|Home | About | Journals | Submit | Contact Us | Français|
8-Chloroadenosine (8-Cl-Ado) is a ribonucleoside analogue that is currently in clinical trial for chronic lymphocytic leukemia. Based on the decline in cellular ATP pool following 8-Cl-Ado treatment, we hypothesized that 8-Cl-ADP and 8-Cl-ATP may interfere with ATP synthase, a key enzyme in ATP production. Mitochondrial ATP synthase is composed of two major parts; FO intermembrane base and F1 domain, containing α and β subunits. Crystal structures of both α and β subunits that bind to the substrate, ADP, are known in tight binding (αdpβdp) and loose binding (αtpβtp) states. Molecular docking demonstrated that 8-Cl-ADP/8-Cl-ATP occupied similar binding modes as ADP/ATP in the tight and loose binding sites of ATP synthase, respectively, suggesting that the chlorinated nucleotide metabolites may be functional substrates and inhibitors of the enzyme. The computational predictions were consistent with our whole cell biochemical results. Oligomycin, an established pharmacological inhibitor of ATP synthase, decreased both ATP and 8-Cl-ATP formation from exogenous substrates, however, did not affect pyrimidine nucleoside analogue triphosphate accumulation. Synthesis of ATP from ADP was inhibited in cells loaded with 8-Cl-ATP. These biochemical studies are in consent with the computational modeling; in the αtpβtp state 8-Cl-ATP occupies similar binding as ANP, a non-hydrolyzable ATP mimic that is a known inhibitor. Similarly, in the substrate binding site (αdpβdp) 8-Cl-ATP occupies a similar position as ATP mimic ADP-BeF3 −. Collectively, our current work suggests that 8-Cl-ADP may serve as a substrate and the 8-Cl-ATP may be an inhibitor of ATP synthase.
The nucleoside analogues currently used in the clinic are largely DNA-directed and act by inhibiting DNA synthesis . These analogues either incorporate into DNA and/or affect metabolic enzymes such as ribonucleotide reductase , purine nucleoside phosphorylase , and adenosine deaminase [4, 5] to perturb intracellular deoxynucleotide pools, resulting in decreased DNA synthesis in cells. In contrast, 8-chloroadenosine (8-Cl-Ado) contains a ribose sugar and is unique because it is RNA-directed. The advantage of RNA-directed agents such as 8-Cl-Ado is that they provide a valuable strategy for targeting quiescent cancers that do not actively synthesize DNA. Two RAID (Rapid Access to Interventional Development) contracts were awarded by the National Cancer Institute for the development of 8-Cl-Ado, and the drug is being tested in the first Phase I clinical trial for the treatment of chronic lymphocytic leukemia (CLL), an indolent leukemia. Ultimately, this agent will be used for other quiescent malignancies such as multiple myeloma and solid tumors.
In preclinical pharmacological studies using multiple myeloma (MM) cell lines and primary leukemia cells, 8-Cl-Ado was demonstrated to be phosphorylated into its triphosphate form (8-Cl-ATP) , which inhibits transcript synthesis by incorporation into RNA  and by inhibition of polyadenylation . In addition, a decline in intracellular ATP is observed in cells treated with 8-Cl-Ado, however, by unknown mechanisms. Studies using cell lines that are proficient or deficient in adenosine kinase and in vitro kinase assays demonstrated that 8-Cl-Ado is monophosphorylated (8-Cl-AMP) by adenosine kinase [6, 9]. 8-Cl-AMP is further metabolized to 8-Cl-ADP and 8-Cl-ATP, which are presumed to be catalyzed by monophospho-kinase and diphospho-kinase, respectively. We previously demonstrated that 8-Cl-ATP accumulates at high levels (>400 µM) [6, 10] while ATP levels decline to 50% of control after 6 h of treatment. The endogenous ATP concentration was reduced from ~1.7 mM to 0.65 mM a 12-hr incubation with 8-Cl-Ado in cell lines, and similar phenomenon was observed in primary leukemia cells obtained from the peripheral blood of patients with CLL .
The majority of cellular ATP is synthesized using respiratory chain oxidative phosphorylation by ATP synthase, which is the last enzyme in the respiratory chain . ATP synthase catalyzes the synthesis of ATP from recycling ADP [12–15], and it has been associated with a number of human diseases including cancer (for a review, see ). Studies suggest that cell surface ATP synthase can function as a receptor for ligands involved in several cellular processes including regulation of cell proliferation and differentiation, and immunological tumor recognition [17, 18]. Elevated expression of ATP synthase on endothelial cell surfaces has been reported to play an important role during angiogenesis, and angiostatin action is in part via the inhibition of cell surface ATP synthase . As such, recent developments have highlighted ATP synthase as a potential cancer target for therapeutics. Aurovertin B, an ATP synthase inhibitor, is currently under investigation for the treatment of breast cancer and has been shown to inhibit proliferation and induce apoptosis . More recently, monoclonal antibodies against ATP synthase was reported to inhibit proliferation and colony formation in human vascular endothelial cells and reduce tumor growth in xenograft models [21, 22].
In order to understand the function and mechanism of ATP synthase, X-ray crystal structures of both FO and F1 domains have been obtained by others [12–15, 23]. ATP synthase is composed of two major parts: the FO domain, an intermembrane proton channel, and the F1 domain, the catalytic complex. The nucleotide binding/catalytic sites are within F1 domain, which is composed of five different subunits with stoichiometry α3β3γδε. The crystal structures demonstrate that α and β subunits are structurally similar, and each subunit consists of three domains: a small N-terminal domain, a nucleotide binding domain and a helical C-terminal domain [12, 13, 23]. Both α and β subunits bind nucleotides but only the β subunit participates in the catalysis. It has been found that the three catalytic sites in the three β subunits have different affinities for nucleotides, and these together with other results have been used to propose a binding change mechanism of β units during the catalytic cycle of the enzyme [12–15]. It was proposed that rotation of the central FO stalk interconverted the three binding sites of F1 domain from open (low affinity binding) to tight (high affinity binding), from tight to loose (intermediate affinity binding), and from loose to open [23, 24]. The structures of the F1-ATPase provided insight into the binding change mechanism: the different conformations of β-subunits account for the conformations proposed to occur during the catalytic cycle [12, 13, 23].
We hypothesized that 8-Cl-ADP may serve as a substrate of ATP synthase and that 8-Cl-ATP may be an inhibitor to this key enzyme based on the following prior observations. First, the structural similarities between Ado and 8-Cl-Ado or ATP and 8-Cl-ATP. Second, the use of 8-Cl-Ado or 8-Cl-ATP by the same enzymes as Ado and ATP, and third, a decline in cellular ATP pool after incubation of cells with 8-Cl-Ado. Using biological, biochemical, and computational molecular modeling studies our current work tested this hypothesis.
All experiments were conducted using an exponentially growing multiple myeloma (MM) cell line, MM1.S [25, 26]. All cells were routinely tested for Mycoplasma infection using a commercially available kit (Invitrogen, Carlsbad, CA).
8-Cl-Ado was purchased initially from Bio Log (La Jolla, CA) and then obtained from Dr. V. Rao at the Drug Development Branch of the NCI. [3H]8-Cl-Ado and [3H]Ado were purchased from Moravek Biochemicals (Brea, CA). 8-Cl-ATP was custom-synthesized by BioLog (La Jolla, CA). Oligomycin was purchased from Sigma-Aldrich (St. Louis, MO). Coformycin (CF) and deoxycoformycin (dCF) were obtained from Dr. Robert Schultz at the NCI (Bethesda, MD). All other chemicals were reagent grade.
Oxygen consumption is measured using a Hansatech Oxytherm (Hansatech Instrument, England). Drug treated cells are placed in sealed respiration chamber containing 1 mL of fresh culture medium pre-equilibrated with 21% oxygen. The sample temperature was regulated with a thermostat control and stirred with a micro-stirrer [27, 28]. Oxygen levels were measured polarographically using a Hansatech oxygen electrode disc and the manufacturer’s software.
To quantitate the phosphorylated metabolites of Ado and 8-Cl-Ado, MM cells were incubated with oligomycin (2 µg/mL) for 30 min and then 10 µM [3H]Ado or [3H]8-Cl-Ado was added. The adenosine deaminase inhibitors CF and dCF (0.1 µM) were added to prevent deamination.. The cellular nucleotides were extracted using the perchloric acid extraction procedure  and analyzed on a gradient that separates free nucleoside, mono, di, and triphosphate forms .
For non-radioactive material, the cellular extracts were applied to an anion-exchange Partisil-10 SAX column as described before . The column eluate was monitored by UV absorption at 256 nm, and 8-Cl-ATP was identified by comparing its retention profile and absorption spectrum with those of an authentic standard. The intracellular concentration of nucleotides contained in the extract was calculated from a given number of cells of a determined mean volume. The cell number was determined using Coulter counter (Coulter Electronics, Hialeah, FL). This equipment is attached to a channelizer which was used to estimate the mean volume of cells in a given cell population. This volume was used to quantitate the concentration of nucleotides. The lower limit of sensitivity of this assay was 10 pmol in an extract of 5 × 106 cells corresponding to a cellular concentration of 1 µM. For radioactive 8-Cl-Ado a Radiomatic Flow-through HPLC system (Packard, Downers Grove, IL) was used. The eluate from the anion exchange column passed through an automatic radiometric detector along with liquid scintillation fluid (Ultima Flo, Packard) and tritium counts were recorded for each radioactive peak.
The ATP synthase structures (1BMF  and 2CK3 ) were obtained from Protein Data Bank (PDB ). In these crystal structures, ADP and ANP (an non-hydrolyzable ATP analogue) were observed at the interfaces of tight binding subunits (αdpβdp) and loose binding subunits (αtpβtp), respectively [23, 24]. As such, both subunits from 1BMF were extracted to conduct the modeling studies. ADP and 8-Cl-ADP were docked into the catalytic site between αdpβdp while αtpβtp was used to probe its interactions with ATP/8-Cl-ATP. In order to study the possible inhibitory effect on the enzyme, 8-Cl-ATP was also docked into the αdpβdp subunit of 2CK3 in which ADP-BeF3 −, an analogue of ATP, showed inhibition to ATP synthase . ADP and ANP were extracted from the crystal structure of 1BMF, followed with the correction of their chemical structures and the addition of all hydrogens. Since there is no crystal structure of the protein complexed with 8-Cl-ADP or 8-Cl-ATP, these two chemical structures were built based on the crystal structures of ADP and ANP, respectively. The structures were then used as our starting conformations in the docking process. Hydrogens were loaded to the complexes and all of the lone pairs were subjected to removal from both the ligands (ADP/8-Cl-ADP and ANP/ATP/8-Cl-ATP) and the receptors. Two docking packages, GOLD 3.2  and AutoDock 3.0 , were employed to dock the ligands into the receptor catalytic sites. For the docking process a box with a dimension of 20Å × 20Å × 20Å was created, centered at the geometrical center of the original ADP or ANP molecules accordingly. For the αdpβdp subunit of 2CK3, the geometrical center of ADP-BeF3 − was utilized for modeling. During docking, the receptors were treated as rigid bodies while the ligands were flexible, and Mg2+metal ion was kept in all docking as part of the binding pocket. All water molecules were removed from the receptors except for those in the αdpβdp subunit of 2CK3 in which a structural water molecule was maintained. All structures were prepared using MOE (Chemical Computing Group, Montreal, Quebec, Canada) , and all figures were created using PyMol (DeLano Scientific LLC, Palo Alto, CA) .
All statistical analyses were performed using GraphPad Prism software (GraphPad Software, Inc. San Diego, CA). Values of p ≤ 0.05 were considered as statistically significant.
The majority of ATP in cells is synthesized through oxidative phosphorylation that involves the mitochondrial respiration chain, which consists of complexes I-V. To determine whether 8-Cl-Ado affects complex I-IV, we compared O2 consumption in whole cells when cells were treated with either Ado or 8-Cl-Ado. As expected, there was a change in O2 consumption when cells were incubated with Ado or 8-Cl-Ado compared to control cells (Figure 1A and B). However the change was not different between the two substrates Ado (Figure 1A) and 8-Cl-Ado (Figure 1B).
To determine the role of ATP synthase in the accumulation of 8-Cl-ATP, oligomycin was used to inhibit the activity of this enzyme. When cells were treated with Ado, the accumulation of [3H]ATP was 200, 400, and 1900 µM at 0.5, 1 and 3 h, respectively (Figure 2A). When cells were pretreated with oligomycin prior to addition of [3H]Ado, there was a decrease in formation of [3H]ATP at all time points tested. Generally the decrease was about 50% of the control cells (p = 0.012, Figure 2A).
There have been numerous studies on the Warburg effect, examining the phenomenon that tumor cells have a high glycolysis rate, which is responsible for the generation of ATP. To establish the possible contribution of glycolysis in the formation of ATP or 8-Cl-ATP, we employed the use of deoxyglucose (dGlu), an inhibitor of glycolysis. Even using high levels of dGlu (10 mM), cells treated with dGlu still accumulated ~80% of ATP and 8-Cl-ATP levels of untreated cells (Supplemental Figure 1), suggesting that ATP synthase is driving ATP source in our system. As expected, pre-treatment of cells with both dGlu and oligomycin resulted in 80% decline in ATP or 8-Cl-ATP accumulation, however with these treatments, cells were undergoing apoptosis.
Treatment of cells with [3H]8-Cl-Ado resulted in 150 –170 µM [3H]8-Cl-ATP at 0.5 h. This concentration remained fairly similar at 1 and 3 h (Figure 2B). Hence, compared to ATP accumulation from [3H]Ado, 8-Cl-ATP concentrations were 10-times lower. However, as observed in the oligomycin and Ado incubated cells, [3H]8-Cl-ATP accumulation was also declined. The decrease was about 65% at all three time points (p = 0.001, Figure 2B). Similar to 8-Cl-Ado, incubation with 8-Cl-cAMP (serves as a precursor to 8-Cl-Ado) also resulted in a 50% decrease in 8-Cl-ATP after treatment with oligomycin (p = 0.042, Figure 2C).
To further determine if the inhibition of phosphorylation of [3H]Ado and [3H]8-Cl-Ado by oligomycin was only at the triphosphate level, we measured accumulation of the monophosphate, diphosphate, and triphosphates of Ado or 8-Cl-Ado. As shown in Figure 2D, the primary decrease was in triphosphate formation. [3H]ATP formation was decreased by 50% while [3H]8-Cl-ATP accumulation was declined by 60% of control cells. Monophosphate or diphosphate remained either similar or slightly increased in cells treated with oligomycin (Figure 2D).
In contrast to a decrease in triphosphate formation from 8-Cl-Ado or 8-Cl-cAMP after treatment of cells with oligomycin, accumulation of arabinosylcytosine triphosphate was not reduced when untreated or oligomycin-treated cells were incubated with arabinosylcytosine (Supplemental Figure 2). These data suggest that oligomycin treatment specifically affected ADP to ATP or 8-Cl-ADP to 8-Cl-ATP formation.
To determine whether 8-Cl-ADP is a substrate for ATP synthase, first we docked ADP into the binding site of the ATP synthase (1BMF) with GOLD and AutoDock. Both programs obtained the best docked poses with root mean square distance (RMSD) of 0.31 Å and 0.75 Å, respectively. With 8-Cl-ADP, these values were 0.43 Å by GOLD and 0.72 Å by AutoDock. Comparison between docked 8-Cl-ADP and crystal ADP revealed that the binding modes were similar except a small shift of the whole molecule (Figure 3A). Most portions of the structure were well superimposed; in particular the sugar moiety was almost completely overlapped (with RMSD 0.03 Å), and there was only a very small deviation along the diphosphate group. The most obvious difference was observed around the adenine group: the angle between 8-Cl-ADP and ADP purine rings was about 22.60°, and there was a big bump on the surface of 8-Cl-ADP (Figure 3A). Initial inspection of the binding pocket demonstrated that along the direction of hydrogen at 8-position on adenine, there existed a concave region which provided enough space to accommodate –Cl group (Figure 3B). As further indicated in Figure 3C, the –Cl surface indeed protruded into the small concave region and its volume just fitted in the space.
Detailed distance analysis of the starting conformation of 8-Cl-ADP (built from the ADP crystal structure in 1BMF) revealed that the distance between the –Cl group and the hydroxyl 5’-oxygen of ADP was only 1.61Å away (Figure 4A). However, in the final docked conformation, the distance was 2.44Å (Figure 4A). The analysis of the original ADP crystal structure showed that the distance was about 2.15Å between the 5’-oxygen of ADP and the hydrogen at 8- position. Similarly we found that the distance between the 5’-carbon and the 8-position -Cl was 3.03Å but in the starting conformation the distance was 2.65Å (Figure 4A). We also noticed that in ADP the distance between 8- position hydrogen and the carbon was 3.08Å.
As part of our studies, hydrogen bonding interactions were also analyzed in the docked and crystal complexes. There was no tremendous difference of the hydrogen bonding patterns in the two structures, except that an additional hydrogen bond was formed between 8-Cl-ADP and the α-subunit Arg373 (Figure 4B). No significant binding affinity change was observed based on our modeling prediction.
To determine if intracellular 8-Cl-ATP can inhibit ATP synthase and hence the conversion of ADP to ATP, cells were treated with 8-Cl-Ado to accumulate 8-Cl-ATP. After washing the cells free of 8-Cl-Ado, they were incubated with [3H]Ado and [3H]AMP, ADP, and ATP were measured in previously untreated (control) or 8-Cl-Ado treated cells (Figure 5A–C). Incubation of cells with 8-Cl-Ado resulted in about 450 µM 8-Cl-ATP with a 50% decline in ATP pool. Both in control and 8-Cl-Ado treated cells there was a rapid accumulation of [3H]AMP, ADP, and ATP with ATP being the highest metabolite. In the 8-Cl-Ado pretreated cells, there was a significant increase in AMP formation (p = 0.015, Figure 5A), no change in ADP accumulation (p = 0.082, Figure 5B) and major decline in ATP (p = 0.0009, Figure 5C). Cells treated with Ado alone accumulated about 220 µM of ATP at 0.5 hr. This increased in a time-dependent fashion and reached to 720 µM at 3 hr. Cells pretreated with 8-Cl-Ado showed a decline in ATP accumulation which was 160 µM at 0.5 hr and 420 µM at 3 hr (Figure 5C). These data suggest that while similar levels of ADP were present in untreated or 8-Cl-Ado treated cells, 8-Cl-ATP accumulation was substantially reduced in 8-Cl-Ado pretreated cells suggesting inhibition of the ADP to ATP conversion step of ATP synthase by 8-Cl-ATP.
To further test inhibition of ATP synthase by 8-Cl-ATP, we did two sets of molecular modeling. First, 8-Cl-ATP was docked to the loose binding site (αtpβtp) of the enzyme and second to the tight binding site (αdpβdp). In the loose binding state of αtpβtp subunits, 8-Cl-ATP occupies very similar binding mode as ANP, which is usually used as an analogue of ATP to trap the enzyme in a structure closely related to the ATP-bound state because it cannot be hydrolyzed as ATP would be. As is shown in Figure 6A, 8-Cl-ATP fits the binding pocket very well. Similar to the ADP binding observed, there is also a small pocket in this loose binding conformation to accommodate the chlorine atom (compared to the smaller hydrogen atom). In contrast, there was relatively large shift for the sugar ring of 8-Cl-ATP from the crystal structure of ANP sugar moiety (Figure 6A). Hydrogen bond analysis (data not shown) demonstrated a few pattern changes of the interactions, but obviously the binding affinity did not change significantly.
We also docked 8-Cl-ATP into the tight binding site (αdpβdp) based on the crystal structure of F1 domain of 2CK3; the docked 8-Cl-ATP occupies a very similar position and orientation of ADP-BeF3 −, the original inhibitor in 2CK3 (Figure 6B). Three catalytically essential residues[23, 24] α-Arg373, β-Arg189 and β-Lys162 are closer to the nucleotide in the docked complex than the ligand in the ADP-BeF3 −-F1 structure, creating a tighter binding interface for the nucleotide (Figure 6B).
Quantitatively, no large difference between the bindings of ADP/8-Cl-ADP and ATP/ANP/8-Cl-ATP to the enzyme was observed based on our modeling (Figure 6C). AutoDock predicted both ADP and 8-Cl-ADP had about -10 kcal/mol docked energy (Figure 6C), and for ATP/8-Cl-ATP, the binding was not as strong (below -7 kcal/mol) as ADP. With GOLD, similar results were obtained, although the fitness scores of 8-Cl-ADP/8-Cl-ATP were relatively lower than their corresponding natural substrates (Figure 6C).
F1-FO ATP synthase is a key manufacturing site of ATP. Given the critical role of ATP synthase, major efforts have been made to probe its protein structure and function. The availability of the structural information about ATP synthase F1 domain enabled us to conduct molecular docking studies of 8-Cl-Ado. Such methods have been widely used in drug discovery as well as in protein-ligand binding and interactions [35–40]. The structure of the F1 unit has been studied by a variety of methods [23, 41–44]. The first atomic resolution F1 subunit crystal structure (1BMF)  was employed in our study since it is used frequently as a reference structure and believed to represent a conformation in the active catalytic cycle [23, 24]. The crystal structures and ATPase reaction using F1 have been reported in detail because this portion of the enzyme could be isolated, is soluble, and functional in vitro, while the total ATP synthase requires proton motive force to run this enzymatic machinery as a rotary pump [45–47]. Due to this, ATP synthase requires either mitochondrial preparation , reconstitution in liposomes, and/or isolation from the cell surface, which makes biochemical studies challenging . In the present investigation, we relied on molecular modeling as well as whole cell assays using 8-Cl-Ado, Ado, and pharmacological inhibitor of the enzyme, oligomycin.
Due to the similarity between ADP and 8-Cl-ADP, we hypothesized that 8-Cl-ADP can bind to the ATP synthase tight binding site (αdpβdp) of F1 catalytic domain and act as an alternate substrate. Several lines of evidence suggest a possibility for this postulate. First, the ATP synthesis requires functional activity of complex I through IV for complex V (ATP synthase) to be active. None of these first four complexes on the mitochondrial membrane were affected by 8-Cl-Ado (Figure 1), suggesting that the last complex may be the target for this inhibition. Second, as observed for ATP formation from ADP, the accumulation of 8-Cl-ATP from 8-Cl-ADP was decreased by oligomycin, an inhibitor of FO subunit of ATP synthase (Figure 2). Third, decrease in 8-Cl-ATP accumulation is not due to decrease in ATP (a phosphate donor) as ara-CTP accumulation was not inhibited by oligomycin treatment. Fourth, 8-Cl-cAMP which serves as a prodrug for 8-Cl-Ado  also showed a decline in 8-Cl-ATP formation after oligomycin treatment. Finally, as detailed below, molecular modeling suggests that the chlorinated ADP binds well at the substrate binding site of the ATP synthase.
Molecular docking results on 8-Cl-ADP was very similar to ADP except a small shift of the whole structure and an obvious protrusion of the –Cl group towards a small pocket formed by residues Ala158, Gly159, Gly161 and Val164, all of which have relatively short side chains (Figure 6C). The protrusion was both due to the large –Cl group as well as the 22.60° tilting of the purine ring compared to ADP crystal structure. The modeling indicated likely conformational changes in the 8-Cl-ADP structure to accommodate the C8-chlorine atom. In the starting 8-Cl-ADP conformation the distance between the chlorine atom and the 5’-oxygen of ADP was only about 1.61Å (Figure 4A), which would most likely cause clashes. In the final docked conformation the distance increased to 2.44Å, which is closer to the same distance in the original ADP crystal structure (2.15Å). Similarly, the distance between the 5’-carbon atom and the C8-position chlorine (3.03 Å) became more comparable with that of ADP (3.08 Å), indicating that the docked conformation was optimized by shifting the adenine ring to reduced the geometric deficiency caused by C8-substitution. In addition, the chlorine and oxygen interactions become evident in the docked structure. The distance between the Gly159 carbonyl oxygen and the chlorine is 3.94Å compared to 4.43Å in the starting conformation (Figure 4A), and the angle of the C-Cl-O is about 150°. According to a recent study , the interaction between carbon-bonded halogens (Cl, Br and I) and electronegative atoms (O, N and S) were not marginal and the attractive natures of these types of interactions were due to strong electrostatic effects. Thus the C-Cl-O interaction here most probably further stabilized the binding of 8-Cl-ADP to ATP synthase.
Although the replacement of C8-hydrogen with chlorine produced a small shift of the binding conformation, it did not dramatically influence the hydrogen bonding interactions. In particular, with residue Lys162 the distance between the β-phosphate and ε-amino group only changed about 0.01 Å (Figure 4B). There is less than 0.4Å change of the distance between the oxygen atoms of 8-Cl-ADP β-phosphate group and Lys162/Arg373, placing the interactions within hydrogen bonding distance. Other than those, the diphosphate group bridging oxygen (–P-O-P-) is about 2.93Å away from one of the guanidinium nitrogen atoms of Arg373, which is predicted to be involved in hydrogen bonding interactions. Interestingly, perhaps compensated by the distance change of the other two hydrogen bonds, this additional hydrogen bond does not lead to large binding affinity change (Figure 4B and Figure 6C), indicating that ADP and 8-Cl-ADP could be competitively bound to ATP synthase F1 domain.
Based on the similarity between ATP and 8-Cl-ATP, we hypothesized that 8-Cl-ATP can bind to the tight binding site (αdpβdp) and loose binding conformation (αtpβtp), therefore acting as an inhibitor for the function of the enzyme. Cellular accumulation of ATP was affected in cells that were loaded with 8-Cl-ATP (Figure 5C) without much effect on AMP and ADP formation. These whole cell data suggest that 8-Cl-Ado metabolites may decrease conversion of intracellular ADP to ATP.
Docking of 8-Cl-ATP to ATP synthase demonstrated that 8-Cl-ATP occupied a very similar binding mode as ANP, with the exception of a relatively large shift in the sugar ring of 8-Cl-ATP. This conformational deviation did not generate geometrical clashes, and thus 8-Cl-ATP fit in the pocket complementarily (Figure 6A). The large shift might be the overall effect of the replacement of 8-position hydrogen with chlorine as well as the substitution of nitrogen in -P-N-P- with oxygen. Analysis of the hydrogen bonding demonstrated a few changes (Figure 6C), however, the binding affinity did not change dramatically suggesting that like ANP, 8-Cl-ATP may inhibit ATP synthase.
Recently Walker and co-workers reported that F1-ATPase was inhibited by ADP-BeF3 −, which is also an ATP analogue . Interestingly, when 8-Cl-ATP was docked into the tight binding site (αdpβdp) of 2CK3, it was observed to form tighter interactions with the enzyme. ATP synthase was inhibited by its co-crystallized ligand ADP-BeF3 − via mimicking ATP structure, and residue α-Arg373 could sense the presence/absence of the γ-phosphate of the ligand . The docked 8-Cl-ATP occupies very similar position and orientation of ADP-BeF3 − (Figure 6B). However, 8-Cl-ATP seems closer to the catalytically essential residue [23, 24] (α-Arg373, β-Arg189, β-Lys162 and β-Glu188) in the docked complex than ADP-BeF3 − in the crystal structure, and thus may lead to strong interactions with ATP synthase (Figure 6B). These observations provide an explanation why the endogenous ATP was dramatically decreased with the increase of 8-Cl-ATP.
Based on our calculations, the binding affinities of ADP/ATP and 8-Cl-ADP/8-Cl-ATP to the enzyme were similar (Figure 6C). This result suggested that 8-Cl-ADP and 8-Cl-ATP might be the competitors of ADP and ATP, respectively. This supports our experimental findings that the endogenous ATP pool usually declined 40–50% several hours after 8-Cl-Ado was added to the system. With results from GOLD, similar conclusions were obtained, although the fitness scores of 8-Cl-ADP/8-Cl-ATP were lower than their correspondent natural substrates, which can be due to the different implementation of the two docking packages.
Additionally as expected, the calculated binding energy is much higher when 8-Cl-ATP was docked into the tight binding site of ATP synthase F1 domain, comparable to the original bound ligand ADP-BeF3 −, which was found inhibiting the enzyme. Previously identified inactivators of F1 ATPase include covalent inhibitors such as 4-Cl-7-nitrobenzofurazan and dicyclohexylcarbodiimide [51, 52], and non-covalent inhibitors such as azide , IF1 , efrapeptin , polyphenolic phytochemicals  including resveratrol . The efrapeptin prevents the open-form of beta subunit to close; preventing substrate binding to the catalytic site , whereas aurovertin B, appears to prevent closure of the catalytic site . Oligomycin, another antibiotic, affects the FO, membrane intrinsic subunit of the enzyme. In contrast to these inactivators, 8-Cl-Ado appears to serve as both alternate substrate (as diphosphate) and inhibitor (as triphosphate), making it a unique inhibitor of the enzyme. It is interesting to note that treatment with a related analog, N6-furfuryladenosine, also results in marked ATP depletion in various cancer cell lines, however it is unknown at the present time whether its action is via ATP synthase inhibition [59, 60].
Our studies presented here demonstrate that predictions from computational molecular modeling can be further validated by experimental methods in cellular systems, which provide a unique mechanism of action for 8-Cl-Ado. Our data shed light on the complex mechanism by which 8-Cl-ATP accumulation results in decline in cellular bioenergy. Cancer cells rely highly on cellular bioenergy metabolism and 8-Cl-Ado is in the clinic as an anticancer agent, therefore, our current results provide insight into the mechanism of action for this novel ATP analogue.
This work is supported in part by grant CA85915 from the National Cancer Institute, Department of Health and Human Services, a Translational Research Award from Leukemia and Lymphoma Society of America, and a University of Texas M. D. Anderson Cancer Center Start-up Fund to SZ.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.