Phage N15 Telomere Resolution

The Escherichia coli prophage N15 exists as a linear DNA molecule with covalently closed ends. Purified N15 protelomerase TelN is the only protein required to convert circular DNA substrates to the linear form with hairpin termini. Within the center of the telomerase occupancy site tos, the target for TelN is the 56-bp telRL consisting of the central 22-bp palindrome telO and two 14-bp flanking inverted sequence repetitions. DNase I footprinting of TelN-telRL complexes shows a segment of ∼50 bp protected by TelN. Surface plasmon resonance studies demonstrate that this extended footprint is caused by two TelN molecules bound to telRL. Stable TelN-target DNA complexes are achieved with telRL; however, the additional sequences of tos stabilize the TelN-target complexes.TelO alone is not sufficient for specific stable complex formation. However, processing can occur, i.e. generation of the linear covalently closed DNA. Within the context oftelRL, sequences of telO are involved in specific TelN-telRL complex formation, in processing itself, and/or in recognition of the processing site. The sequence of the central (CG)3 within telO that is part of a 14-bp stretch proposed to have Z-DNA conformation is essential for processing but not for formation of specific TelN-telRLcomplexes. The concerted action of both TelN molecules at the target site is the basis for telomere resolution. Capturing of reaction intermediates demonstrates that TelN binds covalently to the 3′-phosphoryl of the cleaved strands.

N15 is a lambdoid, temperate Escherichia coli phage that has double-stranded DNA with cohesive ends packaged in the virion (1). The complete nucleotide sequence of the 46-kb genome is known (2). Upon infection of E. coli, the DNA circularizes via the cos ends. In contrast to other temperate phages, N15 does not integrate into the host genome. Lysogeny is established by processing the telomere resolution site telRL to form a linear genome with covalently closed, hairpin-like ends telL and telR (see "Discussion"). The proposed telomerase occupancy site tos (1) contains a series of inverted repeats centered on the 56-bp palindromic telRL. This palindrome or part of it could form a potential cruciform structure that may function as the real substrate of the telomerase TelN. In vivo replication of linear N15 plasmids requires three N15-derived components as follows: the origin of replication ori, the repli-cation initiation protein RepA, and TelN protein for resolution of the replicated telomeres (3). Ori probably is located immediately upstream of repA. repA is essential in N15 lytic replication and for maintenance of the linear prophage. Both lytic and lysogenic replication is independent of at least dnaA, dnaJ, dnaK, grpE, and recA (1,4,5). For telomere resolution, in vitro purified TelN and a telRL substrate are sufficient (6). In contrast to eukaryotic telomerases, TelN seems to be purely proteinaceous, lacking DNA synthesizing activity. The cleavingjoining activity of TelN resembles a specialized type of integrase that uses the same DNA molecule as donor and as acceptor.
Phage N15 belongs to the small group of systems known to replicate as linear DNA with hairpin ends. Such linear genomes are generally of eukaryotic origin, but a few from eubacteria are known: Yersinia bacteriophage PY54 1 ; a Klebsiella plasmid (7); the plasmids and genomes of Borrelia (8,9) and Agrobacterium tumefaciens (10,11); poxviruses (e.g. vaccinia virus (12)), iridopoxvirus (e.g. African swine fever virus (13)), and parvovirus (e.g. minute virus of mice (14)); and the genomes and certain plasmids of mitochondria and plastids (15)(16)(17). In none of these cases is replication clearly understood. However, several models for replication of linear DNA with hairpin ends exist (18). In one group, initiation of replication occurs at the hairpin ends by site-specific cleavage of one strand within the hairpin ends. The 3Ј-hydroxyls are then used as primers for strand displacement DNA synthesis. The other group contains an internal ori where -type replication initiates bidirectionally resulting in circular head-to-head and tail-to-tail dimers. These intermediates would then be processed by a cleaving-joining reaction, the telomere resolution, to regenerate covalently closed hairpin ends in the progeny molecules. Recently, telomere resolution was demonstrated in vivo for the linear Borrelia burgdorferi plasmid lp17 using telomere substrates (19). For phage N15, even in vivo circular dimeric replication intermediates, resolution of these molecules via TelN telomerase activity was described (3). This demonstrates that one completed round of replication results in two progeny molecules as is the case for most replication mechanisms except the rolling circle type.
N15 serves as a model system to study replication of linear DNA with hairpin ends that allows the use of E. coli extracts circumventing pathogenic organisms or cells that are difficult to grow. This is the first report describing the architecture of the complex of a telomere resolvase with the cognate telomere resolution site, the nucleotide sequence requirements for target recognition and processing, and the properties of a captured reaction intermediate. We unravel binding of the protelomerase TelN to the target using fragment retention on gels and * This work was supported by Grant LA 672/3-4 from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Overexpression and Purification of Proteins-TelN wild type protein was overproduced and purified using the procedure described previously (6). The same procedure was used to purify TelN Y424F to nearhomogeneity. In both cases fraction V (6) was used.
DNA Techniques-Standard techniques for DNA manipulation were used (21). The nucleotide sequence of recombinant DNA was verified by the dideoxynucleotide chain termination method (22).
Construction of Plasmids-By using synthetic oligonucleotides, we constructed the plasmids pJD105R-sym, pJD105A20TT21A, pJD105A-T-sym, pJD105(TA) 7 , pJD105(CG) 7 , pJD105S, pJD105⌬8, pJD105⌬8.1, pJD105⌬8.2 and pJD105⌬8.7, which all carry a telRL derivative (Table  II). The oligonucleotides were annealed to form the telRL region with the desired mutation or deletion and HindIII and BamHI restriction sites at the fragment ends. The resulting double-stranded fragment was inserted as HindIII-BamHI fragment into pBR329 that had been prepared accordingly (6). The pBR329 vector fragment was also religated without any insert, to form pBR329⌬1. The plasmid pJD101Y424F was constructed using PCR-based mutagenesis as described (23) with synthetic primers (Table II) and pJD101 as template. The construction of several plasmids used in this study was described previously (Table I).
The nucleotide sequence of all DNA fragments inserted into vector plasmids was confirmed (see above).
Transmission Electron Microscopy-For mapping the binding sites of TelN by electron microscopy, the plasmids containing various TelN targets, pJD104 (harboring tos), pJD105 (telRL), or pJD106 (telO) (6), were linearized with either AatII or AseI. In the case of pJD104 SphI was used instead of AseI. The linearized DNA was purified by electrophoresis on 0.7% agarose gels and subsequent gel extraction. 0.1 pmol of the plasmid was then incubated with 1-3 pmol of TelN Y424F protein in 10 l of buffer containing 20 mM Tris-HCl (pH 7.6), 25 mM potassium glutamate, 0.1 mM EDTA, 1 mM dithiothreitol for 30 min at 30°C. For direct adsorption to mica (24), the complexes were cross-linked for 10 min with 0.2% glutaraldehyde. 1 l was diluted 30-fold in 10 mM Tris-HCl (pH 7.6), 10 mM MgCl 2 . Micrographs were taken with a Philips EM400. DNA molecules with protein bound were traced on 16-fold enlarged negatives on a LM4 digitizer (Brü hl, Nü rnberg, Germany). The protein binding positions were determined as described (25).
Preparation of Radiolabeled DNA Fragments-The sources for the TelN target fragments were plasmids pJD104, pJD105, pJD105 A20TT21A, pJD106, and pBR329⌬1. The latter served as a control. Plasmids except pJD104 were digested with the restriction endonucleases DraI and AccI. pJD104 was digested by DraI, BamHI, and SphI. From each of the plasmids except pJD104 five fragments were obtained. The fragment labeled D in Fig. 3A contained the TelN target telRL. pJD104 yielded seven fragments, with fragment D (Fig. 3A) carrying tos. The fragments obtained were treated with shrimp alkaline phosphatase and 5Ј-labeled using T4 polynucleotide kinase and [␥-32 P]ATP as described (21).
Fragments for DNase I footprint analyses were prepared by PCR amplification of fragments containing telRL. To achieve selective labeling of only one strand, one of the two primers (Table II) was 32 P-5Јlabeled. Plasmids pJD105 or pJD105A20TT21A (Table I) served as templates for generating a 159-bp fragment (pJD105 base pairs 212-370) that carry the wild type telRL sequence or telRL A20TT21A, respectively. Reaction mixtures for PCR contained 50 pmol of each the labeled and the unlabeled primers (Table II), 0.05 pmol of template, and 2 units of Deep Vent polymerase. The PCR fragments were purified by phenol extraction and ethanol precipitation.
Fragment Retention on Gels-In 20 l of buffer containing 20 mM Tris-HCl (pH 7.6), 0.05 mg/ml bovine serum albumin, 5 mM MgCl 2 , 2-5 nM DNA fragments were incubated at 37°C for 30 min with 0 -5 M TelN Y424F. The reaction products were analyzed by electrophoresis on a 3.5% non-denaturing polyacrylamide gel as described previously (26). The gel was autoradiographed using the phosphor storage technology (27). The resulting image was documented using an Amersham Biosciences Storm PhosphorImager 820 and quantified by the use of the ImageQuant software version 5.0 (Amersham Biosciences).
DNase I Footprint Analysis-In 10 l of buffer containing 20 mM Tris-HCl (pH 7.6), 0.05 mg/ml bovine serum albumin, 80 M spermine, 5 mM CaCl 2 , 25 mM potassium glutamate, 0.1 mM EDTA, and 1 mM dithiothreitol, 5 pmol of PCR fragment was incubated with 0 -50 pmol of TelN protein for 30 min at 30°C. DNase I cleavage reactions were performed essentially as described (28,29). 1.5 l of DNase I (0.0015 units, RNase-free, Roche Molecular Biochemicals) was added to the TelN-DNA complexes. Following 10 min at 37°C, the DNase I was inactivated by adding 50 l of a mixture of 1% SDS, 200 mM NaCl, 20 mM EDTA (pH 8.0), 25 ng/l salmon sperm DNA. The reaction products were phenol-extracted and precipitated with ethanol containing 100 g/ml glycogen. The pellet was dissolved in 5 l of buffer (40 mM Tris-HCl (pH 8.8), 2 mM EDTA), mixed with 2 l of formamide/dye solution, heated for 3 min at 95°C, and loaded onto a 6% (w/v) polyacrylamide-urea sequencing gel (30). The cleavage pattern was visualized by autoradiography as described for the fragment retention assay. To generate a fragment pattern for the nucleotide sequence, dideoxy cycle sequencing was performed using the same conditions, templates, and radiolabeled primers as for the generation of the PCR fragments, except that Vent (exo Ϫ ) polymerase was applied as described (31).
Surface Plasmon Resonance Studies-Surface plasmon resonance was determined using a Biacore 2000 instrument (Biacore AB). Two DNA fragments were formed using synthetic oligonucleotides (Table II). One fragment consists of telRL within the original N15 nucleotide context, whereas the sequence of the other one was random. To differentiate between target-specific and nonspecific binding of TelN, 173 RU 2 of each fragment biotinylated at one end was immobilized at the streptavidin matrix in separate chambers of an SA sensor chip according to the manufacturer's recommendations. 250 l of TelN Y424F in various concentrations in buffer consisting of 20 mM Tris-HCl (pH 7.6), 5 mM MgCl 2 , 50 g/ml bovine serum albumin, 0.005% P20 surfactant (Biacore) with or without calf thymus DNA were injected for 375 s over the immobilized DNA. Then the complexes formed were washed with 120 l of the same buffer for 180 s to allow for dissociation. The flow rate always was 40 l/min. All runs were performed at 25°C. Changes in plasmon resonance correspond to a change in mass bound to the chip and are measured in resonance units (1 RU corresponds to 1 ng/mm 2 for proteins). Resonance response for protein is higher than that observed for DNA (32). For double-stranded DNA it was reported that RU protein ϭ 0.73 RU dsDNA (33). Therefore, the values measured were corrected using this factor. Data were evaluated according to Ref. 33 using the BIAevaluate software version 3.0 (Biacore AB).

RESULTS
The TelN-binding Site in tos Is telRL-Telomere resolution requires recognition and binding to the target DNA by TelN before processing via transesterification could take place. To study the recognition and binding step separated from telomere resolution, it was necessary to uncouple binding and transesterification. Hence, if not otherwise stated, for all protein-DNA binding studies a mutant TelN protein was used, in which the proposed active site Tyr-424 was replaced by a phenylalanine (6). TelNY424F has lost its transesterification activity but retained its specific DNA binding ability. To locate the binding site of TelN on the DNA, complex formation with three substrates, tos, telRL, and telO, was followed by electron microscopy. Tos (telomerase occupancy site) has a size of 310 bp and is defined by the inverted sequence repetitions R1/L1 and R2/L2 flanking telRL (1). The repeats R3/L3 enclose the palindromic telomere resolution site telO forming telRL (56 bp). The tos substrate used here consists of 563 bp including neighboring N15-derived base pairs (6) ( Table I). Knobs on the DNA indicate likely representative examples of complexes of TelN with tos ( Fig. 1). Plotting the positions of the knobs versus the contour length of the DNA yielded histograms demonstrating the specificity of binding. As indicated by the peak positions in the histograms, site-specific complex formation was detected with tos and telRL only (Fig. 2, A and B). The positions of the dots in TelN-telO complexes were scattered randomly along the DNA and did not map preferentially to a unique region showing that these complexes arose from non-sequence specific binding (Fig. 2C). From these data we conclude that tos has a unique binding site for TelN. Because telRL (56 bp) alone is sufficient for binding and the positions of the peaks on tos and telRL match perfectly, this binding site is telRL. TelRL consists of two inverted sequence repetitions, R3 and L3 (14 bp each), and the central palindrome telO (22 bp). The data on TelN-telRL ( Fig. 2B) and TelN-telO ( Fig. 2C) complexes show in addition that the sequences R3 and L3 are essential for specific complex formation, because telO alone failed to yield specific complexes. To learn more on the specificity of binding, the complex formation was studied by fragment retention in gel electrophoresis.
Tos Is the Most Efficient Binding Substrate for TelN-A mixture of equimolar amounts of restriction fragments was incubated with increasing amounts of TelNY424F and subsequently electrophoresed on a polyacrylamide gel. Representative images of the fragment retention experiments (Fig. 3A) show that the target fragment D containing tos or telRL was bound specifically by TelN, because in the presence of TelN the amount of fragment D at its original position in the gel was reduced. Simultaneously, a new band migrating much slower appeared just below the wells of the gel. This band most likely represents the newly formed protein-DNA complexes. No such specific retention of fragment D devoid of any N15 DNA was observed with the vector DNA (Fig. 3A, lanes m-o). Also for telO no specific binding was detected (Fig. 3A, lanes j-l) confirming the data obtained by electron microscopy (Fig. 2C). The clear indication for nonspecific binding in the telO and vector experiment is that the K d(app) values of fragment D with and without telO are identical (1300 nM; Fig. 3B and Table III). Furthermore, the largest band alters its position first, and upon increase of the protein amount each of the fragments migrates slower. To answer the question if binding of wild type and mutant TelN is comparable, we used a non-processed telRL derivative (telRLA20TT21A, Fig. 3A, lanes g-i). The almost identical dissociation constants of 290 and 310 nM ( Fig. 3B and Table III) demonstrate that both proteins TelN and TelNY424F bind telRLA20TT21A with similar affinities indicating that the TelN Tyr to Phe mutation does not influence binding dramatically. Interestingly, a more than 2-fold difference of the K d (app) values was observed in comparison to telRL. The TelN-telRL complex was apparently more stable than TelN/telRLA20TT-21A. Thus the difference in stability is probably because of the AT to TA mutation. The most stable complexes were formed with tos (K d(app) ϭ 28 nM). There is an almost 5-fold increase in stability of the TelN-tos complexes compared with telRL complexes (K d(app) ϭ 130 nM, Table III) suggesting that the additional invert sequence repetitions, R1/L1 and R2/L2 present in tos considerably contribute to this effect. It should be kept in mind that TelN has a high affinity to DNA in general, i.e. it binds DNA in a non-sequence-specific fashion. To obtain preliminary data on the protein-complex architecture we used DNase I footprinting.
TelN Covers a 50-bp Stretch of telRL-DNase I footprints on TelN-telRL complexes demonstrate that a large stretch of nu- cleotides is protected against nuclease attack on both strands (Fig. 4). The protection pattern suggests that ϳ50 bp are inaccessible to DNase I cleavage. The protection pattern shows one hypersensitive band on each strand. The footprint of the upper strand is divided in alternating stronger and weaker protected segments. This pattern is only rudimentary on the lower strand. Mapping the data onto the telRL sequence demonstrates a mirror symmetric arrangement of the hypersensitive bands (Fig. 5) TelN). B, equilibrium binding of tos, telRL, and telO to TelN Y424F. Fragment D from A was quantified as described under "Experimental Procedures." TelN concentrations from 0 to 2.5 M were used. For each of the plasmids, the values plotted were averaged from at least two independent experiments. The apparent equilibrium dissociation constant K d(app) for telRL A20TT21A was determined using both TelN wt protein and TelN Y424F, respectively.
to facilitate access of TelN to its target. A random DNA sequence of the same length served as reference for base-line adjustment. The highest value determined was 450 RU corresponding to 1.64 TelN molecules per telRL target (Fig. 6A). In the presence of competitor DNA the value decreased to 1.46 TelN molecules per telRL target (Fig. 6B). We interpret the data that average two molecules of TelN bound to one telRL molecule. To analyze which region in the target sequence de-termines TelN binding and processing specificity, we have generated mutant substrates.
Affinity of telRL for TelN Is Modulated by the Repeats R3/L3-Reduction in binding affinity was already observed when tos (Table III, K d(app) ϭ 28 nM) was truncated to telRL (K d(app) ϭ 130 nM) by the removal of R1/L1 and R2/L2. Symmetric stepwise reduction of the length of the 56-bp telRL sequence led to a gradual loss of the specific DNA binding ability. At the substrate size of 36-bp specific binding was no longer observed (data not shown). The complete removal of one arm (R3) of the invert sequence repetition also abolished the sequence-specific interaction of this truncated telRL substrate (Fig. 7,  pDB105⌬8). These observations indicate that the inverted repeats R1/L1, R2/L2, and also R3/L3 are responsible for increasing the binding affinity to the telomere resolution site telO. The introduction of symmetry into telRL (pJD105R-sym) outside of telO at the 4 bp (15, 17, 40, and 42, see

FIG. 5. Schematic representation of TelN-protected nucleotides of telRL.
The nucleotide sequence of telRL and the flanking regions is shown. Brackets denote nucleotides protected from DNase I attack. Vertical lines above and below the sequence indicate nucleotide positions whose 5Ј-phosphodiester bonds are hypersensitive to DNase I cleavage. The inverted repeat R3/L3 of telRL is marked by arrows, and the inner palindrome telO is indicated by a double arrow. Numbers correspond to the length of telRL (56 bp). The symmetry center is given by a vertical line within the sequence.
FIG. 6. Two TelN molecules bind specifically to telRL. TelN Y424F was passed over short stretches of double-stranded DNA containing either telRL or a random nucleotide sequence (Table II)  basis for the original asymmetry in the otherwise palindromic telRL sequence, does not have any significant influence on binding or processing (Fig. 7). Consequently, further analysis concentrated on the substrate properties of telO.
TelO Is Sufficient for Telomere Resolution-Binding of telO by TelN occurs at the level of nonspecific interactions (Fig. 2C, Table III, and Fig. 7). Therefore, no telomere resolution was detected at TelN concentrations applied for telRL processing. However, at TelN concentrations of ϳ50-fold higher than used for telRL, the telO substrate (pJD106) is processed weakly indicating that telO contains all sequence requirements needed for telomere resolution in vitro. Because of the lack of specific binding of telO the question of what is needed for telO recognition and processing in terms of nucleotide sequence was addressed by using mutations within telO in the context of telRL (Fig. 7).
TelO Is Essential for Specific Recognition and Processing--Binding and processing studies were initiated with telRLA-20TT21A, a target mutation containing an asymmetric sequence situation in telO (see above and Fig. 7). This derivative formed less stable complexes with wild type TelN than wild type telRL (Table III), and processing could not take place anymore (Fig. 7). These data already indicated that the sequence in the central palindrome is of major importance in telomere resolution. Restoring symmetry in the A20TT21A mutation, by introducing A36TT37A, reduced specific binding further indicating that telO apparently is involved in recognition and binding. Not surprisingly, the nucleotide sequence also determines whether processing could occur or not. The telO core sequence (20 bp) most likely lies between bases pairs 19 and 38 because derivative pJD105⌬8 still was processed but pJD105⌬8.1, 1 bp shorter on each side, was not. Replacement of part of the core sequence, proposed to have Z-DNA conformation (1) by a symmetric but otherwise random sequence (pJD105S , Table II) was inert in binding and processing. Extension of the central (CG) 3 to (CG) 7 by symmetric replacement of the flanking TATA region by CGCG also proved to behave inertly in binding, recognition, and processing (pJD105[CG] 7 ; see Fig. 7). Surprisingly, the replacement of the central (CG) 3 by TATATA (pJD105[TA] 7 ) retained full binding capacity, but processing of this derivative was considerably altered.
Processing of the circular (TA) 7 substrate by TelN yielded several diffuse bands upon agarose gel electrophoresis, the characters of which were not readily interpreted. Inspection of the sample by electron microscopy indicated that the majority of the DNA consisted of circular molecules decorated with at least one dot, each most likely representing a covalently attached TelN molecule. The TelN-DNA molecules may represent intermediates of the resolution reaction (data not shown). Processing of linear (TA) 7 substrate, i.e. linearized by cleavage with the restriction enzyme NruI prior to the TelN reaction, yielded two clear sharp bands of the expected size when treated with proteinase K before electrophoresis (data not shown). The finding indicated that processing of the (TA) 7 substrate occurred at or around the predicted telomere resolution site. The larger of the two fragments was selected for further analysis by end labeling.
TelN Binds Covalently to the 3Ј-Ends of the Telomere Resolution Site-Labeling of the 3Ј-end or the 5Ј-end demonstrates free accessibility, i.e. a protein bound to the respective end would prevent attachment of the label. Linearization of the (TA) 7 substrate pJD105(TA) 7 and subsequent processing with TelN yielded two fragments. The longer one (3239 bp) was isolated and 32 P-labeled at the 3Ј-ends or the 5Ј-ends. If TelN is attached to the 3Ј-end, only one side of this fragment is expected to be 3Ј-labeled, whereas both 5Ј-ends should carry the label (Fig. 8A). Further cleavage with a second restriction enzyme resulted in two fragments of 2227 and 1012 bp, only the longer one being detectable by autoradiography following separation by gel electrophoresis (Fig. 8B). 3Ј-Labeling showed that only one of the fragment ends carried the label, i.e. the other end was blocked most likely by covalent attachment of TelN (Fig. 8B). 5Ј-32 P-labeling indicated that both 5Ј-ends were accessible to the label (data not shown). These observations demonstrate that telomere resolution for the (TA) 7 substrate went as far as the first cleavage reaction that yielded the covalent adduct of DNA and TelN as the important reaction intermediate. No covalently closed hairpin ends were detected by alkaline gel electrophoresis (data not shown). Obviously this finding provides a way of dissecting the transesterification reactions in telomere resolution.
Our results demonstrate that the inverted sequence repeti- FIG. 7. Mutational analysis of sequence requirements of the TelN target telRL. The nucleotide sequence of telRL is shown. Nucleotide positions deviating from perfect 2-fold symmetry are given in gray letters. The inverted repeat R3/L3, the inner palindrome telO, and both arms telR and telL are indicated. Numbers correspond to the length of telRL (56 bp). telRL mutants are represented by bars, base pairs exchanged are in red letters within the bars. Supercoiled substrate (right column) was incubated with TelN and analyzed by agarose gel electrophoresis as described before (6). 0.15 pmol of each plasmid was treated with 100 pmol (7700 units) of TelN protein. Substrate processing was considered to take place, if any linear product could be observed (column P). TelN binding was considered to occur specifically, if TelN Y424F bound to the substrate with a K d(app) at least 2-fold below that observed for pBR329⌬1 (column B). ϩ or Ϫ, suitable or not suitable, respectively, for binding or processing. (ϩ), reduced amount of bound or processed substrate. tions R1/L1 to R3/L3 are required for formation and stabilization of specific TelN-target DNA complexes but are dispensable for DNA processing. The sequence of the central 6-bp segment of telO is essential for recognition of the DNA processing site and/or processing, but it apparently does not contribute to specific binding (pJD105[TA] 7 , Fig. 7). However, sequences flanking these central 6 bp add to the formation of specific TelN-target DNA complexes and are essential for recognition of the processing site. Therefore, the nucleotide sequence of telO determines (i) specific binding leading to stable complexes and (ii) recognition of the processing site where cleavage takes place.

DISCUSSION
The 56-bp telRL located within tos is the binding site for TelN, because the protein forms specific and stable complexes with telRL (Figs. 2 and 3). Because telO, the central 22-bp palindrome of telRL (Fig. 5), is no target for stable complex formation, the pair of inverted sequence repetitions R3/L3 flanking telO may provide essential recognition sequences for the protein. In line with this interpretation, it is also possible that R3/L3 alters the DNA geometry of telO slightly so extended protein-DNA contacts may result in stable complex formation. telO is located in the center of tos (310 bp) and flanked by three sets of inverted repeats R1/L1 to R3/L3 (1). The apparent dissociation constant determined for tos-TelN complexes is approximately 5-fold lower than obtained for telRL-TelN complexes ( Fig. 2A and Table III). How could this difference be explained?
The central problem that specific DNA-binding proteins have to solve is how to overcome the kinetic barrier of finding the cognate binding site among a large excess of nonspecific DNA. One solution to greatly facilitate this task is one-dimensional diffusion along the DNA until the target sequence is reached (34). The high unspecific DNA binding ability observed for TelN (Table III) supports the idea of binding to DNA randomly and then sliding along the double strand until the specific site has been recognized. In addition, the repeats R1/L1 and R2/L2 embedding telRL might serve as association points where the protein preferably but loosely binds and slides along the DNA until it hits telRL, where stable complex formation takes place.  These repeats located in the nearest vicinity of the telomere resolution site are unique on the 46-kb N15 chromosome, where tos occupies much less than 1%. Furthermore, the sets of sequence repetitions within tos may favor a locally limited structural alteration at telRL that consequently improves DNA-TelN interactions at the telomere resolution site to yield more stable complexes. If the stabilization of TelN-tos complexes we observed using a linear DNA fragment is actually caused by such conformational changes, telomere resolution in a supercoiled target may even be more favorable. Which of the two possibilities is most conceivable?
A convincing model for replication of linear genomes with covalently closed ends proposes the existence of circular headto-head (telLLЈ) and tail-to-tail (telRRЈ) dimers as replication intermediates. These replicated telomeres telLLЈ and telRRЈ are completely symmetrical. The circular intermediates would then be processed by a cleaving-joining reaction, the telomere resolution, to regenerate covalently closed hairpin ends in the progeny molecules (18,35). Recently telomere resolution using such a substrate containing the replicated telLLЈ was demonstrated in vivo for the linear B. burgdorferi plasmid lp17 (19). Ravin et al. (3) demonstrated that a linear N15 mini-plasmid with covalently closed ends replicates in vivo via a dimeric circular intermediate, which then TelN resolves into two monomeric progeny molecules. N15 lysogens strictly replicate as a linear plasmid. The current knowledge of the N15 lifecycle is summarized schematically in Fig. 9.
In the cell, the preferred conformation of circular doublestranded DNA is a negatively supercoiled form I (F I). In the last replication step telomere resolution at both telLLЈ and telRRЈ results in two single linear DNA molecules with covalently closed hairpin ends telR and telL. Therefore, it is likely that structural alteration in a superhelical substrate would allow establishment of a specific DNA-TelN complex with the most suitable geometry for the cleaving-joining reaction. Such alterations might consist in cruciform extrusions, establishment of Z-conformation, or local topological changes, for instance local unwinding.
The 56-bp telRL sequence contains 2 bp flanking the central 22-bp palindrome telO on each side, which interrupt the 2-fold rotational symmetry (see Fig. 5). This asymmetry allowing differentiation between telR and telL is obviously dispensable for telomere resolution in vivo for N15 (3) and for the Borrelia linear plasmid lp17 (19) and also in vitro for N15 (Fig. 7, see below). In line with the N15 telRL, telomeres from Borrelia (chromosome and linear plasmid) also contain such slight asymmetries (see e.g. Ref. 9), whereas the telomere resolution site of the Yersinia enterocolitica phage PY54 is completely symmetric. 1 Cleavage in telO occurs either in a staggered way or directly in the middle of the 22-bp sequence (Fig. 5), the latter option being rather unlikely for sterical reasons (6). Two TelN molecules are bound to telRL as demonstrated by using surface plasmon resonance (Fig. 6). Hence, one TelN molecule is supposed to bind to and occupy each arm of telRL, the region that contains both ends of the strands destined to be connected. The frictional ratio of 1.8 suggests an overall elliptic shape of TelN that may explain the extended footprint covering the complete telRL target site (Fig. 5). Telomere resolution within the 22-bp telO would not per se require TelN to cover additionally the flanking repeats R3/L3. If cleaving-joining occurs in a staggered way, each of the bound TelN molecules probably stabilizes the cleaved strands to fix the acceptor ends on the complementary strand for executing the joining reaction. However, even when we applied a telO containing plasmid substrate for the TelN-mediated telomere resolution under a variety of conditions, we never detected any intermediate. The cleavingjoining reaction proceeded fast to completion in the presence of either telRL or telO. Because the TelN concentration required for processing telO is about 50-fold higher than for telRL, we  hypothesize that in the cell the biological function of the repeats R3/L3 of telRL is to capture TelN to obtain the local enzyme concentration required for telomere resolution.
telO, the palindromic central part of telRL, contains the site for telomere resolution. Which portion of telO is required for the cleaving-joining reaction catalyzed by TelN? Analysis of telRL deletion derivatives indicate that nucleotides 19 -38 (see Fig. 7) are essential for TelN to bind the substrate, cleave both strands, form the hypothetical transient TelN-DNA intermediate which conserves part of the chemical energy, and complete the telomere resolution by the second transesterification yielding a linear molecule with covalently closed ends. Our results support the idea that one TelN molecule interacts with each half of telRL or telO indicative for a concerted action of both protein molecules in binding to the telomere resolution side and/or telomere resolution. However, the direct interaction of two TelN molecules on the target remains to be elucidated.
What are the main determinants of telO for telomere reso-lution? A telRL plasmid substrate in which the perfect symmetry of telO is disturbed by the exchange of two nucleotides at positions 20 and 21 (A20TT21A) was still able to complex specifically with TelN, albeit with reduced stability (Table III). However, telomere resolution was abolished completely (Fig.  7). Restoring symmetry by complementary nucleotide exchange in the second arm of telO reduced stability of the complexes and the ability to process remained abolished. Hence, not only the sequence is important but also substrate symmetry in this region of telO is an additional determinant for telomere resolution. The telRL derivative carrying a central (CG) 7 instead of TATA(CG) 3 TATA was no longer a substrate for binding or processing (Fig. 7). The TATA tetra-nucleotides therefore are recognized by TelN to locate the precise cleavage site or to position the enzyme in the appropriate geometry. Interestingly, replacement of (CG) 3 by (TA) 3 that yielded a (TA) 7 stretch retained wild type binding capacity, but processing of the previously linearized substrate (Fig. 8) was altered toward the accumulation of homogeneous reaction intermediates, i.e. cleaved substrate with covalently attached TelN. Hence it is clear that the central hexamer (CG) 3 plays its major role in substrate processing, especially in the second transesterification, because the first one being the cleavage reaction still occurs in (TA) 7 . The symmetrical intermediate carrying one TelN molecule at each half of the telomere resolution site further promotes the hypothesis of a concerted action of two TelN molecules (see above). The notion of the concerted action is further supported by the finding that reactions on substrate halves (pJD⌬8.7) or on substrates with one inactivated arm (pJD105A20TT21A) did not take place (Fig. 7).
In the (TA) 7 intermediate, TelN was bound to the 3Ј-end of the cleaved strands (Fig. 8), indicating that the enzyme exhibits mechanistic similarities to integrase and Flp recombinase. These enzymes belonging to the integrase family of site-specific recombinases (36) bind covalently to the 3Ј-phosphoryl of the cleaved strand via a phosphotyrosyl linkage (37,38). Three short motifs named box A, B, and C conserved in these proteins (36) are also present in TelN. The mutation Y424F in the proposed catalytic domain box C (VXXEXLGHXXXXTXXHYK) is inert in telomere resolution indicating that Tyr-424 might indeed function as the key residue in catalysis. As the study of reaction intermediates is often complicated by virtue of their transient and unstable nature, our (TA) 7 mutant will serve as a highly potent tool to address this question, because the intermediates could be isolated in sufficient amounts to determine the corresponding phosphoaminoacyl residue. The (TA) 7 substrate will also help to precisely locate between which nucleotides both strands are cleaved.
Because the TelNY424F mutant protein is inert in processing but retains the specific binding capacity, we suspect that the protein consists of distinct domains, one of them being responsible for DNA binding. For the reasons mentioned above TelN most likely functions as a specialized tyrosine recombinase for which donor and acceptor strands reside in the same DNA molecule. FIG. 9. Life cycles of phage N15. The phage injects its linear double-stranded DNA into the cell, where it circularizes via the cos ends. During lysogeny, TelN processes this circular genome at the telomere resolution site telRL. A stabile linear reaction intermediate obtained by using a mutant telRL derivative contains TelN covalently attached to each 3Ј-end as denoted in brackets. Joining of the ends results in the linear prophage with covalently closed hairpin ends telL and telR. This linear molecule replicates via dimers (3). The mechanism of induction toward lysis is unknown. Detailed knowledge of the lytic replication following infection is also missing.