A cruciform-dumbbell model for inverted dimer formation mediated by inverted repeats
A cruciform-dumbbell model for inverted dimer formation mediated by inverted repeatsChing-Tai Lin, Yi Lisa Lyu and Leroy F. Liu*
Department of Pharmacology, UMDNJ-Robert Wood Johnson Medical School, 675 Hoes Lane, Piscataway, NJ 08854, USA
Received April 29, 1997;Revised and Accepted June 22, 1997
ABSTRACT
Small inverted repeats (small palindromes) on plasmids have been shown to mediate a recombinational rearrangement event in Escherichia coli leading to the formation of inverted dimers (giant palindromes). This recombinational rearrangement event is efficient and independent of RecA and RecBCD. In this report, we propose a cruciform-dumbbell model to explain the inverted dimer formation mediated by inverted repeats. In this model, the inverted repeats promote the formation of a DNA cruciform which is processed by an endonuclease into a linear DNA with two hairpin loops at its ends. Upon DNA replication, this linear dumbbell-like DNA is then converted to the inverted dimer. In support of this model, linear dumbbell DNA molecules with unidirectional origin of DNA replication (ColE1 ori) have been constructed and shown to transform E.coli efficiently resulting in the formation of the inverted dimer. The ability of linear dumbbell DNA to transform E.coli suggests that the terminal loops may be important in bypassing the requirement of DNA supercoiling for initiation of replication of the ColE1 ori.
INTRODUCTION
In Escherichia coli, RecA is known to be central for homologous recombination (1 -3 ). However, recent studies have demonstrated efficient RecA-independent homologous recombination on both plasmids and chromosomes (4 -11 ). The RecA-independent homologous recombination is independent of the function of known recombination enzymes (9 ,11 ,12 ). Analysis of the RecA- independent recombination on plasmids has revealed complexity of the system. In addition to the expected deletion product of recombination between direct repeats, head-to-tail dimeric products have also been observed (6 ,8 -11 ). The frequency of formation of these various recombination products also depends on the distance separating the homologous DNA sequences (13 ) and the presence of DNA sequences (possibly serving as spacers between the direct repeats and an unknown cis-element on the plasmid) distant to the homologous sequences (4 ). Models of sister-strand exchange during DNA replication have been proposed to explain the formation of these dimeric products (6 ,9 ,11 ).
Recent studies of RecA-independent recombination between inverted repeats have also revealed the formation of a dimeric recombination product (5 ). The studies on recombination between inverted repeats were facilitated by construction of an HPH/tet cassette on pBR322. The HPH/tet cassette functions as a genetic switch controlling expression of the tet gene depending on the orientation of the P fragment (promoter-containing fragment). Recombination between the two inverted H fragments, which changes the orientation of the P fragment and thereby activates expression of the functional tet gene, can be readily monitored by tetracycline selection (5 ). Unlike the dimeric recombination products of direct repeats which are head-to-tail dimers with 1+2 and 1+3 structures (the numbers refer to the number of the repeat units on the head-to-tail dimer; e.g. 1+2 has a total of three repeat units with one repeat located diagonally from the other two tandem repeat units) (6 ,8 -11 ), the dimeric recombination product of inverted repeats is exclusively head-to-head with two pairs of giant inverted repeats, resembling certain gene amplification products in drug-resistant cells such as the double minute (DM) chromosomes in mammalian cells, the inverted dimer containing the DFR1 gene in yeast and the H-circles in Leishmania (14 -19 ).
Similar to the replication models proposed for direct repeat- mediated formation of dimeric recombination products, a reciprocal- strand-switching (RSS) model involving DNA replication has also been proposed for the formation of the inverted dimer from plasmids containing short inverted repeats (5 ). However, there has been no direct evidence supporting this model. In the present communication, we consider an alternative model (Fig. 1 ) which can also explain the formation of the head-to-head dimer satisfactorily. In this model, the inverted repeats are presumed to undergo a structural transition to form a DNA cruciform at a frequency depending on a variety of conditions known to favor this structural transition (20 ,21 ). The cruciform is then processed by an endonuclease which cuts diagonally at the Holliday junction. The resulting linear DNA, which is in the form of a dumbbell, can be further processed by replication to form the head-to-head dimer. The simplicity of this cruciform-dumbbell model has prompted us to test the aspects of this model using in vitro engineered dumbbell DNA containing the ColE1 ori. Theoretically, dumbbell DNA containing the ColE1 ori is not expected to replicate in E.coli for the following two reasons; first, ColE1 ori is known to require negative supercoiling for initiation of DNA replication (22 -25 ). The linear dumbbell DNA is not expected to be supercoiled by gyrase. Second, ColE1 ori is unidirectional (26 ,27 ) and only one arm of the dumbbell DNA is expected to be replicated through initiation at ColE1 ori. We show in the present communication that dumbbell DNA can transform E.coli efficiently. The sequences of the terminal loops appear to be important for transformation. The potential roles of the terminal loops in replication of linear dumbbell DNA are discussed.
MATERIALS AND METHODS
Enzymes and reagents
Klenow polymerase (large DNA polymerase fragment) was purchased from GIBCO-BRL. T4 DNA ligase was from NEB. Restriction enzymes were from several commercial sources. Escherichia coli DNA gyrase was a gift from Dr Martin Gellert (NIH, MD). Escherichia coli DNA topoisomerase I was a gift from Dr James C.Wang (Cambridge, MA).
Construction of dumbbell DNA in vitro
As shown in Figure 1 , the fragment containing the SV40 origin and the neomycin-resistant gene on pHPH-2 was obtained from BamHI digestion of pMAMneo (Clontech, CA). After polymerase fill-in, the blunted fragment was cloned into the SspI site of pHPH (5 ). Plasmid pHPH was derived from pBR322 and contained the HPH/tet inverted repeats cassette. As reported previously, the HPH/tet cassette, which consists of a flipped Ptet promoter fragment including part of the tet gene (the P fragment) flanked by inverted repeats (the two H fragments) can mediate efficient RecA-independent recombination/rearrangement resulting in the exclusive formation of a special inverted dimer (5 ). The HPH/tet cassette is basically a genetic switch controlling transcription of the functional tetracycline gene, depending on the orientation of P fragment. The inverted dimer, pID-IP, was generated by transforming pHPH-2 into E.coli DH5[alpha] (recA-) followed by selection with tetracycline. This inverted dimer contains a functional tetracycline gene due to the inversion of the flipped P fragment and therefore cells containing the inverted dimer can be readily selected for by resistance to tetracycline.
pID-IP* was constructed from pID-IP by destroying one of the two identical NdeI sites on the inverted dimer. This was accomplished by partial digestion with NdeI, followed by gel purification of the full length linear DNA, and religation after Klenow polymerase fill-in of the cohesive ends. The resulting plasmid, pID-IP*, therefore, contains only a single NdeI site. The dumbbell molecules were generated by linearization of 5 [mu]g of pID-IP* with NdeI, followed by alkaline denaturation (0.1 N NaOH) and rapid renaturation (neutralization with 0.1 N HCl plus 100 mM Tris-HCl pH 7.6) at 37oC for 10 min.
To flip the direction of one P fragment in pID-IP*, the pID-IP* DNA was digested with AatII (Fig. 6 ). Following phenol-CHCl3 extraction and ethanol precipitation, the digested DNA was ligated and transformed into E.coli DH5[alpha]. The clone with the P fragment flipped was identified by restriction enzyme analysis and designated pID-PP* (inverted dimer with parallel P fragments). Dumbbell DNA molecules derived from pID-PP* were prepared in the same manner as described above for pID-IP*.
Transforming dumbbell molecules into E.coli
Dumbbell molecules were digested with EcoRV to reduce the residual linear or supercoiled molecules and then transformed into E.coli DH5[alpha] by heat-shock at 42oC for 30 s following incubation of DNA with competent cells at 4oC for 30 min. The competent cells were made by incubation in 100 mM CaCl2 at 4oC for 20 min. Transformation frequency was obtained from the colony number after plating the transformation mixture on LB (10 g/l Bacto-tryptone, 5 g/l Bacto-yeast extract and 10 g/l NaCl) plates supplemented with 100 [mu]g/ml ampicillin. Plasmid DNAs isolated from transformants were analyzed by restriction enzyme digestion. The inverted dimers recovered from dumbbell transformants have both NdeI sites inactivated and are referred to as pID-IP** and pID-PP**, respectively.
Generation of supercoiled dumbbell DNA in vitro
The dumbbell DNA molecules derived from pID-IP* were circularized by renaturation at 65oC for 10 min. Circularization was achieved through an intramolecular interaction between the two complementary hairpin loops. As shown in Figure 5 A, the circularized dumbbell DNA (CDB) contained a two-base gap which was filled in by Klenow polymerase and ligase in vitro. The closed-circular dumbbell DNA (CCDB) was then treated with 40 U E.coli DNA gyrase and/or 5 U E.coli topoisomerase I ([omega] protein) in a reaction mixture (containing 100 mM KCl, 40 mM Tris pH 7.5, 10 mM MgCl2, 0.5 mM EDTA and 30 [mu]g/ml BSA) at 37oC for 60 min. The reaction was terminated by addition of Proteinase K and SDS (final concentrations of 200 mg/ml and 1%, respectively) and further incubated at 37oC for 15 min.
RESULTS
In vitro engineering of linear dumbbell DNA
Our strategy for preparing dumbbell DNA is schematically shown in Figure 2 A. We took advantage of the special head-to-head dimer (pID-IP) (inverted dimer with inverted P fragments) generated due to recombinational rearrangement of the HPH/tet cassette-containing plasmids. The pID-IP used in our current studies was isolated from tetracycline-resistant clones of pHPH-2, an HPH/tet cassette-containing plasmid (see Fig. 1 and Materials and Methods for details). In order to prepare dumbbell DNA, we eliminated one of the two NdeI sites on the pID-IP DNA (see Materials and Methods). The resulting plasmid, pID-IP*, was then converted to full-length linear DNA by NdeI digestion. Upon alkali denaturation and rapid renaturation, the majority of the linear pID-IP* DNA was converted to the dumbbell (DB) form as shown in Figure 2 A. Because the two HPH/tet cassettes on the pID-IP* DNA are in the inverted orientation, the two hairpin loops of the dumbbell DNA are expected to be complementary in their DNA sequences. The structure of dumbbell molecule was confirmed by restriction enzyme analysis (Fig. 2 B). PstI digestion of the linear pID-IP* DNA resulted in six bands, and five bands (two of them have twice the amount of DNA due to repeated DNA sequences) are expected based on the restriction map of NdeI linearized pID-IP* (Fig. 2 B, lane b). The extra band (marked with * to the left of the gel) in Figure 2 B was generated from residual supercoiled and nicked pID-IP*. The dumbbell DNA was expected to be digested by PstI into four fragments. Upon denaturation/renaturation, the presumed dumbbell DNA indeed gave only four major bands (Fig. 2 B, lane d). The presence of the hairpin loops at the ends of the dumbbell DNA was further confirmed by EcoRV digestion which was expected to cut within the double-stranded P segment of the HPH/tet cassette of pID-IP*. Since the P segment of the HPH cassette of the dumbbell DNA was located at the single-stranded hairpin loops, no digestion was expected. Indeed, as shown in Figure 2 B (compare lanes l and m), EcoRV did not digest the dumbbell DNA. As a positive control, EcoRV cut the NdeI-linearized pID-IP* DNA into three bands (Fig. 2 B, compare lanes j and k). Similarly, supercoiled pID-IP* DNA was digested by EcoRV into two bands under identical conditions (Fig. 2 B, compare lanes h and i). These experiments confirmed the structure of the dumbbell DNA.
Table1 . The transformation frequency of linear, dumbell and supercoiled DNA
Conditions
Coloniesa
Sc pID-IP*b
linear pID-IP*b
linear pID-PP*c
Control
-EcoRV
11 300 +- 70
400 +- 120
100 +- 50+
EcoRV
40 +- 10
2 +- 1
4 +- 3
D/R
-EcoRV
10 180 +- 210
1370 +- 360
200 +- 140
+EcoRV
370 +- 20
1020 +- 40
100 +- 40
Sc, supercoiled; D/R, denaturation/renaturation. aTransformation efficiency was the average from four independent experiments using 100 ng DNA. bThe structure of pID-IP* is shown in Figure 1A. Linear pID-IP* was obtained by digestion with NdeI. cThe structure of pID-PP* is shown in Figure 6A.
Dumbbell DNA can efficiently transform E.coli
DISCUSSION
The cruciform-dumbbell (CD) model for the formation of inverted dimers (giant palindromes) from plasmids containing short inverted repeats (small palindromes) (Fig. 1 ) is different from the proposed reciprocal-strand-switching (RSS) model (5 ) in the initial steps of forming the dumbbell-like intermediates. According to the RSS model, reciprocal switching of the leading and lagging strands at the repeated sequences during DNA replication, followed by cutting of the Holliday junction, results in the dumbbell-like replication intermediate. In the current CD model, we propose that a palindrome-cruciform transition, followed by cutting of the Holliday junction, results in the dumbbell-like DNA intermediate. In both models, replication of the dumbbell-like DNA intermediate leads to the formation of the inverted dimer (the giant palindrome). While the CD model appears to be at least equally attractive, there are two conceptually difficult steps; one is the formation of the cruciform from the inverted repeats, and the other is initiation of DNA replication from the linear dumbbell-like DNA intermediate. The present work is aimed at addressing the latter question.
In the RSS model, the dumbbell-like DNA intermediate is already pre-initiated. In the CD model, the dumbbell-like DNA is most likely not pre-initiated. This is based on the following considerations; if the palindrome/cruciform transition occurs on the pre-initiated (replicating) DNA molecules, the cruciforms are most likely located behind the replication forks (the positive supercoiling in front of the replication forks would prevent palindrome/cruciform transition). Cutting of the cruciforms located behind the replication forks eventually leads to the formation of linear dumbbell DNA which cannot be supercoiled by gyrase. It has been well documented that linear or relaxed DNA with the ColE1 origin of DNA replication cannot initiate DNA replication in vivo and in vitro (22 -25 ). Efficient initiation of DNA replication of the ColE1 origin requires a negatively supercoiled DNA template. It is therefore questionable whether the linear dumbbell-like DNA intermediate proposed in the cruciform-dumbbell model is able to initiate DNA replication. In order to test whether the linear dumbbell-like DNA can replicate in E.coli, we have engineered these dumbbell DNA molecules in vitro and used them for transformation studies. Surprisingly, while normal linear DNA does not transform E.coli, the linear dumbbell DNA (derived from pID-IP*) transforms E.coli ~9% as efficiently as supercoiled DNA (Table 1 ). It is unclear why the linear dumbbell DNA transforms E.coli efficiently. However, it appears that the DNA sequence and/or the complementarity of the two hairpin loops of the dumbbell DNA are important for efficient transformation (Table 1 , compare linear pID-IP* with linear pID-PP* following D/R). Based on our in vitro studies, the dumbbell DNA can circularize via intra-molecular loop-loop interaction forming a paranemic joint between the two complementary loops (Fig. 4 ). The circularized dumbbell DNA can also be negatively supercoiled by DNA gyrase in vitro following covalent closure of the two-base gap by polymerase/ligase (Fig. 5 ). It seems possible that these reactions can also readily occur in vivo to convert the dumbbell DNA into a supercoiled DNA for initiation of DNA replication. It is also possible that E.coli DNA topoisomerase I can further stabilize the supercoiled circular dumbbell DNA by converting the paranemic joint into a plectonemic joint. Such a conversion (paranemic to plectonemic conversion) is not expected to affect segregation of replicated DNA since the plectonemic joint can be converted into double-stranded interlocks and further resolved by topoisomerase IV and to a lesser extent by DNA gyrase (28 -34 ).
The dumbbell DNA prepared from pID-PP* contains two identical (non-complementary) hairpin loops and thus cannot undergo intramolecular circularization. As expected, the transformation efficiency is ~10 times lower. We have also constructed a plasmid, pID-PP'*, in which the two P fragments have completely different sequences. Again, the dumbbell DNA prepared this plasmid had a seven times lower transformation efficiency (unpublished results). While these results support the supercoiling explanation for efficient dumbbell transformation, we were puzzled by the rather high residual transformation frequency of these dumbbell DNA molecules derived from pID-PP* which cannot circularize. Furthermore, analysis of the transformants has revealed an abundance of a tetrameric plasmid DNA product (0% for pID-IP, 20% for pID-PP* and 35% for pID-PP'*).
While the mechanism for the formation of the tetrameric plasmid remains unclear, we can offer two possible explanations for the residual transformation efficiency of dumbbell DNA with non-complementary or different loops. These dumbbell DNAs can undergo intermolecular loop-loop interactions to form pseudo-supercoiled DNA substrates for initiation of DNA replication. Alternatively, the single-stranded loops can initiate DNA replication thereby bypassing the requirement of negative supercoiling. Priming of single-stranded DNA, especially at the regions with stem-loops, has been well documented (35 -37 ). At present, we are uncertain about the exact mode of initiation of DNA replication by the dumbbell DNA. However, the ability of dumbbell DNA to transform E.coli provides partial support to the proposed CD model for the formation of giant palindromes from small palindromes in E.coli. The CD model may also be applied to explain the formation of inverted dimers (giant palindromes) during gene amplification in eukaryotic systems (14 ,16 -19 , 38 -45 ). Further studies are necessary to verify aspects of the proposed model.
ACKNOWLEDGEMENTS
We would like to thank Dr Carton W.Chen for helpful discussions and Mr Pu Duann and Ms Maarika Liivak for their help in various stages of this work. This work was supported by NIH grant GM27731.
REFERENCES
1 Kowalczykowski, S. C., Dixon, D. A., Eggleston, A. K., Lauder, S. D. and Rehrauer, W. M. (1994) Microbiol. Rev., 58, 401-465.
17 Huang, T. and Campbell, J. L. (1995) J. Biol. Chem., 270, 9607-9614.MEDLINE Abstract
18 Ouellette, M., Hettema, E., Wust, D., Fase-Fowler, F. and Borst, P. (1991) EMBO J., 10, 1009-1016.MEDLINE Abstract
19 White, T. C., Fase-Fowler, F., van Luenen, H., Calafat, J. and Borst, P. (1988) J. Biol. Chem., 263, 16977-16983.
20 Lilley, D. M., Gough, G. W., Hallam, L. R. and Sullivan, K. M. (1985) Biochimie, 67, 697-706.
21 Bowater, R. P., Chen, D. and Lilley, D. M. (1994) Biochemistry, 33, 9266-9275.
22 Marians, K. J. (1992) Annu. Rev. Biochem., 61, 673-719.
23 Marians, K. J., Minden, J. S. and Parada, C. (1986) Prog. Nucleic Acids Res. Mol. Biol., 33, 111-140.
24 Ogasawara, N., Seiki, M. and Yoshikawa, H. (1979) Nature, 281, 702-704.MEDLINE Abstract
25 Wood, D. C. and Lebowitz, J. (1984) J. Biol. Chem., 259, 11184-111847.
26 Lovett, M. A., Katz, L. and Helinski, D. R. (1974) Nature, 251, 337-340.
27 Martin-Parras, L., Hernandez, P., Martinez-Robles, M. L. and Schvartzman, J. B. (1991) J. Mol. Biol., 220, 843-853.MEDLINE Abstract
28 Adams, D. E., Shekhtman, E. M., Zechiedrich, E. L., Schmid, M. B. and Cozzarelli, N. R. (1992) Cell, 71, 277-288.
29 Hiasa, H. and Marians, K. J. (1994) J. Biol. Chem., 269, 16371-16375.MEDLINE Abstract
30 Minden, J. S. and Marians, K. J. (1985) J. Biol. Chem., 260, 9316-9325.
31 Minden, J. S. and Marians, K. J. (1986) J. Biol. Chem., 261, 11906-11917.
32 Peng, H. and Marians, K. J. (1995) J. Biol. Chem., 270, 25286-25290.MEDLINE Abstract
33 Zechiedrich, E. L. and Cozzarelli, N. R. (1995) Genes Dev., 9, 2859-2869.
34 Peng, H. and Marians, K. J. (1993) Proc. Natl. Acad. Sci. USA, 90, 8571-8575.MEDLINE Abstract
35 Harrington, C. and Perrino, F. W. (1995) Nucleic Acids Res., 23, 1003-1009.MEDLINE Abstract
36 Langer, U. and Alonso, J. C. (1994) FEMS Microbiol. Lett., 119, 123-128.MEDLINE Abstract
37 Miao, D. M., Honda, Y., Tanaka, K., Higashi, A., Nakamura, T., Taguchi, Y., Sakai, H., Komano, T. and Bagdasarian, M. (1993) Nucleic Acids Res., 21, 4900-4903.
38 Ouellette, M. and Borst, P. (1991) Res. Microbiol., 142, 737-746.MEDLINE Abstract
39 Hyrien, O., Debatisse, M., Buttin, G. and de Saint Vincent, B. R. (1988) EMBO J., 7, 407-417.MEDLINE Abstract
40 Passananti, C., Davies, B., Ford, M. and Fried, M. (1987) EMBO J., 6, 1697-1703.MEDLINE Abstract
41 Ruiz, J. C. and Wahl, G. M. (1988) Mol. Cell. Biol., 8, 4302-4313.
42 Ruiz, J. C. and Wahl, G. M. (1990) Mol. Cell. Biol., 10, 3056-3066.
43 Butler, D. K., Yasuda, L. E. and Yao, M. C. (1995) Mol. Cell. Biol., 15, 7117-7126.
44 Yasuda, L. F. and Yao, M. C. (1991) Cell, 67, 505-516.
45 Butler, D. K., Yasuda, L. E. and Yao, M. C. (1996) Cell, 87, 1115-1122.
*To whom correspondence should be addressed. Tel: +1 732 235 4592; Fax: +1 732 235 4073; Email: lliu@umdnj.edu
