Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. May 12, 1998; 95(10): 5505–5510.

In vitro site-specific integration of bacteriophage DNA catalyzed by a recombinase of the resolvase/invertase family


The genome of the broad host range Streptomyces temperate phage, [var phi]C31, is known to integrate into the host chromosome via an enzyme that is a member of the resolvase/invertase family of site-specific recombinases. The recombination properties of this novel integrase on the phage and Streptomyces ambofaciens attachment sites, attP and attB, respectively, were investigated in the heterologous host, Escherichia coli, and in an in vitro assay by using purified integrase. The products of attP/B recombination, i.e., attL and attR, were identical to those obtained after integration of the prophage in S. ambofaciens. In the in vitro assay only buffer, purified integrase, and DNAs encoding attP and attB were required. Recombination occurred irrespective of whether the substrates were supercoiled or linear. A mutant integrase containing an S12F mutation was completely defective in recombination both in E. coli and in vitro. No recombination was observed between attB/attB, attP/attP, attL/R, or any combination of attB or attP with attL or attR, suggesting that excision of the prophage (attL/R recombination) requires an additional phage- or Streptomyces-encoded factor. Recombination could occur intramolecularly to cause deletion between appropriately orientated attP and attB sites. The results show that directionality in [var phi]C31 integrase is strictly controlled by nonidentical recombination sites with no requirement to form the topologically defined structures that are more typical of the resolvases/invertases.

In site-specific recombination a recombinase interacts with a specific site in the DNA, brings the sites together in a synapse, and then catalyzes strand exchange so that the DNA is cleaved and religated to opposite partners (1, 2). The reaction can result in integration, inversion, or resolution/excision depending on the position and orientation of the recombination sites, their interactions with recombinase, and the presence or absence of accessory factors or sites. Site-specific recombinases in bacteria fall into one of two very distinct families, the λ integrase-like enzymes and the resolvase/invertases, on the basis of amino acid sequence similarities and their different mechanisms of catalysis (13). Recombination by members of the λ integrase family (e.g., λ integrase, P1 Cre-loxP) is well understood and involves the formation and resolution of a Holliday junction intermediate during which the DNA is transiently attached to the enzyme through a phosphotyrosine linkage (46). The resolvase/invertase family of enzymes (e.g., Tn3 or γδ resolvases, Mu Gin invertase) act via a concerted, four-strand staggered break and rejoining mechanism during which a phosphoserine linkage is formed between the enzyme and the DNA (2, 7). The crystal structure of γδ resolvase bound to a cleavage site reveals a unique arrangement of the catalytic and DNA-binding domains in that they bind to different faces of the helix (8). Although two models have been proposed (911), the structure of the synapse and the changes in the conformation of resolvase that bring about strand exchange are still a mystery (12).

Recently a new subgroup of the resolvase/invertases has been identified that have the resolvase/invertase catalytic domain in their N-terminal regions but are much larger in molecular mass (between 50.7 kDa and 82 kDa; refs. 1320) than, and extremely diverged from, the typical enzymes of the family (≈20 kDa; refs. 2 and 21). This subgroup includes enzymes involved in excision of DNA during spore (13) or heterocyst (14) development, excision of transposons (15, 16), or integration and excision of bacteriophage genomes (1720). The ability of some of this group of enzymes to catalyze integration, i.e., the integrases from the lactococcal phage TP901–1, and actinophages R4 and [var phi]C31, is a significant departure from the usual activities of the resolvase/invertases (2).

The recombination sites acted on by this new subgroup also differ fundamentally from the more typical resolvase/invertases. In the latter, the two DNA sites involved in recombination are identical for each system. In Tn3, the res sites are 114 bp, and in Mu, the gix sites are 30 bp (2, 22). The sequences are not symmetrical and therefore sites can be in direct or inverse orientation. The reactions catalyzed by Tn3 resolvase and Gin, however, are topologically defined, with precise arrangements of the components in the synaptic complex and in the strand-exchange reaction itself (2, 9, 10, 22, 23). Thus, in the resolvase system only res sites that are in direct repeat recombine, and in the Gin system only inverted sites recombine; recombination is totally blocked if a synapse with the “wrong topology” is formed because of sites being in the incorrect orientation. In the λ integration system (5, 24), the recombination sites have different sequences, i.e., the 25-bp attB, chromosomal attachment site, and the 240-bp attP, phage-encoded attachment site, and recombination results in hybrid sites attL and attR. The sites have a 15-bp of “core” sequence identity. Accessory factors are required for integration (integration host factor, or IHF) and excision (IHF, Fis, and Xis), which also can be described as attP/B and attL/R recombination, respectively. Crossover sites for the enzymes belonging to the high molecular weight subgroup of the resolvase/invertases lie within very short regions of identity between substrate sites [3 bp for [var phi]C31 integrase (ref. 20; Fig. Fig.1);1); 5 bp for CisA (13, 25); 5 bp for XisF (14); 12 bp for R4 Sre (19); 2 bp for Tn4451 TnpX (15, 16)]. These probably represent core-like sequences from longer, nonidentical, recombination sites. If this group of enzymes uses the same catalytic mechanisms as the typical resolvase/invertases, then, in theory, only 2 bp of sequence identity between the two sites is required. Recent work by Crellin and Rood (16) showed that mutation of the 2-bp GA sequence at the center of the recombination sites for TnpX confirmed the need for both sites to have the same sequence where the proposed staggered break occurs.

Figure 1
Organization of the region encoding int in [var phi]C31 and its comparison with the attB sequence from S. ambofaciens (adapted from ref. 20). (a) Organization of attP and int from [var phi]C31 and restriction sites relevant to this study. The black box ...

Here we describe the properties of the [var phi]C31 integrase expressed in Escherichia coli and acting on plasmid-encoded att sites and in an in vitro reaction by using purified integrase. The results show that directionality in [var phi]C31 integrase is strictly controlled by nonidentical recombination sites with no requirement to form the topologically defined structures that are more typical of the resolvase/invertases. Furthermore, for recombination between attP and attB, the enzyme does not require any accessory proteins or supercoiled substrates. The properties of this integrase are discussed in comparisons with other, well studied, site-specific recombination systems.


Strains and Plasmid Constructions.

E. coli DH5α [F′/endA1 hsdR17 (rK mK+) supE44 thi-1 recA1 gyrA(NalR) relA1Δ(lacIZYA-argF) U169 deoR ([var phi]80dlacΔ(lacZ)M15); ref. 26] was used routinely as a host for plasmid constructions, transformations, and plasmid preparations (27). E. coli BL21(DE3) [F ompT (lon) hsdSB (rB mB) DE3 lysogen; a λ prophage carrying the T7 RNA polymerase gene; ref. 28] was used as a host for overproduction of wild-type and mutant integrases. Plasmids used are listed in Table Table1.1. Detailed information on their construction can be obtained from the communicating author. The crossover site in the 0.5-kbp attB-containing fragment in pHS21, pHS23, pHS34, and pHS44 is approximately in the middle of the fragment whereas the crossover site in the 0.5-kbp attP-containing fragment in pHS20, pHS22, pHS33, and pHS44 is 112 bp from the EcoRI site shown in Fig. Fig.1.1. pHS282 was constructed by insertion into pGEM7 of an 84-bp amplification product obtained by PCR using oligonucleotides HS3 (5′-AGGTCTCGAGAAGCGGTTTTCGG) and HS4 (5′-CGCCCTAGGTGTCATGTCGGCGACCC) designed by using the published attP site (20). The overexpression plasmid, pHS62, containing the int gene inserted downstream of the T7 promoter was constructed as follows: A 2-kbp BsrI-Tth111I fragment containing int was inserted into pT7–7 (31) to form pHS61. Primers were designed by using the int sequence (20) for PCR amplification of the N-terminal region to place the start codon downstream of the ribosome-binding site for maximal expression in pT7–7. The primers used were HS1, 5′-AGGTCATATGGACACGTACGCGGGTGC-3′; and HS2, 5′-CCGGCCCCCGTCGCGCTCGA-3′. Amplification yielded a 148-bp fragment that was cut with NdeI and EcoRI, and a 55-bp fragment was used to replace the NdeI-EcoRI fragment from pHS61 to give pHS62. The S12F mutation in the int gene (in pHS63) was generated by using a mutagenic primer, HS11 (5′-AGGCATATGGACACGTACGCGGGTGCTTACGACCGTCAGTTCCGCGAG) in place of HS1, and the resulting amplification product was used to replace the NdeI-EcoRI fragment in pHS61 to give pHS63. The PCR-derived fragments were sequenced to ensure that the correct sequences were present.

Table 1
Plasmids used

In Vitro Recombination.

Integrase was purified from E. coli BL21(DE3) containing either pHS62 or pHS63 by ammonium sulfate precipitation, ion exchange chromatography, and heparin agarose affinity chromatography (H.M.T., S. E. Wilson, and M.C.M.S., unpublished data). Approximately 1 μg each of substrate DNAs was mixed with 1 μg of approximately 90% pure integrase in a volume of 100 μl. Final reaction conditions were 20 mM Tris, pH 7.5/100 mM NaCl/1% glycerol/0.1 mM EDTA, incubated at 30°C. The reactions were stopped by phenol extraction and ethanol precipitation, and the pellets were resuspended in 50 μl of 1× restriction buffer (27). After restriction, 10 μl of the reaction products was separated by electrophoresis on 0.8% 1× TBE (90 mM Tris/64.6 mM boric acid/2.5 mM EDTA, pH 8.3) agarose gels. DNA for use in the recombination reactions was purified by alkaline/lysis and polyethylene glycol precipitation (27). To prepare linearized substrates, plasmids were cleaved with ScaI and the linear DNA was purified after electrophoresis in a 0.8% 1× TBE agarose gel.


The [var phi]C31 Integrase Catalyzes the Integration Reaction in E. coli.

Integrase expressed from the tac promoter in pHS33, a derivative of the autonomously replicating defective λ vector, pZMR100 (30), catalyzed recombination between an attP site inserted downstream of the int gene in pHS33 and the attB site on a compatible plasmid (pHS21) in E. coli DH5α (Fig. (Fig.2).2). Fragments (approximately 0.4 kbp) containing the products of the attP/B recombination, i.e., attL and attR, were inserted into pGEM7 vectors to form pHS50 and pHS52, respectively, and sequenced to confirm that the recombination in E. coli yielded the same products as is normally obtained in Streptomyces ambofaciens (20). The int gene carrying a frameshift mutation at the unique NcoI site failed to yield recombinants, indicating the dependence of the recombination on the expression of [var phi]C31 integrase (data not shown). The ability of integrase to catalyze attP/B recombination in E. coli demonstrated that no other Streptomyces-encoded accessory proteins were required. Constructs in which the attP site from pHS33 and the attB site from pHS21 had been switched (i.e., pHS34 and pHS20) and when introduced together into E. coli also gave rise to recombinants (Fig. (Fig.2),2), confirming that the location of the attP site with respect to the int gene did not affect its activity. Although the precise extent of DNA required for recombination at the attP site has not been determined, a plasmid, pHS282, containing 84 bp of DNA encoding the attP crossover site (located centrally within the insert) was recombinationally active in E. coli containing pHS34 (Fig. (Fig.2).2).

Figure 2
Recombination by [var phi]C31 integrase between attP and attB sites in E. coli DH5α recA. (a) Restriction maps of plasmids used and expected recombinant products. E. coli containing pHS33 and pHS34 encode attP and attB, respectively, ...

In Vitro Recombination Catalyzed by [var phi]C31 Integrase.

From the experiments above it seemed highly likely that the recombination reaction would occur in vitro without any additional proteins other than integrase. The int gene was overexpressed and purified by ion-exchange and heparin agarose affinity chromatography (H.M.T., S. E. Wilson, and M.C.M.S., unpublished data). A purified extract containing approximately 90% integrase in 20 mM Tris, pH 8.0/1 M NaCl/1 mM EDTA was obtained that contained little or no DNases. Standard conditions for recombination involved mixing substrate attP and attB DNA with integrase in Tris buffer, pH 7.5, and incubation at 30°C. A time course of the in vitro reaction suggested that as much as 50% of the initial substrate DNA recombined to form attL and attR after 10–16 hr incubation (Fig. (Fig.33b). Addition of 10 mM MgCl2 did not stimulate or inhibit recombination as measured after 16 hr of incubation (Fig. (Fig.33c), and addition of 10 mM EDTA had no apparent inhibitory effect (not shown). Both supercoiled and linearized DNAs were suitable substrates for recombination between attP and attB (Fig. (Fig.33d).

Figure 3
Recombination by [var phi]C31 integrase between attP and attB sites in vitro. (a) Restriction maps of plasmids used and expected recombinant products. Detection of the recombination products in bd was by restriction with EcoRI followed by agarose ...

During strand exchange a recombinase belonging to the resolvase/invertase family transiently forms a phosphoserine linkage between the enzyme and the 5′ end of a cleaved DNA substrate at the crossover site. The active site serine residues are found very close to the N-terminal end and are surrounded by other well conserved residues (21). In the [var phi]C31 integrase, Ser-12 is most likely to be the active-site serine (21). A mutation was introduced in the int gene to change Ser-12 to a phenylalanine (S12F). The mutant protein, encoded by pHS63, could not catalyze recombination between attP and attB sites either in E. coli or in the in vitro reaction by using purified protein obtained via the same procedure as the wild-type protein (Fig. (Fig.33c), although binding of S12F integrase to attP and attB was unaffected (not shown).

In the Absence of Other Factors [var phi]C31 Integrase Only Catalyzes Recombination Between attP and attB.

Could the [var phi]C31 integrase catalyze excision, i.e., the reaction between attL and attR? We use here the same system to describe the organization of the attP, B, L, and R sites as is used for the λ attachment sites (5, 24). The attB site therefore is BOB′, attP is POP′, where B and B′ and P and P′ describe the left and right arms at each site and O describes the homologous core. After recombination attL therefore is BOP′ and attR is POB′. The attL and attR sites were isolated from the products of attP/attB recombination in E. coli as approximately 400-bp fragments containing 112 and 196 bp of phage DNA, respectively. The attL- and attR-containing fragments were inserted into pGEM7 to produce pHS55 in the orientation BOP′–POB′, i.e., as they would occur in the integrated prophage. To prevent any constraints on recombination because of attL and attR located too close, pHS55 contained a 1.5-kbp fragment derived from the region encoding the repressor in [var phi]C31. No recombination was observed between the attL and R sites in pHS55 either in E. coli containing a compatible plasmid expressing integrase (pHS15; not shown) or in vitro (Fig. (Fig.4).4). Similarly, no recombination was detected in E. coli or in vitro between attL and attR when located on different plasmids, or between attP or attB and attL or attR (Fig. (Fig.4).4). Using plasmids of different sizes to facilitate detection of the recombinant form, no recombination was detected between two attP sites or between two attB sites (Fig. (Fig.4).4).

Figure 4
[var phi]C31 integrase only catalyzes attP/B recombination. Plasmids encoding attP (pHS20 or pHS22), attB (pHS21 or pHS23), attL (pHS50), or attR (pHS52) or attL and attR together (pHS55) were used as substrates for recombination with [var phi]C31 ...

Intramolecular Recombination Between attP and attB.

Although in nature the location of the attP and attB sites on the same DNA molecule is unlikely to occur, we looked for any constraint on recombination under these conditions that might be indicative of a topological influence on formation of the synapse. A substrate containing the attP and attB sites in so-called direct orientation (BOB′:POP′) was constructed (pHS44) and recombination was observed in E. coli (also containing pHS15) and in vitro. Recombination between the attP and attB sites in the same molecule is expected to give two circular molecules, only one of which contains a replication origin (attR). In E. coli the attL-containing recombination product is lost (Fig. (Fig.55b, lanes 1–3). In vitro circles were detected after restricting the products of recombination, by using pHS44 as a substrate, with an enzyme (BamHI) that does not digest the attR-containing product (Fig. (Fig.55b, lane 5). The circles then were cleaved with ScaI to give the expected linear fragments containing attR (Fig. (Fig.55b, lane 7). The migration properties of the circles suggests that they are relaxed monomers, dimers, trimers, and higher multimers of the attR-containing DNA. We presume that the multimers form by recombination between attP and attB on different plasmids before deletion events occurring by intramolecular recombination. Resolution of two directly repeated res sites by resolvase results in singly linked (catenated) supercoiled circular products (9, 10). Curiously, little if any of the attR circles obtained by recombination between attP and attB in pHS44 (Fig. (Fig.55b) appeared to be supercoiled, although this may be an artifact because of nuclease contamination of the integrase preparation.

Figure 5
Intramolecular recombination between attP and attB. (a) Restriction maps of pHS44 and the expected recombinant products. (b) Restriction analysis of parental and recombinant products after extraction of plasmid DNA from E. coli (lanes 1–3) or ...


The [var phi]C31 integrase is a member of a new subfamily of resolvase/invertases that have, on the basis of amino acid sequence comparisons, the resolvase/invertase catalytic domain at their N termini but that are much larger than the typical members of the family (1320). Consistent with this classification is the observation that mutation of the putative nucleophile, Ser-12, of [var phi]C31 integrase completely abolished recombination both in vivo and in vitro (Fig. (Fig.33c) whereas previously mutation of Tyr-174 and Tyr-181 (chosen because tyrosine is the nucleophile in enzymes belonging to the λ integrase family of recombinases) had no effect on recombination (20). Similar data were obtained after mutation of Ser-15 of TnpX, which prevented excision of Tn4451, but mutation of tyrosines thought to form part of a λ integrase-like catalytic domain had no effect on excision (16). [var phi]C31 integrase displays unusual properties for a recombinase of the resolvase/invertase family. First, it does not display a strong preference for the position or orientation of recombination sites because it caused integration and deletion depending on the positions of the attP and attB sites (Figs. (Figs.225). Second, the reaction does not depend on supercoiling because linear substrates recombined in vitro (Fig. (Fig.3).3). These properties are reminiscent of certain mutants of Gin invertase that, although selected to be Fis-independent, simultaneously gained the ability to integrate and delete DNA between appropriately placed gix sites (32, 33). The Fis-independent Gin mutants demonstrated that strand exchange is not dependent on the topologically defined complex of wild-type Gin.

Whereas the resolvase/invertases have evolved to control the direction of recombination via topologically defined nucleoprotein structures, [var phi]C31 integration is strictly controlled by nonidentical recombination sites (Fig. (Fig.4;4; refs. 20 and 34). This aspect of [var phi]C31 site-specific recombination is reminiscent of λ recombination but with significant differences that probably reflect the formation of very different synaptic structures. In λ, the minimal attP sequence is 240 bp in length and an accessory factor, IHF, is required for recombination (5, 24). At the λ attP site, which must be supercoiled, integrase and IHF bind to sites on the arms of attP (arm-type binding sites) at some distance from the core sequence (the site of crossing over). λ integrase, which has two DNA-binding motifs, then can contact the low-affinity, core-type binding sites immediately flanking the core sequence (5, 24). The synapse is formed when this attP–integrase–IHF complex or “intasome” combines with the much shorter (25-bp) attB site (5, 24). In the [var phi]C31 system we envisage that (like γδ and Tn3 resolvase, Gin, etc.) integrase binds the imperfect inverted repeats that flank the crossover sites in attP and attB (refs. 20 and 34; Fig. Fig.1).1). This model accounts for the smaller attP site (which can be equal to or less than 84 bp) and the lack of accessory factors required for attP/B recombination. Asymmetries in integrase binding to att sites would determine the directionality because only integrase bound to the attP/B combination of sites could form a productive synapse. We currently are studying these interactions and other features of the att sites that might contribute to the control of directionality in this system. We expect that the excision reaction catalyzed by [var phi]C31 integrase (i.e., attL and attR recombination) will require accessory proteins, probably phage-encoded, thus resembling excision catalyzed by XisF in Anabeana (35) or plasmid pSM19035 monomerization by the β protein (36). The structure and activities of [var phi]C31 and other members of this new group of recombinases show that the resolvase/invertase family is much more diverse than originally thought (2).


We are grateful to S. Kuhstoss and M. Brawner for gifts of plasmids. This work was funded by Medical Research Council Grant G9502658.


This paper was submitted directly (Track II) to the Proceedings Office.

Abbreviation: IHF, integration host factor.


1. Sadowski P. J Bacteriol. 1986;165:341–347. [PMC free article] [PubMed]
2. Stark W M, Boocock M R, Sherratt D J. Trends Genet. 1992;8:432–439. [PubMed]
3. Argos P, Landy A, Abremski K, Egan J B, Haggard-Ljunquist E, Hoess R H, Kahn M L, Kaliones B, Narayana S V L, Pierson L S, III, et al. EMBO J. 1986;5:433–440. [PMC free article] [PubMed]
4. Nunes-Duby S, Azaro M A, Landy A. Curr Biol. 1995;5:139–148. [PubMed]
5. Landy A. Annu Rev Biochem. 1989;58:913–949. [PubMed]
6. Guo F, Gopaul D N, Van Duyne G D. Nature (London) 1997;389:40–46. [PubMed]
7. Hatfull G F, Grindley N D F. In: Genetic Recombination. Kucherlapati R, Smith G R, editors. Washington, DC: Am. Soc. Microbiol.; 1988. pp. 357–396.
8. Yang W, Steitz T A. Cell. 1995;82:193–207. [PubMed]
9. Stark W M, Boocock M R, Sherratt D J. Trends Genet. 1989;5:304–309. [PubMed]
10. Stark W M, Sherratt D J, Boocock M R. Cell. 1989;58:779–790. [PubMed]
11. Rice P A, Steitz T A. EMBO J. 1994;13:1514–1524. [PMC free article] [PubMed]
12. Oram M, Szczelkun M D, Halford S E. Curr Biol. 1995;5:1106–1109. [PubMed]
13. Sato T, Samori Y, Kobayashi Y. J Bacteriol. 1990;172:1092–1098. [PMC free article] [PubMed]
14. Carrasco C D, Ramaswamy K S, Ramasubramanian T S, Golden J W. Genes Dev. 1994;8:74–83. [PubMed]
15. Bannam T L, Crellin K, Rood J I. Mol Microbiol. 1995;16:535–551. [PubMed]
16. Crellin P K, Rood J I. J Bacteriol. 1997;179:5148–5156. [PMC free article] [PubMed]
17. Mahairas G G, Sabo P J, Hickey M J, Singh D C, Stover C K. J Bacteriol. 1996;178:1274–1282. [PMC free article] [PubMed]
18. Christiansen B, Bronsted L, Vogensen F K, Hammer K. J Bacteriol. 1996;178:5164–5173. [PMC free article] [PubMed]
19. Matsuura M, Noguchi T, Yamaguchi D, Aida T, Asayama M, Takahashi H, Shirai M. J Bacteriol. 1996;178:3374–3376. [PMC free article] [PubMed]
20. Kuhstoss S, Rao R N. J Mol Biol. 1991;222:897–908. [PubMed]
21. Leschziner A E, Boocock M R, Grindley N D F. Mol Microbiol. 1995;15:865–870. [PubMed]
22. Van der Putte P, Goosen N. Trends Genet. 1992;8:457–462. [PubMed]
23. Kanaar R, van der Putte P, Cozzarelli N R. Proc Natl Acad Sci USA. 1988;85:752–756. [PMC free article] [PubMed]
24. Nash H A. Trends Biochem Sci. 1990;15:222–227. [PubMed]
25. Stragier P, Kunkel B, Kroos L, Losick R. Science. 1989;243:507–512. [PubMed]
26. Woodcock D M, Crowther P J, Doherty J, Jefferson S, Decruz E, Noyerwiedner M, Smith S S, Michael M Z, Graham M W. Nucleic Acids Res. 1989;17:3469–3478. [PMC free article] [PubMed]
27. Sambrook J, Fritsch E F, Maniatis T. Molecular Cloning: A Laboratory Manual. 2nd Ed. Plainview, NY: Cold Spring Harbor Lab. Press; 1989.
28. Studier F W, Rosenberg A H, Dunn J J, Dubendorff J W. Methods Enzymol. 1990;185:60–89. [PubMed]
29. Kuhstoss S, Richardson M A, Rao R N. Gene. 1991;97:143–146. [PubMed]
30. Rogers M, Ekaterinaki N, Nimmo E, Sherratt D J. Mol Gen Genet. 1986;205:550–556. [PubMed]
31. Tabor S, Richardson C C. Proc Natl Acad Sci USA. 1985;82:1074–1078. [PMC free article] [PubMed]
32. Klippel A, Cloppenborg K, Kahmann R. EMBO J. 1988;7:3983–3989. [PMC free article] [PubMed]
33. Klippel A, Kanaar R, Kahmann R, Cozzarelli R. EMBO J. 1993;12:1047–1057. [PMC free article] [PubMed]
34. Rausch H, Lehmann M. Nucleic Acids Res. 1991;19:5187–5189. [PMC free article] [PubMed]
35. Ramaswamy K S, Carrasco C D, Fatma T, Golden J W. Mol Microbiol. 1997;23:1241–1249. [PubMed]
36. Rojo F, Alonso J C. J Mol Biol. 1994;238:159–165. [PubMed]
37. Smith M C M, Owen C. Mol Microbiol. 1991;5:2833–2844. [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...