Local DNA Sequence Control of Deletion Formation in Escherichia coli Plasmid pBR 322 Ujala DasGupta , ' Kathleen Weston - Hafer

The specificity of deletion formation was studied using tests involving reversion of palindromic insertion mutations. Insertions of a TnS-related transposon at 13 sites in the ampicillin-resistance (amp) gene of plasmid pBR322 were shortened to a nested set of perfect palindromes, 22, 32 and 90 bp long. We monitored frequencies of reversion to Amp', which is the result of deletion of the palindrome plus one copy of the flanking 9 bp direct repeats (which had been formed by transposition). Revertant frequencies were found to depend on the location and the sequence of the palindromic insert. Changing a 45-kb interrupted palindrome to a 22-bp perfect palindrome stimulated deletion formation by factors of from fourfold to 545-fold among the 13 sites, while elongation of the perfect palindrome from 22 to 90 bp stimulated deletion formation by factors of from eightto 18,000-fold. We conclude that deletion formation is strongly affected by subtle features of DNA sequence or conformation, both inside and outside the deleted segment, and that these effects may reflect specific interactions of DNA processing proteins with template DNAs. ELETIONS are frequent among both spontaD neous and induced mutations. Insights into mechanisms by which they arise have come from the mapping and sequencing of frameshift mutations and deletions, and from studies of transposon excision. In studies that preceded the advent of direct DNA sequencing it was deduced that frameshift mutations occur preferentially in repeated DNA sequences. It was proposed that they arise by replication errors in which misaligned pairing of complementary template and nascent DNA strand sequences result in the failure to copy, or the repeated copying of, one or a few bases of the template (STREISINGER et al. 1966; STREISINGER and OWEN 1985). Analyses of forward mutations in an F' episome indicated that deletions less than 200 bp long are formed preferentially, and that the endpoints of these and also larger (>700 bp) deletions usually coincide with direct repeats of 4 bp or more. These data were also explained on the basis of replication errors due to slipped mispairing of complementary sequences analogous to the frameshift mispairing model (ALBERTINI et al. 1982; FARABAUGH et al. 1978; GALAS 1978). The endpoints of deletions in other replicons including small multicopy plasmids also generally coincide with direct repeats (COLLINS, VOLCKAERT and NEVERS 1982; JONES, PRIMROSE and EHRLICH 1982; LOPEZ et al. 1984). Short, weakly Abbreviations: amp, tet and kan, genes encoding resistance to ampicillin, tetracycline and kanamycin, respectively; superscript r, resistance: superscript F, sensitivity; bp, base pair(s); kb, kilobase pair(s). ' Present address: Biophysics Division, Indian Institute of Chemical Biology, Calcutta, 700032, India. Genetics 115: 41-49 Uanuary, 1987) matched, inverted repeats were also found near certain deletion endpoints. This led to a suggestion that palindromes are also deletionogenic (ALBERTINI et al. 1982; GLICKMAN and RIPLEY 1984; SCHAAPER, DANFORTH and GLICKMAN 1986). This putative intrinsic instability of short palindromes is surprising in light of the importance of palindromes in regulatory sites and stable (transfer and ribosomal) RNAs (PLATT 1981; NOLLER 1984; CALENDAR and GOLD 1985). Other studies showed that sites nicked or cleaved by proteins such as DNA gyrase or the M13 gene 2 product are deletion hotspots (MARVO, KING and JASKUNAS 1983; IKEDA, KAWASAKI and GELLERT 1984; IKEDA 1986; MICHEL and EHRLICH 1986). It is likely therefore, that deletions arise by several different mechanisms, some involving DNA replication, and others DNA breakage. Useful information on deletion formation has also come from studies of the reversion of Tn5and TnlOinduced insertion mutations, in which reversion results from loss of the transposon plus one copy of the flanking 9-bp direct repeats (precise excision). This excision was not correlated with transposition to new sites, nor dependent on element-specific transposition functions, and thus was considered a type of spontaneous deletion event (EGNER and BERG 198 1 ; FOSTER et al. 1981; NAG et al. 1985; J . RIKKERS and D. E. BERG unpublished data). In other tests, shortening the approximately 1.5-kb long inverted repeats of T n 5 or TnlO or reversing them (to give direct repeats) markedly decreased reversion frequencies, thus indicating that long inverted repeats are deletionogenic (EGNER 42 U. DasGupta, K. Weston-Hafer and D. E. Berg and BERG 1981; FOSTER et al. 1981). The deletion of Tn5 and TnlO is more frequent in F' episomes than in the bacterial chromosome, and is due to F' transfer among cells in the population (BERG, EGNER and LOWE 1983; SYVANEN et al. 1986). Because F' transfer results in the formation of long single stranded regions, and earlier tests had shown Tn5 to be extremely unstable in single-stranded fd phage (HERRMANN et al. 1978), we concluded that single strandedness as well as palindromy promotes the formation of large deletions, perhaps by replication errors as diagrammed in Figure l A (EGNER and BERG 198 l). The simplest versions of models such as those in Figure 1 imply that direct and inverted repeats will be important determinants of deletion frequency, and of the location of deletion endpoints. However, the stabilities of Tn5 insertion mutations also seemed to be strongly dependent on the insertion site (EGNER and BERG 1981; BERG, EGNER and LOWE 1983; NAG et al. 1985). This indicated that other features of DNA sequence, less distinctive than direct and inverted repeats, contribute to the formation of large deletions. Tests based on the reversion of insertion mutations are analogous to the classical reversion analyses of point mutations (MARON and AMES 1983), and we have continued to use this approach to learn how the location and sequence of DNA segments affect their stabilities. In the present studies we focused on the formation of small deletions using palindromic inserts in the amp gene of plasmid pBR322. Our results indicate that frequency of deletion of a segment depends on its location and sequence, and is independent of recA function. MATERIALS AND METHODS General procedures: Bacterial growth, plasmid DNA extraction, restriction analysis and cloning, and DNA sequencing procedures have been described (BIRNBOIM and DOLY 1979; MANIATIS, FRITSCH and SAMBROOK 1982; MAXAM and GILBERT 1980; NAG et al. 1985). Bacterial strains and plasmids: AI1 bacterial strains used are derivatives of Escherichia coli K 12. Strain MC 106 1 (from M. CASADABAN) was the host for most plasmid constructions and reversion tests; GM1 19 (dam-) (ARRAJ and MARINUS 1983) was used to grow DNAs to be cleaved with the methylation-sensitive enzyme BclI. The isogenic strains AB1 157 (recilC) and JC2924 (recA56) (from A. J. CLARK via H. HUANG) were used to test the recA-independence of deletion events. The plasmids used were pBR322 (Amp' Tet') (MANIATIS, FRITSCH and SAMBROOK 1982) and the pBR322 derivative pGCl (Amp' but Tet') (MYERS, LERMAN and MANIATIS 1985). The sequence of the polylinker segment of ou r isolate of pGCl (from R. MYERS) had three differences from the published sequence: four instead of three contiguous Gs in the XhoI-Sal1 interval; five instead of four contiguous Cs in the BamHI-BglII interval; and three instead of two contiguous Gs in the Ncol-EcoRI interval. Plasmids with revertible insertion mutations: Mutations in the pBR322 amp gene were generated by NAG et al. (1 985) using the Tn5-related inverse transposon diagrammed in Figure 2. This transposon, like Tn5, generates 9-bp direct repeats of target sequences during insertion; Amp' revertants result from deletion of the entire insert plus one copy of the direct repeats of amp gene sequences. Tests of 34 insertion mutants revealed reversion frequencies ranging from IO-'" to IO-' (NAG et al. 1985). Thirteen mutants including six representative of high and low reverting insertions were chosen for tests of the effect of position and palindrome length on deletion frequency. The pBR322 coordinates and sequences at the sites of insertion in amp are given in Table 1. In vitro manipulation of insertion mutations. While both Tn5-wild type and the T n 5 based inverse transposon used to generate insertion mutations contain terminal inverted repeats of the 1534-bp IS50 elements, the inverse transposon contains BclI and BglII restriction sites conveniently close to its ends (Figure 2). Plasmids containing uninterrupted 22and 32-bp palindromes were generated by digestion with BclI and BglII, respectively (these enzymes do not cleave pBR322). An uninterrupted 90-bp palindrome was generated by the in vitro insertion of duplicate copies of a short segment from plasmid pGCl into the central BglII site of the 32-bp palindrome (Figure 3A). The sequences of the palindromic inserts are given in Table 2. The sequence of the 90-bp palindrome at A-10 was verified by DNA sequencing (MAXAM and GILBERT 1980) using DNA fragments end-labeled at the central XhoI site. The presence of the 90-bp insert at 1 1 of the other 12 sites was verified by electrophoresis of DNA fragments end-labeled at BglII sites within the inverted repeats as in NAG et al. (1985). The palindrome first constructed at site A-24 was found by end-labeling and DNA sequencing to contain an internal 5-bp deletion, and is hereafter designated A-246. A perfect 90-bp palindrome at this site was also constructed and verified. Measurement of deletion frequencies. The frequencies of revertants observed in a population will reflect the rate at which reversion occurs, and also differences in growth rate that favor one plasmid or cell type over another (PHADNIS and BERG 1985). T o assess the effects of selection we estimated reversion frequencies using fluctuation tests (LURIA and DELBRUCK 1943). For ease, young single colonies were used (Figure 4) instead of the traditional liquid cultures. Bacteria carrying the plasmids to be tested for reversion to Amp' were streaked on an L-agar tetracycline plate and grown at 37" until the colonies were just visible (about 7 hr of growth) and contained about 5 X 1 O5 viable cells per colony. The fraction of colonies which contained revertant subclones and relative revertant frequencies were estimated by streaking entire colonies on L-agar ampicillin plates (Figure 4). These control experiments showed that the quantitation of deletion frequency described below was not seriously biased by selection. To quantitate the frequencies of deletion of an insert plus one copy of its direct repeats, the reversion of the Amp' mutations to Ampr was selected as follows: Tet' Amps bacteria were streaked on L-agar tetracycline plates (12 Ng/ml) and incubated overnight at 37 O . Young single colonies were inoculated from these plates into 2 ml L-broth plus tetracycline, and grown to stationary phase (1 4 hr). Aliquots were then spread on L-agar ampicillin plates (250 Fg/ml) and incubated at 37". This represents about 32 generations of clonal growth from the single founding cell.