Identification of cytoskeletal elements enclosing the ATP pools that fuel human red blood cell membrane cation pumps

Abstract

The type of metabolic compartmentalization that occurs in red blood cells differs from the types that exist in most eukaryotic cells, such as intracellular organelles. In red blood cells (ghosts), ATP is sequestered within the cytoskeletal–membrane complex. These pools of ATP are known to directly fuel both the Na⁺/K⁺ and Ca²⁺ pumps. ATP can be entrapped within these pools either by incubation with bulk ATP or by operation of the phosphoglycerate kinase and pyruvate kinase reactions to enzymatically generate ATP. When the pool is filled with nascent ATP, metabolic labeling of the Na⁺/K⁺ or Ca²⁺ pump phosphoproteins (E_Na-P and E_Ca-P, respectively) from bulk [γ-³²P]-ATP is prevented until the pool is emptied by various means. Importantly, the pool also can be filled with the fluorescent ATP analog trinitrophenol ATP, as well as with a photoactivatable ATP analog, 8-azido-ATP (N₃-ATP). Using the fluorescent ATP, we show that ATP accumulates and then disappears from the membrane as the ATP pools are filled and subsequently emptied, respectively. By loading N₃-ATP into the membrane pool, we demonstrate that membrane proteins that contribute to the pool’s architecture can be photolabeled. With the aid of an antibody to N₃-ATP, we identify these labeled proteins by immunoblotting and characterize their derived peptides by mass spectrometry. These analyses show that the specific peptides that corral the entrapped ATP derive from sequences within β-spectrin, ankyrin, band 3, and GAPDH.

The advent of techniques to isolate and characterize intracellular organelles has made it possible to establish that metabolic pathways and their associated intermediates are commonly localized to membrane-enclosed compartments (1, 2). Although mammalian red blood cells (RBCs) have no intracellular membranes or organelles, previous work has documented that the ATP that fuels the Na⁺/K⁺ (ATP1A1) and Ca²⁺ (ATP2B1) pumps resides within a structurally distinct compartment that constitutes the preferred source of ATP for both types of pumps (3–6). This membrane-associated ATP pool can be filled with exogenous ATP in either hemoglobin-free porous ghosts or inside-outside membrane vesicles (IOVs) in ways that allow control of the contents of the ATP pool. One method of filling this pool is to incubate the ghosts or IOVs with bulk ATP; another is to run reactions with the membrane-bound ATP synthesizing enzymes phosphoglycerate kinase (PGK) and pyruvate kinase (PK) in the forward direction (7–9). In all of these methods, the ghosts and IOVs are washed free of loading substrates, leaving only the ATP entrapped within the membrane-associated pools to energize the pumps. The ATP thus sequestered is unavailable for reaction with added hexokinase plus glucose (8, 9). Moreover, not only do the Na⁺ and Ca²⁺ pumps use the same pools of ATP, but the pools can be emptied either by running either pump forward or running the PGK reaction backward (7, 9). Pool ATP alone also has been shown to support a variety of Na⁺ and Ca²⁺ pump functions once thought to be energized by bulk ATP (9).

The hallmark test of whether or not ATP is in the confined pool is to evaluate the ability of either pump to form its respective phosphointermediate (E-P) on utilization of [γ-³²P]-ATP in the presence of either Na⁺ or Ca²⁺ (7, 9, 10). If nonradioactive ATP is first entrapped within the pool, then neither of the pumps’ ³²P-labeled E-P is seen until the unlabeled pools of ATP are emptied. In the case of IOVs, instead of E-Ps, ATP pool-dependent ²²Na⁺ or ⁴⁵Ca²⁺ uptake (transport) by their respective pumps can be measured (8, 9).

In recent studies reported in intact RBCs, the changes in nucleotide metabolism (i.e., concentrations of ATP, ADP, and AMP) were measured under conditions of marked stimulation of the Ca²⁺ pump (11). The patterns of changes in nucleotide concentrations were consistent with the requirement for nucleotide sequestration within the cytoskeletal–membrane complex, distinct from the nucleotides contained within the cytoplasm. Thus, these results complement the foregoing evidence for membrane pools of ATP. Importantly, that study was the first to identify ATP pools in intact cells, where previously their presence could only be inferred (12).

The present study not only confirms, by the use of different methods, the primary hallmarks for characterizing the membrane pool of ATP in ghosts as mentioned above, but also extends the analysis by identifying some of the cytoskeletal components that confine the pooled ATP. One method used to label the latter components involved an antibody raised against 8-azido-ATP (N₃-ATP), a photoactivatable ATP analog (13). N₃-ATP (obtained from Affinity Photoprobes) was entrapped within the ATP pool in the dark. On exposure to UV irradiation, the N₃-ATP–labeled proteins were detected on Western blot analysis after solubilization and gel electrophoresis. Another method involved the use of 8-azido-[α-³²P]-ATP with subsequent identification of the labeled proteins by phosphoimaging. In other experiments, N₃-ATP–labeled pool proteins were identified by mass spectrometry (MS). Using these methodologies, we have unequivocally identified β-spectrin, ankyrin, glyceraldehyde-3-phosphate dehydrogenase, and band 3 as components that corral the ATP pools present in the cytoskeletal–membrane complex in RBC ghosts.

Results

Effects of Pool ATP on Pulse-Labeling of Na⁺ and Ca²⁺ Pump E-Ps.

Central to the experiments described below, we felt it necessary to repeat the basic characteristics of loading and emptying the ghost pool of ATP as described above, because that work had been carried out in a different laboratory with different analytical methods. Hemoglobin-free porous (frozen-thawed) RBC ghosts were prepared from freshly drawn blood as described previously (9). These ghosts are permeable to all added constituents. Of essential importance, after all pool manipulations and incubations, the ghosts were thoroughly washed to remove all nonpool solutes (including ATP and its derivatives) from inside and outside of the ghosts. Assay for pulse-labeling of the E_Na-P and E_Ca-P were carried out by incubation at 0 °C with ∼2 μM [γ-³²P]-ATP for 20 s. All reactions were stopped by exposure to trichloroacetic acid (TCA) (Fig. 1). The precipitates were washed, solubilized in SDS, and then separated by SDS/PAGE electrophoresis. The relative radioactivity of the separated bands was assessed by phosphoimaging (Cyclone Plus Storage System; PerkinElmer). This method differs from previously used methods (9) in which total TCA precipitates were counted, thus including interference from other possible labeled constituents. The phosphoimages give precise resolution of the E-Ps of the Na⁺ and Ca²⁺ pumps (Fig. 1 A and B). Thus, it is clear that when the membrane pools contain unlabeled ATP, pulse-labeling of both the E_Na-P and E_Ca-P with [γ-³²P]-ATP is inhibited until the ATP pools have been emptied. The results presented in Fig. 1B confirm that the Na⁺ and Ca²⁺ pumps use the same pools of ATP (9).

Fig. 1.

Effect of loading the ATP pool in ghosts with nonradioactive ATP on the ability to detect [γ-³²P]-labeled E_Na-P and E_Ca-P by phosphoimaging. The ATP pool in ghosts was filled with nonradioactive ATP and then either emptied or left filled, as indicated. After addition of [γ-³²P]-ATP and the desired cations, the E_Na-P and E_Ca-P were detected by SDS/PAGE, followed by phosphoimaging. The relative intensities of these bands reflect the ability to label the E_Na-P and E_Ca-P with a 20-s pulse of 2 μM [γ-³²P]-ATP. (A) Lanes 1 and 2 show labeling of the E_Mg-P and E_Na-P under conditions with all ATP in the pools removed by previous washing. Lanes 3 and 4 show the same experiment as in lanes 1 and 2, except with the membrane pools filled with unlabeled ATP before the addition of [γ-³²P]-ATP. Lanes 5 and 6 and lanes 7 and 8 show the E-P values when the pools of ATP have been depleted either by running the Na⁺/K⁺ pump forward with Na⁺ and K⁺ (lanes 5 and 6) or by running the PGK reaction backward (lanes 7 and 8). Note that when the ATP pools are filled with unlabeled ATP (lanes 3 and 4), E_Na-P and E_Mg-P are of similar intensity; however, if the ATP pools are emptied, thereby allowing [γ³²P]-ATP to label the pumps, then the relative intensity of the band representing the E_Na-P is greater than its Mg²⁺ counterpart. Quantitation ratios of the intensities of the bands for each pair of lanes were 2.9 for lanes 1 and 2, 0.88 for lanes 3 and 4, 2.2 for lanes 5 and 6, and 1.6 for lanes 7 and 8, with an SEM of ±3% for n = 2. (B) These lanes show the same band intensity relationships as in A for both the E_Na-P (lower bands) and the E_Ca-P (upper bands), even when both types of E-Ps are labeled in the same incubation. Note that when the pools are filled with ATP (lanes 5–8), labeling of the E-Ps with [γ-³²P]-ATP is inhibited. Moreover, when the ATP pool is emptied by running the Na⁺/K⁺ pump forward (lanes 1–4), the calcium pump also can be labeled with [γ-³²P]-ATP, indicating that the Na⁺/K⁺ and Ca²⁺ pumps share the same ATP pools. The phosphoimages shown here are representative of multiple experiments.

Visualization of ATP Pool Filling and Emptying Using a Fluorescent ATP Analog.

The fluorescent analog used was 2′ (or 3′)-O-(2,4,6-trinitrophenyl) (TNP)-ATP or TNP-ADP. TNP-ATP is not a substrate for the Na⁺ pump (14, 15), but it does act like ATP, inhibiting the pumps by ∼76% when incorporated into the membrane pools (using the same protocol as with ATP). Evidence that TNP-ATP can fill the ATP pools derives from the fact that ghosts incubated with TNP-ATP cannot phosphorylate the Na⁺/K⁺ pump with [γ-³²P]-ATP until the pool is emptied of TNP-ATP. Thus, the relative values of the ³²P-labeled pump E-Ps measured on a 20-s exposure to [γ-³²P]-ATP were 1.05 ± 0.03 pmol ³²P/mg protein after preincubation without ATP, 0.13 ± 0.01 pmol ³²P/mg protein after preincubation with ATP, and 0.35 ± 0.04 pmol ³²P/mg protein after preincubation with TNP-ATP. Moreover, when TNP-ADP was preincubated with the ghosts and converted to TNP-ATP by running the PGK reaction forward, the level of the ³²P-labeled pump phosphoimtermediate was 0.42 ± 0.01 pmol ³²P/mg protein.

Fig. 2A shows a fluorescent image of porous ghosts when the pools are filled with TNP-ATP. The TNP-ATP incorporated into membrane pools has a punctate appearance, localized on the membrane, with the interior of the ghosts empty. When the pool of TNP-ATP is emptied by running the PGK reaction backward (Materials and Methods), the relative fluorescence is markedly decreased (Fig. 2B). Fig. 2 C and D shows control experiments in which the PGK reaction is dependent on its full complement of substrates, including NADH and 3-phosphoglycerate (PGA), to remove the pool of TNP-ATP. In Fig. 2C, one of the substrates (PGA) needed to run the reaction backward is omitted, and in Fig. 2D, both substrates (PGA and NADH) required in the reaction are omitted. In these two controls, membrane-associated fluorescence is similar to that shown in Fig. 2A. These results demonstrate the dynamic nature of the filling and emptying of the ATP pools.

Fig. 2.

Dynamic imaging by confocal microscopy of the loading and emptying of membrane pools of ATP using the fluorescent ATP analog TNP-ATP. ATP pools in porous ghosts were either loaded or loaded and then emptied of TNP-ATP by running the PGK reaction backward. Ghosts were then observed by confocal microscopy. (A) Ghosts with their ATP pools filled with TNP-ATP. Because a large ensemble of ghosts is shown, the plane of focus passes through the center of most ghosts, showing the fluorescent ATP only on the cell periphery where the plane of focus coincides with the membrane. However, in a few ghosts, the plane of focus passes through part of the membrane, revealing pool ATP in the mid region of the cell. The fact that ghosts exhibit only punctate fluorescence in a ring on the cell periphery indicates that the TNP-ATP–loaded pools are localized to the membrane (12). (B) The entrapped TNP-ATP is labile and susceptible to removal by running the PGK reaction backward. (C and D) Controls, with one of the substrates (PGA) needed to run the PGK reaction backward omitted (C) and both substrates (PGA and NADH) omitted (D). In D, the fluorescent images of the loaded ghosts are stable even after continued incubation in the absence of both substrates. These images are similar to ones as studied by Hoffman et al. (12).

Identification of Pool Corral Components by Photolabeling with N₃-ATP.

To identify the membrane proteins that form the compartment containing the entrapped ATP, the ATP pool was filled with N₃-ATP, and photolabeling of pool corral components was initiated by UV illumination. To identify each photolabeled component, an antibody was raised against N₃-ATP, as described in Materials and Methods, and used in immunoblots to stain those proteins photolabeled with N₃-ATP. As shown in Fig. 3, the anti–N₃-ATP antibody readily detected BSA photolyzed in the presence of N₃-ATP (lane 6), but not unlabeled BSA (lane 5) or unlabeled RBC ghosts (lane 7).

Fig. 3.

Identification of ghost proteins labeled with pool-associated N₃-ATP. Here the ghosts were filled or emptied of N₃-ATP, as described in Materials and Methods. After washing, the membranes were photolyzed and separated by SDS/PAGE. Equal amounts of ghost proteins (30 μg) were loaded in all lanes. Lanes 1–7 were transferred to nitrocellulose and immunoblotted with an antibody to N₃-ATP, whereas lane 8 was stained with Coomassie blue. Lane 1 shows pools filled with N₃-ATP; lanes 2 and 3, pools first filled with N₃-ATP and then emptied by either running the Na⁺ pump forward or running PGK reaction backward, respectively. In lane 4, pools are filled with N₃-ATP in the presence of excess unlabeled ATP to block any specific ATP-binding sites. Lanes 5–7 show the specificity of the N₃-ATP antibody, demonstrating that it labels only BSA photolyzed with N₃-ATP (lane 6), not unlabeled BSA (lane 5) or unlabeled RBC ghosts (lane 7). Lane 8 shows a Coomassie blue stain of ghost proteins after separation by SDS/PAGE. The proteins tentatively identified are indicated with lettered lines. Possible protein candidates for the labeled bands are A, β-spectrin and/or ankyrin; B, Ca²⁺ pump (M_r ∼140 kDa) or fragments of spectrin/ankyrin; C, α-subunit of the Na⁺/K⁺ pump (M_r ∼110 kDa); D, band 3; E, actin; and F, GAPDH.

Fig. 3, lane 1, shows Western blots of N₃-ATP–labeled proteins after the pools were filled with N₃-ATP and the entrapped N₃-ATP was photoactivated. As shown in lane 1, proteins that migrate near β-spectrin and/or ankyrin (M_r ∼250 kDa), the Ca²⁺ pump (M_r ∼140 kDa), or fragments of spectrin/ankyrin, the α-subunit of Na⁺/K⁺ pump (M_r ∼110 kDa), band 3, actin, and GAPDH are labeled by this protocol. Lane 2 shows a control experiment in which the pool N₃-ATP was removed before exposure to UV irradiation by running the Na⁺ pump forward in the dark (with Na⁺ and K⁺). Importantly, labeling of all membrane proteins except actin (M_r ∼42 kDa) was prevented by this procedure, suggesting that all of the labeled proteins except possibly actin participate in fencing of the ATP pools. Lane 3 is a similar control to that shown lane 2, but with the pool N₃-ATP emptied by running the PGK reaction backward in the dark (with NADH and PGA) before exposure to UV irradiation. Lane 4 shows a very different control, in which both nonspecific and pool-specific ATP binding sites were blocked by coincubation with a 20-fold excess of unlabeled ATP. In this latter case, photolabeling of the actin band was prevented as well. Inhibition of actin labeling in this latter result was obviously expected, given that actin contains an ATP-binding site that will bind ATP even in the absence of a membrane-associated pool (16, 17). Based on a comparison with the staining pattern of RBC membrane proteins (lane 8), the foregoing data collectively suggest that membrane/cytoskeletal-associated ATP pool components might include β-spectrin, ankyrin, the Ca²⁺ pump, the Na⁺/K⁺ pump, band 3, and GAPDH.

Analysis of ATP-Associated Membrane and Cytoskeletal Proteins Labeled with [α-³²P]-N₃-ATP.

To validate the foregoing antibody staining of ATP pool components photolabeled with N₃-ATP, we turned to analysis of ghost proteins photolabeled with [α-³²P]-N₃-ATP after using, following an analogous protocol, the radiolabeled form of the ATP analog to either fill or empty the ATP pools of ATP. Thus, autoradiographs were analyzed instead of Western blots. Fig. 4, lanes 1–4, shows autoradiographs of the labeled proteins after separation by SDS/PAGE, and lane 5 shows a Coomassie blue-stained control. Examination of lane 1 reveals that proteins that migrate near spectrin, ankyrin, the Ca²⁺ pump, the α-subunit of Na⁺/K⁺ pump, band 3, and actin are all radiolabeled by [α-³²P]-N₃-ATP. The ghosts in lane 2 were emptied in the dark of entrapped [α-³²P]-N₃-ATP by running the Na⁺/K⁺ pump forward with Na⁺ plus K⁺ before exposure UV irradiation. Lanes 3 and 4 are controls in which the pool was filled (lane 3) with or emptied of (lane 4) the radiolabeled ATP analog, but with neither lane exposed to UV irradiation. It is evident that, except for actin, the only membrane proteins labeled are those exposed to the conditions described for lane 1.

Fig. 4.

Labeling of ATP pool-associated membrane proteins with [α-³²P]-N₃-ATP. This was accomplished following the same protocol used for N₃-ATP. Labeled proteins were separated by SDS/PAGE and analyzed with a Cyclone phosphoimager. Lanes 1–4 show phosphoimages of ghost proteins after separation. Lane 5 is the Coomassie blue-stained counterpart. The proteins tentatively identified are indicated with arrows. Possible protein candidates for the labeled bands are: A, β-spectrin and/or ankyrin; B, Ca²⁺ pump (M_r ∼140 kDa) or fragments of spectrin/ankyrin; C, α-subunit of the Na⁺/K⁺ pump (M_r ∼110 kDa); D, band 3; and E, actin.

MS Identification of Specific Peptides of ATP Pool-Associated Membrane Proteins Labeled with N₃-ATP.

We first evaluated whether the liquid chromatography–tandem MS (LC-MS/MS) protocol used (Materials and Methods) would provide the necessary sensitivity to detect the peptides of ghost proteins. For this purpose, a crude trypsin digest of ghost proteins was loaded onto a C18 column of an Eksigent Ultra2D NanoLC system, which was coupled inline to a hybrid dual-cell linear trap-orbitrap mass spectrometer (LTQ-Orbitrap Velos; Thermo Fisher) for peptide sequencing. Peptides from all ghost membrane/cytoskeletal proteins were detected (Table S1).

To determine which ghost proteins might be involved in corralling the ATP entrapped within the membrane pools, we first photolabeled the pool-associated elements with N₃-ATP as described earlier (Fig. 3). We then isolated the ATP-derivitized peptides by affinity chromatography with anti-ATP antibody-coated beads and/or a polymer-based metal ion capturing (PolyMAC) reagent, as described in Materials and Methods. To detect N₃-ATP labeled peptides by MS, we found it necessary to first treat the samples with alkaline phosphatase to remove negatively charged phosphates from the labeled peptides. This step gave us the ability to identify specific peptides labeled with N₃-ATP when the ATP pool was first filled with the photoactivatable ATP.

Peptides of β-spectrin, ankyrin, band 3, and actin (Table 1) were found to contain the extra mass attributed to N₃-ATP labeling. The ATP labeling on β-spectrin is at the extreme COOH terminus, near the tetramerization site and just beyond the ankyrin- binding site (18, 19). Ankyrin is labeled in its ZU-5 domain, within the spectrin-binding sequence of ankyrin (20, 21). Band 3 is labeled specifically at a glycolytic enzyme-binding site that can associate with GAPDH, aldolase, phosphofructokinase, and lactate dehydrogenase. Actin is labeled on amino acids proximal to its known N₃-ATP–binding site (17). GAPDH seen on the immunoblots of labeled proteins (Fig. 3) was not detected in our MS/MS analyses. However, for unknown reasons, only 9% of GAPDH peptides in fresh preparations of trypsin-digested ghosts could be detected by this procedure (Table S1), suggesting that GAPDH peptides may be photolabeled, as indicated in Fig. 3, but not readily detected by MS. Although proteins corresponding to the molecular weights of the catalytic subunits of both the Na⁺/K⁺ and Ca²⁺ pumps were seen in the immunoblots of N₃-ATP–labeled proteins, neither of these polypeptides was identified by MS.

Table 1.

N₃-ATP–labeled peptides of proteins that entrap the membrane-associated pool of ATP

Protein description	Accession no.	Peptide sequence
Spectrin β chain	P11277	2124-SSWESLQPEPSHPY-2137
Ankyrin	P16157	961-TPPPLAEEEGLSDR-974
Ankyrin	P16157	1069-LCQDYDTIGPEGGSLK-1084
Band 3	P02730	361-GLDLNGGPDDPLQQTGQL-378
Actin	P68133	121-QIMFETFNVPAMYVAIQAVLSLYASGR-147

The letters in bold type in each sequence represent the amino acids directly labeled with N₃-ATP.

Discussion

Previous work has established that ATP is sequestered within compartments residing in the membrane–cytoskeleton complex in human RBC ghosts. The ATP so entrapped serves as the preferential and proximal substrate for both the Na⁺/K⁺ and Ca²⁺ pumps. The main results reported in this paper identify several of the membrane and cytoskeletal components that serve to corral the local pools of ATP, including β-spectrin, ankyrin, GAPDH, and band 3 (Figs. 3 and 4 and Table 1). The methods used to identify these proteins involve first filling the pool with N₃-ATP, then after photolysis, identifying the proteins in SDS polyacrylamide gels or isolating the N₃-ATP–labeled peptides and determining their identities by MS. Clearly, we have not identified all of the elements involved in the corral, including some unknown protein bands labeled in Figs. 3 and 4, but not detected by MS. Moreover, we detected no peptides from either the Na⁺/K⁺ or Ca²⁺ pump on our MS analyses, even though polypeptides corresponding to their respective molecular weights (∼110 kDa for the Na⁺/K⁺ pump and ∼140 kDa for the Ca²⁺ pump) are labeled with the N₃-ATP in Figs. 3 and 4. We presume that this inability to detect the two pumps by MS derives from (i) the poor capacity of their peptides to ionize in the mass spectrometer, as demonstrated by our inability to detect any peptides from either pump in MS analyses of unmodified membranes (Table S1); (ii) their relatively low copy numbers in the ATP pool complex; or (iii) the stringent threshold that we set for inclusion of an N₃-ATP–labeled peptide in our MS dataset.

Several aspects of our results merit further comment. First, it is clear that the entrapped ATP cannot be tightly bound, because strongly immobilized ATP cannot serve as a substrate for either the glycolytic enzymes, PGK and PK, or the Na⁺/K⁺ or Ca²⁺ pump. Nevertheless, it is likely that the entrapped ATP, estimated as 100–600 molecules per pool (7), resides in a hydrophobic environment, given that the fluorescence of TNP-ATP is largely quenched in aqueous solution, but becomes more intense as solvent polarity declines (14). The strong fluorescence of TNP-ATP entrapped in the ghost membrane pool (Fig. 2) is consistent with a hydrophobic ATP compartment.

Our results also raise the question of how the corrals that harbor the ATP pool relate to the known structures of the membrane–cytoskeletal complexes (22–24). The major RBC membrane complexes are of two types, a junctional complex and an ankyrin complex, with some overlap in the constituents of the two (25, 26). The prime distinctions between the two complexes lie in the fact that the former contains actin, adducin, and protein 4.1, whereas the latter contains ankyrin. Our results clearly demonstrate that N₃-ATP labels ankyrin in its ZU5 domain, which contains the binding site for β-spectrin (20) and lies adjacent to the ankyrin-binding site for band 3 (27, 28). N₃-ATP also labels β-spectrin in a flexible region near the site responsible for association of αβ-spectrin dimers to tetramers and not too distant from the ankyrin-binding site on β-spectrin (18–20). In addition, N₃-ATP labels band 3 near the junction of its cytoplasmic and membrane-spanning domains within a peptide recently found to include a second binding site for GAPDH. Because this peptide resides close to the docking site of ankyrin on band 3 (27, 28), and because ankyrin is known to bind the cytoplasmic domain of the Na⁺ pump (29), it can be argued that all of the prominent N₃-ATP–labeling sites on the membrane except actin appear to reside within or near the ankyrin–band 3 complex, as depicted in Fig. 5.

Fig. 5.

Possible arrangement of membrane components that form the ATP pools that fuel the cation pumps. Membrane components that were photolabeled with N₃-ATP and identified by MS are shown in color, and other membrane components implicated in the pool’s architecture but not identified by MS are shown in gray. The locations of the major N₃-ATP–labeled peptides in ankyrin (residues 961–974), band 3 (residues 361–378), and β-spectrin (residues 2124–2137) are marked with an asterisk.

The fact that N₃-ATP labels actin, which resides at the junctional complex at least ∼30 nm from the ankyrin complex, warrants further discussion. Although an actin-associated ATP pool cannot be unequivocally excluded by our data, the fact that labeling of actin was not prevented by emptying N₃-ATP from the pool by running either the Na⁺ pump forward or the PGK reaction backward (Fig. 3, lanes 2 and 3) suggests that actin labeling is not mediated by N₃-ATP within the pool. We suggest instead that N₃-ATP bound to the ATP-binding site on actin is responsible for this prominent labeling. This conclusion is supported by the observation that addition of excess unlabeled ATP competitively blocks N₃-ATP labeling of actin in porous ghosts (Fig. 3, lane 4).

Another problem concerns the locations of the two enzymes (PGK and PK) that provide ATP for the pool. Although recent data from our laboratory suggest that PK, like many other glycolytic enzymes, binds to band 3, no information exists on the location of PGK on the membrane. Nevertheless, association of PGK with the membrane appears likely, given the report that an antibody to PGK inhibits the activity of purified PGK, but not PGK bound to the RBC membrane (30), indicating that PGK could be tightly associated with pool ATP in the RBC membrane.

Finally, it is not known what other functions might be associated with the RBC membrane ATP pool. Possible candidates include providing fuel for an ATP-dependent glucose transporter (31), energizing a flippase (32), or supplying the ATP for the deoxygenation-promoted ATP release pathway (33, 34). With regard to the latter, we have preliminary evidence that TNP-ATP entrapped in the ATP pool of resealed ghosts is released to the outside of the ghosts on deoxygenation.

Materials and Methods

Loading and Emptying the Membrane Pools of ATP in Porous Ghosts.

Two incubations, where indicated, were needed, the first to fill the pools and the second to empty them of entrapped ATP. Thus, whether or not the pools were empty could be determined by a pulse chase experiment with 2 μM [γ³²-P]-ATP for 20 s at 0 °C, as described above. All incubations and ATP type pool manipulations were carried out using the methods and solutions described previously (7, 9).

The first incubation was carried out for 30 min at 37 °C in solution (A), containing 10 mM Tris (pH 7.5), 40 mM NaCl, 2 mM MgCl₂, and 0.25 mM EDTA. In addition, for those ghosts in which the pools were to be loaded with ATP, solution (A) also contained 1.5 mM ATP. The ghosts were washed (always with 17 mM Tris, pH 7.5) and then exposed to a second incubation for another 15 min at 37 °C in solution (A), but this time the buffer also contained either 10 mM KCl (to run the Na⁺/K⁺ pump forward) or solution (B) (5 mM MgSO₄, 17.5 mM NaHCO₃, 20 mM cysteine, 50 mM glycine, 5 mM 3-phosphoglyceric acid, and 0.25 mM NADH; pH 7.0), to run the phosphoglycerate (PGK) reaction backward. In other experiments, the ATP pools also could be filled by running the PGK reaction forward. This was done using solution (C), containing 0.83 mM Triose-P, 0.415 mM NAD, 0.25 mM ADP, 5.0 mM MgSO₄, 50.0 mM Na₂HPO₄/NaH₂PO₄, and 132 mM glycine. The ghosts were always washed to remove all bulk substrates before attempts to pulse-label the Na⁺/K⁺ pump or Ca²⁺ pump with [γ-³²P]-ATP. After pulse-labeling, the reaction was stopped by adding a solution containing 2.5% TCA, 1 mM ATP, and 1 mM K₃PO₄. The precipitate was washed with this solution before preparation for analysis by phosphoimaging.

For the results shown in Fig. 1B, the experiments were performed by loading the pool with ATP as before, by first incubating at 37 °C for 30 min in modified solution (A) containing 1.5 mM ATP, 15 mM Tris (pH 7.5), 12 μM MgCl₂, 50 mM NaCl, or 50 mM NMDG. The second incubation, to unload the pooled ATP, was carried out as described above for running the Na⁺ pump forward with Na⁺ plus K⁺. The ghosts were then washed before pulse-labeling with [γ-³²P]-ATP to measure Na⁺ as well as Ca²⁺ phosphoproteins, as described previously.

Fluorescence Microscopy.

The ghost’s ATP pools were either loaded with TNP-ATP or loaded, then emptied of TNP-ATP following the same protocol as described for the loading manipulations with ATP. The ghosts were washed and imaged with an Olympus FluoView FV1000 inverted confocal microscope with a 60× oil-immersion objective.

Production and Use of a Polyclonal Antibody to N₃-ATP.

Keyhole limpet hemocyanin photolabeled with N₃-ATP (13) was used as an antigen. UV photolysis was carried out in PBS buffer (pH 7.4) at room temperature. The antigen suspended in Gold Adjuvant Reagent (Sigma-Aldrich) was injected into rabbits at regular intervals over a 4-mo period. The separated serum was stored at −80 °C until use. Specificity of the antibody was determined by Western blot analysis (Fig. 3).

Photolabeling and Detection of Pool-Related Protein Components by N₃-ATP.

The ghost’s ATP pools were loaded with N₃-ATP in the dark following the same protocol as for ATP. The ghosts were washed before exposure to UV irradiation, which resulted in photolabeling of the proteins residing adjacent to the entrapped N₃-ATP. The ghosts were solubilized with 2% SDS, and protein concentration was determined using the Micro BCA Assay Kit (Thermo Scientific). Ghost proteins were separated by SDS/PAGE, with the N₃-ATP–labeled proteins detected by Western blot analysis using the anti–N₃-ATP antibody and the Thermo Scientific Enhanced Chemiluminescence Kit.

Radiolabeling and Detection of ATP Pool-Associated Protein Components.

These experiments were carried out using the same protocol described above for N₃-ATP–labeled protein components. After separation by SDS/PAGE, the [³²P]-labeled proteins were detected by phosphoimaging.

Purification of N₃-ATP–Photolabeled Peptides.

ATP pool components of ghosts were photolabeled with N₃-ATP, as described above, with 2% C₁₂E₈ (Affymetrix) in 100 mM Tris-HCl (pH 7.5) used to solubilize the ghosts. The samples were then diluted 20 times with 100 mM Tris-HCl and precleared with rabbit preimmune IgG immobilized on Protein A/G PLUS-Agarose (Santa Cruz Biotechnology) following the manufacturer’s protocol. N₃-ATP–labeled proteins were purified by immobilized anti–N₃-ATP antibody on Affi-Gel Hz hydrazide gel (Bio-Rad), then trypsin-digested to prepare a mixture of labeled and unlabeled peptides. PolyMAC reagent functionalized with Ti (IV) ions (Tymora Analytical) was used to separate N₃-ATP–labeled peptides from unlabeled peptides as described previously (35). Alkaline phosphatase (New England BioLabs) was added to remove the 5′-phosphate from N₃-ATP, and then removed from the peptide mixture using a 10-kDa cutoff protein concentrator (Millipore). Peptide samples were desalted using a Sep-Pak C18 desalting column (Waters) before MS analysis. In some studies, total membrane protein extracts were first digested with trypsin (35), and then N₃-ATP–labeled peptides were isolated as described above.

LC-MS/MS Analysis.

Tryptic peptides were analyzed by LC-MS/MS on a high-resolution hybrid duel-cell linear ion trap orbitrap mass spectrometer (LTQ-Orbitrap Velos; Thermo Fisher) interfaced with an Eksigent Ultra 2D NanoLC HPLC system. Peptide samples were dissolved in 8 μL of 0.1% formic acid and injected into a capillary column (75 μm i.d. and 12 cm bed length) packed with 5 μm of C18 Magic beads resin (Michrom). Peptides were then eluted at 300 nL/min using a 0–100% acetonitrile gradient over 90 min. The electrospray ionization emitter tip was generated on the column with a laser puller (model P-2000; Sutter Instruments). The mass spectrometer was operated in the data-dependent mode in which a full-scan MS (from m/z 300–1,700 with a resolution of 60,000 at m/z 400) was followed by 20 MS/MS scans of the most abundant ions. Ions with a charge state of +1 were excluded. The mass exclusion time was 90 s.

Peptide Identification.

The LTQ-Orbitrap raw files were searched directly against a Homo sapiens database with no redundant entries (67,250 entries; human International Protein Index v.3.64) using the SEQUEST algorithm on Proteome Discoverer version 1.2 (Thermo Fisher) or Sorcerer IDA server version 2.5.6 (Sage-N Research). Proteome Discoverer created DTA files from the raw data with a minimum ion threshold of 15 and absolute intensity threshold of 50. Peptide precursor mass tolerance was set at 10 ppm, and MS/MS tolerance was set at 0.8 Da. Search criteria included a static modification of cysteine residues of +57.0214 Da, a variable modification of +15.9949 Da to include potential oxidation of methionine residues. N₃-ATP–photolabeled peptides were identified based on the additional mass associated with the N-adenosine adduct (+280.092 Da) or the N-adenine adduct (+148.050 Da). False discovery rates were set for 1%.