Palindromic DNA (DNA Secondary Structure/Addition and Deletion Mutations/Deletion Mutation Specfflcity/Evolution) BARRY W

Proc. Nati. Acad. Sci. USA Vol. 81, pp. 512-516, January 1984 Genetics Structural intermediates of deletion mutagenesis: A role for palindromic DNA (DNA secondary structure/addition and deletion mutations/deletion mutation specfflcity/evolution) BARRY W. GLICKMAN AND LYNN S. RIPLEY Laboratory of Genetics, National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709 Communicated by Matthew Meselson, September 21, 1983

ABSTRACT A model is presented for deletion mutations ic or quasipalindromic DNA sequences participate in dele- whose formation is mediated by palindromic and quasipalin- tion formation through the juxtaposition of deletion end dromic DNA sequences. It proposes that the self-complemen- points. The inherent self-complementarity of palindromic tarity of palindromes allows the formation of DNA secondary DNA sequences permits the formation of cruciform or hair- structures that serve as deletion intermediates. The structures pin structures in nucleic acids, precisely juxtaposing other- juxtapose the end points of the deletion and thus direct dele- wise distant bases (6). Repeated DNA sequences are not re- tion specificity. While misaligned DNA intermediates that ex- quired. plain deletion termini occurring in repeated DNA sequences The termini of the E. coli lacd deletion S86 cannot be jux- have been described, no explanations have been offered for taposed by a misalignment involving repeated sequences. deletion termini occurring in other sequences. The DNA sec- However, the deletion termini are located precisely at the ondary structures whose formation is mediated by palindrom- ends of a quasipalindromic DNA sequence that includes the ic sequences appear to explain many of these. In this paper, entire deletion. The formation ofa DNA hairpin or cruciform secondary-structure intermediates are described for a series of structure in the wild-type DNA sequence places the deletion spontaneous deletions of known sequence in the lad gene of termini immediately adjacent to one another, rather than Escherichia coli. The model is supported by its failure to pre- separated by their normal linear distance of 27 base pairs dict structures that canjuxtapose simulated deletion termini in (Fig. 1A). This structural intermediate would produce the the lacI gene. We have found a strong association between pal- S86 deletion if it served as a substrate for excision or as a indromic sequences and repeated sequences at lacl deletion template for DNA synthesis (Fig. 1B). The latter process termini that suggests the joint participation of repeated and might be mediated by either DNA polymerase or DNA li- palindromic DNA sequences in the formation of some dele- gase. A ligation across such a hairpin stem might also be tions. Sequences of deletions in other organisms also suggest responsible for deletions if the hairpin served as a substrate the participation of palindromic DNA sequences in the forma- for nucleases that removed aberrantly aligned regions from tion of deletions. ordinary double-stranded DNA. Whatever the precise molecular mechanism(s) responsible Local DNA sequences can strongly influence the frequency for the production of deletions from such DNA structures, of mutation (1). The frequent association of deletion termini the prediction of the model is that the deletion termini are with repeated DNA sequences (2-4) suggests that deletion sequence determined and coincide with the end points of the mutations are no exception. The DNA repeats have been palindromic sequences. Secondary structures whose forma- postulated to mediate complementary base pairing between tion is mediated by palindromes precisely coincide with the two misaligned DNA strands. The subsequent fixation of the deletion termini of S86 (Fig. lA) and 556 (Fig. 1C) and thus interstrand misalignment by DNA replication or recombina- offer an explanation for the formation of these deletions. tion generates a deletion or addition mutation. This model for deletion mutagenesis depends upon the presence of re- Quasipalindromes and Repeats at Deletion Termini peated sequences and is analogous to the model for frameshift mutations proposed by Streisinger et al. (5) for the pro- A previous analysis of the specificity of deletions in the lacI duction offrameshift mutations in locally repeated DNA se- gene demonstrated that DNA repeats were associated with 7 quences. of 12 spontaneous deletions (3). Our analysis of these same Among spontaneous deletion mutations of known se- deletions revealed that, among the 7 deletions associated quence in the Escherichia coli lacI gene (3), however, nearly with repeats, the specific termini of 5 are also predicted by half (5 of 12) have termini in sequences that are not repeated. palindromic sequences. Fig. 2 illustrates DNA hairpins that The origins of these deletions have thus been mysterious. juxtapose the termini of the S74, S23, and S32 deletions. The Our examination of the DNA sequences that are neighbors quasipalindromes responsible for the hairpins, like those in- to deletion termini has revealed the presence of quasipalin- volved in the S86 and S56 deletions, can precisely explain dromic sequences. Misalignments whose formation is medi- the deletion ends. In other words, even were the repeated ated by palindromes are capable ofjuxtaposing deletion ter- sequence absent, the same deletion specificities (termini) are mini, providing explanations for the specific end points of predicted. Also shown in Fig. 2 is deletion S10. The specific many lacI deletions. termini of this deletion are predicted only by repeats. The fact that S86 and S56 can be explained by palindromic Deletion Specificity Directed by Quasipalindromic Sequences sequences in the absence ofrepeats and S10 by DNA repeats in the absence of a palindrome suggests that either palin- In its simplest form, our model proposes that DNA second- dromes or repeats can direct deletion formation. However, ary structures whose formation is potentiated by palindrom- as discussed below, the striking coincidence of both palindromic and repeated sequences at 5 of the 12 deletion end can The publication costs of this article were defrayed in part by page charge points suggests the cooperation of both components. As payment. This article must therefore be hereby marked "advertisement" be seen in Fig. 2, the DNA hairpins permitted by the palin- in accordance with 18 U.S.C. §1734 solely to indicate this fact. dromes can be substantially stabilized by interstrand hydro- 512 Downloaded by guest on September 26, 2021 Genetics: Glickman and Ripley Proc. NatL. Acad Sci. USA 81 (1984) 513

A G-A G A peats. Alternatively, it has been suggested that the role of c C G G the palindrome might be to bring the repeats responsible for AG IT, GI this deletion into substantially closer proximity (7). A A G A C * G C .G Deletion Specificity and Complex DNA Secondary Structures , 650 C 690 5-G-C-A-A-T-C-A-A-A-T-C -T-G-G -A--T-G- 3 The models discussed above provide specific explanations for the origins of 9 of the 12 lacI deletions whose sequences B 5- -3 5- have been determined. The termini of the three remaining 3- deletions are associated with repeats of no more than two c A A bases and do not lie precisely at the ends of quasipalindromic 5'- -3' T I sequences. However, both interstrand and intrastrand mis- 3.'....'r5 T-A T-A alignments of a more complex type offer a means by which

'I termini may be juxtaposed and hence could C CC these deletion C G.C- 715 explain their specificity. The basis for these complex mis- T-A T A com- . 3' _A*T alignments resides in the presence of nearby sequences 69021 plementary to the novel joint created by the deletion. At the 5U- A-G-C--C-C-A-A-A3' novel joint, the DNA sequence immediately to the left of the FIG. 1. Production of deletions by palindromic DNA sequences left terminus of the deletion is joined to the sequence imme- at deletion termini. (A) A DNA secondary structure aligning the diately to the right of the right terminus. The nearby se- ends of the lacI deletion S86 by means of a palindromic sequence. In quence that is complementary to this novel joint permits a this and all subsequent figures the numbering of lacI sequences is misalignment thatjuxtaposes the deletion termini. When this from Farabaugh et al. (3) and the deletion termini are indicated by nearby sequence is a repeat of the novel joint, an interstrand arrows. (B) One mechanism for deletion formation; the DNA sec- misalignment mediates the structure; when the sequence is ondary structure forms within a gapped DNA molecule and this in- related to the novel joint, an intrastrand mis- trastrand misalignment is fixed by DNA synthesis across the stem of palindromically the hairpin, generating the deletion. (C) A DNA secondary structure alignment mediates the structure. aligning the ends of the lad deletion S56 by means of a palindromic The lacI deletions S24 and S42 may arise from structures sequence. that involve interstrand misalignments whose formation is mediated by repeats of the sequences at their novel joints. gen bonding due to the presence of repeats. Alternatively, The S24 ends are juxtaposed by a sequence 53 base pairs the interstrand misalignments permitted by repeated se- away, which is a repeat of 7 bases at the novel joint (3 to the quences can be viewed as being stabilized by intrastrand hy- left and 4 to the right of the deletion) (Fig. 3). Similarly, the drogen bonding due to the palindromes. In the case of S10, in S42 termini are juxtaposed by a sequence 25 base pairs re- which repeated sequences can define the deletion termini, moved by an interstrand misalignment permitting the pairing there is substantial potential for hydrogen bonding within the of 9 out of 10 bases at the novel joint (Fig. 3). deleted sequence. Here, then, intrastrand hydrogen bonding A complex structure that specifies the termini of the S120 might contribute to the stability of the structural intermedi- deletion (Fig. 3) depends upon an intrastrand misalignment ate but does not define the specificity of the deletion. Similar involving a 6-base sequence that is palindromic to the se- potentials may exist in much larger deletions as well. One E. quence at the novel joint. This complex structure not only is coli deletion extending between the lacZ and lad genes con- consistent with the specific end points of the deletion but tains a large quasipalindromic sequence that might substan- also may be substantially stabilized by additional hydrogen tially stabilize the interstrand misalignment mediated by re- bonding. There is substantial potential for hydrogen bonding

S23 A- X 6 6

S74 A A S32 330~~~~~~~~25 .C GC 5 C'~ C' -A-CG- C-C-C-3- 60- A A- A 'CC 310 A-C-C-A-C-C-A5 J C -C G :.A_CAC

80 -A SIC) 340 C-A -C A--TA----GC-GC T^ 8 -A-C CCAAC- AA-AT - A C ACCCTG C A-A-A-A-C-A G ! G CG 4O T A A C C C T1- -A A- I A 320) 34040 G-T ^^CCAA T T CA-- T ----- 3'CA c . 6 C G (^-C-C-^-^-1-6-T-G--C-A-C-T- S1tti 10 -T /fGC ~ ~ ~ 3 1640 320 - 6 - -A A A A A A C~~~~~~~~~-C-A-C-A-C-A' A A T CACG C CACGCG C- --GCC310 A-C-C-C-C-CG-C-T

FIG. 2. Structural intermediates thatjuxtapose deletion termini that lie in repeated DNA sequences. All four deletions may be explained by interstrand pairing mediated by repeats. Alternatively, the specificity of the S74, S23, and S32 deletions is explained by palindromic sequences. These structures have substantial intrastrand hydrogen bonding that juxtapose the termini. Deletion SJO can form intrastrand hydrogen bonds, but it is the interstrand bonds that direct the specificity of the mutation. DNA repeats are boxed. Downloaded by guest on September 26, 2021 514 Genetics: Glickman and Ripley Proc. NatL Acad Sd USA 81 (1984)

S24 700 CO S120 ,C-T, C C-310 G T CC T T AT C.C C-C 5-T-G-G-A- G ,A-A-A-C-C-3' CT T GC 7A C TC A C G G T T G CT 3'-C-G-A-C-C C-T-A-G-S' CC 770

C' i I S42 IC ,,TA' A CC330A T -95o TCCC CT ACCC CCC T T C A C A C C r AC ;T 407 HAT 130 C A C 965 285285 ,-C A-T-C-3' 5'-T-G-G-A-C, C A C- G- T-G- 3' 5'-C C C CC ATTAACT 3'-C C CC C-CT A C T T C A G C GG A C G A C C C C G C - C- C 3'-A--A-A-A T-T-7-T-C-5C 2852C 300 9 50 2 FIG. 3. Complex secondary structure explanations for specific termini of lad deletions S24, S42, and S120. Formation of the S42 and S24 deletions is directed by the interstrand misalignments due to the complementarity of nearby sequences to the noveljoint of the deletion. The 71- base-pair S120 deletion can be explained by formation of a complex secondary structure mediated by intrastrand misalignments involving complementary bases external to the deletion. Intriguingly, this intrastrand misalignment can be further stabilized by a nearly perfect repeat (6 of 7 bases) in a manner analogous to that described for the deletions in Fig. 2. DNA repeats are boxed. within the deleted sequence as seen previously for SO (Fig. permit interstrand misalignments when complementarity is 2). Moreover, as with the stabilized hairpins described in provided by a repeat of the novel joint or intrastrand mis- Fig. 2, a repeat (in this case nearly perfect, 6 of 7 bases) alignments when complementarity is provided by a palin- stabilizes the hairpin upon which the deletion termini are drome (e.g., see Fig. 3). juxtaposed. The thorough characterization ofpotential structural intermediates requires that all potential sources of both inter- Methods of DNA Sequence Inspection strand and intrastrand hydrogen bonding by considered be- fore the preferred intermediate can be selected. Each inves- The essential feature of all the misaligned DNA structures tigator will necessarily have to decide upon parameters such that we propose is their ability to juxtapose deletion end as how long or how perfect a DNA repeat or palindrome points. Although all the structures portrayed in the figures must be or, in the case of complex deletions, how close to were optimized for maximal stability by computer, their abil- the deletion it must be. The parameters we have used were ity to juxtapose deletion termini was determined by eye. mainly determined empirically as described in the next sec- Thus, the model as proposed may be applied without the aid tion (e.g., see Table 1). of a computer to any DNA sequence in which a deletion is known. Tests of lacI Deletion Specificity During the examination of DNA sequences for explanations of deletion specificity, we developed a routine ap- The characteristics of all 12 spontaneous lad deletions con- proach. Initially, deletions are inspected to see iftheir occur- sidered here and the character of the misalignments that rence involves DNA repeats, remembering that the misalign- juxtapose their termini are shown in Table 1. Substantial re- ment model requires that one copy of the repeat be retained peats, palindromes, or both provide the potential for mis- while the other copy of the repeat and the intervening se- alignments in each case and thus strongly suggest the in- quence is lost. Next, each deletion sequence is inspected to volvement of inter- and intrastrand misalignments. We have identify palindromic components that might juxtapose the undertaken several comparisons that demonstrate a substan- deletion termini. In the absence of repeated sequences, the tial enrichment of sequences capable of producing such mis- deletion endpoints are unique and the presence (or not) of a alignments at these deletion termini compared to the fre- palindrome is immediately obvious (e.g., see Fig. 1). When quency of such sequences lying throughout the lacI gene. DNA repeats consistent with the deletion are identified, the Coincidence of DNA repeats and palindromes at deletion inspection for palindromic sequences that juxtapose the de- termini. Three out of four lacI deletions explicable by re- letion termini is carried out in several steps. The internal peats are also explicable by palindromes. Although the sam- portions of the deletion most distal to the first copy of the ple is small, the result is unexpected. We have attempted to repeat are examined for palindromic sequences complemen- evaluate its significance by comparing this very high fre- tary to the repeated sequence. Because the structural inter- quency to the frequency of such coincidences in the entire mediate may alternatively involve the misalignment of the lacI gene. The number of repeats of five or more bases that second copy of the repeat on the complement of the first, the are separated by distances that permit the deletion of be- most proximal bases of the deletion are also inspected for tween 10 and 150 base pairs is 79. Among these, 5 (6.4%) palindromic complementarity to the second copy of the re- coincide with palindromes of four or more base pairs. Not peat. This inspection process continues because the struc- only is 6.4% lower than 75%, but the recovery of 3 of the 5 tural intermediate may use only a portion of the repeat for examples of coincident repeated and palindromic sequences interstrand bonding, saving the remaining bases for intra- in a sample of only 12 deletion sequences in a gene of 1,080 strand bonding (e.g., see Fig. 2). (Note: When deletions in- base pairs strongly suggests the association of coincident re- volve repeats, although the sequence of the deletion is not peated and palindromic DNA sequences with favored dele- ambiguous, the physical end point can lie at any position tion sites. within a repeat. The number of combinations of end points These coincidences are not unique to lacI. For example, that must be examined as potential structural intermediates two deletions in the lysozyme gene of bacteriophage T4 exceeds the number of repeated bases by one.) whose sequences were determined (8) have termini at coinci- Structures explaining complex deletions differ from those dent palindromic and repeated DNA sequences (9). described above in that the misalignments are mediated by Structural intermediates do not juxtapose random points in DNA sequences external to the deletion. Nearby sequences DNA sequences. Misaligned DNA structures whose forma- complementary to the novel joint created by the deletion tion was mediated by repeats, palindromes, or combinations Downloaded by guest on September 26, 2021 Genetics: Glickman and Ripley Proc. Natl. Acad. Sci. USA 81 (1984) 515

Table 1. Summary of DNA sequences that juxtapose the termini of spontaneous lacI deletion mutants No. of bases paired in DNA No. of bases paired (and their distances misalignments mediated by from the novel joint) in complex structures Size, repeats and palindromes at the mediated by repeats and palindromes No. of times base deletion termini complementary to the novel joint Deletion isolated pairs Repeats Palindromes Repeats Palindromes S74 2 75 8 7of8 S23 1 123 8 4 of 5 S32 2 22 5 4 S10 2 20 8 0 S86 1 27 0 5 of 6 556 1 25 2 8of12 S24 1 13 2 0 7 (53)* 5 (65) S42 1 13 1 0 lOt (25)* 5 (15) S120 1 71 0 0 0 6(26)* DNA misalignments serve to juxtapose deletion termini and direct the formation of specific end points. DNA repeats permit formation of interstrand misalignments, which have been proposed to account for both frameshifts (5) and deletions (3) lying in repeats. Palindromes permit formation of intrastrand misalignments capable of juxtaposing the termini of many hitherto unexplained deletions. In the case of complex structures the misalignment involves sequences external to the deletion that are complementary to the novel joint created by the deletion. Repeats permit interstrand misalignments while palindromes permit intrastrand misalignments. Distance is measured from the base adjacent to the proximal end of the deletion to the first base of the nearby sequence upon which the misalignment occurs. *Indicates the complex structures illustrated in Fig. 3. tThis includes one GOT apposition; if it is disallowed, the number is 6. thereof were found for all the sequenced spontaneous lacI deletion end points, despite their distances in linear DNA, deletions (Table 1). Because the deletion sample size is might be joined through metabolic events in which mis- small, and because the various DNA structures associated aligned DNA structures participate as either substrates or with the termini have different expected frequencies, we templates. We have described in general terms potential in- evaluated the significance of the perfect correlation between teractions between secondary structures in DNA whose for- sequenced termini and DNA structures by determining the mation is mediated by palindromes and metabolic processes frequency of similar structural intermediates at the termini of that could result in deletion. Indeed, some evidence exists "random" deletions. Simulated deletions were created by for the operation of enzymes of DNA metabolism upon pal- using a Hewlett-Packard random number generator to define indromic structures as both substrates and templates. Hair- the 3' end on the nontranscribed strand and independently pins can be substrates for nucleases (10-12). Deletion muta- the size of the deletion. All simulated deletions lie entirely tions recovered after cloning procedures are consistent with within the lacI coding sequence and are between 10 and 100 the in vitro formation of hairpins and the subsequent diges- base pairs long. tion of their unpaired loops by S1 nuclease (13). Analogous Structural intermediates were then sought for 20 random in vivo results might be expected from nucleases responsible deletions. None of the termini were juxtaposed by repeated for the removal of aberrant DNA structures or the resolution or palindromic sequences of five or more base pairs at the of recombinant DNA structures (12). Misaligned quasipalin- ends of the deletion; at best, one set of end points was dromic DNA sequences appear to serve as templates in the brought to within a single base pair by a quasipalindrome. production of certain yeast frameshift mutations (14). There Thus, it is clear that the high frequency of repeated or palin- is abundant in vitro evidence that DNA polymerases behave dromic sequences (or both) at deletion ends in the lacI gene aberrantly when confronted with templates containing hair- is not expected on a random basis. pins (15-17), and an association exists between an extensive We also sought complex structural explanations for the quasipalindromic sequence and an anomalous strong muta- same 20 random deletions. Complex structures were re- tor effect of an "antimutagenic" polymerase (18). quired to permit either an inter- or intrastrand misalignment Studies that identify the specific enzymatic processes in- mediated by at least 6 base pairs that were located within 100 volved in deletion mutagenesis should substantially improve base pairs of either end of the deletion. Even with these com- our ability to expand the general models for secondary struc- paratively nonrestrictive rules for identifying complex structure involvement in deletion specificity to more specific tures, only 4 of the 20 simulated deletions could be ex- models that permit predictions of deletion mutagenesis fre- plained. This is in striking contrast to the 3/3 frequency of quencies. For example, we might then be able to distinguish complex structures that juxtaposed sequenced lacI deletion whether the higher frequency of lacI deletions associated termini not explained by other structures. with the concomitant occurrence of palindromes and repeats Each of our tests supports the conclusion that the sample is due to improved structural stability of the intermediate or of sequenced spontaneous deletion termini in the lacI gene is instead is due to the independent occurrence of mutagenesis nonrandom with respect to DNA sequences having the po- due to both repeat-mediated mechanisms and palindrome- tential to precisely juxtapose the deletion end points. We mediated mechanisms at a single site. conclude that the structural intermediates predicted by the Quantitative predictions of deletion frequencies mediated model provide an attractive explanation for deletion muta- by DNA structures involving repeats, quasipalindromes, or tion specificity. both are complicated by the diverse aspects of DNA metabolism and DNA topology likely to influence the frequency Discussion with which the DNA secondary structures form. Misaligned structures are not favored in double-stranded DNA but can Our DNA secondary-structure model predicts the participa- be expected in single-stranded or supercoiled DNA. Thus, tion of a variety of structural intermediates in deletion for- the formation of a DNA hairpin is likely to depend not only mation. It provides a unified basis for understanding how upon its intrinsic hydrogen bonding potential but also upon Downloaded by guest on September 26, 2021 516 Genetics: Glickman and Ripley Proc. NatL Acad ScL USA 81 (1984)

&C-A, *^ Is

A-C-C-A-T-T A-C-A C-AI-TC-C-A T-A-A FIG. 4. Deletion of a quasipalindrome might account for the C C difference in length between the 3' untranslated messages of S.C chicken 3- and p'-globin genes. The structure is analogous to the ACT lad deletions S86 and S56 (Fig. 1). The left-hand box indicates the C *C termination codons at the ends of the translated portions of each 'C C message. The right-hand box indicates the sequence thought to ! T It direct polyadenylylation of the message. The hairpin structure C-C represents a computer-fitted optimal structure. An alignment is shown for the P-Gkbn sequenceT-AACTAC C A CFTAc A CCACCA ACACCTGTAAC AAT remaining chicken 3-globin and p'-globin se- p'-Gbin sequence 5'- A TC CCAC[CC CLC6CTCCCAC ACCCAC;C;Ci CA TC[ T 1 quences; identical bases are indicated by stars. the processes that influence the formation of single-stranded gratefully acknowledge the assistance of F. Harms and T. Wertz in or supercoiled intermediates (19). These processes are likely the computer analysis of potential DNA secondary structure. to include many aspects of DNA replication, recombination, and repair. 1. Benzer, S. (1961) Proc. Nail. Acad. Sci. USA 47, 403-415. We have also found examples among eukaryotic DNA se- 2. Stewart, J. W. & Sherman, F. (1974) in Molecular qnd Envi- quences of deletion mutations that may have arisen as a con- ronmental Aspects of Mutagenesis, eds. Prakash, L., Sher- man, F., Miller, M. W., Lawrence, C. W. & Taber, J. W. sequence of a DNA secondary structure intermediate. The (Thomas, Springfield, IL), pp. 102-127. presence of deletion termini in repeated DNA sequences has 3. Farabaugh, P. J., Schmeissner, U., Hofer, M. & Miller, J. H. previously been noted in the globin genes (4). Most of these (1978) J. Mol. Biol. 126, 847-863. deletions are too small to be the product of the deletion of an 4. Efstradiadis, A., Posakony, J. W., Maniatis, T., Lawn, R. M., entire hairpin, and we have turned to transcribed but non- O'Connel, C., Spritz, R. A., DeRiel, J. K., Forget, B. G., translated portions of these genes to look for deletions. Weissman, S. M., Slightom, J. L., Blechl, A. E., Baralle, These nontranslated sequences are believed to be particular- F. E., Shoulders, C. C. & Proudfoot, N. J. (1980) Cell 21, 653- ly rich in deletion and addition mutations (20). An example 668. a 5. Streisinger, G., Okada, Y., Emrich, J., Newton, J., Tsugita, of how secondary-structure intermediate might have pro- A., Terzaghi, E. & Inouye, M. (1966) Cold Spring Harbor duced the base sequence difference in the 3' untranslated se- Symp. Quant. Biol. 31, 77-84. quences of chicken , and p'-globin messengers is shown in 6. Gierer, A. (1966) Nature (London) 216, 667-669. Fig. 4. Alignment of the two sequences between the termina- 7. Albertini, A. M., Hofer, N., Calos, M. P. & Miller, J. M. tion of the coding sequence and the A-A-T-A-A-A hexanu- (1982) Cell 29, 319-328. cleotide believed to direct polyadenylylation showed that the 8. Owen, J. E., Shultz, D. W., Taylor, A. & Smith, G. R. (1983) p'-globin sequence was 5Q bases shorter than the /-globin J. Mol. Biol. 165, 229-248. sequence. An attempt to explain this difference by formation 9. Ripley, L. S. & Glickman, B. W. (1983) Cold Spring Harbor of deletions mediated by repeats employed three indepen- Symp. Quant. Biol. 47, 851-861. dent events in which portions of both repeats were retained 10. Panayotatos, N. & Wells, R. D. (1981) Nature (London) 289, 1- 466-470. within the alignments (21). Our examination of the longer 11. Lilley, D. M. (1981) Nature (London) 292, 380-382. globin sequence revealed a quasipalindrome having substan- 12. Porter, A. G., Barber, C., Carey, N. H., Hallewell, R. A., tial complementarity and juxtaposing potential deletion end Threlfall, G. & Emtage, J. S. (1979) Nature (London) 282,471- points separated by 54 base pairs (Fig. 4). Deletion of the 477. hairpin could account for the substantial difference in length 13. Mizuuchi, K., Kemper, B., Hays, J. & Weisberg, R. A. (1982) between the two sequences. The suggestion is attractive be- Cell 29, 357-365. cause it accounts for the difference in a single mutational 14. Ripley, L. S. (1982) Proc. Natl. Acad. Sci. USA 79, 4128- event. However, it must be viewed with caution because it 4132. 15. Sherman, L. A. & Gefter, M. L. (1976) J. Mol. Biol. 103, 61- assumes that the difference arose by deletion and not by in- 76. sertion and that it is not a cloning artifact (12). 16. Ikoku, A. S. & Hearst, J. E. (1981) J. Mol. Biol. 151, 245-259. Misalignment of DNA sequences provides opportunities 17. Kaguni, L. S. & Clayton, D. A. (1982) Proc. Natl. Acad. Sci. for the production ofduplications as well as deletions. Dupli- USA 79, 983-987. cations of DNA sequences between repeats have been noted 18. Ripley, L. S., Glickman, B. W. & Shoemaker, N. B. (1983) (22), and duplications and other complex rearrangements Mol. Gen. Genet. 189, 113-117. have been observed in association with palindromic DNA 19. Wells, R. D., Goodman, T. C., Hillen, W., Horn, G. T., sequences We have discussed events in Klein, R. D., Larson, J. E., Muller, U. R., Neuendorf, S. K., (23-27). duplication Panayotatos, N. & Stirdivant, S. M. (1980) Prog. Nucleic Acid more detail elsewhere (9). Res. Mol. Biol. 24, 167-267. The focus of our discussion has been the DNA secondary- 20. Van Ooyen, A., van den Berg, J., Mantei, N. & Weissman, C. structure intermediates of deletion. We conclude that dele- (1979) Science 206, 337-344. tion specificity is largely sequence directed: The association 21. Roninson, I. B. & Ingram, V. M. (1981) Proc. Nail. Acad. Sci. of palindromic DNA sequences with deletion end points is USA 78, 4782-4785. consistent with their potential to juxtapose deletion end 22. Edlund, T. & Normark, S. (1981) Nature (London) 292, 269- points. Tests ofthis model are now possible through manipu- 271. lation of neighboring DNA sequences and identification of 23. Collins, J. (1980) Cold Spring Harbor Symp. Quant. Biol. 45, the of DNA metabolism for deletions. 409-416. aspects responsible 24. Vockaert, G., Tavernier, J., Derynck, R., Devos, R. & Fiers, Palindromic or quasipalindromic sequences are also impor- W. (1981) Gene 15, 215-223. tant in determining the specificity and frequency of several 25. Ghosal, D. & Saedler, H. (1978) Nature (London) 275, 611- other types of spontaneous and induced mutation (14, 18, 617. 28). 26. Besemer, J., Gortz, G. & Charlier, D. (1980) Nucleic Acids Res. 8, 5825-5833. 27. Samulski, R. J., Berns, K. I., Tan, M. & Muzycka, N. (1982) We thank M. Zuker for his DNA secondary structure and align- Proc. Nail. Acad. Sci. USA 79, 2077-2081. ment computer programs and the statistical analysis of the produc- 28. Todd, P. A. & Glickman, B. W. (1982) Proc. Natl. Acad. Sci. tion of "complex" DNA structures in random DNA sequences. We USA 79, 4123-4127. Downloaded by guest on September 26, 2021