Reverse Transcriptases from Bacterial Retrons Require Specific Secondary Structures at the 5’-End of the Template for the cDNA Priming Reaction*

Multicopy single-stranded DNA (msDNA) is a pecu- liar molecule consisting of a single-stranded DNA that is branched out from an internal G residue of an RNA molecule (msdRNA) via a 2’,5’-phosphodiester link-age. The genetic unit required for msDNA synthesis is designated “retron” and consists of msr (a gene for msdRNA), msd (a gene for msDNA), and a gene for reverse transcriptase (RT) in a single operon. To date, four different msDNAs have been isolated from Escherichia coli. They do not share any primary sequences in either RNA or DNA. To elucidate the specificity of bacterial RT for msDNA synthesis, the msr-msd region from retron-Ec67 was introduced into E. coli cells producing RT-Ec73, or the msr-msd region from re- tron-Ec73 into E. coli cells producing RT-Ec67. In both cases, msDNA was not synthesized. However, when the msdRNA coding regions (rnsr) for retron-Ec67 and -Ec73 were mutually exchanged and the chimeric genes were introduced into E. coli cells producing either RT-Ec67 or RT-Ec73, it was thus found that msDNA was produced only when msr and RT were from the same retron. Requirement of the msr region for msDNA synthesis by RT was further investigated by mutations in the msr region for retron-Ec67. These analyses

The genetic unit required for msDNA synthesis is designated "retron" and is a 1.3-2.5-kilobase DNA fragment integrated in the chromosomal DNA and consists of a single operon ( Fig. 1) (see Inouye (1991, 1992a) for 2684 reviews). The operon is composed of the coding region for msdRNA (msr) and the coding region for the single-stranded DNA of msDNA (msd), followed by an open reading frame for RT. In the primary transcript from the operon, there are inverted repeats, one in the msr region near the 5'-end of the primary transcript (a2 in Fig. 1) and the other downstream of the msd region (a1 in Fig. 1). As a result, these inverted repeats allow the transcript to form a stable stem structure placing the branched G residue within the msr region at an end of the stem, The priming reaction of msDNA is considered t o be initiated from the 2"OH group of this G residue, followed by cDNA synthesis using the same RNA transcript as a template. The absolute requirement of the inverted repeats, a1 and a2, has been demonstrated by mutational analysis of these regions, indicating that the formation of the stem structure immediately upstream of the branched G residue is essential for msDNA synthesis (Hsu et al., 1989). In addition, the first base used for the priming reaction has been shown to be variable and to be determined by the complementarity of the base on the template strand corresponding to the first base of msDNA (Hsu et al., 1989. In this study, we found that RT from retron-Ec67 cannot complement RT from retron-Ec73 for the production of msDNA-Ec73 or vice versa. Subsequently, by exchanging the rnsr regions downstream of the branched G residue between the two retrons, we found that msDNA is produced only when the msr region and RT are derived from the same retron. From these results together with further mutational analysis in the msr region, we propose that individual bacterial RTs require their own specific secondary structures located at the 5'-end of the template RNA molecule for cDNA synthesis and for the 2'-OH priming reaction. This presents an interesting similarity to retroviral RTs, which require specific tRNA molecules for the cDNA priming reaction for individual retroviral DNA synthesis (Varmus and Swanstron, 1985).

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-E. coli CL83 (Lerner and Inouye, 1990), which is an recA derivative of JM83, was used for all experiments in this study. Since this strain is a K12 derivative, it does not have a retron for msDNA production (Lampson et ai,, 1989). CL83 cells harboring plasmids were grown in L-broth (Miller, 1972) supplemented with appropriate antibiotics. pSP65 (Boehringer Mannheim) was used to clone the 0.6-kilobase BamHI-Hind111 fragment containing the msr-msd region of retron-Ec67, and the 0.7-kilobase HincII fragment for the msr-msd region of retron-Ec73 was cloned into pUC9 (Vieira and Messing, 1982). PUCK19 (Ohshima et al., 1992) and pBR322 were used for cloning polymerase chain reaction (PCR) fragments of the msr-msd region. For complementation analysis, plasmids carrying only RT genes were used. The 1059-bp X6aI fragment, which contains the RT-Ec73 gene, was cloned from retron-Ec73 (Sun et al., 1991) into the unique XbaI site of pGB21ppp5 , a pSClOl derivative harboring the spectinomycin-resistant gene. One of the resulting plasmids carrying the RT gene in the same orientation as the lpp-lac promoter was designated pRT-73. pRT-67  contains the RT-Ec67 gene in ~GB2lppP.~. pC1-1EP5 containing retron-Ec67 (Lampson et al., 1989) and p73-Hc0.7 containing the 0.8-kilobase HincII fragment containing the msr-msd region of retron-Ec73 (Sun et al., 1991) were used for PCR to construct chimeric msr-msd genes.

~" G G U T C G C J~C C C T A -~'
S " t~m u o r c w u w -3~

5"tCPgaTCCTTATGCACCTT-3'
FIG. 2. Schematic diagram for synthesis of chimeric genes with msr67-msd73 and msr73-msd67 by PCR. A, msr and msd and their orientations are shown by wide open and filled arrows on thin lines, which represent retron-Ec67, and/or -Ec73, respectively. The branched G residue is circled. Small arrows with numbers and letters are primers used for PCR and are positioned at the locations having DNA sequences to complement primer sequences. Wavy and thick lines attached at the ends of 67c, 67d, 73c, and 73d indicate 15base sequences derived from other retrons (see text for details). B, nucleotide sequences of primers used for PCR. Boldface and lightface upper-case letters represent nucleotides of retron-Ec67 and -Ec73, respectively. Lower-case letters represent nucleotides added to create restriction sites, which are underlined. and 73e and 730 were also synthesized as shown in Fig. 2 ( A and B). These sets were used to synthesize three PCR fragments (73AB, 73CD, and 73EF) using p73-Hc0.7 as a template.
Fragments 67AB and 73AB contain one of the inverted repeat sequences (a2) from retron-Ec67 and -Ec73, respectively. The a2 sequence corresponds to the region immediately upstream of the branched G residue. Fragments 67AB and 73AB also contain the promoters for retron-Ec67 and -Ec73, respectively. Fragments 67EF and 73EF include the msDNA coding region (msd) and the other inverted repeat sequence (al) from retron-Ec67 and -Ec73, respectively. Fragments 67CD and 73CD contain part of the msr regions (the sequences downstream of the branched G residue) from retron-Ec67 and -Ec73, respectively, that are flanked with the 15-hp retron-Ec73 and -Ec67 sequences, respectively. These 15-bp sequences are derived from immediately upstream and downstream of the part of the msd regions used (see Fig. 2 A ) . From both retron-Ec67 and -Ec73 DNAs, three PCR fragments (see Fig. 2 A ) were individually synthesized by PCR and purified by polyacrylamide gel electrophoresis.
These msr-msd fragments in PUCK19 were under the control of the lac promoter from PUCK19 and also their own promoters. To avoid the transcription from the lac promoter, the chimeric msr-msd regions were cleaved from the clones by EcoRI and RamHI digestion, and the resulting fragments were recloned into pBR322. In the resultant plasmids, the msr-msd regions were able to be transcribed only from their own promoters. These plasmids were designated pBR(67-67) for msr67-msd67, pBR(73-73) for rnsr73-mqd73, pBR(67-73) for msr67-msd73, and pBR(73-67) for mvr73-m.vd67.
For PCR, 0.1 pg of template DNA (pC1-1EP5 or p73-Hc0.7) was added to 100 pl of PCR mixture containing 50 mM Tris-HCI (pH 9.0), 1.5 mM MgCI,, 20 mM ammonium sulfate, 0.2 mM each dNTP, and 50 pmol of each oligonucleotide primer (Fig. 2R). After heat denaturation of the reaction mixture prepared as described above (100 "C, 3 min) followed by quick chilling on ice, 3 units of Hot TubTM DNA polymerase (Amersham Corp.) were added to the reaction mixture. PCR was performed for 25 cycles in a programmable thermal controller (Md Research, Inc.) under the following conditions: 94 "C, 1 min; 55 "C, 2 min; and 72 "C, 2 min. The resulting PCR fragments were purified by polyacrylamide gel electrophoresis and cloned into the SmaI site of pUCK19. The cloned PCR fragments were sequenced by the dideoxy chain termination method (Sanger et al., 1977) with the modified T7 DNA polymerase (Sequenase, United States Biochemical Corp.).
Preparation of RNA and Detection of msDNA by R T Extension-Total RNA fraction was prepared from exponentially growing cells by the method of Chomzynski and Sacchi (1987). R T extension reaction was carried out using avian myeloblastosis virus R T (Molecular Genetic Resources) and [a-"PIdCTP as described previously (Lampson et al., 1989). The labeled samples were electrophoresed on a n 8 M urea, 4% polyacrylamide gel.
Labeling and Sequencing of msDNA-msDNA was prepared by the alkaline SDS lysis method used for plasmid DNA (Birnboim and Doly, 1979). For labeling and sequencing of msDNA, msDNA was prepared from a 300-ml overnight culture of CL83 cells harboring a R T gene and a fused msr-msd region. The msDNA fractions were treated with 80 pg/ml RNase A for 15 min a t 37 "C and subjected to 10% polyacrylamide gel electrophoresis. After staining the gel with ethidium bromide, the bands corresponding to msDNA were cut out, and msDNAs were electroeluted. About 0.3 pg of msDNA was labeled with "'P a t the 3'-end using [n-"'P]ddATP and terminal deoxynucleotidyltransferase (International Biotechnologies, Inc.) as follows. The msDNA was added to the labeling mixture (100 pl) containing 0.14 mM potassium cacodylate (pH 7.2). 30 mM Tris, 1 mM CoCI2, and 0.2 mM fi-mercaptoethanol. After heat denaturation (100 "C, 5 min) followed by quick chilling on ice, [n-"'PIddATP and 30 units of terminal deoxynucleotidyltransferase were added, and the final reaction mixture was incubated a t 37 "C for 60 min. The reaction was stopped by adding 25 mM EDTA and the same volume of isopropyl alchohol, and then the msDNA was precipitated by centrifugation. T h e concentrated samples were then applied to a preparative sequencing gel (6% polyacrylamide gel in 8 M urea). After electrophoresis, each band was cut out, and the labeled msDNAs were eluted from the gel with elution buffer containing 0.3 M sodium acetate (pH 7.0), 0.1% SDS, and 1 mM EDTA. After incubation a t 37 "C overnight, the eluted msDNAs were precipitated with ethanol. The DNA sequences of the msDNAs were determined by the method of Maxam and Gilbert (1980). Construction of Mutations in msr Region of Retron-Ec67-Mutagenesis in the msr67 region were constructed by PCR in two steps using synthetic oligonucleotides containing mutated sequences as primers by the method of Ho et al. (1989). In the first PCR, two fragments were amplified using pUCK(67-67) as a template. One was made with the 67a primer (Fig. 2R) and an oligonucleotide containing a mutated sequence, and the other with the 67f primer (Fig. 2R) and the oligonucleotide complementary to that used for mutagenesis. The second PCR was performed using the two PCR fragments isolated as described above as templates with 67a and 67f as primers. After digestion with EcoRI and RamHI, the final PCR fragments (mvr67-msd67) were cloned into pRR322. msDNAs were isolated from cells transformed with both pRT-67 and pBR322 carrying the mutant msr67-msd67 fragments. msDNA production was assayed by polyacrylamide gel electrophoresis as shown in Fig. 6.
Other Materials"T4 DNA ligase was purchased from Bethesda Research Laboratories. The Klenow fragment of E. coli DNA polymerase I was from Boehringer Mannheim. Restriction endonucleases were from Bethesda Research Laboratories, Boehringer Mannheim, or New England BioLabs, Inc. [a-"'PIdCTP, [a-"PIddATP, and [a-"S]dATP were from Amersham Corp.

RESULTS
RTs Are Not Exchangeable between Retron-Ec67 and -Ec73"In a retron, the msr-msd region and the downstream R T gene are under the control of a single promoter located upstream of the msr-msd region and produce msDNA ( Fig. 1) (Lampson et al., 1989). However, the RT gene from retron-Ec67 can be expressed under a separate promoter to complement the function of the msr-msd region of retron-Ec67 to synthesize msDNA-Ec67 . This was achieved by double transformation of cells with p67-BH0.6 for the msrmsd region of retron-Ec67 in a high-copy number plasmid (pSP65) and with pRT-67 for the RT gene from retron-Ec67 in a low-copy number plasmid (pGB21ppp') under the lpp-lac promoter . The production of msDNA by the cells harboring both p67-BH0.6 and pRT-67 was clearly demonstrated by the RT extension method as shown in Fig. 3 (lane 2). The pattern of the RT extension products (lane 2 ) 3. msDNA production from cells harboring various combinations of msr-msd and RT gene from retron-Ec67 and -Ec73. Total RNA fraction was prepared from the cells harboring different retrons, and msDNA was detected by RT extension as described under "Experimental Procedures." The labeled samples were treated with RNase A and analyzed on a 8 M urea, 4% polyacrylamide gel. Lane 1, RT-extended msDNA from E. coli strain CL83 (Lerner and Inouye, 1990) harboring pCI-EP5 (Lampson et al., 1989) for msDNA-Ec67; lane 2, that from cells harboring both p67-BH0. was almost identical to that of the products from the cells harboring pC1-1EP5 that contained the entire retron-Ec67 (lane 1). The size of the major product was -107 bases in length (including an extra 4-base RNA attached at the 5'-end of DNA) (as shown by dot in both lunes 1 and 2).
Requirement of RT and msr Region from Same Retron-The results described above suggest that there is a specific requirement(s) within the structure of the msr-msd region for an individual RT to synthesize msDNA. RT may require specific primary and/or secondary structures within either msr or msd or both for msDNA synthesis. To examine this possibility, we constructed chimeric msr-msd genes between retron-Ec67 and -Ec73 using PCR, the msr region in retron-Ec67 was replaced with that of retron-Ec73 and vice versa as described under "Experimental Procedures." The resulting plasmids used for the experiments were pBR(67-73) for msr67-msd73 andpBR(73-67) for msr73-msd67 (see Fig. W).
pBR(67-67) for msr67-msd67 and pBR(73-73) for msr73-msd73 were also used as controls. Fig. 4 ( C and D ) shows the putative structures of the chimeric msDNAs expected from the structures of the chimeric msr-msd regions as well as the structures of msDNA-Ec67 (Fig. 4A) (Lampson et al., 1989) and msDNA-Ec73 ( Fig.   4B) (Sun et al., 1991). The putative secondary structures of the primary transcripts from the chimeric msr-msd regions are also shown in Fig. 5A for msr67-msd73 and in Fig. 5B for msr73-msd67 on the basis of the biosynthetic mechanism of msDNA ( Fig. 1) (see Dhundale et al. (1987); also see reviews by Inouye and Inouye (l991,1992a, 199213)). The synthesis of msDNA is initiated from the 2'-OH group of the G residue at the end of the al-a2 stem structure (circled in Fig. 5). For the first base (T residue) linked to the G residue by a 2',5'phosphodiester linkage, the A residue at position 126 for msr67-msd73 and the A residue at position 136 for msr73-msd67 are used as templates. msDNA is then elongated from left to right by RT using the RNA strand as templates.
The msDNA production by the plasmids constructed as described above were examined in E. coli cells producing either R T from retron-Ec67 (RT-Ec67) or from retron-Ec73 (RT-Ec73). For this purpose, E. coli CL83 (Lerner and Inouye, 1990) was first transformed with pBR(67-73), pBR(73-67), pBR(67-67) or pBR(73-73). The transformed cells were then retransformed with the second plasmids, pRT-67 for RT-Ec67 or pRT-73 for RT-Ec73. msDNAs were isolated from the doubly transformed cells by the alkaline SDS method, treated with RNase A, and separated on a 10% polyacrylamide gel. As shown in Fig. 6, msDNA production was observed only in the following combinations: msr67-msd67 and RT-Ec67 (lune 2), msr73-msd73 and RT-Ec73 (lane 8), msr67-msd73 and RT-Ec67 (lane 4 ) , and msr73-msd67 and RT-Ec73 (lanes 10 and 12). The results are summarized in Table I, and it is evident that msDNAs were synthesized only when msr and R T were from the same retron. It should be noted that msDNAs without msdRNA attached migrate abnormally on nondenaturing gels because of their extensive secondary structures as can be seen in msDNA-Ec67 (Fig. 6, lune 6) and msDNA-Ec73 ( l a n e 8). In particular, msDNA-Ec73 has an extensive secondary structure (Fig. 4 B ) (Sun et al., 1991), which is considered to cause faster migration on the gel. A minor band that migrated faster than the major band for msDNA-Ec73 (lune 8) was found to be processed between residues 6 and 7 of msDNA from its DNA sequence (data not shown).

FIG. 5. Proposed secondary structures of primary transcripts of msr-msd regions from pRR(67-73) and pBR(73-67).
A , structure of the msrG7-msd73 transcript from pRR(67-73); R, structure of the msr73-msd67 transcript from pRR(73-67). The branched G residue is circled. The inverted repeat sequences (a1 and a2) that form a stem structure in the primary RNA transcript  are indicated by the arrows. Arrowheads indicate positions where DNA synthesis terminates.
T h e dotted lines between the two gels indicate the same base. Lane N , the same samples without any sequencing reaction. The band in this lane migrated slower than the other lanes (G. G + A. C + 7 ' . and C) because of piperidine treatment during the sequencing reaction (Furuichi et al., 1987a). msDNAs were not detected for msr67-msd73 with RT-Ec73 and for msr73-msd67 with RT-Ec67 (data not shown). These results are consistent with the results in Fig. 6, demonstrating that msDNA is produced only when RT and the msr region are from the same retron.
msDNA-(67-73)Ll and -L2 were found to consist of 75 and 74 bases, respectively. The entire DNA sequence of msDNA-(67-73)L2 is shown in Fig. 8B and is composed of the 68-base sequence identical to that of residues 1-68 of msDNA-Ec73 (Fig. 4B) plus the 6-base sequence at the 3'-end identical to the 3'-end sequence of msDNA-Ec67 (Fig. 4A). The DNA sequence of msDNA-(67-73)Ll was longer by a C residue at the 3'-end than that of msDNA-(67-73)L2; otherwise, the DNA sequences of msDNA-(67-73)Ll and -L2 were identical (data not shown). Thus, the structure of msDNA-(67-73)Ll is identical to that predicted for msDNA-(67-73) (Fig. 4C). These results indicate that termination of msDNA for msDNA-(67-73) was less stringent than that for msDNA-Ec67; sometimes termination occurs earlier by 1 base or cannot occur at the regular site to extend the DNA strand almost all the way to the branched G residue. These termination sites are indicated by the arrowheads on the primary transcript for msr67-msd73 in Fig. 5A.
The DNA sequence of msDNA-(73-67)-band 3, the major band of the three bands in Fig. 7 (lune 3 ) , is shown in Fig.   8C. It consists of the 60-base sequence identical to the sequence from residues 1 to 60 of msDNA-Ec67 (Fig. 44) plus the 6-base sequence at the 3'-end (Fig. 8C). This 6-base sequence is identical to the &base sequence of the 3'-end of msDNA-Ec73 (Fig. 4B) plus an additional A residue at the 3'-end. This 3'-end A residue most likely resulted from the extension of msDNA synthesis by 1 base at the 3'-end, which adds an A residue (see Fig. 4, B and D). Thus, the structures of msDNA-(73-67)-band 2 and -band 1 are likely to have further extensions at the 3'-end (2 A residues and 3 A residues for band 2 and 3 DNAs, respectively). These termination sites are indicated by the arrowheads on the primary transcript for msr73-msd67 in Fig. 5B. Otherwise, the structure of msDNA-(73-67) is identical to that predicted in Fig. 40.
Requirement of Secondary Structure of msdRNA for msDNA Synthesis-The results described above demonstrated that the msr region is essential for msDNA synthesis. msdRNA coded by the msr region contains extensive inverted sequences that are believed to form stable secondary structures as shown in Fig.'4. Thus, we next examined the importance of the putative secondary structures of msdRNA for msDNA synthesis using msDNA-Ec67 (Fig. 4A). Using PCR, various mutations were incorporated in the putative secondary structure of msDNA-Ec67 as depicted in Fig. 9.
The 2-base substitution mutation (67r-1) that changed 2 U residues at positions 18 and 19 to 2 C residues did not alter the amount of msDNA produced (Table 11). This mutation is believed to tighten the stem structure by changing two U-G pairs to two C-G pairs (Fig. 9). The G to C substitution mutation at position 24 (67r-2) that destroyed the G-C base pair in the stem (see Fig. 9) resulted in a substantial reduction of msDNA production (Table 11). On the other hand, the resumption of the base pair in the same position of the stem by altering the C residue at position 38 to a G residue in the msr region containing mutation 67r-2 (67r-3) (see Fig. 9) restored the normal level of msDNA production (Table 11). This result indicates that msdRNA indeed forms the secondary structure proposed in Fig. 9.
We also constructed three deletion mutations: deletion of the U residue at position 25 (67r-4), deletion of 2 U residues at positions 25 and 26 (67r-5), and deletion of 2 U residues of 67r-5 together with 2 A residues at positions 36 and 37 (67r-6). The first two mutations cause 1 and 2 mismatched A residues in the stem for mutations 67r-4 and 67r-5, respectively (see Fig. 9), whereas mutation 67r-6 shortens the stem structure by 2 base pairs. None of these mutations were able t o produce msDNA (Table 11). Furthermore, the flipping of the entire inverted repeat sequences in the msr region (67r- Mutations made in the msdRNA portion of msDNA-Ec67 are shown by the arrows. The region mutated is boxed for each mutation. The structure of msDNA-Ec67 is from Lampson et al. (1989). The mutations were made in pBR(67-67) harboring the msr67-msd67 region by PCR as described under "Experimental Procedures." 7) also resulted in no msDNA production ( Fig. 9 and Table  11). The msdRNA with this mutation still maintains a stem structure of the same length as that of the wild-type msdRNA and of a similar stability as that of the wild-type msdRNA (Fig. 9).
The 67r-7 mutation flips not only the stem structure, but also the sequence of the loop from AAU to UAA. The conversion of this UAA sequence of the 67r-7 mutation back to AAU (67r-8) (see Fig. 9) did not restore msDNA production (Table  11). The RNA sequence of the loop was also found to be essential for msDNA synthesis; the AAU sequence of the wild-type loop to AGU (67r-9), UAA (67r-10), and CCA (67r-11) resulted in no msDNA production ( Fig. 9 and Table 11).
The opening of the stem structure on the top of the stem by altering the G-C base pair to the G-G mismatch (67r-12) also resulted in no msDNA production (Table 11). This mutation changed the 3-base loop of msdRNA-Ec67 to a 7-base loop by reducing the stem length by 2 base pairs. Taken together, the mutational analysis of the secondary structure reveals that there is a very stringent requirement of the secondary structure in msdRNA for msDNA synthesis.

DISCUSSION
This study demonstrates that the specificity of msDNA synthesis by bacterial RTs is determined by the structure of the msr region of a retron; between two retrons, the msd regions encoding the single-stranded DNA are exchangeable, whereas the msr regions encoding msdRNAs are not. TO initiate msDNA synthesis from the 2'-OH group of an G residue located at the 5'-end region of a template RNA, a stem structure has to be formed between two inverted repeat sequences within a template, one immediately upstream of the cDNA initiation site (the branched G residue) at the 5'end region of a template (a2 sequence) (see Fig. 5 , A and B ) and the other at the 3'-end region of the same template (a1 sequence). It has been demonstrated that formation of the stem structure is essential, although no requirement for specific sequences immediately upstream of the branched G residue in the stem structure is needed (Hsu et al., 1989). This -is also confirmed by this study since the stem structure from retron-Ec73 can be used for msDNA synthesis by RT-Ec67. For the initiation of retroviral cDNA synthesis, a specific tRNA is required for the priming reaction. It appears that a retroviral RT generally recognizes a specific tRNA to unfold it, allowing the tRNA to hybridize to the primer-binding site (Varmus and Swanstron, 1985) that is formed in the viral RNA and is complementary to the 3'-end of the primer tRNA.
It has been suggested that the interaction between template RNA forming secondary structures and primer tRNA is important for the initiation of reverse transcription in all type C and D retroviruses (Aiyar et al., 1992). The cDNA priming reaction is initiated from the A residue of the 3'-end CCA sequence of the tRNA. In the msDNA priming reaction, one can consider that the a1 sequence corresponds to the retroviral primer-binding site sequence and the a2 sequence to the 3'end sequence of the primer tRNA. However, unlike the tRNA primer, the mismatched G residue immediately downstream of the a2 sequence is used for the msDNA priming reaction (see Fig. 5, A and B ) . Furthermore, in contrast to the retroviral tRNA primer, there is no requirement for a secondary structure upstream of the a2 primer sequence. Instead, a specific RNA sequence (msdRNA) downstream of the a2 sequence was found to be required for msDNA synthesis. The mutational analysis of the msdRNA for msDNA-Ec67 revealed that the secondary structure formed by the msdRNA is essential for the production of msDNA-Ec67. The secondary structure formed in msdRNA-Ec67 consists of a 12-bp stem structure with a 3-base loop. We found that there were hardly any mutations in the secondary structure that did not affect msDNA production, indicating that RT-Ec67 stringently requires not only the 12-bp stem structure of the unique sequence, but also the AAU loop sequence for the synthesis of msDNA-Ec67. This stringent requirement for the secondary structure is most likely to be necessary for the msDNA priming reaction from the 2"OH of the G residue at the 3'-end of the a2 sequence (see Fig. 5A). Note that like retroviral RTs, RT-Ec67 is able to elongate a DNA chain using a nonspecific RNA or DNA template with the use of a nonspecific RNA or DNA primer in a cell-free system (Lampson et al., 1990).
Our finding that both bacterial and retroviral RTs strictly require a specific secondary structure for the cDNA priming reaction is intriguing particularly in light of the recent elucidation of the three-dimensional structure of HIV-1 RT (Ar-nold et al., 1992;Kohlstaedt et al., 1992). Since both RTs are evolutionarily related (Xiong and Eickbush, 1990), the threedimensional structures of bacterial RTs are considered to be also similar to that of HIV-1 RT, in which the anticodon and dihydrouridine stems and loops of tRNA are proposed to bind to the 51-kDa subunit (p51) of a heterodimer of p51 and p66 (the 66-kDa subunit). Interestingly, the polymerase domain of p66 forms a large cleft, whereas the four polymerase subdomains of p51 occupy completely different positions, resulting in no cleft in the p51 structure (Kohlstaedt et al., 1992). In HIV-1 RT, there is a sequence of -110 amino acid residues between the amino-terminal DNA polymerase domain and the carboxyl-terminal RNase H domain. This sequence, designated the connection domain, forms a subdomain that plays essential roles not only for the p51 conformation, but also for the p51 and p66 subunit interaction. Except for RT-Ec67, all the other known bacterial RTs lack the connection domain (see Inouye (1992a, 1992b) for reviews), which raises an interesting question as to whether bacterial RTs form a dimer and, if so, how the structure equivalent to p51 is formed without the connection domain.