Learn more: PMC Disclaimer | PMC Copyright Notice
Trinitrophenyl derivatives bind differently from parent adenine nucleotides to Ca2+-ATPase in the absence of Ca2+
Associated Data
Abstract
Trinitrophenyl derivatives of adenine nucleotides are widely used for probing ATP-binding sites. Here we describe crystal structures of Ca2+-ATPase, a representative P-type ATPase, in the absence of Ca2+ with bound ATP, trinitrophenyl-ATP, -ADP, and -AMP at better than 2.4-Å resolution, stabilized with thapsigargin, a potent inhibitor. These crystal structures show that the binding mode of the trinitrophenyl derivatives is distinctly different from the parent adenine nucleotides. The adenine binding pocket in the nucleotide binding domain of Ca2+-ATPase is now occupied by the trinitrophenyl group, and the side chains of two arginines sandwich the adenine ring, accounting for the much higher affinities of the trinitrophenyl derivatives. Trinitrophenyl nucleotides exhibit a pronounced fluorescence in the E2P ground state but not in the other E2 states. Crystal structures of the E2P and E2 ∼ P analogues of Ca2+-ATPase with bound trinitrophenyl-AMP show that different arrangements of the three cytoplasmic domains alter the orientation and water accessibility of the trinitrophenyl group, explaining the origin of “superfluorescence.” Thus, the crystal structures demonstrate that ATP and its derivatives are highly adaptable to a wide range of site topologies stabilized by a variety of interactions.
Trinitrophenyl (TNP)-nucleotides (1) are often used for probing the structure of ATP-binding sites and conformational changes arising from nucleotide binding (2, 3), and for measuring the affinity of ATP by competition experiments (2, 4). It is a preferred ATP analogue for photochemical crosslinking with azide derivatives (5). These applications utilize the enhancement of fluorescence or absorption of visible light of the TNP group upon binding to a protein (6). Because of its sensitivity, competition with ATP/ADP has been a valuable means for examining mutational effects on nucleotide affinity (7). Thus, TNP nucleotides have been widely used with F1 (8), myosin (1), and P-type ATPases (2–5, 7, 9, 10), among others.
Nonetheless, whether TNP derivatives are good mimics of authentic adenine nucleotides (AxPs) may be questionable. In several proteins TNP nucleotides have much higher affinities than the genuine AxPs. For instance, TNP-ATP is a high affinity (nM) antagonist of P2X receptors, which have IC50 for ATP (or AMPPCP) in the μM range (11). The affinity is at least one order of magnitude higher in the E2 states of Ca2+-ATPase (3, 12) and Na+, K+-ATPase (4), representative P-type ATPases. Furthermore, TNP-AMP binds to Ca2+-ATPase similarly to or even more strongly than TNP-ATP (12), in marked contrast to AxPs. Thus, a substantially different binding mode of TNP derivatives is suggested. Although more than 20 entries are registered in the Protein Data Bank (PDB) for Ca2+-ATPase (reviewed in ref. 13), no structure with a bound TNP nucleotide exists. In fact, only three crystal structures have been published with bound TNP nucleotides. They are a bacterial histidine kinase CheA (14), an ABC transporter HlyB (15), and an adenylyl cyclase (16), in which TNP nucleotides bind to the proteins in a somewhat distorted manner but otherwise similarly to AxPs (or their close analogues).
Here we report crystal structures of Ca2+-ATPase from skeletal muscle sarcoplasmic reticulum (SERCA1a) with bound TNP-ATP, -ADP, -AMP, as well as ATP and ADP in the absence of Ca2+ (E2-state) and stabilized with thapsigargin (TG), a very potent inhibitor (17). Ca2+-ATPase consists of three cytoplasmic domains designated A (actuator), N (nucleotide binding), and P (phosphorylation), and 10 transmembrane helices (18). Previous crystal structures in the presence of Ca2+ have shown that ATP (nonhydrolysable analogue adenosine 5′-(β,γ-methylene) triphosphate (AMPPCP) (19, 20) and adenosine 5′-(β,γ-imido)triphosphate (21) were used in place of ATP) binds near the hinge between the N and P domains and cross-links them. The triphosphate chain is extended with no Mg2+ bound and the adenine ring stacks with Phe487 in the N domain. In the absence of Ca2+, AMPPCP takes a folded conformation due to bound Mg2+ (22, 23). Here we show that the binding mode of TNP-AxPs to the ATPase in the E2-state is distinctly different, with the TNP ring stacking with Phe487 and the adenine ring sandwiched between two arginines.
TNP-nucleotides exhibit greatly enhanced fluorescence (“superfluorescence”) (24) in certain reaction intermediates of Ca2+-ATPase. Superfluorescence is characteristic of the E2P ground state and not observed with the transition state analogue (25). Hence TNP-AMP bound structures were also determined for 
(E2P ground state analogue) and 
(E2 ∼ P transition state analogue), to explore the origin of the superfluorescence. The crystal structures show that the orientation of the TNP ring and the water accessibility vary with the arrangement of the three cytoplasmic domains, thereby explaining the different levels of fluorescence in different reaction intermediates.
Results
Binding Mode of ATP and ADP in E2(TG).
Crystals of Ca2+-ATPase in the E2(TG), 
and 
forms were generated in the presence of TG by microdialysis. Nucleotides were introduced by placing the dialysis buttons in buffers containing nucleotides. Because the overall structural changes were small, molecular replacement based on the atomic model built for E2(TG + BHQ) (PDB ID code 2AGV) (26) went smoothly (Table S1).
The crystal structure with bound ATP in the absence of Ca2+ stabilized with TG [abbreviated as E2·ATP(TG)] was determined at 2.15-Å resolution (Fig. 1A; Fig. S1 for stereo) and is virtually identical to that with bound AMPPCP described previously (PDB ID code 2DQS) (22). The adenine ring of ATP is inserted into the binding cavity in the N domain of Ca2+-ATPase, flanked by Phe487 on one side and Leu562 on the other. This binding mode is the same, but the stacking of the adenine ring with the Phe487 side chain is not as optimal as in the E1·AMPPCP crystal structure, now refined to 2.5-Å resolution (Fig. 1B). The orientation is ∼25° offset from parallel, primarily due to a different orientation of the Phe487 main chain. The binding cavity is locally distorted (Fig. 2), due to salt bridges formed between Arg489 (N domain) and Asp203 (A domain) and between Glu486 (N) and His190 (A).
Binding mode of ATP (or AMPPCP) and TNP-ATP in Ca2+-ATPase. (A) ATP in the absence of Ca2+ but in the presence of thapsigargin [E2·ATP(TG)], pH 6.1, at 2.15-Å resolution; (B) AMPPCP in the presence of Ca2+ (E1·AMPPCP) at 2.5-Å resolution; (C) TNP-ATP in the absence of Ca2+ but in the presence of thapsigargin [E2·TNP-ATP(TG)] at 2.15-Å resolution. The violet net in C represents an |Fobs|-|Fcalc| electron density map (contoured at 3σ), before introducing TNP-ATP into the atomic model. The N domain appears green and the P domain yellow. Small red spheres represent water molecules; cyan spheres, Ca2+; green spheres, Mg2+. Small cyan disks represent main chain amide. Broken lines in blue show hydrogen bonds and metal coordination. A stereo version of this figure is presented as Fig. S1.
Local distortion of the adenine binding site of Ca2+-ATPase in the absence of Ca2+. Superimposition of the atomic models for E2·ATP(TG) (atom color) and E1·AMPPCP (green for the N domain, pink for the A domain; AMPPCP, orange). Broken lines in blue show hydrogen bonds and coordinations of Mg2+.
In the E1·AMPPCP crystal structure, AMPPCP takes an extended conformation with no Mg2+ bound (Fig. 1B) (19, 20). In E2·ATP(TG), if Mg2+ is present, the triphosphate chain takes a folded conformation and coordinates Mg2+ with all three phosphate groups (Fig. 1A). This position of Mg2+ was confirmed by substituting Mg2+ with Mn2+ and calculating an anomalous difference map for Mn2+ and an |Fobs|-|Fobs| difference Fourier map (i.e., |Fobs[E2·ATP(Mn2+)]|-|Fobs[E2·ATP(Mg2+)]|) (Fig. S2). This Mg2+ is also coordinated by three water molecules, two of which are fixed by hydrogen bonds with Glu439 (Fig. 1A). The carboxyl group of Glu439 itself is more than 3 Å away from the Mg2+. This coordination geometry is different from that reported by Jensen et al. (23) for the AMPPCP-bound structure (PDB ID codes 2C88 and 2C8K) at 2.8 to 3.1-Å resolution. The γ-phosphate points away from Asp351, the residue phosphorylated during the catalytic cycle, due to the position of Mg2+ but is stabilized by a water molecule hydrogen bonded to Gln202 in the A domain. Thus, Mg2+ will enhance the binding of ATP to E2(TG) (Kd = 150 μM without Mg2+ and 100 μM with Mg2+) (23). Nonetheless, the affinity is still much lower than that in the presence of Ca2+ (3 μM) (27).
The difference maps also located another Mg2+ in the P domain. It is coordinated by Asp703 and Asp707 (Fig. 1A), instead of Asp351 as in E1·AMPPCP and 
. [However, the metal ion in the E1·AMPPCP crystal structure is most likely to be Ca2+ rather than Mg2+ (22, 28).] The affinity of this Mg2+ appears low because the coordination geometry is far from an ideal octahedral one.
The lower affinity for ATP of E2(TG) reflects the lack of direct interactions with residues in the P domain. In the E1·AMPPCP crystal structure, AMPPCP crosslinks the N and P domains by forming many hydrogen bonds (Fig. 1B): They include those between ribose and Arg678 (P), α-phosphate and Arg489 (N), β-phosphate and Asp627 (P)/Arg560 (N), and γ-phosphate and Thr353 (P)/Gly626 (P) (19, 20). In E2·ATP(TG), most of these salt bridges are absent (Fig. 1A) because the three cytoplasmic domains are arranged differently. Instead, the triphosphate chain of ATP is stabilized by Mg2+-coordination. Consequently, the folded form, with the γ-phosphate turned away from Asp351, is preferred (Fig. 1A). Then, when the cytoplasmic headpiece opens in transition to E1, ATP binding affinity rises because local distortion is relieved, in preparation for efficient phosphoryl transfer. If the energy cost of distortion is large, relief of distortion (i.e., opening of the headpiece) alone will stabilize the system and drive the reaction cycle forward.
The crystal structure of the ADP-bound form was determined at 2.3-Å resolution. Binding geometry of ADP is essentially the same as ATP, but the position of the β-phosphate is different because the coordination of Mg2+ is possible only with the α- and β-phosphates. Here the Glu439 side chain is fully extended (Fig. S3).
Binding of TNP-ATP in E2(TG).
TNP nucleotides cannot bind to the ATPase in the same mode as ATP, because the TNP ring would collide with Pro518, Ala517, and Cys561 (Fig. S4). Instead, the Phe487 side chain stacks with the TNP ring in an ideal geometry (Fig. 1C). Moreover, the adenine ring is now sandwiched by the long side chains of Arg489 (on the N domain) and Arg678 (P domain), both in fully extended conformations. One of the three NO2-groups on the TNP ring forms a hydrogen bond with the Lys515 side chain. No residues form direct hydrogen bonds with the adenine ring, but the Glu680 carboxyl and the Val679 carbonyl fix a water molecule that is hydrogen bonded to N6 of the adenine ring (Fig. 1C).
The β-phosphate is stabilized by the Thr353 hydroxyl and Gly626 amide, in a way very similar to the γ-phosphate in the E1·AMPPCP crystal structure (Fig. 1B). Because of this, the triphosphate chain is extended and takes a zigzag conformation. The γ-phosphate is free with no Mg2+ bound, even though excess Mg2+ was present in the crystallization medium. It comes close to Asp351, even closer than in E1·AMPPCP, and a water molecule bridges them (small arrow in Fig. 1C).
Thus, TNP-ATP bridges the N and P domains, similar to AMPPCP in the E1·AMPPCP crystal structure. Likewise Arg560 as in E1·AMPPCP bridges Thr441 (N) and Asp627 (P), further stabilizing N and P domain interactions (Fig. 1C). In E2·ATP, Arg560 bridges only Thr441 and the β-phosphate (Fig. 1A).
Binding of TNP-ADP and TNP-AMP in E2(TG).
The binding mode of TNP-ADP in E2(TG) is essentially the same as that of the triphosphate form (Fig. S3), and the β-phosphate occupies the same position as the γ-phosphate in E1·AMPPCP, suggesting that TNP-ADP might be hydrolyzed even in the absence of Ca2+.
The binding mode of TNP-AMP in E2(TG) (Fig. 3A) is somewhat different from that of TNP-ATP in that the α-phosphate group of TNP-AMP, stabilized by the Thr353 hydroxyl and the Gly626 amide, occupies the position of the β-phosphate of TNP-ATP/ADP. The positions of the TNP group and the adenine ring are slightly shifted so that the whole molecule inclines toward Thr353, which forms a hydrogen bond with the α-phosphate. This orientation is stabilized by a hydrogen bond between Glu439 and a NO2 group in the TNP ring, in addition to the one between the Lys515 and another NO2 group.
Binding of TNP-AMP in various intermediates (or analogues) of Ca2+-ATPase in the absence of Ca2+. (A) E2(TG), (B) 
, an E2 ∼ P transition state analogue, and (C) 
, an E2P ground state analogue. The A, N, and P domains are colored pink, green, and yellow, respectively. Blue broken lines show hydrogen bonds. Small red spheres represent water molecules; small cyan disks, main chain amide. Carbon atoms in the TNP group are colored gray.
Binding of TNP-AMP in E2P and E2 ∼ P Analogues.
TNP-AMP has been used for probing structural changes around the nucleotide binding site of Ca2+-ATPase. There is particular interest in the origin of the superfluorescence of TNP-AMP in the E2P ground state. 
is a stable analogue of the E2P ground state, and superfluorescence is observed even in the presence of TG (25). This is in marked contrast to 
, a stable transition state analogue, in which TNP-AMP exhibits little fluorescence (25). Hence we soaked TNP-AMP into 
and 
crystals. These crystals diffracted to at least 2.6-Å resolution, which allowed us to visualize tightly bound water molecules as well as Mg2+.
The orientation of the TNP ring in 
is similar to that in E2(TG), although Arg678 no longer sandwiches the adenine ring, due to a different position of the P-domain (Fig. 3B). However, the orientation in 
is distinctly different, being 60° offset from parallel (Fig. 3C and Fig. S5). This appears to be an unfavorable orientation, but probably not as much as one would think, as the hydrogen bond with Lys515 remains and, furthermore, Thr441 and Arg560 form three unique hydrogen bonds with another NO2 group. Even Arg174 in the A-domain forms a hydrogen bond with the ribose.
The environment of the TNP ring varies substantially and this explains the origin of superfluorescence. In 
, there is a tight pocket around the TNP ring, excluding water molecules (Fig. 4C). TNP-AMP itself is inserted somewhat more deeply than in 
. Here, Arg174 on the A domain, backed by Ile188 also on the A domain, lines a part of the upper surface of the binding pocket. In 
and E2(TG) (Fig. 4A and B), there is a relatively large open space on the opposite side of the TNP ring to Phe487, allowing access of water molecules (Fig. 2 and and4).4). Thus, a very hydrophobic environment is realized only in 
, consistent with superfluorescence in the E2P state.
(B), and 
(C). Prepared with PyMol (The orientation of the TNP ring appears to be determined by the position of Arg174 on the A domain. During the transition from E1P → E2P → E2·Pi, the A domain rotates approximately 110° around an axis approximately perpendicular to the membrane (29). As a result, the position of Arg174, located on the A–N interface, changes drastically. In the 
transition, the A domain rotates 20° (30) and Arg174 moves approximately 2.3 Å at the Cα atom. This seemingly small change is enough to alter the hydrogen bonding partner of Arg174. In the 
crystal structure, Arg174 NH2 forms both a hydrogen bond with the ribose and cation–π interactions with Phe487 and the TNP ring of TNP-AMP (Fig. 3C). When TNP-AMP is absent, the Arg174 side chain occupies a different position, being fixed by hydrogen bonds with the Ser186 hydroxyl and the main chain carbonyl (29). In 
, Arg174 forms hydrogen bonds with one of the NO2 groups on the TNP ring, fixing the TNP ring in a different orientation (Fig. 3B).
The position of the phosphate group also varies (Fig. 3). In E2(TG), the phosphate group is stabilized by the Thr353 hydroxyl and the Gly626 amide (Fig. 3A). In contrast, in 
, it is stabilized by a hydrogen bond with the Cys561 amide (Fig. 3C), pointing away from Asp351, the phosphorylated residue. In none of the crystal structures is Mg2+ bound to TNP-AMP.
Discussion
Crystal Structures of Other Proteins with Bound TNP Nucleotides.
The PDB has only three structures with bound TNP nucleotides: They are CheA with TNP-ATP (PDB ID code 1I5D) (14); HlyB (an ABC transporter) with TNP-ADP (PDB ID code 2PMK) (15); and adenylyl cyclase with TNP-ATP (PDB ID code 2GVD) (16). In all of them, TNP derivatives bind to the proteins with somewhat distorted but similar modes to the original AxPs. TNP-ATP binds with a > 500× higher affinity to CheA, > 50× to HlyB, and > 1,000× to adenylyl cyclase. TNP-ATP binds to CheA with the TNP ring inserted into a hydrophobic pocket adjacent to the adenine binding site, thereby conferring a higher affinity (14). In adenylyl cyclase the adenosine moiety is flipped so that the TNP ring is stabilized by several hydrogen bonds (16, 31). The HlyB crystal structure with bound TNP-ADP shows virtually an identical binding mode (15). Thus, the binding mode of TNP-AxPs described here is distinctly different from those in previous reports.
Origin of the High Affinity Binding of TNP Nucleotides to Ca2+-ATPase.
TNP modification confers a much higher (> 50 times) affinity to adenine nucleotides for Ca2+-ATPase. Reported Kd for TNP-ATP is 0.1–0.2 μM (3, 12) and that for ATP is around 10 μM [Kd = 10.1 μM (12); 5–9 μM (22); 4 μM (3)] in the absence of Ca2+. This higher affinity arises primarily because both adenine and TNP moieties are stabilized by hydrophobic interactions with Phe487, Arg489, and Arg678 (Fig. 1C). Stabilization of the phosphate chain by hydrogen bonds with the Thr353 hydroxyl and the Gly626 amide may also contribute. The β-phosphate is stabilized in TNP-ATP and -ADP, whereas the α-phosphate is stabilized in TNP-AMP (Fig. 3A). It is evident that TNP-ATP would have a lower affinity, because the γ-phosphate in TNP-ATP is located fairly close to the cluster of aspartates (i.e., Asp351, Asp703, and Asp707) around the phosphorylation site (Fig. 1C). However, it is obscure why the affinity of TNP-AMP is even higher than that of TNP-ADP. In fact, Seebregts and McIntosh (5) reported that TNP-8N3-AMP has a Kd of 0.1 μM and TNP-8N3-ADP/ATP have 0.2 μM Kd. Suzuki et al. (12) reported Kds of 7.62 nM and 156 nM for TNP-AMP and -ATP, respectively. Unfortunately, as far as we know, there is no literature that reports on the affinity of all three under the same conditions.
The observation that TNP-AMP binds to 
and 
with high affinity (32) suggests that the primary origin of the high affinity of TNP-AxPs is hydrophobic stabilization of the TNP ring. In these two complexes, there is no direct stabilization of the adenine ring (Fig. 3 B and C). Yet the surface of the binding pocket of the TNP ring is complementary, in particular, in 
(Fig. 4C).
Explanation of Photolabeling Results.
McIntosh reported that TNP-8N3-AMP and -ATP label Lys492 in Ca2+-ATPase (33). Inesi reported that TNP-2N3-AMP is much more efficient than the 8N3 derivative in labeling Lys492 (34). For a TNP-8N3-AxP to be able to label Lys492, the adenine ring has to flip with respect to the ribose (Fig. 1C), which will cause a considerable steric problem. In contrast, it is evident that photoaffinity labeling of Lys492 is impossible in the canonical binding mode of AxP, as an azido group attached at the 8 position, and Lys492 cannot come closer than 6 Å. Thus these reports completely agree with the crystal structure presented here.
Caviers and colleagues used TNP-8N3-ADP photoaffinity labeling to identify ATP binding residues in Na+, K+-ATPase (35). With native Na+, K+-ATPase, labeling occurred on the N-domain residue Lys480, which is equivalent to Lys492 in Ca2+-ATPase, indicating that TNP-ADP binds to Na+, K+-ATPase similarly to Ca2+-ATPase. With the FITC-modified enzyme, labeling occurred in a highly conserved region in the P domain (Ala714-Lys728 in Ca2+-ATPase) just prior to the cytoplasmic end of the M5 helix (35). In the crystal structures, there was no electron density suggesting a nucleotide in this position, although the concentration of TNP-ADP was as high as 5.6 mM.
Fluorescence of TNP-AMP in 
and 
.
Nakamoto and Inesi reported that TNP nucleotides fluoresce strongly in E1P (36). Suzuki and colleagues have shown that fluorescence from TNP-AMP in 
is low whereas that in 
very high and proposed that the A domain in 
is located between those seen in 
and 
(32). The position of TNP-AMP in either E2(TG) or 
would be incompatible with its binding in E1P, if the structure around the phosphorylation site were the same as in the 
crystal structure (19, 20), because the adenine ring would collide with the P domain. The adenine binding site is fairly open to bulk solvent in the crystal structure, and the fluorescence of TNP-AMP will be correspondingly weak in 
. Even though the side chain of Arg174 on the A domain is fairly free to move, the A domain should be located within a rather restricted range for a hydrogen bond to be formed with the ribose. In fact, addition of TNP-AMP to 
slowly converts it into E1·2Ca2+, judging from proteolysis patterns (32). Thus, our crystal structures corroborate the idea that the position of the A domain and the structure of the phosphorylation site in E1P are different from those in the 
crystal structure (32).
Conclusion
The crystal structures described here demonstrate that AxPs and TNP-AxPs are highly adaptable and can bind to an ATP binding site in widely different modes, stabilized by π–π interaction, amine-aromatic sandwiching, and hydrogen bonding. Mg2+ may also alter the orientation of the whole nucleotide through its coordination of the polyphosphate chain and protein residues. These various interactions provide means for adjusting the affinity for ATP in different steps of a reaction cycle. The binding modes of TNP-AxPs shown here, with the adenine ring stabilized in distinctly different ways, suggest that authentic AxPs may bind similarly, although with lower affinity. A further corollary of this study is that derivatives or inhibitors specific to a particular protein may be designed by considering the molecular topology of an area substantially larger than the canonical adenosine binding site.
Methods
Crystallization.
Crystals of Ca2+-ATPase were prepared from affinity purified enzyme by microdialysis. E2(TG) (26), 
(30), and 
(30) crystals were prepared as before. Nucleotides were introduced into crystals by placing dialysis buttons used for crystallization overnight into the flash freezing buffer that contained one of the following: 10 mM ATP, 10 mM ADP, 10 mM TNP-ATP, 5.6 mM TNP-ADP, or 2.5 mM TNP-AMP. Mg2+ concentration was 1 mM higher than that of nucleotide.
Data Collection and Structure Determination.
Crystals were picked with nylon loops in a cold room and flash-frozen in cold nitrogen gas. 
crystals were left to stand for 2.5 min for dehydration after removing excess buffer (30). Diffraction data were collected from crystals cooled to 100 K at BL41XU of SPring-8 with an ADSC Q315 CCD detector, and processed with Denzo and Scalepack (37). All the E2-nucleotide structures were determined by molecular replacement using CNS (38) starting from the models published previously without nucleotides (26, 30). The atomic model of E1·AMPPCP was refined with REFMAC including TLS refinement (39). Statistics of the diffraction data and refinement are given in Tables S1 and S2.
Acknowledgments.
We thank Drs. N. Shimizu (Japan Synchroton Radiation Research Institute) and T. Tsuda (now at Gakushuin University) for data collection at BL41XU of SPring-8, and Y. Norimatsu (now at Fujitsu) for structure refinement. We are indebted to Dr. D. B. McIntosh for his help in improving the manuscript. This work is a part of an ongoing long-term project (2009B0025) at SPring-8 and was supported in part by a Specially Promoted Project Grant (to C. T.) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.
Footnotes
The authors declare no conflict of interest.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1017659108/-/DCSupplemental.
Data deposition: Coordinates and structural factors have been deposited in the Protein Data Bank, www.pdb.org (PDB ID codes 3AR4 [E2·ATP(TG)], 3AR3 [E2·ADP(TG)], 3AR7 [E2·TNP - ATP(TG)], 3AR6 [E2·TNP - ADP(TG)], 3AR5 [E2·TNP - AMP(TG)], 3AR9 [
], 3AR8 [
], and 3AR2 (E1·AMPPCP).