Regulation of AMP Deaminase by Phosphoinositides*

AMP deaminase (AMPD) converts AMP to IMP and is a diverse and highly regulated enzyme that is a key component of the adenylate catabolic pathway. In this report, we identify the high affinity interaction between AMPD and phosphoinositides as a mechanism for regulation of this enzyme. We demonstrate that endogenous rat brain AMPD and the human AMPD3 recombinant enzymes specifically bind inositide-based affinity probes and to mixed lipid micelles that contain phosphatidylinositol 4,5-bisphosphate. Moreover, we show that phosphoinositides specifically inhibit AMPD catalytic activity. Phosphatidylinositol 4,5-bisphosphate is the most potent inhibitor, effecting pure noncompetitive inhibition of the wild type human AMPD3 recombinant enzyme with a K i of 110 nm. AMPD activity can be released from membrane fractions by in vitro treatment with neomycin, a phosphoinositide-binding drug. In addition, in vivo modulation of phosphoinositide levels leads to a change in the soluble and membrane-associated pools of AMPD activity. The predicted human AMPD3 sequence contains pleckstrin homology domains and (R/K)X n (R/K)XKK sequences, both of which are characterized phosphoinositide-binding motifs. The interaction between AMPD and phosphoinositides may mediate membrane localization of the enzyme and function to modulate catalytic activity in vivo.

AMPD is a diverse and highly regulated enzyme located at a branchpoint in the adenine nucleotide catabolic pathway and is important in regulating nucleotide pools. AMPD is also a component of the purine nucleotide cycle, an energy-generating pathway reportedly operative in many animal tissues (reviewed in Ref. 12). The AMPD1 gene encodes human isoform M and rat isoform A (13); the AMPD2 gene encodes the human isoform L and rat isoform B (14,15); and the AMPD3 gene encodes the human isoform E and the rat isoform C (16,17). A single AMPD gene has also been identified in yeast (18). All human AMPD isoforms contain similar C-terminal regions and substantially divergent N-terminal domains. The AMPD1 isoform is found almost exclusively in skeletal muscle, whereas the AMPD2 and AMPD3 isoforms are widely expressed in many tissues and cells (19,20), including mammalian brain (21,22). AMPD activity is highly regulated through interactions with other proteins (23)(24)(25), phosphorylation (26,27), and small molecules (10, 11, 28 -32). Regarding the latter, polyphosphates (32) and inositol (1,2,3,4,5,6)-hexakisphosphate (10) inhibit AMPD, whereas inositol (1,3,4,5)-tetrakisphosphate (Ins(1,3,4,5)P 4 ) modestly stimulates AMPD activity (11). AMPD activity is also modulated by lipids. In particular, nonskeletal muscle AMPD activities are regulated by fatty acids, fatty acyl coenzyme As, phosphatidic acid, and phosphatidylcholine (PC) with K i values ranging from 10 to 100 M (33-39). The high concentrations required for these modulatory effects suggest that these lipid interactions may be low affinity or nonspecific. However, the interaction with lipids may prove to be physiologically relevant. Whereas hydropathy analysis suggests that AMPD isoforms do not contain putative transmem-brane spanning domains, the AMPD3 enzyme can associate with erythrocyte membrane fractions (40,41). In this report, we provide several independent lines of evidence that the specific, high affinity interaction between the AMPD3 isoform and phosphoinositides constitutes an important mechanism for enzyme regulation and localization.
Expression and Purification of Human AMPD3 Recombinant Enzymes-Wild type and N-terminally truncated (⌬M90) human AMPD3 recombinant enzymes were produced in Sf9 (Spodoptera frugiperda) insect cells using a baculoviral expression system and partially purified by phosphocellulose chromatography as described previously (42).
AMPD Assay-AMPD activity was determined using a phenol/hypochlorite reaction (11). The reaction contained 25 mM sodium citrate, pH 6.0, 50 mM potassium chloride, 10 mM AMP, and samples were incubated at 37°C for 10 min, followed by addition of 2.5 ml of 100 mM phenol, 200 mM sodium nitroprusside in H 2 O, 2.5 ml of 125 mM sodium hydroxide, 200 mM dibasic sodium phosphate, 0.1% sodium hypochlorite. Absorbance was measured at 625 nm. Absolute ammonia was determined with ammonium sulfate. For kinetic determinations, AMPD activity was assayed using a high pressure liquid chromatography method to separate substrate (AMP) from product (IMP) as described previously (42).
Unilamellar Mixed Micelle Assay-Binding of AMPD to unilamellar mixed micelles was performed as described in Ref. 43. 400 g of PE was added to various concentrations of other phospholipids and dried under N 2 . The lipids were resuspended in 1 ml of 180 mM sucrose, pelleted at 10,000 ϫ g for 15 min, and redissolved in 1 ml of 50 mM HEPES, 1 mM EDTA, and 1 mM EGTA. Human AMPD3 recombinant protein (2 g) was added to the micelles for 30 min at 25°C, and centrifuged at 400,000 ϫ g for 40 min. The pellets were separated by SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with the anti-AMPD3 antibody.
PC12 Cell Culture-PC12 cells were cultured on 10-cm culture plates and grown to confluency in cell culture media (RPMI, fetal bovine serum, horse serum, penicillin/streptomycin, and L-glutamine). Cells were collected by trituration, centrifuged at 1000 ϫ g for 5 min, and lysed by sonication, and the supernatants and membranes were separated by centrifugation at 12,000 ϫ g for 5 min.
Phosphoinositide Labeling-Phosphoinositide levels were determined as described in Ref. 44. PC12 cells were preincubated with 10 mM taurine for 2 h, and then 30 Ci of [␥-32 P]ATP was added to the culture media and incubated at 37°C for 3 h. Lipids were extracted with 1 ml of chloroform-methanol (1:2 v/v) and 0.25 ml of 10 mM EDTA, pH 7.4, and centrifuged at 1000 ϫ g for 5 min. The organic layer was removed, washed with 0.125 ml of 2.4 M HCl and 1 ml of CH 3 OH-H 2 0 (1:1 v/v) and centrifuged for 5 min. The organic layer was removed, dried under N 2 , and dissolved in chloroform-methanol (2:1 v/v). Lipids were separated using Whatman (Maidstone, UK) silica gel plates with chloroform/ methanol/4 M NH 4 OH (90:65:20 v/v/v).

Endogenous Brain AMPD Binds Inositide Affinity Probes-
Using an aminopropyl-InsP 4 resin to purify inositide-binding proteins from rat brain, we previously identified a protein at approximately 80 kDa that eluted in fractions containing the clathrin adaptor/assembly protein AP-2 (3,5,6,45). Published reports have shown that AMPD, an enzyme family whose native purified subunit molecular masses are between 66 -88 kDa, is regulated by inositol polyphosphates (10,11). Therefore we tested whether AMPD activity was enriched in fractions containing the 80-kDa protein eluted from the aminopropyl-InsP 4 resin (Fig. 1A). All of the AMPD activity bound to the resin, and a symmetrical peak of AMPD activity was eluted in the fractions containing AP-2 and the 80-kDa protein. Similar to AP-2, the 80-kDa protein in the peak activity fractions incorporated an [ 125 I]ASA-InsP 4 photoaffinity label (Fig. 1B, odd fractions). Photolabeling was specific because including 30 M unlabeled Ins(1,3,4,5)P 4 displaced the label (Fig. 1B, even fractions). By batch purification, the aminopropyl-InsP 4 resin gave a 71-fold purification with a specific activity of 917.1 nmol/ min/mg and an 86% yield. Immunoblot analysis of batch eluted aminopropyl-InsP 4 eluate with an anti-AMPD3 serum (Fig. 1C) indicated a single strongly immunoreactive band at approximately 80 kDa. Because both AMPD2 and AMPD3 are expressed in brain (21,22) and all the AMPD activity was recovered in a single peak, the AMPD2 and AMPD3 isoforms appear to behave similarly in this purification. Thus, comparable with other high affinity inositide-binding proteins, endogenous rat brain AMPD is effectively purified using an inositide affinity resin (3-9, 45, 46).
Human AMPD3 Recombinant Enzyme Binds Phosphoinositides-Recent advances in recombinant expression of human AMPD cDNAs have provided larger quantities of higher purity enzymes than can be obtained from endogenous sources (42). Therefore, the human AMPD3 recombinant enzyme was used to determine the specificity and affinity of inositide binding in AMPD. AMPD was specifically labeled using a high efficiency [ 3 H]BZDC-Ins(1,3,4,5)P 4 photoprobe (Fig. 2, A-C, Total). A variety of inositides and phospholipids were tested for displacement of the photolabel. PtdIns(4,5)P 2 and PtdIns(4)P were the most potent inhibitors of photolabeling, with 50% displacement of the label (IC 50 ) effected by addition of 250 nM (Fig. 2, A and  B). PtdIns(3,4,5)P 3 was weaker, with an IC 50 between 1-2 M (Fig. 2B). Ins(1,3,4,5)P 4 and inositol (1,2,3,4,5,6)-hexakisphosphate were substantially less effective and displaced the photolabel only when added at 10 M (Fig. 2B). Phosphatidylinositol (PI) and PS displaced the photolabel when added at concentrations greater than 10 M (Fig. 2C), whereas PE or PC did not displace the label even at 30 M. AMPD could also be photolabeled with a [ 3 H]BZDC-PtdIns(4,5)P 2 probe (Fig. 2D). Lower concentrations of PtdIns(4,5)P 2 were required to effect displacement in this label (IC 50 ϭ 100 nM) presumably because lower concentrations of the [ 3 H]BZDC-PtdIns(4,5)P 2 label were used. These data show that the human AMPD3 recombinant enzyme binds phosphoinositides with high affinity, with both the inositol head group and glycerolipid moieties essential for high affinity binding.
To ensure that differences in lipid accessibility or micellar size were not the basis for the apparent specificity differences observed, we examined the interaction of AMPD3 recombinant protein with unilamellar mixed lipid micelles. Binding of AMPD to mixed micelles that contained PE as the core lipid and increasing concentrations of either PtdIns(4,5)P 2 , PS, PI, or PC, is shown in Fig. 3. A fraction of the human AMPD3 recombinant enzyme consistently associated with the PE micelle pellet, presumably through a low affinity interaction with PE that is present in high concentrations in the micelles. However, addition of PtdIns(4,5)P 2 to the PE micelles produced a substantial increase in the binding of AMPD to the micelles. The increase in AMPD associated with the lipid micelle was concentration-dependent and saturable with half-maximal association observed at approximately 200 nM. In contrast, addition of PS, PI, or PC to the PE micelles did not lead to any additional increase in association of AMPD compared with PE alone, indicating that the enhanced binding was selective for PtdIns(4,5)P 2 . These data demonstrate that the human AMPD3 recombinant enzyme recognizes PtdIns(4,5)P 2 in a mixed micelle with an affinity and selectivity similar to that determined in the photoaffinity labeling.
Phosphoinositides Inhibit AMPD Activity-We next examined the effect of phosphoinositides on AMPD catalytic activity. Phosphoinositides potently inhibit human AMPD3 recombinant activity in a dose-dependent manner (Fig. 4). PtdIns(4,5)-FIG. 1. Aminopropyl-InsP 4 affinity resin purification of AMPD from rat brain. A partially purified rat brain fraction was applied to an aminopropyl-InsP 4 resin and eluted with the NaCl gradient shown. A, individual fractions were assayed for AMPD activity. B, pairs of fractions were first pooled (fractions1 and 2, fractions 3 and 4, etc.) and then split into two samples for photoaffinity labeled using [ 125 I]ASA-InsP 4 . Photolabel only (odd fractions) or photolabel plus 30 M unlabeled Ins(1,3,4,5)P 4 (even fractions) was used to assess binding specificity. C, immunoblot analysis of crude brain extract, partially purified (heparin-agarose), and batch aminopropyl-InsP 4 resin purified fractions probed with an anti-AMPD3 antiserum (42). Apparent subunit molecular mass was estimated according to the migration of prestained protein markers. A representative purification profile, photolabeling, and immunoblot, performed in four independent experiments, is shown.  3. AMPD binds to PtdIns(4,5)P 2 in unilamellar mixed micelles. Recombinant AMPD3 cosedimentation with mixed lipid micelles was assessed by immunoblot analysis and quantified by densitometry. PE was the core lipid, and increasing concentrations of PtdIns(4,5)P 2 , PS, PC, or PI were added as indicated. Sedimentation assays were done in three independent experiments, and the averaged values are shown. P 2 was the most effective inhibitor of enzyme activity, with an IC 50 of approximately 100 nM (Fig. 4A, open symbols). The only other phospholipids that inhibited activity in the submicromolar range were phosphoinositides (Fig. 4, B and C). PtdIns(3)P, PtdIns(4)P, PtdIns(3,4)P 2 , and PtdIns (3,4,5) were between 1.5and 3-fold less potent than PtdIns(4,5)P 2 but did not effect complete inhibition even at 3 M. Other phospholipids, such as PI, PC, and PS, produced either no or only slight (Ͻ20%) inhibition when the concentration was increased to 3 M (Fig.  4, C and D). Inositol-1,4,5-trisphosphate, Ins(1,3,4,5)P 4 , and inositol (1,2,3,4,5,6)-hexakisphosphate had no effect on activity, even at 30 M (Fig. 4D). Endogenous rat brain AMPD is also potently inhibited by phosphoinositides (Fig. 4A, closed symbols). Phosphoinositides were approximately 2-fold less effective on endogenous brain AMPD than the recombinant enzyme, with an IC 50 for PtdIns(4,5)P 2 of 250 nM. This decreased affinity may be a result of interference by other inositidebinding proteins in the preparation or may be a consequence of N-terminal proteolytic cleavage of endogenous AMPD (see below). These data indicate that interaction with phosphoinositides leads to potent inhibition of human AMPD3 recombinant and endogenous rat brain activity, with phosphoinositides more than 500 times more effective than other phospholipids or inositol polyphosphates in inhibition of AMPD activity.
A kinetic analysis of inhibition by PtdIns(4,5)P 2 on the human AMPD3 recombinant activity was performed by varying PtdIns(4,5)P 2 and substrate concentrations. A Lineweaver-Burk plot of the inhibition data (Fig. 5A) shows decreasing V max values but no effect on the K m in the presence of increasing concentrations of PtdIns(4,5)P 2 , indicative of noncompetitive inhibition. Moreover, a linear replot of slope versus PtdIns(4,5)P 2 concentration (Fig. 5B) demonstrates pure noncompetitive inhibition. The K i for PtdIns(4,5)P 2 inhibition of the human AMPD3 recombinant activity is 110 Ϯ 41 nM (n ϭ 5) (Fig. 5).
Mammalian AMPD isoforms have conserved C-terminal catalytic domains and divergent N-terminal regions. To determine whether the N terminus contributes to phosphoinositide recognition, we tested effects of phosphoinositides on a truncated human AMPD3 recombinant enzyme that is missing 89 Nterminal amino acids (⌬M90AMPD3). The activity of the Nterminally truncated protein is also inhibited noncompetitively by phosphoinositides but with a 5-fold lower affinity (data not shown). The K i for PtdIns(4,5)P 2 with the ⌬M90AMPD3 recombinant enzyme is 540 Ϯ 260 nM (n ϭ 4). These data suggest that the N terminus of the AMPD3 isoform is necessary to confer the high affinity for PtdIns(4,5)P 2 inhibition but that domains outside of the N terminus also interact with phosphoinositides.
Membrane Association of AMPD May Involve Phosphoinositides-To investigate whether AMPD may interact with endogenous phosphoinositides within the membrane, we examined the effects of neomycin, a phosphoinositide-binding drug (47), on AMPD activity in membrane fractions. Addition of neomycin to the membrane fraction, followed by recentrifugation, released AMPD activity from the membranes, resulting in a 2-3-fold increase in activity in the supernatant fraction (Fig.  6A, black bars). Furthermore, addition of neomycin directly to the membrane fraction, without recentrifugation, led to a significant increase in the total AMPD activity, suggesting that AMPD activity is inhibited or conformationally restricted while associated with membranes (Fig. 6A, gray bars). These data are consistent with endogenous AMPD association with membranes involving phosphoinositide interactions.
To investigate further whether phosphoinositide interactions may mediate membrane association of AMPD, we incubated cells in the presence of taurine, which increases the intracellular levels of PtdIns(4,5)P 2 and PtdIns(4)P (44). Incubation with taurine led to an approximately 2-fold increase in the level of total phosphoinositides in PC12 cells (Fig. 6B). In the taurine-treated cells, AMPD activity in the supernatant fraction was 24% lower compared with the control (Fig. 6C), whereas activity associated with the membrane was 31% higher than in untreated cells (Fig. 6D). We also tested the effects of neomycin, which has been documented to block interactions between PtdIns(4,5)P 2 and its target proteins (47). Addition of neomycin to PC12 cells led to a significant increase in the supernatant AMPD activity, consistent with release of AMPD from the membrane (data not shown). Together these data suggest that modulation of endogenous phosphoinositide levels can produce a change in the distribution between the soluble and membrane-associated pools of AMPD.
Phosphoinositide-binding Domains in AMPD-One characteristic motif that has been shown to mediate phosphoinositide binding in many proteins, including pleckstrin, spectrin, PLC␦, and Bruton's tyrosine kinase, is a pleckstrin homology (PH) domain (1,2,48). Analysis of the predicted AMPD3 sequence indicates that it contains two tandem PH domains in the Cterminal half (Fig. 7A). Interestingly, one of the proposed catalytic regions of AMPD lies within the second PH domain (Fig.  7C). The second AMPD3 PH domain also contains conserved basic residues (denoted by the asterisks) in several loops, which have been shown to be essential for inositide binding in the PLC␦ PH domain (48). The AMPD2 sequence contains similar PH domains (data not shown). In addition to the PH domain, another motif has been implicated in phosphoinositide binding: the highly basic domain (K/R/H)X 3-6 (K/R/H)X(K/R)(K/R), which is found in the actin-binding proteins profilin, gelsolin, and cofilin, as well as the synaptic vesicle protein synaptotagmin (49 -51). In the AMPD3 isoform there are two (R/K)X 5 (R/ K)XKK sequences that fit this consensus (Fig. 7B). One is present within the first PH domain, and the other lies in the N terminus (Fig. 7C). These sequence data identify candidate phosphoinositide-binding domains in the AMPD3 enzyme. DISCUSSION In this report, we demonstrate that phosphoinositides are specific regulators of AMPD. Six lines of evidence support this conclusion. First, we identified AMPD in a screen for high affinity inositide-binding proteins expressed in mammalian brain. Second, AMPD was specifically photolabeled by inositide-based photoprobes and bound to PtdIns(4,5)P 2 in mixed lipid micelles. Third, endogenous rat brain AMPD and human AMPD3 recombinant enzymes were potently inhibited by phosphoinositides. PtdIns(4,5)P 2 was the most effective inhibitor and displayed pure noncompetitive inhibition with a K i of 110 nM. Fourth, AMPD activity was released from and disinhibited in membrane fractions by neomycin, a PtdIns(4,5)P 2 -binding drug. Fifth, in vivo modulation of phosphoinositides led to a change in AMPD pools. Finally, several characterized phosphoinositide-binding domains are present in the predicted AMPD3 amino acid sequence.
The C-terminal region of AMPD2 and AMPD3 isoforms contains two PH domains, structural domains shown to mediate phosphoinositide binding in a variety of proteins (48). Interestingly, the second PH domain in AMPD contains several conserved basic residues shown to be critical for inositide binding in the PLC␦ PH domain (48). In addition, the divergent Nterminal region of the AMPD3 isoform may also participate in phosphoinositide binding because the ⌬M90AMPD3 recombinant enzyme displayed a 5-fold lower affinity for PtdIns(4,5)P 2 than the wild type activity. In AMPD3, both N-terminal and C-terminal domains contain an (R/K)X n (R/K)XKK sequence, a motif that has been implicated in phosphoinositide binding in several actin-binding proteins and synaptotagmin (49 -51). These sequence data, together with the fact that the AMPD3 isoform has a very high affinity for PtdIns(4,5)P 2 , support the likelihood that several domains in the enzyme contribute to this interaction.
Binding to phosphoinositides may contribute to localizing AMPD to membranes, regulating catalytic activity, sequestering PtdIns(4,5)P 2 , and/or sequestering a mobilizable pool of AMPD. Phosphoinositide binding is essential for membrane association of several PH domain-containing proteins, including PLC␦ and protein kinase B/Akt kinase (48). A function for phosphoinositide binding in membrane association of AMPD is supported by the demonstration that neomycin treatment can release the enzyme from membranes, and treatments that lead to increases in phosphoinositide levels also lead to increased AMPD in membrane fractions. Recent studies have shown that PtdIns(4,5)P 2 is present in plasma membrane and internal membrane compartments, including endoplasmic reticulum, Golgi, and nuclear membranes (2). Subcellular fractionation studies in brain indicate that AMPD activity is associated with multiple membrane compartments (53). Whether phosphoinositide binding is sufficient for localizing AMPD to specific membrane compartments and whether additional domains or protein-protein binding is required are important issues for future studies.
The interaction between AMPD and phosphoinositides may contribute to sequestering the enzyme in an inactive pool. In this regard, interaction between phosphoinositides and AMPD would be analogous to phosphoinositide regulation of the actinbinding protein profilin. Following receptor-activated, phospholipase C-mediated PtdIns(4,5)P 2 hydrolysis, profilin is released from the membrane and can interact with actin monomers to promote actin polymerization (49,50). Similarly, AMPD may be released from membranes and activated following PtdIns(4,5)P 2 hydrolysis or by altering the accessibility of PtdIns(4,5)P 2 via translocation of other phosphoinositide-binding proteins or activation of PtdIns(4,5)P 2 metabolizing enzymes (1, 2). In addition, interactions between AMPD and molecules that enhance or diminish the affinity or access to phosphoinositides could lead to changes in AMPD membrane FIG. 7. Identification of candidate phosphoinositide-binding domains in AMPD. A, alignment of two PH domains in AMPD3 with the characterized PH domains in pleckstrin (Plec), dynamin (Dyn), PLC␦1, spectrin (Spec), and Bruton's tyrosine kinase (Btk) (48). Alignments and assignments of beta strands (␤), loops (L), and the ␣ helix (␣) are exactly as those defined in Ref. 48. Amino acids that share similarities by Dayhoff criteria, in six out of seven sequences, are shaded. The amino acids in bold italic type are residues shown to be involved in phosphoinositide binding (48). Asterisks denote the position of conserved basic resides in AMPD to those in PLC␦1. B, AMPD3 contains two sequences that are homologous to phosphoinositide-binding domains shown in Ref. 51 and identified in profilin (Pro) (49), synaptotagmin (Syn) (50), and gelsolin (GelI and GelII) (51) and proposed in cofilin (Cof) (51). The AMPDC sequence begins at amino acid 471 and AMPDN sequence starts at amino acid 63. C, position of candidate phosphoinositide-binding motifs in the human AMPD3 protein.
association and activity. For example, inositol polyphosphates that bind to but do not inhibit AMPD may act as competitive antagonists of phosphoinositides for AMPD. Competition between inositol polyphosphates and phosphoinositides has been proposed in regulation of PLC␦, AP-2, Bruton's tyrosine kinase, and synaptotagmin (1-3, 45, 46, 48, 50).
A mobilizable pool of AMPD may be required for a variety of purposes, for regulation of adenylate pools, to compete with cytosolic 5Ј nucleotidases in modulation of adenosine levels, or for ammonia generation. Investigation of the cellular function of phosphoinositide binding to AMPD requires additional studies that appear warranted based on the results presented here.