The genetic transformation machinery: composition, localization, and mechanism


  • Jean-Pierre Claverys,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France; and
    2. Laboratoire de Microbiologie et Génétique Moléculaires, Université de Toulouse, UPS, Toulouse, France
    Search for more papers by this author
  • Bernard Martin,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France; and
    2. Laboratoire de Microbiologie et Génétique Moléculaires, Université de Toulouse, UPS, Toulouse, France
    Search for more papers by this author
  • Patrice Polard

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France; and
    2. Laboratoire de Microbiologie et Génétique Moléculaires, Université de Toulouse, UPS, Toulouse, France
    Search for more papers by this author

  • Editor: Chris Bayliss

Correspondence: Jean-Pierre Claverys, Université Paul Sabatier-CNRS, LMGM, Bât. IBCG, 118 route de Narbonne, F-31062 Toulouse Cedex 9, France. Tel.: +33 561 33 59 11; fax: +33 561 33 58 86; e-mail:


Natural genetic transformation is widely distributed in bacteria. It is a genetically programmed process that is inherent to the species. Transformation requires a specialized membrane-associated machinery for uptake of exogenous double-stranded DNA. It also requires dedicated cytosolic proteins, some of which have been characterized only recently, for the processing of internalized single-stranded DNA fragments into recombination products. A series of observations made in Bacillus subtilis and Streptococcus pneumoniae led to the recent emergence of a picture of a unique, highly integrated machine localized at the cell poles. This dynamic machine, which we propose to name the transformasome, involves both membrane and cytosolic proteins, to internalize, protect, and process transforming DNA. This review attempts to summarize these recent observations with special emphasis on the early stages in DNA processing.


Although there is no sexual reproduction in bacteria, natural genetic transformation and conjugation are regarded as parasexuality (Maynard Smith et al., 1991; Kohiyama et al., 2003). Both processes mediate the transport of single-stranded DNA (ssDNA) into the cytosol (Chen et al., 2005). However, conjugation is a plasmid-encoded process and conjugation-mediated chromosome transfer has been documented only in a very limited number of species (e.g. Escherichia coli, Salmonella). On the other hand, bacterial transformation is widely distributed in nature (Johnsborg et al., 2007) and relies on a process that is inherent to the species (i.e. it does not depend on the presence of mobile or extrachromosomal genetic elements). Transformation thus constitutes a bacterium-programmed mechanism for genetic exchange. Besides encoding the entire set of proteins required for binding of exogenous double-stranded DNA (dsDNA), for internalization of ssDNA fragments, and for homologous integration of ssDNA into the chromosome, many transformable species have elaborated complex regulatory circuits to tightly regulate the expression of this set of genes. They express it in response to environmental signals to become transiently ‘competent’ for genetic transformation. Signals inducing competence vary from species to species. For example, competence is induced in response to nutrient limitation in Bacillus subtilis or to the presence of antibiotics in Streptococcus pneumoniae (the Pneumococcus) (Claverys et al., 2006).

Until recently, transformation was believed to involve the uptake of environmental DNA (Chen et al., 2005). However, there is now evidence that at least some bacterial species do not rely for transformation on the accidental release of DNA but can trigger active release. In S. pneumoniae, release involves the killing of noncompetent pneumococci by competent cells, a phenomenon called pneumococcal fratricide (Claverys et al., 2007). It was recently shown that killing is not limited to S. pneumoniae but that related species such as Streptococcus mitis and Streptococcus oralis can also be attacked by competent pneumococci, and that competent cells of S. mitis also display the ability to attack S. mitis, S. pneumoniae, and S. oralis (Johnsborg et al., 2008). The existence of a different DNA release mechanism, which does not invoke cell lysis, has also been documented in Neisseria gonorrhoeae (Hamilton & Dillard, 2006). Thus a picture emerges of a highly integrated process involving not only DNA uptake and processing machineries but also active mechanisms for the gathering of transforming DNA. This integrative view is reinforced by recent work in B. subtilis and S. pneumoniae suggesting that the DNA uptake and processing machineries constitute a unique multiprotein polar complex. This large machine involving both cytosolic and membrane proteins mediates the binding and uptake of ssDNA, its protection from endogenous nucleases, and its recombination with the recipient chromosome.

These recent observations are the subject of this review (Note that hereafter ‘genetic transformation’ refers only to chromosomal transformation. We do not consider plasmid transformation as its genetic requirements are complicated by issues, which are out of the scope of this review, like the reconstitution of replicons from incoming ssDNA fragments and/or the initiation of plasmid replication.). As DNA uptake and the regulation of competence have been abundantly reviewed in the recent period, we place emphasis on the early stages in DNA processing through a combined discussion of old relevant literature and recent data.

DNA uptake


As mentioned above, the ability to transport DNA into the cytosol is displayed only by cells that have transiently differentiated to competence. The mechanism of DNA uptake was characterized in both B. subtilis and S. pneumoniae long before the identification of any uptake gene. Because this mechanism was described in great details in the earlier literature [Dubnau (1999) and references therein], it is presented only briefly here. In both species, dsDNA is first bound to the cell surface without base sequence preference. Remarkably, ssDNA displays c. 200-fold reduced transforming activity compared with dsDNA in S. pneumoniae (Miao & Guild, 1970), but the limiting step(s) is not known. Fragmentation of dsDNA occurs upon binding, which generates dsDNA fragments with an average length of 6 kb in S. pneumoniae and 13.5–18 kb in B. subtilis (Dubnau, 1999). In both species, ssDNA fragments are then transported across the membrane (Fig. 1a) whereas the nontransported strand is degraded; degradation products (5′-nucleotides, nucleosides, and free bases for B. subtilis; short oligonucleotides for S. pneumoniae) are released into the medium. Internalized ssDNA becomes immediately resistant to externally added nucleases and can be recovered from lysed cells.

Figure 1.

 A DNA uptake machine is assembled at the cell poles. (a) Uptake of exogenous dsDNA directly produces ssDNA in competent cells. Accumulation of ComGA-GFP was observed in a Bacillus subtilis strain carrying an ectopic copy of comGA-gfp [reproduced with permission from Fig. 1b (Hahn et al., 2005)]. Numbers within triangles refer to steps in the process as discussed in the text. CW, cell wall; M, membrane. (b) In vegetative cells creation of ssDNA from a dsDNA break can be achieved by the concerted action of DNA helicases and exonucleases (e.g. Escherichia coli RecBCD or RecQ-RecJ). Bsu, B. subtilis; Eco, E. coli; Spn, Streptococcus pneumoniae.

A 3′→5′ polarity of ssDNA entry was established in S. pneumoniae (Méjean & Claverys, 1988) whereas no clear polarity was reported for B. subtilis (Dubnau, 1999). In S. pneumoniae, degradation of the nontransported strand occurred with the opposite polarity. The rate of uptake and of degradation appeared similar, 90–100 nt s−1 at 31 °C (Méjean & Claverys, 1993), suggesting (but by no means demonstrating) that transport of one strand is coupled with degradation of its complement. A transport rate of 80±10 nt s−1 at 37±2 °C was recently determined for B. subtilis (Maier et al., 2004). In both species, transport rates were determined (and are indicated) for individual DNA molecules.

The DNA uptake machinery

This machinery was thoroughly characterized mostly through the work of Dubnau and coworkers in B. subtilis with additional insights coming from studies in S. pneumoniae (for reviews, see Chen & Dubnau, 2004; Chen et al., 2005). Briefly, the machinery diagrammed for S. pneumoniae in Fig. 1a comprises a transformation pseudopilus, a structure evolutionarily related to type IV pili (Dubnau, 1999; Chen et al., 2005, 2006), made of the major pilin-like protein, ComGC, and minor pilin-like proteins not represented in the figure (ComGD, ComGE, ComGG); a dsDNA receptor protein, ComEA; an endonuclease identified in S. pneumoniae only, EndA; a polytopic membrane protein considered as the transmembrane channel for ssDNA, ComEC; and an ATP-binding protein, ComFA. Similarly to type IV pili assembly, the pseudopilins are processed by a prepilin peptidase (ComC in B. subtilis, Ccl in S. pneumoniae; not shown) and assembled into the pseudopilus with the aid of the polytopic membrane protein, ComGB, and the traffic NTPase, ComGA.

The mechanism of transport can be visualized as follows (Chen & Dubnau, 2004; Chen et al., 2005). Initial binding of exogenous DNA to the surface may involve interaction of the negatively charged DNA backbone with positively charged residues in the pseudopilus itself (Craig & Li, 2008). In B. subtilis, a surface endonuclease, NucA, introduces double-stranded cleavages into the bound DNA (Provvedi et al., 2001). In agreement with a primary role of NucA in generating DNA termini positioned for engagement with the transport machinery, this endonuclease is not required when cleaved DNA is used (Provvedi et al., 2001). The nuclease responsible for the initial introduction of single-strand nicks in S. pneumoniae has not been identified, but EndA could be responsible for cleavage opposite the initial nick (Dubnau, 1999). Retraction (through disassembly) of the pseudopilus would then allow the exogenous DNA to traverse the peptidoglycan (step 1 in Fig. 1a), possibly driven by the proton motive force (Maier et al., 2004; Chen et al., 2005), thereby conveying dsDNA into contact with ComEA (step 2). The dsDNA receptor would then deliver dsDNA to EndA (step 3) to degrade one strand, thus permitting transport of its complement through the transmembrane channel, ComEC (step 4). It is proposed that ComFA is acting as an ATP-dependent strand translocase to mediate ssDNA internalization.

All components of the DNA uptake machine are encoded by genes belonging to the competence regulon in both species (Claverys et al., 2006) with one remarkable exception, EndA. Historically, this protein was the first key component for DNA uptake to be identified. This endonuclease, localized in the membrane (Lacks & Neuberger, 1975; Rosenthal & Lacks, 1980), is expressed in noncompetent cells and is not induced at competence (Claverys et al., 2006). It is required for transport and for degradation of the nontransported strand but not for the initial binding of DNA at the cell surface (Lacks et al., 1975). It was suggested that EndA is fixed asymmetrically near the pore and that degradation of the nontransported strand occurs by successive endonucleolytic cleavages (Rosenthal & Lacks, 1980). The counterpart of EndA in B. subtilis remains unknown. Interestingly, besides its role in DNA uptake, EndA contributes to virulence by allowing pneumococci to escape from neutrophil extracellular traps, which are made of DNA (Beiter et al., 2006). This feature probably accounts for its expression outside competence.

Degradation of the nontransported strand was found to be abolished in the absence of ComGC or ComGA (Bergéet al., 2002), consistent with the view that despite its membrane location, EndA cannot access transforming DNA unless the transformation pseudopilus is present to facilitate peptidoglycan crossing. Degradation was also abolished in the absence of ComEA, hence the proposal that ComEA is required to deliver dsDNA to EndA (Bergéet al., 2002). Interestingly, degradation was still active in the absence of ComEC, which suggests that neither the pseudotransformation pilus nor the dsDNA receptor ComEA rely on the presence of the transmembrane channel for proper functioning (at least as concerns the delivery of dsDNA to EndA and the degradative action of EndA).

Generating ssDNA from dsDNA, a prerequisite for homologous recombination

The processing of dsDNA into a 3′-ssDNA overhang is a general prerequisite for homologous recombination in living organisms (Fig. 1b). In the paradigm E. coli, this processing relies on the concerted action of a DNA helicase and a DNA nuclease, for example RecBCD or RecQ-RecJ (Fig. 1b). Because production of 3′-ssDNA is achieved by the DNA uptake machinery (Fig. 1a), it is not surprising that the RecBCD-like proteins RexAB in S. pneumoniae (Halpern et al., 2004) or AddAB in B. subtilis (Fernandez et al., 2000) are not required for chromosomal transformation. Moreover, no RecQ homologue is encoded in the genome of S. pneumoniae (Bolotin et al., 2004). The DNA uptake machinery can thus be regarded as a functional equivalent of the dsDNA processing apparatus operating in vegetative cells.

Whatever the mechanism for the production of ssDNA (Fig. 1), this intermediate can then be used as a substrate by RecA, the universal bacterial recombinase, which was first identified and characterized in the model bacterium E. coli, or its counterpart RAD51 in Eukaryotes or RadA in Archaea (Sung & Klein, 2006). The loading of RecA onto ssDNA, which is required to promote the search for homology, to initiate homology-dependent pairing, and to catalyze DNA strand exchange, will be discussed further below.

Fate of transforming DNA


DNA recovered from transformed pneumococci immediately after uptake is devoid of transforming activity when assayed on competent cells. This transient loss of transforming activity termed eclipse (Ephrussi-Taylor, 1960) is due to the fact that, as mentioned above, ssDNA has much less transforming activity than dsDNA. Investigation of the kinetics of recovery from eclipse (as described in the legend to Fig. 2a) indicated that complete restoration of donor marker activity occurred at 30 °C within <20 min after uptake (Fig. 2a). Single-site markers (i.e. point mutations) generally showed 50% recovery in <7 min at this temperature (Ghei & Lacks, 1967; Shoemaker & Guild, 1972).

Figure 2.

 Processing of internalized ssDNA. (a) Recovery from eclipse. In the initial transformation, competent cells were transformed with chromosomal DNA carrying a mal+ (closed circles) and an str-r marker (open circles) for 5 min at 30°C (DNA uptake was terminated with DNAse I). Recovery of these markers was assayed by extracting total DNA from transformed cells, which had been incubated at 30°C various times after uptake, and transforming a test culture (secondary transformation). Reproduced from Ghei & Lacks (1967, fig. 2) with permission. (b) Analysis of the fate of radioactively labelled donor DNA (in the presence of 100 μM HpUra to prevent reincorporation of degraded nucleotides). Adapted from Bergéet al. (2003, fig. 3). (c) Kinetics of integration (circles) and of degradation (open triangles) of radioactively labelled DNA (5 min uptake at 30°C). Adapted from Méjean & Claverys (1984, fig. 3d).

Processing of radioactively labelled ssDNA

Recovery from eclipse was interpreted as the restoration of a double-stranded structure by integration into the resident dsDNA. The use as donor DNA in S. pneumoniae of a radioactively labelled specific fragment (Méjean & Claverys, 1984; Bergéet al., 2003) allowed the monitoring of the fate of internalized ssDNA and of the physical integration of radiolabel into the homologous chromosomal region (Fig. 2b). Briefly, competent cells were exposed to DNA for 3 min (at 30 °C) and uptake of exogenous DNA was terminated with DNAse I; incubation was continued at 30 °C and total DNA extracted from transformed cells was digested with restriction enzymes and analyzed by gel electrophoresis. The kinetics of physical integration into the resident chromosome (monitored through the labelling of specific chromosomal restriction fragments) were strikingly parallel to the kinetics of disappearance of ssDNA, a finding consistent with the view that ssDNA is the precursor for homologous recombination (Fig. 2c). The kinetics of formation of recombinants also appeared very similar to those for recovery from eclipse (compare Fig. 2a with c) providing support to the view that the latter results from the formation of donor–recipient heteroduplexes.

The eclipse nucleoprotein complex

Early evidence for the existence of a nucleoprotein complex was obtained in S. pneumoniae through the study of eclipse DNA. Using hydroxylapatite chromatography, Morrison showed that donor DNA strands recovered from cell lysates during eclipse did not behave like naked ssDNA but were bound to a protein (Morrison, 1977, 1978). This nucleoprotein complex was termed eclipse complex (EC). EC, which arises intracellularly, not in lysates (Morrison, 1977), was reported to contain a single [35S]methionine-labelled protein with an apparent molecular mass (MM) of 15.7 or 19.5 kDa (Morrison & Baker, 1979; Morrison et al., 1979). This protein is one of the major species identified in the pulse-labelling experiments establishing that during pneumococcal competence protein synthesis is restricted to a few specific polypeptides (Morrison & Baker, 1979). Historically, this observation constitutes the first evidence for the presence of a competence-induced protein complexed with internalized ssDNA, although the identity of the protein was established only some 30 years later.

Competence-induced proteins involved in DNA processing

Five competence-induced proteins are known to be required for the processing of transforming DNA in S. pneumoniae: CoiA, DprA, RadA, RecA, and SsbB (Dagkessamanskaia et al., 2004; Peterson et al., 2004). These proteins, except RadA, are also induced in B. subtilis competent cells (Berka et al., 2002; Hamoen et al., 2002; Ogura et al., 2002). The universal DNA strand-exchange protein RecA is essential for chromosomal transformation both in B. subtilis (Dubnau et al., 1973) and in S. pneumoniae (Martin et al., 1992).

DprA was first identified in Haemophilus influenzae. A transposon insertion in dprA did not affect the uptake of DNA in a form resistant to externally added DNAse but nearly abolished chromosomal transformation (Karudapuram et al., 1995). Because in Gram-negative bacteria, dsDNA becomes nuclease-resistant as it passes through the outer membrane, these data could not distinguish between a role of DprA in DNA translocation through the cytoplasmic membrane or in the processing of internalized ssDNA into recombinants. In S. pneumoniae, the efficient uptake of radioactively labelled DNA in dprA mutant cells ruled out the former possibility (Bergéet al., 2002). Inactivation of dprA results in >104-fold reduction in transformation rate (Bergéet al., 2003; Mortier-Barrière et al., 2007). As concerns CoiA, a 100-fold reduction in transformation was observed in its absence (Desai & Morrison, 2006; Mortier-Barrière et al., 2007). A similar reduction was recently reported for radA mutant cells (Burghout et al., 2007).

In B. subtilis, inactivation of dprA (also called smf) resulted in a 50-fold (Ogura et al., 2002) to 100-fold (Berka et al., 2002) reduction in chromosomal transformation and inactivation of coiA (also called yjbF) reduced transformation by at least 10-fold (Hamoen et al., 2002). Inactivation of radA (also called sms) was reported to reduce chromosomal transformation by 10-fold (Carrasco et al., 2002) to 37-fold (Krüger et al., 1997) (Note that B. subtilis and S. pneumoniae RadA belong to the bacterial Sms family, which is only distantly related to bacterial RecA and archaeal RadA families.).

A facilitating, rather than essential, contribution is made by SsbB since the reduction in the yield of recombinants in ssbB mutant cells of S. pneumoniae, which was initially reported to be 10–30-fold (Campbell et al., 1998), was subsequently found to be only three- to fivefold (Bergéet al., 2003; Morrison et al., 2007; Mortier-Barrière et al., 2007). Similarly, in B. subtilis, inactivation of ssbB (previously called ywpH) reduced transformation between five- (Ogura et al., 2002) and 10-fold (Berka et al., 2002). SsbB is, with SsbA, one of two paralogous ssDNA-binding proteins present in S. pneumoniae (Fig. 3). SsbA, which is essential for cell viability (Thanassi et al., 2002), is not induced in competent pneumococci. Bacillus subtilis is similarly equipped with two SSBs but both genes are induced in competent cells (Fig. 3).

Figure 3.

 SSB homologues of the naturally transformable species Bacillus subtilis, Streptococcus pneumoniae, and Haemophilus influenzae. EcSSB (the Escherichia coli SSB paradigm) comprises two domains, an N-terminal globular domain (c. 110 aa) involved in tetramerization and in nonspecific interaction with ssDNA; and a C-terminal unstructured domain (also called the acidic tail because it is enriched in acidic amino acids), which is the site for specific protein–protein interactions with various partners involved in DNA metabolism (Lecointe et al., 2007; Lu & Keck, 2008; and references therein). Bacterial SSBs display high sequence identity in the N-terminal domain and a similar tip at the C-terminus, the hexapeptide DDD(L/I)PF. Percent identity (over the number of amino acids between parenthesis) indicated within the N-terminal domain rectangle are given with respect to EcSSB. Percent identity between B. subtilis and S. pneumoniae proteins are indicated vertically on the left side of the figure (Note that BsSsbA was previously called SSB.). Fold increase in expression in competent cells (expressed as % of the maximum value observed in the competence regulon in the corresponding experiment) are indicated on the right; values were calculated from the following references: B1, B2, Type 1 and type 2 comparisons, respectively (Berka et al., 2002); D, Dagkessamanskaia et al. (2004); O, Ogura et al. (2002); P, Peterson et al. (2004); R, Redfield et al. (2005); NA, not applicable; NI, noninduced.

Protection of incoming ssDNA from nucleases

Resistance to nucleases of ssDNA in EC was assayed in vitro with DNAse I, Neurospora endonuclease, nuclease P1, and pneumococcal EndA (Morrison & Mannarelli, 1979). In all cases, the amount of enzyme required to release a given fraction of EC DNA was 50–1000 times that required to release the same fraction of naked control ssDNA. These data suggest that ssDNA in EC is protected from endogenous nucleases. A direct indication that once internalized transforming ssDNA may require protection from nucleases came from the observation that 1 min after uptake, radiolabelled donor DNA was already degraded in dprA or recA mutant cells (Bergéet al., 2003).

Identification of SsbB as the major component of EC

Because the apparent MM of the competence-induced protein in the EC was estimated between 15.7 and 19.5 kDa, it seemed more likely to be SsbB (MM, 14 925 Da) than DprA (MM, 31 063 Da) or RecA (MM, 41 950 Da). To identify the component(s) of the EC, purification of the complex through hydroxyapatite was scaled up and protein components were detected both by Western blotting with specific antisera, as well as by nonspecific protein-staining methods. The major protein component of EC was thus identified as SsbB (Morrison et al., 2007). However, there was another band of lower intensity cofractionating with ssDNA in this complex, which comigrated in sodium dodecyl sulfate-polyacrylamide gel electrophoresis with recombinant pneumococcal SsbA. The possible significance of the detection of SsbA with EC as well as the lack of detection of DprA and RecA are discussed in the Interplay section.

A need for alleviating the SSB barrier in transformation?

The molecular characterization of the S. pneumoniae EC suggests that SsbB (and possibly SsbA) interact with internalized ssDNA at an early stage during DNA processing. It has been shown in E. coli that when SSB covers an ssDNA template, it prevents the loading of RecA by steric hindrance. The same observation has been reported with SsbA and RecA from S. pneumoniae and from B. subtilis (Steffen et al., 2002; Carrasco et al., 2008). Dedicated proteins that overcome the SSB barrier and facilitate the loading of RecA onto ssDNA have been characterized in E. coli. These proteins, RecF, RecO, and RecR (Umezu et al., 1993), are widely conserved in bacteria (Rocha et al., 2005). Similarly, the recombinases RAD51 (from Eukaryotes) or UvsX (from bacteriophage T4) cannot access ssDNA coated by RPA or Gp32, the SSB-like proteins in Eukaryotes and in T4, respectively (Beernink & Morrical, 1999). Alleviation of the RPA or Gp32 barrier is carried out by dedicated proteins RAD52 and UvsY, respectively. Altogether, Rec(F)OR, UvsY, and RAD52 define the family of recombination mediator proteins (RMPs) (Beernink & Morrical, 1999) (Table 1). The mechanism of alleviation involves direct and specific protein–protein interactions between the SSB protein and its cognate RMP (Kantake et al., 2002) as well as between the RMP and its recombinase (Beernink & Morrical, 1999; Sung & Klein, 2006) (Table 2).

Table 1.   RMPs required to alleviate SSB barriers and to load recombinases onto ssDNA
OrganismsSaccharomyces cerevisiaeBacteriophage T4Escherichia coli (bacteria)Bacillus subtilis/Streptococcus pneumoniae
Biological roleHomologous recombination (general)Recombinational repairGenetic transformation
General function
Table 2.   Biochemical activities of DprA compared with other RMPs
ActivityS. cerevisiae/T4/E. coliS. pneumoniae, B. subtilis
Interaction with DNAss>dsss>ds*
Annealing complementary ssDNAYesYes
Interaction with recombinaseYesYes
Stimulate recombinaseYesYes
Interaction with SSBYesYes
Overcome SSB inhibitionYesYes
ATP requiredNoNo

The early binding of SsbB to incoming ssDNA thus raised the question of the RMP in charge of its displacement to allow the loading of RecA. Although Rec(F)OR appeared as obvious candidates, the findings that these proteins are dispensable for transformation of S. pneumoniae (A. Baudis, C. Granadel, J.P. Claverys & B. Martin, unpublished data) and that recO and recR mutants of B. subtilis display a less than fourfold reduction in chromosomal transformation (Fernandez et al., 2000) suggested that another RMP might be operating in competent cells.

DprA, a transformation-dedicated RMP

Recent work (Mortier-Barrière et al., 2007) revealed that DprA displays all activities that are the hallmark of RMPs (Table 2).

DprA interacts with RecA

The first evidence for a functional interplay between RecA and DprA came from the finding that both proteins were equally needed for the protection of incoming ssDNA (Bergéet al., 2003). The existence of an interaction between the two proteins was then demonstrated by the specific cocapture of DprA during the selective purification of C-terminal His-tagged RecA from S. pneumoniae competent cells (Mortier-Barrière et al., 2007). This physical interaction was confirmed by yeast two-hybrid assays (Y2H) for both B. subtilis and S. pneumoniae homospecific protein pairs. Cross-species interactions were also detected by Y2H between SpDprA (when there is ambiguity, protein names are preceded by two letters to identify the species, Bs for B. subtilis, Ec for E. coli, and Sp for S. pneumoniae) and BsRecA as well as between BsDprA and SpRecA.

DprA is a widely distributed bacterial protein present in c. 85% of 317 completely sequenced bacterial genomes (Mortier-Barrière et al., 2007). Examination of the highly conserved 205-residue domain they share, called pfam02481 (see the Pfam database of protein families;, revealed no clue as to its function. Recombinant S. pneumoniae and B. subtilis DprA proteins were therefore purified from E. coli in a soluble form to allow their biochemical characterization. Both proteins appeared monomeric in gel filtration experiments, although Y2H indicated that they could self-interact.

DprA interacts with ssDNA and protects it from nucleases

Both BsDprA and SpDprA exhibit a prominent ssDNA-binding activity, essentially no detectable affinity for linear dsDNA but significant binding to supercoiled DNA. In line with its ssDNA-binding activity, SpDprA protects ssDNA (including the 5386-nt-long circular ssDNA molecule of bacteriophage ΦX174) from the action of various endo- and exonucleases (Mortier-Barrière et al., 2007). Further analysis of ssDNA binding showed that DprA binds in a co-operative manner. By contrast with E. coli SSB, however, DprA does not seem to open secondary structures in ssDNA but rather occupy accessible regions, as judged from electron microscopy images of condensed nucleoprotein complexes produced by DprA on long circular ssDNA substrates. In classical gel-shift experiments using short ssDNA substrates (c. 80 nt long), DprA produced large nucleoprotein complexes that could not enter polyacrylamide gels. Bound radiolabelled ssDNA oligonucleotide could be released from the complexes by addition of cold oligonucleotides, showing that these complexes are dynamic.

DprA promotes single-strand annealing

The three prototype RMPs have the ability to anneal complementary ssDNA (Luisi-DeLuca & Kolodner, 1994; Mortensen et al., 1996; Kantake et al., 2002). RAD52 and EcRecO can also anneal ssDNA that is complexed with their cognate ssDNA-binding protein RPA and SSB, respectively (Sugiyama et al., 1998; Kantake et al., 2002). DprA was shown to display ssDNA annealing activity but its capacity to do so with SSB-coated ssDNA was not examined (Mortier-Barrière et al., 2007). It was suggested that the ability of this protein to juxtapose DNA substrates could greatly facilitate the subsequent action of RecA by increasing the local concentration of potential recombination substrates (Mortier-Barrière et al., 2007). It is of note that this property may not be shared by all DprAs because simple ssDNA annealing was not observed with BsDprA when assayed under the same conditions (M. Velten & P. Polard, unpublished data).

DprA promotes the loading of RecA on ssDNA

As heterospecific interactions were detected between SpDprA and BsRecA, and because a similar degree of conservation exists between BsRecA, SpRecA, and EcRecA, the latter, which was readily available in pure form, was used to investigate whether SpDprA could stimulate RecA loading onto ssDNA in vitro. On its own, EcRecA cannot form long nucleofilaments because secondary structures inhibit RecA polymerization. It is well known that SSB favor RecA propagation by melting out regions of secondary structure that otherwise impede the binding of RecA (Kowalczykowski & Krupp, 1987). DprA was found to promote the loading of EcRecA onto ΦX174 ssDNA. In contrast to SSB, DprA was incorporated into the nucleofilaments. The presence of DprA in the nucleofilaments was demonstrated using high-resolution atomic force microscope imaging in native conditions as well as an immunoaffinity labelling procedure and electron microscopy. The mixed nucleofilaments were shown to promote strand exchange with a supercoiled homologous DNA template (Mortier-Barrière et al., 2007).

DprA alleviates the SSB barrier

To further document the parallel between DprA and RMPs, transmission electron microscopy was used to investigate the capacity of DprA to interact with ssDNA precoated with E. coli SSB. As DprA could bind onto SSB–ssDNA complexes, its ability to promote assembly of EcRecA nucleofilaments on SSB-coated ssDNA (which normally prevents the loading of RecA) was examined. The presence of DprA completely overcame the inhibitory effect of SSB on EcRecA loading (Mortier-Barrière et al., 2007). In addition, DprA suppressed the inhibitory effect of SSB on the ssDNA-dependent ATPase activity of EcRecA. As RecA hydrolyzes ATP upon dissociation of ssDNA-RecA nucleofilaments, this result provided independent evidence that DprA promotes the nucleation of EcRecA onto SSB-coated ssDNA. Altogether, these results led to the proposal that DprA is a transformation-dedicated RMP (Mortier-Barrière et al., 2007).

To summarize, the universal recombinase RecA is a central element that is recruited to participate to different pathways (e.g. recombinational repair or genetic transformation) through direct interaction with specific partners (Table 1). None of the canonical RecA-loading machineries characterized in E. coli, RecBCD or Rec(F)OR, seem to be crucial for chromosomal transformation in B. subtilis or in S. pneumoniae. Instead in these organisms, competent cells presumably rely on transformation-dedicated proteins, for example DprA and SsbB, to ensure the loading of RecA on internalized ssDNA and the subsequent processing of transforming DNA.

Interplay of DprA, RecA, and SSBs: open questions

Interplay in transformable species containing a unique SSB

Haemophilus influenzae possesses only one SSB (Fig. 3), the expression of which is increased during competence (Redfield et al., 2005). This is also the case of Helicobacter pylori and Campylobacter jejuni (Lindner et al., 2004). In these bacteria, a rather simple model of DNA processing in which the loading of RecA on internalized transforming DNA is dependent on the ability of DprA to act as an RMP that alleviates the SSB barrier would easily apply (Note that it is implicit in this model that SSB is the first protein to bind to ‘nascent’ ssDNA i.e. ssDNA emerging from the transmembrane channel ComEC; Fig. 1). Extrapolation from data obtained in vitro with SpDprA and the E. coli RecA and SSB proteins (Mortier-Barrière et al., 2007) to H. influenzae is tempting in view of the high similarity between H. influenzae and E. coli proteins (73% and 60% identity for RecA and SSB, respectively).

Interplay in the homospecific system?

However, many transformable species including B. subtilis and S. pneumoniae are equipped with two SSBs, SsbA, and SsbB (Fig. 3). Does the above-mentioned model readily apply in that case? Additional experiments are required because it is already known that SpSsbB differs from SpSsbA with respect to its binding to ssDNA (Grove et al., 2005; Grove & Bryant, 2006). It is also known that SpRecA differs from EcRecA in that its ssDNA-dependent ATP hydrolysis activity is completely inhibited by SpSsbA (as well as by E. coli SSB) (Steffen et al., 2002). The same conclusion was recently attained with BsRecA and BsSsbA (Carrasco et al., 2008). It would therefore be important to establish whether SpSsbB prevents the loading of SpRecA onto ssDNA. If this protein turns out to compete with SpRecA, it would be equally important to determine whether DprA can alleviate the SpSsbB barrier. Although the major EC component is SpSsbB, some SpSsbA was also present at the position of EC during hydroxylapatite chromatography (Morrison et al., 2007). If further work reveals that SpSsbA is also involved in processing of incoming ssDNA, it may become relevant to investigate whether DprA can also alleviate the SpSsbA barrier. These considerations illustrate the need to explore the functioning of the homospecific system to document the interplay between SpRecA and SpSsbB/SpSsbA.

Possible role(s) of transformation-dedicated SSBs

In both B. subtilis and S. pneumoniae, the transformation-dedicated SSB, SsbB, is not essential for chromosomal transformation as judged from the three- to fivefold only reduction in transformation observed in its absence. Nevertheless, the identification of SpSsbB as the major component of EC (Morrison et al., 2007) suggests that it is directly involved in the processing of transforming DNA. What could then be the role(s) of SpSsbB? We discuss hereafter four possibilities.

  • 1At an early stage, SpSsbB might assist the loading of RecA on nascent ssDNA, independently of or in concert with DprA. Results of in vitro experiments with SpDprA, EcRecA, and SSB established the formation of pure RecA nucleofilaments when the loading of RecA on SSB-coated ssDNA was stimulated by DprA. In contrast, mixed DprA-RecA nucleofilaments were obtained in the absence of SSB (Mortier-Barrière et al., 2007). As stated above, it is necessary to characterize in vitro the interplay of SpDprA, SpRecA, and SpSsbB to establish whether these observations apply to the homospecific system.
  • 2The existence of SpSsbB–ssDNA complexes (i.e. EC) suggests that an alternative early function of SpSsbB might be in the protection and storage of ssDNA in competent cells (the reservoir model; Fig. 4, left part). In this model, DprA could interact first with nascent ssDNA and favor the immediate loading of RecA to initiate the formation of recombinants (Fig. 4, right part). When DprA and SpRecA molecules engaged in processing become available again, SpRecA could access the SpSsbB–ssDNA reservoir, assuming that DprA can alleviate the SpSsbB barrier to load SpRecA.
  • 3At a later stage in transformation, SpSsbB could prevent the loading of SpRecA on the displaced parental strand (i.e. the nonpaired strand in D-loops) once donor–recipient heteroduplexes are formed, much as documented with SSB and EcRecA (Lavery & Kowalczykowski, 1992), and with SpSsbA and SpRecA (Steffen et al., 2002).
  • 4At any stage during the process, SpSsbB might favor the recruitment of additional specific protein effectors on internalized ssDNA and/or on recombination intermediates. The latter hypothesis is suggested by the physical interaction observed between SsbB and CoiA, both in B. subtilis and in S. pneumoniae. Although the function of CoiA is unknown, the two-log reduction in transformation in its absence indicates that it plays an important role. As formation of EC readily occurs in coiA mutant cells (Desai & Morrison, 2007), the protein is presumably involved at a late stage in processing.
Figure 4.

 Diagrammatic representation of the interplay of DprA, RecA, and SsbB. For clarity, only two pathways are illustrated, the direct loading of RecA on nascent ssDNA assisted by DprA and the loading of SsbB on incoming ssDNA to create a reservoir for further access by RecA/DprA. See text for alternative possibilities.

Are both SSBs involved in the processing of internalized ssDNA?

The existence of SsbB does not necessarily exclude SsbA from the transformation process. The identification of SpSsbA at the position of EC during hydroxylapatite chromatography (Morrison et al., 2007) suggests that this protein may also bind incoming ssDNA. However, one cannot exclude that the presence of SpSsbA simply reflects the capture of replication intermediates (on which the essential SsbA is likely to be bound) that presumably behave similarly to transformation intermediates during hydroxylapatite chromatography. Further analysis of EC in ssbB mutant cells and hydroxylapatite chromatography analysis of extracts from noncompetent cells as well as from nontransformed competent cells (i.e. cells that did not internalize DNA) could establish whether SsbA participates in the process of transformation.

If both SSBs turned out to be involved in the processing of transforming DNA and to fulfill specific roles, which feature could account for this specificity? An obvious possibility would lie in their distinct C-terminal domains (Fig. 3). It is well known that SSBs specifically interact with some partners through their C-termini [e.g. the documented interaction of BsSsbA with PriA, RecG, and RecQ (Lecointe et al., 2007) or of E. coli SSB with exonuclease I (Lu & Keck, 2008 and references therein)]. SsbB might have evolved a protein partnership differing from that of SsbA, thus enabling each SSB to contribute to different aspects of DNA processing during transformation.

Why do deletions of dprA or recA result in ssDNA degradation if SSB is protecting it?

The identification of SpSsbB in EC suggests an early interaction of this protein with nascent ssDNA. However, this conclusion is not consistent with the other finding that in the absence of RecA or DprA, ssDNA is very rapidly degraded (Bergéet al., 2003). The latter observation suggests that SpSsbB cannot access ssDNA when any of the two proteins is missing. A possible explanation would be that SpSsbB requires the assistance of DprA or RecA for proper subcellular localization. Alternatively, a DprA/RecA-controlled switch between protective and degradative DNA modes of uptake might exist. DprA/RecA could thus control the activity of a nuclease located at or near the entry pore. This switch would allow toggling of the uptake machinery into a degradative mode of uptake, possibly contributing to the replenishment of nucleotide pools. Switching might occur when the amount of ssDNA internalized becomes in excess over the processing capacity of DprA/RecA. This possibility would not necessarily be antagonistic with the SpSsbB–ssDNA reservoir model as SpSsbB could protect only a fraction of internalized ssDNA.

An alternative model would be that nucleation of SpRecA aided by DprA occurs on nascent ssDNA emerging from the transmembrane channel ComEC, followed by polymerization of RecA toward the 3′-extremity. DprA/RecA would thus prevent DNAses from accessing the 3′-end of the incoming ssDNA molecules. SpSsbB might then be required to protect the middle of internalized ssDNA molecules from access by an endonuclease.

A transformation machine

Interactions between proteins involved in DNA transport were first obtained during the characterization of the uptake machinery. More recently, interactions between proteins involved in the processing of transforming DNA were also detected, suggesting the existence of a DNA processing machinery. Moreover, during the last 4 years, a picture of a large multiprotein polar complex involving both DNA-processing proteins, which are primarily cytosolic, and DNA transport proteins, which are mostly membrane bound, has slowly emerged. We collected from the literature the available information (listed in Table 3 with the corresponding references) supporting this picture and assembled it into a static interaction network shown in Fig. 5 (bottom). This information is based on molecular and biochemical data (Steffen et al., 2002; Morrison et al., 2007; Mortier-Barrière et al., 2007) on observations showing a preferential polar localization of four processing proteins, CoiA, DprA, RecA, and SsbB, as well as on the demonstration of the colocalization of the same four processing proteins with ComGA and/or ComFA (Hahn et al., 2005; Kidane & Graumann, 2005; Kramer et al., 2007; Tadesse & Graumann, 2007). The latter point is illustrated by the colocalization of SsbB and ComGA in B. subtilis cells coexpressing a ComGA-CFP and an SsbB-YFP constructs (Fig. 5, top). Information on possible interactions was also deduced from results regarding stability or localization dependencies of proteins in the machine (Kidane & Graumann, 2005; Kramer et al., 2007; Tadesse & Graumann, 2007), from results of Förster (or fluorescence) resonance energy transfer (FRETS) analyses (Kramer et al., 2007), from the existence of genetic relationships (Bergéet al., 2003), from pull-down experiments (Kramer et al., 2007; Mortier-Barrière et al., 2007), and finally from Y2H (Mirouze, 2007; Mortier-Barrière et al., 2007).

Table 3.   Direct and indirect evidence for interactions between components of the transformasome
Type of evidence
Molecular or
Polar localization or
Stability or localization
CoiA (4)(4)    
ComFA (2)(4)    
ComGA (2), (3), (9)(2)    
DprA (4), (9)(4), (9)    
RecA (3), (9)(3)    
SsbB (2)     
ComEC-ComEA  (4)  (4) 
ComFA-CoiA (4) (4)   
ComFA-ComEC  (4)    
ComFA-DprA (4) (4)   
ComFA-RecA   (4)   
ComGA-CoiA (4) (4)   
ComGA-ComFA (2)(4)(4)   
ComGA-DprA (4), (9)     
ComGA-RecA (3)(3)(3), (4)   
ComGA-SsbB (2) (4)   
CoiA-SsbB (4) (4)  (5)
CoiA-RecA (4)     
DprA-RecA (4) (4)(1)(7)(5), (7)
DprA-SsbB (4)(4)(4)   
RecA-SsbB   (4)   
Figure 5.

 Interaction network suggesting the existence of a transformasome. (Top) Illustration of the colocalization of SsbB and ComGA in Bacillus subtilis cells coexpressing a ComGA-CFP and an SsbB-YFP constructs [reproduced with permission from Hahn et al. (2005, fig. 3e). (Bottom) The network was constructed using available information (see Table 3) inferred from (b) molecular or biochemical data, (c) colocalization of fluorescent signals, (d) stability or localization dependencies, (f) FRET analysis, (g) the existence of genetic relationships, (p) pull-down data, or (y) Y2H. Note that this network does not include genetic and biochemical data concerning interactions between components of the DNA uptake machine [see Dubnau (1999); Chen & Dubnau (2004); Chen et al. (2005)].

It is of note that some of the fusion proteins used in these studies were nonfunctional (e.g. B. subtilis SsbB-YFP; Hahn et al., 2005). It should also be made clear that the mechanistic significance of several of the interactions illustrated in this Fig. 5 remains to be established. Nevertheless, altogether these data provide support to the existence of a unique machine to internalize, protect, and process transforming DNA into recombination products.

To refer to this unique but dynamic entity comprising the DNA uptake and the DNA processing, we propose to use the word ‘transformasome’, despite its previous use with a different meaning in the mid-1980s. This name was introduced by Smith and coworkers to designate specialized membranous structures (vesicles) that protect DNA during H. influenzae transformation (Kahn et al., 1983). A search through PubMed revealed that, since its introduction, this term was used in only four publications, from 1983 to 1986, and that no other publication involving the study of these specialized structures had appeared since then. In view of this situation, we think that the change of meaning we propose is acceptable.

Perspectives on the transformasome: from mechanism to interconnection(s) with the cell cycle

Some progress in our understanding of the processing of transforming DNA can be expected from the identification and the characterization of nucleoprotein complexes and/or protein assemblies that would depend on the internalization of ssDNA (DNA processing ‘in progress’). New insights into the interplay of DprA, SsbB, and RecA may also give a clue as to the raison d'être of transformation-dedicated RMP and SSB. They could help answer the question of the selective pressure(s) that resulted in the evolution of dedicated proteins in place of the use of proteins already present in vegetative cells (e.g. RecFOR and SSB), and they could possibly reveal whether the dedicated proteins fulfill different biological roles, for example genetic exchanges (for DprA) vs. recombinational repair (for RecFOR). The unraveling of the role of specific proteins (e.g. CoiA or RadA) in late processing, as well as the possible identification of resolvases and/or ligases that would be required for the resolution of transformation intermediates (whereas none has been identified so far in any species) can also be anticipated.

Besides these aspects, an appealing perspective resides in the investigation of the subcellular localization of the transformasome in S. pneumoniae, as initiated in B. subtilis (Hahn et al., 2005; Kidane & Graumann, 2005). Does transforming DNA uptake take place at the cell poles in cocci similarly to rod-shaped cells, which would suggest that this preferential location is a general feature in naturally transformable species? It would also be interesting to establish whether chromosomal integration events take place close to the site of DNA entry or in a more central subcellular location. Dynamic threads of RecA-GFP have been observed in transformed cells of B. subtilis (Kidane & Graumann, 2005), but it remains unknown whether they result from the trafficking of internalized ssDNA from the cell poles toward the nucleoid or from chromosome subdomains moving toward the poles, to allow completion of homologous recombination.

An even more appealing perspective lies in the characterization of the possible relationship between the transformasome and the cell cycle. There is already a hint in B. subtilis that ComGA, a protein essential for assembling the transformation pseudopilus, also functions as a checkpoint to limit growth until competence is over (Haijema et al., 2001). This question is even more important in the case of S. pneumoniae, which develops competence in exponentially growing (i.e. actively replicating/dividing) cells.


A long-standing question is that of the biological role(s) of natural transformation. Among the selective pressures that maintain competence for DNA uptake, which of food gathering, chromosome repair or gene exchange is a major factor? Several weaknesses of the food gathering hypothesis have been already discussed (Dubnau, 1999; Claverys et al., 2006). Although the present-day role of transformation is presumably not nutritional, degradation of ssDNA could play an important role for repair, via temporary replenishment of nucleotide pools (Guiral et al., 2007). Concerning the DNA-for-repair hypothesis, the original proposal that internalized ssDNA could be used as template for the recovery of damaged chromosomal regions (Bernstein et al., 1985) was called into question (Dubnau, 1999). However, the alternative model that internalized homologous DNA could contribute to survival because D-loops formed during integration could prime new replication forks to help bypass chromosomal lesions (Guiral et al., 2007) would deserve examination. The third possibility, that natural transformation plays a role in generating genetic diversity (Dubnau, 1999), is particularly attractive in the case of S. pneumoniae (Guiral et al., 2007), which, considering only the 90 capsular biosynthetic loci characterized in the species (Bentley et al., 2006), can readily access by transformation >1.8 Mbp alternative coding (i.e. almost the equivalent of an individual genome size c. 2.2 Mbp). Progress in the knowledge of the transformasome will probably bring additional arguments to fuel the debate on the biological role of transformation. Already, the existence of a unique integrated machine – the transformasome – to internalize, protect, and process transforming DNA into recombination products in itself provides support to the DNA-for-repair and DNA for genetic diversity hypotheses.


We thank Nathalie Campo and Marc Prudhomme for critical reading of the manuscript, the laboratory members for stimulating discussions, and Chris Bayliss for helpful suggestions. We also wish to thank the members of our teams who have contributed over the years to develop research on genetic transformation of S. pneumoniae. Our partnership with the laboratories of Philippe Noirot (Laboratoire de Génétique Microbienne, INRA, Jouy-en-Josas, France) and of Eric Le Cam (Laboratoire de Microscopie Moléculaire et Cellulaire, Institut Gustave Roussy, Villejuif, France) was instrumental for the characterization of DprA. This work was supported in part by grants from the Ministère Déléguéà la Recherche et aux Nouvelles Technologies (Programme Microbiologie 2003–2004; ref: RB/CD/2003/09/001) and from the Agence Nationale de la Recherche (projet n_ BLAN06-3_141806).


Patrice Polard is the 2nd corresponding author for this article.