cGMP and glutathione-conjugate transport in human erythrocytes

The roles of the multidrug resistance-associated proteins, MRP1, MRP4 and MRP5


S. B. Hladky, Department of Pharmacology, University of Cambridge, Cambridge, CB2 1PD, UK. Fax: + 44 1223 334040, Tel.: + 44 1223 334019, E-mail:


The nature of cGMP transport in human erythrocytes, its relationship to glutathione conjugate transport, and possible mediation by multidrug resistance-associated proteins (MRPs) have been investigated. MRP1, MRP4 and MRP5 are detected in immunoblotting studies with erythrocytes. MRP1 and MRP5 are also detected in multidrug resistant COR-L23/R and MOR/R cells but at greatly reduced levels in the parent, drug sensitive COR-L23/P cells. MRP4 is detected in MOR/R but not COR-L23/R cells. Uptake of cGMP into inside-out membrane vesicles prepared by a spontaneous, one-step vesiculation process is shown to be by a low affinity system that accounts for more than 80% of the transport at all concentrations above 3 µm. This transport is reduced by MRP inhibitors and substrates including MK-571, methotrexate, estradiol 17-β-d-glucuronide, and S(2,4-dinitrophenyl)glutathione (DNP-SG) and also by glibenclamide and frusemide but not by the monoclonal Ig QCRL-3 that inhibits high-affinity transport of DNP-SG by MRP1. It is concluded that the cGMP exporter is distinct from MRP1 and has properties similar to those reported for MRP4. Furthermore the evidence suggests that the protein responsible for cGMP transport is the same as that mediating low-affinity DNP-SG transport in human erythrocytes.


adenosine 5′-O-(3-thiotriphosphate)




estradiol 17-β-d-glucuronide


reduced glutathione


horseradish peroxidase




(3-([[3-(2-[7-chloro-2-quinolinyl]ethenyl)phenyl}-{(3-(dimethylamino-3-oxopropyl)-thio}-methyl]thio) propanoic acid


multidrug resistance-associated protein

PNGase F

peptide N-glycosidase F


31–8220 bisindolylmaleimide


4-acetamido-4′-isothiocyano-2,2′-disulfonic stilbene IX methanesulfonate

Active transport of the cyclic nucleotide cGMP across human erythrocyte membranes can be demonstrated using intact cells [1] or inside-out membrane vesicle preparations [2–4]. In the studies using inside-out membrane vesicles, the active uptake of cGMP was found to be saturable with two components, one of high-affinity (Km 2–5 µm) [2,5] and another of low-affinity (Km 170 ± 50 µm) [5]. Two components have also been described for the transport of another organic anion, a glutathione conjugate S(2,4-dinitrophenyl)glutathione (DNP-SG), in human erythrocytes [6,7].

Two members of the multidrug resistance-associated protein (MRP) transporter family, MRP1 and MRP5 have been detected previously in human erythrocyte membranes [8,9], and transport by MRP1 has been conclusively shown to account for the high-affinity component of DNP-SG transport [10–12]. MRP4 and MRP5 have been shown to transport the cyclic nucleotides, cAMP and cGMP [9,13,14] and it has been suggested that MRP5 mediates the high-affinity component of the cGMP transport [15]. However, the same group has questioned this identification [16] and recently it has been shown that when expressed in HEK293 cells, MRP4 and MRP5 mediate low-affinity transport of cyclic nucleotides [17].

The aim of the present study was to investigate the nature of cGMP transport in human erythrocytes, its relationship to glutathione conjugate transport, particularly to the low-affinity DNP-SG component, and its possible mediation by MRP4 and/or MRP5. The present work provides evidence from immunoblotting studies that both MRP5 [9] and MRP4 are expressed in human erythrocytes. Using inside-out membrane vesicles prepared by a spontaneous, one-step vesiculation process, we identify a low affinity component for the cGMP transport which accounts for more than 80% of the transport at all concentrations above 3 µm. This transport is reduced by a range of inhibitors and substrates for MRPs including MK-571, methotrexate, E217βG, and DNP-SG and also by glibenclamide and frusemide. We show that this cGMP exporter is distinct from MRP1 and has characteristics similar to those reported for MRP4. The evidence suggests that the protein responsible for cGMP transport is the same as that mediating low-affinity DNP-SG transport in human erythrocytes.

Experimental procedures


[8-3H]cGMP (specific activity 13.9 Ci·mmol−1) was obtained from Amersham Biosciences, [glycine-2-3H]GSH (specific activity 40 Ci·mmol−1) and [3H]glibenclamide (specific activity 44.7 Ci·mmol−1) were obtained from New England Nuclear, respectively.

M5I-1 mAb against MRP5 was a kind gift of R. J. Scheper (Free University, Amsterdam, the Netherlands); anti-MRP4 mAb was a kind gift of G. D. Kruh (Fox Chase Cancer Centre, Philadelphia, PA, USA); QCRL-3 mAb was purchased from Signet Laboratories, USA. M5I-1 and anti-MRP4 mAbs have been previously described [18,19].

4-Acetamido-4′-isothiocyano-2, 2′-disulfonic stilbene (SITS), adenosine 3′,5′-cyclic monophosphate (cAMP), adenosine 5′-O-(3-thiotriphosphate) (ATP-γ-S), adenosine triphosphate (ATP), 4-aminopyridine, aprotinin, 1-chloro-2,4-dinitrobenzene, clotrimazole, creatine kinase, creatine phosphokinase, daunorubicin, dideoxyforskolin, isobutylmethylxanthine (IBMX), doxorubicin, estradiol 17-β-D-glucuronide, forskolin, glibenclamide, glutathione (reduced form, GSH), glutathione S-transferase, guanosine 3′,5′-cyclic monophosphate (cGMP), imidazole, indomethacin, leupeptin, lithocholic acid 3-sulphate, methotrexate, pepstatin A, probenecid, taurocholic acid, tetraethylammonium chloride, Triton X-100, Tween 20, verapamil and vincristine were all obtained from Sigma Chemicals. Calcein was purchased from Molecular Probes. Staurosporine and bisindolylmaleimide IX methanesulfonate (Ro 31–8220) were obtained from Calbiochem. (3-([[3-(2-[7-chloro-2-quinolinyl]ethenyl)phenyl}-{(3-(dimethylamino-3-oxopropyl)-thio}-methyl]thio) propanoic acid, MK-571, was a generous gift of M. Turner (Merck-Frosst Center for Therapeutic Research, Quebec, Canada). Peptide N-glycosidase F (PNGase F) was purchased from Promega.

Drugs were prepared in 10 mm Tris/HCl (pH 7.4) for GSH, ATP, ATP-γ-S, cGMP, taurocholic acid, imidazole, vincristine and MK-571 or in 66% dimethyl sulfoxide/34% water for glibenclamide, SITS, methotrexate, verapamil, indomethacin, E217βG and clotrimazole. The final concentration of the dimethyl sulfoxide did not exceed 0.5% in each experiment. GSH stock solutions (adjusted to pH 7.4) were freshly prepared on the day of each experiment.

[3H]DNP-SG was synthesized enzymatically as previously described [12,20]. The purity of the 3H-labelled DNP-SG was determined by thin-layer chromatography on silica gel plates [(0.25 × 40 × 80) mm, Alugram®SIL G/UV254, Macherey-Nagel, Germany] using n-propanol:water (7 : 3, v/v) as solvent [21].

Cell lines

COR-L23/R and MOR/R are MRP1-overexpressing, multidrug-resistant, human large-cell lung tumour lines produced by doxorubicin selection [22,23]. All cells were cultured on plastic in growth medium containing RPMI-1640 medium supplemented with 10% (v/v) foetal bovine serum, glutamine (2 mm), penicillin (100 IU·mL−1) and streptomycin (100 µg·mL−1) (complete RPMI-1640) in a 5% CO2 humidified incubator at 37 °C. The L23/R and MOR/R sublines were maintained in the presence of 0.2 µg·mL−1 and 0.4 µg·mL−1 doxorubicin, respectively. The cells were kept in drug-free medium for at least 48 h before use in experiments. Cells were passaged when they became confluent. RNA protection assay of the doxorubicin-resistant COR-L23/R and MOR/R cells [24] shows that these cells: do not express P-glycoprotein; over-express MRP1 when compared to the doxorubicin-sensitive controls; and express MRP4 at a low level. MRP5 is expressed at very low level in COR-L23 cells but at high level in MOR cells.

Preparation of inside-out human lung tumour cell membrane vesicles

Membrane vesicles from human lung tumour cells were prepared according to a method described previously [25] in the presence of protease inhibitors (5 µg·mL−1 leupeptin, 2 µg·mL−1 aprotinin, 80 ng·mL−1 pepstatin A). Briefly, cells were lysed in ice-cold hypotonic buffer (1 mm Tris/HCl, pH 7.4) for 30 min at 4 °C. Following centrifugation at 100 000 g for 30 min at 4 °C, the resulting pellet was homogenized vigorously with a Teflon hand homogenizer in buffer containing 10 mm Tris/HCl, 250 mm sucrose, and protease inhibitors, layered over 38% (w/v) sucrose in 10 mm Tris/HCl and centrifuged at 100 000 g for 30 min at 4 °C. The membranous material in the layer at the interface with the sucrose was collected, washed and centrifuged at 100 000 g for 30 min at 4 °C. The resulting pellet was re-suspended in transport buffer (10 mm Tris/HCl, 250 mm sucrose, pH 7.4), and stored in aliquots at −80 °C.

Preparation of inside-out human erythrocyte membrane vesicles

Fresh venous blood was drawn from donors into tubes containing EDTA or heparin and processed immediately. There were five donors, each of whom gave informed consent, two of northern European origin, one southern European, one Chinese, and one Sri-Lankan. Membrane vesicles were prepared by a spontaneous, one-step vesiculation process as previously described [26–28] with minor modifications. Briefly, red blood cells were washed three times with 5 vols. of isotonic medium (80 mm KCl; 70 mm NaCl; 0.2 mm MgCl2; 10 mm Hepes; 0.1 mm EGTA, pH 7.5). Higher concentrations of EGTA (0.5–3 mm) and high pH (8.5) interfere with the vesiculation process [26]. The buffy coat and topmost cell layer were removed after each wash. The packed cells were then lysed by addition to 90 vols. of ice-cold solution L (2 mm Hepes and 0.1 mm EGTA, pH 7.5) and subsequently centrifuged at 40 000 g for 20 min at 4 °C. The supernatant was removed and the pelleted ghosts were re-suspended in ice-cold solution L. This step was repeated twice. After the last wash, the pellets were re-suspended by addition of half the original packed cell volume of cold solution L and incubated at 37 °C for 30 min resulting in spontaneous formation of spectrin-actin-free vesicles [28]. After incubation, the suspension was washed with solution L and the resulting pellet resuspended in 10 mm Tris/HCl (pH 7.4). The protein concentrations of the vesicle samples were determined using the BCA (bicinchoninic acid) protein assay (Pierce). Membrane vesicles were frozen and stored at −80 °C until use.

Measurement of membrane vesicle sidedness

The proportion of inside-out vesicles in the membrane preparations was assessed by determining the accessibility of the ectoenzyme acetylcholinesterase, and the endoenzyme glyceraldehyde-3-phosphate dehydrogenase to their substrates. Triton X-100 was used to disrupt the permeability barrier and expose latent markers. The determination of enzyme activities was performed colorimetrically [29,30]. The assays were modified by exchange of all phosphate solutions with 10 mm Tris/HCl (pH 7.4) for the assays involving membrane vesicles prepared from human erythrocytes. The pH optimum of glyceraldehyde 3-phosphate dehydrogenase activity is about 8.4 [31], but the activity in the present study was determined at pH 7.4 to obtain comparable conditions in the assays of sidedness and transport. Generally 30–37% of the vesicles were inside-out.

Vesicle uptake studies

ATP-dependent uptake of radiolabeled cGMP or DNP-SG into erythrocyte membrane vesicles was measured by a rapid filtration technique [20]. Thawed membrane vesicles were diluted in buffer and 50 µg protein added to a buffer system (55 µL final volume) containing 1 mm ATP, 10 mm MgCl2, 10 mm creatine phosphate, 100 µg·mL−1 creatine kinase, 10 mm Tris/HCl (pH 7.4) and 3.3 µm[3H]cGMP or 3 µm[3H]DNP-SG or 254 µm[3H]DNP-SG. Aliquots (20 µL) were taken from the mixture after 15 min in the case of cGMP uptake, after 30 min with 3 µm[3H]DNP-SG uptake, and after 45 min with 254 µm[3H]DNP-SG uptake, diluted in 1 mL of ice cold stop solution (10 mm Tris/HCl, pH 7.4) and subsequently filtered through nitrocellulose filters (Whatman 0.2 µm pore size, presoaked overnight in 3% (w/v) bovine serum albumin. The filters were rinsed with 3 mL of ice-cold stop solution and the tracer retained on the filter was determined by liquid scintillation counting.

All transport data are presented as the difference between the values measured in the presence of ATP and those measured in the presence of the nonhydrolysable ATP analogue, ATP-γ-S. The ATP regenerating system (10 mm creatine phosphate, 100 µg·ml−1 creatine kinase) was present in both cases. Uptake of the substrate was expressed relative to the protein concentration of the membrane vesicles, and all data were corrected for the amount of radiolabelled substrate bound to the filter in the absence of vesicle protein. The substrate and inhibitor concentrations are given in the respective figure legends. Tested compounds were added from a stock solution in the appropriate solvent [10 mm Tris/HCl (pH, 7.4), dimethyl sulphoxide or ethanol, with the latter two solvents at a final concentration < 0.5% v/v], identical concentrations of the vehicle being used in control samples.

Curve fitting and statistics

Data are reported as mean ± s.e.m. Estimates of maximum uptake rates and apparent dissociation constants were obtained by least squares fits to the data using either the solver in Microsoft excel® or kaleidagraph® (Synergy Software, Reading, PA, USA). Measurements of uptake vs. substrate concentration were fitted assuming that transport occurs as the sum of two processes each described by a Hill equation [32]:


where c is the concentration, Umax the maximum uptake, Kd the apparent dissociation constant and n is the Hill coefficient, here expected to lie between 1 and 2. The data were fitted to minimize the sum of squared proportional deviations,


fits to inhibition curves were to equations of the form:


where I is the concentration of the inhibitor, IC50 is the inhibitor concentration producing 50% inhibition of the inhibitable component, U0 is the uptake in the absence of the inhibitor, Unoninh is the uptake which cannot be inhibited and ni is the Hill coefficient for the inhibitor. For simple competition, ni = 1. Unless stated otherwise, KaleidaGraph® was used to fit the data using the variances at each concentration, σi, as the weights:


fits with different numbers of fitting parameters were compared using an F-test on the ratio of the variance associated with the reduction in degrees of freedom to the variance of the fit with the smaller number of degrees of freedom [33],


where d.f. is the number of degrees of freedom (= number of data points − number of fitting parameters). The improvement in fit is labelled ‘significant’ if the probability from the F-test is less than 0.05. Fits with the same number of degrees of freedom were compared with each other using the likelihood ratio:


where ΔSSE is the difference in SSE for the two fits and the noise variance was estimated as σ2 = MinSSE/d.f. with MinSSE equal to the value of SSE for the best fit.

SDS/PAGE and Western blotting

Membrane vesicle or crude lysate proteins (5–40 µg) were separated through 7.5% (w/v) polyacrylamide and subsequently transferred onto Hybond ECL nitrocellulose membranes (Amersham Biosciences). Each membrane was then incubated in blocking buffer [5% (w/v) milk powder in 0.1% NaCl/Pi/Tris/Tween (25 mm Tris/HCl, pH 7.4, 150 mm NaCl, 0.1% Tween 20)] overnight at 4 °C prior to the addition of the primary Ig (M5I-1, 1 : 40 dilution; anti-MRP4, 1 : 300 dilution). The positions of the MRP proteins on the membranes were visualized using the enhanced chemiluminescence horseradish peroxidase (HRP) detection system (Amersham Biosciences). The secondary antibodies used were the HRP-conjugated rabbit anti-(rat IgG) Ig (1 : 2000 dilution for M5I-1) or HRP-conjugated rabbit anti-(mouse IgG) Ig (1 : 2000 dilution for anti-MRP4).

Membrane proteins from the human erythrocytes were N-deglycosylated by treatment with PNGaseF as follows: Briefly, membrane vesicles from human erythrocytes (40 µg) were first denatured at 100 °C for 10 min in the presence of 0.5% SDS and 1% β-mercaptoethanol, followed by incubation at 37 °C for 1 h in the presence of 50 mm sodium phosphate (pH 7.5), 1% of the nonionic detergent Nonidet P-40 (NP-40), and 2000 units of PNGase F (New England Biolabs). PNGase F is an amidase which cleaves between the innermost N-acetylglucosamine (GlcNAc) and asparagine residues of high mannose, hybrid and complex oligosaccharides from N-linked glycoproteins [34]. PNGase F hydrolyzes nearly all types of N-glycan chains from glycopeptides/proteins.


ATP-dependent uptake of cGMP into human erythrocyte membrane vesicles

The rate of ATP-dependent uptake of 3.3 µm[3H]cGMP at 37 °C into inside-out erythrocyte membrane vesicles was approximately constant for more than 30 min at about 10 pmol·mg−1·min−1(Fig. 1A). Uptake of [3H]cGMP in the absence of ATP but in the presence of the nonhydrolysable ATP analogue, ATP-γ-S, was less than 5% of the uptake in the presence of ATP. The amount of [3H]cGMP taken up by these vesicles was approximately twofold lower (44 ± 3%, mean ± SEM, n = 4) when measured with NaCl/Pi (140 mm NaCl; 3 mm KCl; 10 mm Na2HPO4; and 1.8 mm KH2PO4, pH 7.4) than with the usual low osmolality transport buffer. Such a difference is to be expected as the higher osmolality should decrease the volume of the intravesicular space.

Figure 1.

ATP-dependent uptake of cGMP into inside-out membrane vesicles prepared from human erythrocytes. (Top) Uptake of 3.3 µm[3H]cGMP was measured in the presence of 1 mm ATP or the nonhydrolysable analogue ATP-γS. (Middle) The variation of uptake rate with concentration of cGMP. (Bottom) Haynes–Wolfe plot of the data for low concentrations. In this type of plot a single, simple saturable component of uptake (Hill coefficient = 1) would yield a straight line. The fitted constants for the curves in (Middle) and (Bottom) are given in Table 1. The dotted curves are drawn for a single, simple saturable component of transport (Hill coefficient = 1); the dashed curves for a single component described by a Hill equation with Hill coefficient = 1.09, and the solid curve for two components, each described by a Hill equation with Hill coefficients of 2 for the high affinity, low capacity component and 1.3 for the low affinity, high capacity component. Data for the four highest concentrations were determined in three independent experiments from one preparation of vesicles. All other data points represent at least three experiments and two vesicle preparations.

ATP-dependent uptake of[3H]cGMP was determined at cGMP concentrations in the range 0.5–300 µm (Fig. 1B,C). To test whether the uptake occurs via a single component, the data were fitted assuming two components each described by a Hill equation (see Eqn 1) as shown in Fig. 1 and Table 1. The data imply that there is a large (Umax > 300 pmol·mg−1·min−1), weakly cooperative (n ≈ 1.1–1.4) low affinity component with dissociation constant, Kd2, in the range 50–85 µm and suggest that there may also be a second, much smaller high affinity component with Kd1, in the range 0.5–2.5 µm. However, this latter component, which may correspond to the uptake observed previously [2,5], contributes less than 20% of the uptake even for a low concentration, 3.3 µm, of cGMP.

Table 1. Fitting parameters for the uptake of [3H]cGMP into one-step, inside out erythrocyte membrane vesicles shown in Fig. 1. The maximum uptake rates, Umax1 and Umax2, the dissociation constants, Kd1 and Kd2, and the Hill coefficients, n1 and n2 are defined as indicated in Eqn (1). The data were obtained using two different vesicle preparations. As no differences were observed between the two, the data were combined without scaling. The residual value of the sum of squared proportional deviations, SSE (see Eqn 2), is shown for each fit. For each column except the first the variance ratio (see Eqn 5) has been calculated relative to the column immediately to the left. The fit obtained with the constraints n1 = 1 and n2 = 1 is not shown as the fitted value of Umax1 was 0.000. These data imply (F-test on the variance ratio, P = 0.0004) that the low affinity component shows cooperativity, n2 > 1, and are consistent with the presence of a high affinity component (F-test, P = 0.002), but do not specify its characteristics.
Fitted constantConstraints
Umax1 = 0
n2 = 1
Umax1 = 0n1 = 1n1 < = 2
Umax1/pmol·mg−1· min−1007.12.7
Umax2/pmol·mg−1· min−1551389293304
P 0.00040.0020.187

ATP-dependent uptake of DNP-SG into inside-out, human erythrocyte vesicles

When inside-out erythrocyte membrane vesicles were incubated at 37 °C with 3 µm[3H]DNP-SG, the uptake in the presence of 1 mm ATP increased linearly in time for at least 60 min while uptake when ATP was replaced by ATP-γ-S was almost negligible [12]. ATP-dependent uptake of [3H]DNP-SG in human erythrocyte vesicles was determined over a broad concentration range (0.44–1000 µm) (Fig. 2). As for cGMP, the data for DNP-SG were analyzed using a two component Hill equation. The results of fits with several different restrictive assumptions are shown in Table 2. The quantitative fitting (variance ratio test) confirms, as is obvious by eye, that the transport occurs via at least two components. To explore the range of acceptable values of the Hill coefficient for the low-affinity component, n2, least squares fits were obtained for specified values of n2 (Table 2). Acceptable fits (likelihood ratio = 0.05 compared to the best fit) were obtained for values of n2 between 1 and 1.48. Over the range from 1 to 1.4, the low-affinity dissociation constant decreases from 82 to 65 µm while that for the high-affinity component varies from 0.5 to 2 µm.

Figure 2.

Rate of ATP-dependent uptake of [3H]DNP-SG into inside-out erythrocyte membrane vesicles. The dotted curve is drawn for a single simple saturable component of uptake, the solid curve for two simple saturable components. The dashed and dash-dot curves are drawn for two components each obeying a Hill equation with the constraints that n2 = 1.4 or n2 = 2, respectively. The data are plotted directly (Top) and as a Haynes–Wolfe plot (Bottom). The fitted constants are described in Table 2. These data are not consistent with a single-component of uptake, but cannot unambiguously determine the properties of two components when provision is made for the possibility that more than one substrate molecule may interact with the transporter at a time.

Table 2. Fitting parameters for uptake of [3H]DNP-SG into one-step, inside out erythrocyte membrane vesicles. The maximum uptake rates, Umax1 and Umax2, the dissociation constants, Kd1 and Kd2, and the Hill coefficients, n1 and n2 are as defined in Eqn (1). The data were collected in three series using different vesicle preparations. To allow simultaneous fitting of all three sets of data, all data in the first set are scaled by multiplication by AF and all data in the second by CF. For fits of the two component Hill equation, there are 22 remaining degrees of freedom. The one and two component fits are compared with each other using an F-test on the variance ratio (Eqn 5). The two component fit is significantly better. The various constrained two component fits are compared with the fit for n1 > = 1, n2 > = 1 using the likelihood ratio (Eqn 6). These data are consistent with any value of n2 between 1 and 1.48 (LR = 0.05).
Fitting constantConstraints
Umax1 = 0;
n2 > = 1
n1 > = 1;
n2 > = 1
n1 > = 1;
n2 > = 1.2
n1 > = 1;
n2 > = 1.4
n1 > = 1;
n2 > = 2
VR 25.9   
P 2 × 10−7   
LR 10.340.0880.002

In agreement with previous work [10,11] the high affinity component of DNP-SG transport in these vesicles is most likely mediated by MRP1 [12]. Strong evidence in support of this comes from the observation that the uptake rate of 3 µm DNP-SG is reduced by at least 80% by QCRL-3 [12], an MRP1-specific conformational-dependent monoclonal Ig [35].

To investigate the low affinity component the DNP-SG concentration was increased to 254 µm. The inhibition by QCRL-3 was then only 40 ± 5% (n = 6). This result and the complete inhibition observed at low DNP-SG concentrations suggests that there is some low affinity transport that is not inhibited by QCRL-3 and is thus not mediated by MRP1. On this basis, the low affinity process should account for no more than 20% of the uptake observed at 3 µm. This requirement is consistent with the uptake measurements provided n2 = 1.4. All the data are consistent with high affinity transport via MRP1 (Kd = 2 µm, Umax = 20 pmol·mg−1· min−1) and low affinity, weakly cooperative transport via a second transporter (Kd = 65 µm, Umax = 196 pmol· mg−1·min−1 and n2 = 1.4).

Interrelations between cGMP and DNP-SG uptake into human erythrocyte membrane vesicles

To explore the relations between cGMP and DNP-SG transport, the ability of each to inhibit transport of the other was investigated. ATP-dependent uptake of 3 µm DNP-SG was not affected by the presence of cGMP at concentrations up to 500 µm(Fig. 3A) suggesting that the high affinity DNP-SG transporter, MRP1, does not transport cGMP. This finding was further supported by the observation that the MRP1-specific Ig, QCRL-3, produced negligible inhibition of the ATP-dependent uptake of cGMP (uptake rate for 3.3 µm cGMP with 10 µg·mL−1 QCRL-3 was 98 ± 5% of control, n = 3).

Figure 3.

Effect of cGMP on the DNP-SG transport in human erythrocytes and vice versa. (Top) ATP-dependent uptake of 3 µm[3H]DNP-SG (30 min at 37 °C) was not affected by cGMP at concentrations up to 500 µm. (Middle) ATP-dependent uptake of 254 µm[3H]DNP-SG (30 min at 37 °C) was partially inhibited by cGMP. The fitted curve corresponds to two components: a noninhibitable component of 2943 ± 115 pmol·mg−1 and a component of 4110 ± 144 pmol·mg−1 inhibited by cGMP with an IC50 of 133 ± 18 µm (mean ± SEM). (Bottom) ATP-dependent uptake of 3.3 µm[3H]cGMP (15 min at 37 °C) was inhibited by DNP-SG as a single component described by a Hill equation with U0 = 123 ± 1 pmol·mg−1, IC50 = 82 ± 2 pmol·mg−1, and a Hill coefficient of 1.25 ± 0.02 (χ2 = 208). The fit of the Hill equation was significantly better (variance ratio test, 17 data points, three parameters in the Hill equation, P = 0.003) than the fit to a simple competition curve (U0 = 135 ± 1 pmol·mg−1, IC50 = 47 ± 1 pmol·mg−1, χ2 = 395).

On the other hand, ATP-dependent uptake of DNP-SG at high concentrations was inhibited by cGMP (Fig. 3B). The fitted curve for uptake at 254 µm DNP-SG has two components: a noninhibitable uptake (30 min) of 2943 ± 115 pmol·mg−1 and a component of 4110 ± 144 pmol·mg−1 inhibitable by cGMP with an IC50 of 133 ± 18 µm. These components are plausibly attributed to the high and low-affinity components of DNP-SG transport, respectively.

In order to investigate whether the cGMP transport is also affected by increasing concentrations of DNP-SG, uptake of 3.3 µm[3H]cGMP in inside-out membrane vesicles was measured in the presence of DNP-SG in the range of 0.5–800 µm (Fig. 3C). DNP-SG was able to inhibit all of the cGMP transport detectable at this concentration suggesting that it occurs via a single, DNP-SG inhibitable component. The solid curve is a plot of a Hill equation (see Eqn 3) with IC50 82 ± 2 µm and a Hill coefficient of 1.25 ± 0.02 µm.

Effect of MRP inhibitors, substrates, and modulators on cGMP uptake into human erythrocyte membrane vesicles: A number of compounds that are known to interact with one or more MRPs were tested for their ability to inhibit cGMP transport (see Fig. 4 and section below entitled ‘cGMP transport is inhibited by anion transport inhibitors, PKC inhibitors and IBMX’). MK-571, a leukotriene D4 (LTD4) receptor antagonist, which has been shown to inhibit transport by MRP1 [36,37], MRP2 and MRP3 [38] and MRP4- [39] but not MRP5-mediated cGMP transport [9] completely inhibited the [3H]cGMP uptake in vesicles with an IC50 value of 0.38 ± 0.01 µm (Fig. 4A) suggesting that this transport is not mediated by MRP5. cAMP (Fig. 4B) which is transported by MRP4 and MRP5 [9,13], inhibited with IC50 = 296 ± 26 µm. The ratio of this constant to the apparent dissociation constant for cGMP, 50–80 µm (Table 1), is similar to the ratio, 6, for MRP4 [13], but is much smaller than the ratio, 380, for MRP5 [9]. Glibenclamide (Fig. 4C), an agent known to bind to various ABC proteins [40,41] including the sulphonylurea receptor [42,43], was effective in inhibiting the cGMP transport in human erythrocyte vesicles at micromolar concentrations. Substantial inhibition was produced by methotrexate and E217βG, established MRP4 substrates, by indomethacin which is known to inhibit transport by MRP1 and MRP2 [44], and by clotrimazole (Fig. 4D) an imidazole-derived antifungal agent which inhibits MRP1 mediated transport [12]. Imidazole, the backbone molecule of clotrimazole had no effect. Taurocholic acid, an established substrate for MRP1, MRP2, and MRP3, inhibited but only at concentrations sufficiently high (> 200 µm, Table 3) that it may be acting in a ‘detergent like’ manner.

Figure 4.

Inhibition of uptake of 3.3 µm cGMP by(A) MK-571,(B) cAMP,(C) glibenclamide and(D) clotrimazole. (A) Inhibition by MK-571. The curve is the best fit assuming simple competition, U0 = 130 ± 2 pmol·mg−1, IC50 = 0.38 ± 0.01 µm, χ2 = 95. The Hill equation provides a closer fit with, U0 = 120 ± 2 pmol·mg−1, IC50 = 0.48 ± 0.02 µm, ni = 1.10 ± 0.02 and χ2 = 70, but the improvement is not significant (F = 2.7, P = 0.13). The best fit allowing for noninhibitable uptake assigns a negative value (−1.4 pmol·mg−1) to the noninhibitable component. (B) Inhibition by cAMP. The curve shows the best fit for simple competition with U0 = 108 ± 5 pmol·mg−1, IC50 = 296 ± 26 µm, χ2 = 11. Fits of the Hill equation and of simple competition plus a noninhibitable component of uptake are almost superimposed on that shown (and improvements in fit were not-significant with F and P-values for the variance ratio relative to simple-competition of 0.03 and 0.86, and 0.24 and 0.64, respectively). (C) Inibition by glibenclamide. The curve is the best fit for simple competition U0 = 80 ± 1 pmol·mg−1, IC50 = 2.8 ± 0.1 µm, χ2 = 250. Fits of the Hill equation and of simple competition plus a noninhibitable component of uptake did not produce significant improvements in the fit (F and P-values 3.26 & 0.12 and 5.0 & 0.07, respectively). The best fit for the noninhibitable component was 3.4 ± 0.3 pmol·mg−1. (D) Inhibition by clotrimazole. The curve is the best fit for simple competition U0 = 97 ± 1 pmol·mg−1, IC50 = 24 ± 1 µm, χ2 = 86. Fits of the Hill equation and of simple competition plus a noninhibitable component did not produce significant improvements in the fit (F and P-values of 4.4 & 0.1 and 4.7 & 0.1, respectively. The best fit value of the Hill coefficient was less than 1 (0.72).

Table 3. Effect of MRP substrates, inhibitors and modulators on the ATP-dependent uptake of [3H]cGMP by inside-out human erythrocyte membrane vesicles. Data represent mean ± SEM of n experiments. The control uptake of [3H]cGMP (addition of dimethylsulfoxide only) for indomethacin, methotrexate and E217βG was 77.3 ± 5.2 pmol·mg−1. The control uptake for the remaining drugs was 136.5 ± 2 pmol·mg−1.
CompoundConcentration (µm)[3H]cGMP uptake (% control)n
GSH500101.2 ± 0.83
1000108.3 ± 7.16
2000100.1 ± 1.93
400098.0 ± 1.23
Vincristine100106.3 ± 7.63
200100.6 ± 3.23
+ 1 mm GSH100116.4 ± 1.13
Calcein10073.9 ± 1.83
30038.5 ± 3.43
+ 1 mm GSH10074.4 ± 1.83
Indomethacin2095.9 ± 3.73
507.9 ± 5.63
Methotrexate27525.2 ± 2.83
3753.5 ± 1.13
17βE2G6512.6 ± 3.74
Taurocholic acid20080.1 ± 2.13
35043.1 ± 1.13

Reduced glutathione (GSH, pH 7.4), in the range of 0.5–4 mm, neither enhanced nor inhibited cGMP uptake (Table 3). This contrasts with the effect of 1–5 mm GSH to stimulate uptake of DNP-SG in human erythrocyte vesicles [12] but is consistent with the lack of effect of GSH on MRP4 and MRP5 mediated transport in transfected HEK293 cells [17].

The cationic vinca alkaloid, vincristine, and the organic anion, calcein, which are established MRP1 substrates, were also tested for their ability to inhibit cGMP uptake. Calcein inhibited cGMP uptake (about 25% inhibition at 100 µm and about 60% at 300 µm) but vincristine itself had no effect even at 200 µm (Table 3). It is known that several cationic MRP1 substrates, including vincristine, require GSH for their transport and that vincristine inhibits the high affinity DNP-SG transport in the presence but not in the absence of GSH [12,45,46]. Thus the effect of vincristine on the cGMP uptake was also tested in the presence of 1 mm GSH but no inhibition was observed. Given that MRP1-mediated transport of calcein in whole cells can be modulated by the level of GSH present [47], the effect of calcein on cGMP uptake into erythrocyte membrane vesicles was also tested in the presence of 1 mm GSH but no additional inhibition was observed (Table 3).

These results are all consistent with cGMP transport via MRP4 while the strong inhibition produced by MK-571 and the relative potency of cAMP appear to be incompatible with transport via MRP5.

CGMP transport is inhibited by anion transport inhibitors, PKC inhibitors and IBMX

Because substrates for MRPs are often organic anions, inhibitors that block ion transport were tested (Table 4). The anion transport inhibitors frusemide, niflumic acid, phloridzin, SITS and probenecid all reduced the rate of cGMP uptake (Table 4). By contrast the potassium channel blockers, 4-aminopyridine, tetraethylammonium chloride and CsCl, had no observable effect though BaCl2 did (Table 4). Verapamil, a calcium channel blocker, an inhibitor of P-glycoprotein, and a general though weak inhibitor of MRPs in vesicular drug uptake studies, reduced cGMP transport in the presence or absence of 1 mm GSH. Two protein kinase C inhibitors, staurosporine and Ro 31–8220 [48], were also tested for their ability to block cGMP transport in human erythrocytes. Staurosporine has recently been shown to bind directly to several ABC transporters [49] in addition to preventing phosphorylation of these transporters in intact cells [50]. Staurosporine at 10 µm completely inhibited the cGMP uptake while Ro 31–8220 at 10 µm showed only weak inhibition. Forskolin, an activator of adenylyl cyclase, inhibited while its inactive analogue, 1,9-dideoxyforskolin, had no effect at the same concentration. IBMX which is structurally related to cGMP and currently used as a nonspecific phosphodiesterase inhibitor, inhibited transport. All of these effects are compatible with cGMP transport by a member of the MRP family.

Table 4. Effect of ion channel inhibitors on the ATP-dependent uptake of [3H]cGMP by inside-out human erythrocyte membrane vesicles. Data represent mean ± SEM of n experiments. The control level of [3H]cGMP uptake with dimethylsulfoxide (n = 16) for furosemide, SITS, probenecid, staurosporine and Ro 31–8220 was in the range of 77–82 pmol·mg−1 protein. The control level of [3H]cGMP uptake with ethanol (n = 18) for phloridzin, niflumic acid, and verapamil was in the range of 90–105 pmol·mg−1 protein. The control level of [3H]cGMP uptake for the remaining drugs was in the range of 129–135 pmol mg−1 protein.
Compound Concentration (µm)[3H]cGMP uptake (%control)n
Anion channel inhibitors
 Frusemide   0.595.3 ± 2.36
   190.7 ± 3.16
   537.1 ± 2.18
  1019.4 ± 2.66
  206.6 ± 0.94
  50 6.9 ± 1.39
 Niflumic acid   1105.2 ± 2.14
   545.6 ± 3.76
  1021.4 ± 1.56
  205.4 ± 0.53
  505.6 ± 1.19
 Phloridzin   1109.8 ± 3.35
   599.9 ± 4.75
  1075.2 ± 4.27
  5028.3 ± 2.310
 SITS  2003.1 ± 0.93
 Probenecid  20011.3 ± 0.73
  4008.1 ± 1.13
Cation channel inhibitors
 BaCl2 40064.7 ± 3.23
 100047.9 ± 1.63
 CsCl 400100.1 ± 2.73
 1000102.4 ± 3.13
 4-Aminopyridine 50087.7 ± 4.63
 Tetraethylammonium Cl1000087.3 ± 3.43
 Verapamil  5066.8 ± 1.73
 Verapamil + 1 mm GSH  5071.3 ± 3.53
PKC inhibitors
 Staurosporine  102.9 ± 0.73
 Ro31–8220  1074.2 ± 2.33
 Colchicine  5098.4 ± 3.08
 Imidazole 10094.1 ± 0.83
 Dideoxyforskolin 10095.3 ± 2.23
 Forskolin 10066.7 ± 1.63
 IBMX 10016.5 ± 1.34

Immunodetection of MRP4 and MRP5 proteins in human erythrocytes and COR-L23/R cells

To identify candidate proteins that could possibly mediate the cGMP transport, immunoblot analysis was performed on membrane vesicles from human erythrocytes using monoclonal antibodies against MRP5 [18] and MRP4 [19]. The anti-MRP5 Ig, M5I-1, specifically detected an intact band at 190 kDa which shifted to 160 kDa after treatment with peptide N-glycosidase F (PNGaseF) (Fig. 5A) suggesting that MRP5 is N-glycosylated. A protein with the same apparent molecular mass was also detected in doxorubicin-resistant MOR/R and COR-L23/R lung tumour cells but at greatly reduced level in the doxorubicin-sensitive COR-L23/P lung tumour cells (Fig. 5B). The anti-MRP4 Ig detected an intact band at 170–180 kDa in human erythrocytes and MOR/R cells but not in COR-L23/R cells (Fig. 5B).

Figure 5.

Immunodetection of MRP4 and MRP5 in human erythrocytes and COR-L23 cells. Inside-out membrane vesicles prepared from human erythrocytes, COR-L23/P and COR-L23/R cells, and crude lysates from MOR/ADR cells were size fractionated on 7.5% SDS/PAGE, blotted and immunostained with (A) M5I-1 mAb, and (B) anti-MRP4 mAb, detecting MRP5 and MRP4, respectively. Membranes (40 µg) from human erythrocytes were also treated with PNGase F to remove the N-glycans and then immunostained with M5I-1 (A). The amount of protein loaded per lane is indicated at the bottom of each blot. Arrows mark the immunodetected band for MRP4 and MRP5. Doxorubicin-resistant lung adenocarcinoma cell line MOR/R was used as a positive control; doxorubicin-sensitive human large-cell lung tumour cell line COR-L23/P was used as a negative control.


It is now well recognized that inside-out membrane vesicles prepared from human erythrocytes can take up both glutathione-conjugates, such as DNP-SG, and cyclic nucleotides, e.g. cGMP, by rapid ATP-dependent transport processes. In the present study, uptake of cGMP is shown to consist primarily of a low-affinity component with a maximum uptake (Umax) of 300–400 pmol·mg−1·min−1 and a dissociation constant (Kd) in the range of 50–82 µm and possibly also a second high affinity component of uptake contributing less than 20% of the total trasnport even at low concentrations. However, MK-571, glibenclamide, DNP-SG, clotrimazole and cAMP all inhibit this uptake as if there were only a single component of transport (Figs 3C and 4). If present, the high-affinity component may correspond to the high-affinity transport previously reported. In those studies [2,5] ATP-dependent cGMP uptake into inside-out membrane vesicles from human erythrocytes was found to have two components, a high affinity uptake with Umax1 = 0.2–0.4 pmol·mg−1·min−1 and Kd1 = 2.4–4.7 µm and a low affinity uptake with Umax2 = 1.6pmol·mg−1·min−1 and Kd2 = 170 µm. The maximum uptake rates in these studies were very low, being just over threefold higher than the background [2]. The maximum uptake rate in the current study is two orders of magnitude higher. The reasons for this remarkable difference are unclear though there may be several factors involved. These include different osmolalities of the solutions used to measure uptake, differences in the methods of vesicle preparation, and possibly even differences in the profile of transporters present on the red cell membranes from different donors.

In the present study, the vesicles were resealed and assayed in low osmolality solution. This contrasts with the previous study where the vesicles were resealed at low osmolality but assayed in NaCl/Pi with an osmolality 10–100-fold higher. To check these solutions as possible influencing factors, cGMP uptake in the current study was also measured using solutions as employed in the previous studies. Under these conditions, uptake rates were lower, but only by a factor of two, i.e. too small an effect to explain alone the discrepancies in results between the two studies.

Another possible explanation for the discrepancy is the presence of traces of calcium in the previous studies. It was shown there that calcium can inhibit cGMP transport with an IC50 of about 40 µm and the inclusion of 100 µm EGTA was found to increase uptake rates by 100% [4] Even so no chelators were included in most of those studies either during vesicle preparation or during uptake measurements. The vesicle preparations used in the present work were produced by a one-step spontaneous vesiculation method in the presence of 100 µm EGTA [26–28].

Further differences relate to the percentage of inside-out vesicles generated. The previous study used the procedure of Steck and Kant [30] which yields a much higher percentage (typically > 60%) of inside-out vesicles than are routinely produced by spontaneous vesiculation (usually 30–37%). However, these differences would be expected to produce lower uptake rates in the vesicles generated by spontaneous vesiculation, not higher rates as observed here. Whether there are inhibitory elements on the vesicle membranes that are stripped off more effectively by spontaneous vesiculation or stimulatory elements removed in the lengthy Steck and Kant procedure remains to be determined. Certainly it has been proposed that the spontaneous vesiculation process may remove restrictive links with cytoskeletal elements [28] allowing more lateral mobility and thus the possibility of alterations in associations between membrane proteins. The additional possibility that different donors possess different profiles of transporters on their red cell membranes is currently being explored.

In contrast to uptake of cGMP, ATP-dependent uptake of DNP-SG into inside-out erythrocyte membrane vesicles in the current study clearly possesses more than one component. This has been noted also by Akerboom et al. [7] and Pulaski et al. [10]. The observations are consistent with the presence of two components, one, a low-capacity high-affinity component (Kd 1–10 µm) that predominates at concentrations below 5 µm and another low-affinity component (> 100 µm) that is dominant at high concentrations.

The high-affinity component of DNP-SG transport, identified previously as being mediated by MRP1 [10–12], appears not to be involved in transport of cGMP (see below). However, the low affinity component of DNP-SG transport is inhibited by cGMP and may well be the transporter that mediates low-affininty cGMP uptake. DNP-SG can indeed inhibit cGMP transport. Furthermore, the apparent dissociation constant for DNP-SG uptake, kd2, and the IC50 for DNP-SG-mediated inhibition of cGMP uptake are approximately equal, i.e. 65 µm compared with 82 µm. Though the two values for cGMP uptake and for cGMP inhibition of DNP-SG uptake differ, i.e. 50–82 µm compared with 133 µm, it remains possible that a single transporter accounts for both uptakes as the interaction between cGMP and DNP-SG is likely to be more complex than simple competition. Further evidence for greater complexity is provided by the curve fits with Hill coefficients greater than 1.

As a wide variety of MRP transport inhibitors block cGMP and DNP-SG transport (Table 3 and Fig. 4), members of the MRP family are very plausible candidates for the transporters mediating ATP-dependent uptake of cGMP and DNP-SG into inside-out erythrocyte vesicles. MRP1 and MRP5 have previously been detected on the membranes of human erythrocytes [8,9]. MRP1 is known to mediate the high-affinity transport of DNP-SG [10–12] and MRP4 and MRP5 are known to transport cGMP [9,13,17]. In the present study in addition to confirming the presence of MRP1 and MRP5 in erythrocyte membranes, we have detected the presence of MRP4 (see Fig. 5). Although no cross-reactivity tests have been reported for anti-MRP4, the present finding that the anti-MRP4 Ig does not detect an immunoreactive band in COR-L23/R cells which express MRP1 and MRP5 suggests that this Ig does not cross-react with MRP1 or MRP5.

Several lines of evidence show that MRP1 is most unlikely to be the transporter responsible for the transport of cGMP. It has already been shown that cGMP does not inhibit transport of LTC4, an established high-affinity MRP1 substrate [4]; cGMP does not inhibit MRP1-mediated transport of DNP-SG; and, verapamil inhibits cGMP uptake without any requirement for GSH (this study) while GSH is needed for its inhibition of MRP1-mediated transport [12,51]. Most convincingly, uptake of 3.3 µm cGMP is unaffected by the presence of the conformation-dependent monoclonal Ig against MRP1, QCRL-3, at 10 µg·mL−1 which has been shown to block MRP1 mediated high-affinity uptake of DNP-SG [12].

Both MRP4 and MRP5 have been shown to transport cGMP, though the apparent dissociation constants have proven controversial. The original reports indicated relatively high-affinity trasnport with Km values in the range of 2–10 µm[9,13], but the transport is clearly much lower affinity in HEK293 cells transfected with MRP4 or MRP5 [17]. To examine further the possible role of the MRPs in erythrocytes, several established MRP substrates, inhibitors, and modulators were tested for their ability to block the [3H]cGMP uptake into inside-out vesicles. MK-571, a leukotriene receptor antagonist, inhibited the cGMP transport with an IC50 value of 0.38 ± 0.01 µm. This appeared to be the most potent of the inhibitors tested in the present study. MK-571 has been shown to inhibit MRP1, MRP2, MRP3 and MRP4 mediated transport [9,36,38,39] but has been reported to have no affect at concentrations of up to 50 µm on cGMP transport attributed to MRP5 [9].

Methotrexate and E217βG were found to inhibit the cGMP uptake in human erythrocyte vesicles with E217βG at 65 µm inhibiting about 90% and methotrexate inhibiting about 75% at 275 µm and completely at 375 µm. These compounds are established MRP4 substrates with Km values around 220 µm for methotrexate [19,39], and 30 µm for E217βG [39,3,4]. At present, methotrexate and E217βG appear to be substrates of MRP4 [13,39] and not of MRP5.

cAMP, which has been shown to be transported by both MRP4 and MRP5 [9,13], inhibited all of the cGMP uptake with an estimated IC50 value of 315 ± 70 µm. The ratio of this value to the apparent Kd value for cGMP, about 5, is consistent with the value, 5, reported for MRP4 [13] but not the value, 380, reported for MRP5 [9].

Taurocholic acid, an established substrate for MRP1, MRP2, and MRP3, had only a weak inhibitory effect on the cGMP uptake (about 20% at 200 µm and about 60% at 350 µm). Interpretation of such inhibition is difficult given that taurocholic acid would probably be acting as a detergent at high concentrations.

Glibenclamide was found to be a very potent inhibitor of cGMP transport in the human erythrocyte membrane vesicles with an estimated IC50 value of 1.9 ± 0.1 µm. This was the second most potent of the inhibitors tested against the cGMP transporter. Glibenclamide has been shown to interact with several ABC proteins including the sulfonylurea receptor [42,52], CFTR [40], and more recently, P-glycoprotein and MRP1 [41,53], blocking their function. Given that glibenclamide inhibits many ABC transporters, its action serves merely to point to an ABC transporter as being responsible for cGMP transport but not to identify which ABC transporter it is.

Several other compounds including forskolin, an activator of adenylate cyclase, and IBMX, a nonspecific phosphodiesterase inhibitor with structural similarity to cGMP, also inhibited the cGMP uptake. Forskolin at a tested concentration of 100 µm inhibited the cGMP transport by about 35% while its inactive analogue, 1,9-dideoxyforskolin, had no effect at the same concentration. Forskolin has previously been shown to inhibit the cGMP transport in other types of erythrocyte vesicle preparations [2] but the exact mechanism of its action remains unknown. It is possible it could interact with a drug-binding site on the cGMP transporter which bears structural homology to adenylyl cyclase, as it has been previously proposed for the effect of forskolin on P-glycoprotein [54]. Another structurally related compound, IBMX, currently used as a nonspecific phosphodiesterase inhibitor, proved an effective inhibitor of cGMP transport at 100 µm. The PKC inhibitors, staurosporine and Ro 31–8220, inhibited the cGMP uptake. Staurosporine completely inhibited the cGMP uptake at 10 µm while at the same concentration Ro 31–8220 inhibited by about 25%. The greater potency of staurosporine is consistent with a recently proposed mechanism for its action: interaction with the ATP binding sites of the transporter inhibiting energy-dependent drug efflux activity [49].

The major characteristics of the cGMP transporter in human erythrocytes, as described in the present study, are inhibition by MK-571, glibenclamide, E217βG, methotrexate, DNP-SG and cAMP. This inhibitor profile matches well with that for MRP4. At present, five distinguishing features exist between human MRP4 and MRP5. First, MRP4 is inhibited by MK-571 [39] while MRP5 is not [9]. Second, MRP4 transports methotrexate [39] but MRP5 does not confer resistance to this drug suggesting no transport [55]. Third, human MRP4 transports 17βE2G [13] but human MRP5 appears not to do so [9]. Fourthly, cAMP and cGMP interact with MRP4 at similar concentrations [13] while the apparent KD for cAMP with MRP5 is much higher than that for cGMP [9].

In summary, using inside-out membrane vesicles prepared from human erythrocytes by a spontaneous, one-step vesiculation process, a dominant low affinity component of cGMP transport with Kd value in the region of 50–80 µm has been identified. This transport is completely inhibitable by MK-571, glibenclamide, clotrimazole, cAMP and DNP-SG, consistent with the idea of a single transporter being responsible. It is also markedly inhibited by methotrexate. All of these and cGMP block low affinity [3H]DNP-SG transport in human erythrocytes and so the low affinity transport of DNP-SG and the transport of cGMP may be mediated by one and the same transporter. The characteristics of transport indicate that MRP1, which mediates the high-affinity transport of DNP-SG, is not the protein responsible. The properties of the cGMP transport are similar to those for MRP4.


We would like to thank Dr G. Kruh for the anti-MRP4 Ig, Dr G. Scheffer and Dr R. J. Scheper for the M5I-1 Ig and Dr M. Turner for the MK-571.