key: cord-325043-vqjhiv7p
authors: Gorbalenya, Alexander E.; Blinov, Vladimir M.; Donchenko, Alexei P.; Koonin, Eugene V.
title: An NTP-binding motif is the most conserved sequence in a highly diverged monophyletic group of proteins involved in positive strand RNA viral replication
date: 1989
journal: J Mol Evol
DOI: 10.1007/bf02102483
sha: 
doc_id: 325043
cord_uid: vqjhiv7p

NTP-motif, a consensus sequence previously shown to be characteristic of numerous NTP-utilizing enzymes, was identified in nonstructural proteins of several groups of positive-strand RNA viruses. These groups include picorna-, alpha-, and coronaviruses infecting animals and como-, poty-, tobamo-, tricorna-, hordei-, and furoviruses of plants, totalling 21 viruses. It has been demonstrated that the viral NTP-motif-containing proteins constitute three distinct families, the sequences within each family being similar to each other at a statistically highly significant level. A lower, but still valid similarity has also been revealed between the families. An overall alignment has been generated, which includes several highly conserved sequence stretches. The two most prominent of the latter contain the socalled “A” and “B” sites of the NTP-motif, with four of the five invariant amino acid residues observed within these sequences. These observations, taken together with the results of comparative analysis of the positions occupied by respective proteins (domains) in viral multidomain proteins, suggest that all the NTP-motif-containing proteins of positive-strand RNA viruses are homologous, constituting a highly diverged monophyletic group. In this group the “A” and “B” sites of the NTP-motif are the most conserved sequences and, by inference, should play the principal role in the functioning of the proteins. A hypothesis is proposed that all these proteins posses NTP-binding capacity and possibly NTPase activity, performing some NTP-dependent function in viral RNA replication. The importance of phylogenetic analysis for the assessment of the significance of the occurrence of the NTP-motif (and of sequence motifs of this sort in general) in proteins is emphasized.

Structural (sequence) motifs thought to be identifiers of certain protein activities are among the main tools used in the functional and evolutionary interpretation of protein sequence data (Doolittle 1986a,b; Hodgman 1986) . Because these motifs are short sequence stretches, and usually include amino acid residues frequent in proteins (i.e., Gly, Ala, Ser, and some others), the presence of such a motif in a protein sequence is, as a rule, in itself not statistically significant. Thus, it is important to work out some additional criteria for evaluation of such observations.

One of the most widespread sequence motifs is implicated in an activity crucial to the function of a great variety of proteins, namely purine nucleotide binding followed in most cases by hydrolysis of the 8-3' phosphate bond. This motif was first recognized by Walker and coworkers in several ATP-and GTPutilizing enzymes (Walker et al. 1982; Gay and Walker 1983) . It consists of two separate units designated "A" and "B" sites; the "B" site is located in the polypeptide chains C-proximally relative to the "A" site. For the "A" site the following consensus sequence was proposed: GXXXXGK-(T) XXXXXXI/V, and the "B" consensus was R/K-XXXGXXXL***D, where X stands for any amino acid residue, and * for a hydrophobic residue (Walker et al. 1982) . Results of subsequent analyses of a variety of NTP-utilizing proteins (reviewed by Halliday 1984; M611er and Amons 1985; Doolittle 1986a) suggest much more liberal consensus formulas, namely G/AXXXXGKT/S for the "A" site, and an Asp residue preceded by five residues, three of which are hydrophobic, for the "B" site. Hereafter we accept these loose definitions for the "A" and "B" consensus sequences; taken together, they are designated "NTP-motif." In fact, in recent studies, protein sequences were searched for the "A" consensus alone as the "B" consensus in its loosest form is obviously too degenerate to be unequivocally recognized, except in a family of diverged proteins (see below).

For adenylate kinase, Escherichia coli Tu factor, p2 lras oncoprotein, SV40 T antigen, and some other proteins, there is experimental evidence that the NTP-motif, or at least a larger segment of a protein encompassing it, is involved in NTP binding and/ or cleavage (Clertant and Seif 1984; Jurnak 1985; La Cour et al. 1985; Fry et al. 1986 ). More specifically, the "A" site has been implicated directly in the binding of the pyrophosphate moiety of NTP, whereas the Asp residue of the "B" site appears to interact with the magnesium cation complexed with the same phosphate groups (M611er and Amons 1985; Bradley et al. 1987) . All these observations make it an attractive idea to search sequences of functionally uncharacterized proteins for the presence of the NTPmotif to the end of prediction of NTPase activity, or at least NTP-binding capacity.

Following this line, we screened the protein sequences of positive-strand RNA viruses (the largest class of viruses, whose single-stranded genomic RNA also serves as the mRNA for the synthesis of viral proteins) and identified the "A" consensus in nonstructural proteins of several viral families (Gorbalenya et al. 1985) . In some of these proteins the presence of this consensus has been independently noticed by other workers too (Argos and Leberman 1985; Doolittle 1986a; Dever et al. 1987; Domier et al. 1987) . Also, the NTP-binding capacity of one of these proteins, p 126 of tobacco mbsaic virus, has been demonstrated experimentally quite independently (Evans et al. 1985) . We proposed that such a capacity should be characteristic of all the NTPmotif-containing proteins of positive-strand RNA viruses. On the other hand, the validity of such predictions in general has been disputed (Argos and Leberman 1985; Doolittle 1986a) . Moreover, Doolittle (1986a) identified the "'A'" consensus in several proteins reported to be devoid of NTP-binding properties.

In the present study we undertook a more systematic investigation of the primary structures of the proteins of positive-strand RNA viruses containing the "A" consensus of the NTP-motif. We demonstrate that the NTP-motif-containing proteins of positive-strand RNA viruses (including 8 proteins identified in the previous paper and 13 proteins of viruses whose genomes have been sequenced since then) constitute three monophyletie families that can be brought together into a higher rank taxon. In these homologous proteins the "A" and "B" sites of the NTP-motif are the most strictly conserved sequences and, by inference, should be of principal functional importance, presumably constituting parts of NTPase catalytic centers.

Protein sequences were extracted from the current literature (for references see Table 1 ). The initial screening of the sequences of positive-strand RNA viral proteins for the presence of the "A" consensus sequence of the NTP-motif was performed by use of the program SRCH designed to screen protein sequences for defined amino acid residue strings (motifs). The selected sequences were further analyzed manually for the presence ofcandidate "B" sequences C-proximal to the "A" site. Sequences of the viral proteins containing the NTP-motif were compared by use of the programs DIAGON (Staden 1982) and OPTAL (Pozdnyakov and Pankov 1981) ; the latter program was modified and adopted for multiple sequence alignment as described below. All the programs were written in FORTRAN and run on an ES-1060 computer.

The program OPTAL, based on the original algorithm of Sankoff (1972) , performs optimal alignment of pairs of protein sequences or stepwise alignment of multiple sequences. According to the Sankoff algorithm, a series of cumulative similarity matrices for the compared sequences is created. In the present work, for the calculation of the elements of these matrices, weights of amino acid residue pairs were taken from the mutation rate scoring matrix MDM78 (Staden 1982) . To accelerate calculations, only those elements of the matrices enclosed within a diagonal window, the width of which was chosen to be equal to V5 of the length of the compared sequences, were computed; it has been shown that this window width is sufficient in most cases for generation of the optimal alignment (Pozdnyakov and Pankov 1981) . At the first step of the optimal alignment generation, a series of locally optimal alignments with q = 0, 1, 2, . . . ,qmax gaps was obtained. In practice, for sequence lengths of up to 250 residues dealt with in this work, qm~, was chosen to be equal to 15. This exceeds the gap number most frequently observed in related protein sequences (about four gaps per 100 residues; see Doolittle 1981) and should guarantee the generation of the optimal alignment. For selection of the best of the alignments with a given q value and concomitant assessment of its statistical significance, the following Monte Carlo procedure was employed. Twenty-five pairs of "random" sequences were generated by scrambling the real compared sequences, and the above alignment procedure was simulated for each pair. The mean score, Sq ~d and the standard deviation, aq, were calculated separately for alignments with different gap numbers. For each of the locally optimal real alignments, the deviation ofthe observed score from the mean value for the randomized sequences was calculated in SD units: Rezaian et al. 1985 Ahlquist et al. 1984 Cornelissen et al. 1983 Goelet et al. 1982 Strauss et al. 1984 Takkinen 1986 Gustafson and Armour 1986 Bouzoubaa et al. 1986 Bouzoubaa et al. 1987 Boursnell et al. 1987 For picorna-and potyviruses, the numbering of complete polyproteins is indicated; for alphaviruses, the numbering of the nonstructural polyproteins is indicated; for CPMV the numbering of the polyprotein encoded by RNA B is indicated; and for BNYVV RNA 1 the numbering of the entire high-molecular weight product is indicated. The dendrograms were designed to visualize the procedure of multiple sequence alignment in the order of decreasing similarity between proteins; they cannot be automatically regarded as evolutionary trees. For abbreviations of viruses see Table 1 . The values refer to the conserved domains, as indicated in Table 1 , and, for the proteins of the 1 st family, to the N-terminal subdomains. Dq = Sq r'l -Sq~"d/aq.

The alignment with the maximal D value was considered optimal for the selected window width. For multiple sequence alignment, a generalization of this procedure was employed. To align two sets of m and n prealigned sequences, cumulative similarity matrices were created as before, but for the calculation &their elements, values W~j = ~w~, i.e., the combined weights of all possible pairs of residues (m. n total) in the ith position of the set n and the jth position of the set m are used instead of the weights of individual pairs of residues. A weight of 10 was ascribed to a pair of two gaps, and a weight of zero to a pair of gaps with any residue. The procedures of alignment generation and the choice of the optimal alignment were as described, but, for the generation of "random" sequence sets, "columns" of residues occupying each position in the real sets of aligned sequences were jumbled.

Preliminary comparative analysis of the amino acid sequences of the NTP-motif-eontaining proteins of positive-strand RNA viruses by use of the programs DIAGON and OPTAL (see Methods) revealed three distinct families and some additional proteins in whose close relatives the motif was not conserved. Within each family, all the proteins contained stretches at least 120 residues long that were similar to each other at a statistically highly significant level. For most pairs, the observed alignment scores exin this table. Question marks indicate that the real size of the respective proteins is not known; the large proteins presented in the table may in fact be processed. In those cases where sequences of several serotypes (strains) of a single virus species were available (specifically, for several picornaviruses and TMV), only one sequence was included. An exception is rhinovirus serotypes 14 and 2 with sequences that are substantially different, pr = product The NTP-motif-containing segments displaying statistically significant similarity within each family (see Table 1 ) are shown in black; they were aligned by the "A" sites of the NTP-motif (see text). The tricorna-and tobamovirus proteins are multidomain proteins, with the N-terminal domains similar to each other and to alphavirus nsPl protein (not indicated; Ahlquist et al. 1985) . In the bottom of A and B the respective patterns ofevolutionarily conserved amino acid residues are shown (designated "consensus 1" and "consensus 2"). Invariant residues are capitalized. Dots stand for variable residues (or gaps) within conserved residue clusters; the lengths of variable regions between these clusters are indicated by bracketed numbers. Asterisks denote the amino acid residues constituting the "A" site and the proposed Mg2+-binding D residue of the "B" site.

ceeded the mean scores for randomized sequences by at least 5 SD (see Methods). Such a level of sequence similarity between proteins is usually regarded as serious evidence for their monophyletic origin (Doolittle 1981 (Doolittle , 1986a Dayhoffet al. 1983) .

To obtain optimal group alignments for each family, the sequences were aligned in order of decreasing similarity (Fig. 1) . As is evident from the figure, the significance of the multiple sequence alignments was quite high for each of the families. The families were numbered 1 st, 2nd, and 3rd in order of decreasing sequence divergence between the presently recognized members 9 For part of the proteins constituting the 1st and the 2nd families, analogous sequence comparisons (but with no reference to the NTPmotif) were performed previously by other workers and meaningful similarities have also been observed Goldbach 1986) . Very recently, Domier et al. (1987) compared the sequences of the proteins of the 2rid and the 3rd families and noticed the presence of the "A" consensus of the NTP-motif. The I st family includes the NTP-motif-containing proteins (domains) of alpha-, tobamo-, tricorna-, furo-, and coronaviruses as well as the putative product of the hordeivirus RNA ~ open reading frame (ORF) 2, totalling 10 proteins (Table 1) . These proteins vary greatly in their size ( Fig. 2A) , genomic positions of the respective coding sequences, and modes of expression. For the proteins of this family, a statistically significant similarity was observed within a fragment of about 250 amino acid residues ( Fig. 2A) . This fragment contains 21 highly conserved residues (of which 14 are invariant) divided between seven clusters of unequal size (or individual residues). The first and third conserved clusters en-compass the "A" and "B" sites of the NTP-motif, respectively. The N-proximal five clusters of conserved amino acid residues in these proteins are separated from the sixth and seventh clusters by a variable region of about 60-90 residues. In fact, the conserved domain appears to be further divided into two subdomains, the N-terminal one containing the NTP-motif, and the C-terminal one of totally unknown function. Two specific points are worth noting. First and most remarkably, two segments of the furovirus genome encode two NTP-motif-containing proteins only distantly related (as compared to other members of the family) to each other; this is demonstrated by direct pairwise comparison of their sequences (data not shown). Second, the inclusion of the coronavirus NTP-motif-containing domain within the 1 st family is tentative, as the level of its sequence similarity to the other proteins of this family is not much higher than that between different families (see below). Also, the distance between the "A" and "B" sites of the NTP-motifis much longer in the coronavirus protein than those in the other proteins of this family. Nevertheless, all the amino acid residues invarianl in the latter are conserved in the coronavirus protein also (see below), justifying its inclusion in this family.

The 2nd family of NTP-motif-containing proteins includes picornaviral proteins 2C and comoviral protein p58, totalling nine proteins (Table 1) . These proteins are much more uniform in their size (Fig. 2B) , genomic positions of the respective genes, and mode of expression than those of the 1st family. The region of the most prominent similarity spans the central domain of about 130 amino acid residues; this domain contains 45 conserved residues (23 invariant), more or less evenly distributed (Fig.  2B , consensus 2). The "A" and "B" sites of the NTPmotif are located near the N-terminus and in the middle of the conserved domain, respectively.

The 3rd family includes CI proteins of two potyviruses (Table 1 and Fig. 2C ). These proteins are very similar to each other, having more than 50% identical amino acid residues. Thus, derivation of a consensus, like those derived for the other two families, made little sense. The "A" and "'B" sites of the NTP-motifare located in the N-terminal parts of CI proteins.

Comparison of the prealigned sequences of the three families of NTP-motif-containing proteins by the multiple alignment version of OPTAL yielded highly significant alignment scores for all three possible pairs (Fig. 1) . However, the final alignment of the sequences of the three families generated by this 261 program (not shown) was not quite satisfactory because the "B" sites of the NTP-motif, as well as some other clusters of residues that seemed good candidates for the conserved regions, did not coincide (although it must be pointed out that the "A" sites did match). Presumably, this might be due to different lengths of spacers separating these regions in the proteins of the three families. Thus, an overall alignment has been generated by manual fitting of the computer alignments of the three sets of sequences so as to maximize residue coincidence conserved within individual families (Fig. 3) . In this alignment five amino acid residues are strictly invariant, four additional residues are common to the 2nd and 3rd families, and three residues are conserved in the 1st and 3rd families. In addition, several positions in all, or nearly all, the sequences are occupied by functionally related residues (Fig. 3,  consensus) . All in all, a certain degree of conservation was observed in about 40% of the positions of the alignment (highlighted in Fig. 3 and further characterized in the legend to this figure) .

Strikingly, four of the five invariant residues are located within the "A" and "B" sites of the NTPmotif (Fig. 3) . These sites and short sequence stretches surrounding them also contain a considerable number of additional coincidences and similar sequence replacements between proteins of different families. Thus, the "A" and "B" consensus sequences and short adjacent segments are the most similar portions of the NTP-motif-containing proteins of positive-strand RNA viruses.

Of additional interest is a comparison of the positions of the NTP-motif-containing proteins (domains) in viral multidomain proteins; this approach is illustrated in Fig. 4 . Only two stretches of similar amino acid sequences are common to all viruses analyzed in this study: (1) the conserved region of the RNA polymerase Morozov and Rupasov 1985; Koonin et al. 1987 Koonin et al. , 1988 , and (2) the NTP-motif-containing domain (this paper). Viruses, with proteins that constitute the 2nd and 3rd families characterized above, possess an additional protein sequence of significant similarity, i.e., the proteases of picorna-, como-, and potyviruses Franssen et al. 1984; Carrington and Dougherty 1987; Domier et al. 1987) . In all viruses with nonsegmented genomes (with the probable exception of coronaviruses), in CPMV B polyprotein, and in the furovirus RNA 1 product (p237), the proteins (domains) containing similar sequence stretches are positioned in the same order within multidomain proteins, namely N-NTP-motif-containing domain-(protease)-polymerase-C (Fig. 4) . In coronaviruses the polymerase has not yet been identified. However, the results of our preliminary analysis indicate that the polymerase do- An overall alignment of the evolutionarily conserved segments of the vi~l NTP-moti~containing proteins of the three families. Only partial sequences of the conserved regions (Table 1 ) were aligned; they encompass the N-terminal subdomains of the proteins of the 1 st family, the sequences of the 2nd family without the five N-terminal amino acid residues, and complete sequences of the 3rd family. The residue numbers shown above the alignment are arbitrary; the numbering begins from the first residues of the aligned stretches and includes gaps. The sets of sequences of the three families aligned by the program OPTAL are separated by blank lines. Dots denote conservative positions. These are defined here as positions occupied by similar amino acid residues in at least 50% of the sequences of each of any two of the three families. The upper, middle, and lower rows of dots indicate the conservative positions of the 1st, 2rid, and 3rd families, respectively. Thus, if a given position in the alignment contains dots, say, in the upper and lower rows, this indicates the conservation of residues (in the above sense) between the 1st and 3rd families, and so on. Similar residues are defined as those belonging to one of the following groups: A, V, I, L, M, and F (hydrophobic); F, Y, and W (aromatic); G and A (small); S and T (hydroxy-); K, R, and H (basic); D, E, N, and Q (acidic and their derivatives); C and P have no similar residues. The pattern of highly conserved residues is shown under the aligned sequences, designated "cons" for consensus. Uppercase letters correspond to invariant residues, and lowercase letters to those conserved in two out of three families; in the latter ease, where a similar residue was conserved in the 3rd family, it was also indicated. * = a hydrophobic residue. The "A'" (positions 6-13 in the alignment) and "B" (positions 93-98) sites of the NTP-motif are denoted by horizontal bars above and below the alignment. For viruses with segmented genomes, the specific designations of the RNA segments encoding the NTP-motif-containing proteins are given in parentheses. main also resides in F2, but its position relative to the NTP-motif-containing one is reversed as compared to the "canonical" array described above (unpublished observations). Anyway, this single exception certainly does not invalidate the general trend for the specific positioning of these domains in viral rnultidomain proteins.

Comparative analysis of the amino acid sequences of all positive-strand RNA virus RNA polymerases provides a strong case for their monophyletic origin . We believe that the sequence similarity between the NTPmotif-containing proteins, together with their similar localization in viral multidomain proteins, indicate that they also constitute a monophyletic group.

In the course of the present study we screened all the available protein sequences of positive-stand RNA viruses for the presence of the NTP-motif. Also, some additional searches have been made: (1) domains occupying positions similar to those of the NTP-motif-containing ones in viral multidomain proteins were searched for the possible presence of degenerate forms of the motif; and (2) partially sequenced proteins were tested for similarity to the NTP-motif-eontaining proteins.

The "A" consensus sequence has been found in the C-terminal part of AIMV RNA polymerase, in the capsid protein of yellow fever virus (a flavivirus), in NS1 proteins of four flaviviruses, and in the F1 polyprotein of IBV; also, a second "A" sequence (besides the one included in our alignment) is present in the furovirus p237. In the first three instances the consensus sequence was not conserved in the relatives of the respective proteins, suggesting that its occurrence was most likely fortuitous. In the last two cases, the absence of other coronavirus and furovirus protein sequences precluded this type of analysis, leaving the significance of these observa- . Similar sequence stretches are also joined by sloped lines. The NTP-motif-containing protein ("NTPase"), the RNA-dependent RNA polymerase (polymerase), and the protease (the latter identified in picorna-, como-, and potyviruses) are designated. Nominations of specific proteins are given above each rectangle. Other designations are: ~, sites of proteolytic processing; ~, leaky termination codons [of two alphaviruses included, nsP4 is expressed only in SNBV via a leaky termination codon (Takkinen 1986) ]. All the information is given only for the NTP-motif-containing proteins (domains), the polymerases, and the parts of multidomain proteins enclosed between.

tions uncertain. However, it should be noted that the segments of F1 and of p237 encompassing the consensus sequence bear no significant sequence similarity to the viral NTP-motif-containing domains described above (unpublished observations). Flavivirus protein NS3, which occupies a position similar to that of alphavirus nsP2 in the polyproteins of these viruses, contains an "A" consensus sequence with a single deviation and a "B" sequence strikingly similar to those of the 1 st and 3rd families of viral NTP-motif-containing proteins. Comparison of the three available sequences of NS3, those of yellow fever, West Nile, and dengue 2 flaviviruses Castle et al. 1986; Yaegashi et al. 1986) , demonstrated strict conservation of these sequences. A more detailed analysis that we recently performed revealed statistically significant similarity between the putative NTP-binding domains of NS3 and those of potyviral proteins (unpublished observations). It seems quite plausible that NS3 may have some degree of evolutionary and functional relatedness to the group of viral proteins described in this paper.

A striking similarity has been detected between the C-terminal sequence of the protein p 120 encoded by BSMV RNA ~ [for which only a partial sequence has been reported (Rupasov et al. 1986) ] and the C-terminal subdomain of the 1 st family of viral NTP-motif-containing proteins. Although the N-terminal part of the p120 sequence is not yet known, in all other proteins of this family, invariably the two subdomains are observed together. Thus, the hordeivirus genome, like the furovirus genome, probably encodes two NTP-motif-containing proteins in two genomic segments .

In all other complete protein sequences of positive-strand RNA viruses reported, namely those of black beetle virus (a nodavirus), carnation mottle virus, and RNA bacteriophages, the consensus sequences of the NTP-motif have not been observed.

The NTP-motif was first introduced by Walker et al. (1982) and was subsequently employed for localization of putative catalytic sites and for prediction of NTP-binding capacity in numerous proteins. However, for reasons mentioned in the Introduction, the validity of the whole approach remained rather uncertain. In the present study we demonstrate that in a highly diverged group including similar proteins of positive-strand RNA viruses, the consensus sequences of the NTP-motif constitute the most strictly conserved stretches, encompassing four of the five invariant amino acid residues. Moreover, the NTP-motif-containing domain is one of the two most conserved domains revealed upon an overall comparison of the sequences of this class of virus proteins. This strongly suggests that this protein domain possesses NTP-binding capacity and possibly NTPase activity, presumably supplying some NTP-dependent function(s) that is of vital importance for viral reproduction. This hypothesis is in agreement with the available experimental data implicating these proteins in viral RNA replication and with the reported NTP-binding capacity of TMV p126 (Evans et al. 1985) , although direct testing is certainly warranted. In fact, there is experimental evidence that clearly, though indirectly, demonstrates the importance of the NTP-motif in viral RNA replication, Recently several poliovirus mutants resistant to or dependent on guanidine, a potent inhibitor of RNA replication of some picornaviruses, have been thoroughly studied (Pincus et al. 1986 (Pincus et al. , 1987 . They all mapped to the 2C protein, with the amino acid replacements located in the proximity of the "A" and "B" sites of the NTPmotif, or near the conserved Asn residue in the 183rd position of the segments of 2C aligned in this paper (Fig. 3) . Dever et al. (1987) have recently proposed a consensus for GTP-binding domains that includes, in addition to the "A" and "B" sites of the NTP-motif, a third highly conserved sequence element thought to determine the specificity for guanosine. They identified this sequence in the 2C protein of one serotype of FMDV (but not of the other picornaviruses) and suggested that this protein should possess specific GTP-binding capacity, as opposed to other picornaviral 2C proteins. However, it would be unprecedented for proteins so closely related to have different specificities for nucleotides. In our 265 opinion, it is much more likely that, within groups of highly similar NTP-motif-containing proteins such as picornaviral 2C, the substrate specificities and other principal properties should be identical. On the other hand, when considering more distantly related proteins, such as those belonging to the three distinct families described above, one cannot exclude the possibility that such proteins might differ significantly in their activities and functions in viral reproduction.

It seems premature to discuss at length the possible significance of the present observations for understanding the evolution of positive-strand RNA viruses. Two trends, however are obvious. First, NTP-motif-containing proteins are nearly ubiquitous among eukaryotic positive-strand RNA viruses. The existing classification of these viruses (Matthews 1982) includes about 30 families (groups), or somewhat more, taking recent developments into consideration. For 13 of these, complete, or nearly complete genomic sequences are available. Proteins containing the typical NTP-motif were observed in nine families (Table 1 ); in addition, viruses of one family (Flaviviridae) probably possess a functionally related protein with a deviant motif. It is tempting to speculate that an NTP-dependent function supported by the amino acid residues constituting the NTP-motif may be indispensable for positivestrand viral RNA replication; in some cases this function may be supplied by cellular proteins. In this context it is compelling that the RNA replicase of single-stranded RNA bacteriophages contains the translation elongation factor Tu, an NTP-motifcontaining GTPase, as one of its subunits (reviewed by Blumenthal 1979) . Second, it appears that the sequence diversity of the NTP-motif-containing proteins as revealed here does not precisely reflect the "phenotypic" diversity of viruses that forms the basis for the existing classification. Of the nine virus families (groups) having NTP-motif-containing proteins, six contribute members to the 1 st family of proteins (see above), two to the 2nd, and one to the 3rd family. Thus, the 1st family covers a very broad range of viral groups differing greatly in their genomic strategies and biological properties. It is anticipated that sequencing ofgenomes of new viral groups will add new members to this family.

The NTP-motif (or the "A'" consensus alone) has been identified in an extremely large class of NTPbinding proteins, mostly NTPases (although it should be noted that the presence of this motif is not an absolute prerequisite for NTP-binding capacity). The NTP-motif-containing proteins include a large group of GTPases, namely the RAS family, G proteins, transducins, and some of translation initiation and elongation factors (Dever et al. 1987 and references therein). Also belonging to this class are numerous proteins involved in bacterial DNA synthesis, recombination, and repair, and in membrane transport (Doolittle et al. 1986; Finch et al, 1986a,b; Higgins et al. 1986; Husain et al. 1986; Yin et al. 1986; Gilchrist and Denhardt 1987) , proteins implicated in multidrug resistance in mammalian cells (Chen et al. 1986; Gros et al. 1986) , and several NTP-utilizing enzymes of DNA viruses (Gorbalenya et al. 1985; Anton and Lane 1986; Doolittle 1986a; Astell et al. 1987, and references therein) . From this incomplete list it is obvious that the presence of the NTP-motif brings together numerous proteins with extremely diverse functions. It must be emphasized that many NTP-motif-containing proteins do not bear statistically significant similarity to each other (of. Argos and Leberman 1985; Doolittle 1986a ) and the existence of distinct monophyletic groups of such proteins (excluding very closely related, such as, for example, different RAS species) is not obvious a priori. Nevertheless, the NTP-motif-containing proteins of positive-strand RNA viruses do constitute such a family, whereas the GTPases probably constitute another.

Although widespread, NTP-motif-containing proteins are not strictly ubiquitous in all biological species. Specifically, this motif could not be found upon screening of the protein sequences of two large viral classes, negative-strand RNA viruses and retroid viruses (unpublished observations). Thus, the presence of proteins of this class in the majority of eukaryotic positive-strand RNA viruses appears to be a nontrivial observation, given their small genome size.

As for the value of the NTP-motif as a predictor of protein function, we believe that searching amino acid sequences for this motif (and conceivably for other sequence motifs of this kind) may be a very powerful methodology, if accompanied by phylogenetic analysis.

During preparation and reviewing of this manuscript, important relevant information became available. Genome sequences of viruses of three more groups that encode NTP-motif-containing proteins were determined. These are tobacco rattle virus [a tobravirus (Hamilton et al. 1987) ], white clover mosaic virus and potato virus X [two potexviruses (Forstcr et al. 1988; Krayev et al. 1988 )], and tomato black ring virus [a nepovirus (C. Fritsch, personal communication)]. The presumptive NTPbinding domain of the tobravirus is closely related to that of TMV and clearly belongs to the 1 st family of viral NTP-motifocontaining proteins described above. The genomes of potexviruses each encode two NTP-motif-containing proteins. These proteins also beiong to the 1 st family, but their inclusion in the alignment further loosens the consensus. Interestingly, in some positions of the potexvirus proteins, residues otherwise invariant in the 1 st family are replaced by those characteristic of the 2rid family. The nepovirus NTP-motif-containing protein belongs to the 2nd family. Thus, the new data appear to confirm our prediction that sequencing of genomes of viruses belonging to new groups should add members mainly to the 1st family of NTPmotif-containing proteins. Also, the genome sequence of southern bean mosaic virus, a sobemovirus, has been determined (Wu et al. 1987) . The authors claimed that it encoded a presumptive NTPbinding domain. However, a more detailed analysis indicates that this domain probably fulfills an entirely different function, namely the protease one, with its sequence being strikingly similar to those ofpicornaviral proteases (Gorbalenya et al. 1988a ). Thus, sobemoviruses may lack an NTP-motif-conraining protein, which is similar to other positivestrand RNA viruses of small genome size (namely nodaviruses, carnation mottle virus, and RNA phages).

Comparison of the sequences of the 1st family of viral NTP-motif-containing proteins with those of several bacterial helicases revealed highly significant similarity, suggesting an RNA helicase function for these proteins Gorbalenya et al. 1988b,c; Hodgman 1988) .

Nucleotide sequence of the brome mosaic virus genome and its implications for viral replication

Sindbis virus proteins nsPI and nsP2 contain homology to nonstructural proteins from several RNA plant viruses

Thenucleotide sequence of the coding region of tobacco etch virus genomic RNA: evidence for the synthesis of a single polyprotein

Non-structural protein 1 of parvoviruses: homology to purine nucleotide using proteins and early proteins of papovaviruses

Homologies and anomalies in primary structural patterns of nucleotide binding proteins

Similarity in gene organization and homology between proteins of animal picornaviruses and a plant comovirus suggest common ancestry of these virus families

Structural and functional homology ofparvovirus and papovavirus polypeptides

Q~ RNA replicase and protein synthesis elongation factors EF-Tu and EF-Ts

Completion of the sequence of the genome of the coronavirus avian infectious bronchitis virus

Nucleotide sequence of beet necrotic yellow vein virus RNA-2

Nucleotide sequence of beet necrotic yellow vein virus RNA-1

Consensus topography in the ATP binding site of the SV40 and polyomavirus large tumour antigens

Small nuclear inclusion protein encoded by a plant potyvirus genome is a protease

The complete nucleotide sequence of the RNA coding for the primary translation product of foot and mouth disease virus

Primary structure of the West Nile flavivirus genome region coding for all nonstructural proteins

RoninsonIB (1986) Internal duplication and homology with bacterial transport proteins in the mdrl (P-glycoprotein)gene from multidrug-resistant human cells

A common function for polyomavirus large-T and papillomavirus E 1 proteins

Homology between the proteins encoded by tobacco mosaic virus and two tricornaviruses

Complete nucleotide sequence of alfalfa mosaic virus RNA 1

Establishing homologies in protein sequences

GTP-binding domains: three consensus sequence elements with distinct spacing

The nucleotide sequence of tobacco vein mottling virus RNA

Potyviral proteins share amino acid sequence homology with picorna-, comoand caulimoviral proteins

Similar amino acid sequences: chance or common ancestry?

Protein sequence data banks: the continuing search for related sequences

Of URFs and ORFs. A primer on how to analyze derived amino acid sequences

Domainal evolution ofa prokaryotic DNA repair protein and its relationship to active transport proteins

Photoaffinity labeling of a viral induced protein from tobacco

Complete nucleotide sequence of the Escherichia colt recB gene

Complete nueleotide sequence ofrecD, the structural gene for the a subunit of exonuclease V of Escherichia coil

The complete nucleotide sequence of the potexvirus white clover mosaic virus

Zimmern D (1984) Homologous sequences in non-structural proteins from cowpea mosaic virus and picornaviruses

ATP-binding site of adenylate kinase: mechanistic implications of its homology with ras-encoded p21, F~-ATPase, and other nucleotide-binding proteins

Homology between human bladder eacrinoma oncogene product and mitochondrial ATP-synthase

Escherichia colt rep gene: sequence of the gene, the encoded helicase, and its homology with uvrD

Nucleotide sequence of tobacco mosaic virus RNA

Molecular evolution of plant RNA viruses

Prediction of nucleotide-binding properties of virus-specific proteins from their primary structure

Two segments of barley stripe mosaic virus genomic RNA encode two homologous proteins which probably possess NTPase activity

Sobemovirus genome appears to encode a serine protease related to cysteine proteases ofpicornaviruses

A conserved NTP-motif in putative helicases

A novel superfamily of nuclcoside triphosphatebinding motif containing proteins which are probably involved in duplex unwinding in DNA and RNA replication and recombination

Mammalian multidrug resistance gene: complete eDNA sequence indicates strong homology to bacterial transport proteins

The complete nucleotide sequence of RNA/3 from the type strain of barley stripe mosaic virus

I984) Regional homology in GTP-binding protooncogene and elongation factors

The complete nucleotide sequence of tobacco rattle virus RNA-1

Hermodson MA (1986) A family of related ATP-binding subunits coupled to many distinct biological processes in bacteria

The elucidation of protein function from its amino acid sequence

A new superfamily ofreplicative proteins

Sequences ofEscherichia coli uvrA gene and protein reveal two potential ATP binding sites

Structure of the GDP domain of EF-Tu and location of the amino acids homologous to ras oncogene proteins

Primary structural comparison of RNA-dependent polymerases from plant, animal and bacterial viruses

EvolutionofRNA-dependentRNApolymerases of positive strand RNA viruses

Evolution of RNA-dependent RNA polymerases of positive strand RNA viruses: a comparison of phylogenetic trees generated by different methods

ClarkBFC (1985) Structural details of the binding ofguanosine diphosphate to elongation factor Tu from Escheriehia eoti as studied by X-ray crystallography

Genome of coxsackievirus B3

The nucleotide sequence of cowpea mosaic B RNA

Classification and nomenclature of viruses

Phosphate-binding sequences in nucleotide-binding proteins

On the possibility of a common origin of the genes encoding the RNA polymerases of bacterial, plant and animal positive strand RNA viruses

Primary structure and gene organization of human hepatitis A virus

The nucleotide abd deduced amino acid sequences of the eneephalomyocarditis viral polyprotein coding region

Analysis of the complete nucleotide sequence of the picornavirus Theiler's murine encephalomyelitis virus (TMEV) indicates that it is closely related to cardioviruses

Guanidine-selected mutants of poliovirus: mapping of point mutations to polypeptide 2C

Guanidine-dependent mutants of poliovirus: identification of three classes with different growth requirements

Accelerated method for comparing amino acid sequences with allowance for possible gaps. Plotting optimum correspondence paths

Molecular cloning of poliovirus eDNA and determination of the complete nucleotide sequence of the viral genome

Nueleotide sequence of cucumber mosaic virus RNA 1

Nucleotide sequence of yellow fever virus implications for flavivirus gene expression and evolution

Nucleotide sequence of 3'-terminal regions of barley stripe mosaic virus RNAs 1 and 3

Matching sequences under deletion/insertion constraints

Human rhinovirus 2: complete nucleotide sequence and proteolytic processing signals in the capsid protein region

An interactive graphics programme for comparing and aligning nucleic acid and amino acid sequences

The complete sequence of a common cold virus: human rhinovirus 14

Complete nucleotide sequence of the genomic RNA of Sindbis virus

Complete nucleotide sequence of the nonstructural protein genes of Semliki Forest virus

Distantly related sequences in the a-and ~-subunits ofATP synthase, myosin, kinases and other ATP-requiring enzymes and a common nucleotide binding fold

Sequence and organization of southern bean mosaic virus genomic RNA

Partial sequence analysis of cloned dengue virus type 2 genome

Nucleotide sequence of the Escherichia coli replication gene dnaZX

Acknowledgments. The authors are deeply grateful to Professor V.I. Agol for constant interest and encouragement, to Dr. K.M. Chumakov for help with some of the computer programs, to Dr. S.Y. Morozov for useful discussions, and to Drs. C. Fritsch and S.Y. Morozov for communicating their sequence data prior to publication.

 I  f  I  f  I  I  I  f  f  I  I  I  I  I  I  f  f  I  f  I  I  I  I  I  C