J Mol Evol (1995) 40:564-569

jou oFMOLECULAR N [EVOLUTION © Springer-VedagNew YorkInc. 1995

Nucleic Acid-Binding Metabolic Enzymes: Living Fossils of Stereochemical Interactions? Nikos C. Kyrpides, 1 Christos A. Ouzounis2 1 Institute of Molecular Biology and Biotechnology, Heraklion, Greece 2 European Molecular Biology Laboratory, Meyerhofstrasse 1, D-69012 Heidelberg, Germany Received: 3 June 1994 / Revised and Accepted: 3 August 1994

Abstract. Recently, a series of intriguing observations expanded the list of a number of metabolic enzymes known to be associated with various forms of nucleic acids, including single- and double-stranded DNA, cognate and noncognate RNAs, and specific tRNAs. There is no clear reason why such a phenomenon should take place in contemporary cell physiology, or, further, why such a property has evolved at all. Sixteen known cases are presented in an attempt to delineate any common features of these enzymes. Apart from their ancient nature, as judged by their wide distribution and their participation in fundamental biochemical pathways, it appears that these enzymes do not share any structural or functional characteristics. Given that most of these proteins require nucleotide-based cofactors for their activity, it is proposed that they may represent genuine molecular fossils of the transition from an RNA to a protein world. Their nucleic acid-binding properties are in keeping with previously proposed hypotheses regarding the origins and evolution of nucleotide-based cofactors. The mode of interaction between these proteins and their nucleic acid substrates remains unclear, but it may represent an extended form of stereochemical interactions that have been proposed for the origins of the genetic code. Key words: Metabolic e n z y m e s - - Stereochemical interactions - - RNA world

Correspondence to: C.A. Ouzounis

Introduction: A Very Old Class of Nucleic Acid-Binding Metabolic Enzymes Many proteins bind nucleic acids due to an intrinsic affinity for them, a feature from which their unique functional roles stem. Such proteins are transcription factors, developmental regulators, oncogene-associated proteins and of course enzymes involved in various genetic processes (Sigler 1991). Their interactions with genetic molecules are fundamental for the control of gene expression, and their emergence and universal presence can be reasonably attributed to natural selection. Recently, however, a whole " n e w " class of nucleic acid-binding proteins has been (re)discovered. These proteins are mostly single-domain metabolic enzymes, which appear to play double roles, both in the catalysis of their substrates as well as in transcription, translation, or transport of nucleic acids. Some of these cases have been recently reviewed as "RNA-binding proteins" (Hentze 1994). In this review, it was proposed that this "surprising, new" property suggests some structural relationship in these apparently unrelated enzymes, and a role for (di)nucleotide-binding domains in protein/RNA interactions (Hentze 1994). Here we present a different interpretation, discarding the argument for functionally equivalent domains. We suggest instead that this phenomenon is the first convincing manifestation of some form of ancient stereochemical interactions between enzymes and nucleic acids. These (long anticipated!) interactions are now demonstrated--albeit accidentally--for proteins least expected to interact with nucleic acids. We

565 discuss the characteristics of these enzymes, their ancestral nature, and the possibility that their nucleic acidbinding properties have been conserved over billions of years of biochemical evolution. Sixteen metabolic enzymes that bind nucleic acids are presented below. A general discussion of the common features of these enzymes in terms of reactions, metabolic pathways, coenzymes and substrates, and distribution among organisms follows. No distinction is made initially with respect to the type of nucleic acid binding in regulating gene expression (Maloy and Stewart 1993), but whether this division may bear any evolutionary significance is discussed. It is fortunate that for more than half of these proteins, their tertiary structures are known, allowing an analysis and prediction of interaction sites with nucleic acids in the future.

Sixteen Metabolic Enzymes That Bind Nucleic Acids The identified enzymes that bind DNA or RNA (mRNA, tRNA, or certain modified polyribonucleotides) are as follows, listed in order of their discovery. 1. 1,4-c~-D-glucan:l,4-c~-D-glucan 6-c~-(1,4-a-glucano)-transferase (or: 1,4-c~-glucan branching enzyme) (2.5S RNA) (Korneeva et al. 1979) 2. Glyceraldehyde-3-phosphate dehydrogenase (various) (Perucho et al. 1980; Singh and Green 1993) 3. Lactate dehydrogenase (ss-DNA) (Perucho et al. 1980) 4. Proline dehydrogenase (PutA) (ds-DNA) (Menzel and Roth 1981) 5. Biotin-protein ligase (BirA) (ds-DNA) (Cronan 1989) 6. Catalase (cognate mRNA) (Clerch et al. 1991) 7. Aconitase (mRNA) (Gray et al. 1993) 8. Thymidylate synthetase (cognate mRNA) (Chu et al. 1991) 9. Dihydrofolate reductase (cognate mRNA) (Chu et al. 1993) 10. Hexokinase (ds-DNA?) (Prior et al. 1993) 11. Nucleoside diphosphate (NDP) kinase (ds-DNA) (Postel et al. 1993) 12. Cl-tetrahydrofolate synthase (ss-DNA) (Wahls et al. 1993) 13. 3-oxoacyl-CoA thiolase (mRNA) (Nanbu et al. 1993) 14. Glutamate dehydrogenase (noncognate mRNA) (Preiss et al. 1993) 15. N A D + - d e p e n d e n t i s o c i t r a t e d e h y d r o g e n a s e (mRNAs) (Elzinga et al. 1993) 16. Fatty acid synthase (ss- and ds-DNA) (K~islin and Heyer 1994) From a mere handful of cases back in 1992 (while the first three cases had long been neglected), the list ex-

panded during the last year to 16 such observations, for various proteins and organisms.

Questions: Which, How, and Why? First, we briefly present the current knowledge for each case (which?); then we concentrate on the common characteristics of these enzymes (how?); and finally we examine some possible evolutionary scenarios (why?). We will see that these proteins form a wide spectrum of metabolic enzymes, with very few common properties.

Which ? 1. Probably the first demonstration for an enzyme to bind an RNA molecule was the identification of a 31-nucleotide (n0-1ong RNA bound to the 1,4-~glucan branching enzyme (1,4-o~-gBE) from rabbit muscle (Komeeva et al. 1979). This element has been shown to be essential for the activity of the protein, and therefore may not strictly be considered a nucleic acid interacting with the enzyme in a regulatory manner--it is a genuine coenzyme. However, this may signify a bridging element in the postulated continuum between nucleotide coenzymes and nucleic acid macromolecules. (See below.) Other references to enzymes containing ribonucleotides can be found in the same report (Korneeva et al. 1979). It is not clear whether any of these cases has been followed up in the succeeding years. 2. Curiously, a tRNA-binding protein from human cells has been identified as the enzyme glyceraldehyde-3-phosphate dehydrogenase (G3PD) (Singh and Green 1993), a glycolytic enzyme. It appears that G3PD recognizes specific tRNAs only. The most interesting aspect of this observation is that NAD + (the cofactor of G3PD) inhibits tRNA binding, suggesting a common site of interaction. The authors suggest that G3PD might be involved in transport of tRNAs from nucleus to cytoplasm (Singh and Green 1993). The earliest observation that G3PD binds homologous (cognate) and heterologous single-stranded (ss) DNA and RNA, however, was published 14 years ago (Perucho et al. 1980). Inhibition of nucleic acid binding by NAD and NADH had also been reported in that study (Perucho et al. 1980). This was perhaps the first proposition based on experimental evidence that enzymes with nucleotide-based cofactors have nucleic acid-binding properties. Similar experiments were performed for lactate dehydrogenase. (See below.) 3. In the same study cited above, lactate dehydrogenase was also found to bind ss-DNA (Perucho et al. 1980). In addition, similar inhibition of DNA bind-

566 ing was observed when NADH was added (Perucho et al. 1980). The interpretation was that a similar dehydrogenase dinucleotide-binding domain is responsible for nucleic acid binding. This proposal has never been experimentally proven. It is still not known how and why this interaction takes place. 4. A protein coded by the putA gene in enterobacteria is a bifunctional membrane-associated proline dehydrogenase/pyrroline-5-carboxylate dehydrogenase (PutA) and was found to regulate its expression antogenously (Menzel and Rotb 1981). PutA catalyzes the oxidation of proline to glutamate and is encoded by a single polypeptide. This enzymatic reaction requires FAD as a cofactor. Genetic and biochemical evidence suggest that PutA is an autogenous repressot of its own gene and the put operon (reviewed in Maloy and Stewart 1993). 5. Another similar case is biotin-protein ligase, coded by the birA gene in enterobacteria. This protein catalyzes the ligation of biotin to the biotin carboxy carrier protein via a biotinoyl-AMP intermediate (reviewed in Maloy and Stewart 1993). Mainly genetic evidence suggests that the BirA protein is a transcriptional repressor of genes for biotin biosynthesis (Cronan 1989). 6. Rat catalase has been found to bind its cognate mRNA and thus enhance mRNA stability (Clerch et al. 1991). A lung protein has been identified that forms complexes with catalase mRNA in a reversible form, controlled by redox agents (Clerch and Massaro 1992). The properties and identity of this protein are not known. Catalase is a protoporphyrin IX heme-binding protein that catalyzes the breakdown of hydrogen peroxide to water and oxygen with very high efficiency. Again, its role in mRNA binding is unknown. 7. The human IRE-BP (iron-responsive elementbinding protein) is a mRNA-binding protein with a well-established role in iron metabolism. It exerts its action through posttranscriptional and translational control over the transferrin receptor and the ferritin and erythroid 5-aminolevulinate synthase mRNAs, respectively. Recently, however, it was shown to be homologous with mitochondrial aconitase and identical to cytosolic aconitase (Gray et al. 1993). Aconitase is a well-studied enzyme of the TCA cycle of known three-dimensional structure. The active site coordinates the Fe-S cluster and participates in substrate recognition and catalysis. Enzymatic and RNA-binding activities are mutually exclusive, and interconversion between them is determined by intracellular iron concentration. Modeling by homology suggests that the RNA-binding site of IRE-BP is in spatial proximity to the aconitase active-site cleft (Basilion et al. 1994). 8. Human thymidylate synthase (TS) has been shown

to be autoregulated translationally (Chu et al. 1991). Thymidylate synthase catalyzes the conversion of dUMP to dTMP and dihydrofolate. The autoregulatory interaction is believed to play an important role in TS translation. Recently, it was found that this interaction is mediated by a 36-nucleotide-long RNA sequence region in the 5'-end of the TS mRNA (Chu et al. 1993). Interestingly, this region contains the AUG codon of the mRNA (Chu et al. 1993). In addition to the 5'-end region, another proteinbinding site was localized within the protein-coding region, a fact generating additional difficulties for proposals of coevolutionary optimization of proteins and RNA (Hentze 1994). 9. Human dihydrofolate reductase (DHFR), a key enzyme in purine biosynthesis, has been examined in vitro for cognate mRNA binding (Chu et al. 1993). It was found that not only does DHFR interact with its own mRNA, but this interaction is inhibited when DHFR is incubated with its substrates or an inhibitor, methotrexate (Chu et al. 1993). It is not clear how the interaction takes place, and whether the active site participates in mRNA binding, or substrate binding leads to a conformational change in DHFR, disabling it to bind its cognate mRNA. R is not known whether this interaction occurs in vivo and has a physiological role. 10. The expression of the RAG1 gene of Kluyveromyces lactis, coding for a glucose transporter, appears to be regulated transcriptionally (or posttranscriptionally) by three trans-acting genes, RAG4, RAG5, and RAG8 (Prior et al. 1993). It has been shown that RAG5 codes for the only hexokinase coded by the genome of this organism (Prior et al. 1993). Hexokinase (HXK) is a well-known ATP-dependent kinase involved in glycolysis. The fact that HXK in K. lactis seems to control the transcription of a glucose carrier gene is unexpected, but it may well have an important physiological role. The mechanism of this regulation remains unknown (Prior et al. 1993). It is not clear either whether RAG5 hexokinase is involved directly in DNA binding. 11. PuF (purine-binding factor) is a DNA-binding protein that binds to a nuclease hypersensitive element (NHE) on the human c-myc P1 promoter and is necessary for efficient P1 and P2 transcription initiation in vivo. Cloning of PuF revealed a 99% sequence identity to the human nm23-H2 gene (both the coding sequence and the 3'-UTR of PuF and nm23-H2 are identical) (Postel et al. 1993). The nm23-H2 gene has a nucleoside diphosphate (NDP) kinase function. NDP kinases synthesize GTP and other nucleotide triphosphates by transferring a phosphate group from ATP to the corresponding diphosphates. Thus the interesting possibility arises that the PuF/ nm23-H2 protein may act both as a transcription

567 factor that activates myc and as a NDP kinase providing the nucleoside triphosphates needed for DNA synthesis, before cell division. 12. "Serendipitously discovered" are the singlestranded (ss) DNA binding properties of C 1tetrahydrofolate (THF) synthases from various species (Wahls et al. 1993). Eukaryotic C1-THF synthases are trifunctional enzymes with three activities (synthetase, cyclohydrolase, and dehydrogenase). In bacteria, these properties are independently possessed by monofunctional or bifunctional enzymes. The C1-THF synthase from Schizosaccharomyces pombe is a trifunctional enzyme that has been shown to have sequence-independent ss-DNAbinding properties (Wahls et al. 1993). Using various enzymes, it was shown that the ss-DNA-binding property is confined within the synthetase domain of C1-THF synthases (Wahls et al. 1993). The role of this interaction potential is unknown. 13. Searching for factors that bind to the AU-rich element of the Y-UTR of sarcotoxin HA mRNA in Sarcophaga peregrina has led to the isolation of a protein with striking sequence similarity to the rat and yeast 3-oxoacyl-CoA thiolase, suggesting that this protein is a Sarcophaga thiolase (Nanbu et al. 1993). Furthermore, the purified Sarcophaga protein was tested for thiolase activity and was shown to have the same substrates and similar specific activity as the mitochondrial thiolase from rat (Nanbu et al. 1993). Finally, the rat mitochondrial thiolase was also tested for and was shown to bind specifically to AU-rich sequences in vitro (Nanbu et al. 1993). The authors suggested that these thiolases might participate in the metabolism of some cytoplasmic and mitochondrial RNAs by interacting with their pyrimidine-rich region or poly(A)-tract. 14. Bovine glutamate dehydrogenase (GDH) was shown to specifically bind to cytochrome c oxidase liver (L)-form mRNA (Preiss et al. 1993). For this particular substrate, no other proteins, including G3PD and lactate dehydrogenases (various forms), could bind the L-form cytochrome c oxidase mRNA, while GDH exhibited a high affinity for this transcript as compared to homopolymer, tRNA, and total cytosolic RNA binding (Preiss et al. 1993). GDH is a key enzyme involved in the deamination of glutamate to c~-ketoglutarate, a reaction linking various metabolic pathways. The role for this interaction remains also unknown. 15. The yeast NAD+-dependent isocitrate dehydrogenase (IDH), an enzyme of the tricarboxylic acid (TCA) cycle, was shown to bind 5'-untranslated regions (UTRs) of mitochondrial mRNAs (Elzinga et al. 1993). The authors suggest that IDH may be part of a general control network that links TCA cycle with gene expression in mitochondria (Elzinga et al. 1993). However, details of the mode of interaction

with mitochondrial mRNAs, as well as the physiological role of this binding property, remain obscure. 16. Finally, during this year, the fatty acid synthase (FAS) from Schizosaccharomyces pombe was shown not only to bind ss- and ds-DNA but also to mediate strand exchange in vitro (K~tslin and Heyer 1994). The authors suggest that it is unlikely that this function exists in vivo, an assumption remaining to be tested.

How? The obvious questions that arise from the above presentation of long-known and recently identified nucleic acid-binding metabolic enzymes are: Do these proteins have any common characteristics in terms of their (1) reactions, (2) pathways, (3) coenzymes and subtrates, and (4) distribution in organisms? If the answers to the above questions prove to be all negative, then we are faced with a very general phenomenon that requires a theoretical explanation. We may subsequently ask: How do these enzymes bind nucleic acids without specialized domains? Or, could these intermolecular forces, when they become known, unveil some aspects on the origins of these interactions? As far as the reactions catalyzed by the abovementioned enzymes are concerned, it is immediately obvious that they all belong to ancient pathways for energy metabolism, and nucleotide and amino acid biosynthesis (Table 1), while covering a broad range of mechanisms and pathways. This suggests that the nucleic acidbinding properties of these proteins have been conserved through long evolution, rather than discovered recently in individual cases. In fact, if this postulate is true, then similar nucleic acid-binding properties should be present in homologous proteins of the above-mentioned enzymes. If that does not hold true, then these properties should be regarded as acquired characteristics of singular cases late in evolution (Hentze 1994). Yet, for some of these cases, the nucleic acid-binding property is already known to be present in homologues, like aconitase, C aTHFS, and 3-oxoacyl-CoA thiolase. It is remarkable that these enzymes form certain groups as far as metabolic pathways are concerned. For instance, aconitase catalyzes the second step in TCA cycle and IDH the third one. If nucleic acid-binding properties are confined only in particular primordial pathways such as the TCA cycle and glycolysis, the prediction would be that, in the above case, at least the enzymes in reactions 1 and 4 (citrate synthase and c~-ketoglutarate dehydrogenase, respectively) also bind some forms of nucleic acid in similar fashion. Even more interestingly, most, though not all, of these enzymes have nucleotide-based derivatives as cofactors (Table 1). As mentioned above, this phenomenon appears to be an echo of the succession of biological sys-

568 Table 1. Nucleicacid-binding metabolic enzymes

Enzyme

Pathway

Coenzyme

Binds

1,4-a-gBE G3PDa LDHa PutA BirA Catalase Aconitase TS" DHFRa HXK NDP kinase CI-THFS Thiolase GDH IDH FAS

Glycogen biosynthesis Glycolysis, reaction 6 Fermentation Amino acid catabolism Biotin metabolism H202 breakdown TCA cycle, reaction 2 Thymine synthesis Nucleotide synthesis Glycolysis, reaction 1 NTP synthesis Amino acid synthesis Fatty acid metabolism Amino acid synthesis TCA cycle, reaction 3 Fatty acid synthesis

31-nt 2.5S RNA NAD÷ NAD+ FAD Biotinoyl-AMP Protoporphyfin IX Fe-S cluster dUMP NADPH ATP ATP ATP None? NAD(P)+ NAD(P)+ NADPH

Coenzyme DNA, RNA, tRNA DNA ds-DNA ds-DNA Cognate mRNA IRE (RNA) Cognate mRNA Cognate mRNA ds-DNA? ds-DNA ss-DNA Noncognate RNA Noncognate RNA Noncognate RNA ss- and ds-DNA

a Canonical cases: participate in ancient pathways, have nucleotide coenzymes,bind cognate mRNAs and manifest possible or proven inhibition of mRNA binding by substrates or inhibitors, as predicted (Kyrpides and Ouzounis 1993)

terns from an RNA-based world to a protein-based world (White 1976), as far as catalysis is concerned. The involvement of nucleotide-based cofactors in nucleic acidbinding metabolic enzymes only strengthens the proposal that this phenomenon is a relic of primordial interactions but is not sufficient to explain the property of nucleic acid binding. The proposed succession from ribonucleic acid to protein frameworks is manifested best in the wide occurrence of nucleofide-based cofactors that are still being used in catalysis (White 1982). An extreme prediction would be that all proteins that bind similar cofactors should have similar nucleic acid-binding properties. Finally, most of these proteins are distributed across all major domains of life, from Escherichia coli to Homo sapiens (data not shown). This ubiquitous presence again argues in favor of their ancient origins, and provides additional support for the primordial nature of these interactions.

Why ? Following our argument, it is evident that although these enzymes have very little in common on the basis of reactions and pathways, they represent ancient protein families, with ancestral genes present before the divergence of prokaryotes and eukaryotes. Further, the involvement of nucleotide-based coenzymes in their reactions seems to be a genuine shadow of an ancient R N A world, as was first predicted by White in 1976. The correct interpretation of the White hypothesis, however, is that these reactions were first catalyzed by R N A molecules, which were stabilized--and later superseded by proteins (White 1982). Thus, R N A binding represents the primitive state, while cofactor binding is an acquired property (and not vice versa). The fossils of this associ-

ation manifest themselves today as pyridine-based cofactors in these enzymes, and thus RNA/cofactor-binding sites are indeed expected to coincide. The new knowledge emerging from these examples is that some of these bifunctional enzymes not only bind a form of nucleic acid, but actually their cognate m R N A molecules. This autoregulatory self-association presumably represents an earlier, more primitive state, while diversification of protein/nucleic acid interactions may represent a later evolutionary step. The case of DNAbinding metabolic enzymes, among which "canonical" interactions do not seem to be present, perhaps reflects the genetic transition from the R N A to the D N A world, and later the functional transition from cis- (auto) to

trans-regulation. Thus, the importance of this resurrection is twofold: first, it provides evidence that specific stereochemical constraints are involved in binding of R N A (or DNA) to enzymes, and second, it strongly indicates that autoregulation may be a primitive mode of interactions which resulted in (instead of originated from) the "canonical" cases of these enzymes (RNA binding, autoregulation, competitive inhibition).

A n o t h e r F o r m of S t e r e o c h e m i c a l Interaction?

Almost one out of ten proteins is known to bind various forms of nucleic acids (C. Ouzounis, unpublished). However, searching for fossils of stereochemical interactions, one is always burdened by cases optimized during evolution to fulfill very specialized roles by binding to various nucleic acids. The few examples of nucleic acida s s o c i a t e d proteins that bind their o w n m R N A s (Kyrpides and Ouzounis 1993) may not be very convincing. The case of nucleic acid-binding metabolic enzymes, however, can be considered as an extreme man-

569 ifestation of stereochemical interactions, raising questions about their origins. It m a y w e l l be that there is an inherent interaction potential b e t w e e n ancient proteins and their m R N A s that has not b e e n a m p l i f i e d by e v o l u tion and therefore is not a historical chance event. C o g nate interactions are directly related to the p r o b l e m o f the origins o f c o d i n g s e q u e n c e s (H. B. W h i t e III, pers. c o m m . ) , w h i c h still remains an o p e n and u n r e s o l v e d issue (C. O u z o u n i s and N. Kyrpides, unpublished). Natural selection alone cannot sufficiently explain the fact that various m e t a b o l i c e n z y m e s b i n d nucleic acids. It is possible that this type o f interaction i n v o l v e s a f o r m of s t e r e o c h e m i c a l specificity similar in nature to the one b e t w e e n (anti)codons and a m i n o acids, p r o p o s e d for the o r i g i n s o f the g e n e t i c c o d e ( M e l c h e r 1974; R o o t Bernstein 1982). Therefore, an alternative possible v i e w regarding the origins o f such m e c h a n i s m is that these ancient proteins n e v e r escaped this m o d e o f interaction and that they still exhibit s o m e affinity for their cognate nucleic acids, or other regulatory elements (Kyrpides and O u z o u n i s 1993). This " e x t e n d e d s t e r e o c h e m i c a l hypothe s i s " (Kyrpides and O u z o u n i s 1 9 9 3 ) - - a s applied to autoregulation has correctly predicted the inhibition of cognate m R N A binding by substrates and inhibitors of e n z y m e activity (Kyrpides and O u z o u n i s 1993). W e expect the appearance o f m o r e such cases, and h o p e that a g e n e r a l i z e d solution to this intriguing p h e n o m e n o n will be proposed. Acknowledgments'. We thank Barry Hall (University of Rochester, New York, USA) for his support, Harold B. White III (University of Delaware at Newark, Delaware, USA) and Emile Zuckerkandl (Institute of Molecular Medical Sciences, Palo Alto, CA, USA) for their helpful comments, and finally Elsa Maniataki (IMBB, Heraklion, Greece) for her constructive criticisms.

References Basilion JP, Rouanlt TA, Massinople CM, Klausner RD, Burgess WH (1994) The iron-responsive element-binding protein: localization of the RNA-binding site to the aconitase active-site cleft. Proc Natl Acad Sci USA 91:574-578 Chu E, Koeller DM, Casey JL, Drake JC, Chabner BA, Elwood PC, Zinn S, Allegra CJ (1991) Autoregulation of thymidylate synthase messenger RNA translation by thymidylate synthase. Proc Natl Acad Sci USA 88:8977-8981 Chu E, Takimoto CH, Voeller D, Greta JL, Allegra CJ (1993a) Specific binding of human dihydrofolate reductase protein to dihydrofolate reductase messenger RNA in vitro. Biochemistry 32:4756-4760 Chu E, Voeller D, Koeller DM, Drake JC, Takimoto CH, Maley GF, Maley F, Allegra CJ (1993b) Identification of an RNA binding site for human thymidylate synthase. Proc Natl Acad Sci USA 90: 517-521 Clerch LB, Iqbal J, Massaro D (1991) Am J Physiol 260:L428-L433

Clerch LB, Massaro D (1992) Oxidation-reduction-sensitive binding of lung protein to rat catalase mRNA. J Biol Chem 267:2853-2855 Cronan J (1989) The E. coli bit operon: transcriptional repression by an essential protein modification enzyme. Cell 58:427-429 Elzinga SDJ, Berdnarz AL, Oosterun KV, Dekker PJT, Grivell LA (1993) Yeast mitochondrial NAD÷-dependent isocitrate dehydrogenase is an RNA binding protein. Nucleic Acids Res 21:5328-5331 Gray NK, Quick S, Goosen B, Constable A, Hiding H, Kuhn L, Hentze M (1933) Recombinant iron-regulatory factor functions as an ironresponsive-element-binding protein, a translational repressor and an aconitase. Eur J Biochem 218:657-667 Hentze (1994) Enzymes as RNA-binding proteins: a role for (di)nucleotide-binding domains? Trends Biochem Sci 19:101-103 Kaslin E, Heyer W-D (1994) Schizosaccharomyces pombe fatty acid synthase mediates DNA strand exchange in vitro. J Biol Chem 269:14103-14110 Komeeva GA, Petrova AN, Venkstern TV, Bayev AA (1979) Primary structure of the nucleic acid from the 1,4-0~-glucan branching enzyme. Eur J Biochem 96:339-347 Kyrpides NC, Ouzounis CA (1993) Mechanisms of specificity in mRNA degradation: autoregulation and cognate interactions. J Theor Biol 163:373-392 Malty S, Stewart V (1993) Autogenous regulation of gene expression. J Bacteriol 175:307-316 Melcher G (1974) Stereospecificity of the genetic code. J Mol Evol 3:121-140 Menzel R, Roth J (1981) Regulation of the genes for proline utilization in Salmonella typhimurium: autogenous repression by the putA gene product. J Mol Biol 148:21-44 Nanbu R, Kubo T, Hashimoto T, Natori S (1993) Purification of an AU-rich RNA binding protein from Sarcophaga peregrina (flesh fly) and its identification as a thiolase. J Biochem 114:432-437 Perucho M, Salas J, Salas ML (1980) Study of the interaction of glyceraldehyde-3-phosphate dehydrogenase with DNA. Biochim Biophys Acta 606:181-195 Postel EH, Berberich SJ, Flint SJ, Ferrone CA (1993) Human c-myc transcription factor PuF identified as nm23-H2 nucleoside diphosphate kinase, a candidate suppressor of tumor metastasis. Science 261:478--480 Preiss T, Hall AG, Lightowlers RN (1993) Identification of bovine glutamate dehydrogenase as an RNA-binding protein. J Biol Chem 268:24523-24526 Prior C, Mamessier P, Fukuhara H, Chert XJ, Wesolowski-Louvel M (1993) The hexokinase gene is required for transcriptional regulation of the glucose transporter gene RAG1 in Kluyveromyces lactis. Mol Cell Biol 13:3882-3889 Root-Bemstein RS (1982) On the origin of the genetic code. J Theor Biol 94:895-904 Sigler PB (1991) Protein-nucleic acid interactions. Curr Opin Struct Biol 1:61-62 Singh R, Green MR (1993) Sequence-specific binding of transfer RNA by glyceraldehyde-3-phosphate dehydrogenase. Science 259:365368 Wahls WP, Song JM, Smith GR (1993) Single-stranded DNA binding activity of Cl-tetrahydrofolate synthase enzymes. J Biol Chem 268: 23792-23798 White HB III (1976) Coenzymes as fossils of an earlier metabolic state. J Mol Evol 7:101-104 White HB III (1982) Evolution of coenzymes and the origin of pyridine nucleotides. In: Everse Jet al. (eds) The pyridine nucleotide coenzymes. Academic Press, pp 1-17