABCB1) is a broad-spectrum efflux transporter that limits substrate xenobiotic exposure in the body, thus playing a crucial role in drug pharmacokinetics and biodisposition [16]. It has been shown to stereoselectively transport numerous drug enantiomers in several in vitro and in vivo studies [17][18][19][20]. The potential stereoselective transport of TMD enantiomers by an ABC efflux drug transporter as well as a preferential transport of TMD and M1 have not yet been investigated. These may play non-negligible roles in the biodisposition of each compound and, therefore, the final expected central analgesic effect of the drug racemate and its active opioid metabolite M1. Several opioid analgesics have been shown to be substrates as well as inhibitors of P-gp in studies using the in vitro Caco-2 monolayer model [21][22][23].
Measurement of transepithelial electrical resistance (TEER). TEER was checked every 5 days during the 21-day monolayers maturation. Prior to bidirectional transport studies, medium was removed from both apical and basolateral chambers and monolayers were rinsed three times with the transport buffer Hank's balanced salt solution (HBSS) supplemented with 25 mM N-(2-hydroxylethyl)piperazine-N¢-2ethane-sulfonic acid (HEPES) (Gibco BRL) and pH-adjusted to 7.4 with 0.5 M NaOH. Cells were equilibrated in the same buffer for 1 hr and the integrity of each monolayer was checked by measuring its transepithelial electrical resistance (TEER) with a Millicell-ERS ohmmeter (Millipore Corp., Bedfort, MA). Resistance was also checked immediately after the transport experiments.
Transmission electron microscopy and Western blotting of P-gp. Transmission electron microscopy and Western blotting of P-gp were performed in our earlier study in the context of the validation of our Caco-2 cell model. Histological examination showed a continuous, differentiated cell monolayer presenting microvillus on the apical cell surface, interdigitations, numerous desmosomes (maculae adherents) and tight junctions. Western blot analysis revealed a C219 antibody-reactive band of 170 kD corresponding to P-gp expression [24].
Transport studies. The bidirectional transport of the two enantiomers of TMD and M1 (1-100 lM) was investigated at pH 6.8/7.4 and 7.4/7.4 in the presence and absence of cyclosporine A, at respectively 1-100 and 4-400 times the plasma concentrations (Cmax) that produce a clearcut analgesic effect after a 100 mg dose of TMD in vivo [25]. The pH gradient condition (6.8/7.4) was used to examine whether pH-mediated efflux mechanisms are involved in TMD and M1 transport across pH gradients that exist in intestinal and kidney transport [26,27]. The bidirectional transport of all molecules was assessed at the lowest concentration (1 lM) with GF120918 at physiological pH (7.4/7.4) to confirm P-gp interaction with the drugs tested. Indeed, GF120918 does not inhibit the multi-resistance proteins (MRPs) [28]. It may inhibit the breast cancer resistance protein (BCRP) but this transporter is expressed in very low extent in Caco-2 cells [29]. For the P-gp inhibition studies, the inhibitor was present in both chambers. Cyclosporine A and GF120918 were used at 10 and 4 lM, respectively. Samples (150 ll for TMD and 100 ll for M1) were removed from each receiver chamber at various times (30, 60, 90, 120 and 180 min) and replaced with buffer to maintain constant volumes. The 3-hr transport studies were performed at a constant agitation rate (50 rpm) using a circular shaker (type SSM1, Stuart Ò ) in an incubator (37°, 5% CO 2 and humidified atmosphere). pH 7.4/7.4 condition: As described for the pH 6.8/7.4 condition except that (A) and (B) buffers and drug solutions were made with HBSS supplemented with 25 mM HEPES pH 7.4.
After the transport studies, all aliquots were stored at -20°until analysis.

Paracellular transport of [ 3 H]-mannitol. Apical to basolateral permeability (A-B) of [ 3 H
]-mannitol, a radio-labelled paracellular marker (activity: 0.5 lCi/ml), was measured over 3 hr to monitor the integrity of monolayers tight junctions. Samples taken from the basolateral side (100 ll) at various times were counted in a Pico-Fluor 15 medium using a liquid scintillation counter (Packard Instruments).
Analytical method. Analysis of (+)-TMD, ())-TMD and M1 was performed by liquid chromatography coupled to the fluorescence detector. A stereo-selective method was applied for (+)-TMD and ())-TMD determination using a Chiral AGP column (100 · 4.0 mm), as previously described [30] with slight modifications. The mobile phase consisted of a mixture of acetonitrile and phosphate buffer, 100 mM pH 7.0 (2/98) and was delivered at 0.9 ml/min. M1 analysis was performed using a MN phenyl column (70 · 4.0 mm). The mobile phase consisted of a mixture of acetonitrile and orthophosphoric acid, 50 mM (15/85) pH 3.0, and was delivered at 0.8 ml/ min. For both TMD and M1, fluorescence was measured with emission and excitation wavelengths set at 275 and 308 nm, respectively. In all cases, 50 ll samples were directly injected into the HPLC system. Method performances in terms of reproducibility, repeatability and linearity were assessed before analysis (data not shown). Rhodamine 123 analysis was performed as previously described [24].
In all cases, the samples consisted of aliquots removed from the receiver chambers (drug solution in the aqueous buffer HBSS). No additional treatment was needed. No extraction was required and the samples were directly injected into the HPLC system without need for an internal standard.
Calculations. TEER was calculated from the following equation [31]: where the TEER mono is the cell monolayer and polycarbonate porous membrane resistance, TEER blank the polycarbonate porous membrane resistance and A the polycarbonate porous membrane surface area (1.13 cm 2 ).
Apical to basolateral (P app (A -B)) and basolateral to apical (P app (B -A)) apparent permeability coefficients were calculated according to Artursson [32] using the following equation: Where dQ/dt (lg/min) is the permeability rate of the drug, calculated from the regression line of the time points of sampling, A is the surface area of the monolayer (cm 3 ) and C 0 the initial drug concentration in the donor chamber (lg/l).
Karlsson [33] suggested the involvement of a drug efflux transporter in the investigated Caco-2 cells if the efflux ratio (TR ¼ P app (B -A)/ P app (A -B)) is >2 and if a decreased secretory transport rate (P app (B -A)) is observed in the presence of an inhibitor of this transporter. For a compound with an efflux ratio of 1.5-2.0, a positive effect of the inhibitor confirm the implication of the efflux transporter [34].
Statistics. The Unpaired Student test was used for statistical comparison of the transport rates in each direction, the transport rate in the presence and absence of the P-gp inhibitor for a particular direction, the transport efflux ratios in relation to pH conditions and the transport efflux ratio difference in relation to P-gp inhibitor addition (Xlstat version 5.0). A P value of <0.05 was considered significant.  Transepithelial transport of (+)-TMD, (-)-TMD and M1. Transepithelial transport of (+)-TMD and ())-TMD. Under the pH 6.8/7.4 condition, a significant efflux ratio (TR) of the two enantiomers was observed, with a 1.9 to 2.7-fold higher transepithelial transport from the basolateral to apical side (B-A) than from the apical to basolateral side (A-B). A-B transport was not affected by cyclosporine A, however, although not statistically significant, a trend towards a decrease of the influx (A-B) (up to 13%) was observed. A statistically significant increase of the efflux (B-A) (up to 18%, at all concentrations except 100 lM), and thus of the efflux ratio, was observed in the presence of cyclosporine A (P < 0.05) (table 1, fig. 3).

Integrity of
At physiological pH, the differential transport was nearly abolished (0.9 < TR < 1.2). As for the pH gradient condition, the A-B transport was not significantly affected by cyclosporine A. However, an overall statistically significant increase of the efflux (B-A) (up to 10%), with a slight increase of the efflux ratio, was observed in the presence of cyclosporine A, although this was less pronounced than under the pH gradient condition (P < 0.05) (table 1, fig. 3).
Transepithelial transport of M1. Under the pH 6.8/7.4 condition, a significant efflux ratio was observed with a 1.8 to 2.3-fold higher transepithelial transport from the basolateral to apical side (B-A) than from the apical to basolateral side (A-B). The A-B transport was not affected by cyclosporine A, except for a statistically significant decrease observed at 1 lM. A statistically significant increase of the efflux (B-A) (up to 26%) at all concentrations, and thus of the efflux ratio, was observed in the presence of cyclosporine A (P < 0.05) (table 1, fig. 4).
At physiological pH, the differential transport was nearly abolished (0.9 < TR < 1.1). The A-B transport was not affected by cyclosporine A. An overall statistically significant increase of the efflux (B-A) (up to 14%), although less pronounced than at the pH gradient condition, was observed (P < 0.05) (table 1, fig. 4).
At physiological pH, the B-A transport of M1 was not modulated by GF120918 at the lowest concentration (1 lM), however, a statistically significant decrease of the influx (A-B) (20%) was observed (P < 0.05) (table 2).

Discussion
Our study demonstrates that the two enantiomers of TMD as well as its active opioid metabolite M1 are not P-gp substrates. Indeed, a significant secretory transport of the three Table 1. Apparent permeability coefficients, P app (cm/sec.), of (+)-TMD, ())-TMD and M1 in the presence and absence of cyclosporine A, at pH 6. 8 compounds was observed in the presence of a pH gradient (TR: 1.8-2.7; P < 0.05), suggesting the involvement of an efflux transporter. However, the observed polarised transport was abolished at physiological pH. Furthermore, no effect of the first generation and non-specific P-gp inhibitor cyclosporine A and the more potent and specific inhibitor third generation GF120918, in the way of an increase of the absorptive flux (P app (A-B)) and a decrease of the secretory flux (P app (B-A)), at both pH conditions was observed. Our P-gp positive control rhodamine 123 showed a polarised transport across the Caco-2 monolayers in the way of a net active secretion as well as a 80% decrease of its secretory transport and 38% increase of the absorptive one in the presence of the P-gp inhibitor cyclosporine A. Consequently, TMD enantiomers and M1 are not P-gp substrates, at respectively 1-100 and 4-400 times the maximal plasma concentrations (Cmax) that produce a clear-cut analgesic effect after a single oral dose of 100 mg TMD racemate [25].  Table 2. Apparent permeability coefficients, P app (cm/sec.), of (+)-TMD, ())-TMD and M1 in the presence and absence of GF120918 at pH 7.4/7.4. The efflux ratio shown here is the net basolateral to apical flux direction (TR ¼ P app (B-A)/P app (A-B)). *P < 0.05. Significance of the inhibitor effect for a given transport direction.

Drug concentration € P-gp inhibitor
Values are the mean € SD of three experiments. These experimental results are in accordance with the results of pharmacogenetic clinical studies evaluating the role of P-gp in TMD kinetics. In fact, although a role of the P-gp polymorphism C3435T in the pharmacokinetics of TMD has been suggested in a small clinical trial involving patients with selected CYP2D6 and MDR1 polymorphisms, no statistically significant difference in kinetics was found between the 3435CC, CT and TT groups, independently of CYP2D6 polymorphisms, nor was influence of the G2677T/A polymorphism observed. A significant difference between the pooled CC + CT and TT patients was found only in the CYP2D6 poor metabolizers, and concerned the Cmax parameter only. M1 pharmacokinetics was found to be MDR1 polymorphism-independent [36].
The Caco-2 cell model is one of the most commonly used high throughput and cost-effective in vitro models for evaluating permeability and P-gp interaction potential of drug candidates. This model is recommended by the Food and Drug Administration (FDA) for the identification of substrates, inhibitors and inducers of P-gp through bidirectional transport studies [37,38]. The calibration of the model with appropriate reference probes and a thorough understanding of the rationale underlying the caveats associated with this model allow the results to be put in the proper perspective [24,39]. The Caco-2 cell monolayers are moreover a good model for pH-gradient transport studies for the following reasons: (i) the monolayers remain intact at apical pH values that cover the entire physiological range of the intestinal tract; (ii) they lack a protective mucus layer, which means that the pH at the cell surface rapidly equilibrates and becomes equal to the bulk pH of the applied buffer solution [40].
In our study, (+)TMD, ())-TMD and M1 permeability in the apical-to-basolateral (A-B) direction (absorptive flux) decreased approximately by 50% when apical pH was reduced from pH 7.4 to 6.8, resulting in a significant efflux ratio of the drugs (TR: P app (B-A)/P app (A-B): 1.8-2.7; P < 0.05). TMD and M1 are weak bases. This pH dependency of TMD and M1 absorptive flux is not explained by the passive diffusion of the unionised form according to the pH-partition theory. Indeed, the pKa of TMD is 9.4 and pH values in our study are among 2 and 2.4 units below its pKa. Hence, 100% of TMD is in the ionised form at both pH 6.8 and 7.4. M1 pKa is unknown but it is expected to be close to that of TMD. Therefore, the polarised transport of (+)-TMD, ())-TMD and M1 observed under a pH gradient is probably related to the contribution of specific efflux transport system(s). The overall effect of the P-gp inhibitors used in our study on the transepithelial transport of the two enantiomers of TMD and M1 was a statistically significant increase of the secretory flux (up to 18% and 26%, respectively), with an increase of the efflux ratio (TR), at both pH conditions. A higher increase of the secretory flux of TMD enantiomers was observed with GF120918 at physiological pH as compared with cyclosporine A (17% versus 10%). Moreover, a significant decrease of the absorptive flux of M1 (approximately 20%) was observed in the presence of GF120918 while no effect was observed with cyclosporine A. This apparent paradoxical effect of P-gp inhibitors suggests the involvement of uptake drug transporter(s) in the transmembrane transport of TMD enantiomers and M1.
The gastrointestinal tract and the kidney provide a unique environment of cellular transport due to the presence of proton-based efflux pumps such as the Na + -H + and the organic cation/H + exchangers [26,27]. The Na + -H + antiporter has been characterized in Caco-2 cells [41] and studies using the Caco-2 cell model showed an active secretory transport of weak bases such as celiprolol, loperamide, zolmitriptan, diphenydramine and amphetamine derivatives by protonbased efflux pumps [33,[42][43][44][45]. H + antiport systems might therefore be involved in the limitation of the gastrointestinal absorption of TMD as well as renal excretion of TMD and M1. The use of specific inhibitors such as amiloride, or cimetidine could help confirming this hypothesis [33,[42][43][44][45]. Apart from efflux drug transporters, Caco-2 cells express several uptake transporters. The most relevant are the organic anion transporting polypeptide OATP2B1/SLCO2B1, the peptide transporter PEPT1/SLC15A1, the organic cation/carnitine OCNT2/SLC22A5 as well as the monocarboxylic acid transporters MCT1/SLC16A1 and MCT5/ SLC16A5 and the human peptide transporter HPT1/ CDH17 [46,47]. PEPT1 is inhibited by cyclosporine A [48]. Moreover, it has been shown to transport endogenous opioid peptides and consequently might be relevant in the oral bioavailability of synthetic opioids [49]. TMD and M1 cross easily the blood-brain barrier [25]. They appear to cross the cellular membrane efficiently and to accumulate in tissues, and these pharmacokinetic properties are in concordance with our study results. Indeed, a comparative study of Caco-2 monolayer permeability and human drug absorption suggests that compounds with P app > 10 · 10 )6 cm/sec. can be classified as well-absorbed compounds (70-100%) [50]. Since TMD (pKa: 9.4) and M1 are almost 100% protonated at both pH conditions tested, a passive diffusion mechanism may not fully explain TMD and M1 high tissue distribution. The involvement of uptake transporters seems therefore relevant in this case. More specific in vitro models are required to support our observations.
In conclusion, (+)-TMD, ())-TMD and the major analgesic metabolite O-desmethyl-TMD are not P-gp substrates. However, proton-based efflux pumps may be involved in limiting the gastrointestinal absorption of TMD enantiomers and as well as enhancing TMD enantiomers and M1 renal excretion. Moreover, our investigations suggest the possible involvement of uptake carriers in their transepithelial transport across biological membranes. The hypothesis advanced, which might be of importance in the prediction of clinically significant drug interactions, needs to be investigated.