An 18-kDa domain of a glycosylphosphatidylinositol-linked NAD:arginine ADP-ribosyltransferase possesses NAD glycohydrolase activity.

Transfection of NMU (rat mammary adenocarcinoma) cells with NAD:arginine ADP-ribosyltransferase cDNAs from Yac-1 murine lymphoma cells or rabbit muscle increased NAD glycohydrolase and ADP-ribosyltransferase activities. The ADP-ribosyltransferase activity was released from transformed NMU cells by phosphatidylinositol-specific phospholipase C (PI-PLC) and hence glycosylphosphatidylinositol (GPI)-anchored, whereas the NAD glycohydrolase (NADase) activity remained cell-associated. By gel permeation chromatography, the size of the PI-PLC-released transferase was approximately 40 kDa and that of the detergent-solubilized NADase was approximately 100 kDa. Using polyclonal antibodies against rabbit muscle transferase on Western blots, approximately 18- and approximately 30-kDa band were visualized among proteins from the NADase fractions and 38-40-kDa bands with protein from the transferase fractions. Incubation of blots with [32P]NAD led to the incorporation of radioactivity into the immunoreactive transferase bands of 38 kDa and the immunoreactive NADase band of approximately 18 kDa. These data suggest that proteolysis of ADP-ribosyltransferase synthesized in transformed NMU cells might result in the formation of aggregates of an 18-kDa NAD glycohydrolase. A fusion protein with glutathione S-transferase linked to the amino terminus of Yac-1 transferase, from which the amino-terminal 121 amino acids had been deleted (GST-Yac-1-delta121), exhibited NADase, but not transferase, activity. The size of the recombinant fusion protein was similar to that of the proteolytic fragment seen in NMU cells transformed with transferase cDNA. These results are compatible with the conclusion that the NAD glycohydrolase activity was generated in NMU cells by proteolysis of ADP-ribosyltransferase, with release of a carboxyl-terminal fragment that possesses glycohydrolase but not transferase activity, i.e. the carboxyl-terminal portion of the transferase can exist as a catalytically active NADase.

In murine cytotoxic T cells, ADP-ribosylation of a 40-kDa surface protein (p40) was associated with the inhibition of p56 lck tyrosine kinase, which exists in a complex with p40, and suppression of cytotoxic T cell proliferation and cytotoxic activity (17,18). The ADP-ribosyltransferase involved in this reaction has not been identified although candidate genes have been identified, including Yac-1. A second ADP-ribosyltransferase (Yac-2) was cloned and characterized from mouse lymphoma cells (14). In contrast to the muscle and Yac-1 transferases, the Yac-2 enzyme although membrane-bound is apparently not GPI-anchored and exhibits significant basal NAD glycohydrolase activity (14). Based on the PI-PLC sensitivity of the cytotoxic T cell transferase, Yac-2 may not be responsible for the activity.
Structural analysis of bacterial toxin ADP-ribosyltransferases has demonstrated three regions of similarity (regions I, II, and III) that are believed to form part of the active site (19). For some of the toxins, region I contains a critical histidine or arginine (20 -22); region II is composed of closely spaced aromatic and hydrophobic amino acids (23,24); region III contains a catalytic glutamic acid (20,(25)(26)(27)(28)(29)(30). In other toxins, a serine replaces the region II domain (24). Similar to findings with the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: National Institutes of Health, 10  bacterial toxins, site-directed mutagenesis of the rabbit muscle transferase identified glutamates 238 and 240 in the putative region III as essential for enzyme activity (31). These data, along with the alignment of the deduced amino acid sequences of the cloned vertebrate transferases, are consistent with the hypothesis that the enzymes have a common mechanism of NAD binding and catalysis.
The muscle and lymphocyte ADP-ribosyltransferases catalyze the ADP-ribosylation of arginine, agmatine, and other simple guanidino compounds (8,13). The NAD glycohydrolase activity of these enzymes is only a small percentage of maximal transferase activity in the presence of optimal concentrations of guanidino compounds as ADP-ribose acceptors. We report here that rat mammary adenocarcinoma (NMU) cells transformed with the rabbit skeletal muscle ADP-ribosyltransferase cDNA expressed significant NAD glycohydrolase as well as transferase activity. Membrane-bound NAD glycohydrolase activity was associated with a 100-kDa protein, an apparent oligomer of a proteolytic fragment of the transferase, and a truncated form of the transferase, synthesized in Escherichia coli, exhibited NADase activity.
Expression of Proteins in NMU Cells-NMU cells were transfected with 15 g of the pMAMneo vector (control) or vector containing the full-length or truncated transferase cDNAs using the calcium phosphate precipitation method (33), and stable transformants were selected with Geneticin (0.5 mg/ml). Following induction of protein expression with dexamethasone (1 M) for 24 h, cells were washed with Dulbecco's phosphate-buffered saline (PBS) and incubated for 60 min in PBS (0.7 ml) with or without phosphatidylinositol-specific phospholipase C (PI-PLC, 0.5 unit). After collecting the PBS fraction, the cells were washed with PBS, trypsinized, and lysed in 0.5 ml of lysis buffer (10 mM Tris, pH 8.0, 1 mM EDTA) by repeated freeze-thawing. After centrifugation (100,000 ϫ g, 1 h) of the lysate, the supernatant (Supernatant, 0.5 ml) was collected and the membranes (Pellet) were suspended in 0.5 ml of PBS.
Gel Permeation Chromatography of ADP-Ribosyltransferase and NA-Dase on Ultrogel AcA 44 -NMU cells transformed with wild-type Yac-1 or rabbit muscle transferase cDNAs were incubated in PBS without or with PI-PLC as described above. The PBS fraction was applied to a column (2 ϫ 120 cm) of Ultrogel AcA 44, equilibrated with buffer A (20 mM Tris, pH 8.0, 1 mM EDTA, 100 mM NaCl, 1% Triton X-100). Samples of the eluate were assayed for transferase activity (Fig. 1A). Membranes from the transformed NMU cells were suspended in buffer A, stirred for 18 h at 4°C, and centrifuged (100,000 ϫ g, 1 h). The supernatant was applied to a column (2 ϫ 120 cm) of Ultrogel AcA 44, equilibrated, and eluted with buffer A. Samples of fractions were assayed for NAD glycohydrolase activity (Fig. 1B). Fractions containing maximal activity were pooled and used for activity measurements.
Immunoreactivity of ADP-Ribosyltransferase and NADase-Proteins in fractions from gel filtration that contained ADP-ribosyltransferase or NAD glycohydrolase activities were subjected to SDS-polyacrylamide gel electrophoresis (SDS-PAGE) in a 12% gel (Tris glycine) and transferred to nitrocellulose. The membranes were incubated first with polyclonal anti-rabbit muscle ADP-ribosyltransferase antibodies (9) and then with anti-rabbit IgG-horseradish peroxidase conjugate. Immunoreactive proteins were detected by enhanced chemiluminescence.
Labeling of ADP-Ribosyltransferase and NADase with [ 32 P]NAD-ADP-ribosyltransferases have been detected by zymographic assays using overlays with [ 32 P]NAD (34,35). In this highly sensitive procedure, proteins released from transformed NMU cells with PI-PLC, which contained transferase activity, and residual membrane proteins containing NAD glycohydrolase activity, were subjected to SDS-PAGE in a 14% gel and transferred to nitrocellulose. The membrane was incubated in 50 ml of buffer containing 25 mM Tris, pH 7.5, 100 mM NaCl, 0.3% bovine serum albumin, and 0.3% Tween 20 for 1 h at room temperature before addition of [ 32 P]NAD (100 Ci) to 25 ml of the buffer and further incubation for 5 h. The membrane was washed five times in the same buffer without bovine serum albumin and exposed to Kodak X-Omat film at Ϫ80°C for 18 h.
ADP-Ribosyltransferase Assay-Assays, in a final volume of 300 l, contained 50 mM potassium phosphate, pH 7.5, 20 mM agmatine, and 0.1 mM [adenine-U- 14  Protein Assay-Protein concentration was determined by bicinchoninic acid assay or Bio-Rad assay with bovine serum albumin as standard.

Synthesis of Full-length Yac-1 and Rabbit Muscle Transferase cDNAs in NMU Cells-NMU cells transformed with the
Yac-1 and rabbit muscle transferase cDNAs expressed ADPribosyltransferase and NAD glycohydrolase activities (Table I). Since these activities were not detected in the vector-transformed cells, both were attributed to transformation of the cells with transferase cDNA. From the transformants, ϳ93 (Yac-1) or 86% (rabbit muscle) of the transferase activity was released into the medium (PBS fraction) by treatment with PI-PLC. In contrast, after treatment of cells with PI-PLC, most of the NAD glycohydrolase activity was associated with the membranes. Although the ADP-ribosyltransferase expressed in transformed NMU cells was a GPI-anchored protein as previously characterized (9), the NAD glycohydrolase activity retained in the membrane fraction appeared not to be GPI-anchored.
Characterization of the ADP-Ribosyltransferase and NAD Glycohydrolase Activities-To elucidate the mechanisms responsible for the generation of both ADP-ribosyltransferase and NAD glycohydrolase activities in the transformed NMU cells, samples of PBS and pellet fractions, following treatment with PI-PLC, were applied to a gel permeation column. As shown in Fig. 1A, the transferase activity eluted from the column after ovalbumin (45 kDa), consistent with a protein of 38 kDa. The NAD glycohydrolase activity, solubilized with Triton X-100, eluted between thyroglobulin (670 kDa) and bovine serum albumin (68 kDa) and had an estimated molecular mass of ϳ100 kDa (Fig. 1B). The Yac-1 and rabbit muscle NAD glycohydrolase and ADP-ribosyltransferase activities, separated by column chromatography, were evaluated kinetically. Equivalent amounts of protein from vector-transformed NMU cells, subjected to gel filtration chromatography, did not demonstrate transferase or NAD glycohydrolase activities (data not shown). The apparent K m values for NAD in the ADP-ribosyltransferase reaction assayed with 20 mM agmatine were similar for the Yac-1 and rabbit muscle transferases, as were the K m values for NAD glycohydrolase activity, which were ap-proximately one-third those of the respective transferases (Table II).
Expression of Transferase and NAD Glycohydrolase Activities from Mutant Transformants-In contrast to NMU cells transformed with the full-length rabbit muscle transferase, cells transformed with truncated transferase cDNAs, lacking the amino-and carboxyl-terminal signal sequences (pM5Ј3ЈT) or the amino-terminal sequence (pM5ЈT), exhibited minimal transferase activity (Table III). Cells transformed with a transferase cDNA lacking the carboxyl terminus (pM3ЈT) released nearly all of the transferase activity into the culture medium. The carboxyl-terminal truncation may prevent the PI-PLCsensitive linkage from being formed, thus leading to secretion. In contrast, cells transformed with the full-length construct did not show an increased level of transferase in the medium. The NAD glycohydrolase activity of pM5ЈT-transformed cells was similar to that of control cells (vector alone), whereas activities of pM5Ј3ЈT-and pM3ЈT-transformants were, respectively, ϳ8.5 and ϳ4 times those of cells transformed with vector alone. The inability to detect a significant increase in NAD glycohydrolase activity with the cells transformed with the pM5ЈT may reflect enhanced degradation due to the presence of the carboxylterminal hydrophobic signal sequence, as was shown with other GPI-linked proteins (37). Similar to activities in cells transformed with the wild-type full-length transferase cDNA, the NAD glycohydrolase activity of cells transformed with the carboxyl-terminal truncated cDNA was present in the membrane fraction. These data suggest that proteins synthesized from cDNA lacking the carboxyl-terminal signal sequence or having the full-length sequence may be retained by the cells; these intracellular proteins exhibit primarily NAD glycohydrolase activity.
Immunoreactivity of Transferase and NADase Released from Full-length Rabbit Muscle Transferase Transformant-Proteins eluted from the Ultrogel AcA 44 column that possessed ADP-ribosyltransferase and NAD glycohydrolase activities were separated by SDS-PAGE under reducing (with DTT) or nonreducing (without DTT) conditions (Fig. 2). In fractions containing transferase, 38 -40-kDa immunoreactive proteins were observed under both nonreducing (lane 1) and reducing (lane 2) conditions. In fractions containing NAD glycohydrolase activity, two immunoreactive proteins (ϳ18 and ϳ30 kDa) were observed without (lane 3) or with (lane 4) DTT. Immunoreac-TABLE I Effect of PI-PLC on release of ADP-ribosyltransferase and NAD glycohydrolase activities from NMU cells transformed with the wild type Yac-1 or rabbit muscle transferase cDNAs NMU cells transformed with the Yac-1-or rabbit muscle transferase cDNAs or the pMAMneo vector were incubated for 60 min at 37°C in PBS without or with phosphatidylinositol-specific phospholipase C (PI-PLC, 0.5 units). After collecting the PBS fraction (0.7 ml), the cells were treated with lysis buffer and centrifuged (100,000 ϫ g, 1 h). The soluble fraction (supernatant (Sup), 0.5 ml) was collected, and the membranes were suspended in PBS (pellet, 0.5 ml). The PBS, supernatant, and pellet fractions were assayed for ADP-ribosyltransferase and NAD glycohydrolase activities. Total proteins (mg) were 0.43 and 0.42 (Yac-1-cDNA-transformed cells, ϪPI-PLC and ϩPI-PLC, respectively); 0.45 and 0.47 (rabbit muscle transferase cDNA-transformed cells, ϪPI-PLC and ϩPI-PLC, respectively); 0.47 and 0.48 (vector-transformed cells, ϪPI-PLC and ϩPI-PLC, respectively). Experiments were performed in duplicate. Data are means Ϯ S. E. (n ϭ 4). ND, not detectable.  7 and 8). These data are consistent with the hypothesis that NAD glycohydrolase activity results from proteolysis of the transferase.

Radiolabeling of Transferase and NADase with [ 32 P]NAD-
The ADP-ribosyltransferase released from transformed NMU cells following PI-PLC treatment and NAD glycohydrolase solubilized from the pellet fraction with detergent were subjected to SDS-PAGE, transferred to nitrocellulose, and incubated with [ 32 P]NAD (Fig. 3). 38-kDa proteins in the PI-PLC-released fraction (lane 2) and ϳ18 kDa proteins from the membrane fraction (lane 4) were detected by autoradiography. Radiolabel was not detected with proteins from vector-transformed cells (lanes 1 and 3).
Expression of GST-Yac-1 Fusion Proteins-Analyses of several bacterial toxin ADP-ribosyltransferases by x-ray crystallography, photoaffinity labeling with NAD ϩ , and site-directed mutagenesis demonstrated three regions with amino acid sequence similarities that form the catalytic site (24). The mammalian ADP-ribosyltransferases (e.g. rabbit (8) and human (9) muscle and mouse lymphocyte (13,14) transferases) also possess amino acid sequences similar to those in the active sites of the bacterial toxin transferases. In the bacterial toxin transferases, two structures have been proposed (24). In one, region I contains a nucleophilic arginine or histidine. Region II has closely spaced aromatic and hydrophobic amino acids, and region III is characterized by active-site glutamic acid, which is critical for enzyme activity. In the second, the region II domain is replaced with a serine (24). Amino acid sequence alignment and computer analysis of the mouse and rat RT6 proteins, which are T cell alloantigens that possess NAD glycohydrolase (rat) and ADP-ribosyltransferase (mouse) activities, suggest that, as in the bacterial toxin transferases (38), an analogous arginine or histidine (Arg-174 of Yac-1) and an acidic amino acid-containing region (Glu-233, 235 of Yac-1) are important for enzyme activity.
In the eukaryotic transferases, the catalytic site appears to be located in the carboxyl-half of the protein. To determine whether an ϳ18-kDa NAD glycohydrolase, as found in the cDNA-transformed NMU cells, could be generated from an ADP-ribosyltransferase, perhaps by release of a stable catalytic domain, truncated GST-Yac-1 transferase fusion proteins were synthesized in E. coli. Truncated fusion proteins, partially purified by glutathione-Sepharose 4B chromatography, were assayed for transferase and NAD glycohydrolase activities (Table  IV). Neither enzyme activity was detected in proteins from vector-transformed E. coli separated by glutathione-Sepharose chromatography. The GST-Yac-1 construct that lacks the hydrophobic amino-and carboxyl-terminal signal sequences (23 and 37 amino acids from amino-and carboxyl-terminal ends, respectively) exhibited ADP-ribosyltransferase activity; it had, however, negligible NADase activity. The GST-Yac-1-␦121 protein, which lacks 121 amino acids from the amino terminus, exhibited NAD glycohydrolase but negligible ADP-ribosyltransferase activity. The ratio of specific activities of NAD glycohydrolase and ADP-ribosyltransferase was 0.004 for GST-Yac-1-5Ј3ЈT and 183 for GST-Yac-1-␦121. The ratio of enzyme activities demonstrates that GST-Yac-1-␦121 has NAD glycohydrolase but not ADP-ribosyltransferase activity. The estimated molecular mass of a truncated Yac-1 enzyme corresponding to GST-Yac-1-␦121 is ϳ18.8 kDa. These results are consistent with previous data that NADase activity results from proteolysis of transferase and the generation of a stable, enzymatically active carboxyl-terminal fragment.

TABLE II
Kinetic constants for ADP-ribosyltransferase and NAD glycohydrolase activities obtained from Yac-1-and rabbit muscle transferase cDNA-transformed NMU cells Transformed NMU cells were incubated in PBS containing PI-PLC (0.5 unit). The PBS was collected, and the cells were lysed and centrifuged. Membranes were suspended in PBS and solubilized in Triton X-100 as described under "Methods." PBS and the solubilized membrane fractions were loaded on a gel filtration column as described under "Experimental Procedures." Fractions containing transferase and NAD glycohydrolase activities were used for determination of

DISCUSSION
Rat mammary adenocarcinoma (NMU) cells transformed with rabbit muscle and Yac-1 ADP-ribosyltransferase cDNAs possessed GPI-linked ADP-ribosyltransferase activity that was released from intact cells by PI-PLC. Transformed cells, however, also exhibited membrane-associated NAD glycohydrolase activity, not released by PI-PLC, which on solubilization and gel permeation chromatography appeared to be larger than the transferase. Based on the data from denaturing gels, however, it was concluded that this higher molecular weight species was an aggregate of an ϳ18-kDa protein or, alternatively, an aggregate containing the 18-kDa protein associated with other cellular components. The ADP-ribosyltransferases generated from transformants lacking amino-or and carboxyl-terminal signal sequences (pM5ЈT and pM5Ј3ЈT, respectively) should be TABLE III ADP-ribosyltransferase and NAD glycohydrolase activities in mutant rabbit muscle transferase cDNA-and vector-transformed NMU cells ADP-ribosyltransferase and NAD glycohydrolase activities were partially purified from NMU cells transformed with truncated transferase cDNAs as described under "Methods." Amounts of proteins (mg) were 1 and 1.1 (full-length, without and with PI-PLC, respectively); 1 and 0.97 (pM5Ј3ЈT, without and with PI-PLC, respectively); 0.97 and 1 (pM5ЈT, without and with PI-PLC, respectively); 0.93 and 0.91 (pM3ЈT, without and with PI-PLC, respectively); 1.1 and 1.05 (vector alone, without and with PI-PLC, respectively). Experiments were performed in duplicate. Data are means Ϯ S.E. (n ϭ 3). ND, not detectable. Sup, supernatant.  2. Immunoreactivity of proteins from transformed NMU cells. NMU cells transformed with the rabbit muscle ADP-ribosyltransferase cDNA were incubated with PI-PLC. PBS and membrane fractions, prepared as outlined in the legend to Fig. 1, were subjected to SDS-PAGE without (lanes 1, 3, 5, and 7) or with DTT (lanes 2, 4, 6, and 8) in 12% gel, transferred to nitrocellulose, and incubated with antirabbit muscle ADP-ribosyltransferase antibodies. Lanes 1 and 2, proteins (2 g) from the PBS fraction possessing ADP-ribosyltransferase activity after gel filtration without or with DTT, respectively; lanes 3 and 4, membrane proteins (20 g) possessing NAD glycohydrolase activity, separated by gel filtration without or with DTT, respectively; lanes 5 and 6, samples (20 g) from the PBS fraction, incubated without or with DTT, respectively; lanes 7 and 8, membrane proteins (100 g) from NMU cells transformed with the pMAMneo vector, without or with DTT, respectively. Positions of protein standards (kDa) are indicated on the left. unable to enter the endoplasmic reticulum (ER). Despite low levels of cell-associated transferase activity in cells expressing these truncated forms of the transferase, significant levels of membrane-associated NAD glycohydrolase activity resistant to release by PI-PLC were found. The transferase activity from pM3ЈT-transformant, however, was secreted into the medium due to the absence of the carboxyl-terminal hydrophobic signal sequence, which is essential for GPI anchoring (39). This protein would be transported into the ER but lacks the structure required for addition of the GPI anchor and retention on the cell surface. These data are consistent with the hypothesis that some of the ADP-ribosyltransferase produced by transformed NMU cells is processed to an NAD glycohydrolase. This observation was made with full-length transferase, a portion of which was retained as an NADase, and with the amino-truncated forms, both of which due to the absence of a signal sequence, were retained and proteolyzed. Reactivity of the 18-kDa protein with anti-transferase antibodies and radiolabeling of these proteins with [ 32 P]NAD support the conclusion that the NAD glycohydrolase arises from degradation of the ADP-ribosyltransferase.
Proteins destined for GPI anchoring have characteristic hydrophobic amino-and carboxyl-terminal signal peptides. The amino-terminal sequence is required for insertion of the nascent protein into the ER, whereas the carboxyl-terminal sequence is important for attachment of the mature protein to a GPI anchor through which it can be retained in the membrane (39). Field et al. (37) suggested that nascent proteins with unprocessed carboxyl-terminal signal peptides are present in micelle-like aggregates, which are retained in the ER and are subject to proteolytic degradation. The same process may have led to proteolytic degradation of the transferase and generation of the fragment possessing NADase activity.
To establish that a stable, catalytically active carboxyl-terminal fragment can be generated, a truncated transferase linked to GST was synthesized. The truncated recombinant Yac-1 ADP-ribosyltransferase protein (GST-Yac-1-␦121) exhibited NAD glycohydrolase but not transferase activity. The proteolyzed transferase demonstrated NAD glycohydrolase activity similar to that of the truncated GST-Yac-1-␦121 fusion protein. Because proteolysis generates an ϳ18-kDa fragment containing NADase activity and a GST-fusion protein containing an ϳ18-kDa carboxyl-terminal region of the transferase cDNA exhibits NADase activity, it appears that this portion of the protein is capable of forming a stable, catalytically active domain. This domain possesses NADase but not transferase activity, suggesting that the amino-terminal half of the protein is needed for effective interaction with the ADP-ribose acceptor. The fact that this domain is active is consistent with the model proposed for the rat and mouse RT6 proteins, in which arginine 146 rather than histidine 91 is the hydrogen-bonding amino acid involved in the formation of the catalytic site (38). In the former model for the catalytic site, a much smaller region of the protein is needed for activity than is the case with the latter. The synthesis of an active truncated protein also supports the hypothesis that the carboxyl-terminal half of the transferase can fold to form a stable domain. Since proteolysis of the transferase generated a stable NADase of a size similar to that of the recombinant truncated protein, it may be that, even in the intact transferase, the NADase domain is separated from the amino half of the protein and that proteases cleave between the two regions.