Molecular Microbiology (2005) 55 (6), 1896–1910 doi:10.1111/j.1365-2958.2005.04517.x© 2005 Blackwell Publishing Ltd Blackwell Science, LtdOxford, UKMMIMolecular Microbiology0950-382XBlackwell Publishing Ltd, 2005 ? 2005 55 618961910 Original Article M. tuberculosis f Rv1 integration and excisionL. A. Bibb, M. I. Hancox and G. F. Hatfull Accepted 7 December, 2004. *For correspondence. [email protected]; Tel. ( + 1) 412 624 4350; Fax ( + 1) 412 624 4870. Integration and excision by the large serine recombinase ffff Rv1 integrase Lori A. Bibb, Maria I. Hancox and Graham F. Hatfull* Pittsburgh Bacteriophage Institute and Department of Biological Sciences, University of Pittsburgh, Pittsburgh, PA 15260, USA. SummaryThe Mycobacterium tuberculosis prophage-like ele-ment ffff Rv1 encodes a site-specific recombination sys-tem utilizing an integrase of the serine recombinasefamily. Recombination occurs between a putative attP site and the host chromosome, but is unusual in thatthe attB site lies within a redundant repetitive element(REP13E12) of which there are seven copies in the M. tuberculosis genome; four of these elementscontain attB sites suitable for ffff Rv1 integration in vivo .Although the mechanism of directional control oflarge serine integrases is poorly understood, arecombination directionality factor (RDF) has beenidentified that is required for ffff Rv1 integrase-mediatedexcisive recombination in vivo . Here we describedefined in vitro recombination reactions for both ffff Rv1integrase-mediated integration and excision andshow that the ffff Rv1 RDF is not only required for exci-sion but inhibits integrative recombination; neitherreaction requires DNA supercoiling, host factors,or high-energy cofactors. Integration, excision andexcise-mediated inhibition of integration require sim-ple substrates sites, indicating that the control ofdirectionality does not involve the manipulation ofhigher-order protein–DNA architectures as describedfor the tyrosine integrases.Introduction The genomes of both Mycobacterium tuberculosis H37Rvand CDC1551 contain two prophage-like elements f Rv1and f Rv2, and while related elements are present in Mycobacterium bovis AF2122/97 (Cole et al ., 1998; Fleis-chmann et al ., 2002; Garnier et al ., 2003) they are absentfrom M. bovis BCG (Mahairas et al ., 1996). These ele-ments are approximately 10 kilobases in length and con-tain 14–15 open reading frames, many of which code forputative phage proteins such as capsid subunits, proheadproteases and integrases (Cole et al ., 1998; Hendrix et al ., 1999; Fleischmann et al ., 2002). However, the cod-ing capacity of these prophage-like elements is likelyinsufficient for the production of tailed phage virions, andno particles containing f Rv1 or f Rv2 DNA have beendescribed. f Rv1 and f Rv2 differ in that f Rv2 encodes a tyrosineintegrase and f Rv1 encodes a large serine recombinase(Hendrix et al ., 1999). The f Rv1 element is present indifferent chromosomal locations in the two completelysequenced M. tuberculosis genomes, suggesting that itencodes a functional system for excision and integration(Bibb and Hatfull, 2002; Fleischmann et al ., 2002; Garnier et al ., 2003), and the functionality of this site-specificrecombination system has been confirmed by demon-strating that a non-replicating plasmid containing the inte-grase gene and an attP site (reconstructed from theprophage attL and attR sites) efficiently transforms M.bovis BCG (Bibb and Hatfull, 2002). Integration can occurat four attB sites, each located within the repetitive ele-ment REP13E12, and two of these are the sites occupiedin M. tuberculosis H37Rv and CDC1551 (Bibb and Hatfull,2002). There are seven REP13E12 copies in the M. tuber-culosis genome although no integration events wereobserved within the other three REP13E12 elements, pre-sumably resulting from sequence differences in the attB sites that are deleterious to integration (Bibb and Hatfull,2002). The efficiency of f Rv1 integration is approximately100-fold lower in Mycobacterium smegmatis than it is inBCG, although the events observed occur within aREP13E12-related element (Bibb and Hatfull, 2002). Thelower efficiency of M. smegmatis integration is not due tothe lack of a BCG-specific host factor, because introduc-tion of a permissive BCG attB site into M. smegmatis elevates the frequency of integration approximately 300-fold (Bibb and Hatfull, 2002).The f Rv1 attP and attB sites share a 12 bp commoncore within which strand exchange occurs (Bibb and Hat-full, 2002). The attP site has inverted symmetry centredbetween positions 4 and 5 of the common core, indicatingthat the 5 ¢ -TG dinucleotide at core positions 4 and 5represents the central dinucleotide about which strandcleavage occurs. The attB site #6 has only limited invertedsymmetry that is also centred about the putative centraldinucleotide. However, inverted symmetry is greater inM. tuberculosis f Rv1 integration and excision 1897 © 2005 Blackwell Publishing Ltd, Molecular Microbiology , 55 , 1896–1910 attB site #1 than in #6 although both sites are used atsimilar frequencies in vivo (Bibb and Hatfull, 2002). Theorganization of these sites is discussed below (see Figs 4and 5). f Rv1 integrase is a 469 residue protein of which the150 N-terminal residues are related to the catalyticdomain of a large group of site-specific recombinases thatincludes transposon resolvases and DNA invertases;each of these contains a catalytic serine located near theN-terminus (position 10 in gd resolvase) (Hatfull andGrindley, 1988; Bibb and Hatfull, 2002; Smith and Thorpe,2002). Over 30 members of this large serine-recombinasegroup have been described, although they vary greatly inlength (441–772 amino acids) and in the sequence of thesegments C-terminal to the putative catalytic domains(Smith and Thorpe, 2002). The Bxb1 serine integrase hasbeen demonstrated to cleave its DNA substrates to gen-erate 2-base 3 ¢ extensions and to form covalent DNA–protein linkages, and it is likely that all serine-integrasesuse this catalytic mechanism, which is also common totransposon resolvases and DNA invertases (Ghosh et al .,2003; Kim et al ., 2003). However, the Bxb1 and f C31serine integrase systems have both been shown torequire small, simple DNA substrates for attP and attB –in contrast to the complex sites used by