Evolution and Regulatory Role of the Hexokinases

Home , Fucokinase, Gluconokinase, Mannokinase, Ribokinase, Ribulokinase, Xylulokinase

Biochimica et Biophysica Acta 1401Ž. 1998 242±264

View metadata, citation and similar papers at core.ac.uk brought to you by CORE Review provided by Elsevier - Publisher Connector Evolution and regulatory role of the hexokinases MarõaÂÂ Luz Cardenas a, Athel Cornish-Bowden a,), Tito Ureta b a Institut FederatifÂÂ Biologie Structurale et Microbiologie, Laboratoire de Chimie Bacterienne, Â Centre National de la Recherche Scientifique, 31 chemin Joseph-Aiguier, 13402 Marseille Cedex 20, France b Departamento de Biologõa,Â Facultad de Ciencias, UniÕersidad de Chile, Casilla 653, Santiago, Chile Received 19 September 1997; revised 24 November 1997; accepted 27 November 1997

Keywords: Evolution; Hexokinase; Glucokinase

Contents

1. Introduction ...... 243 2. General characteristics of glucose-phosphorylating enzymes ...... 243 2.1. Isoenzymes ...... 243 2.2 Sugar specificity ...... 243 2.3 Specificity of the putative ancestral hexokinase...... 245 2.4 Hexokinases and transporters ...... 246 2.5 Nucleotide triphosphate specificity ...... 248 2.6 Molecular mass...... 248 2.7 Hexokinase DŽ. or glucokinase? ...... 249 3. Functional organization of the hexokinases ...... 250 3.1 Inhibition by glucose 6-phosphate...... 250 3.2 Two active sites on hexokinase B: kinetic consequences ...... 252 3.3 Inhibition of hexokinase C by excess glucose ...... 253 3.4 Functional adaptation of the rat isoenzymes ...... 254 4. Similarities in primary structure of hexokinases ...... 255 4.1 Rat isoenzymes ...... 255 4.2 Low molecular mass hexokinases ...... 257 4.3 Phylogenetic relationships between hexokinases ...... 258 4.4 Structural similarities with other proteins ...... 259 5. Acknowledgements ...... 261 References ...... 261

) Corresponding author. Fax: q33-491-71-89-14; E-mail: [email protected]

1. Introduction isoenzymes were first reported in rodent liverwx 14±18 , but appear to be characteristic of all animals, includ- The hexokinase-catalysed phosphorylation of glu- ing the humanwx 19,20 . Isoenzymes have also been cose by ATP occurs in all eukaryotic cells as the first found in green plantswx 21 and in several, but not all, step in the utilization of glucose, and the reaction is invertebrate species so far examinedwx 7 . Vertebrate also widespread in prokaryotic cells; the subsequent tissues contain up to four isoenzymes, designated as steps vary, as the glucose 6-phosphate formed in the hexokinases A, B, C and Dwx 18 on the basis of their first step may have different metabolic fates in differ- electrophoretic mobility; the alternative names hexok- ent types of cell and in different physiological condi- inases types I, II, III and IV, respectively, given by tions. Much information on hexokinases from differ- Katzen et al.wx 22 , are in widespread use; hexokinase ent species has accumulated since the pioneering DŽ. or IV is often called `glucokinase', but, as we work of Meyerhofwx 1 on yeast hexokinase, and there shall discuss in Section 2.7, this name is unfortunate are reviews on various aspects, including kinetics, because the four isoenzymes do not differ in their structure, and geneticswx 2±9 . specificity for glucose. Glucose is the preferred substrate of the hexoki- Comparisons between the isoenzymes from vari- nases, but they can also phosphorylate other hexoses ous sources has led several groups to suggest that to varying degrees, as recognized by the recom- evolution of the vertebrate hexokinases involved du- mended name of hexokinaseŽ ATP:D-hexose 6-phos- plication and fusion of an ancestral hexokinase of 50 photransferase, EC 2.7.1.1. . Only a few species, es- kDa that resembled the present-day yeast hexokinases pecially bacteria, are known to contain true glucoki- and mammalian hexokinase D in sizewx 4,7,9,23±26 . nasesŽ ATP:D-glucose 6-phosphotransferase, EC We shall discuss this question in Section 4. 2.7.1.2. , i.e., enzymes specific for glucosewx 7 . Hexokinases from different species differ in 2.2. Sugar specificity molecular mass and tissue distribution, and the enzyme often exists as a mixture of isoenzymes that It is widely believed that specificity has increased differ in kinetic characteristics and molecular mass. during evolution, i.e., that the more specific enzymes In this review, we first give a broad picture of the evolved from less specific ancestral proteinswx 27±29 , general characteristics of the hexokinases in different but one should not expect from this that the simpler phyla; we discuss whether a single kinetic model can modern organisms will have less specific enzymes be used to rationalize the kinetic and regulatory than the more complex organisms, because all mod- properties of the hexokinases; finally, we examine ern organisms have evolved to a high degree of similarities in primary structures and on the basis of efficiency in occupying particular ecological niches. sequence alignments propose a phylogenetic tree and As we will discuss in Section 3.4, at least three of the model of molecular evolution to explain it. four hexokinase isoenzymes in the rat have kinetic properties that correlate well with the functions of the organs in which they predominate, and there is no 2. General characteristics of glucose-phosphorylat- reason to doubt that a similar degree of adaptation ing enzymes exists in other organisms. Explanation for any differences in enzyme specificity between different types 2.1. Isoenzymes of organisms should therefore be sought in differences in their present-day requirements and not in Multiple hexokinases were first demonstrated in any supposed closeness to their ancestors. yeastwx 10 , which has three isoenzymes, hexokinases Examination of the glucose phosphorylating en- wx PIIIŽ. or A and P Ž. or B , and glucokinase 2,11±13 . zymes across the phylaŽ. Table 1 shows substantial The existence of distinct isoenzymes was shown not variation in hexose specificity, with the more specific only in diploid strains, in which hexokinases PI and enzymes found generally in the simpler organisms; PII could have been dismissed as allelic variants, but animals lack enzymes with high specificity for phos- also in haploid strainswx 12,13 . In animals, hexokinase phorylation of glucose, mannose or fructose at posi- 244 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264

Table 1 Specificity of hexokinases in different organisms Taxon High specificity Intermediate specificity Low specificity Archaeaa Pyrococcus furiosus wx30 Bacteriab Streptococcus mutans wx31 Leuconostoc meserentoides wx32 Escherichia coli wx33 E. coli wx34 Streptomyces Õiolaceruber wx35 Bacillus stearothermophilus wx36 E. coli wx37 Zymomonas mobilis wx38 Aerobacter aerogenes wx39 Rhodospirillum rubrum wx40 Pseudomonas saccharophila wx41 BreÕibacterium fuscum wx42 Lower eukaryotes Euglena gracilis wx43 Saccharomyces cereÕisiae w11,44 xSac. cereÕisiae wx45 Dictyostelium discoideum wx46 Torulopsis holmii wx47 Neurospora crassa wx48 Candida sp.wx 49 Saprolegnia litoralis wx40 Entamoeba histolytica wx50 Sap. litoralis wx40 Trypanosoma equiperdum wx51 Green plants Peaswx 52 Wheat w 53,54 x Animals Lobsterwx 55 Vertebrates w 4,56 x aThe Archaea were formerly called Archaebacteria. The more recent name reflects recognition that these organisms constitute a domain of life distinct from the bacteriawx 57 . bThe Bacteria were known as Eubacteria during the period when the Archaea were regarded as members of the same kingdomwx 57 . tion 6, whereas bacteria, apart from E. coli wx33 , have 2-deoxyglucose; however, unlike them and all other no unspecific hexokinases, and all of those that have known hexokinases it uses ADP as phosphate donor been described are either highly specific or at least Ž.and hence has AMP as a product and has no more specific than the enzymes of higher animals. detectable activity with ATP or other potential donors Prokaryotes and lower eukaryotes typically contain such as GDP or pyrophosphate. The authors suggest a series of specific hexokinases that each act on one that the specificity for ADP is related to the ability of hexose only, normally glucose, mannose or fructose; the organism to activate sugars after a period of in Pse. saccharophila, for example, glucokinase, starvation, i.e., in conditions of very low energy fructokinase and mannokinase exist as separate en- charge. By contrast, the thermoacidophile Sulfolobus zymeswx 41 , and various other organisms have spe- solfataricus appears to use ATP, but the enzyme cific glucokinases, such as E. coli wx34 , Z. mobilis responsible was studied only in cell homogenates, wx38,58 , B. stearothermophilus w36,59,60 x , Myxococ- and although these homogenates could also phospho- cus coralloides wx61 , S. mutans wx31 , A. aerogenes rylate fructose there was no information about whether wx39 , Bre. fuscum wx42 , D. discoideum wx46 . Some of the same enzyme was involvedwx 64 . these highly specific glucokinases, such as those of A specific glucokinaseŽ. EC 2.7.1.2 and a fructoki- S. mutans wx31 , A. aerogenes wx39 and D. discoideum naseŽ. EC 2.7.1.4 appear to exist also in Eug. gracilis wx46 , are not even inhibited by fructose or mannose, a strain Z grown either under autotrophic or hetero- further contrast with vertebrate `glucokinase', as dis- trophic conditionswx 43 . Unfortunately, the author of cussed in Refs.wx 62,63 . A few bacterial enzymes are this work did not mention whether other sugars, such somewhat less specific, such as the mannofructoki- as mannose, were tested, mannofructokinase and nase of L. meserentoides wx32 and the mannoglucoki- mannoglucokinase being not uncommon. The glucok- nase of Str. Õiolaceoruber wx35 . inase has a low affinity for glucose, with a reported

The recently described glucokinase from the hy- K m of 8 mM, reminiscent of the half-saturation con- perthermophilic archaeon P. furiosus wx30 is unusual centration value for the vertebrate `glucokinase', and and intriguing. Like the bacterial glucokinases, it is also of the value attributed to the N-acetylglucosa- highly specific for glucose, inactive with fructose, mine kinases that have often been misidentified as mannose or galactose, and only weakly active with glucokinasewx 65±67 . Nevertheless, it appears certain M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 245 that Eug. gracilis has no unspecific hexokinases with naseswx 31,39 ; one may ask whether this implies similar characteristics to those of plants and animals. major structural differences between the active sites The yeast Sac. cereÕisiae contains unspecific hex- of vertebrate and bacterial hexokinases. Z. mobilis okinases, two isoenzymes hexokinase PIII and P , that glucokinase is a special case, as it requires inorganic can phosphorylate keto- and aldohexosesŽ fructose, phosphate for activitywx 73 , suggesting that the active glucose and mannose. wx 2,45 , and a third isoenzyme, site for glucose, the first substrate, only becomes glucokinase, that is much more specific for aldohex- available after phosphate binding. No corresponding oseswx 11,13,44 ; it appears not to phosphorylate fruc- activation exists with eukaryotic hexokinases, and tose in vivowx 11 , and although it can phosphorylate when inorganic phosphate has an effect it is one of fructose in vitro, it does so with a high K m Ževen inhibition. For example, in rat hexokinase A phos- though the K m for glucose, about 0.03 mM, is the phate has a double action, counteracting glucose 6- lowest of the three isoenzymes. and a limiting rate phosphate inhibition at low concentrations, but be- only 0.4% of that for glucose, making it far more coming an inhibitor at higher concentrations; in hex- specific for glucoseŽ. in comparison with fructose okinase B it is always an inhibitorwx 26 . than the other two isoenzymes. This makes yeast a What selective advantages can explain why bacte- very different case from that of the rat discussed in ria have adopted highly specific hexokinases in their Section 2.7. metabolism with no isoenzymes, whereas animals use Unspecific kinases able to act on a wide variety of isoenzymes with broad specificity? This question has hexoses are broadly distributed, not only in verte- no clear answer at present, but there are several brates, but also in plant tissues such as wheat germ related points that may shed light on it in the future, wx53,54 , pea seeds wx 52,68 and the endosperm of such as the specificity of sugar transporters and the castor oil beans Ž.Ricinus communis wx69 , and are the availability of different sugars in the prebiotic rule in insects and echinodermsŽ for references, see medium. Microorganisms generally experience dras- Ref.wx 7. . Although specific kinases have been de- tic changes in the composition of the media, both scribed in Schistosoma mansoni wx70 and Echinococ- qualitative and quantitativeŽ changes from which ver- cus granulosus wx71 , confirmation of the reported tebrate cells are generally protected. , and, in addition, substrate specificity would be welcome. sometimes occupy very restrictive ecological niches, Vertebrate hexokinases typically act on mannose, but the short life-time of a typical bacterial cell fructose and 2-deoxyglucose as well as glucose, the implies several generations with little change in the preferred substrate; hexokinase D, the so-called `glu- range of sugars available. These two considerations cokinase', is no exception. In the rat, the most thor- will produce a higher selective pressure for speci- oughly studied organism, the four isoenzymes have ficity than that experienced by vertebrate cells. essentially the same relative specificity for glucose and fructosewx 72 , despite frequent suggestions to the 2.3. Specificity of the putatiÕe ancestral hexokinase contrary, as we shall discuss in Section 2.7. In summary, the highly specific hexokinases are One can imagine the ancestral prebiotic `cells' found in bacteria and unicellular eukaryotes, while functioning with limited enzyme resources, i.e., with the non-specific hexokinases are characteristic of a small number of different enzymes with a very higher eukaryotes. The non-specific hexokinase in E. broad specificity allowing them to react with a wide coli wx33 does not fit this pattern, and its subunit range of related substrates; at the same time, they molecular mass of 25 kDa is also unusual for a must certainly have been much less efficient as cata- bacterial hexokinase, as we shall discuss in Section lysts than present-day enzymes. Subsequent evolution 2.6. would have decreased the degree of substrate ambi- It would be useful to know whether the specific guity and improved catalytic efficiency, but because bacterial hexokinases are homologous with the verte- of constraints in the transition state it is not easy to brate enzymes. High specificity is not itself unusual, evolve an enzyme with both high activity and broad but substrate analogues are usually inhibitors, whereas specificity. Consequently, the degree of substrate am- this is not the case for some of the bacterial glucoki- biguity of all modern enzymes may well be much 246 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 less than that of primitive enzymes, so vertebrate does not reflect the properties of the early ancestral hexokinases ought to be considerably more specific enzymes, which were probably not only less specific than the enzymes present in the ancestral cells. than the present bacterial enzymes, but also less It is far from certain which sugars were available specific than the present vertebrate enzymes. It would in prebiotic conditions, but some predictions can be be interesting to know how much the hexokinases of made on the basis of the present knowledge of chem- present-day anaerobes differ from those of aerobic ical synthesis and stability of compounds. The abun- bacteria. dance of carbohydrate in the world today and its availability as a foodstuff for animals and microor- 2.4. Hexokinases and transporters ganisms results from the activities of plants, and does not reflect the situation at the origin of life, whether Glucose entry into most cells is mediated by facili- one assumes a prebiotic soup or any other type of tated diffusion, the transporter being an integral model.Ž In recent years alternatives to the classical membrane proteinŽ see Ref.wx 83 for a review. . Cer- model of the origin of life have been proposedwx 74 . tain hexokinases have been suggested to be involved However, these should not alter the validity of the in some way in the transport of glucose into the cell argument we now develop.. Žfor references, see Ref.wx 9. . This function would It is generally accepted that the primaeval soup presumably involve association with the plasma was relatively rich in amino acidswx 75 , carbohydrates membrane, though we do not know of any direct being much less commonwx 76,77 . Organic com- evidence for this. It is true that an unidentified hexok- pounds different from sugars may have been used inase in rat hepatomaŽ. but not in hepatocytes has preferentially in fermentation by early bionts. The been reported to associate with the plasma membrane prebiotic synthesis of sugars could have occurred wx84 , but in our view this has not been sufficiently through the formose reaction or Butlerow reaction studied to allow definite conclusions to be drawn. In Ž.polymerization of formaldehyde to sugars , which mammalian cells several different proteins are in- yields numerous sugars with no selectivity, producing volved with the transport of glucose and fructose. In almost all possible pentoses and hexoses, including the case of glucose there are at least five such branched chain sugarswx 78,79 . Under prebiotic condi- proteinswx 83,85 , Glut1, Glut2, Glut3, Glut4 and Glut7 tions, this reaction is not particularly efficientwx 80,81 , ŽGlut5 being a fructose transporter and Glut6 being and in view of the instability of sugars in aqueous non-functional. which share with the hexokinases the solutions at high temperatureswx 77 their concentra- characteristics of differing in tissue distribution and tions in prebiotic conditions must have been ex- glucose affinity. Glut2 has the lowest affinity for tremely low. Even aldohexoses like glucose and man- glucose Ž.K m about 15 mM and Glut1, Glut3 and nose, which appear to be more stable than aldopen- Glut4 have higher affinity, with K m about 2±5 mM. toses such as ribosewx 77 , are still quite unstable, They can transport also other sugars as fructose especially as the primitive ocean cannot have been Ž.Glut2 , 2-deoxyglucose Ž Glut1, Glut2, Glut3 and exactly neutral.Ž The instability of the pentoses and Glut4.Ž and galactose Glut1 and Glut3 . . In general, hexoses together with the low efficiency of the for- expression of the different transporters correlates with mose reaction for production of sugars suggests that expression of particular hexokinase isoenzymes: Glut1 glycolysis must have appeared relatively latewx 82 , as correlates with hexokinase A, Glut2 with hexokinase the sugars would not have initially been available in D, and Glut4 with hexokinase B, but at present there large enough quantities.. is no evidence that this is due to common mecha- In these conditions, selective pressure is likely to nisms of genetic regulation. have favoured enzymes with a relatively high affinity The case of Glut2 is especially striking: although it for sugars, but low specificity. Afterwards, the sugar is also expressed in other cells, it is highly expressed composition would have progressively changed as a in hepatocytes and b-cells of pancreas, the two types consequence of metabolic activity, inducing new se- of cells where the presence of hexokinase D has been lective pressures. This suggests that the high hexose established beyond doubtwx 62 . Moreover, it has a specificity of some of the present bacterial enzymes half-saturation concentration for glucose similar to M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 247

Table 2 The molecular dimensions of hexokinases Taxon Organism Isoenzyme Molecular massŽ. kDa Native Subunit Archaea P. furiosus wx30 Glucokinase 93 47 Bacteria B. stearothermophilus wx36 Glucokinase 67 34.5 S. mutans wx31 Glucokinase 41 24 E. coli wx33 `Glucokinase' 49 24.5 wx Fungi Sac. cereÕisiae PIII , P 87 102 51 Sac. cereÕisiae `Glucokinase'wx 88 aggregates 51 Protozoa Try. brucei wx89 Hexokinase 295 50.3 wx Plants Wheat germ LIII , L 53,54,90 50 50 wx Wheat germ HIII , H 53 100 Nematodes Hymenolepsis diminuta wx91 Hexokinase 98 Ascaris suum wx92 Hexokinase 100 100 Insects Drosophila wx93 Hexokinase A 47 47 Echinoderms Starfishwx 94 Hexokinase 50 50 Chordates Lampreywx 95 Hexokinase 90 90 Rat Hexokinase Awx 96 98 98 Rat Hexokinase Bwx 23 96 96 Rat Hexokinase Cwx 97,98 99.5 98 Rat Hexokinase Dwx 99 49 49

that of hexokinase D; both have low affinity for functional implications for mammals. Glucokinase glucose compared with other transporters and other from B. stearothermophilus is much smaller in size hexokinases. This raises the question of whether the than mammalian `glucokinase'Ž. hexokinase D , as hexokinase isoenzymes and the glucose transporters shown in Table 2, and has very different kinetic may be homologous, with associations between par- properties, especially in relation to specificity of sug- ticular isoenzymes and particular transporters to form ars and nucleotidesŽ. see Section 2.5 . functional units, with close physical interaction, that Lachaal and Jungwx 86 suggested that bacterial have co-evolved. Evidence will be needed, however, glucokinases and rat hexokinase D may share the to give credence to this idea, despite some attempts to same structural organization, but this seems most find interactions between transporters and hexokinase implausible on the basis of the kinetic properties and isoenzymes. the sequence comparisons shown in Figs. 3 and 5 For example, Lachaal and Jungwx 86 studied the Ž.below . They were probably influenced by the interaction between human Glut1 and glucokinase nomenclatural confusion, which implies a specificity from B. stearothermophilus in an in vitro system for the rat enzyme that it does not possess, and by the with the transporter incorporated into lipid vesicles. reported 22% sequence identity between Z. mobilis Although they observed binding, the affinity was glucokinaseŽ which is similar in size to the B. rather low, with a dissociation constant of 5 mMat stearothermophilus glucokinase. and rat hexokinase least an order of magnitude greater than the concen- Dwx 100 . However, the paper wx 100 did not make it tration of the enzyme in the cell. In addition, the clear how the value of 22% was obtained. The pro- maximum effect occurred in unphysiological condi- gram Align 1 shows only 14% identity, whereas tionsŽ pH 4 and rather high concentrations of ADP FastA 2 gives a somewhat different alignment with and glucose 6-phosphate. , and so the result must be treated with caution. In any case, the great differ- 1 Accessed through the Internet at ences in structure between bacterial and mammalian http:rrmolbiol.soton.ac.ukrcomputeralign.html. hexokinases, with no evident sequence similarityŽ see 2 Accessed through the Internet at Section 4.2. , makes it difficult to regard it as having http:rrgenome.eerie.frrfastaralign-query.html. 248 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264

17% identity. In both cases, these percentages refer stearothermophilus wx36 , in both cases isolated as Ž.as is usual to the longer sequence, which has 464 dimers; there are numerous other bacterial sugar ki- residues. However, if we recalculate them as percent- nases with molecular masses in the range of 32±37 ages of 327, the length of the Z. mobilis sequence, kDa, including fructokinases of Rhizobacterium legu- they become 20% and 24%, respectively, in better minosarum wx104 and Vibrio alginolyticus wx105 , fruc- agreement with the value of 22% given in Ref.wx 100 . tomannokinase of Fusobacterium mortiferum wx106 , This suggests that a more conventional way of ex- and ribokinase of E. coli wx107 . pressing the degree of identity found would be 15.5%, Apart from the hexokinase of the parasitic round- indicating no particular relationship. worm A. suum, with a reported subunit molecular mass of 100 kDawx 92 , fungal and invertebrate hexok- 2.5. Nucleotide triphosphate specificity inases have a monomer molecular mass of 50 kDa wx7 , although in a few cases they dimerize readily, as Although eukaryotic hexokinases prefer ATP as in yeast. Vertebrate hexokinases are monomers of the nucleotide substrate, bacterial enzymes appear 100 kDa, with the exception of hexokinase D, a less specific and ITP is a relatively good substrate. monomer of about 50 kDa that does not dimerize For example, glucokinase from B. stearother- wx99,108 . Thus, apart from the 35-kDa bacterial en- mophilus, which does not phosphorylate mannose, zymes the molecular masses fall in the geometric galactose, fructose, 2-deoxyglucose, glucosamine or series 25:50:100. For more details see Ref.wx 7 . xylose at concentrations of 1.0 mMŽ the apparent K m It was tempting to speculate from the molecular for glucose being 0.52 mM. , has K m values of 0.06 masses that the hexokinases of present-day organisms mM for ATP and 0.6 mM for ITP with similar Vmax might have derived from an ancestral gene specifying valueswx 36 . By contrast, rat hexokinase D, despite its a protein of about 25 kDa that is still present in some wx much broader sugar specificity, has K m 24-fold bacteria 7 . The initial proposal was that the ances- higher and Vmax 8-fold lower for MgITP than for tral gene duplicated and fused to give a new gene for MgATPwx 101 . With the other isoenzymes MgITP a hexokinase of about 50 kDa, this gene being still appears to be also a poor substratewx 102 . present in fungi and invertebrates, but that a further The specificity of the glucokinase from P. furiosus duplication and fusion at about the time of the origin for ADPwx 30 has been mentioned alreadyŽ Section of the invertebrates produced a gene for a protein of 2.2. . Another unusual example is provided by the about 100 kDa. For the 100-kDa hexokinases there is glucokinase from Propionibacterium shermanii,an good evidence for gene duplication and fusion, as we anaerobic bacterium; this enzyme, which has a shall discuss, but this hypothesis fails to account for molecular mass of 30 kDa, can use ATP as phosphate the 50-kDa hexokinase D of vertebrates. This isoen- donor, but is much more specific for polyphosphate zyme may be a relic of an invertebrate gene that wx103 . escaped the duplication±fusion event at the origin of vertebrates, and a similar enzyme may still therefore 2.6. Molecular mass be present in invertebrates; alternatively, the hexokinase D gene could have appeared afterwards as the As illustrated in Table 2, most native hexokinases consequence of the even fission of a gene coding for from different sources show molecular masses of 50 a 100-kDa enzyme. Against the first hypothesis, no or 100 kDa, with subunits, when they exist, also with enzyme similar to the vertebrate hexokinase D has molecular masses of 50 or 100 kDa. The archaeal ever been found in invertebrate organisms, but we are hexokinases also fall into this patternwx 30,64 , but not aware of any systematic search for such an smaller values have been found for several eubacte- enzyme. ria; for example, the enzymes of S. mutans wx31 and The possibility of gene duplication and fusion E. coli wx33 , have subunit molecular masses of 24 raises the question of whether the change from 50 kDa. A different glucokinase with a 34.5-kDa subunit kDa to 100 kDa corresponded to the acquisition of also exists in E. coli wx34 ; others of about 33 kDa new functions. For a long time, and until relatively have been purified from Z. mobilis wx58 and B. recently, it was believed that this change corre- M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 249 sponded to the appearance of an allosteric site re- Similar statements made in Ref.wx 118 have clearly sponsible for the strong inhibition by glucose 6-phos- misled other authors, such as Kengen et al.wx 30 and phate observed with 100-kDa vertebrate hexokinases, Hsieh et al.wx 119 , who quote them in contexts show- but not with hexokinase D or yeast hexokinases, all ing that they took the name `glucokinase' to have of them of 50 kDa. However, this attractive hypothe- implications of specificity. Other examples may be sis has now been disproved, as we shall discuss in found in the older literature: for example, Ishikawa et Section 3.1. al.wx 60 included the rat-liver enzyme among hexokinases that are `very specific for glucose'. Jang et al.wx 21 have recently studied two isoen- 2.7. Hexokinase D() or glucokinase? zymes from the green plant thale cress Ž Arabidopsis thaliana), and report that their sequences are, ``simi- Vertebrate hexokinase DŽ. or hexokinase type IV lar to those of the mammalian glucokinase and yeast has the typical specificity of a vertebrate hexokinase: hexokinase proteins but distinct from those of the it is not specific for glucose, but catalyses the phos- mammalian hexokinase proteins''. Given that all of phorylation of fructose, mannose and 2-deoxyglucose the mammalian hexokinase proteins are similar, it is as wellwx 62,63,72,102,109±113 . The practice of call- clearly impossible for this conclusion to be correct ing it `glucokinase' is very misleading, because it Ž.see Section 4.2, and Fig. 6 in particular ; such a suggests a correspondence with the true glucokinases statement could only have been made by authors of bacteria, or at least with the somewhat specific under the impression that mammalian `glucokinase' glucokinase from yeast. The only common point is not a hexokinase. Errors such as this arise directly between yeast glucokinase and vertebrate hexokinase from the nomenclatural confusion. D is the molecular massŽ and not even this similarity Hexokinase D differs from the other vertebrate applies to the true glucokinases of bacteria. . Both the hexokinases in having different promoters in liver sugar specificity and the kinetic characteristics are and pancreas, the tissue-specific alternative promoters quite different: vertebrate hexokinase D has a much allowing tissue-specific regulation of the gene. No higher half-saturation concentration for glucose than corresponding regulation has been described for the any of the yeast enzymes and is cooperative with hexokinases of low K m. Hexokinase D is encoded in respect to glucosewx 114,115 , a property not shown by the ratŽ. and in the human by a single-copy gene that any of the yeast isoenzymes. has 11 exonsŽ reviewed in Refs.wx 120,121. . The The name has more than just semantic importance, transcripts present in the liver and b-cells share despite suggestions to the contrarywx 116 , because sequences encoded by the last nine exons. The se- there is abundant evidence that researchers who are quences of the liver and b-cell mRNA differ at their X not specialists in vertebrate hexokinases have been 5 -ends because the gene has two promoters, one of misled into thinking that vertebrate hexokinase D which functions in b-cells and the other in the liver. corresponds to the real glucokinase in bacteria and As translation is initiated within the tissue-specific have created errors and confusion in discussions of first exon, the sequences of amino acids 1±15 of the hexokinase evolutionŽ e.g.,wx 21,117,118. . Griffin et liver and b-cell isoforms are different. al.wx 117 , for example, express surprise that, ``the rat Hexokinase D activity was not found in the livers glucokinase was most closely related to the C-termi- of 13 species from six orders of birdswx 122 , nor in 13 nal halves of the rat hexokinases whereas the yeast species of higher reptiles of the order Squamata glucokinase was most closely related to the yeast wx123 . However, no studies were performed in pan- hexokinases . . . This observation suggests that the creas and, as the expression of the hexokinase D gene duplication leading to rat glucokinase was a separate is tissue-specificwx 124 , the gene could be silent in Ž.and more recent event from the duplication leading liver but expressed in pancreas. This is a relevant to the yeast glucokinase, implying that the glucoki- problem because in the rat and humanŽ and presum- nase activities arose twice in evolution.'' Such a ably in all mammals. the enzyme appears to play an statement makes sense only on the assumption that its important role in insulin secretion, acting as a glucose authors think that hexokinase D is a true glucokinase. sensorwx 125 . Alternatively, the gene may be also 250 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 inactive in the pancreas of Aves and Squamata, glu- 50-kDa isoenzyme hexokinase D is not inhibited by cose-sensing being achieved by a different mecha- glucose 6-phosphate at physiological concentrations, nism resulting in the known differences between but only by fructose 6-phosphate via a complex with blood sugar levels in birds as compared to mammals; a separate regulatory proteinwx 128 , it was tempting to the fact that total pancreatectomy in birds, lizards and speculate that this regulatory protein could have snakes is followed by a prolonged period of hypogly- evolved similarly, with an allosteric site for fructose caemiawx 126 is also pertinent. The same questions 6-phosphate rather than for glucose 6-phosphate; the apply to the absence of hexokinase C from the livers duplication that produced the 100-kDa hexokinases of birds and higher reptiles. In this case, however, after subsequent fusion could have occurred without almost nothing is known about the role of this isoen- fusion to produce two separate molecules, hexokinase zyme in the utilization of glucose in any species. D corresponding to the C-terminal catalytic half of What has become of the silenced gene? Is it now a hexokinases A, B and C, and the regulatory protein pseudogene or is it merely permanently repressed? for hexokinase D corresponding to the N-terminal These questions could easily be answered with the regulatory half. This hypothesis has been very attrac- techniques now available, and the same ones apply to tive, but it can now be ruled out completely, for the several mammalian speciesŽ ruminants, felids and several reasons, as discussed more fully elsewhere bats. that lack hexokinase Dwx 127 . Considerable wx62,129 . The most serious objection to it is that clarification of the roles of the different hexokinase sequence data for the regulatory proteinwx 62,128,130 isoenzymes can be expected if the appropriate experi- do not support it at all, as its sequence appears to be ments are carried out. unrelated to that of any known protein. Until fairly recently, therefore the strong inhibition of hexokinase A by glucose 6-phosphate was thought 3. Functional organization of the hexokinases to be due to interaction with an allosteric site in the N-terminal half of the molecule, whereas the C-terminal half was responsible for all the catalytic activity. 3.1. Inhibition by glucose 6-phosphate ŽIt is perhaps worth remarking that the allosteric

hypothesis for the low-Km hexokinases has a long Strong inhibition by glucose 6-phosphate is often history, being proposed by Crane and Solswx 131 well said to be a striking property of the 100-kDa verte- before the term allosteric itself was introduced. . brate hexokinases that is not shared by the 50-kDa Nevertheless, doubts were raised by the fact that enzymes, whether hexokinase D or the yeast hexoki- some invertebrate 50-kDa hexokinases, such as nases. This idea, together with the evidence that the starfish hexokinasewx 94,95 , seem to be inhibited by catalytic activity was entirely in one half of the glucose 6-phosphate. 100-kDa molecule, suggested that the active site pre- It now seems clear from work of different groups sumably present in the other half at the time of the using different approaches that the C-terminal half of original gene duplication had evolved to become a hexokinase A provides both activities, not only catal- regulatory site: this would bind glucose 6-phosphate ysis but also high sensitivity to inhibition by glucose and account for the strong inhibition, and the absence 6-phosphate. For example, White and Wilsonwx 25 of the second half would explain why similar inhibi- reported that the C-terminal fragment of 51 kDa tion is not reported for the 50-kDa hexokinases. obtained by digestion with trypsin retained full cat- Consistent with this idea, only one catalytic site alytic activity, had separate sites for hexoses and appears to exist in hexokinase A, located in the hexose 6-phosphates, and was very sensitive to inhi- C-terminal half. Mutations of residues in the N-termi- bition by glucose 6-phosphate. They accordingly sug- nal half that correspond to catalytic residues in the gested a new view of the evolutionary relationship C-terminal half do not affect the catalytic activity of between the hexokinases whereby the sensitivity to the enzyme, and studies of molecular fragments pro- glucose 6-phosphate arose before gene duplication duced by proteolysis or genetic manipulation are and fusion. According to this view, the ancestral likewise consistentŽ references in Ref.wx 9. . As the 50-kDa ancestor of the 100-kDa mammalian hexoki- M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 251 nases resembled the present-day starfish hexokinase forms. It has long been known that the presence of wx94,95 more closely than the yeast enzyme. this effector causes hexokinase A to dissociate from The results of the report by White and Wilsonwx 25 the membranewx 135±138 , an observation that has have been confirmed by other groups. Magnani et al. been the basis of a simple and rapid purification of wx132 have reported the functional expression of the hexokinase A from brainwx 139 . Arora et al. wx 133 cDNA for C-terminal human hexokinase A in E. proposed that the N-terminal half might serve as a coli, and showed the recombinant half-hexokinase spacer between the membrane and the catalytic C- with only one glucose-binding site and only one terminal half of the enzyme, allowing this to interact putative ATP-binding site to be catalytically active more easily with the next glycolytic enzyme, i.e., and sensitive to inhibition by glucose 6-phosphate, phosphoglucose isomerase. exactly like the complete enzyme. They again con- An alternative to explain the inhibition by glucose cluded that the C-terminal half of human hexokinase 6-phosphate is to suppose that it is ordinary product A contains both the catalytic site and the regulatory inhibition. If this were the case one would expect site, and they argued that the evolutionary relation- some correlation between the inhibition constants Glc6P ships between the hexokinases should be modified to K i for glucose 6-phosphate and the correspond- Glc allow for the appearance of a regulatory site before ing half-saturation concentrations K 0.5 for glucose the gene duplication. across the range of isoenzymes, and Table 3 accord- Arora et al.wx 133 have evaluated the role of the ingly compares these. As the data come from work of N-terminal half of tumour hexokinase A by two different groups under non-identical conditions the different approaches: first, they used site-directed orders of magnitude rather than the exact values mutagenesis to assess the requirements of conserved should be considered. However, for each enzyme we Glc Glc6P amino acid residues predicted to interact with glu- have used K 0.5and K i values from the same cose; second, they characterized the overexpressed N- research groups. For some of the rat isoenzymes Glc and C-terminal halves of the enzyme. Replacing Ser- more recent values of K 0.5 are available, but using 155, Asp-209 and Glu-260, residues in the N-termi- these instead of those listed in Table 3 leaves the Glcr Glc6P nal half predicted to interact with glucose, with ala- ranking of K 0.5K i ratiosŽ right-hand column of nine residues did not affect either the catalytic activ- Table 3. essentially unchanged. ity or the inhibition by glucose 6-phosphate. The Analysis of these results shows that there is no overexpressed C-terminal polypeptide is catalytically difference between 50- and 100-kDa species in rela- active and shows the same inhibition pattern as the tion to glucose 6-phosphate inhibition, and the differ- complete 100-kDa parent enzyme for inhibition by ence between hexokinase D and the other three mam- 1,5-anhydroglucitol 6-phosphate, an analogue of glu- malian isoenzymesŽ especially hexokinase A, which cose 6-phosphate. Kinetic analysis revealed that this has essentially the same ratio. is much less than one analogue is linearly competitive with the substrate would suppose from the common view that hexoki- ATP both in the intact enzyme and in the C-terminal nase D is not subject to inhibition by glucose 6-phos- half, with K i of 88 mM for the intact enzyme and phate. Of course, the inhibition of hexokinase D by 118 mM for the C-terminal half. However, the N- glucose 6-phosphate must be physiologically irrele- terminal half was also able to bind ATP and glucose vant, but that is not the point here: what is important Glcr Glc6P 6-phosphate. is that as K 0.5K i value for hexokinase D is Arora et al.wx 133 concluded that the N-terminal similar to the others there is no reason to postulate part of hexokinase A might modulate the binding of any special explanation for the strong inhibition of the enzyme to mitochondria. There is a highly con- the low-Km isoenzymes. Rather one should see the served dodecapeptide at the beginning of the N- weak binding of glucose 6-phosphate to hexokinase terminal sequencewx 117 responsible for the interac- D as a natural consequence of the much weaker tion with a porinwx 134 . On the other hand, there is a binding of glucose to this isoenzyme than to the binding site for glucose 6-phosphate that controls the others. binding of the brain enzyme to mitochondria, regulat- All of these considerations make it unreasonable to ing the relative levels of soluble and membrane-bound exclude the possibility that all enzymes catalysing the 252 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264

Table 3 Inhibition of different hexokinases by glucose 6-phosphate a b Glc c Glc6P c Glcr Glc6P Hexokinase Molecular massŽ. kDa K0.5 Ž.mM K i Ž.mM K 0.5K i Rat hexokinase Cwx 97 100 0.007 0.92 0.008 H. diminuta wx91 50 0.09 5.2 0.02 Rat hexokinase Dwx 140 50 5 60 0.08 Lobster hexokinase IIwx 55 50 0.08 0.8 0.1 wx Wheat hexokinase LII 141 50 2.5 16.5 0.15 Rat hexokinase Awx 97 100 0.045 0.21 0.21 Starfishwx 95 50 0.045 0.07 0.64 A. suum wx92 100 4.7 3.4 1.38 Rat hexokinase Bwx 97 100 0.23 0.16 1.44 Sea urchinwx 95 50 0.35 0.06 5.8 aThe enzymes are ordered according to the ratios shown in the right-hand column. bTo facilitate comparison the molecular masses shown are approximateŽ. for more precise values see Table 2 . c Glc6P For each hexokinase, both kinetic constants are from the same group; all were obtained from kinetic experiments apart from the Ki wx for wheat hexokinase LII 141 , which was obtained fluorimetrically. phosphorylation of hexoses follow essentially the terminal half, discussed in Section 4.1, argues against same mechanism, and that the range of types of the maintenance of a functional active site in this half kinetic behaviour derive from quantitative differences of the molecule. in rate constants. Nonetheless, recent work by two different groups BlasquezÂ et al.wx 142 have reported that the yeast wx143±145 has produced just such a surprise in the hexokinases, PIII and P , but not the third isoenzyme case of hexokinase B, which does appear to have a Ž.glucokinase , are sensitive to inhibition by trehalose functional active site in the N-terminal half of the 6-phosphate. This could play a role in the regulation molecule. Not only complete human hexokinase B, of hexokinase activity, as the intracellular concentra- but also its N- and C-terminal halves, have been tion of trehalose 6-phosphate has been estimated to expressed as separate molecules in E. coli wx143 . be in a range close to the observed inhibition con- When extracted, each of the three molecules proved stant. From the point of view of evolution it will be to be catalytically active, with a specific activity for interesting to have information about the effect of the complete enzyme of 147 unitsrmg, not very this ester phosphate in hexokinases of other species. different from the sum of the separate values for the N- and C-terminal halvesŽ 94 and 60 unitsrmg,

respectively. ; the K m values for glucose were similar 3.2. Two actiÕe sites on hexokinase B: kinetic conse- for the three cases, 0.34, 0.46 and 0.51 mM, respec- quences tively, with somewhat greater differences between the

three K m values for ATP, 1.02, 0.78 and 3.8 mM, For many years it appeared that despite the dimer- respectively. Ardehali et al.wx 143 concluded that the like structures of the 100-kDa mammalian hexoki- activity of the whole enzyme might be a combination nases each contained only a single active site capable of the activities of the two halves. of catalysing the phosphorylation of glucose. This Although this interpretation appears reasonable, wx still seems to be true of hexokinases A and C 25,26 , some caution is appropriate because unless the K m and it would be surprising if it turns out not to be: in values are the same the kinetics of a mixture of the case of hexokinase A so much work has been Michaelis±Menten activities should not follow done that it is hard to believe that the existence of a Michaelis±Menten kinetics exactly, as pointed out by second catalytic site on the molecule could have Tsai and Wilsonwx 144 in the context of their similar escaped all efforts to detect it until now; hexokinase results; why, therefore, were deviations from C has been less intensively studied, but the very rapid Michaelis±Menten kinetics not detected for hexoki- accumulation of amino acid substitutions in its N- nase B in the past, even though the known non- M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 253

Michaelis±Menten behaviour of hexokinases C and is absent from many tissues, or present only in low Dwx 63 had led various investigators to look for them? quantities, and because its physiological role is less Calculation of the appearance of the plots expected clear. The distinctive feature of this isoenzyme, which from the parameters given shows that the expected occurs in many other vertebrates as well as mammals deviations from Michaelis±Menten kinetics with re- wx4 , is the substrate inhibition by excess glucose. This spect to glucose are so small that they would be may be related to the product inhibition of the other virtually undetectable. Similar calculationsŽ with the two 100-kDa isoenzymes by glucose 6-phosphate, same conclusion. were done by Tsai and Wilson and it is worthwhile to examine whether variants on a wx144 with their own data. They studied complete common mechanism can account for both phenom- molecules of hexokinase B in which mutations in one ena. or other half eliminated one or other active site, and Let us postulate that the ancestral hexokinases they found less variation in the kinetic constants Ž K m followed a random-order mechanism in which either values for glucose all about 0.14 mM, and values of glucose or MgATP could bind to the enzyme as a

0.64, 1.09 and 0.45 mM for the K m for ATP with the first substrate; the binary complex produced could wild-type, C-terminal active and N-terminal active then react with the other substrate to produce a molecules, respectively. than was found with the catalytically active ternary complex. In fact, the half-moleculeswx 143 considered above. mechanism followed by yeast hexokinase has a sub- wx The somewhat larger differences between the K m stantial random component 2 , and, although hexoki- values for ATP suggests that somewhat greater devia- nase D appears to react very largely with glucose tions from Michaelis±Menten kinetics with respect to binding as first substrate, there is evidence that the ATP might have been observed for the complete binary complex of the enzyme with MgATP can also enzyme. However, they would still be small, and if lead to productswx 152 . This example suggests an they had been observed they could have been at- evolutionary trend towards a more ordered pathway tributed to the difficulties of controlling the concen- in which the less favoured binary complex either tration of the true substrateŽ MgATP2y. without af- disappears from the mechanism or becomes a dead- fecting other ions liable to affect the kineticswx 62 . end complex unable to continue along the catalytic More recent studies by Tsaiwx 146 , at present pub- route. If hexokinase C differed from the other isoen- lished only in abstract, set out to reconcile the exis- zymes in having a preferred order with MgATP tence of two active sites on hexokinase B with the before glucose, the same evolutionary hypothesis observation that it can be inactivated by a single would predict inhibition by glucose in a dead-end molecule of N-bromoacetylglucosaminewx 147 , imply- reaction. Moreover, the inhibitory concentration of ing that glucose binds with a 1:1 stoichiometry. He glucose should vary with the MgATP concentration, proposes that although glucose can bind to either of as indeed is observed: with hexokinase C of ascites the two sites to give a catalytically active complex, tumour the concentration of glucose needed for inhi- the conformational change that accompanies glucose bition increases as the MgATP concentration in- binding affects both sites and thus prevents binding creaseswx 153Ž. Fig. 1 . For example, at 0.2 mM of more than one glucose molecule at a time. It MgATP glucose concentrations above 0.1 mM inhibit remains to be seen whether this type of model can strongly, whereas at 10 mM MgATP the inhibition is account for the flux-ratio data for hexokinase Bwx 148 , weaker, with a maximum rate at 0.2±0.5 mM glu- which have generated some controversywx 149±151 cose. The behaviour of frog hexokinase C is similar without leading to any fully convincing conclusion. wx154 . The corresponding phenomenon for the hexokinase 3.3. Inhibition of hexokinase C by excess glucose isoenzymes that prefer to bind glucose first would be substrate inhibition by MgATP, but this has not been The mammalian hexokinases A, B and D, the reported. However, inhibition by glucose 6-phosphate characteristic isoenzymes of brain, muscle and liver, is competitive with respect to MgATP, not, as one respectively, have been studied much more thor- might naively expect, with glucose, suggesting that oughly than hexokinase C, probably because this last glucose 6-phosphate can bind to the MgATP binding 254 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264

Some detailed kinetic studies of Z. mobilis glucokinasewx 73 reveal some major differences in kinetic behaviour from that observed with the mammalian enzymes, as noted in Section 2.2; in addition, as discussed in Section 4.2, there is little or no evidence of structural similarity between the bacterial and eukaryotic enzymes. If this is maintained it suggests that little could be learned about the mechanisms of the eukaryotic enzymes by studying the kinetics of those of bacteria.

3.4. Functional adaptation of the rat isoenzymes

Fig. 1. The effect of ATP on the inhibition of hexokinase C by Hexokinase A is the predominant isoenzyme in rat excess glucose. The figure shows data for the dependence of the brain, hexokinase B predominates in muscle, and rate of glucose phosphorylation catalysed by hexokinase C at three different concentrations of ATPwx 153 . Adapted from Ref. hexokinase D in hepatocytes and pancreatic islets. wx153 . The kinetic properties of these three isoenzymes are well adapted to the presumed roles of glucose phosphorylation in the different locations, as may be seen site or, at least, to the same conformation to which in Fig. 2. Both hexokinases A and B are virtually MgATP binds. When MgATP binds in this site it saturated at glucose concentrations in the normal does not prevent glucose from binding to produce a physiological range for the blood, which is indicated catalytically active ternary complex, but when glu- by the shaded region of the figure, and thus their cose 6-phosphate binds to the MgATP site, glucose is excluded and inhibition results. In the case of hexokinase D, glucose is clearly the first substrate, with only about 1% of the reaction occurring with MgATP binding first. Thus, MgATP binds to the free enzyme with much less affinity than glucose and glucose 6-phosphate does likewise, and no inhibition by either MgATP or glucose 6-phosphate is evident at low physiological concentrations. The general hypothesis is therefore that if MgATP can bind as first substrate then glucose 6-phosphate can compete with it for the binding site of the g-phosphate and prevent catalysis. An apparent difficulty with this idea is that the isotope-exchange studies of Gregoriou et al.wx 148,152 provided no evidence for the existence of dead-end complexes. Fig. 2. The correlation of kinetic properties with physiological This difficulty is, however, apparent and not real, roles. The figure compares the kinetic properties of the four isoenzymes of hexokinase found in the rat. The physiological because dead-end reactions are completely invisible range in the blood is indicated by the shaded region. The three to the flux-ratio method used in these studies, i.e., the insets show the curves redrawn with a linear instead of a logarith- fact that they were not observed provides no evidence mic scale of glucose concentration, and provide a qualitative either for or against their existence. illustration of the different kinetics: hexokinases A and B follow Light could perhaps be shed on these ideas by Michaelis±Menten kinetics; hexokinase C is inhibited by excess glucose; the curve for hexokinase D is somewhat sigmoid, with a kinetic studies of the 25-kDa bacterial hexokinases, Hill coefficient of 1.6, this sigmoidicity accounting for the greater such as those of S. mutans and E. coli, but unfortu- steepness of the curve for hexokinase D in the main part of the nately this information is not at present available. figure. M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 255 kinetic activity is largely unaffected by variations 4. Similarities in primary structure of hexokinases within this range; because of this saturation deviations from Michaelis±Menten behaviour would have 4.1. Rat isoenzymes no significant effect, and in fact both isoenzymes follow Michaelis±Menten kinetics. However, when Close structural relationships of hexokinase D with the availability of glucose is pathologically low, it is the other vertebrate hexokinases have long been well more important to satisfy the glucose needs of the documented. For instance, the amino acid composi- brain than those of other tissues, and it is significant tions of the four isoenzymes were found to be strik- that the low K m of hexokinase A allows it to retain ingly similar, especially those of hexokinases B and more than 70% of full activity at glucose concentra- Dwx 4,158 , and the analysis of the peptide maps of tions as low as 0.1 mM. If, as we have argued in hexokinases B and D confirmed the resemblance, as a Section 3.1, the strength of product inhibition by significant number of common peptides were found glucose 6-phosphate is largely determined by the wx159 . More direct support for similarity between the affinity for glucose, then the low K m values for sequences of the N- and C-terminal halves came glucose are well adapted to the need for hexokinases later, from isolation and sequencing of several tryptic A and B to be sensitive to accumulation of glucose peptides from hexokinase C and comparison of these 6-phosphate. with yeast hexokinasewx 160 . As seen also in Fig. 2, the kinetic behaviour of When the sequences became known of hexoki- hexokinase D is very different, but this also accords nases Awxwx 161 , B 162 , and C wx 163 , as well as those well with the needs of the liver and pancreas to of hexokinase D from both rat liverwx 164 and human respond to variations in the blood-glucose concentra- pancreaswx 165 , the general similarity between all the tion. The small degree of sigmoidicity in hexokinase rat enzymes was confirmedŽ though without any spe- Dwx 114,115 , coupled with its relatively low affinity cial similarity between hexokinases B and D. , and the for glucose, allows it to have a middle range of hypothesis that the 100-kDa hexokinase could be the activity almost exactly the same as the physiological result of gene duplication and fusion was reinforced. range of blood glucose concentrations. If a side-effect The sequence similarities are illustrated in Fig. 3, of the low affinity for glucose is that hexokinase D is which represents the dot plots corresponding to the virtually unaffected by glucose 6-phosphate at any four rat hexokinases. Several important points are reasonable physiological concentration then this also evident in this figure. First, the 100-kDa hexokinases corresponds with the probable physiological needs. are all dimer-like, i.e., in each case the N- and This picture will probably need to be complicated C-terminal halves of the molecule resemble one an- somewhat in the light of recent observations on the other, as is clear from the strong secondary diagonals translocation of hexokinase D between the nucleus in each comparison of a 100-kDa molecule with any and the cytoplasmwx 155±157 , but it would be prema- other. However, none of the enzymes shows any ture to attempt this now. convincing evidence of a 25-kDa unit, and if there The kinetic behaviour of hexokinase C is much was ever a gene duplication and fusion of a 25-kDa harder to rationalize. However, isoenzymes inhibited molecule sufficient mutations have since accumulated by excess glucose are widely distributed in different to obliterate any evidence for it. The N-terminal half specieswx 56 , and so it is reasonable to assume that it of hexokinase C diverges more than the C-terminal has some function, even if it is not very obvious at half from the other hexokinases, and hexokinase D, present what it is. The effect of the negative slope in with a molecular mass half those of the other three, the physiological range of glucose concentrations resembles the C-terminal half of a 100-kDa hexoki- must be to flatten somewhat the curve representing nase more than the N-terminal half. In general, the the total activity of the 100-kDa hexokinases, in the N-terminal halves show more evidence of divergent limit making their composite effect virtually indepen- evolution than the C-terminal halves. dent of the glucose concentration. However, in most The same characteristic is evident in a treeŽ Fig. tissues the activity of hexokinase C appears too low 4a.Ž constructed from the aligned sequences treating for any such effect to be significant. the N- and C-terminal halves of the 100-kDa hexoki- 256 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264

Fig. 3. Comparison of the amino acid sequences of the four rat isoenzymes of hexokinase. Each dot represents an identical residue in the sequences compared, but a dot is only shown if such an identity is the centre of a region of nine residues containing at least four identities. The squares along the main diagonal show comparisons of each sequence with itself, and the secondary diagonals in the comparisons for hexokinases A, B and C show that each of the 100-kDa hexokinases is `dimer-like', i.e., that its N-terminal half has a high degree of similarity to its C-terminal half. The absence of further diagonals intermediate between those that can be seen indicates a lack of evidence that the 50-kDa sequences are derived from duplication of a putative 25-kDa ancestral sequence.

nases as separate sequences. . Apart from an anomaly ancestral molecule, but the N-terminal halves have with hexokinase C that we shall consider shortly, evolved faster, about 25% faster in the case of hexok- there is a high degree of self-consistency, with a well inases A and B, and more than twice as fast in the defined branch pointŽ. point 1 in the tree at which case of hexokinase C. This interpretation implies that hexokinase D diverges from hexokinases A and B at the separation of hexokinase D from the other verte- about the same time that the N- and C-terminal brate hexokinases occurred around 700 Myr ago, i.e., halves of hexokinases A and B diverge from one that it is much more ancient than the radiation that another. Both halves of hexokinases A and B suggest produced the different vertebrate classesŽ about 350 that the divergence of these two isoenzymes from one Myr ago. . Comparisons of the hexokinase D se- another started more recently. quences of Xenopus laeÕis and mammals made by Hexokinase C fits badly into this tree: the N-termi- Veiga-da-Cunha et al.wx 168 indicate about 80% iden- nal half suggests that it started to separate from the tity, consistent with this interpretation. other three isoenzymes at an early momentŽ. point 2 , More rapid evolution of the N-terminal half of the but the C-terminal half indicates equally clearly that 100-kDa isoenzymes accords well with the evidence this separation was a much more recent eventŽ point discussed in Section 3.1 that the catalytic site of 3. than the divergence of hexokinase D, possibly at hexokinase A is in the C-terminal half. This was also about the same time that hexokinases A and B started believed to be true of hexokinases B and C until the to diverge from one another. recent studies that showed hexokinase B to have two The tree can be rationalizedŽ. Fig. 4b if hexokinase catalytic sites, one in the N-terminal half and the D and the C-terminal halves of the 100-kDa isoen- other in the C-terminal halfwx 143±145 . The rapid zymes have evolved at a consistent rate since the evolution of the N-terminal half of hexokinase C is original duplication and fusion produced the 100-kDa quite consistent with the recent evidence that it has M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 257 only a single active sitewx 26 . Its rapid evolution may hexokinaseswx 21 , comparison with one another and be related to its lack of a clear physiological role, as with two rat isoenzymes and yeast glucokinaseŽ Fig. its structure may have been less constrained during 6. shows that the data are entirely consistent with evolution than the others. what one would expect from the phylogenetic separation between green plants, mammals and yeast, and 4.2. Low molecular mass hexokinases from the fact that the two plant hexokinases are similar in sizeŽ. 496 and 502 residues to yeast glu- Fig. 5 illustrates dot plots comparing the sequences cokinaseŽ. 499 residues : the two plant proteins re- of hexokinase D, the three yeast isoenzymes and the semble one another very closely, and resemble any E. coli glucokinase with subunit molecular mass of mammalian sequence about as much as they resemble 34.5 kDa. There is significant similarity between all any yeast sequence, or as much as the yeast and of the 50-kDa enzymes: the most similar are yeast mammalian sequences resemble one another. There is isoenzymes PIII and P , but yeast glucokinase is much easily recognizable similarity between A. thaliana more similar to the other yeast isoenzymes than it is hexokinase 1, for example, and either half of hexoki- to rat hexokinase D. Thus rat hexokinase D and yeast nase A; this similarity is particularly obvious between glucokinase do not form a natural group; if anything residues 65 and 355 of the plant sequence. By con- hexokinase D is more similar to yeast hexokinase PI than it is to yeast glucokinase. This feature has surprised some authorswx 117,118 , who are misled by the misnomer of `glucokinase' for rat hexokinase D. Fig. 4. The hexokinase C anomaly.Ž. a The tree shown was Despite the contention that the two hexokinases obtained by the UPGMAŽ unweighted pair-group method with from the plant A. thaliana are similar to rat hexoki- arithmetic averages. wx 166 from the sequences of the rat and nase D but distinct from the mammalian 100-kDa human isoenzymes of hexokinase aligned as given by Fothergill- Gilmore and Michelswx 167 , each half of each of the 100-kDa sequences being treated as a separate protein. The scale of amino acid substitutions measures distances from ancestors to descendant, and the numbers must be doubled to obtain descendant-to- descendant distances, e.g., the junction at 12 substitutions between the N-terminal halves of human and rat hexokinase A indicates that these sequences differ at 24 loci. The numbered branchpoints illustrate the anomaly with hexokinase C: although there is a well defined pointŽ. labelled 1 indicating that the two hexokinase D sequences differ from any of nine of the other half-sequences at about 140 loci, the sequence of the N-terminal half of hexokinase C differs at more than 200 loci from any of the other sequencesŽ. point labelled 2 , whereas the C-terminal half differs from the corresponding hexokinase A and B sequences at about 100 lociŽ. point labelled 3 . As these values differ by much more than statistical scatter they suggest that hexokinase C separated from hexokinases A and B both before and after the separation of hexokinase D. The model shown inŽ. b is an attempt to rationalize the results ofŽ. a , by supposing that after duplication produced the 100-kDa molecules the C-terminal halves of these molecules, together with the 50-kDa hexokinase D molecules, accumulated substitutions at the rate of about 0.055 per MyrŽ. indicated by lines of normal thickness , whereas the N-terminal halves of hexokinases A and B evolved fasterŽ 0.070 per Myr; thicker lines. and that of hexokinase C much faster Ž.0.125 per Myr; very thick lines . The scale of time is calibrated by assuming that the rat and human lineages separated about 65 Myr B.P. The human hexokinase C descendant is shown in parentheses because, although this enzyme is known to existwx 20 , it has not been purified and no sequence data are available. 258 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 trast, the C-terminal thirds of the plant enzymes show almost no recognizable similarity with the rat or yeast enzymes, and it is noticeable that this is also the region where the yeast and rat sequences differ the most from one another. Despite what is claimed in Ref.wx 21 , the 50-kDa hexokinases do not form a natural group from which the 100-kDa hexokinases are excluded. E. coli glucokinase, with a subunit molecular mass of 34.5 kDawx 34 , shows no similarity to the other hexokinases of low molecular massŽ. Fig. 5 , and the same is true of S. coelicolor glucokinasewx 169Ž not illustrated. . This does not exclude the possibility of homology with the 50-kDa hexokinases, but if any exists the many mutations during the long period since the separation of prokaryotes and eukaryotes have eliminated any evidence for it. As we shall Fig. 6. Comparison of mammalian, plant and yeast hexokinase discuss in Section 4.4, other bacterial hexokinases do sequences. Dots are drawn according to the same criteria as in apparently have non-hexokinase homologues among Fig. 3. The isoenzymes labelled Plant 1 and 2 are hexokinases 1 the enzymes of higher organisms. and 2 from A. thaliana wx21 . A partial determination of the first 10 residues from the N-terminal of the ADP-dependent glucokinase from the archaeon P. furiosus indicated no this limited information does not allow any conclu- similarity with other hexokinases or with any other sions to be drawn. sequences in the SwissProt databasewx 30 . However, 4.3. Phylogenetic relationships between hexokinases

Fig. 7 shows an attempt to accommodate all of the sequence relationships into a single model for the evolution of the hexokinases. Some points seem to be uncontrovertible, despite suggestions otherwise in the literature. In each of the kingdoms where isoenzymes are found these isoenzymes have arisen independently, even though the yeast isoenzymes are relatively ancient, their relationships are with one another and not with the isoenzymes in other kingdoms. In particular, vertebrate hexokinase D has nothing apart from its size and the unfortunate misnomer of `glucokinase' to connect it with yeast glucokinase, and still less with any bacterial glucokinase. Although the bacterial enzymes are tentatively included in Fig. 7, there is in reality very little evidence to suggest that they are related, beyond the fact that they catalyse the same reactions, and some have a size suggestive of Fig. 5. Comparison of 50-kDa hexokinase sequences. Dots are relationship to the putative 25-kDa ancestor of the drawn according to the same criteria as in Fig. 3. The bottom line Ž shows comparisons with the 34.5-kDa glucokinase from E. coli. eukaryotic hexokinases although as discussed in Sec- The absence of recognizable diagonals indicates that this enzyme tion 2.6, most of the known bacterial hexokinases are does not resemble any of the others. in the 32±37-kDa range. . Moreover, this ancestor is M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 259

Fig. 7. The phylogeny of hexokinases. The tree attempts to summarize all of the comparisons discussed in the text. In accordance with the tree drawn in Fig. 4b, the duplication that produced the 100-kDa animal hexokinases is shown as occurring no earlier than the branchpoint that separated the 50-kDa hexokinase D from the other isoenzymes. Because of the need to show different amounts of detail in different parts of the tree there has been no attempt to draw it with a consistent scale of time. The line leading to the bacterial hexokinases is discontinuous because there is no direct evidence that any of these is homologous with the eukaryotic enzymes. itself purely hypothetical, as no suggestion of gene the two halves of the gene and in the hexokinase D duplication remains in the 50-kDa sequences that can gene. However, corresponding data are not available be examined today. for other hexokinase genes, and consequently this Before the amino acid sequences became known information provides no basis for preferring any par- the suggestion of a special similarity between hexoki- ticular model of hexokinase evolution. If the model nases B and D seemed to be most easily accommo- of Fig. 7 is correct, we should certainly expect the dated by a model in which the 50-kDa hexokinase D genes for hexokinases A and C to have the same is descended from a fragment produced by fission of intron organization as that reported for hexokinase B, a 100-kDa ancestorwx 4 . This possibility is not dis- as Printz et al.wx 145,171 have suggested, and it proved by any data now available, but the original would not be surprising if more distantly related reason for proposing it no longer applies, and no genes shared it as well. trace of the other 50-kDa fragment presumably produced by the fission has been found. It seems more 4.4. Structural similarities with other proteins parsimonious, therefore, to suppose that hexokinase D separated from the 100-kDa isoenzymes before the Although, as discussed in Section 4.2, there is no duplication and fusion that produced them, and Fig. 7 convincing evidence of homology between bacterial is drawn in accordance with this interpretation. Anal- glucokinases and the hexokinases of higher organ- ysis of the gene organization of hexokinase B isms, the fructokinases of several bacteria appear to wx170,171 showed the same arrangement of introns in be related to a human enzyme, adenosine kinase, with 260 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 two markedly similar regions in the sequenceswx 172 ; A major part of their argument was based on a for example, residues 294±305 of the human enzyme misconception, however, because the conclusion that are identical at 11 out of 12 sites to residues 243±256 evolution for glucose specificity has arisen three times of fructokinase from V. alginolyticus wx105 . Neither is based on the supposition that the mammalian hex- of the two consensus sequences identified in Ref. okinases D are specific for glucose. If this error is wx172 occurs in any of the hexokinase sequences corrected one is left with very sparse data: yeast represented in Figs. 3, 5 and 6, however, and so it is glucokinase, the least thoroughly studied of the yeast not possible to relate this discovery to the phyloge- hexokinases, provides the only evidence that glucose nies that we have been considering. specificity has arisen independently in eukaryotes and There is also evidence for three-dimensional simi- prokaryotes, and a single divergent fructokinaseŽ that larities between hexokinases and other kinds of pro- of Z. mobilis. provides the only evidence that fruc- tein even though no relationships between the corre- tose specificity has arisen independently in two sponding sequences can be detected by standard branches of the prokaryotic phylogeny. methods. On the one hand, Kabsch et al.wx 173 recog- As we discussed in Section 3.3, the variations in nized that the ATP-binding site of actin had some kinetic behaviour exhibited by the different forms of features in common with the corresponding region of hexokinase can probably all be accommodated within yeast hexokinase, and similar observations were made a single type of mechanism, in which the reaction in comparing actin with the 70-kDa heat shock pro- proceeds through a ternary complex that can be tein Hsc70wx 174,175 . Bork et al. wx 176 subsequently formed by binding of glucose and MgATP in either compared the three-dimensional structures of all three order, but with a preferred order of substrate binding proteins and confirmed that the parts of the molecules Ðmore marked for some hexokinases than for oth- concerned with binding and hydrolysis of ATP were ers. Wilson and Schwabwx 178 recently reexamined very similar, especially so for actin and Hsc70, but the mechanism in the light of the recent discovery of with essentially the same domain organization in all the similarity in structure with actin. They pointed three proteins. out that although most discussion of MgATP binding The three-dimensional structures allowed align- to the yeast hexokinase molecule had focused on a ment of the sequences and detection of motifs that region of the structure known as the large lobe, there could not have been found by studying the sequences was reason to believe that the other region, the small alone. The same motifs proved to be present in rat lobe, was also involved. If MgATP were to bind hexokinase D, several other sugar kinasesŽ fucokinase before glucose this would probably cause the two from E. coli, glycerokinase from E. coli and B. lobes to move closer together, producing a structure subtilis, gluconokinase from B. subtilis, xylulokinase lacking glucose but otherwise similar to the reactive from E. coli and ribulokinase from Salmonella ty- ternary complex with both substrates bound. As glu- phimurium. and several proteins involved in the cose binds to a site deep within the active-site cleft it prokaryotic cell cyclewx 176 . Subsequently two phos- would not have easy access to it if the structure were phatases from E. coli, exopolyphosphate phosphatase partially closed already. Hence binding of MgATP and guanosine pentaphosphate phosphatase, were before glucose would not prevent catalysis altogether found to belong to the same superfamilywx 177 . but would be less favourable than the alternative Bork et al.wx 176 examined whether the similarity order with glucose binding first. of the sugar kinases to these other proteins could Three-dimensional structures are not yet available have arisen by convergent evolution, but concluded for any of the mammalian hexokinases, but the se- Žon the basis of qualitative rather than detailed statis- quence similarities imply that the crystallographic tical arguments. that the unusually complicated struc- studies of the PIII and P isoenzymes of yeast hexoki- ture of the region of similarity was unlikely to have nase constitute a valuable resource for modelling the arisen twice independently and that, therefore, the mammalian enzymes and predicting which are the various proteins were homologous. In a later article functionally important residues. This needs to be wx118 , they argued that convergent evolution had done cautiously, however, especially in relation to nonetheless occurred within the sugar kinase group. hexokinase D, because the markedly different kinetic M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 261 behaviour from that of the yeast isoenzymes must be wx18 C. Gonzalez,ÂÂ T. Ureta, R. Sanchez, H. Niemeyer, Biochem. Biophys. Res. Commun. 16Ž. 1964 347±352. related to some significant differences in structure. wx The recent demonstrationwx 179,180 that two proteins 19 J. Brown, D.M. Miller, M.T. Holloway, G.D. Leve, Sci- ence 155Ž. 1967 205±207. Ðone of them admittedly artificialÐcan share 50% wx20 P.A. Rogers, R.A. Fisher, H. Harris, Biochem. Genet. 13 sequence identity yet have well defined but entirely Ž.1975 857±866. different three-dimensional structures underlines the wx21 J.-C. Jang, P. Leon,Â L. Zhou, J. Sheen, Plant Cell 9Ž. 1997 need for such caution. A detailed study of the general 5±19. wx question of the relationship between sequence align- 22 H.M. Katzen, D.D. Soderman, H.M. Nitowski, Biochem. Biophys. Res. Commun. 19Ž. 1965 377±382. ments and three-dimensional structures has appeared wx wx 23 M.J. Holroyde, I.P. Trayer, FEBS Lett. 62Ž. 1976 215±219. recently 181 . wx24 D.T. Vowles, J.S. Easterby, Biochim. Biophys. Acta 566 Ž.1979 283±295. wx25 T.K. White, J.E. Wilson, Arch. Biochem. Biophys. 174 Acknowledgements Ž.1989 375±393. wx26 J.E. Wilson, Biochem. Soc. Trans. 25Ž. 1997 103±107. wx27 R.A. Jensen, Ann. Rev. Microbiol. 30Ž. 1976 409±425. We are grateful to Drs. Patricio RodrõguezÂ and wx28 D.E. Koshland Jr., Fed. Proc. 35Ž. 1976 2104±2111. Linda Fothergill-Gilmore for assistance in locating wx29 H. Kacser, R. Beeby, J. Mol. Evol. 20Ž. 1984 38±51. structural information on hexokinases, and to Drs. wx30 S.W.M. Kengen, J.E. Tuininga, F.A.M. de Bok, A.J.M. Yves Quentin and Juli PeretoÂ for useful discussions. Stams, W.M. de Vos, J. Biol. Chem. 270Ž. 1995 30453± 30457. wx31 E.V. Porter, B.M. Chassy, C.E. Holmlund, Biochim. Bio- phys. Acta 709Ž. 1982 178±186. References wx32 V. Sapico, R.L. Anderson, J. Biol. Chem. 242Ž. 1967 5086±5092. wx1 O. Meyerhof, Biochem. Z. 183Ž. 1927 176±215. wx33 Y. Fukuda, S. Yamaguchi, M. Shimosaka, K. Murata, A. wx2 S.P. Colowick, in: P.D. BoyerŽ. Ed. , The Enzymes, 3rd Kimura, Bull. Agric. Biol. Chem. 48Ž. 1984 2541±2548. edn., Vol. 9B, Academic Press, New York, 1973, pp. wx34 D. Meyer, C. Schneider-Fresenius, R. Horlacher, R. Peist, 1±48. W. Boos, J. Bacteriol. 179Ž. 1997 1298±1306. wx3 T. Ureta, in: C.L. MarkertŽ. Ed. , Isoenzymes IIIÐDevel- wx35 B. Sabater, J. Sebastian,Â C. Asensio, Biochim. Biophys. opmental Biology, Academic Press, New York, 1975, pp. Acta 284Ž. 1972 406±413. 575±601. wx36 H. Hengartner, H. Zuber, FEBS Lett. 37Ž. 1973 212±216. wx4 T. Ureta, Comp. Biochem. Physiol. B 71Ž. 1982 549±555. wx37 J. Sebastian,Â C. Asensio, Biochem. Biophys. Res. Com- wx5 Z. Lobo, P.K. Maitra, Genetics 86Ž. 1977 724±744. mun. 28Ž. 1967 197±202. wx6 J.E. Wilson, in: R. BeitnerŽ. Ed. , Regulation of Carbohy- wx38 H.W. Doelle, Eur. J. Appl. Microbiol. 14Ž. 1982 241±246. drate Metabolism, Vol. 1, CRC Press, Boca Raton, FL, wx39 M.Y. Kamel, D.P. Allison, R.L. Anderson, J. Biol. Chem. 1985, pp. 45±85. 241Ž. 1966 690±694. wx7 T. Ureta, C. Medina, A. Preller, Arch. Biol. Med. Exp. 20 wx40 J.A. Delvalle, C. Asensio, BioSystems 10Ž. 1978 265±282. Ž.1987 343±357. wx41 N.J. Palleroni, R. Contopoulou, M. Doudoroff, J. Bacteriol. wx8 R.J. Middleton, Biochem. Soc. Trans. 18Ž. 1990 180±183. 71Ž. 1956 202±207. wx9 J.E. Wilson, Rev. Physiol. Biochem. Pharmacol. 124Ž. 1995 wx42 N. Saito, J. Biochem.Ž.Ž. Tokyo 57 1965 367±371. 65±198. wx43 G. Lucchini, Biochim. Biophys. Acta 242Ž. 1971 365±370. wx10 K.A. Trayser, S.P. Colowick, Arch. Biochem. Biophys. 94 wx44 P.K. Maitra, Methods Enzymol. 42Ž. 1975 25±30. Ž.1961 177±181. wx45 E.A. Barnard, Methods Enzymol. 42Ž. 1975 6±20. wx11 J.M. Gancedo, D. Clifton, D.G. Fraenkel, J. Biol. Chem. wx46 P. Baumann, Biochemistry 8Ž. 1969 5011±5015. 252Ž. 1977 4443±4444. wx47 M. Ueda, S. Iwasawa, N. Miyata, K. Ahiko, Agric. Biol. wx12 F.C. Womack, M.K. Welch, J. Nielsen, S.P. Colowick, Chem. 46Ž. 1982 2637±2643. Arch. Biochem. Biophys. 158Ž. 1973 451±457. wx48 R. Lagos, T. Ureta, Eur. J. Biochem. 104Ž. 1980 357±365. wx13 P.K. Maitra, J. Biol. Chem. 245Ž. 1970 2423±2431. wx49 M. Hirai, E. Ohtani, A. Tanaka, S. Fukui, Biochim. Bio- wx14 D.G. Walker, Biochem. J. 84Ž. 1962 118P±119P. phys. Acta 480Ž. 1977 357±366. wx15 D.G. Walker, Biochim. Biophys. Acta 77Ž. 1963 209±226. wx50 R.E. Reeves, F. Montalvo, A. Sillero, Biochemistry 6 wx16 E. Vinuela,Ä M. Salas, A. Sols, J. Biol. Chem. 238Ž. 1963 Ž.1967 1752±1760. PC1175±PC1177. wx51 E.L. Risby, J.R. Seed, J. Protozool. 16Ž. 1969 193±197. wx17 C. Sharma, R. Manjeshwar, S. Weinhouse, J. Biol. Chem. wx52 J.F. Turner, Q.J. Chensee, D.D. Harrison, Biochim. Bio- 238Ž. 1963 3840±3845. phys. Acta 480Ž. 1977 367±375. 262 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264

wx53 J.C. Meunier, J. Buc, J. Ricard, FEBS Lett. 14Ž. 1971 wx84 P. Emmelot, C.J. Bos, Biochim. Biophys. Acta 121Ž. 1966 25±28. 434±436. wx54 T.J.C. Higgins, J.S. Easterby, Eur. J. Biochem. 45Ž. 1974 wx85 G.W. Gould, G.D. Holman, Biochem. J. 295Ž. 1993 329± 147±160. 341. wx55 M.R. Stetten, P.K. Goldsmith, Biochim. Biophys. Acta 657 wx86 M. Lachaal, C.Y. Jung, J. Cell Physiol. 156Ž. 1993 326± Ž.1981 468±481. 332. wx56 T. Ureta, J. Radojkovic,ÂÂÂÄ R. Lagos, V. Guixe, L. Nunez, wx87 J.J. Schmidt, S.P. Colowick, Arch. Biochem. Biophys. 158 Arch. Biol. Med. Exp. 12Ž. 1979 587±604. Ž.1973 458±470. wx57 C.R. Woese, O. Kandler, M.L. Wheelis, Proc. Natl. Acad. wx88 P.K. Maitra, Z. Lobo, Mol. Cell. Biochem. 18Ž. 1977 Sci. U.S.A. 87Ž. 1990 4576±4579. 21±27. wx58 R.K. Scopes, V. Testolin, A. Stoter, K. Griffiths-Smith, wx89 O. Misset, O.J.M. Bos, F.R. Opperdoes, Eur. J. Biochem. E.M. Algar, Biochem. J. 228Ž. 1985 627±634. 157Ž. 1986 441±453. wx59 C.K. Goward, R. Hartwell, T. Atkinson, M.D. Scawen, wx90 J.C. Meunier, J. Buc, A. Navarro, J. Ricard, Eur. J. Biochem. J. 237Ž. 1986 415±420. Biochem. 49Ž. 1974 209±223. wx60 H. Ishikawa, T. Maeda, H. Hikita, Biochem. J. 248Ž. 1987 wx91 R.W. Komuniecki, L.S. Roberts, Comp. Biochem. Physiol. 13±20. B 57Ž. 1977 45±49. wx61 F. Gonzalez,ÂÂ A. Fernandez-Vivas, J.M. Arias, E. Montoya, wx92 S.C. Supowit, B.G. Harris, Biochim. Biophys. Acta 422 Arch. Microbiol. 154Ž. 1990 438±442. Ž.1976 48±59. wx62 M.L. Cardenas,Â Glucokinase: Its Regulation and Role in wx93 D. Moser, L. Johnson, C.Y. Lee, J. Biol. Chem. 255Ž. 1980 Liver Metabolism, R.G. Landes, Austin, TX, 1995. 4673±4679. wx63 M.L. Cardenas,Â Biochem. Soc. Trans. 25Ž. 1997 131±135. wx94 Y. Mochizuki, S.H. Hori, J. Biochem.Ž.Ž. Tokyo 81 1977 wx64 M. De Rosa, A. Gambacorta, B. Nicolaus, P. Giardina, E. 1849±1855. Poerio, V. Buonocore, Biochem. J. 224Ž. 1984 407±414. wx95 Y. Mochizuki, Comp. Biochem. Physiol. 70BŽ. 1981 745± wx65 J. Davagnino, T. Ureta, J. Biol. Chem. 255Ž. 1980 2633± 751. 2636. wx96 J.E. Wilson, Biochim. Biophys. Acta 276Ž. 1972 568±571. wx66 M.B. Allen, J.L. Brockelbank, D.G. Walker, Biochim. wx97 L. Grossbard, R.T. Schimke, J. Biol. Chem. 241Ž. 1966 Biophys. Acta 614Ž. 1980 357±366. 3546±3560. wx67 M.L. Vera, M.L. Cardenas,Â H. Niemeyer, Arch. Biochem. wx98 C.L. Wright, A.S. Warsy, M.J. Holroyde, I.P. Trayer, Biophys. 229Ž. 1984 237±245. Biochem. J. 175Ž. 1978 125±135. wx68 J.F. Turner, L. Copeland, Plant Physiol. 68Ž. 1981 1123± wx99 M.J. Holroyde, M.B. Allen, A.C. Storer, A.S. Warsy, 1127. J.M.E. Chesher, I.P. Trayer, A. Cornish-Bowden, D.G. wx69 J.A. Miernyk, D.T. Dennis, Arch. Biochem. Biophys. 226 Walker, Biochem. J. 153Ž. 1976 363±373. Ž.1983 458±468. wx100 W.O. Barnell, K.C. Yi, T. Conway, J. Bacteriol. 172 wx70 E. Bueding, J.A. Mackinnon, J. Biol. Chem. 215Ž. 1955 Ž.1990 7227±7240. 494±505. wx101 D.V. Pollard-Knight, A. Cornish-Bowden, Biochem. J. 245 wx71 M. Agosõn,Â L. Aravena, Biochim. Biophys. Acta 34Ž. 1959 Ž.1987 625±629. 90±102. wx102 D.G. Walker, Essays Biochem. 2Ž. 1966 33±67. wx72 M.L. Cardenas,Â E. Rabajille, H. Niemeyer, Biochem. J. wx103 N.F.B. Phillips, P.J. Horn, H.G. Wood, Arch. Biochem. 222Ž. 1984 363±370. Biophys. 300Ž. 1993 309±319. wx73 R.K. Scopes, D.R. Bannon, Biochim. Biophys. Acta 1249 wx104 G.J. Fennington Jr., T.A. Hughes, Microbiology 142Ž. 1996 Ž.1995 173±179. 321±330. wx74 B.E.H. Maden, Trends Biochem. Sci. 20Ž. 1995 337±341. wx105 G.L. Blatch, R.R. Scholle, D.R. Woods, Gene 95Ž. 1990 wx75 L.E. Orgel, The Origins of Life, Chapman & Hall, London, 17±23. 1973, p. 175. wx106 J. Thompson, N.Y. Nguyen, S.A. Robrish, J. Bacteriol. 174 wx76 E. Broda, The Evolution of the Bioenergetic Processes, Ž.1992 3227±3235. Pergamon, Oxford, 1975. wx107 J.N. Hope, A.W. Bell, M.A. Hermodson, J.M. Grourke, J. wx77 R. Larralde, M.P. Robertson, S.L. Miller, Proc. Natl. Acad. Biol. Chem. 261Ž. 1986 7663±7668. Sci. U.S.A. 92Ž. 1995 8158±8160. wx108 M.L. Cardenas,Â E. Rabajille, H. Niemeyer, Arch. Biochem. wx78 P. Decker, H. Schweer, R. Pohlmann, J. Chromatogr. 244 Biophys. 190Ž. 1978 142±148. Ž.1982 281±291. wx109 H. Niemeyer, T. Ureta, L. Clark-Turri, Mol. Cell. Biochem. wx79 R. Shapiro, Orig. Life Evol. Biosphere 18Ž. 1988 71±85. 6Ž. 1975 109±126. wx80 N.W. Gabel, C. Ponnamperuma, NatureŽ. London 216 wx110 D. Pollard-Knight, A. Cornish-Bowden, Mol. Cell. Ž.1967 453±455. Biochem. 44Ž. 1982 71±80. wx81 C. Reid, L.E. Orgel, NatureŽ.Ž. London 216 1967 455. wx111 M.D. Meglasson, M. Schinco, F.M. Matschinsky, Diabetes wx82 P.H. Clarke, S.R. Elsden, J. Mol. Evol. 15Ž. 1980 328±333. 32Ž. 1983 1146±1151. wx83 G.I. Bell, C.F. Burant, J. Takeda, G.W. Gould, J. Biol. wx112 A. Cornish-Bowden, M.L. Cardenas,Â Trends Biochem. Sci. Chem. 268Ž. 1993 19161±19164. 16Ž. 1991 281±282. M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 263 wx113 A. Cornish-Bowden, M.L. Cardenas,Â Trends Biochem. Sci. sell, J.M. May, D.K. Granner, J. Biol. Chem. 271Ž. 1996 17Ž. 1992 59. 1849±1852. wx114 H. Niemeyer, M.L. Cardenas,Â E. Rabajille, T. Ureta, L. wx144 H.J. Tsai, J.E. Wilson, Arch. Biochem. Biophys. 329Ž. 1996 Clark-Turri, J. Penaranda,Ä Enzyme 20Ž. 1975 321±333. 17±23. wx115 A.C. Storer, A. Cornish-Bowden, Biochem. J. 159Ž. 1976 wx145 R.L. Printz, H. Osawa, H. Ardehali H., S. Koch, D.K. 7±14. Granner, Biochem. Soc. Trans. 25Ž. 1997 107±112. wx116 H.G. Hers, Trends Biochem. Sci. 17Ž. 1992 59. wx146 H.J. Tsai, FASEB J. 11Ž. 1997 A897. wx117 L.D. Griffin, B.D. Gelb, D.A. Wheeler, Genomics 11 wx147 B.A. Connolly, I.P. Trayer, Eur. J. Biochem. 93Ž. 1979 Ž.1991 1014±1024. 375±385. wx118 P. Bork, C. Sander, A. Valencia, Protein Sci. 2Ž. 1993 wx148 M. Gregoriou, I.P. Trayer, A. Cornish-Bowden, Eur. J. 31±40. Biochem. 134Ž. 1983 283±288. wx119 P.-C. Hsieh, B.C. Shenoy, D. Samols, N.F.B. Phillips, J. wx149 N.J. Ganson, H.J. Fromm, J. Biol. Chem. 260Ž. 1985 Biol. Chem. 271Ž. 1996 4909±4915. 12099±12105. wx120 M.A. Magnuson, Diabetes 39Ž. 1990 523±527. wx150 M. Gregoriou, I.P. Trayer, A. Cornish-Bowden, Eur. J. wx121 P.B. Iynedjian, Biochem. J. 293Ž. 1993 1±13. Biochem. 161Ž. 1986 171±176. wx122 T. Ureta, S.B. Reichberg, J. Radojkovic,Â J.C. Slebe, Comp. wx151 M.B. Bass, H.J. Fromm, Arch. Biochem. Biophys. 256 Biochem. Physiol. B 45Ž. 1973 445±461. Ž.1987 708±711. wx123 T. Ureta, J.C. Slebe, J. Radojkovic,Â C. Lozano, Comp. wx152 M. Gregoriou, I.P. Trayer, A. Cornish-Bowden, Biochem- Biochem. Physiol. B 50Ž. 1975 515±524. istry 20Ž. 1981 499±506. wx124 M.A. Magnuson, J. Cell Biol. 48Ž. 1992 115±121. wx153 J. Radojkovic,Â T. Ureta, Biochem. J. 242Ž. 1987 895±903. wx125 F.M. Matschinsky, Diabetes 39Ž. 1990 647±652. wx154 T. Ureta, J. Biol. Chem. 251Ž. 1976 5035±5042. wx126 M. Gabe, TraiteÂÂ de Zoologie, in: P.P. GrasseŽ. Ed. , Vol. wx155 L. Agius, M. Peak, Biochem. Soc. Trans. 25Ž. 1997 145± 14, Fasc. 3, Masson, Paris, 1970, pp. 1333±1399. 150. wx127 T. Ureta, J. Radojkovic,ÂÂ S. Zepeda, V. Guixe, Comp. wx156 J.J. Guinovart, A.M. Gomez-Foix,Â S. Seoane, J.M. Biochem. Physiol. B 70Ž. 1981 225±236. Fernandez-Novell,ÂÂ D. Bellido, S. Vilaro, Biochem. Soc. wx128 E. Van Schaftingen, M. Detheux, M. Veiga da-Cunha, Trans. 25Ž. 1997 157±160. FASEB J. 8Ž. 1993 414±419. wx157 L. Niculescu, M. Veiga-da-Cunha, E. Van Schaftingen, wx129 M. Detheux, A. Vandercammen, E. Van Schaftingen, Eur. Biochem. J. 321Ž. 1997 239±246. J. Biochem. 200Ž. 1991 553±561. wx158 A. Cornish-Bowden, J. Theor. Biol. 65Ž. 1977 735±742. wx130 M. Detheux, J. Vandekerckhove, E. Van Schaftingen, FEBS wx159 M. Darby, I.P. Trayer, Eur. J. Biochem. 129Ž. 1983 555± Lett. 321Ž. 1993 111±115, corrected in FEBS Lett., 339, 560. 312Ž. 1994 . wx160 F. Marcus, T. Ureta, Biochem. Biophys. Res. Commun. wx131 R.K. Crane, A. Sols, J. Biol. Chem. 210Ž. 1954 597±606. 139Ž. 1986 714±719. wx132 M. Magnani, M. Bianchi, A. Casabianca, V. Stocchi, A. wx161 D.A. Schwab, J.E. Wilson, Proc. Natl. Acad. Sci. U.S.A. Daniele, F. Altruda, M. Ferrone, L. Silengo, Biochem. J. 86Ž. 1989 2563±2567. 285Ž. 1992 193±199. wx162 A.P. Thelen, J.E. Wilson, Arch. Biochem. Biophys. 286 wx133 K.K. Arora, C.R. Filburn, P.L. Pedersen, J. Biol. Chem. Ž.1991 645±651. 268Ž. 1993 18259±18266. wx163 D.A. Schwab, J.E. Wilson, Arch. Biochem. Biophys. 285 wx134 K.K. Arora, M. Fanciulli, P.L. Pedersen, J. Biol. Chem. Ž.1991 365±370. 265Ž. 1990 6481±6488. wx164 T.L. Andreone, R.L. Printz, S.J. Pilkis, M.A. Magnuson, wx135 I.A. Rose, J.V.B. Warms, J. Biol. Chem. 242Ž. 1967 D.K. Granner, J. Biol. Chem. 264Ž. 1989 363±369. 1635±1645. wx165 S. Nishi, M. Stoffel, K. Xiang, T.B. Show, G.I. Bell, J. wx136 J.E. Wilson, J. Biol. Chem. 243Ž. 1968 3640±3649. Takeda, Diabetologia 35Ž. 1992 743±747. wx137 D.M. Parry, P.L. Pedersen, J. Biol. Chem. 258Ž. 1983 wx166 P.H.A. Sneath, R.R. Sokal, Numerical Taxonomy, Free- 10904±10912. man, San Francisco, CA, 1973, pp. 188±308. wx138 D.M. Parry, P.L. Pedersen, J. Biol. Chem. 265Ž. 1990 wx167 L.A. Fothergill-Gilmore, P.A.M. Michels, Prog. Biophys. 1059±1066. Mol. Biol. 59Ž. 1993 105±235. wx139 A.C. Chou, J.E. Wilson, Methods Enzymol. 42Ž. 1975 wx168 M. Veiga-da-Cunha, S. Courtois, A. Michel, E. Gosselain, 20±25. A. Van Schaftingen, J. Biol. Chem. 271Ž. 1996 6292±6297. wx140 A.C. Storer, A. Cornish-Bowden, Biochem. J. 165Ž. 1977 wx169 S. Angell, E. Schwarz, M.J. Bibb, Mol. Microbiol. 6 61±69. Ž.1992 2833±2844. wx141 T.J.C. Higgins, J.S. Easterby, Eur. J. Biochem. 65Ž. 1976 wx170 K. Kogure, Y. Shinohara, H. Terada, J. Biol. Chem. 268 513±516. Ž.1993 8422±8424. wx142 M.A. Blasquez,Â R. Lagunas, C. Gancedo, J.M. Gancedo, wx171 R.L. Printz, S. Koch, L.R. Potter, R.M. O'Doherty, J.J. FEBS Lett. 329Ž. 1993 51±54. Tiesinga, S. Moritz, D.K. Granner, J. Biol. Chem. 268 wx143 H. Ardehali, Y. Yanot, R.L. Printz, S. Koch, R.R. White- Ž.1993 5209±5219. 264 M.L. CardenasÂ et al.rBiochimica et Biophysica Acta 1401() 1998 242±264 wx172 J. Spychala, N.S. Datta, K. Takabayashi, M. Datta, I.H. wx177 J. Reizer, A. Reizer, M.H. Saier Jr., P. Bork, C. Sander, Fox, T. Gribbin, B.S. Mitchell, Proc. Natl. Acad. Sci. Trends Biochem. Sci. 18Ž. 1993 247±248. U.S.A. 93Ž. 1996 1232±1237. wx178 J.E. Wilson, D.A. Schwab, FASEB J. 10Ž. 1996 799±801. wx173 W. Kabsch, H.G. Mannherz, D. Suck, E.F. Pai, K.C. wx179 S. Dalal, S. Balasubramanian, L. Regan, Nature Struct. Holmes, Nature 347Ž. 1990 37±44. Biol. 4Ž. 1997 548±552. wx174 K.M. Flaherty, C. DeLuca-Flaherty, D.B. McKay, Nature wx180 M. Groû, K.W. Plaxco, Nature 388Ž. 1997 419±420. 346Ž. 1990 623±628. wx181 R.A. Abagyan, S. Batalov, J. Mol. Biol. 273Ž. 1997 355± wx175 K.M. Flaherty, D.B. McKay, W. Kabsch, K.C. Holmes, 368. Proc. Natl. Acad. Sci. U.S.A. 88Ž. 1991 5041±5045. wx176 P. Bork, C. Sander, A. Valencia, Proc. Natl. Acad. Sci. U.S.A. 89Ž. 1992 7290±7294.