Tobacco Documents Online

© I$~3 F_~e~=r S~i=nc~ Ik~bI~h~rs B.V. All ri~.ht~ re.fred 0165-1218193/$~D&0~ • MUTGEN 00~02 Multiple pathways of deletion formation in Escherichia coli Elias Balbinder Department of Biochemi.~try, Biophysics and Genetics and Colorado Cancer Center, Univ.ershy of Colorado Health Sciences Center, 4200 East 9thAvenue, P.O. Box B 121, Denver, CO 80262, USA (Received 24 June 1992) (Revision received 29 December 1992) (Accepted 11 January 1~93) Keyword~: Deletions; Palindromes; SOS repair; Tnl0 excision; RecBC; Mutant selection for deletion enhancement We are investigating the mechanisms for deletion formation through the use of mutants which alter deletion frequency together with well characterized systems for deletion detection. We report here on three mutations which were isolated for their ability to stimulate deletions in plasmid pMC874 (dli mutations). The mutation rec-2251 (formerly known as dlil) is a new allele of recBCD, a group of genes coding for the polypeptide components of the major recombination enzyme complex in E. coli; the second one, dli2 may be a new allele of uvrD, which codes for DNA heliease lI; and the third one, dli3, has the phenotype of a mismatch repair mutation. Here we compare the effects of mutations in SOS-repair genes to those of the dli mutations on three different deletion events: (a) the deletion of short (60-100-bp) palindromic and non-palindromic inserts in derivatives of plasmid pBR325; (b) larger (600-800-bp) deletions in plasmid pMC874; and (e) the excision of the Tnl0 transposon from chromoso- mal sites. Our results indicate that some form of SOS processing stimulates the loss of palindromes but not non-palindromes in plasmid pBR325 derivatives, and that RecA is necessary for UV-induced excision of Tnl0 but this event is inhibited by UmuCD or its homolog MueAB. Each of the dli mutations showed unique effects on different classes of deletions. Mutation rec-2251 stimulated specifically deletions in pMC874 but had no effect on the deletion of non-palindromes in pBR325, and reduced the incidence of the other deletion events tested including loss of palindromic inserts in pBR325 as well as Tnl0 excision. Mutation dli2, on the other hand, stimulated all deletions tested to varying extents, while dli3 did not affect markedly deletion formation in pBR325 plasmids but had a large stimulatory effect on both deletions in plasmid pMC874 and Tnl0 excision. These results reveal that (a) some SOS-repair functions participate in deletion formation, (b) mutations selected for altering the incidence of one class of deletions may have totally different effects on other deletion events, and (e) the differences in mutant Correspondence: Dr. Elias Balbinder, Department of Bio- chemistry, Biophysics and Genetics and Colorado Cancer Center, University of Colorado Health Sciences Center, 4200 East 9th Avenue, P.O. Box B 121, Denver, CO 80262, LISA. 40000107

194 behavior may result in part from the ability of some pathways to discriminate among different deletion intermediates such as hairpins or cruciforms formed by palindromic sequences vs. transient seconday structures stabilized by direct repeats flanldng non-palindromic sequences. Although the e~stence of large genetic alter- ations such as deletions, duplications, inversions, translocations, etc. has been known for more than fifty years, interest in their study has increased over the last decade because of their role in cancer and genetic disease (Callahan and Camp- bell, 1989; DeKlein, 1987; Yunis, 1986). These alterations are generally considered to arise by poorly defined processes akin to "illegitimate re- combination" (Franklin, 1967), meaning events which differ from homologous recombination in not requiring extensive genetic homology or recA function. It is clear today that the situation is much more complex, and that genetic rearrange- ments can arise through a variety of mechanisms, none of which are yet well understood. Spontaneous deletions in prokaryotie replicons apparently can occur either through the resolu- tion of unstable transient intermediates formed in the normal course of DNA metabolism (AI- bertini et al., 1982; Ripley and Glickman, 1983), from intermoleeular recombination events (Di- anov etal., 1991; Mazin et al., 1991; Singer and Westlye, 1988), or as the result of errors in the movement of transposable elements (Kleckner et al., 1979). Deletions of the first two classes occur predominantly between direct repeats, which are often flanking inverted repeats (palindromes). In this case, it has been proposed that deletions result from the resolution of unstable structural intermediates such as hairpins or erueiforms (Al- bertini et al., 1982; Balbinder et al., 1989; Das- Gupta etal., 1987; Foster et al,, 198l; Gliekman and Ripley, 1984; Jones et al., 1982; Meulien et al., 1981; Ripley and Glickrnan, 1983; Singer and Westlye, 1988; Weston-Hafer and Berg, 1989, 1991). Cruciform structures in plasmids have ac- tually been shown to exist in rive (Zheng et al., 1991), but other mispaired intermediates have proven difficult to demonstrate. The misalign- ment mutagenesis model (Albertini et al., 1982; Glic'kman and Ripley, 1984) proposes that tran- sient deletion intermediates can form on single- stranded stretches of DNA as the result of slip- page during replication and that these intermedi- ates are resolved through several mechanisms, namely: (a) recombination by enzyme systems promoting rearrangements at sequence repeats; (b) misreplication across the base of misaligned secondary structures; and (c) nudease excision of the transient intermediate. There is some evi- dence for (a) and (b) above (Dianov et al., 1991; Mazin etal., 1991; Singer and Westlye, 1988; Brunier et al., 1988) but not for (e). It is possible that all three mechanisms may participate in the formation of different, or even the same, deletion events. Another model accounts for deletions generated by double-stranded DNA breaks be- tween direct repeats. This postulates that exonu- dease processing of these breaks produces single-stranded tails, and the eventual matching of short repeats within the latter followed by ligation restores a eovalently closed DNA molecule missing the segment between the re- peats (Conley et al., 1986; Pulak and Anderson, 1988). These models apply to deletions arising via intramoleeular events. However deletions result- ing from intermolecular homologous recombina- tion between long (20 bp or more) direct repeats have been reported to occur in bacteriophage T4 and some ColE1 plasmids (Dianov et al., 1991; Mazin etal., 1991; Singer and Westlye, 1988). A special case of deletions brought about by inter- molecular recombination seems to occur through the introduction by DNA gyrases and topoisom- erases of single- and double-strand breaks at sites of very little sequence homology, i.e. 3 bp (Ikeda etal., 1984; Ikeda, 1986). There are eases of deletions in the same repli- con mediated by different mechanisms depending on the length of the direct repeats, or other parameters. To give a few examples. (A) Brunier et al. (1989) reported that deletion of inverted repeats in engineered single-stranded 40000108

DNA plasmids in E. coil occurred by a repliea- tive slippage mechanism between flankSng direct repeats 9 bp long, but by a breakage znd reunion recombination proce~ss when the direct repeats were 18 bp or longer. (B) The formation of deletions between direct repeats of at least 20 bp in pBR327 derivatives was always associated with dimerization of pins- mid DNA (Dianov et al., 1991; Mazin et al., 1991), but this apparently occurred by different mechanisms, depending on the length of homol- ogy. In the case of ve~j long direct repeats (165- 401 bp), deletion and the accompanying dimeriza- tion occurred mainly by unequal crossing-over in a Rec+ background, while with smaller direct repeats (21 and 42 bp) the deletion frequency was greatly increased in recB- recC- backgrounds. This was interpreted as resulting from recombi- nation events taking place while the plasmids were undergoing rolling circle replication, which is inhibited by RecBC nuclease (Cohen and Clark, 1986; Mazin et al., 1991). (C) There are three different events in Tnl0 excision from chromosomal sites (Foster et al., 1981): precise excision, in which the entire trans- posen is removed between 9-bp direct repeats to restore the wild-type sequence; near-precise exci- sion which eliminates most of the transposon except for a 50-bp remnant between short repeats internal to Tnl0, and excision of the 50-bp rem- nant between the 9-bp direct repeats. Mutations isolated for stimulating Tn10 excision from the chromosome (tex mutations) stimulated precise and near precise excision, but had no effect on the excision of the short, remnant sequence (Lundblad and Kleckner, 1982). (D) Enzymes which control superhelical ten- sion can affect the deletion incidence of palin- dromic, but not of non-palindromic sequences. Sinden et al. (1991) found that deletion frequen- cies of palindromic fragments cloned into the EcoR1 site of pBR325 were increased by factors between 10 and 25 in strains carrying the topo- isomerase mutation topAIO, and decreased by factors of 2-10 in strains carrying the gyrase mutation g~,rB225. In contrast deletion frequen- cies of non-palindromic inserts were not affected by these mutations. In summary, it appears that deletions as well as other genetic rearrangements 195 are the re.suit of a "grab bag of processes" (Low, 1988) which normally participate in various as- pects of DNA metabolism and which are yet to be identified. A fruitful approach to elucidate deletion path-. ways is to fo~us on specific events where the structural parameters are knoxvn or can be con- trolled, and use mutants to see how they affect these events. This would include tmoxvn mutants in genes for DNA metabolism, as well as new mutants isolated by their ability to alter the inci- dence of deletion. A number of such mutants have been obtained using different selection sys- tems. Not surprisingly the different .approaches appear to have implicated different genes. With the exception of the uup mutations (Hopkins et al., 1983) selected by the stimulation of transpo- son Tn5 excision on an F' plasmid, and an as yet unidentified gene between xth and pnc whose deletion stimulates the occurrence of 1.3-kbp deletions in a derivative of plasmid pBR322 (Yi, Stearns and Demple, 1988), all other mutations thus isolated have turned out to be new alleles of already known genes. In general these have been of two types: (a) null alleles in which gene func- tion is lost, and (b) novel alleles giving enzymes with new properties. The recovery of null alleles indicates the presence of functions which nor- mally act to prevent the formation of deletions. ]n this group we can mention some of the rex mutations of Lundblad and Kleekner (1982, 1985) selected for enhancement of transposon Tnl0 excision (texB, texC, texD and texE) which fall genes for mismatch repair functions; a mutation in topB, the gene for topoisomerase III which eliminates the activity of this enzyme and en- hances deletions in an F'lac plasmid (Whoriskey et al., 1991; Schofield et al., 1992); and the xth- pnc deletion mentioned above (Yi, Stearns and Demple, 1988). In the second group, namely alle- les giving enzymes with new properties, are the tex,4 mutations (Ltmdblad and Kleekner, 1982; Lundblad et al., 1985) which are novel alleles of recBC, and the new alleles of sbcB, the geue for exonuelease I isolated by Allgood and Silhavy (1991) for their stimulation of deletions in plas- mid pZ157. The presence of mutations of this type indicates that some functions which normally do not participate in deletion formation have 40000109

been altered by mutation in such a way that they now take an active part in forming these rear- rangements. Although the isolation of mutants by their ability to alter deletion incidence has yielded, thus far, mostly phenomenological information, these mutants have been extremely useful in identifying poss~le pathways and will be of great value for the future study of the molecular mech- anisms responsible for deletions. One fruitful ap- proach at this time is to determine the effects of different mutations on structurally well-defined deletions. In this paper I review some of our work in progress to show how a combined approach using mutations together with well-characterized deletion substrates can b~. used to identify path- ways for specific classes of deletions. We will discuss first the isolation and characterization of mutations which enhance deletion incidence in plasmid pMC874 (dli, for "deletion increase'" mu.tations) and these will then be compared, for their effects on a variety of deletions, to muta- tions in SOS repair. Isolation and characterization of mutations which stimulate deletions in plasmid pMC874 -- A new recBC allele Plasmid pMC874 (Fig. 1) was constructed by Casadaban et al. (1980). This plasmid carries pMC874 Fig. I. Structure of plasmid p!k~C874 and two Lac+ revertants. The diagram of p},~C,.q74 has been slightly modified from Casadaban et al. (19S0) and the structures of pEB7 and pEBI9 have been arrived at by restrict{on enzyme analysis (Fig. 2). Both pEB7 and pEBI9 give a Lac+Kms phenot~e, indicating the deletion of a fragment which joins the "/onr promoter to the promoterless lac ~peran. Based on restriction anal~[s data, pEB7 has a delet[on of about ~0t} bp and pEB19 a deletion of about 700-8t30 bp plus an insertion of cquivaIent size opposite the deletion containing a second Pstl site. Both deletions in pEB7 and pEB19 eliminate the BamHl restriction site. 40000110

40000111

TABLE 1 PLEIOTROPIC PHENOTY'PES OF all MUTANTS AND CONTROLS Strains Geaob~e Phage Reeomb. Mutator * T42- Ared- gam- P1 trar~sd. Rif~ ~ Rifr EB265 WT - s ND 1 (34) EB3L3 rec-2251 (dill) s-m el 0.02 0.2 EB3Z5 dli2 - - ND 4.7 EB335 dli3 l-el ND 24.5 AFT3Z5 WT - s 1 1 (20) EB934 rec-2251 (P1 323 × AFT325) s-m el < 0.01 0.3 V74 recC343 - - 1 ND JC54Ol recB21C22 s-m el ND ND ND. not done: WT, wild-type. Plaque sizes of mutant phages on different hosts are indicated as per Schultz et al. (1983), as follows: el, extra larg/~; I, large; m, medium; s, small: -, no visible plaques. * Relative frequencies of Rift with WT = 1- Absolute freq. X 10-~ in parentheses. carry it, but gives rise to Lac÷ papillae at very low frequency (!.6 × 10-9) on McConkey's agar (Balbinder, 1988). Plasmids in these papillae carry deletions which join the kmr promoter to the lacZ gcne in pMC874 (Fig. 1). We know from restriction enzyme analysis (Fig. 2) that there are at least three deletion-associated events in plus- mid pMC874 and there may be more. The three events identified thus far are: (a) a single 600-bp deletion (pEB7, Fig. 1); (b) a deletion of about 700-800 plus an insertion of equivalent size on the opposite side of the plasmid (pEBI9, Fig. 1), and (c) dimerization of the plasmid accompanied by a deletion of about 700-800 bp (Fig. 2). The deletion in plasmid pEB7 could be the result of an intramolecular event, while pEB19 and the dimer plasmids may represent intermolecular events. 31 dli mutations were selected for a high papil- lation phenotype (Balbinder, 1988). 3 of these, which we tentatively designated dlil, dli2 and dli3 have been studied thus far. Our results indi- cate that d]il is a new recBC allele while dli2 and dli3 are not. The dlil mutation has been mapped by bacteriophage Pl-mediated transduction and found to be closely linked to thyA and arg~ at 60.3 rain on the E. coli chromosome, a segment that includes the recB and recC genes. It has been renamed rec-2251. The other two muta- tions, dli2 and dli3 did uot cotransduce, with thyA or argA. The phenotypes of the strains containing the mutations dlil, dli2 and dli3 (strains EB323, 325 and 335, respectively) are shown in Tables 1 and 2. In Table 1 we shmv how these strains plate certain bacteriophage mutants, their recombinational ability and whether they possess a mutator phenotype. In Table 2 we show how they affect the frequency of Lac-~ Lae+ deletions in plasmid pMC874 and the excision of the transposon Tnl0 from a chromosomal site, ga176::TnlO (Tex phenotype). Table 1 also in- cludes several controls, as follows. (1) Strain EB934 is a transductant from the cross used to TABLE 2 RELATIVE FREQUENCIES OF Lac- ~ Lac ~" DELE- TION IN PLASMID pMC.874 AND Tnl0 EXCISION FROM A CHROMOSOMAL SITF, (ga176:'.TulO) IN dli MU- TANTS Absolute frequencies, expressed as the average number of papillae per plate for cells growing eonfluently on MeConkel, agar, are shown in parentheses. Strain Gcnotype Lac- ~ Lac+ Gal- ~ Gal+ pMC874 ga176::TnlO EB265 WT 1 (5.3) I (126) EB323 rec-2251 72 0.13 V74 recC.~43 ND 4 EB325 dli2 3 > 24 EB335 dli3 63 > 17 40000112

replace recBCD + in strain AFT325 (Amundsen et al., 1990) with rec-2251; (2) strain AFT325; (3) EB265, which is strain MCI0fI0 (Casadaban and Cohen, 1980) carrying plasmid pMC874 and the ancestor of EB323, 325 and 335 (Balbinder, 1988); (4) JC5491 0,Villets and Clark, 1969) is a double mutant for the classical null alleles recB21 and recC22; (5) V74 carries the ter.A mutation recC343 (Lundblad et al., 1984). RecBCD is a multifunc- tional enzyme complex composed of the products of three genes, recB, recC and recD. Null muta- tions in recBC result in the loss of the enzyme's nuelease activities and a decrease in recombina- tional capability, the residual recombination be- ing due to the recF pathway (Smith, 1988; Taylor, 1988). All the strains in Table 1 were tested for their ability to plate the bacteriophages T4 2- and A red-garn- (Amundsen et al., 1990), re- combinational ability as determined by phage Pl-mediated transduction, and mutator phenotye (Lundblad and Kleckner, 1982) measured by the spontaneous mutation frequency from rifampicin sensitivity to resistance (Rif~ ~ Rift). The ability of an E. coli host to plate the bacteriophage mutants T4 2- and A red-gain- is a useful test to for ReeBCD nuelease activity (Lundblad and Kleekner, 1982; Amundsen et al., 1990). The DNA of phage T4 2- mutants is rapidly degraded by RecBCD enzyme so these fail to make plaques on recBC + cells but do on recBC null mutants (Amundsen et al., 1990). 3. red-gain- can grow well in recBC- mutants but not in wild-type E. coli because in the latter the RecBCD enzyme degrades concatemeric DNA, the only source of progeny lambda DNA and the A gain gene prod- uct prevents this from happening (Smith, 1983). The red- mutation eliminates the red recombi- nation pathway, which allows the production of some lambda progeny in recBCD + strains (ref. above). The null alleles recC22 and recB21 are deficient in recombination as well as all nuclease activities tested (Taylor, 1988). The phenotype conferred by mutation rec-2251 (strains EB323 and EB934) resembles that of the recB21 recC22 strain JC5491 in ability to plate T42- and A red-gam- (Table 1), and in having a decreased recombinational ability as reported by other workers (Willets and Clark, 1969; Taylor, 1988). Also, like null recBC-, rec-2251 differs from texA alleles in these respects (Lundblad and Kleetmer, 1982; Lundblad et al., 1984). The latter are un- able to plate A red-garn- and have normal re- combinational ability (strain V74, Table 1) and a normal spontaneous mutation frequency, i.e. no mutator phenotype (refs. above) while rec-2251 acts as an antimutator, i.e. it shows a decrease in spontaneous mutation frequency from Rifs to Rifr. As shown in Table 2, the rec-2251 mutation also differs from recC343 in having a negative Tex phenotype, i.e. it inhibits Tnl0 excision while recC343 stimulates it, as expected (Lundblad et al., 1984). As shown by strain EB934 (Table 1) the pleiotropic phenotype conferred by rec-2251, i.e., ability to plate phages T42- and X red-gara-, reduced recombinational ability and antimutator phenotype, cosegregates intact in transductional crosses. The rec-2251 mutation is the first case of a new recBC allele recovered by selecting for enhancement of a deletion event since the rex,4 mutations were isolated (Lundblad and Kleckner, 1982). The other two dli mutations, dli2 and dli3, have not been mapped yet but, as mentioned above, we know that they are not new recBC alleles. Interestingly however, their phenotypes (Tables 1 and 2) resemble other rex mutations corresponding to genes in methylation directed mismatch repair. The dli2 mutation resembles the texB/uorD mutations in the gene coding for helicase 1I in its total inability to plate ~ red- gam-, mutator and Tex phenotypes (Lundblad and Kleekner, 1982, 1985). The mutation dli3 resembles phenotypically the mismatch mutants texC /mutH, texD /rautS and texE /dam (Lundb- lad and Kleckner, 1982, 1985) in its ability to plate ,~ red-gain-, mutator and Tex phenotypes (Tables 1 and 2). Derivatives of pBR325 used to study deletions To control the structural parameters of dele- tions we have constructed derivatives of the plas- mid pBR325 carrying inserts of different size and sequence cloned into the EcoRI site of the chlo- ramphenicol acetyl transferase (cat) gene. This inactivates the gene and makes the cells carrying such plasmids sensitive to chloramphenicol (CmS). Since the only variables in these constructs are 40000113

the size and sequence of the inserts (Fig. 3), differences in deletion frequency (Cm~ Cmr) would likely be the result of these parameters. The construction of the plasmids has been de- scribed in detail (Balbinder et al., 1989; Sinden et al., 1991) but is briefly summarized here to hell) the reader. The relevant features of these plas- raids are shown in Fig. 3. In both pRS1 and pRS4, the insert is a non-palindromic 64-bp HaelIl fragment of plasmid pBR322 with EcoR! linkers added, but they differ in the sequence of the insert termini. In pRSI the insert is flanked by the same 8-bp sequence which includes an EcoRI site, while in pRS4 it carries an additional copy of the 8-bp repeat on the 3' side as a consequence of a 9-bp duplication. Because of the orientation of the insert in both plasmids, the terminal homologies include adjacent sequences to the left of each EcoPd linker which generate terminal 17-bp repeats (18-bp in pRS4) with a 15/17-bp (or 15/18-bp) homology. In oRS4, the TABLE 3 SPONTANEOUS DELETION FREOUENCIES × 10-9 (CmS~Crn*) IN PLASMID pBR325 DERIVATIVES IN DIFFERENT WILD-TYPE E. cofi STRAINS Plasmid Strain SC30-RP MC1000 pRS1 13+ 0.6 2.6_+ 0.8 pRS4 985 _+189 648 5:114 pOCEI5 47 + 15 73 +_ 7 FI4C. 241 + 90 614 ::1:159 FI4S 25 + 9 47 + 9O 9-bp tandem duplication creates a second perfect ll-bp direct repeat xvhich overlaps with the im- perfect 18-bp one. The presence of the over/ap- ping direct repeats in pRS4 increased the dele- tion frequency by more than five hundredfold over pRSI (Table 3), presumably because of the possible formation in pRS4 of multiple deletion intermediates between the terminal direct re- pR$1 pRS4 pOCE15 FJ~. ~. Relevant sequences of the pBR32~-derJvcd plasmids. (A) Sequencc of unique ~coR] site o~ the ca~ ~¢ne (bo~ed). pBR325 8enerafin~ at Ioast o~e such s~te (boxed) at e~ch end. "~e p~'rfect b~rpia str~ctur¢ pot¢~tial[y [ormed in ~OC~15 Js shown. The e~tcnded direct repeats i~ pRS1 and thc ovcrlappia~ direct repeats of pRS4 arc u~d¢rlln~d. M~smatches ig the ~ep¢ats are indicated by asterisks. For more details see text, Balbinder et al., 1989 and Sinden et ai., 1991. 40000114

peats (Balbinder et al., 1989). The other three plasmids, pOCE15, F14C and F14S carI3' palin- dromic inserts differing in size and sequence. Plasmid pOCE15 contains a 60-bp perfect palin- drome consisting of an inverted repeat of a lac operator fragment (Betz and Sadler, 1981; Bal- binder et al., 1989). Plasmids F14C and F14S were constructed by Sinden et al. (1991), and each carries the same 100-bp insert. They differ in that 14C is a perfect palindrome while the center sequences in F14S are non-palindromic (Fig. 3). The spontaneous deletion frequencies of these five plasmids in two "wild-type" strains: SC30RP (Wit'Ida et al., 1982) and MC1000 (Casadaban and Cohen, 1980) are shown in Table 3. The results are similar in both strains and in agreement with previous reports (Balbinder ctal., 1989; Sinden et al., 1991). Effect of SOS processing and dli mutations on the deletion of palindromic and non-palindromic inserts in derivatives of plasmid pBR325 The participation of SOS processing in dNe- t.ion formation is discussed in detail in a separate paper (Balbinder et al., 1993). We want to com- pare here the effects of mutations in SOS-repair genes to those of dli mutations to see whether the same deletions can be brought about by dif- ferent mechanisms. Table 4 summarizes experi- ments presented in detail in the report above, comparing the frequencies of Cm~ ~ Cmr dele- tions in pBR325 derivatives to those of base pair 201 substitutions in the chromosome (measured from the reversion of the mutation trpE65 from Trp- ---> Trp+), in SC30RP-derived strains carr~qng different combinations of alleles for the SOS-re- pair genes recA, lexA and umuC. The results show that deletion of the palindromic inserts was stimulated by overproduction of ReeA*730 and UmuC+ in ~471::Tn5 mutants, which lack LexA repressor and are consequently fully derepressed for the SOS genes (Walker, 1984), but deletion of the non-palindromic inserts was unaffected (com- pare strains Nos. 1 and 4, Table 4). The recA730 allele makes a genetically activated RecA* pro- tein and shows a mutator phenotype, i.e an in- crease in spontaneous mutation frequency (com- pare No. 1 to No. 2 and No. 4, Table 4, and see Witkin et al., 1982; Sweasy et at., 1990). Since the enhancement of palindrome deletion parallels that of Trp '--> Trp+ reversion we concluded that the former is caused by some form of SOS pro- cessing (Balbinder et al., 1993). However, the data indicate that there are differences between deletions and base substitutions. For example re- placement of umuC+ by umuC122::Tn5, a null aline of urnuC, reduced both point mutation and palindrome deletion frequency in plasmid pOCE15, but the former was reduced by a much larger factor than the latter (compare Nos. 4 and 5 in Table 4). Also the overproduction of ReeA+ protein in a lexA71::Tn5 strain increased point mutation Irequency, although not to the same extent as RecA*730, while not affecting deletion frequency noticeably (compare Nos. 1, 3 and 4 in TABLE 4 RELATIVE FREQUENCIES OF SPONTANEOUS Cm~ --* Cm' DELETIONS IN DERIVATIVES OF PLASMID taBR325 ('Fig. 3) AND Trp- ~ Trp+ REVERSIONS IN ISOGBNIC STRAINS CARRYING DIFFERENT COMBINATIONS OF recA, lexA AND umuC ALLELES Absolute frequencies are shown in parentheses and are expressed as the number of Cmr revertants per 109 cells for deletions (see SC30-RP in Table 3), and as the average number of Trp+ revertants per plate for base substitution mutations. Data from Balbinder et al. (1993). No. Relevant genotype Non-palindromos Palindromes Trp- -* Trp + recA lexA umuC pRS1 pRS4 pOCE15 F14C F14S revs. I + + + 1 (1.5) I (985) 1 (47) 1 (241) (1) 25 1 (16) 2 730 + + 1.4 0.9 2.3 1.1 1.4 13 3 + 71::Tn5 + 1.8 0.8 1.5 2.7 0.7 4 4 730 71::Ta5 + 1.5 1.15 14.2 6.2 3 32 5 730 71::Tn5 122::Tn5 ND ND 4.2 ND ND 1.5 40000115

202 Table 4). Our results also indicate that deletion enhancement by SOS processing favors perfect palindromes o~'er imperfect ones (F14C vs. F14S) and perhaps smaller palindromes over larger ones (pOCE15 vs. F14C), and that the sequences of terminal repeats are not important parameters in this process as shown by the lack of enhancement of the deletion of the non-palindromic inserts in pRS1 and pRS4. These observations suggest that the SOS effect depends on the recognition of palindromy vs. non-palindromy, i.e. the configu- ration of the transient intermediates formed on single-stranded stretches of DNA during replica- tion, rather than on the sequences of the terminal homologies or those of the palindromic inserts in pBR325. In Table 5 we show how different deletion events in the pBR325 derived plasmids are af- fected in strains carrying the different dli muta- tions. The mutation rec-2251 (strain EB323) had no major effect on the deletion of non- palindromic inserts but reduced substantially the deletion of both perfect palindromes while that of the imperfect palidrome in F14S was inhibited by only a factor of two. It is interesting that these effects are exactly the reverse of those shown in Table 4 for SOS processing, again suggesting the possibility that certain enzymes can differentiate between deletion substrates based on their see- oudary structure. Table 4 also shows that the texA mutation recC343 and the null alleles recB21 recC22 have no major effect on the deletion of the non-palindromic inserts and, perhaps, inhibit slightly the deletion of the palindromic inserts. In contrast to rec-2251, all deletion events, both of palindromes and non-palindromes, were in- creased albeit to different extents in the presence of dli2 (strain EB325), while dli3 (strain EB335) showed no major effect on palindrome deletions and only minor reductions in the deletion of perfect palindromes. Other deletions -- Effects of SOS functions and of dli mutations on excision of Tnl0 from chro- mosomal sites and of dli mutations on deletions in pMC874 In this section we compare the effect of differ- ent mutations on a variety of deletion events. Although the results are still incomplete, these comparisons suggest ways in which different ge- netic functions could participate in giving rise to various classes of deletions. The d/i mutations were selected for the stimulation of Lac-~ Lac+ deletions in a different replicon, plasmid pMC874, which were larger than the ones in pBR325. Also as xve have seen, deletions in pMC874 could result from more than one event (Figs. 1 and 2). The excision of the transposon Tnl0 from chro- mosomal sites represents a class of deletions which differs from those in pBR325 derivatives and pMC874 in every structural parameter: size of the deletion, size and sequences of the termi- nal direct repeats, location on a different repli- TABLE 5 EFFECTS OF dli MUTATIONS ON DELETION FREQUENCIES IN pBR325 DERIVED PLASMIDS Strain Allele Plasmids Cms ~ Cmr Non-palindromes Palindromes pRS1 pRS4 pOCE15 FI4C F14$ EB265 WT 1 (2.6) 1 (648) 1 (73) l (614) 1 (47) JC5491 recB21recC22 1.15 2.3 0.7 0.55 0.55 V74 recC343 0.4 1.1 0.2 0.3 0.14 EB323 rec-2251 1.5 2.2 0.04 0.05 0.5 EB325 dli-2 22 8.6 3.5 10A 15.4 EB335 dli-3 0.8 2 0.2 0.3 1.04 Relative frequencies of deletion in derivatix'es of plasraid pBR325 in dli and recBC mutants. Absolute frequencies are shown in parenthese.s and expressed as the number of Cmr revertants per 109 cells plated (see MC1000 in Table 3). 40000118