The EMBO Journal vol.14 no.18 pp.4609-4621, 1995

A mutant T7 RNA as a DNA polymerase

Rui Sousa1 and Robert Padilla in dNTP Km (Ricchetti and Buc, 1993) and T7 DNA- directed RNA polymerase (RNAP) can also use RNA as Department of , University of Texas Health Science a template (Konarska and Sharp, 1989). These are not Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, exceptional observations, since it is a general property of TX 78212, USA that they display relaxed template specificity, 'Corresponding author at least in vitro. While template specificity may be relaxed, polymerase substrate specificity is normally extremely We have identified a T7 RNA polymerase (RNAP) stringent. T7 DNAP, for example, displays at least 2000- mutant that efficiently utilizes deoxyribonucleoside tri- fold selectivity for dNTPs over rNTPs, even in Mn2+ . In vitro this mutant will synthesize RNA, buffer, which relaxes the ability of the polymerase to DNA or 'transcripts' of mixed dNMP/rNMP composi- discriminate between dNTPs and ddNTPs (Tabor and tion depending on the mix of NTPs present in the Richardson, 1989). The structural determinants of such synthesis reaction. The mutation is conservative, stringent specificity remain undefined. changes Tyr639 within the active site to phenylalanine We report here the identification of mutant T7 RNAPs and does not affect specificity or overall that display the ability to use dNTPs. The mutations occur activity. Non-conservative mutations of this tyrosine in Tyr639 within motif B (Delarue et al., 1990) of T7 also reduce discrimination between deoxyribo- and RNAP. Two observations impelled us to examine the ribonucleoside triphosphates, but these mutations also substrate discrimination and miscoding properties of these cause large activity reductions. Of 26 mutations of mutants. It had been found that mutations in the corres- other residues in and around the active site examined ponding conserved tyrosine in DNAP I increased mis- none showed marked effects on rNTP/dNTP dis- coding (Polesky et al., 1990; Carrol et al., 1991). It was crimination. Mutations of the corresponding tyrosine also found that transcripts synthesized by a T7 RNAP in DNA polymerase (DNAP) I increase miscoding, Y639F mutant in vivo yielded 33-50% of the protein per though effects on dNTP/rNTP discrimination for the transcript compared with transcripts synthesized by the DNAP I mutations have not been reported. This con- wild-type enzyme (Makarova et al., 1995). The latter served tyrosine may therefore play a similar role in phenotype was unique to the Y639F mutant amongst a many polymerases by sensing incorrect geometry in number of other active site mutants examined for in vivo the structure of the substrate/template/product due to expression and indicated that Y639F transcripts contained inappropriate substrate structure or mismatches. T7 a defect that led to their being inefficiently translated. RNAP can use RNA templates as well as DNA templates These observations implied that mutations in Tyr639 and is capable of both primer extension and de novo might cause increased misincorporation, either increased initiation. The Y639F mutant retains the ability to use mismatch synthesis (miscoding) or incorporation of sub- RNA or DNA templates. Thus this mutant can display strates of inappropriate structure. We have therefore de novo initiated or primed DNA-directed DNA characterized the ability of the Y639 mutants, as well as polymerase, , RNA-directed a large number of other active site mutants, to miscode RNA polymerase or DNA-directed RNA polymerase or to use dNTPs in both Mg2+ and Mn2+ buffers. Our activities depending simply on the templates and sub- results point to a specialized role for Tyr639 in T7 RNAP strates presented to it in the synthesis reaction. (and the corresponding tyrosine in other polymerases) in Keywords: mutagenesis/RNA polymerase/substrate speci- ensuring that substrates to be added to the growing nucleic ficity/transcription acid have the correct structure. They reveal that both transcript and substrate structure affect the efficiency with which the transcript is extended. They show that the Introduction restriction of unprimed initiation to RNAPs is not due to an intrinsic property of ribo- versus , One classification of polymerases relies on but simply to the selectivity of the polymerase active site. their different template specificities (RNA or DNA), They also present researchers with a novel reagent that substrate specificities (rNTPs or dNTPs) and mode of expands the structural range of nucleic acids that can be initiation (de novo or primed). These designations usually enzymatically synthesized in vitro. refer to the template and substrate specificities displayed in vivo during the fulfillment of a polymerase's biological Results function. In vitro, polymerases can display novel activities, albeit with reduced efficiency and/or under non-physio- Structure of the transcripts synthesized by Y639F logical conditions. Escherichia coli DNA-directed DNA and the wild-type enzyme with rNTPs and dNTPS polymerase (DNAP) I, for example, can use RNA as a Figure 1 shows transcription reactions carried out with template, though with a concomitant -100-fold increase the wild-type enzyme or the Y639F mutant polymerase

K Oxford University Press 4609 R.Sousa and R.Padilla

A BIC FIG HTI J IK N|O R|S T|U X geneous sequence abortive transcripts or 59 base run-off DIEI I I LIM PIQ VIW Pol: MWTMu WT Mu WT M.lu WT M. WT M.u WT M. WTM.lWTuWT Mu WT MIu WT Mu WTM.u products made and instead long poly(G) transcripts, as in rGTP rGTP rGTP rGTP rGTP rGTP rGTP rGTP rGTP rGTP rGTP rGTP are NTPs: rATP dATP rATP rATP rATP rATP rATP rATP dATP dATP lane A, made. Adding dATP to reactions lacking rATP rCTP rCTP rCTP dCTP rCTP rCTP rCTP rCTP dCTP dCTP I rUTP rUTP rUTP rUTP rUTPI dTTP dTTP rUTP dTTP does not change the transcripts produced by the wild-type

Rutlo - M enzyme (lane G). However, with the mutant enzyme we observe that addition of dATP (lane H) allows synthesis of a long run-off transcript, as well as synthesis of heterogeneous sequence abortive transcripts that do not co-migrate with the poly(G) transcripts. Observation of an abortive transcript in lane H running near the position of the 4H band in lane D confirms extension of the GGG trimer with an A, but note that the major 4mer transcript in lane H migrates close to, but not precisely with, the major 4mer in lane D or in adjacent lane I. This is consistent with the expectation that these 4mers will have identical sequence and length, but different structure (i.e. ._a rGrGrGrA in lanes D and I, rGrGrGdA in lane H). It should also be noted that some poly(G) transcript .7 ._w synthesis is observed in lane H. For example, in lane H we observe both a heterogeneous sequence 4mer migrating near the _=11111o ..:. 4H position and a smaller amount of 4mer band migrating ,LH 04 :-W 4 at the _w 4G position. When four rNTPs are present (lane C or D) synthesis of poly(G) transcripts is more completely suppressed. This indicates that dATP is utilized by Y639F, X - but not as efficiently as rATP. -,- Ck When rCTP is omitted from the reaction, transcripts .. terminate predominately at the 6mer length, because rCMP is normally first incorporated at position 7 (lanes I and J). Fig. 1. Structure of transcription products produced by Y639F and Addition of dCTP does not allow extension of the wild-type T7 RNAP in the presence of various combinations of rNTPs 6mer and dNTPs. The template was pT75 (Tabor and Richardson, 1985) cut in reactions with the wild-type enzyme (lane K). However, with HindIH so that transcription from its T7 promoter generated a 59 addition ofdCTP to reactions with Y639F allows extension base run-off transcript. Electrophoresis was on a 20% polyacrylamide- beyond the 6mer length and synthesis of the run-off 6 M urea gel. and polymerases were at concentrations of transcript (lane L). Again, the following should be noted: l0e M and NTP concentrations were 0.5 mM (all rNTPs and dTTP), (i) transcripts larger than 6 bases do not 1 mM (dATP, dGTP) or 5 mM (dCTP). [,y-32P]GTP was added to co-migrate with radiolabel the transcription products. Wild-type (WT) or Y639F mutant their counterparts in lane C or D, consistent with the (Mu) polymerases and NTPs used are as indicated. Poly(rG) products expected structural difference despite length and sequence of various sizes are labeled in lane a (2G, 3G, etc.) and heterogeneous identity; (ii) there is more termination at the 6- and 7mer sequence abortive transcripts of different lengths are indicated by 4H, points in lane L than in lane C or D, that 5H, etc. in lane C. Lanes Q-T are a 10-fold longer exposure of lanes indicating Y639F M-P. uses dCTP well, but not as efficiently as it utilizes rCTP. In lanes M and N UTP was omitted from the reactions. Lanes Q-T show a 10-fold longer exposure of lanes M- and a T7 d10 promoter template. Transcription by T7 P. Within the set of four NTPs UTP is unique on this RNAP, like other RNAPs, is characterized by an initial, template, since it first becomes incorporated into the poorly processive 'abortive' phase of transcription during transcript at the 13 base position. This corresponds to a which the short, nascent transcript frequently dissociates transcript length subsequent to the transition from abortive from the ternary complex. When the transcript reaches a to processive transcription. As a consequence of this length of ;-9 bases transcription becomes highly processive transition the ternary complex becomes more stable and the transcript becomes stably associated with the (Martin et al., 1988). Therefore, when transcript extension elongation complex. On this promoter, which initiates is blocked during the processive phase of transcription with GGGAGACCGGAAU, T7 RNAP can also synthesize the stalled ternary complex does not rapidly dissociate long poly(G) ladders (labeled 2G, 3G, etc. in lanes A and (Shi et al., 1988). Instead, it remains stalled on the B) when rGTP is the sole NTP present (Martin et at., template, near the promoter, and blocks re-initiation. For 1988). Lanes C and D display the transcription products this reason we observe a large decrease in the overall typical of run-off reactions containing all four rNTPs: amount of transcription when UTP is omitted from the there are abortive transcripts ranging up to 8 bases in reaction in lanes M and N. A longer exposure of lanes m length and a 59 base run-off product. Note that at the and N (lanes Q and R) does, however, reveal transcription 4mer length the sequences of the poly(G) transcripts and products of the expected structure. When TTP is added the abortive transcripts made in the presence of all four to these reactions (lanes 0, P, S and T) synthesis of the rNTPs (labeled 4H, 5H, etc.) diverge and no longer co- 59 base run-off transcript is observed with both the wild- migrate with the poly(G) transcripts of equivalent size. type and Y639F mutant. The ability of the wild-type When rATP is omitted from the transcription reactions enzyme to extend transcripts with TTP when it is unable (lanes E and F) normal elongation of the initially synthe- to extend transcripts with the other dNTPs is also likely sized GGG trimer cannot occur. There are no hetero- to be related to the unique position at which UTP/ITP 4610 An RNA polymerase as a DNA polymerase first becomes incorporated into the transcript. The amount A B C D E F G H I J K L A B C D E F G H J Superoded Linearized L S Li S Sc Sc L Sc Sc of transcript termination or extension that occurs with a Pol: Mu W.T. Mu W.T. Mutant Wild Type _P,,>:IG' P o F-c P dG P -C particular dNTP depends simply on the relative rates NTPs: Ti cTccPCP A!;TC T -rAc r:GT-crT- cT- c1 r_ G-5 MT GP - 'I," 'A P LTPr U P T C-P of ternary complex dissociation or dNMP incorporation -TPs:C- LO (McClure and Chow, 1980). During abortive transcription complex dissociation must be more rapid than the rate at which the wild-type enzyme incorporates dNMPs into its transcripts. During processive transcription even the expected slow rate of dNMP incorporation by the wild- type enzyme must be competitive with the slow rate of dissociation of the stable elongation complex and an ability of the wild-type enzyme to incorporate dNMPs becomes manifest. Because elongation during the pro- cessive phase oftranscription is fast (-230 bases/s; Golomb and Chamberlin, 1974), while initiation and progression through abortive transcription is slow (Martin and Col- eman, 1987; Martin et al., 1988), elongation of the transcript from 11 to 59 bases is expected to contribute <10% to the time required for transcript synthesis. As a consequence, the phase of transcription during which TTP is incorporated may not be the rate limiting step in synthesis of the run-off transcript, even for the wild-type I*~~~~~~~~~3GtS' ** II;<~~~~~~;-~.r *8 enzyme. These considerations imply that we need not expect to see marked differences in synthesis of a relatively short run-off transcript by Y639F or the wild-type enzyme when T7 P is substituted for UTP. Fig. 2. Effect of dGTP substitution on transcription by the wild-type Lanes U and V show reactions in which two rNTPs and Y639F polymerases. Polymerases, template [supercoiled (Sc) or HindIlI linearized (Li) pT75] and NTPs were as indicated. Polymerase have been substituted by dNTPs (dATP and dCTP) and and template concentrations and electrophoresis conditions as in lanes W and X show reactions in which three rNTPs have Figure 1. (Left panel) Labeling was with [a-32P]rGTP (A, D, G and J) been substituted by dNTPs (dATP, dCTP and TTP). or [a-32P]dGTP (all other lanes) The run-off transcript from the Synthesis of the 59 base run-off transcript is observed HindIll-cut template is indicated in lane F. (Right panel) Labeling was with mutant V and X), indicating with [a-32P]rGTP (A, B, F and G) or [a-32P]dGTP (all other lanes). the polymerase (lanes Poly(rG) and poly(dG) products of various sizes are indicated in lanes that Y639F can carry out synthesis even when three rNTPs A, C and D. Alignment of these transcript patterns with those in lanes are substituted by dNTPs. While the abortive transcript B, C, H and I in the left panel reveals that the added complexity of patterns in lanes V and X may largely be described as a the transcript pattern in the latter set of lanes is due to the presence of combination of the patterns observed in lanes H and a mixture of heterogeneous sequence and poly(G) transcripts. Heterogenous sequence abortive transcripts are indicated in lane C of L, there are some important distinctions. The transcript the left panel (4H and 5H). patterns in lanes V and X reveal that the structure of the transcript, as well as the structure of the NTP, affect the rate at which NMPs are added to the transcript. For (2dG, 3dG, etc.) in lane C do no co-migrate with the example, in lanes V and X there is an increase in poly(rG) transcripts (2rG, 3rG, etc.) in lane A, despite termination at the 6mer length relative to lane 1 [the 6mer length and sequence identity. When a supercoiled template in lanes V and X runs slightly slower than the poly(G) is used, poly(dG) transcripts up to 5 bases in length are Smer in the adjacent lane]. In lanes L, V and X synthesis obtained (lane D, right panel). In lane E rGMP is added of the 7mer involves incorporation of dCMP, but in lanes to reactions which contain only dGTP. rGMP can serve V and X the 6mer contains a dAMP at its 3'-end and in as the initiating, but not elongating, during lane L it contains rAMP. The increase in termination at transcription (Martin and Coleman, 1989). With rGMP the 6mer point in lanes V and X indicates that the presence we therefore ask whether either polymerase can elongate of a dNMP on the 3'-end of the transcript reduces the with dGTP if an rNMP is provided for initiation. Addition ability of the polymerase to incorporate further NMPs. of rGMP to reactions with dGTP further extends the The high level of termination after synthesis of the lengths of the transcripts obtained with the mutant poly- rGrGrGdA 4mer in lane H relative to the observed merase (lane E, right panel). With the wild-type enzyme termination after synthesis of the rGrGrGrA 4mer in very little synthesis is observed in reactions with dGTP lanes C and D similarly indicates that the presence of (lanes H-J, right panel), though the normal pattern of deoxyribonucleotides in the transcript, at least at the 3'- poly(rG) synthesis is observed in the rGTP reactions (lanes end, influences subsequent extension of the transcript. F and G, right panel). Figure 2 reveals the effects of substituting dGTP for When reactions contained three rNTPs and dGTP, rGTP in transcription reactions with the wild-type or synthesis of run-off transcript from the HindIII-cut tem- Y639F polymerases. In reactions containing only dGTP plate is reduced much more than in reactions in which and HindIII-cut pT75 as the template the mutant poly- rGTP is present but other are substituted merase synthesizes poly(dG) transcripts up to 4 bases in by deoxyribonucleotides. For example, there is very little length (lane C, right panel). Note that, consistent with the run-off transcript in lane I of the left panel of Figure 2. assumed structural differences, the poly(dG) transcripts Addition of rGMP increases the amount of run-off tran- 4611 R.Sousa and R.Padilla script made by the mutant enzyme in a reaction containing transcribed region may influence the efficiency with which dGTP (lane H), but the amount of run-off transcript is still Y639F can extend the transcript when using dNTPs. much less than in reactions with rGTP. On a supercoiled Figure 2 also shows that supercoiling, which presumably template (lanes A-C, left panel) high levels of long facilitates unwinding of the template, enhances the activity transcripts are obtained with the mutant enzyme in reac- of Y639F when using dNTPs. To evaluate the effects of tions with dGTP. The wild-type enzyme shows no tran- sequence and single-strandedness in the initially tran- script synthesis in any of the reactions with dGTP, scribed region on the activity of Y639F when using irrespective of whether supercoiled templates or rGMP dNTPs we examined transcription from a set of synthetic are used. promoters which differed in the sequence of their initially Examination of Figure 2 shows that the marked reduc- transcribed regions and in being fully double-stranded tion in run-off transcript synthesis by the mutant enzyme or single-stranded in their initially transcribed regions in reactions with dGTP is not due to a deficit in initiation. (Figure 3). In fact, in all of the reactions with dGTP we observe For most of the promoters tested transcription with abundant synthesis of 2 to -6 base transcripts with rNTPs was not markedly affected by having the initially the mutant enzyme. The low level of run-off transcript transcribed region single-stranded. However, when tran- synthesis means that these short transcripts are being scribing with four dNTPs Y639F was more active on the inefficiently extended to greater lengths. It should also be partially single-stranded promoters. For example, in lane E noted that the abortive transcript pattern seen in lanes B, (partially single-stranded promoter) of Figure 3 transcrip- C, H and I of the left panel in Figure 2 is too complex to tion products are both more abundant and extend to greater be accounted for by presuming one product of each base lengths than in lane D, where the promoter is fully double- length. Aligning the transcripts produced in the presence stranded. A similar comparison may be made between of dGTP only with those produced with dGTP, rUTP, lanes M and L or S and R. Regarding sequence, we found rCTP and rATP allows us to identify the predominant that a promoter which initiated with three Gs was superior abortive transcripts in lanes C, D, H and I of the left panel to a promoter which initiated with two, which was in turn as poly(dG) products. In lane C of the left panel we superior to a promoter which initiated with just one G. indicate the major non-poly(dG) abortives by 4H and 5H. Thus Y639F activity when using dNTPs was greatest on In lanes B, C, H and I of the left panel of Figure 2 we a promoter which initiates GGGAGACC (lanes L and therefore see a pattern similar to that of lane H in Figure M). The initially transcribed region of this promoter 1. The presence of a mixture of normally extended corresponds to the consensus sequence for T7 promoters heterogenous sequence and poly(G) abortive transcripts in the +1 to +6 segment. This was the only promoter indicates that 'normal' transcript extension is inefficient. which when fully double-stranded gave rise to high levels It has been shown that mutations in T7 RNAP that of transcript synthesis with Y639F in reactions containing reduce phosphodiester bond formation rates cause poly(rG) only dNTPs (lane L). Promoters which initiated with two transcript synthesis even when four rNTPs are present Gs (lanes D, E and 0) gave lower levels of transcript (Bonner et al., 1994). It can be seen that the ratio of synthesis in the four dNTP reactions and a promoter which poly(G) to heterogeneous sequence transcripts in lanes B, initiated with one G (lanes H and I) was not utilized by C, H and I of the left panel of Figure 2 is greater than in Y639F in reactions with four dNTPs. Within the initially lane H of Figure 1, indicating a greater deficiency in transcribed region elements other than the number of Gs normal transcript extension when dGTP is substituted for appear to be important. For example, we have found that rGTP than when dATP is substituted for dGTP. Note that the wild-type or Y639F polymerases are less efficient in this occurs even though normal extension of the dGdGdG initial transcription extension on the T7 promoter found trimer in lane I of the left panel of Figure 2 (for example) in the pBS plasmid, which initiates GGGC, than on the would involve addition of a riboAMP, while extension of 010 promoter found in pT75, which initiates with the the rGrGrG trimer in lane H of Figure 1 would involve consensus GGGAGA (data not shown). Another example addition of a deoxyriboAMP, clearly highlighting the role is evident in lanes N and 0, which show the transcripts of transcript structure, as well as substrate structure, in obtained on a partially single-stranded promoter that determining the efficiency of transcript extension. These initiates GGAAAAUU. Like the promoter used in lanes results show that Y639F can initiate and elongate tran- C and E, this promoter initiates with two Gs, but normal scripts with dGTP substituted for rGTP, but normal exten- transcript extension on this promoter is less efficient than sion of the transcripts in the 2-8 base range is impaired, on the promoter that initiates GGACU. In the four rNTP leading to a large increase in the proportion of poly(dG) reactions (lanes C and N) the proportion of short transcripts and short transcripts synthesized. Addition of rGMP to (dimers and trimers) is greater in lane N and we observe serve as the initiating nucleotide and the use of a super- significant amounts of poly(rG) transcripts beyond dimer coiled template both enhance the ability of the mutant length in lane N, but not in lane C. In the four dNTP to extend short transcripts during the initial stages of reactions almost all of the transcripts in lane 0 terminate transcription. at the trimer or tetramer length. Since increasing the number of Gs from one to three enhanced Y639F activity Barriers to initiation and extension of the initial when using dNTPs, we tested a promoter which initiated transcript with dNTPs with a run of 11 Gs (lanes P-S). A potential drawback to Figure 2 reveals that Y639F can efficiently transcribe the such a promoter is that such a long run of Gs could inhibit initial G segment of a promoter that initiates GGGA with the ability of the polymerase to unwind the template. dGTP, but is severely blocked in extending the dG trimer Since T7 promoters with more than three consecutive Gs with A. This implies that the sequence of the initially in the initial region do not occur naturally, it may be that 4612 An RNA polymerase as a DNA polymerase

Fig. 3. Effects of single-strandedness and sequence in the initially transcribed region on the activity of Y639F in reactions with four rNTPs or four dNTPs. Poly(rG) transcripts of various sizes are indicated in lane A. Reactions contained the indicated NTPs and polymerases. Polymerase and promoter concentrations were 10-6 and 10-5 M respectively. NTP concentrations and electrophoresis as in Figure 1. Indicated in some of the lanes are the sequences of different transcripts as deduced from alignment with poly(rG) or poly(dG) ladders synthesized in the presence of rGTP or dGTP only. The synthetic promoter templates used are double-stranded and have the sequence CGAAATTAATACGACTCACTATA in their -23 to -1 regions. The promoters differ in their initially transcribed regions as follows: B-E, T and U, GGACT; F-J, V and W, GAGACCGG; A, J-M, X and Y, GGGAGACC; N, 0 and Z, GGAAAATT; P-S, GGGGGGGGGGGACT. The promoters also differ in being double-stranded (B, D, F, H, J, L, P and R) or single-stranded (other lanes) in their transcribed regions. for other reasons such sequences do not favor initial Table L. Relative selectivitya of Y639F and wild-type polymerases for transcript extension. In fact, we find that this promoter is rNTPs versus dNTPs a poor template. When it is fully double-stranded initiation and extension of the transcript is inefficient with either rATP/dATP rUTP/dTTP rCTP/dCTP rGTP/dGTP or dNTPs (lane R), consistent with the rNTPs (lane P) Y639F 8.5-10 (Mg)b 1.7-1.9 2.4-20.5 0.93-1.6 expectation that a promoter with this sequence would be 0.9-1.0 (Mn)b 0.48-0.76 0.55-0.80 0.98-0.99 difficult to melt. Initiation and transcript extension is Wild-type 72-83 22-25 110-150 51-67 enhanced when this promoter is partially single-stranded 6-14 2.3-2.8 4.3-5.1 4.4-6.3 (lanes Q and S), but while poly(G) transcripts from 2 to aReactions were carried out with all four rNTPs (0.5 mM) in great 7 or more bases in length are abundant, run-off transcripts excess over radioactive rNTPs or dNTPs. Relative selectivity was of the expected length are not predominant products. In determined from the relative percentages of radioactive rNTP versus reactions with four rNTPs the transcript patterns of the dNTP incorporated into RNA. Maximal incorporation was less than wild-type and Y639F polymerases are virtually identical, -30% of total input radioactivity so as to limit effects due to changing not the four rNTP reactions with the NTP concentrations during the experimental time course. The numbers so we do repeat shown give the range for two experiments. bFor each rNTP/dNTP and wild-type enzyme in Figure 3. Lanes T-Z show that the polymerase the upper numbers are those obtained in Mg2+ buffer, the wild-type enzyme is virtually inactive in reactions with lower numbers are from Mn2+ buffer. Polymerases at 1078 M. four dNTPs and the same set of promoters used in lanes Template was supercoiled pT75 at l0e7 M. B-S. determining the efficiency of transcript extension. To Relative selectivity of the mutant and wild-type obtain a quantitative measure of the relative selectivity of polymerases for dNTPs and rNTPs the wild-type and mutant polymerases for dNTPs versus Figures 1-3 present a qualitative analysis of the structure rNTPs under conditions where transcript structure was not of the transcripts produced by the wild-type and Y639F a complicating factor we carried out reactions in which polymerases with various combinations of NTPs. They all four rNTPs were present, but rNTPs or dNTPs were show that Y639F can use dNTPs with high efficiency and used to radiolabel the transcripts (Table I). Under these that both transcript and substrate structure play a role in conditions the unlabeled rNTPs were present in vast excess 4613 R.Sousa and R.Padilla

Table H. Relative activity of Y639F and wild-type polymerase with different rNTP/dNTP mixes Wild-type (Mg2+) Y639F (Mg2+) Wild-type (Mn2+) Y639F (Mn2+) 4 rNTPs 200 200 21-23 9 3 rNTPs, dTTP 11-13 90-96 7-8 4 3 rNTPs, dUTP 9-11 82-91 10-11 7-8 3 rNTPs, dATP 1-2 73 1-2 2-3 3 rNTPs, dCTP 4-9 86 15 7-11 3 rNTPs, dGTP, rGMP <0.5 95-109 1-3 1.4-2.5 3 rNTPs, dGTP <0.5 43-63 0.5-2 3 2 rNTPs, dCTP, dUTP 2-6 27-29 3 4-7 2 rNTPs, dCTP, dTTP 2 30 5-7 5 2 rNTPs, dCTP, dATP <0.5 20-29 0.5-0.9 4-7 2 rNTPs, dTTP, dGTP, rGMP <0.5 13-15 <0.5 3-6 2 rNTPs, dATP, dTTP <0.5 11-14 <0.5 3-4 2 rNTPs, dCTP, dGTP, rGMP <0.5 11-14 <0.5 2-5 2 rNTPs, dATP, dGTP, rGMP <0.5 5-6 <0.5 1-1.5 1 rNTP, dCTP, dATP, dTTP <0.5 12-14 <0.5 0.7-2 1 rNTP, dCTP, dATP, dUTP <0.5 10-11 <0.5 2-3 1 rNTP, dTTP, dCTP, dGTPa <0.5 10-13 <0.5 2-4 1 rNTP, dTTP, dATP, dGTPa <0.5 11 <0.5 1.5-2 1 rNTP, dCTP, dATP, dGTPa <0.5 3-5 <0.5 1-2 4 dNTPs, rGMP <0.5 <0.5 <0.5 0.5 aThese reactions also contain rGMP. Numbers give ranges from two experiments. Template was supercoiled pT75 (10-7 M), polymerases at 10-8 M (in Mg2+ buffer) or l0-7 M (in Mn2+ buffer). rNTPs, rGMP, dTTP were at 0.5 mM; dATP, dGTP were at 1 mM; dUTP was at 2.5 mM, dCTP was at 5 mM. From top to bottom the labeling NTPs were: [a-32P]rGTP, [a-32P]rCTP, [a-32P]rCTP, [a-32P]dATP, [a-32P]dCTP, [a-32P]dGTP, [a-32P]dGTP, [a- P]dCTP, [a-32P]dTrP, [c-32P]dCTP, [a-[P]dTP, dTTP, [a-32P]dCTP, [a-32P]dATP, [a-32P]dTTP, [a-32P]dCTP[a-32P]dTTP, [a-32P]dTTP, [a-X32P]rCTP, [a-32P]dCTP. relative to the labeling NTPs, so labeling dNTPs are active when rCTP and rUTP were substituted by dNTPs, almost always incorporated adjacent to rNTPs and into corresponding to the two added latest to the transcripts of nearly uniform rNMP structure. We found transcript. The wild-type enzyme was inactive with all that on average the mutant enzyme is -20-fold less other two or three dNTP combinations. In the 'two dNTP' selective for rNTPs over dNTPs than the wild-type reactions Y639F was least active in the dGTP, dATP enzyme. We used this assay to screen our collection of reaction, corresponding to the two nucleotides incorporated T7 RNAP active site mutants for increased dNTP utiliza- first during transcription. In the 'three dNTP' reactions tion. In this screen we looked for increased dATP incorp- Y639F was least active in the dGTP, dATP, dCTP reaction, oration in transcription reactions with all of these mutants corresponding to the three nucleotides incorporated first using supercoiled pT75 as the template, four cold rNTPs during transcription. and [32P]rATP or [32P]dATP to label. Since these results In Mn2' buffer both the wild-type enzyme and Y639F were negative with the exception of the Tyr639 mutants, showed a reduction in their sensitivity to substitution of we do not present them here, but the mutants tested in dNTPs for rNTPs, consistent with an expectation of this way are listed in Materials and methods. It has been reduced substrate discrimination in Mn2+ buffer. There reported that use of Mn2+ instead of Mg2+ decreases was, however, also a sharp reduction in overall activity substrate discrimination and increases miscoding for a with Mn2+. We varied Mn2+ concentrations over a wide number of polymerases (Niyogi and Feldman, 1981; Tabor range (20-150 mM in 2-fold dilutions) to determine if an and Richardson, 1989). We therefore examined the rNTP/ optimal Mn2+ concentration that would result in high dNTP selectivity of the mutant and wild-type enzymes in activity could be identified, but we found similar activity Mn2+-citrate buffer (Tabor and Richardson, 1989). With at all Mn2+ concentrations tested (data not shown). Thus Mn2+ the preference of both the wild-type and Y639F while discrimination between rNTPs and dNTPs was less polymerases for rNTPs over dNTPs was markedly reduced. in Mn2+ buffer, Y639F is more active in Mg2+ buffer than in Mn2+ buffer with all NTP combinations examined. Relative activity of the wild-type and Y639F Similarly, the wild-type enzyme exhibits greatly reduced polymerases with different NTP combinations discrimination between rNTPs and dNTPs in Mn2+ buffer, The relative activities of the Y639F and wild-type poly- but is modestly more active in Mn2+ buffer than in Mg2+ merases with supercoiled pT75 as a template and various buffer only for certain combinations of dNTPs and rNTPs. combinations of rNTPs/dNTPs were measured in both Mg2+ and Mn2+ buffers (Table II). In Mg2+ buffer DNA and RNA synthesis on homopolymeric substitution of a single rNTP by a dNTP reduces wild- templates type activity by 20- to >400-fold, but only modestly T7 RNAP will synthesize poly(rG) on poly(dC) reduces the activity of Y639F. The rank order of the effect templates (Ikeda and Richardson 1987; Bonner et al., of a particular dNTP substitution on wild-type enzyme 1994). We measured the activity of the wild-type and activity, dGTP > dATP > dCTP > dT'TP = dUTP, mutant polymerases on poly(dI)-poly(dC) with rGTP, matches the order of their addition to the transcript. With dGTP and dGTP+rGMP (Table III). The activity of T7 the 'two dNTP' reactions the wild-type enzyme was most RNAP on poly(dI).poly(dC) and poly(dC) is especially 4614 An RNA polymerase as a DNA polymerase

Table III. Relative activity on poly(dI)-poly(dC) Wild-type Y639F G640A Y639A Y639S rGTP 1000 1000 240 (200-270) 145 (142-151 ) 48 (47-50) dGTP 7.4 (5.4-12.5) 964 (684-1257) <5 5.3 (4.8-5.5) 0.6 dGTP+rGMP 25 (20-27) 1070 (816-1457) <5 25 (17-30) 4.4 Numbers give mean and range from three experiments. Templates were at 0.2 mg/ml, polymerases at 10-8 M. Labeling NTPs were [a-32P]rGTP, [a-32P]dGTP, [a-32P]rATP and [a-32P]dATP, as appropriate. rNTPs or rNMPs at 0.5 mM, dNTPs at 1 mM. robust. Mutant polymerases that have greatly reduced activity on normal promoter templates still display high activity on poly(dC) templates (Bonner et al., 1994). We therefore characterized two poorly active non-conservative tyrosine mutations on this template (Y639A and Y639S). In Table III we also present results obtained with mutant G640A. Presentation of data for the latter mutant was selected because it is more comparable in activity with the Y639A/S mutants and because it is representative of a mutation which has marked effects on the kinetics of transcription but does not affect substrate discrimination, even though it is directly adjacent to Y639. We find that all the Y639 mutants exhibit reduced substrate discrimina- tion, as demonstrated by the fact that their differential activity in reactions containing dGTP or dGTP+rGMP versus rGTP is less than for the wild-type enzyme or the G640A mutant. In fact, Y639F displays similar activity with rGTP, dGTP or dGTP+rGMP. Since the nucleic acid synthesized by Y639F on poly (dI).poly(dC) using dGTP is presumably composed solely of dNMPs, it is expected to be resistant to alkaline hydrolysis (Schmidt and Tannhauser, 1945). Figure 4 shows the transcription products obtained with the wild- type or Y639F polymerases on poly(dI).poly(dC) before and after treatment with alkali. With dGTP or dGTP+rGMP the wild-type enzyme is poorly active in the synthesis of long transcripts, however, we can observe smears of heterogeneously sized, short transcripts in the reactions with the wild-type enzyme and the dGTP sub- strates (lanes E and F), while Y639F synthesizes higher levels of long transcripts which are retained near the top Fig. 4. Transcription by Y639F and wild-type polymerase with dGTP of these gels (lanes B and C). The presence of these short or rGTP on poly(dI)-poly(dC). Transcription reactions were carried out transcripts indicates that Y639F and the wild-type enzyme with poly(dI)-poly(dC) at 0.2 mg/ml and the indicated NTPs and differ only in the degree to which they can utilize polymerases. Reaction products were left untreated (-) or treated with The can also initiate and extend 1 M NaOH for 5 h at 37°C (+). Polymerase and NTP concentrations dNTPs. wild-type enzyme and electrophoresis as in Figure 1. Labeling was with [a-32P]rGTP transcripts with dGTP, but it is much less processive when (A, D, G and J) or [a-32P]dGTP (other lanes). using dNTPs than Y639F, so its transcripts are much shorter. When these reactions are treated with base, polymerases on poly(rC) with rGTP was 10- to 20-fold degradation of the long transcripts made in the reactions less than on poly(dC) (not shown), but this reduction did with rGTP is observed and the amounts of short RNAs not preclude synthesis of high levels of RNA on poly(rC) (presumably hydrolysis products) increase (lanes G and by using higher polymerase concentrations than were J). In the reactions in which dGTP or dGTP+rGMP were used in the poly(dI)-poly(dC) reactions. When dGTP or used as substrates no degradation of the transcripts by base dGTP+rGMP was used the wild-type enzyme was not treatment is observed, confirming that these transcripts are measurably active on poly(rC), while the activity of Y639F composed of dNMPs. was reduced by only -4-fold (with dGTP+rGMP) or -8- fold (with dGTP). Thus both the wild-type and Y639F T7 RNAP as a reverse transcriptase or RNA polymerases are capable of unprimed RNA-directed RNA replicase which initiates de novo polymerization, while Y639F is also capable of unprimed It has been reported that T7 RNAP can use both RNA reverse transcription. and DNA templates (Konarska and Sharp, 1989). We therefore determined whether the Y639F mutant would DNA- and RNA-primed synthesis of DNA and RNA use dNTPs when transcribing an RNA template [poly(rC), In the assays described so far we have examined de Table IV]. Overall the activity of the wild-type and Y639F novo initiated synthesis. T7 RNAP can also extend RNA 4615 R.Sousa and R.Padilla

some Table IV. Wild-type and Y639F activity on an RNA [poly(rC)] 400- to 600-fold better than rCTP and 200- to 400- template fold better than rATP. Because of their lower activity we can say only that the relative rate of incorporation of 0.5 mM GTP 1 mM dGTP 1 mM dGTP + rGTP on poly(dC) is 184-fold greater than the rate of 0.5 mM GMP incorporation of non-complementary rNTPs for Y639A Wild-type 1000 <0.5 <0.5 and 50-fold greater for Y639S. The use of Mn2> instead Y639F 505 (358-733) 62 (48-80) 116 (90-148) of Mg2' has been reported to increase miscoding for a number of polymerases (Niyogi and Feldman, 1981; Tabor Numbers give mean and range from three experiments. Template was and Richardson, 1989), so we examined the effects of at 0.2 mg/ml, polymerases at 10 M. Labeling NTPs were [a-32P]rGTP and [a-32P]dGTP. Mn2+ on miscoding by wild-type and Y639F polymerases. In Mn2> buffer the G640A, Y639A and Y639S mutants were insufficiently active to allow accurate measures of A B C D E F G H I J Prinrr DNA RNA miscoding. With the wild-type and Y639F polymerases NrPs: - 4 dNTPs 4 rNTPs T 4 dNTPs 4 rNTPs the use of Mn2+ increases miscoding by 20- to 40-fold. Pol: - WT IWTmumu WTmu WTmu However, the apparent rate of miscoding by Y639F remains similar to the wild-type enzyme. On poly(dT) high levels of activity allowing an accurate measure of miscoding frequencies could only be observed for Y639F and the wild-type polymerase in both Mg2+ and Mn2+ buffers. In Mg2+ buffer apparent miscoding ..: rates on poly(dT) were higher than on poly(dC), but were similar for Y639F and the wild-type enzyme. In Mn2+ buffer miscoding rates were increased by -5-fold, on average, but again rates were similar for wild-type enzyme and Y639F. However, on poly(dT) Y639F did show a reproducible -2-fold increase, relative to the wild-type enzyme, in the ratio of the rates of rGTP to rATP incorporation Hompolymer assays have been used previously to measure miscoding by RNAPs (Glazer, 1978; Niyogi and Feldman, 1981; Blank et al., 1986), but it should be Fig. 5. Primed synthesis of DNA and RNA with Y639F and the wild- remarked that they can produce only upper bounds for type polymerase. Transcription reactions contained end-labeled 12 base miscoding frequencies. Measured miscoding rates could DNA or RNA primers of identical sequence (GGACACGGCGAA) reflect contamination of the hybridized to a DNA template (CCCGGGATGGAATGGAGTATT- hompolymeric templates. It CGCCGTGTCCATGGCTGTAAGTATCC). Primer/template is also possible that the transcripts themselves could serve concentration was 105 M. Reactions contained the indicated as templates and support incorporation of rNMPs non- polymerases (1077 M) and NTPs. NTP concentrations and complementary to the original template [i.e. the poly(rG) electrophoresis as in Figure 1. or poly(rA) transcripts made on poly(dC) or poly(dT) could subsequently support synthesis of poly(rC) or poly(rU) primers. We therefore examined the abilities of both transcripts]. Such caveats are less relevant to the miscoding polymerases to carry out DNA- or RNA-primed synthesis observed in Mn2+ buffer, since the change in divalent of DNA and RNA (Figure 5). We found that both the cation increases miscoding but cannot affect template wild-type and Y639F polymerases can extend DNA and composition. However, in Mg2+ buffer we should consider RNA primers with rNTPs, but extension of DNA primers the measured miscoding frequencies to be upper bounds was 2- to 3-fold less efficient than extension of RNA for the true rates of miscoding. Nevertheless, the results primers. Y639F also extended DNA and RNA primers presented in Table V indicate that Y639F does not exhibit with dNTPs, but -4-fold less efficiently than with rNTPs. a gross increase in miscoding, which would manifest itself as a clear increase in the incorporation of non- The Y639F mutant does not exhibit greatly complementary rNMPs on homopolymeric templates. increased miscoding Tyr639 in T7 RNAP corresponds to Y766 in DNAP I and The increased utilization of dNTPs by Y639F is due mutations in Y766 have been shown to cause miscoding. to both a decreased Km and an increased kct for We therefore examined the miscoding properties of the dNTPs wild-type and mutant T7 RNAPs by measuring the relative The Km and kcat of the wild-type and Y639F polymerases incorporation of labeled rGTP, rUTP, rATP and rCTP on with rATP, rITP and five different dNTPs were measured poly(dC) or poly(dT) templates in the presence of excess (Table VI). The Km of the wild-type enzyme for dNTPs unlabeled rGTP or rATP respectively (Table V). An was much higher than previously reported values for the increase in miscoding would be reflected in an increase corresponding rNTPs and varied considerably for different in the rate of incorporation of the non-complementary dNTPs (Ikeda and Richardson, 1987; Patra et al., 1992). NTP into RNA. Notably, the wild-type enzyme Km values correlate with On poly(dC) the wild-type, Y639F and G640A poly- the rNTP/dNTP selectivity values presented in Table I. merases incorporate rGTP into RNA at >1300- to 2000- The selectivity of the wild-type enzyme with ribo- versus fold the rate of rUTP incorporation. rGTP is incorporated deoxyribonucleotides was greatest for CTP, followed by 4616 An RNA polymerase as a DNA polymerase

Table V. Relative rates of incorporation of complementary and non-complementary rNTPs on homopolymeric templates by wild-type and mutant polymerases Wild-type Y639F G640A Y639A Y639S Template poly(dC) GTP/UTP >2000 (Mg2+)a >1760 >1320 >184 >50 53 (Mn2+)a 55 n.d.b n.d. n.d. GTP/CTP 400 550 508 > 184 >50 32 40 n.d. n.d. n.d. GTP/ATP 233 338 388 > 184 >50 9.3 8 n.d. n.d n.d. Template poly(dT) ATP/GTP 170 94 n.d. n.d. n.d. 50 27 n.d. n.d. n.d. ATP/UTP 121 94 n.d. n.d. n.d. 21 20 n.d. n.d. n.d. ATP/CTP >340 >94 n.d. n.d. n.d. 77 60 n.d. n.d. n.d. Numbers are averages from two experiments and reflect the ratio of the percentages of labeled rNTPs incorporated into RNA in reactions in which unlabeled complementary rNTPs were in great excess (0.5 mM) to labeled complementary or non-complementary rNTPs. Templates were at 0.1 mg/ml. Polymerases were at 104 M in Mg2+ buffer and le-7 M in Mn2+ buffer. aThe upper number refers to results obtained in Mg2+ buffer, the lower number to results in Mn2+ buffer. bn.d., not determined. G640A, Y639A, Y639S were poorly active in Mn2+ buffer or on poly(dT) under all conditions.

Table VI. Kinetic constants for the Y639F and wild-type polymerase with rNTPs or dNTPs rATP rITP dTfP dUTP dCTP dGTP dATP Y639F Km 0.063-0.125 0.034-0.094 0.038-0.059 0.052-0.092 0.92-1.6 0.185-0.264 0.20-0.35 kcat (/s) 150-210 180-200 70-110 70-130 50-90 30-60 50-70 Wild-type Km 0.034-0.068 0.029-0.059 0.209-0.262 4.4-9.0 4.3-13.5 0.602-0.701 2.0-5.0 kcat (/s) 190-220 170-230 26-29 25-39 9-14 5-9 6-9 Numbers give ranges from three experiments. The template was supercoiled pT75 at 10-7 M and polymerases were at 10 8 M.

ATP, and was least for UTP, implying that an important to a 'four rNTP' reaction) were obtained for reactions component of the selectivity of the wild-type enzyme for containing three rNTPS and one dNTP: dATP, -3-fold; rNTPs over dNTPs is a much higher Km for dNTPs. For dUTP, -2-fold; dGTP -1.5 fold; dCTP, 1- to 1.5-fold. Y639F the Km values for these dNTPs are from -3- to Because of its poor activity in reactions with dNTPs we -11-fold less, but the rank order of these Km values (dCTP could not determine the corresponding elongation rate Km > dATP Km > dGTP Km > dTTP Km) is the same as reductions for the wild-type enzyme. for the wild-type enzyme. For Y639F kcat values for reactions with different dNTPs were only 2- to 4-fold less Discussion than for dNTPs, while the wild-type enzyme displayed kcat values with dNTPs that were from -6- to -30-fold Our results reveal that mutations of Tyr639 in T7 RNAP less than for dNTPs. reduce the ability of the polymerase to discriminate between rNTPs and dNTPs. A conservative mutation Elongation rates of Y639F in reactions containing a which removes the tyrosine hydroxyl but retains the single dNTP phenolic ring (Y639F) exhibits wild-type activity, but an Elongation rates for Y639F in reactions with four rNTPs average reduction of -20-fold in the selectivity for dNTPs or one dNTP and three rNTPs were determined by over rNTPs (Table I). Non-conservative mutations of this analyzing aliquots, taken at 10 s intervals, from transcrip- tyrosine (Y639A/S) also display decreased rNTP/dNTP tion reactions initiated on supercoiled PT75 by adding discrimination (Table III), but are less active than the NTPs to otherwise complete reaction mixtures (Figure 6). wild-type enzyme. Replacement of an rNTP by a dNTP When analyzed on denaturing agarose gels (Figure 6) the typically reduces Y639F transcript elongation rates by heterogeneously sized transcripts from these reactions are only a factor of two. Tyr639 is conserved in a large resolved as a smear, with the trailing edge of the smear number of DNA-directed RNA and DNA polymerases corresponding to transcripts initiated at t = 0 and from (Delarue et al., 1990). In DNAP I the corresponding which the maximal transcript elongation rate can be tyrosine is Y766 and mutations of this residue to serine determined (Golomb and Chamberlin, 1974; Bonner et al., and phenylalanine have been characterized (Polesky et al., 1994). The following elongation rate reductions (relative 1989, Carrol et al., 1991). The Y766S mutation was alone 4617 R.Sousa and R.Padilla

A B C D El T7 RNAP mutants are so poorly active we draw no conclusions their NTPS- 4rNTP9 3rNTPs 3rNTPs 3rNTPs 3rNTPs regarding miscoding properties except dUTP dCTP dGTP .TPj NTPs [dATP to set upper bounds for their miscoding frequencies that are 5- to 40-fold greater than those of the wild-type or Y639F polymerases. We speculate that this conserved tyrosine may play a 4.8 kb 4.3 kb general role in many polymerases in sensing inappropriate 3.7 kb geometry or structure in the template-nucleotide-primer/ nascent RNA complex within the active site. It may 2.3kb be in both mismatched bases and 1.9 kb . important rejecting nucleotides with inappropriate structure in the sugar 1.4 kb 1.3 kb moiety. Mutation to phenylalanine may reduce discrimina- tion between ribo- and deoxyribonucleotides without affecting miscoding or other kinetic parameters. Non- conservative mutations may decrease substrate discrimina- tion and increase miscoding and can have marked effects Fig. 6. Relative elongation rates of Y639F in four rNTP and three on other kinetic The rNTP + one dNTP reactions. The template was supercoiled pT75 at parameters. miscoding and reduced 5X I0e M. Y639F polymerase was used at a concentration of I0 M. substrate discrimination properties of these mutants are Reactions contained the indicated NTPs. Labeling was with manifest in vivo as an increased mutator phenotype for [a-32P]rATP (lane E) or [a-32P]rGTP (other lanes). After initiation of DNAP I Y766S and a decreased yield of protein per the reactions aliquots were taken at 10 s intervals and analyzed on transcript with T7 RNAP Y639F. The function of this denaturing 1% agarose-formaldehyde gels. The figure shows the 20 s in time point. The bars indicate the positions of X DNA markers. tyrosine insuring incorporation of the 'correct' nucleo- tide is specialized, since characterization of a large number of active site mutants in DNAP I and T7 RNAP has amongst a number of active site mutations characterized so far revealed a misincorporation phenotype only for in that it decreased DNAP I fidelity (increased miscoding). mutations of this tyrosine. Certainly, alternative interpreta- The Y766F mutation displayed wild-type fidelity and tions for the current observations are plausible. For activity. Similarly, the T7 RNAP Y639F mutant displays example, Y766 mutations may increase miscoding in wild-type kinetics (Bonner et al., 1992, 1994; Osumi- DNAP I but may have no effect on dNTP/rNTP discrimina- Davis et al., 1994) and the only effect we can identify for tion and in other polymerases this conserved tyrosine this mutation in T7 RNAP is the reduced substrate may play no role in substrate discrimination or fidelity. discrimination reported here. We speculate that a pheno- Examination of the substrate discrimination properties of type for the DNAP I Y766F mutant might also be DNAP I Y766 mutants and analysis of the structure of revealed if the rNTP/dNTP selectivity of this mutant were transcripts synthesized by Y639F in vivo are necessary to examined. distinguish between alternative interpretations for the While T7 RNAP Y639F showed decreased dNTP/rNTP observations discussed here. selectivity, it did not exhibit increased miscoding, as The mechanism by which this tyrosine contributes to assessed by incorporation of non-complementary NTPs fidelity and/or substrate discrimination is unclear. Relative on hompolymeric templates. In examining the miscoding to the wild-type enzyme the Y639F mutant displays properties of Y639F we sought to determine if the reported both a decreased Km and an increased kcat for dNTP defect in the translatability of transcripts made by Y639F incorporation. Replacing Mg2+ with Mn2+ increases mis- could be due to miscoding by this polymerase. Makarova coding and decreases rNTP/dNTP selectivity for both et al. (1995) found that Y639F was unique amongst a polymerases. However, this information in and of itself is number of T7 RNAP active site mutants tested in that its not particularly revealing of mechanism. Benkovic, transcripts yielded 2- to 3-fold less f-galactosidase protein Johnson and co-workers suggest a mechanism for DNAP per transcript than the wild-type enzyme. If this defect I involving a rate limiting conformational change in the were due to miscoding we would expect an in vivo active site which leads to close contact between amino miscoding frequency of several bases per kb, resulting in acid side chains and the template-dNTP-primer complex the introduction of at least one nonsense codon into (Kuchta et al., 1987; Mizrahi et al., 1985). This isomeriza- the coding sequence of 50-70% of the f-galactosidase tion step can cause steric clashes between protein side messages (most missense mutations would not be expected chains and template-dNTP-primer complexes with 'in- to destroy the P-galactosidase protein). Such a high appropriate' structures, thus limiting the rate at which frequency of miscoding should have been detectable in mispaired dNTPs are utilized. The large, flat ring of the our assays. The fact that it was not implies that some tyrosine may be especially critical in generating these other aspect of the Y639F transcript structure (perhaps geometry-sensing clashes (Carrol et al., 1991). Such a the incorporation of a small fraction of dNMPs in vivo) model seems broad and flexible enough to account for must lead to poor translatability. The fact that Y639F does detection of inappropriate geometries/structures in the not exhibit a gross increase in miscoding is consistent template-NTP-primer/nascent RNA due to inappropriate with the observation that the DNAP Y766F mutation also sugar moiety structure as well as mispairing. In such a does not increase miscoding. It is possible that the Y639A model neither the phenolic ring nor the tyrosine hydroxyl and Y639S mutations in T7 RNAP increase miscoding, need directly contact any part of the substrate NTP. In since the Y766S mutation in DNAP I does so (Polesky fact, Tyr766/639 in DNAP IT7 RNAP is expected to et al., 1989; Carrol et al., 1991), but because these two contact the template strand, not the substrate NTP (Sousa 4618 An RNA polymerase as a DNA polymerase et al., 1993; Steitz et al., 1994; Astatke et al., 1995). or rNMPs. In a primer extension assay we find that both However, DNAP I mutations within the region which the wild-type and Y639F polymerases extend a DNA contacts the template have been found to increase dNTP primer more poorly than an RNA primer, even when the Km (Polesky et al., 1989) and mutations within the template reactions contain only rNTPs. The clearest demonstration binding region of HIV I RT affect utilization of dNTP of the role of transcript structure is revealed by reactions analogs (Tantillo et al., 1994). These results imply that containing dGTP and three rNTPs with a linear, double- changes in the template-protein interface can affect the stranded promoter template that initiates with three Gs. interaction with the NTP. A clear demonstration of this is Y639F will initiate and synthesize the initial dGdGdG the observation by Ricchetti and Buc (1993) that kcat is trimer on such a template, but extends this trimer with similar but dNTP Km is -100-fold greater when DNAP I rNMPs very poorly. In fact, extension by Y639F of an uses an RNA template instead of a DNA template. It is rGrGrG trimer with dATP is more efficient than extension therefore possible that detection of inappropriate geometry of a dGdGdG trimer with rATP. It is unclear over what or structure in the template-NTP-primer/RNA complex length of the transcript sensitivity to replacement of rNMPs due to inappropriate NTP structure occurs indirectly, with dNMPs exists. Certainly, we would expect that the through contacts between the tyrosine and the template polymerase could be sensitive to the structure of the most strand. The relevant aspect of the Y639F mutation may 3' base in the transcript and this is supported by the not be chemical (deletion of hydrogen bonding capacity), observation that adding rGMP to reactions containing but structural (a change in the position of the phenolic dGTP enhances the activity of Y639F 2- to 3-fold. The ring). A role for the tyrosine in excluding water from the presence of dNMPs in the 3 or 4 bases nearest the 3'-end active site may also be relevant. of the transcript could also influence extension. This could While the mechanisms through which the tyrosine occur if the presence of dNMPs in the transcript adversely ensures selectivity in substrate selection are still unclear, affects the geometry of the template-transcript hybrid or the outlined considerations suggest that a change in causes the transcript to be released from the ternary template structure could also have effects on substrate complex more easily. The latter effect is probable, since discrimination. We therefore examined the ability of the DNA-DNA hybrids are less stable than RNA-DNA wild-type and Y639F polymerases to synthesize DNA or hybrids. It may be that during abortive transcription, when RNA on poly(rC). We found that the substrate discrimina- the transcript is small and easily released from the ternary tion properties of these polymerases on this RNA template complex, changes in transcript structure might have larger reflected their substrate discrimination properties on DNA effects on the geometry and stability of the template- templates. Both polymerases could synthesize RNA on transcript complex (most importantly the position of the 3' this RNA template, but only Y639F could synthesize hydroxyl) than they have during processive transcription, DNA on this template, thus presenting a novel polymerase when the transcript makes a more extensive set of contacts activity-unprimed reverse transcription. with the template and/or polymerase which can help hold One question we sought to address in this study was the transcript in the proper position for catalysis (Martin why substitution of multiple rNTPs with dNTPs causes et al., 1988). more than additive reductions in the activity of Y639F. It is therefore the barriers posed by the initial, abortive Wild-type T7 RNAP and the Tyr639 mutants differ not phase of transcription which most limit the activity of in kind, but only in the degree to which they will Y639F when transcribing with multiple dNTPs. In order incorporate dNMPs. In Mg2+ buffer the Y639 mutants to understand more about what template sequences and retain some preference for dNTPs. Partly as a consequence reaction conditions could help overcome these barriers we of this, replacement of a single rNTP with a dNTP leads tested promoters which differed in their initially tran- to modest activity reductions in reactions with Y639F. scribed sequences, as well as different reaction conditions. However, replacement of two, three or four rNTPs with Of five promoter sequences tested the natural 010 promoter dNTPs progressively and exponentially decreases the was best and of the others those with two Gs in the initial activity of Y639F. Analysis of transcription products region were better than those with just one. A promoter synthesized and activities with different combinations of which initiated with 11 consecutive Gs was not superior to two, three or four dNTPs reveals that these large activity the 010 promoter in reactions with four dNTPs. Promoters decreases arise because of increased termination during which were single-stranded in the transcribed region the early, abortive phase of transcription. The modest worked better than those which were fully double-stranded. preference of Y639F for dNTPs can contribute to this The addition of rGMP and the use of a supercoiled increased termination by modestly reducing elongation template to reactions containing dGTP or four dNTPs also rates during this initial, poorly processive phase of tran- increased Y639F activity. Replacement of chloride with scription. Because processivity during this initial phase of glutamate in the reaction did not enhance activity (not transcription is low to begin with, even modest decreases shown), while replacement of Mg2+ with Mn+ decreased in elongation rates can greatly decrease the fraction of substrate discrimination, as well as activity. Under all initiated transcripts which are extended to a length which conditions tested Y639F remained poorly active with four allows the polymerase to escape into the more productive, dNTPs, though our examination of promoter sequence processive phase of transcription (Bonner et al., 1992). In variants was limited. It is possible that a more exhaustive addition, examination of initial transcription products search of promoter variants differing in their initially reveals that transcript structure, as well as substrate transcribed regions could identify a sequence which would structure, contributes to the efficiency of transcript exten- significantly enhance Y639F activity in a 'four dNTP' sion. Figure 1 reveals that transcripts containing dNMPs reaction. are subsequently more poorly extended with either dNMPs When using Y639F as a reagent to synthesize nucleic 4619 R.Sousa and R.Padilla acids of mixed dNMP/rNMP composition the following Synthetic were prepared at the UTHSCSA DNA synthesis facility lessons may be gleaned from our observations. Unless the on an Applied Biosystems DNA synthesizer and purified by HPLC. A synthetic RNA 12mer was from the Midlands Certified Reagent Com- initiating rNTP (rGTP) is replaced, Y639F will remain pany. pT75 (Tabor and Richardson, 1985) and pBS (Strategene sufficiently active to readily allow synthesis of specific Inc.) were purified from Ecoli by alkaline lysis and cesium chloride products from T7 promoters which do not differ greatly gradient centrifugation (Sambrook et al., 1989). Radioactive nucleotides from the optimal, consensus sequence in the initially were from NEN Dupont or ICN. transcribed region. Because T7 RNAP is naturally highly active and processive, even Preparation and purification of mutant polymerases the reductions observed when Construction, expression and purification of the T7 RNAP mutants has Y639F transcribes with three dNTPs do not lower activity been described previously (Bonner et al., 1992). beyond a range feasible for in vitro synthesis. When the initiating rGTP is replaced with dGTP the addition of Transcription reactions rGMP and the use of a supercoiled or partially single- Transcription reactions were carried out in 40 mM Tris-HCl, pH 8.0, stranded template with an optimal sequence in the 15 mM MgCl2 and 5 mM dithiothreitol (DTT) or 20 mM manganese initially citrate, pH 8.0, 5 mM DTT at 37°C. Template, polymerase and NTP transcribed region are important to obtain good activity. concentrations were as indicated in the legends to the figures and tables. When reactions contain four dNTPs the use of rGMP, a Relative activity determinations were made by taking 4 Rl aliquots of partially single-stranded promoter of optimal sequence reactions at 5, 10 and 20 min time points and spotting onto DE81 filter and very high (10- M) polymerase and template con- paper. Unincorporated nucleotides were separated from incorporated centrations nucleotides by washing the filter paper with 0.5 M KH2PO4, pH 7.0, are necessary to obtain significant activity. and retained radioactive nucleotide was quantitated with a Molecular Variations in the initially transcribed region ofthe promoter Dynamics phosphorimager. Radioactive NTPs used were as indicated in may affect the relative activity of Y639F in reactions with the figure and table legends. To evaluate rNTP/dNTP selectivity, reactions different combinations of dNTPs in not entirely predictable were run with four rNTPs and pT75 as template and [a-32P]rNTPs or ways, so it may be important to evaluate the appro- [cX-32P]dNTPs were used to label the transcripts. The relative rate of incorporation of an rNTP versus its cognate dNTP was determined from priateness of particular promoters with particular combina- the relative percentages of labeled rNTP versus dNTP incorporated into tions of dNTPs. DE81-retainable RNA at 5, 10 and 20 min time points. Apparent A major theme of research on nucleic acid polymerases miscoding frequencies were determined similarly, though in this instance over the past several years has been the discovery of the template was a single-stranded homopolymer and the reaction extensive structural similarity between the contained the complementary unlabeled rNTP and complementary majority of [a-32P]rNTP or one of the three non-complementary [a-32P]rNTPs. The these enzymes, even those from functionally different relative percentages of labeled complementary versus non-comple- classes (Kohlstaedt et al., 1992; Jacob-Molina et al., 1993; mentary rNTPs incorporated at 5, 10 and 20 min time points gave an Sousa et al., 1993; Pelletier et al., 1994; Sawaya et al., apparent miscoding rate. The rNTP/dNTP selectivity assay was used to 1994; Steitz et al., 1994). One part of this work has been test the following T7 RNAP mutants for effects on substrate discrimina- the identification of tion: D537S, D537E, S539A, RS5lS, D552S, R627S, K631S, L637A, well-conserved residues. Five amino Y639S, Y639F, Y639A, G640A, F644A, G645A, Q649S, !810S, H811S, acids have been identified as invariant in a large number H811A, D812A, D812E, D812S, D812N, D879E+DF882+DA883, of DNA-directed RNA polymerases (Delarue et al., 1990). D879E+A881T+DA883, D879E+F884+A885, D879E+F882Y, In T7 RNAP these are D537, K631, Y639, G640A and D879E+DA883, D879E+F882W and D879E. D812. A specific, conserved function has been revealed for the two invariant aspartates in Elongation rate determinations coordinating the catalytic These were carried out as described (Golomb and Chamberlin, 1974) Mg2+ ion (Jacob-Molina et al., 1993; Kohlstaedt et al., with some variations (Bonner et al., 1994). 1993; Sousa et al., 1993; Pelletier et al., 1994; Sawaya et al., 1994; Steitz et al., 1994). Our observations imply Determination of NTP Km and kcat a similarly specific and conserved function for Y639 as a Km and kcat were determined as described previously (Patra et al., 1992). sensor of inappropriate geometry or structure in the template-NTP-primer/RNA complex. Acknowledgements The Y639 T7 RNAP mutations present us with, in one sense, the functional unification of polymerases to go This work was supported by ACS grant NP-921 to R.S. along with their structural unification. The active site of wild-type T7 RNAP is forgiving with regard to template References structure (RNA or DNA) or mode of initiation (primed or Astatke,M., Grindley,N.D.F. and Joyce,C.M. (1995) J. Bio. Chem., 270, de novo). A mutation which relaxes the substrate selectivity 1945-1954. of this polymerase further expands the range of activities Blank,A., Gallant,J.A., Burgess,R.R. and Loeb,L.A. (1986) Biochemistry, which it can display in vitro. Depending on the substrates 25, 5920-5928. and templates presented to it, the Y639F T7 RNAP can Bonner,G., Patra,D., Lafer,E.M. and Sousa,R. (1992) EMBO J., 11, act as an RNA- or DNA-directed RNA or DNA polymerase 3767-3775. Bonner,G., Lafer,E.M. and Sousa,R (1994) J. Biol. Chem., 42, 25120- in primed or de novo initiated reactions. Thus it can 25128. display a variety of activities normally associated with Carroll,S.S., Cowart,M. and Benkovic,S.J. (1991) Biochemistry, 30, distinct polymerases, including some entirely novel 804-13. activities, such as de novo initiated reverse transcription Glazer,R.I. (1978) Nucleic Acids Res., 5, 2607-2616. or mixed dNMP/rNMP polymer synthesis. Golomb,M. and Chamberlin,M. (1974) J. Bio. Chem., 249, 2858-2863. Delarue,M., Poch,O., Tordo,N., Moras,D. and Argos,P. (1990) Protein Engng, 10, 461-467. Materials and methods Ikeda,R.A., Richardson,C.C. (1987) J. Biol. Chem., 262, 2800-3808. Jacobo-Molina,A. etal. (1993) Proc. NatlAcad. Sci. USA, 90,6320-6324. Nucleic acids and NTPs Kohlstaedt,L.A., Wang,J., Friedman,J.M., Rice,P.A. and Steitz,T.A. Nucleotides were from Pharmacia or USB/Amersham. Polynucleotides (1992) Science, 256, 1783-1790. were from Pharmacia and the Midland Certified Reagent Company. Konarska,M.M. and Sharp,R.A. (1989) Cell, 57, 423-431. 4620 An RNA polymerase as a DNA polymerase

Kuchta,R.D., Mizrahi,V., Benkovic,P.A., Johnson,K.A. and Benkovic,S.J. (1987) Biochemistry, 26, 8410-8417. Martin,C.T. and Coleman,J.E. (1987) Biochemistry, 26, 2690-2696. Martin,C.T. and Coleman,J.E. (1989) Biochemistry, 28, 2760-2762. Martin,C.T., Muller,D.K., Coleman,J.E. (1988) Biochemistry, 27, 3966-3974. McClure,W.R. and Chow,Y. (1980) Methods Enzymol., 64, 277-297. Mizrahi,V., Henrie,R.N., Marlier,J.F., Johnson,K.A. and Benkovic,S.J. (1985) Biochemistry, 24, 4010-4018. Moroney,S.E. and Picirrilli,J.A. (1991) Biochemistry, 30, 10343-10349. Niyogi,S.K. and Feldman,R.P. (1981) Nucleic Acids Res., 9, 2615-2627. Osumi-Davis,P.A., Sreerama,N., Volkin,D.B., Middaugh,C.R., Woody, R.W. and Woody,A.Y. (1994) J. Mol. Biol., 237, 5-19. Patra,D., Sousa,R. and Lafer,E.M. (1992) J. MoL Biol., 224, 307-318. Pelletier,H., Sawaya,M.R., Kumar,A., Wilson,S.H. and Kraut,J. (1994) Science, 264, 1891. Polesky,A.H., Steitz,T.A., Grindley,N.D. and Joyce,C.M. (1990) J. BioL Chem., 265, 14579-14591. Ricchetti,M. and Buc,H. (1993) EMBO J., 12, 387-396. Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning: A Laboratory Manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Sawaya,M.R., Pelletier,H., Kumar,A., Wilson,S.H. and Kraut,J. (1994) Science, 264, 1930. Schmidt,G. and Tannhauser,S.J. (1945) J. BioL Chem., 159, 83-89. Shi,Y., Gamper,H. and Hearst,J.E. (1988) J. BioL Chem., 263, 527-534 Sousa,R., Lafer,E.M. and Wang,B.-C. (1991) J. Crystal Growth, 110, 237-246 Sousa,R., Chung,Y.J., Rose,J.R. and Wang,B.C. (1993) Nature, 364, 593-599. Steitz,T.A., Smerdon,S.J., Jager,J. and Joyce,C.M. (1994) Science, 266, 2022-2025. Tantillo,C. et al. (1994) J. Mol. Biol., 243, 369-387. Tabor,S. and Richardson,C.C. (1985) Proc. Natl Acad. Sci. USA, 82, 1074-1078. Tabor,S. and Richardson,C.C. (1989) Proc. Natl Acad. Sci. USA, 82, 1074-1078. Received on May 12, 1995; revised on June 27, 1995

4621