The MutT proteins or "Nudix" hydrolases, a family of versatile, widely distributed, "housecleaning" enzymes.

Our studies on the biochemical basis of spontaneous mutations took an interesting and unexpected turn when we discovered that a small region of amino acid homology between the MutT protein of Escherichia coli and the MutX protein of Streptococcus pneumoniae was involved in their nucleoside triphosphatase as well as their antimutator activities (1–3). Computer searches of the data banks revealed that this same small conserved region was present in a number of other proteins in organisms ranging from viruses to humans (2, 4). Most of these proteins containing the signature are coded for by open reading frames (orfs) whose products are either unidentified or of unknown function. We have been attempting, systematically, to identify and characterize enzymatic activities associated with these proteins, and it is now evident that nature has adopted this motif, originally identified as the active site of the nucleoside-triphosphate pyrophosphohydrolase of MutT (5, 6), and adapted it for use in many diverse reactions distinct from its function in the MutT protein. This short review summarizes our present knowledge of those reactions catalyzed by proteins harboring the MutT signature sequence and calls attention to a unique and versatile nucleotide binding and catalytic site. Although it might appear that the enzymes of this family act upon a wide variety of unrelated substrates, those characterized so far all hydrolyze a nucleoside diphosphate linked to some other moiety, X. For convenience, and to correct a misapprehension, we propose the mnemonic “nudix” hydrolase for this family of enzymes to replace the “MutT family.” Currently, this signature sequence is designated the “MutT pattern” in version 13.0 of the PROSITE data base of amino acid sequence motifs (7). This initial classification is misleading, because many, if not most, of these proteins are not involved directly in preventing mutations nor do they catalyze the archetypal nucleoside triphosphate pyrophosphohydrolysis reaction originally described for MutT itself (5, 6).

Our studies on the biochemical basis of spontaneous mutations took an interesting and unexpected turn when we discovered that a small region of amino acid homology between the MutT protein of Escherichia coli and the MutX protein of Streptococcus pneumoniae was involved in their nucleoside triphosphatase as well as their antimutator activities (1)(2)(3). Computer searches of the data banks revealed that this same small conserved region was present in a number of other proteins in organisms ranging from viruses to humans (2,4). Most of these proteins containing the signature are coded for by open reading frames (orfs) 1 whose products are either unidentified or of unknown function. We have been attempting, systematically, to identify and characterize enzymatic activities associated with these proteins, and it is now evident that nature has adopted this motif, originally identified as the active site of the nucleoside-triphosphate pyrophosphohydrolase of MutT (5,6), and adapted it for use in many diverse reactions distinct from its function in the MutT protein. This short review summarizes our present knowledge of those reactions catalyzed by proteins harboring the MutT signature sequence and calls attention to a unique and versatile nucleotide binding and catalytic site. Although it might appear that the enzymes of this family act upon a wide variety of unrelated substrates, those characterized so far all hydrolyze a nucleoside diphosphate linked to some other moiety, X. For convenience, and to correct a misapprehension, we propose the mnemonic "nudix" hydrolase for this family of enzymes to replace the "MutT family." Currently, this signature sequence is designated the "MutT pattern" in version 13.0 of the PROSITE data base of amino acid sequence motifs (7). This initial classification is misleading, because many, if not most, of these proteins are not involved directly in preventing mutations nor do they catalyze the archetypal nucleoside triphosphate pyrophosphohydrolysis reaction originally described for MutT itself (5,6).

The MutT Prototype
Characterization of the E. coli MutT protein, the progenitor of the family, followed from studies designed to elucidate the biochemical basis of the mutT mutator phenotype. Treffers et al. (8) originally described a mutant strain of E. coli, mutT1, with spontaneous mutation frequencies ranging from 100-to 10,000-fold higher than normal. Yanofsky et al. (9) showed that unlike typical defective mutator genes causing a variety of derangements in DNA such as transitions, transversions, frameshifts, etc., mutT causes, exclusively, a single, unidirectional AT 3 CG transversion. This could be explained by either or both of the following base mispairing events during DNA replication.
We were persuaded by the structural arguments of Topal and Fresco (10) that the A ϳ G mispair was the more likely event, and the experiments of Schaaper and Dunn (11) on in vitro DNA replication in extracts of mutT cells support this view. Cloning of the mutT ϩ gene by complementation of the mutT mutator phenotype (5,12) and identification of the lesion in the original Treffers' mutT1 allele as an IS1 insertion in the mutT ϩ gene (5) established that mutT ϩ is directly involved in preventing the enormous increase in the frequency of AT 3 CG transversions. Expression, purification, and characterization of the cloned gene product (5, 6) uncovered a new enzyme, a nucleoside triphosphatase with a preference for dGTP but active on all eight canonical nucleoside triphosphates. The stoichiometry of the reaction is as follows: dGTP ؉ H 2 O 3 dGMP ؉ PP i . Thus the enzyme is a nucleosidetriphosphate pyrophosphohydrolase, which most likely "sanitizes" the nucleotide pool (6) of a mutagenic form of dGTP having a proclivity to mispair with template adenine during replication. Maki and Sekiguchi (33) have reported that 8-oxo-dGTP is the mutagenic form of dGTP.

The MutT Signature
Studies on the structure and function of the MutT nucleoside triphosphatase were greatly enhanced by the discovery of MutX, a homologue of MutT present in S. pneumoniae (2). The mutX ϩ gene can complement a mutT strain of E. coli, and the purified MutX protein has similar nucleoside triphosphatase activity, yet the two enzymes differ markedly in their primary and quaternary structures (3). However, they do share a small region of homology involving about 10 conserved amino acids in a span of about 20. Site-directed mutants showed this region to be important for enzyme catalysis in vitro as well as for antimutator functions in vivo (2). When the human equivalent of E. coli mutT ϩ was cloned and sequenced (13), it was found to share this same homologous region as shown in Fig. 1A.
On the basis of the dGTPase activity of the MutT protein, it was proposed (2, 4) that the conserved motif in the MutT family designated a new catalytic site for the hydrolysis of nucleoside triphosphates, with a preference for dGTP. We elected to test this hypothesis by cloning other open reading frames of unknown function containing the conserved sequence, expressing and purifying the proteins and characterizing their activities. The first of these was orf17, a close neighbor to ruvC on the E. coli genome (14, 15) but not involved in the function of the ruvC-encoded Holliday-junction endonuclease (14). The Orf 17 protein was indeed found to be a nucleoside-triphosphate pyrophosphohydrolase, fulfilling the prophecy (3,16), but its preferred substrate among the canonical nucleotides is dATP: dATP ؉ H 2 O 3 dAMP ؉ PP i . dGTP is the least favored substrate for Orf 17 (relative catalytic efficiency, dATP/dGTP ϭ 10), whereas for MutT, the specificity is reversed (dGTP/dATP ϭ 70). Thus, the original working hypothesis, which included a preference for dGTP as part of the catalytic site, was revised, and we concluded that the conserved signature sequence designated a general nucleoside-triphosphate pyrophosphohydrolase whose specificity was determined by other amino acids outside of the homologous sequence. orf257 was cloned, and the expressed protein was purified to homogeneity. To our surprise (and discomfort), it was inactive on all eight canonical nucleoside triphosphates. We therefore considered the possibility that this family of enzymes catalyzed a general pyrophosphoryl transfer, wherein water was the acceptor for the previous enzymes (18), and that the proper acceptor was missing for Orf 257. Accordingly, we set up an NADH-linked indicator assay to follow, indirectly, the disappearance of nucleoside triphosphates and found that NADH was rapidly degraded in the absence of any added nucleoside triphosphates or putative acceptor. On further investigation (19), it was shown that Orf 257 represents a new member of the dinucleotide pyrophosphatases (EC 3.6.1.9) with a unique specificity for the reduced form of the co-factor (NADH/ NAD ϩ ϭ 100) and with an absence of activity on any of the nucleoside triphosphates. The reaction may be represented as follows: In addition to its unusual specificity, this is the only known metabolic pathway for the generation of NMNH. Thus, Orf 257 was not only an interesting enzyme, but it was also the harbinger of a series of new enzymes, since it broadened the scope of the reactions identified with the conserved sequence. Orf 1.9 -This open reading frame was so named because of its map position in a fragment of DNA from the cps region of E. coli K12 (20). The cps region is involved in the synthesis of the capsular polysaccharide known as M-antigen, or colanic acid, and orf1.9 is near cpsB and cpsG, which code for GDP-mannose pyrophosphorylase and phosphomannomutase, respectively. Orf 1.9 was purified from an expression system (21), and it, too, did not hydrolyze nucleoside triphosphates. Instead, it did have an atypical, sugarnucleotide hydrolase activity catalyzing the following reaction:

Additional Members of the Family
The release of the free sugar from sugar nucleotides is rare in metabolism. Only one other enzyme, yeast GDP-glucose glucohydrolase (EC 3.2.1.42), catalyzes a similar reaction (22), but its properties clearly distinguish it from the Orf 1.9 GDP-mannosyl hydrolase.
Orf 186 -This open reading frame was identified as a hypothetical 21.2-kDa protein in the E. coli genome-sequencing project. It was cloned, expressed, and purified 2 and is most active on Ap 3 A, hydrolyzing it as follows: Ap 3 A ؉ H 2 O 3 ADP ؉ AMP. It has substantial activity on NADH as well but does not hydrolyze nucleoside triphosphates. Rat liver cytoplasmic extracts have an active Ap 3 A triphosphatase (EC 3.6.1.29) (23), and E. coli has a related enzyme (24), but both can be distinguished from Orf 186 by their substrate specificities. E. coli contains an active Ap 4 A tetraphosphatase as well, the product of the apaH gene (25), but it, too, is distinct from Orf 186.
Orf 209 -This protein was also identified as part of the E. coli sequencing project and is located in the 65-68-minute region of the genome. It was cloned, expressed, and purified 3 and is highly specific for ADP-sugars, with its highest activity on ADP-ribose. Unlike Orf 1.9, Orf 209 catalyzes the hydrolysis of the pyrophosphate linkage: ADP-ribose ؉ H 2 O 3 AMP ؉ ribose 5-P. It hydrolyzes ADP-glucose and ADP-mannose at 80 and 40% of the rate, respectively, but it has little or no activity on the corresponding GDP-or UDP-sugars and no activity on the nucleoside triphosphates.
This concludes the list of E. coli genes bearing the conserved sequence uncovered so far. Of the six enzymes identified, two are nucleoside triphosphatases, and the remaining four catalyze the hydrolysis of a variety of substrates. How do these enzymes relate to those in higher organisms?
YSA1-Although several open reading frames from higher organisms harboring the conserved sequence had been deposited in the data banks (see Fig. 1B), the only non-bacterial members of this family for whom enzymatic activities had been described were the nucleoside triphosphatases from human (13), rat (26), and mouse (27), all homologues of E. coli MutT. We have recently cloned and expressed a gene, YSA1, from chromosome II of yeast, having the conserved sequence and which we find is closely related, in function, to Orf 209 of E. coli. 4 It prefers ADP-ribose but hydrolyzes ADP-glucose and ADP-mannose at about half the rate. Nucleoside triphosphates, Ap 4 A, and Ap 3 A are not substrates for the enzyme.
Ap 4 A Tetraphosphatase-This enzyme, an asymmetrical Ap 4 A hydrolase producing ATP and AMP from Ap 4 A, was purified from human placenta and characterized by Lazewska et al. (28). It has recently been cloned from placenta by Thorne et al. (29) and shown to have the conserved motif. Its high specificity for Ap 4 A and inactivity on Ap 3 A clearly distinguish it from Orf 186.
From these last two examples, it is evident that the members of this family are not confined to the bacteria and that additional activities will be identified with other Orfs from the eukaryotes listed in Fig. 1B.

Substrate Specificity
As pointed out in the Introduction, attention was focused on the small region of homology between the MutT and MutX proteins, because biochemical and genetic experiments identified this sequence of amino acids as important for enzymatic activity in vitro and also for the maintenance of normal mutation rates in vivo (2). From these two enzymes, the list of proteins has grown to 13, and many more, no doubt, will follow. Those identified, so far, are shown in Table I, along with their preferred substrates and the reactions they catalyze. All of these enzymes belong in the category, EC 3.6.1, the phosphoanhydrases (30). This large group of enzymes ranges from inorganic pyrophosphatase to plasma membrane calcium-transporting ATPase and includes all enzymes hydrolyzing a pyrophosphate linkage. The enzymes listed in Table I MutT enzyme, its functional homologues, and also Orf 17 make up one subset of the general class in which X is a phosphate group. The other six enzymes listed in Table I adhere to this pattern. They are all specific for nucleoside diphosphates but differ in their requirement for X. For convenience, we will refer to all of these enzymes as nudix hydrolases and to their homologous amino acid region as the nudix motif or nudix signature sequence.

Structure-Function Relationships
In Fig. 1B, a check marks the nudix signature sequences for those proteins so far identified with enzymatic activities. The availability (through genetic engineering) of large quantities of these proteins in a highly purified state provides the opportunity for comparative studies aimed at dissecting out their mechanisms of catalysis and their modes of substrate binding. Most of the studies, so far, have been done with E. coli MutT, because it was the first of the nudix hydrolases purified (5,6).

Tertiary Structure of MutT
The small size of MutT (129 amino acids) and its relative stability (3-6 days at 32°C) along with its abundance in engineered cells (35 mg of pure protein per liter of culture) make it ideal for analysis by heteronuclear multidimensional NMR spectroscopy. The threedimensional solution structure of MutT has recently been completed, and one view of it is shown in Fig. 2 (31). It consists of a 5-stranded, mixed ␤-sheet sandwiched between two ␣-helices connected by long loops. Of special interest are the positions of the amino acids of the nudix signature sequence. Situated mostly in loop I and helix I, they are closely arranged spatially and are readily accessible to the external environment. An examination of Fig. 1B reveals that of the several amino acid identities common to most of the proteins, only 4 amino acids are absolutely conserved in all of them. These are (for MutT) Gly-38, Glu-44, Arg-52, and Glu-57. Recently, the site-directed mutant, E57Q, has been constructed and shown to have at least 10 5 -fold lower activity than wild type (32).
It will be of interest to see if other members of the nudix hydrolase family share the structural and chemical features of the E. coli MutT prototype, especially the architecture of the nudix motif. Koonin (4) has noted that these proteins share a common loophelix-loop motif, and Thorne et al. (29) point out that a predictive analysis of human Ap 4 A hydrolase shows that the sequence 55 AL-RETQEEAG 64 (see Ap 4 Aase in Fig. 1B) is solvent-accessible and has a probability of 71% of being ␣-helical and flanked by loops. Two other members of the nudix family have been crystallized, the Orf 17 nucleoside triphosphatase, 5 and the Orf 1.9 GDP-mannose hydrolase, 6 and so our information on the generality of the structural features of the nudix motif should be forthcoming.

Physiological Function of the Nudix Hydrolases,
"Housecleaning" Enzymes? The enzymes discussed in this review have two features in common. They share a small region of homology herein referred to as the nudix signature sequence, and they all hydrolyze X-linked nucleoside diphosphates. Besides these two similarities, they seem to be widely disparate in their substrate preferences, which include nucleoside triphosphates, coenzymes, nucleotide sugars, and dinucleoside polyphosphates. This would suggest that these enzymes are involved in diverse metabolic pathways. An important clue in establishing the function of a biologic agent is an observable change, a phenotype, associated with its under-or overproduction, and that has been done for only one of the nudix hydrolases, MutT itself. In this case it seems fairly well established that the MutT pyrophosphohydrolase inactivates a potentially mutagenic form of dGTP, thus "sanitizing" (5) the nucleotide pool at the site of DNA synthesis. The very recent discovery of the rest of these nudix hydrolases has precluded extensive studies of their roles in metabolism, and mutants have not been identified that might be tied to some phenotype. In lieu of this, it is interesting to note a common feature of these enzymes; they all seem to hydrolyze potentially hazardous materials, or they prevent the unbalanced accumulation of normal metabolites. In the former category, we may include, in addition to MutT itself, Orf 17, Orf 257, Orf 186, and Ap 4 A hydrolase. For example, the Orf 17 dATPase could possibly play a similar role to that suggested for MutT's action on 8-oxo-dGTP (33) by hydrolyzing the recently reported, 2-hydroxy-dATP (34), that is remove a potentially mutagenic nucleotide from the pool. On the other hand, its role could be to prevent the accumulation of dATP 5   Thus, by analogy to the "housekeeping" or maintenance genes, those coding for the nudix hydrolases could be considered "housecleaning" genes whose function is to cleanse the cell of potentially deleterious endogenous metabolites and to modulate the accumulation of intermediates in biochemical pathways.
Several as yet unspecified proteins are of special interest and on further investigation could reveal novel activities as augured by our experience with those members of the family identified so far. For example, attention is called to proteins 15 and 19 (Fig. 1B). These Orfs are coded for by the antisense strand of the gene for basic fibroblast growth factor and have been identified in Xenopus (42) and humans (43). Also noteworthy is protein 20, from Bartonella bacilliformis, the causative agent of human Oroya fever. This protein is associated with the ability to invade erythrocytes (44). Orf 154 (protein 14) from pSAM2, an integrating plasmid in Streptomyces ambofaciens capable of autonomous replication (45), is most likely the homologue of pur7 involved in puromycin biosynthesis mentioned previously. We have cloned the gene and purified the protein and are currently examining its properties. 7 In addition to its metabolic significance, its presence in an autonomously replicating and integrating plasmid may provide some clue to the widespread dissemination of the nudix genes.
In their interesting article on the origin and evolution of enzymatic species, Petsko et al. (46) argue that nature recruits preexisting catalytic motifs and modifies them to perform related tasks. They state "Getting the chemistry right, it would seem, is the hard part; specificity is relatively easy to deal with later." From its widespread distribution and its diversity of application, it appears that the nudix signature sequence had the "right chemistry" and was one of the primordial catalytic motifs selected and adapted during evolution.