|Home | About | Journals | Submit | Contact Us | Français|
5′-fluorosulfonylbenzonyl 5′-adenosine (FSBA) is an ATP analog that covalently modifies several residues in the nucleotide-binding domains (NBDs) of several ATPases, kinases and other proteins. P-glycoprotein (P-gp, ABCB1) is a member of the ATP-binding cassette (ABC) transporter superfamily that utilizes energy from ATP hydrolysis for the efflux of amphipathic anticancer agents from cancer cells. We investigated the interactions of FSBA with P-gp to study the catalytic cycle of ATP hydrolysis. Incubation of P-gp with FSBA inhibited ATP hydrolysis (IC50= 0.21 mM) and the binding of 8-azido[α–32P]ATP (IC50= 0.68 mM). In addition, 14C-FSBA crosslinks to P-gp, suggesting that FSBA-mediated inhibition of ATP hydrolysis is irreversible due to covalent modification of P-gp. However, when the NBDs were occupied with a saturating concentration of ATP prior to treatment, FSBA stimulated ATP hydrolysis by P-gp. Furthermore, FSBA inhibited the photocrosslinking of P-gp with [125I]-Iodoaryl-azidoprazosin (IAAP; IC50 = 0.17 mM). As IAAP is a transport substrate for P-gp, this suggests that FSBA affects not only the NBDs, but also the transport-substrate site in the transmembrane domains. Consistent with these results, FSBA blocked efflux of rhodamine 123 from P-gp-expressing cells. Additionally, mass spectrometric analysis identified FSBA crosslinks to residues within or nearby the NBDs but not in the transmembrane domains and docking of FSBA in a homology model of human P-gp NBDs supports the biochemical studies. Thus, FSBA is an ATP analog that interacts with both the drug-binding and ATP-binding sites of P-gp, but fluorosulfonyl-mediated crosslinking is observed only at the NBDs.
The ATP Binding Cassette (ABC) family of transport proteins constitutes one of the largest gene families1. The clinical importance of this family is apparent from the fact that of the 48 human ABC transporters, 17 are implicated in human diseases.2 P-glycoprotein (P-gp), the first human ABC transporter to be discovered,3 plays a significant role in multidrug resistance (MDR). MDR is a phenomenon in which cells show resistance to chemically diverse drugs with multiple mechanisms of action and is one of the mechanisms by which tumors become resistant to chemotherapeutic drugs4. Thus, significant effort has been expended over the last two decades to understand the mechanism of P-gp in particular and ABC proteins in general [for reviews see5–8]. P-gp is an energy-dependent molecular pump that utilizes the energy of ATP hydrolysis to move drug-substrates across the cell membrane.9 The enzymatic hydrolysis of ATP and the transport of drug-substrates are tightly coupled phenomena that occur in different domains of the protein.10 The movement of drug-substrates occurs as a consequence of conformational changes in the membrane-spanning transmembrane domains (TMDs); ATP hydrolysis occurs at the intracellular nucleotide-binding domains (NBDs). The development of probes for these two regions of the protein has played a significant role in elucidating the catalytic cycle of P-gp-mediated ATP hydrolysis as well as the mechanism of the transport cycle.
ABC transporters have relatively low affinities (in the micro-molar range) for their drug-substrates as well as for nucleotides.11 These low affinities present technical difficulties that have been enumerated in detail by Senior and colleagues.12 Photoaffinity reagents that can form covalent bonds with the amino acid side chains have proved extremely useful in probing the substrate- and nucleotide-binding sites.13 Over the last few decades there has been a concerted effort to design nucleotide-based affinity labels.14 The general consensus is that labels that incorporate the reactive group into the 5′-hydroxyl group cause minimal perturbation of the adenosine moiety.15 The Colman group at the University of Delaware has taken the lead in synthesizing and characterizing such reagents. One of the most successful reagents to emerge from their work has been 5′-fluorosulfonylbenzonyl 5′-adenosine (FSBA).
The sulfonylfluoride group of FSBA that replaces the phosphate groups of ATP is a reactive functional group that can react with several amino acids.16–18 As the sulfonylfluoride participates in a broad range of reactions, there is a reasonable probability of reaction within any particular active site.15 The sulfonylfluoride acts as an electrophilic agent that has been shown to form covalent bonds with tyrosine, lysine, histidine, serine, and cysteine.15 Thus, FSBA has been used to localize the nucleotide-binding sites, characterize the active site and study the catalytic mechanism of numerous enzymes. These include the cAMP-dependent protein kinase II,19 isocitrate dehydrogenase kinase,20 rabbit muscle pyruvate kinase,21 myosin, multifunctional protein CAD22 and the mitochondrial F1-ATPase.23 This reagent has proved particularly useful in studying the protein kinase family24 and studies show that FSBA produces a stable, isolable product after reacting with the active sites of the enzymes.25 FSBA-protein interactions have been studied extensively, including the effect of FSBA binding and/or crosslinking on enzyme activity and kinetics,26, 27 the use of antibodies and LC-MS to identify the sites of FSBA crosslinking28 and the use of [14C]-labeled FSBA29. Nonetheless, there have been no studies that exploit FSBA to probe the nucleotide-binding sites of ABC transporters.
In this study we show that FSBA interacts with both the nucleotide- and substrate-binding sites of P-gp. In addition, our data demonstrate that FSBA is crosslinked to P-gp. Significantly, FSBA also reverses P-gp-mediated transport of a fluorescent substrate in intact cells. These results show that, as may be expected based on the interactions of FSBA with numerous other proteins (for review, see 14), this nucleotide analog interacts with the NBDs of P-gp. In addition, we demonstrate that FSBA interacts with the substrate-binding sites in the TMDs of this transporter. A similar dual effect has also been observed with disulfiram, although the mechanism is significantly different. Disulfiram, which is not an ATP analog, reacts with the Walker A cysteine residue in each NBD, thus inhibiting ATP hydrolysis, while it also interacts at the substrate-binding sites of P-gp.30, 31 However, in Cys-less P-gp, it interacts only at the transport-substrate site in the TMDs. FSBA, as an ATP analog, either competes with ATP for binding at the NBDs (even in Cys-less P-gp) or binds to another site and has an effect on ATP binding at NBDs. It also interacts at the drug-binding sites, while ATP does not. Using mass spectrometric analysis, we were able to demonstrate that FSBA crosslinks residues in NBDs (for example, K411 in NBD1) as well as in the regions near NBDs. We were not able to detect FSBA crosslinked peptides originating from the TMDs, suggesting that interaction between the FSBA and the substrate-binding site in TMDs is reversible. In addition, this FSBA interaction at the TMDs does not require functional ATP sites, as the IAAP labeling in the Y401A/Y1044A double mutant lacking ATP-binding capacity is still inhibited by FSBA.
[125I] Iodoarylazidoprazosin ([125I]-IAAP), 2,200 Ci/mmol, and [adenine-8-14C]-FSBA (57 mCi/mmol) were obtained from Perkin Elmer Life Sciences (Boston, MA). 8-azido-[α-32P]ATP (15–20 Ci/mmole) and 8-azidoATP were purchased from Affinity Labeling Technologies, Inc. (Lexington, KY). The P-gp-specific monoclonal antibody C-219 was obtained from Fujirebio Diagnostics Inc. (Malvern, PA). All other chemicals were obtained from Sigma-Aldrich Chemical Co. (St. Louis, MO).
KB-3-1 and P-gp-overexpressing KB-V1 cells were maintained in DMEM supplemented with 10% FBS, penicillin, streptomycin and the KB-V1 cells were grown in media containing 1 μg/ml vinblastine.32
Preparation of Crude Membranes From High-Five Insect Cells Infected With Recombinant Baculovirus Carrying the Wild-Type and Mutant Human MDR1Gene. High-Five insect cells (Invitrogen, Carlsbad, CA) were infected with the baculovirus carrying the human MDR1 cDNA (either wild-type or Y401A/1044A or the cys-less mutant where all seven native cysteine residues are replaced with alanine) with a 6X Histidine tag at the C-terminal end as described previously.33 Crude membranes were prepared and stored at −70°C as described previously.33, 34
Human P-gp from crude membranes of High Five insect cells was purified as described previously.33, 35 The crude membranes were solubilized with octyl β-D-glucopyranoside (1.25%) in the presence of 20% glycerol and a lipid mixture (0.1%). Solubilized proteins were subjected to metal affinity chromatography (Talon resin, Clontech, Palo Alto, CA) in the presence of 0.95% octyl β-D-glucopyranoside and 0.04% lipid; 80% purified P-gp was eluted with 200 mM imidazole. P-gp in the 200 mM imidazole fraction was then concentrated (Centriprep-50, Amicon, Beverly, MA) to ~ 0.5 – 1.5 mg/ml and stored at −70°C. P-gp was identified by immunoblot analysis using the monoclonal antibody C21933 and quantified by the Amido Black protein estimation method as previously described.36 Purified P-gp was reconstituted into proteoliposomes by dialysis using a lipid:protein ratio of 10:1 as described.37 The purified Pgp in proteoliposomes exhibited 50 μM verapamil-stimulated ATPase activity in the range of 1100 to 1300 nmol of Pi/min/mg of protein as previously described35, 38.
Crude membranes (10 μg/protein/100 μl) from High Five cells expressing P-gp were incubated at 37°C for 30 min with varying concentrations of FSBA in the presence and absence of sodium orthovanadate (Vi) (0.3 mM) in ATPase assay buffer (50 mM MOPS-KOH, pH 7.5, 50 mM KCl, 5 mM sodium azide, 1 mM EGTA, 1 mM ouabain, 10 mM MgCl2). The reaction was initiated by the addition of 5 mM ATP and incubated for 20 min at 37°C. SDS solution (0.1 ml of 5% SDS) was added to terminate the reaction and the amount of inorganic phosphate released was quantified with a colorimetric reaction, as described previously.11
Crude membranes of High-Five insect cells (50–100 μg of protein) or purified and reconstituted protein (5–10 μg of protein) were incubated in an ATPase assay buffer (50 mM MES-Tris, pH 6.8, 50 mM KCl, 5 mM sodium azide, 1 mM EGTA, 1 mM ouabain and 10 mM MgCl2) containing 10 μM [α-32P]8azidoATP (10 μCi/nmol) in the dark at 4°C for 5 min. The samples were irradiated with a UV lamp assembly (PGC Scientifics, Gaithersburg, MD) fitted with two black light (self-filtering) UV-A long wave F15T8BLB tubes (365 nm) for 10 min on ice (4°C). Ice-cold ATP (10 mM) was added to displace excess non-covalently bound 8azido-[α-32P]ATP. After SDS-PAGE on a 7% NuPAGE gel, the gels were dried and exposed to Bio-Max MR film (Eastman Kodak Co.) at −70°C for 12–24 h. The radioactivity incorporated into the P-gp band was quantified using the STORM 860 PhosphorImager system (Molecular Dynamics, Sunnyvale, CA) and the software ImageQuaNT. One phase decay mode was used to fit the data given in Figure S2 (additional details are given in the legend to Fig. S2).
The crude membranes of High-Five insect cells (60–75 μg protein/100 μl) were incubated at room temperature in ATPase buffer with [125I]-IAAP (5–7 nM) for 5 min under subdued light. The samples were photocrosslinked for 5 min at room temperature followed by electrophoresis and quantification as described previously.34 Modifications to this procedure in specific experiments are detailed in the figure legends.
Purified and reconstituted P-gp (15–40μg protein) was incubated with 100 μM [14C]-FSBA for 30 min at 37°C. The reaction was stopped by the addition of SDS-PAGE loading buffer and the samples were electrophoresed on a 7% NuPage gel at constant voltage. The dried gel was used to quantify the radioactivity incorporated into the P-gp using the STORM 860 PhosphorImager system with a high sensitivity screen and ImageQuaNT software.
Protein bands were cut from polyacrylamide gels and the proteins were in-gel alkylated with 55 mM iodoacetamide, for 45 min at 25°C. Following washing with 50 mM ammonium bicarbonate, the proteins were in-gel trypsin digested for 16 h at 25oC using a Montage In-Gel Digest Kit (Millipore, Billerica, MA) on a 96-well C18 ZipPlate following the manufacturer’s guidelines. Peptides were eluted from the plate and an aliquot of each digestion reaction was analyzed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) on a MALDI micro MX (Waters, Milford, MA) to look for modified peptides. To determine the sites of modification, the remaining peptides were acidified by 1:1 dilution with Diluent Buffer containing 2% acetonitrile, 0.1% formic acid, and 0.05% HFBA in water. Separation of the peptides was performed at 500 nL/min and was coupled to online analysis by tandem mass spectrometry (nLC-ESI-MS/MS) on an LTQ ion trap mass spectrometer (ThermoElectron, San Jose, CA) equipped with a nanospray ion source. From the diluted peptides, 30 μL were loaded onto a 0.1 × 150 mm Magic C18AQ column (Michrom Bioresources, Inc, Auburn, CA) inline after a nanotrap column using the Paradigm MS4 MDLC (Michrom Bioresources, Inc., Auburn, CA). Elution of the peptides into the mass spectrometer was performed with a linear gradient from 95% mobile phase A (2% acetonitrile, 0.5% acetic acid, 97.5% water) to 65% mobile phase B (10% water, 0.5% acetic acid, 89.5% acetonitrile) in 45 minutes and then to 95% mobile phase B in 5 minutes. The peptides were detected in positive ion mode using a mass list containing the masses of the modified peptides identified by MALDI-TOF MS analysis.
A Nucleotide Binding Domain homology model of human P-gp was constructed using the X-ray crystal structure of the ABC transporter haemolysin-B (Hly-B), in complex with ATP/Mg2+, as the template (1XEF.pdb).39 Residues K408, P1045, T1046, R1138 and V1139, missing in the Hyl-B structure, were included in the final model, importing three loops from the X-ray structure of the ABC transporter associated with antigen processing TAP1 (2IXF.pdb).40 The fragments incorporated into the model were the following: PNHPNVQVLQ (residues 27–36 of TAP1-NBD1); PNHPNVQVLQGL (residues 27–38 of TAP1-NBD2) and LTRTPT (residues 117–122 of TAP1-NBD2). Protein residues were in-silico mutated to the P-gp sequence, and the side-chain conformations of the new residues were visually inspected and modified when necessary. The structure was energy-minimized by CHARMm.41 Docking studies of FSBA at the NBDs were carried out using the program Vina-Autodock42 or by manually positioning the FSBA molecule with its adenine group π-stacked to the Y401 and adjusting the torsion angles of rotatable bonds. After ligand docking, the protein-FSBA structures were energy-minimized by CHARMm.
Accumulation assays with rhodamine 123 (0.5 μg/ml) were carried out in P-gp-expressing KB-V-1 cells as described previously.43 For all samples, 10,000 events were counted and analysis was performed with Cell Quest (Becton-Dickinson Immunocytometry systems). The mean fluorescence intensity was calculated using the histogram stat program in Cell Quest software.
FSBA, which is an analog of ATP, has been used to identify and characterize the nucleotide-binding sites of several ATP dependent proteins,15 but had not been used in the study of ABC transporters. The structure of FSBA strongly resembles that of ATP, except for the presence of 5′-fluorosulfonyl-benzoyl in place of the phosphoryl groups of ATP (Fig. 1A,B). We used a flow cytometry based assay to investigate the effect of FSBA on P-gp-mediated drug transport. Flow cytometry using fluorescent transport-substrates of P-gp has been extensively used to measure transport function.44 Cells over-expressing P-gp accumulate less fluorescent substrate, such as the typically used rhodamine 123 than parental cells not expressing P-gp. Treatment of P-gp-over-expressing cells with modulators that interact with P-gp decreases the efflux of the transport-substrate and increases accumulation of the fluorescent substrates to levels seen in the parental cells.44 Fig. 2 clearly shows that KB-V1 cells that over-express P-gp accumulate lower levels of rhodamine 123 than the parental KB-3-1 cells. Treatment of the KB-V1 cells with the P-gp modulator XR9576 increases rhodamine 123 levels comparable to those observed in KB-3-1 cells. When the P-gp-expressing cells were incubated with 1.5 mM FSBA for 45 min at 37°C, the accumulation of rhodamine 123 also increased. The increase in the fluorescence signal was equivalent to that observed in the presence of XR9576, which is a well characterized inhibitor of P-gp-mediated transport.45 To further elucidate the mechanism of FSBA action, we investigated the effect of FSBA on both P-gp-mediated ATP hydrolysis as well as P-gp drug-substrate interactions.
As the structure of FSBA strongly resembles that of ATP (Fig. 1A,B), we determined its effect on ATP hydrolysis by P-gp. Fig. 3A shows that FSBA inhibits P-gp-mediated ATP hydrolysis in a concentration-dependent manner (IC50 = 0.21 ± 0.05 mM mean ± SD, n=3), with the possibility that FSBA has an effect on ATP hydrolysis either by binding directly to the nucleotide-binding domains (NBDs) of P-gp or elsewhere on the molecule. The t1/2 for the inhibition of P-gp-mediated ATP hydrolysis by FSBA is 3.75 ± 1.72 min (mean ± SD, n=3), determined by incubating P-gp with FSBA for increasing lengths of time at 37°C, prior to initiating ATP hydrolysis (Fig. 3B). In addition, we observed that FSBA-mediated inhibition of ATP hydrolysis is temperature-dependent (Fig. 3C). The time and temperature dependence suggests that inhibition of ATP binding alone is not sufficient for FSBA-mediated inhibition; chemical crosslinking is also necessary. Studies on rabbit pyruvate kinase have shown that the inhibition of this enzyme by FSBA can be partially reversed by the reducing agent dithiothreitol (DTT).21 The experiments described above were carried out in the absence of DTT. We found that the addition of DTT prevents the FSBA-mediated inhibition of ATPase activity with an apparent EC50 of 0.07 mM (Supplemental Fig. S1A). Further experiments were carried out to investigate whether DTT could reverse FSBA-mediated inhibition once the covalent crosslink had formed. Treatment with DTT after incubating P-gp with FSBA for 30 min at 37°C shows only a modest reversal of this inhibition (supplemental Fig. S1B: compare III and IV). Thus, if the P-gp is incubated with DTT during FSBA treatment, inhibition of ATPase activity is prevented (Fig S1A); however, once the reaction between FSBA and the amino acid side chains has progressed to completion, it cannot be reversed by DTT (Fig. S1B). Thus, the prevention of the FSBA-inhibitory effect of ATP hydrolysis can be ascribed to a quenching effect of DTT on FSBA, when both are added at the same time to initiate the reaction.
The work described above suggests that FSBA crosslinks to P-gp and this is established more directly with the data presented in Fig. 4. Purified P-gp, reconstituted into proteoliposomes, was incubated with [14C]-FSBA for 30 min at 37°C. The reaction was stopped by the addition of SDS-PAGE loading buffer and the sample was subjected to gel electrophoresis. In Fig. 4A, which is an autoradiogram, we show that the [14C] signal is associated with the P-gp band, demonstrating that the [14C]-FSBA is crosslinked to the P-gp. When the [14C]-FSBA was incubated with P-gp in the presence of 10 mM ATP, however, there was 50% reduction but not elimination of [14C]FSBA crosslinking to P-gp. In addition, we observed comparable levels of [14C]-FSBA crosslinking to wild type P-gp and a cys-less mutant (Fig. 4B), where all seven native cysteine residues have been replaced by alanine. This implies that the cysteine residues are not necessary for the observed crosslinking of FSBA to P-gp.
If FSBA binds to the NBDs, it should be possible to test its effect when these sites are occupied by ATP at saturating concentration. In the experiment shown in Fig. 5A, the NBDs were occupied with saturating concentration of ATP (5 mM) prior to the addition of FSBA. All additions were made on ice to prevent ATP hydrolysis and crosslinking of FSBA. The ATP hydrolysis was initiated by warming the reaction mixture to 37°C. Not only was the inhibition of ATPase activity reversed, but there was a >4-fold stimulation of FSBA-mediated ATP hydrolysis; the concentration required for 50% stimulation was 0.07 ± 0.05 mM (Fig. 5A). Thus, it is plausible that FSBA also interacts with the substrate-binding sites of P-gp and as a consequence stimulates ATP hydrolysis. Moreover, when the NBDs are occupied by 5 mM ATP, FSBA affects the Vmax but not the Km of P-gp-mediated ATP hydrolysis (Fig. 5B), which suggests noncompetitive stimulation of the ATPase reaction (Vmax 58.9/21.8, Km 0.64/0.54 mM in the presence/absence of FSBA, respectively). This observation is consistent with the fact that the nucleotide- and substrate-binding sites are located on different domains of P-gp.9 To permit competition of FSBA and ATP at the NBDs, we incubated FSBA in the presence of ATP for 30 min at 37°C. This incubation was carried out in the absence of Mg2+, permitting both ATP and FSBA to interact with the NBDs. However, ATP hydrolysis would not occur, as Mg+2 is necessary for P-gp-mediated ATP hydrolysis.46 ATP hydrolysis was initiated by the addition of Mg2+ to the reaction mixture. The presence of FSBA results in a ~3-fold increase in the Km (ATP) 1.31/0.44 mM in the presence/absence of FSBA, respectively) (Fig. 5C). This decrease in the affinity of ATP suggests competitive inhibition (i.e., both FSBA and ATP compete for the same or similar sites on P-gp). As it is not possible to prevent FSBA interaction at the drug-substrate site in these experiments, some of the observed kinetic changes may be influenced by its effect on substrate-binding site in TMDs.
Previous studies have demonstrated that the nucleotide analog 8-azido-[α–32P]ATP photocrosslinks specifically to the NBDs of P-gp and can be displaced by unlabeled nucleotides.38 We demonstrate in Supplemental Fig. S2 that FSBA inhibits the photocrosslinking of 8-azido-[α–32P]ATP to P-gp in a concentration-dependent manner with an IC50 of 0.68 ± 0.23 mM (mean ± SD; n=3). Similarly the photoaffinity analog of prazosin, [125I]IAAP, which is transported by P-gp, has been used extensively to study interactions at the substrate-binding sites.47 Drug-substrates of P-gp inhibit the photocrosslinking of [125I]IAAP, but nucleotides such as ATP and ADP have no effect.34 Although FSBA is a nucleotide analog, it inhibits the photocrosslinking of P-gp with [125I]IAAP with an IC50 of 0.17 ± 0.03 mM (Fig. 6A). The addition of ATP had no effect on the inhibitory effect of FSBA (IC50=0.09 ± 0.04 mM; n = 3). Furthermore, similar IC50 value (0.20 ± 0.08 mM) was obtained using the double mutant Y401/1044A, in which the tyrosine residues in A-loops of both NBDs are replaced with alanine, and ATP is incapable of binding to the mutant P-gp (Fig 6B).48 These results suggest that FSBA inhibits photolabeling of P-gp with [125I]IAAP either by directly interacting with substrate-binding sites or by binding elsewhere regardless of its interaction with the NBDs. Moreover, these data indicate that there are two types of interactions between P-gp and FSBA. One is the chemical crosslinking of FSBA to residues within or nearby NBDs affecting ATP sites. The other is an interaction most likely in the TMD region affecting the drug-substrate binding.
When FSBA was incubated with P-gp for 30 min at 4°C and then washed off by centrifugation, there was no inhibition of ATP hydrolysis (Fig 3C: compare the first two bars). On the other hand, if FSBA was incubated with P-gp at 37°C prior to removal of free FSBA by centrifugation, there was a significant inhibition of P-gp-mediated ATP hydrolysis (Fig. 3C: compare the third and fourth bars). These results suggest that FSBA does indeed crosslink to P-gp, but only at 37°C and not at 4°C. On the other hand, P-gp incubated with FSBA at 37°C and washed by centrifugation had no effect on the crosslinking of IAAP (Fig. 6C), suggesting that FSBA fraction that inhibits photolabeling with IAAP is washable and thus it likely interacts with but does not crosslink to the substrate-binding site(s). Similar results were obtained when the mutant P-gp (Y401A/Y1044A), which lacks the capacity to bind nucleotides, was used in lieu of the wild-type protein (Fig. 6D). These data suggest that it is unlikely that crosslinking of FSBA occurs at the drug-substrate-binding sites in the TMDs.
Purified P-gp in proteoliposomes was crosslinked with 2 mM FSBA by incubation at 37°C for 30 min either in the presence or absence of 5 mM ATP). These samples were then electrophoresed and the excised protein bands were subjected to in-gel digestion followed by mass spectrometry analysis, as described in Materials and Methods. Fig. 7 shows an example of the MALDI-TOF mass fingerprinting results. The two panels show a portion of the mass spectrum of the P-gp tryptic digest untreated and after treatment with FSBA. As an example, the peptide at m/z=898.95 is shifted to m/z=1331.85 following FSBA modification. Table 1 lists all the peptides identified with crosslinking to FSBA. Of the seven peptides identified, four were not crosslinked in the presence of ATP. It is noteworthy that no crosslinked peptides were identified in the TMDs, although peptides from the TMDs were observed in the untreated sample. These data demonstrate that FSBA crosslinks to residues within or nearby the NBDs of P-gp, while crosslinking to the TMDs is not detectable.
As there is no available X-ray crystal structure of human P-gp, a NBD homology model of P-gp on the basis of the haemolysin transporter (Hly-B) was built as described in Materials and Methods. The Hly-B structure (1XEF.pdb) was selected as a template for the following three reasons: high similarity between sequences at the NBDs, the fact that the Hly-B protein was crystallized with bound ATP and Mg2+ and the acceptable resolution of the structure (2.5 Å). The molecular modeling studies were carried out with the aim of better understanding the biochemical data presented here: (i) FSBA and ATP compete for the same site at the NBDs of P-gp and (ii) chemical crosslinking of FSBA to residues within or nearby NBDs. The highly similar structures of FSBA and ATP suggest that FSBA binds at the ATP binding site in the same fashion as ATP, with the adenine group π-stacked to the A-loop tyrosine 401 of NBD1 or the Y1044 of NBD2.48, 49 This typical binding mode set the conditions for the modeling studies. The program Vina-autodock was used to predict the possible non-covalent binding of FSBA at the nucleotide-binding sites. In fact, FSBA fits effectively at the NBDs with the adenine group interacting with Y401 (NBD1) or Y1044 (NBD2) through π-π contact, and the fluorosulfonyl moiety at the approximate position typically occupied by the ATP β- and γ-phosphate (see Fig. 8A and 8B). However, this pocket is surrounded by positively charged residues such as Walker A lysine-433 and also a Mg2+ ion, which creates an environment that’s better suited for negatively-charged ligands (ATP) than neutral ones (FSBA).
Further modeling was carried out to investigate other FSBA-protein interactions that better account for the efficient inhibition of ATP hydrolysis. The electrophilic fluorosulfonyl group of FSBA is reactive towards Lewis bases such as the side-chains of lysine, tyrosine, histidines, serine and cysteine,15 and responsible for the chemical crosslinking to the NBDs of P-gp. Therefore, modeling studies having both (i) the FSBA adenine moiety π-stacked to the Y401, and (ii) the fluorosulfonyl group pointed to any basic residue of the fragments identified as crosslinked to FSBA by LC/MS/MS (see Table 1) were performed. One FSBA molecule was successfully modeled with the fluorosulfonyl group pointing towards the amino group of Lys-411 (Fig. S3), thus offering an explanation for the origin of peptide fragment #2 detected by LC/MS/MS (see Figures 7, 8C, 8D and Table 1) and also providing an alternative mode that accounts for the efficient inhibition of ATP hydrolysis by FSBA. Interestingly, when FSBA is modeled at NBD2, the fluorosulfonyl group points to a glutamine residue (Q1054, the equivalent residue of K411 at NBD2), a much weaker Lewis base than lysine, providing a plausible explanation as to why the equivalent fragment of NBD1 at NBD2 is not found crosslinked to FSBA. Although FSBA could not be docked to explain the origin of the rest of the peptide fragments detected by MS, it is important to mention that just a small movement of loops next to the ATP site would make possible docking with the adenine moiety π-stacked to the A-loop tyrosine and the reactive –SO2F group pointing to K1102 (fragment #4 in Table 1), to H398 (fragment #5) and R459 (fragment #6). Although arginine has not been reported frequently as a target for FSBA, it has also been found crosslinked to FSBA.50 The location of FSBA peptides labeled 2 to 7 (Table 1) in the homology model of human P-gp NBDs is shown in Fig. 8C and D. Earlier reports indicated that FSBA could react with several different nucleophilic amino acids.17, 20 While lysine, tyrosine, and histidine residues are part of the sequence of the FSBA-crosslinked peptides, no cysteine is found in any of them. Our results with the Cys-less P-gp (see Fig. 4B) and FSBA-labeled P-gp peptides (Table 1) are also in agreement with these studies.
A homology model of human P-gp on the basis of the recently published X-ray structure of mouse P-gp was also built and subjected to docking analysis. The mouse P-gp sequence is very similar to human P-gp (87% identity) but the X-ray structure reported has two significant disadvantages: low resolution (3.8 Å) and the fact that it was obtained in the absence of nucleotide. Interestingly, we found that FSBA (as well as ATP) cannot be docked in the same way as in the model based on Hly-B structure because the conformation of the Walker A motif is different (Fig. S4). This suggests that a homology model of NBDs of human P-gp based on Hly-B structure (with bound ATP/Mg2+) is the appropriate one for docking studies of FSBA.
In this report, we have characterized in considerable detail the interactions of FSBA with P-gp. The main reason for this is that FSBA is an ATP analog and it has been used extensively to chemically modify and probe the nucleotide-binding sites of proteins that bind and hydrolyze nucleotides.14 This could thus also be a useful reagent to study the NBDs of P-gp as well as other ABC transporters.
The reports in the literature primarily consider FSBA as a nucleotide analog14 for kinases and other ATP-binding proteins. However, our results show that FSBA interacts with NBDs and also has an effect on the substrate-binding sites of P-gp (Fig. 2, ,5A,5A, and and6).6). In intact cells, FSBA, similar to other P-gp inhibitors (XR9576), blocked the efflux of rhodamine 123 (Fig. 2), suggesting that FSBA inhibits P-gp mediated transport in intact cells by interacting at the transport-substrate sites in the TMDs rather than in the NBDs, as one would expect that the NBD sites are occupied by ATP under physiological conditions with the intracellular concentration of ATP in the range of 3–5 mM. Most nucleotides (and nucleotide analogs) are not transport-substrates of P-gp and the transport-substrates do not directly interact with the NBDs.10 This is consistent with a large body of work that shows that ATP hydrolysis and drug-substrate transport occur at different domains.9 The TMDs are anchored in the plasma membrane and interact with the transport-substrates which are generally hydrophobic, while the NBDs, which are physically located in the cytoplasm, interact with hydrophilic nucleotides.9 In FSBA, the benzoylsulfonylfluoride replaces the tri-phosphate moiety of ATP (Fig. 1B). As a result, unlike ATP, which is negatively charged, FSBA is neutral15 and there is the addition of an aromatic benzoyl group. These changes make FSBA more hydrophobic (Log P = 2.64), thus penetrating the membrane domains and providing it with structural features that characterize transport substrates of P-gp. This can explain how FSBA is able to interact with the drug-substrate-binding sites of P-gp. We provide evidence for the effect of FSBA on the substrate-binding sites in Fig. 6A,B. The prazosin analog IAAP has been used extensively as a photoaffinity probe for the substrate-binding sites of P-gp.47 The concentration-dependent inhibition of IAAP binding to P-gp by FSBA (Fig. 6A,B) demonstrates that FSBA interacts directly with the substrate-binding site(s). It has been pointed out previously that the replacement of the phosphate groups of ATP with the sulfonylfluoride group makes FSBA uncharged and a less desirable ATP analog;15 however, these modifications make FSBA more accessible to the substrate-binding site(s) (see above). In addition, there appear to be different energetic requirements for binding to the ATP sites and chemical crosslinking. The former can be demonstrated by the virtually instantaneous displacement of 8-azido-[α-32P]-ATP by FSBA at 4°C. On the other hand, the irreversible FSBA-mediated inhibition of ATP hydrolysis (as a consequence of chemical crosslinking) has a t1/2 of approximately 4 min (Fig. 3B), and this inhibition is strongly temperature-dependent (Fig. 3C). It is clear that while the reversible binding of FSBA to NBDs is temperature-independent, the chemical crosslinking to residues within and nearby NBDs is temperature-dependent.
The sulfonyl fluoride group of FSBA, which replaces the phosphate groups of ATP, is a reactive functional group that can react with several amino acids.16–18 It has been argued that as the sulfonyl fluoride participates in a broad range of reactions, there is a reasonable probability of reaction within any particular active site.15 To distinguish between covalent and non-covalent interactions, P-gp was incubated for 30 min in the absence or presence of FSBA. Reactions were carried out either at 4°C or 37°C and the excess unreacted FSBA was removed by dilution with cold buffer followed by centrifugation. The P-gp samples that were incubated with FSBA at 37°C showed a strong inhibition of ATP hydrolysis (Fig. 3C: compare third and fourth bars), while those incubated at 4°C showed no inhibition (Fig. 3C: compare first and second bars). The control (incubation at 4°C) shows that un-reacted FSBA can be effectively removed by centrifugation and it is only the chemically-crosslinked FSBA that inhibits ATP hydrolysis. The binding of the transport substrate IAAP to wild-type P-gp or mutant Y401A/1044A (that cannot bind to ATP) is not affected by the formation of an FSBA-P-gp adduct (Fig. 6C,D), suggesting that FSBA is not crosslinked to the IAAP binding site. Thus, the interaction of FSBA at the drug-substrate binding pocket in TMDs is reversible and it does not depend on the functional status of ATP sites (Fig. 6A,B).
The covalent crosslinking of FSBA to P-gp is further verified by crosslinking of [14C]-labeled FSBA to purified P-gp reconstituted into proteoliposomes (Fig. 4A). In addition, we used mass spectrometry to identify sites on P-gp labeled with FSBA (Fig. 7). Seven peptides were labeled with FSBA, of which four were protected by pretreatment with 5 mM ATP (Table 1). This is consistent with the observation that incubation with a saturating concentration of ATP prior to addition of FSBA provides only partial protection to labeling by [14C]-FSBA (Fig. 4A). Identification of the peptides labeled with FSBA further suggests that FSBA crosslinks mainly to residues within or near the NBDs (five out of seven fragments are within or near the ATP sites; Table 1, and Figs. 7 and 8C, D), despite interrogating both the nucleotide- and substrate-binding sites (Figs. 3, ,66).
Previous studies have shown that FSBA participates in covalent reactions with various amino acids (for review see 15), including tyrosine, lysine, histidine, serine, and cysteine. The X-ray crystal structures of the NBD domains of several ABC transporters (e.g. MJ0796-1L2T.pdb; Hly-B -1XEF.pdb) show that the ATP molecule is sandwiched between the A-loop, the Walker A & B domains and the H-loop of one NBD and the signature sequence and D-loop of the opposing NBD.39, 51 Docking studies of FSBA at both ATP sites show the sulfonyl fluoride moiety of FSBA (which is the reactive group) is located close to the position normally occupied by the β- and γ-phosphate of ATP (see Figs 8A,B). Therefore, residues such as the cysteine, serine and lysine of Walker A, the histidine of the H-loop and the serine of the signature motif are potential targets for FSBA. However, all those residues (except the H loop-histidine) are not properly oriented for chemical reaction to the partially positively charged sulfur of FSBA. The H loop-histidine seems to be suitably oriented for reaction, but is too far (more than 5Å) from the reactive fluorosulfonyl group. In addition, LC/MS/MS analysis of P-gp labeled with FSBA did not yield any fragment with residues responsible for binding ATP (Fig. 7 and Table 1). On the other hand, modeling studies yielded a reasonable explanation for the origin of fragment #2, and further structural analysis revealed that just small conformational changes at loops close to the ATP sites are necessary to explain the origin of fragments #4, #5, and #6. However, the fact that ATP does not prevent crosslinking of three fragments out of seven suggests that FSBA is also binding to P-gp at other sites near the NBDs (see comments column in Table 1).
Tyrosine kinase inhibitors, which are developed based on their ability to bind to the catalytic site and prevent ATP binding (gleevec, nilotinib, etc.), interact at the drug-substrate sites of P-gp and ABCG2 but not at the ATP sites.52 FSBA is, to our knowledge, the first ATP analog that has an effect on the both the drug-substrate and ATP sites of P-gp, providing a unique and useful tool for biochemical studies of other ABC drug transporters including ABCG2, ABCC1. ABCC2, ABCC4 and ABCC7.
We thank Mr. George Leiman for editorial assistance.
†This research was supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, Center for Cancer Research.
These data include the effect of DTT on FSBA-P-gp interactions (Fig. S1), the inhibition of photolabeling of P-gp with 8-azido [α-32P] ATP by FSBA (Fig. S2), modeling of crosslinking of FSBA to peptide #2 in NBD1 (Fig. S3), and alignment of homology models of NBDs of human P-gp based on HlyB and mouse P-gp structures and docking of FSBA (Fig. S4). This is available free of charge at http://pubs.acs.org.