Proteolysis patterns of epitopically labeled yeast DNA topoisomerase II suggest an allosteric transition in the enzyme induced by ATP binding.

A cloned yeast TOP2 gene was modified to produce yeast DNA topoisomerase II (EC 5.99.1.3) epitopically labeled at its amino or carboxyl terminus. Limited digestion with SV8 endoprotease shows three distinct protease-sensitive sites in each polypeptide of the dimeric enzyme. These sites were mapped by immunostaining of the end-labeled proteolytic fragments resolved by SDS/polyacrylamide gel electrophoresis; two of the mapped locations were confirmed by sequencing the amino ends of two unlabeled peptic fragments. Proteolytic cleavage by SV8 endoprotease at a pair of sites corresponding to the carboxyl sides of Glu-411 and Glu-680 is modulated by the binding of the nonhydrolyzable ATP analogs adenosine 5'-[beta, gamma-imido]triphosphate (5'-adenylyl imidodiphosphate) and adenosine 5'-[gamma-thio]triphosphate: in their absence cleavage occurs predominantly at Glu-411; in the presence of either analog, cleavage occurs predominantly at Glu-680. These results are interpreted in terms of allosteric interdomainal movements in the type II DNA topoisomerase following the binding of ATP.     

Intercalators promote the binding of RecA protein to double-stranded DNA

Ethidium bromide, acridine orange, 4'-(9-acridinylamino)methanesulfon-o-anisidide (o-AMSA), and m-AMSA induce the rapid binding of RecA protein to double-stranded (ds) DNA. The filaments formed appear to retain the drug and are 12.8 nm in diameter with an 8.0-nm pitch. Two classes of drugs have been distinguished: (i) those that bind to RecA protein and induce assembly at low relative concentrations (e.g., ethidium bromide) and (ii) those that do not appear to interact directly with RecA protein and must be present at relatively high drug concentrations to stimulate assembly (e.g., m-AMSA). Ethidium bromide, acridine orange, and quinacrine inhibit RecA protein binding to single-stranded DNA. Addition of ATP to the drug-induced filaments causes the protein to rapidly dissociate from dsDNA, and protein binding to dsDNA diminishes upon extended exposure to room light. We suggest that the structure of the drug-induced filaments may be more typical of the complex that initiates RecA protein assembly along DNA rather than the product of extensive polymerization as induced by adenosine 5'-[gamma-thio]triphosphate.     

ATP-dependent association of nuclear proteins with isolated rat liver nuclei.

In vitro association of Xenopus nucleoplasmin and mammalian nonhistone chromosomal high mobility group 1 (HMG1) protein with nuclei isolated from rat liver was examined. Efficient association of nuclear proteins with isolated nuclei requires ATP, HCO3-, and Ca2+. Association occurred at 33 degrees C but not at 4 degrees C. ATP could be replaced by adenosine 5'-[alpha,beta-methylene]triphosphate (pp[CH2]pA), a nonhydrolyzable ATP analog. pp[CH2]pA associated with nuclei at 33 degrees C and nucleoplasmin and HMG1 rapidly associated with the pp[CH2]pA-bound nuclei at 4 degrees C. Competition studies showed that these associations at both 33 degrees C and 4 degrees C were specific. More than 80% of the bindings of nuclear proteins to the nuclear surface were blocked by wheat germ agglutinin.     

The synapsis event in the homologous pairing of DNAs: RecA recognizes and pairs less than one helical repeat of DNA.

A key step in homologous recombination is the alignment and pairing of homologous DNAs. The Escherichia coli RecA protein initiates pairing by binding to single-strand DNA, forming a helical nucleoprotein filament. We demonstrate that in the presence of the nonhydrolyzable ATP analogue adenosine 5'-[gamma-thio]triphosphate and ADP, RecA can pair a homologous oligonucleotide 15 bases long with a duplex DNA to yield synaptic complexes consisting of the oligonucleotide and duplex DNA stabilized by RecA. RecA can pair as few as eight bases of homology to form such synaptic complexes. The homologous DNAs remain paired to each other upon removal of RecA provided that the length of shared homology is at least 26 base pairs. Based on our findings and the work of others, we propose that in vitro, one helical turn of a RecA nucleoprotein filament containing approximately six RecA monomers and 15 bases of single-strand DNA is the functional unit sufficient to carry out the homology search.     

Nucleotide-stimulated proteolysis of histone H1.

We have identified a proteolytic system that selectively degrades histone H1 in normal human lymphocytes. Treatment of permeabilized human lymphocytes with a series of nucleotides produced a marked decrease in their histone H1 content compared to untreated cells. The nucleotide-stimulated process was selective for histone H1 because gel electrophoresis showed that almost all other lymphocyte protein bands remained constant while histone H1 disappeared. The elimination of histone H1 appears to be the result of proteolysis by a trypsin-like enzyme because it was inhibited by phenylmethylsulfonyl fluoride, antipain, soybean trypsin inhibitor, and diisopropyl fluorophosphate. Proteolysis was stimulated by P1,P4-di(adenosine-5') tetraphosphate, P1,P3-di(adenosine-5') triphosphate, P1,P5-di(adenosine-5') pentaphosphate, adenosine 5'-tetraphosphate, ATP, adenosine 5'-[alpha, beta-methylene]triphosphate, adenosine 5'-[beta, gamma-methylene]triphosphate, ADP, CTP, GTP, UTP, dATP, or pyrophosphate, whereas AMP, adenosine, adenosine diphosphoribose, NAD+, cAMP, or sodium phosphate did not show this stimulation of proteolysis. ATP, [alpha, beta-methylene]ATP, [beta, gamma-methylene]ATP, and pyrophosphate all stimulated proteolysis, suggesting that a pyrophosphate linkage was necessary for this process. Thus, resting human lymphocytes contain a trypsin-like protease that is stimulated by nucleotides or pyrophosphate to selectively degrade histone H1.     

G protein-mediated inhibition of myosin light-chain phosphatase in vascular smooth muscle.

The mechanism of G protein-mediated sensitization of the contractile apparatus of smooth muscle to Ca2+ was studied in receptor-coupled alpha-toxin-permeabilized rabbit portal vein smooth muscle. To test the hypothesis that Ca2+ sensitization is due to inhibition of myosin light-chain (MLC) phosphatase activity, we measured the effect of guanosine 5'-[gamma-thio]triphosphate and phenylephrine on the rate of MLC dephosphorylation in muscles preactivated with Ca2+ and incubated in Ca(2+)- and ATP-free solution containing 1-(5-chloronaphthalene-1-sulfonyl)-1H-hexahydro-1,4-diazepine (ML-9) to block MLC kinase activity. Guanosine 5'-[gamma-thio]triphosphate alone (300 microM) or in combination (3 microM) with phenylephrine decreased the rates of relaxation and dephosphorylation of MLC to about half of control values; this inhibition is sufficient to account for maximal G protein-mediated Ca2+ sensitization of MLC phosphorylation. The rate of thiophosphorylation of MLC with adenosine 5'-[gamma-thio]-triphosphate was not affected by guanosine 5'-[gamma-thio]triphosphate. We suggest that inhibition of protein phosphatase(s) by G protein(s) may have important regulatory functions.     

Nucleoside triphosphate-dependent DNA-binding properties of mos protein.

We have previously shown that the mos gene product, p40mos, produced in Escherichia coli binds ATP and has ATPase activity. In the present study, we investigated the DNA-binding properties of p40mos and two mos deletion mutant proteins. Nitrocellulose blot protein-DNA binding assays showed that p40mos binds DNA in the presence of Mg2+-ATP and certain other nucleoside triphosphates. Ninety percent of the p40mos-bound DNA is dissociated if the complex is washed in the presence of 1 M NaCl or in the absence of ATP. p40mos-DNA binding is not observed in the presence of AMP or the nonhydrolyzable ATP analog adenosine 5'-[beta, gamma-methylene]-triphosphate; however, in the presence of ADP, p40mos binds DNA at 20% of the level that is observed with ATP. An N-terminal-deletion mutant protein, p19mos, has no DNA-binding activity, whereas a C-terminal-deletion mutant protein, p25mos, does. p25mos contains the ATP-binding domain, binds DNA in the presence of either ADP or ATP, and shows 5% and 45% binding (relative to that in the presence of ATP) in the presence of AMP and adenosine 5'-[beta, gamma-methylene]triphosphate, respectively. These results suggest that the N-terminal domain of p40mos is responsible for nucleoside triphosphate-mediated DNA binding. We also observed differential histone-DNA binding in the presence and absence of ATP.     

Effects of adenine nucleotides on the proliferation of aortic endothelial cells.

The effects of adenine nucleotides and adenosine on DNA synthesis and cell growth have been studied in bovine aortic endothelial cells (BAECs). ATP produced a small but significant (+44%) increase of the fraction of BAECs whose nuclei are labeled by [3H]thymidine. This mitogenic effect was mimicked by ADP, the phosphorothioate analogues ATP gamma S and ADP beta S, and the nonhydrolyzable analogue adenosine 5'-(beta, gamma-imido)triphosphate (APPNP), whereas adenosine 5'-(alpha, beta-methylene)triphosphate (APCPP), a selective agonist of P2x-purinoceptors, had no effect at 10 microM and a small one at 100 microM; this profile is consistent with the involvement of P2y-receptors. Adenosine induced a mitogenic response of a magnitude similar to that of ATP. This effect was not reproduced by R-phenylisopropyl adenosine, by 5'-N-ethylcarboxamide adenosine, or by 2',5'-dideoxyadenosine, selective ligands of the A1- and A2-receptors and the P site, respectively, nor was it inhibited by 8-phenyltheophylline, an antagonist of both A1- and A2-receptors. The mechanism of this adenosine action thus remains unclear. ATP and ATP gamma S did not enhance the proliferation of BAECs cultured in the presence of fetal calf serum concentrations ranging from 0.5% to 10%. They inhibited the growth-promoting effect of basic fibroblast growth factor; among the various nucleotides tested, APCPP was the least effective to reproduce the action of ATP, suggesting the possible involvement of P2y-receptors.(ABSTRACT TRUNCATED AT 250 WORDS)     

Interaction of Escherichia coli RuvA and RuvB proteins with synthetic Holliday junctions.

The RuvA, RuvB, and RuvC proteins of Escherichia coli are required for the recombinational repair of ultraviolet light- or chemical-induced DNA damage. In vitro, RuvC protein interacts with Holliday junctions in DNA and promotes their resolution by endonucleolytic cleavage. In this paper, we investigate the interaction of RuvA and RuvB proteins with model Holliday junctions. Using band-shift assays, we show that RuvA binds synthetic Holliday structures to form specific protein-DNA complexes. Moreover, in the presence of ATP, the RuvA and RuvB proteins act in concert to promote dissociation of the synthetic Holliday structures. The dissociation reaction requires both RuvA and RuvB and a nucleotide cofactor (ATP or dATP) and is rapid (40% of DNA molecules dissociate within 1 min). The reaction does not occur when ATP is replaced by either ADP or the nonhydrolyzable analog of ATP, adenosine 5'-[gamma-thio]triphosphate. We suggest that the RuvA and RuvB proteins play a specific role in the branch migration of Holliday junctions during postreplication repair of DNA damage in E. coli.     

Phosphorylation of a high molecular weight DNA polymerase alpha.

Anti-human DNA polymerase alpha murine IgG SJK-287-38 [Tanaka, S., Hu, S.-Z., Wang, T. S.-F. & Korn, D. (1982) J. Biol. Chem. 257, 8386-8390] neutralized DNA polymerase alpha activity from rat embryonic fibroblasts infected with a temperature-sensitive transformation mutant of Rous sarcoma virus (tsLA24). After centrifugation of a crude cytosol fraction from log-phase cells in a 5-20% linear sucrose gradient, polypeptides of Mr approximately equal to 185,000 and 220,000 were immunoprecipitated only from gradient fractions containing DNA polymerase alpha activity. When similar cultures were incubated in medium containing [32P]orthophosphate, it was found that the Mr 220,000 protein was phosphorylated but that the other peptides specific for polymerase alpha activity did not contain detectable amounts of phosphate. Phospho amino acid analysis of the high molecular weight immunoprecipitable proteins indicated that the labeled amino acid was phosphoserine. Incubation of 2.5 units of crude DNA polymerase alpha with 4 units of agarose-immobilized alkaline phosphatase resulted in a nearly complete inhibition of DNA polymerase alpha activity. Subsequent incubation of this preparation with 5 or 50 microM ATP, but not the nonhydrolyzable analog adenosine 5'-[gamma-thio]triphosphate, restored the in vitro DNA polymerizing activity. These results demonstrate that a high molecular weight DNA polymerase alpha (Mr approximately equal to 220,000) is phosphorylated in cultured cells and that this protein is a substrate for a serine kinase rather than the tyrosine-specific protein kinase of Rous sarcoma virus. The results suggest that phosphorylation/dephosphorylation reactions modulate the activity of this polymerase.     

Extracellular ATP is a mitogen for 3T3, 3T6, and A431 cells and acts synergistically with other growth factors.

Extracellular ATP in concentrations of 5-50 microM displayed very little mitogenic activity by itself but it caused synergistic stimulation of [3H]thymidine incorporation in the presence of phorbol 12-tetradecanoate 13-acetate, epidermal growth factor, platelet-derived growth factor, insulin, adenosine, or 5'-(N-ethyl)carboxamidoadenosine. Cultures of Swiss 3T3, Swiss 3T6, A431, DDT1-MF2, and HFF cells were used. The percent of cell nuclei labeled with [3H]thymidine and cell number were also increased. ADP was equally mitogenic, while UTP and ITP were much less active. The effect of ATP was not due to hydrolysis by ectoenzymes to form adenosine, a known growth factor. Thus, the nonhydrolyzable analogue adenosine 5'-[beta, gamma-imido]triphosphate was mitogenic. In addition, it was found that ATP showed synergism in 3T6 and 3T3 cells when present for only the first hour of an incorporation assay, during which time no significant hydrolysis occurred. Furthermore, prolonged preincubation of cells with ATP reduced the mitogenic response to ATP but not to adenosine; preincubation with adenosine or N6-(R-phenylisopropyl)adenosine had the reverse effect. Finally, the effect of adenosine, but not of ATP, was inhibited by aminophylline. We conclude that extracellular ATP is a mitogen that interacts with P2 purinoceptors on the plasma membrane.     

recA protein of Escherichia coli promotes branch migration, a kinetically distinct phase of DNA strand exchange.

The recA protein of Escherichia coli promotes the complete exchange of strands between full-length linear duplex and single-stranded circular DNA molecules of bacteriophage phi X-174, converting more than 50% of the single-stranded DNA into heteroduplex replicative form II-like structures. Kinetically, the reaction can be divided into two phases, formation of short heteroduplex regions (D loops) and extension of the D loops via branch migration. recA protein participates directly in both phases. D loops are formed efficiently in the presence of ATP or the nonhydrolyzable ATP analog adenosine 5'-[gamma-thio]triphosphate, whereas D-loop extension requires continuous ATP hydrolysis. Complete strand exchange requires a stoichiometric amount of recA protein and is strongly stimulated by the single-stranded-DNA-binding protein of E. coli.     

Reconstitution of rat brain mu opioid receptors with purified guanine nucleotide-binding regulatory proteins, Gi and Go.

Reconstitution of purified mu opioid receptors with purified guanine nucleotide-binding regulatory proteins (G proteins) was investigated. mu opioid receptors were purified by 6-succinylmorphine AF-AminoTOYOPEARL 650M affinity chromatography and by PBE isoelectric chromatography. The purified mu opioid receptor (pI 5.6) migrated as a single Mr 58,000 polypeptide by NaDodSO4/PAGE, a value identical to that obtained by affinity cross-linking purified mu receptors. When purified mu receptors were reconstituted with purified Gi, the G protein that mediates the inhibition of adenylate cyclase, the displacement of [3H]naloxone (a mu opioid antagonist) binding by [D-Ala2,MePhe4,Gly-ol5]enkephalin (a mu opioid agonist) was increased 215-fold; this increase was abolished by adding 100 microM (guanosine 5'-[gamma-thio]triphosphate. Similar increases in agonist displacement of [3H]naloxone binding (33-fold) and its abolition by guanosine 5'-[gamma-thio]triphosphate were observed with Go, the G protein of unknown function, but not with the v-Ki-ras protein p21. In reconstituted preparations with Gi or Go, neither [D-Pen2,D-Pen5]enkephalin (a delta opioid agonist; where Pen is penicillamine) nor U-69,593 (a kappa opioid agonist) showed displacement of the [3H]naloxone binding. In addition, the mu agonist stimulated both [3H]guanosine 5'-[beta,gamma-imido]triphosphate binding (in exchange for GDP) and the low-Km GTPase in such reconstituted preparations, with Gi and Go but not with the v-Ki-ras protein p21, in a naloxone-reversible manner. The stoichiometry was such that the stimulation of 1 mol of mu receptor led to the binding of [3H]guanosine 5'-[beta,gamma-imido]triphosphate to 2.5 mol of Gi or to 1.37 mol of Go. These results suggest that the purified mu opioid receptor is functionally coupled to Gi and Go in the reconstituted phospholipid vesicles.     

Assembly of biologically active proheads of bacteriophage lambda in vitro.

Bacteriophage lambda DNA can be packaged in vitro into preformed proheads to generate plaque-forming units. This complex set of reactions is initiated when lambda DNA is mixed with the product of the phage A gene, and proheads. Because proheads are an essential early reactant, the system has potential as an assay for the formation of biologically active proheads. When extracts of cells infected with certain lambda head mutants (for example, B--, C--, Nu3--, and E--) are used as the prohead donor, plaque-forming units are not produced. However, when extracts of E- - and Nu3- - infected cells are first reacted together the combination provides prohead-donor activity to the in vitro packaging system. In vitro assembled, biologically active proheads have the same sedimentation properties and electron micrsocopic appearance as "wild-type" proheads isolated from lambdaA-D- -infected cells. Centrifugation analysis shows that the Nu3- extract contributes gpE, the major capsid protein, to the reaction in the form of monomers or small polymers.     

RecA.oligonucleotide filaments bind in the minor groove of double-stranded DNA.

Escherichia coli RecA protein, in the presence of ATP or its analog adenosine 5'-[gamma-thio]triphosphate, polymerizes on single-stranded DNA to form nucleoprotein filaments that can then bind to homologous sequences on duplex DNA. The three-stranded joint molecule formed as a result of this binding event is a key intermediate in general recombination. We have used affinity cleavage to examine this three-stranded joint by incorporating a single thymidine-EDTA.Fe (T*) into the oligonucleotide part of the filament. Our analysis of the cleavage patterns from the joint molecule reveals that the nucleoprotein filament binds in the minor groove of an extended Watson-Crick duplex.     

Actin filaments mediate Dictyostelium myosin assembly in vitro.

Because myosin thick filaments form in the actin-rich cortex of nonmuscle cells, we have examined the role of Dictyostelium actin filaments in the assembly of Dictyostelium myosin (type II). Fluorescence energy transfer and light-scattering assembly assays indicate that self-association of Dictyostelium myosin into bipolar thick filaments is kinetically regulated by actin filament networks. Regulation is nucleotide dependent but does not require ATP hydrolysis. Myosin assembly is accelerated approximately 5-fold by actin filaments when either 1 mM ATP or 1 mM adenosine 5'-[beta,gamma-imido]triphosphate (AMP-P[NH]P) is present. However, actin filaments together with 1 mM ADP abolish myosin assembly. Accelerated assembly appears to require transient binding of myosin molecules to actin filaments before incorporation into thick filaments. Fluorescence energy-transfer assays demonstrate that myosin associates with actin filaments at a rate that is equivalent to the accelerated myosin assembly rate, evidence that myosin to actin binding is a rate-limiting step in accelerated thick filament formation. Actin filament networks are also implicated in regulation of thick filament formation, since fragmentation of F-actin networks by severin causes immediate cessation of accelerated myosin assembly. Electron microscopic studies support a model of actin filament-mediated myosin assembly. In ADP, myosin monomers rapidly decorate F-actin, preventing extensive formation of thick filaments. In AMP-P[NH]P, myosin assembles along actin filaments, forming structures that resemble primitive stress fibers. Taken together, these data suggest a model in which site-directed assembly of thick filaments in Dictyostelium is mediated by the interaction of myosin monomers with cortical actin filament networks.     

ADP-ribosylation of the Mr 83,000 stress-inducible and glucose-regulated protein in avian and mammalian cells: modulation by heat shock and glucose starvation.

ADP-ribosylation of proteins was analyzed by in vivo labeling of cells with [3H]adenosine, followed by separation of their protein components by two-dimensional isoelectric focusing/NaDodSO4 polyacrylamide gel electrophoresis. We show here that in several cell types of avian and mammalian origin the major [34H]adenosine acceptor in vivo is a polypeptide with a Mr of 83,000 and isoelectric point of approximately equal to 5.3. This polypeptide is identical to one of the stress-inducible and glucose-regulated proteins (here called SP83) previously described in avian and mammalian cells. Snake venom phosphodiesterase digestion of purified 3H-labeled SP83 releases 5'-AMP and a minor fraction of 2'-(5"-phosphoribosyl)-5-AMP. In vitro labeling with [32P]NAD+ of total cell lysates made in the presence of non-ionic detergents also results in incorporation of radioactivity into SP83. Both of these results strongly suggest that the modification is an ADP-ribosylation. Heat shock and glucose starvation of cells induce a rapid and extensive decrease in the incorporation of ADP-ribose into SP83, suggesting that ADP-ribosylation may be important for the regulation of the function of this protein.     

Forskolin: unique diterpene activator of adenylate cyclase in membranes and in intact cells.

The diterpene, forskolin [half-maximal effective concentration (EC50), 5-10 microM] activates adenylate cyclase [ATP pyrophosphate-lyase (cyclizing), EC 4.6.1.1] in rat cerebral cortical membranes in a rapid and reversible manner. Activation is not dependent on exogenous guanyl nucleotides and is not inhibited by guanosine 5'-O-(2-thiodiphosphate) when assayed with adenosine 5'-[beta, gamma-imido]triphosphate as substrate. GTP and GDP potentiate responses to forskolin. The activations of adenylate cyclase by forskolin and guanosine 5'-[beta, gamma-imido]triphosphate p[NH]ppG are not additive, whereas activations by forskolin and fluoride are additive or partially additive. The responses of adenylate cyclase to forskolin or fluoride are not inhibited by manganese ions, whereas the response to p[NH]ppG is completely blocked. Activation of adenylate cyclase by forskolin is considerably greater than the activation by fluoride in membranes from rat cerebellum, striatum, heart, and liver, while being about equal or less than the activation by fluoride in other tissues. Forskolin (EC50, 25 microM) causes a rapid and readily reversible 35-fold elevation of cyclic AMP in rat cerebral cortical slices that is not blocked by a variety of neurotransmitter antagonists. Low concentrations of forskolin (1 microM) augment the response of cyclic AMP-generating systems in brain slices to norepinephrine, isoproterenol, histamine, adenosine, prostaglandin E2, and vasoactive intestinal peptide. Forskolin would appear to activate adenylate cyclase through a unique mechanism involving both direct activation of the enzyme and facilitation or potentiation of the modulation of enzyme activity by receptors or the guanyl nucleotide-binding subunit, or both.     

Cleavage of the Escherichia coli lexA protein by the recA protease.

The recA and lexA proteins of EScherichia coli are involved in a complex regulatory circuit that allows the expression of a diverse set of functions after DNA damage or inhibition of DNA replication. Exponentially growing cells contain a low level of recA protein, and genetic evidence suggests that lexA protein is involved in its regulation, perhaps as a simple repressor. Recent models for recA derepression after DNA damage have suggested that an early event in this process is the proteolytic cleavage of lexA protein, leading to high-level expression of recA. We present several lines of evidence that the specific protease activity of the recA protein, previously described with the lambda repressor as substrate, is capable of cleaving the wild-type lexA+ protein. First, lexA protein can be cleaved in vitro under the same conditions as prevously described for lambda repressor cleavage in a reaction requring both recA protease and ATP or an analogue, adenosine 5'-[lambda-thio]-triphosphate. Second, lexA protein can be observed in vivo as a physical entity after infection with lambda lexA+ transducing phage of host strains containing ittle or no active protease, but not in strains containing high levels of active protease. Finally, infection of host cells containing active protease with a lambda lexA+ transducing phage does not lead to repression of recA, but does so in cells lacking active protease. In all of these conditions the mutant lexA3 protein is largely resistant to inactivation or cleavage; this resistance can explain the dominant phenotype of lexA3 over lexA+. We discuss models for recA derepression and re-establishment of repression which propose that modulation of the protease activity of recA protein regulates both of these transitions.     

RecA protein filaments can juxtapose DNA ends: an activity that may reflect a function in DNA repair.

To further characterize the role of RecA protein-DNA filaments in general recombination and DNA repair, we have examined interactions of these filaments with themselves following formation. When linear double-stranded DNA was incubated with RecA in the presence of Mg2+ and adenosine 5'-[gamma-thio]triphosphate, monomer-length (1n) nucleoprotein filaments were observed. Following continued incubation, filaments having 2n, 3n, ... lengths were observed, indicating that an end-to-end joining of the monomer-length filaments had occurred. When linear single-stranded DNA was covered by RecA protein under several conditions, the ends of the resulting filaments joined together rapidly, producing circular filaments. The end-to-end joining of single-stranded DNA-RecA filaments appeared to require that 3' DNA ends be juxtaposed with 5' DNA ends, because double-stranded DNA molecules having long single-stranded DNA tails with only 3' or 5' termini did not join end-to-end. However, when both 5' and 3' ends were present in the reaction, joining was observed. We suggest that this end-to-end joining activity may help explain the role of RecA protein in both the protection of damaged DNA ends and the repair of double-stranded DNA breaks.