The process of DNA donation for natural transformation of bacteria is poorly understood and has been assumed to involve bacterial cell death. Recently in Neisseria gonorrhoeae we found that mutations in three genes in the gonococcal genetic island (GGI) reduced the ability of a strain to act as a donor in transformation and to release DNA into the culture. To better characterize the GGI and the process of DNA donation, the 57 kb genetic island was cloned, sequenced and subjected to insertional mutagenesis. DNA sequencing revealed that the GGI has characteristics of a horizontally acquired genomic island and encodes homologues of type IV secretion system proteins. The GGI was found to be incorporated near the chromosomal replication terminus at the dif site, a sequence targeted by the site-specific recombinase XerCD. Using a plasmid carrying a small region of the GGI and the associated dif site, we demonstrated that this model island could be integrated at the dif site in strains not carrying the GGI and was spontaneously excised from that site. Also, we were able to delete the entire 57 kb region by transformation with DNA from a strain lacking the GGI. Thus the GGI was likely acquired and integrated into the gonococcal chromosome by site-specific recombination and may be lost by site-specific recombination or natural transformation. We made mutations in six putative type IV secretion system genes and assayed these strains for the ability to secrete DNA. Five of the mutations greatly reduced or completely eliminated DNA secretion. Our data indicate that N. gonorrhoeae secretes DNA via a specific process. Donated DNA may be used in natural transformation, contributing to antigenic variation and the spread of antibiotic resistance, and it may modulate the host immune response.
Natural genetic competence, the ability to take up macromolecular DNA and incorporate it into the genome, is a widespread phenomenon that occurs in both Gram-negative and Gram-positive, as well as pathogenic and non-pathogenic bacteria. There are a number of theories as to why natural transformation has evolved and persisted in so many species, including the use of DNA as a food source, for DNA repair, and to generate diversity (Dubnau, 1999). Neisseria gonorrhoeae is one of 44 known naturally competent bacterial species (Lorenz and Wackernagel, 1994). N. gonorrhoeae, Haemophilus influenzae and many other Gram-negative bacteria take up only species-specific DNA for transformation. For the gonococcus, transforming DNA is recognized by the presence of a 10 bp DNA uptake sequence (Goodman and Scocca, 1988). Uptake of specific DNA is more consistent with a role for transformation in genetic variation or DNA repair rather than in acquisition of nutrients. In fact, natural transformation is the only mechanism by which horizontal transfer of chromosomal markers occurs in N. gonorrhoeae, thus it accounts for the emergence of mosaic alleles for single-copy genes of the gonococcal genome (Koomey, 1998). Transformation is a significant mechanism for pilin antigenic variation in gonococci (Seifert and So, 1988; Gibbs et al., 1989), for generation of new allelic combinations that may contribute to gonococcal porin and Opa diversity (Hobbs et al., 1994; Fudyk et al., 1999), and for the spread of antibiotic resistance markers (Sparling, 1966).
As more bacterial genome sequences become available, there is increasing evidence for the presence of horizontally acquired genomic islands among bacteria. These islands, which may be present on the chromosome of an organism or on plasmids, often confer a selective advantage such as enhanced pathogenicity, metabolism, symbiosis, or ecological fitness (Hacker and Kaper, 2000). Genomic islands are frequently flanked by small directly repeated DNA sequences and often encode secretion systems (Hacker and Kaper, 2000). We previously reported that approximately 80% of N. gonorrhoeae strains carry a gonococcal genetic island (GGI). The 8 kb region of the GGI that had been sequenced showed a lower G + C content than the chromosome of N. gonorrhoeae, suggesting horizontal acquisition (Dillard and Seifert, 2001). The sequence of the GGI is variable, and certain forms were found significantly more often in strains isolated from patients with disseminated gonococcal infection (Dillard and Seifert, 2001).
Mutations in three genes in the GGI resulted in decreased DNA donation for transformation and decreased DNA release into the medium (Dillard and Seifert, 2001; Hamilton et al., 2001). Two of these genes, traG and traH, are similar to type IV secretion system (T4SS) genes. T4SS genes are often identified by their homology to the transfer genes of conjugative plasmids or the Ti plasmid of Agrobacterium tumefaciens (Christie, 2001). GGI traG and traH are similar to genes of the Escherichia coli F-plasmid (Dillard and Seifert, 2001), whose transfer system is a T4SS that facilitates conjugation of the F-plasmid or conjugal transfer of the host cell chromosome by Hfr strains (Lawley et al., 2003). The T4SS of A. tumefaciens has been shown to transfer oncogenic Ti plasmid DNA and several proteins from the bacterium directly into a plant cell, where Ti plasmid gene expression by the host cell results in the development of crown gall disease. A number of human pathogens are known to utilize T4SSs during infection. For example, Bordetella pertussis secretes pertussis toxin via its T4SS encoded by the ptl genes; Legionella pneumophila encodes a T4SS that secretes factors necessary for intracellular survival within macrophages and amoebae; and Helicobacter pylori secretes the CagA protein directly into host cells via a T4SS. T4SSs have also been identified in Rickettsia prowazekii, Brucella spp. and several other pathogenic and non-pathogenic bacterial species (Christie, 2001).
In this manuscript we describe the GGI and demonstrate that it encodes a T4SS in addition to many genes with no known function. Mutations constructed in five putative secretion genes in the island disrupted DNA secretion by N. gonorrhoeae. However, mutations in the gene encoding the putative coupling protein homologue did not eliminate DNA secretion. Additionally, mutation of a gene for a chromosome partitioning protein homologue eliminated DNA secretion, whereas mutations in other genes in or near the GGI had no effect. These data demonstrate that the gonococcus utilizes the GGI-encoded T4SS for the active secretion of chromosomal DNA. Additionally, we describe mechanisms for the gain and loss of this variable genetic island.
Sequence of the GGI
The GGI was cloned from the gonococcal chromosome of strain MS11A using a chromosome walking procedure. This process resulted in the cloning of the previously unknown portion of the GGI in 11 plasmid clones (Fig. 1, Table 2). The DNA sequence revealed that the region encodes 61 open reading frames (ORFs). The most striking characteristic of the sequence is the presence of multiple homologues of T4SS genes (Fig. 1). The first 27.5 kb of the GGI encodes 24 ORFs, 18 of which show significant similarity to the transfer genes of the well known E. coli F-plasmid or other T4SSs (Table 1) and 15 of which are ordered the same as those of the F-plasmid (Fig. 2). Three genes (traD, traI, ltgX) similar to genes of F-plasmid are found in both F-plasmid and the GGI, but in a different relative location. Although the region shows significant similarities to F-plasmid, there are also substantial differences. The GGI carries eight genes interspersed among the transfer genes that show no homology to F-plasmid genes, and F-plasmid carries 19 genes in its transfer region that do not have homologues in the GGI. Genes of the F-plasmid that have no homologues in the GGI include those involved in transcriptional regulation and those that prevent DNA transfer to an F+ strain, so-called surface exclusion genes.
MS11AΔGGI constructed by transformation of HH522 with pARM3
ND500 transformed with pJS3
MS11A transformed with pWSP7; traD Walker box deletion
Table 1. Annotation of ORFs located within the GGI.
Homologue of putative protein
Function of homologue
Salmonella typhi TrwB
Putative docking protein
Xyllela fastidiosa XF1753
Putative nicking enzyme
Escherichia coli plasmid R64 PilT
Bacteroides fragilis OmpA
Outer membrane protein
Pseudomonas resinovorans TrhA
Putative transfer protein
Novosphingobium aromaticivorans pNL1 TraL
Shigella flexneri plasmid R100 TraE
N. aromaticivorans pNL1 TraK
Vibrio cholerae SXT TraB
N. meningitidis DsbC
Protein disulphide isomerase
S. typhi TrhV
Putative transfer protein
N. aromaticivorans pNL1 TraC
N. aromaticivorans pNL1 Orf851
S. flexneri plasmid R100 TraW
N. aromaticivorans pNL1 TraU
E. coli F-plasmid TrbC
S. typhi TrhN
E. coli F-plasmid TraN
V. cholerae TraF
E. coli F-plasmid TraH
E. coli F-plasmid TraG
Pilus assembly/mating-pair stabilization
Phage lambda R
Enterococcus faecalis EF0781
RNA/ssDNA binding protein
E. coli YhaV
Synechocystis sp. SohA
Shigella sonnei plasmid ColIB YccB
Pseudomonas aeruginosa COG0286
Putative DNA methylase
Burkholderi fungorum Bcep3609
Putative DNA methylase
Xanthomonas axonopodis ORF8
Ralstonia solanacearum RSp1626
Haemophilus influenzae HI0420
Borellia burgdorferi BdrW
Repeat containing protein
X. axonopodis XAC2237
Pasturella multocida PM0271
S. typhi TopB
X. fastidiosa XF1778
Single-stranded DNA binding
P. aeruginosa Orf SG99
Anabaena variabilis COG0810
Putative TonB-like transporter
X. fastidiosa XF1784
X. fastidiosa XF1785
Like many of the F-plasmid transfer genes (Frost et al., 1994), 17 of the GGI genes have either overlapping coding regions or are separated by only a few base pairs. Overlapping of coding regions may facilitate the coupled translation of co-transcribed genes and is thought to be a property of horizontally acquired DNA (Lawrence and Roth, 1996). Except for the divergently transcribed genes, most of the transfer region of the F-plasmid is transcribed as a single message from traY to traI (Helmuth and Achtman, 1975). Although we do not have data on the transcription of the GGI transfer genes, the sequence suggests at least three transcripts for the T4SS genes. As traD, traI and yaf are oriented opposite from the remaining genes, these would necessarily be on a separate transcript. The genes from ltgX to ych are in the same orientation and could be in a single transcript similar to that of the F-transfer region. However, a near consensus σ70 type promoter is found preceding traH that would be expected to transcribe from traH to ych. The sequence following ych contains an inverted repeat that could form a transcriptional terminator (Dillard and Seifert, 1997).
In addition to overlapping ORFs, the GGI exhibits other sequence characteristics that suggest it was horizontally acquired from another bacterial species. The G + C content of the GGI (44%) is significantly lower than the remainder of the chromosome (51%). It also shows differences in frequency of certain dinucleotides relative to the rest of the genome, a common characteristic of horizontally acquired DNA (Karlin, 1998). The N. gonorrhoeae genome is known to have an overrepresentation of CG and GC dinucleotides (ρCG* = 1.32 and ρGC* = 1.26) and an underrepresentation of TA dinucleotides (ρTA* = 0.63) (Campbell et al., 1999). The dinucleotide bias is not seen in the GGI for CG (ρCG* = 0.98) and is seen to a lesser extent for GC and TA (ρGC* = 1.18 and ρTA* = 0.74). Furthermore, the sequence has few copies of the gonococcal DNA uptake sequence, a 10-base sequence found frequently in the gonococcal chromosome which, as mentioned previously, is necessary for uptake of DNA during natural transformation (Goodman and Scocca, 1988). There are only six copies of the DNA uptake sequence in the 57 kb GGI (Fig. 1), whereas in the 57 kb flanking the GGI insertion site there are 53 copies (GenBank accession ♯AE004969).
Outside the putative type IV secretion system region of the GGI are 35 ORFs oriented opposite in direction to that of most of the T4SS genes (Fig. 1). Seventeen of these show no similarity to known sequences. Seven are predicted to encode hypothetical proteins of unknown function (Table 1). Another three show similarity to a putative protease (YdcA), a putative N-acetyltransferase (Yeb), and a putative TonB homologue (Yfd). However, the remaining eight are similar to DNA binding or DNA processing proteins. These include a helicase (Yea), topoisomerase (TopB), single stranded binding protein (SsbB), two DNA methylases (Ydg and YdhA), two chromosome partitioning proteins (ParA and ParB) and a cold shock-like protein (CspA).
Phase variation of many proteins of N. gonorrhoeae occurs by slipped-strand mispairing in homonucleotide tracts or other short nucleotide repeats (Belland et al., 1989; Saunders et al., 2000). We found that two putative T4SS genes carried homonucleotide repeats. An A8 repeat occurs near the beginning of traK, within the coding region. A T8 tract occurs within the traL coding region, near the 3′ end. Dinucleotide repeats were found in and around the yfd gene. A (TG)4 repeat occurs between yfeA and yfd. Deletion of a single repeat would result in the fusion of yfeA and yfd into a single ORF. Repeats of (GT)4 and (CT)5 occur near the centre of the yfd coding sequence. Although each of the repeat regions is large enough to predict that variation might occur in the numbers of repeats, further work will be required to determine if Yfd, TraK and TraL are phase-variable.
GGI location and presence in N. gonorrhoeae strains
The GGI in strain MS11A is flanked by a direct repeat of 23 bp. This sequence shows a high degree of similarity to the dif sites of H. influenzae and E. coli(Fig. 3). In E. coli, the dif site is found in the chromosome replication terminus. The E. coli dif site is recognized by the site-specific recombinase XerCD whose function is to separate two chromosomes following an odd number of recombination events during replication (Hill, 1996). Similarly the GGI insert site is located in the gonococcal terminus region (Dempsey et al., 1995). This finding suggests that the GGI may have been inserted into the chromosome by site-specific recombination, potentially using the gonococcal XerCD.
We previously found that GGI genes were present in approximately 80% of gonococcal strains (Dillard and Seifert, 2001). Because the ends of the GGI had not yet been identified, it was not known whether the GGI was located at a single site in all strains, in different places on the chromosome, or on the conjugal plasmids found in some N. gonorrhoeae strains (Eisenstein et al., 1977). Therefore, we examined several commonly used laboratory strains as well as several low-passage clinical isolates of N. gonorrhoeae for the presence of the GGI at the expected chromosomal site (Fig. 4). Polymerase chain reactions (PCRs) specific for the presence (77F to 86R) and absence (73F to hlh-ggiR) of the GGI were performed. Gonococcal strains MS11A, JC1, RD5, VP1, FA19, CV-1, CS7, UT38794, UT39260, NRL8090 and NRL2025 all contain the GGI; strains FA1090, 1291, F62, 24-1, UT38884 and DGI43 do not. All of the strains that had been found previously to contain the internal GGI gene traG (Dillard and Seifert, 2001) were positive for the presence of traD at the junction site, establishing that the GGI is found at the same location in all strains that contain it. Sequence analysis of the dif site and imperfect dif site revealed that these sequences are identical in 12 low passage, GGI+ clinical isolates (data not shown). All strains that were negative for the GGI+-specific PCR gave a product of the expected size for the GGI–-specific PCR, indicating that they do not contain the GGI or other inserted DNA at this location.
Gain and loss of the GGI
We hypothesized that the GGI could be gained or lost from a gonococcal strain by two mechanisms: site-specific recombination at dif or natural transformation followed by homologous recombination. To determine if the GGI might be spontaneously lost from the chromosome by site-specific recombination, we screened for the loss of a marked version of the GGI from strain HH522, which carries a productive PhoA fusion (exp1::mTnCmPhoA). In this screen, loss of the GGI would produce white colonies on XP indicator plates. In multiple attempts, no white colonies were found among greater than 104 colonies, suggesting that the GGI was not lost at a frequency that could be detected by this screen.
The natural competence of the gonococcus suggested that the GGI might be lost through transformation with DNA from a strain lacking the GGI. To test this hypothesis, a plasmid was constructed that carried the dif site and its flanking region from the naturally GGI– strain FA1090. This plasmid, pARM3, was linearized and used to transform the GGI+ strain HH522. Among 704 colonies, eight white (PhoA–) colonies were found; seven of these were also CmS. One maintained chloramphenicol resistance, indicating that it had not lost the mTnCmPhoA or the GGI. Characterization of one of the PhoA–, CmS isolates showed that it had indeed lost the GGI (Fig. 5). Using DNA from this strain, designated ND500, it was possible to amplify a PCR product from the left-flanking DNA to right-flanking gene ung, the GGI–-specific PCR described above. We did not obtain a product when attempting to amplify the GGI-internal gene traD from ND500, nor were we able to amplify from parA to ung. The deletion strain was further analysed by Southern blotting. Chromosomal DNA digested with BsrGI and probed with DNA from the FA1090 dif site and surrounding region gives two bands for MS11A, because this region is separated by the GGI. However in ND500, as in FA1090, only one band is seen, confirming that ND500 has indeed lost the GGI (Fig. 5C).
We used the GGI deletion strain ND500 to test for gain of the GGI by natural transformation. The strain was marked with spectinomycin resistance (ND517) and co-cultured with ND516 (HH522 recA6) with aeration. If the island were introduced into ND517, SpRCmR transformants would arise; however, no SpRCmR transformants were obtained. The coculture was repeated in static liquid culture to minimize DNA fragmentation that might occur in culture. Again, no SpRCmR transformants were obtained, suggesting that gain of the GGI might not occur by natural transformation or that the frequency of the event is too low to detect by these methods.
The absence of the GGI in a minority of strains and the presence of the dif site and dif-like site at the GGI ends suggested that the GGI might be gained or lost by site-specific recombination. To test this hypothesis, a plasmid, pJS3, was constructed to act as a model version of the GGI. This plasmid carries the dif site and a short region of 90 bp internal to the GGI and cannot autonomously replicate in N. gonorrhoeae. The only homologous regions shared by this plasmid and the ND500 chromosome are the 28 bp dif site and the two copies of the 10 bp DNA uptake sequence. To determine if this model version of the GGI would integrate into the chromosome, ND500 was transformed with pJS3 in dimeric form. It should be noted that the process of natural transformation brings in linear DNA so that if no homology is provided by the chromosome or by a second transformation event in the same cell, a circle cannot be reformed. However, transformation with dimeric plasmid allows homologous recombination to form a circular monomer within the cell (reviewed in Stewart and Carlson, 1986). Transformation of ND500 with a dimer of pJS3 yielded erythromycin-resistant transformants, and PCR confirmed that the plasmid had recombined into the dif site in each transformant tested (data not shown). The location of the insertion in one transformant, ND518, was further confirmed by Southern blotting (Fig. 6). Chromosomal DNA from ND518 showed hybridization of pIDN1 (empty vector precursor to pJS3) to the undigested chromosome (lane 3) and to bands of 11 kb and 13 kb in BsrGI-digested DNA (lane 1), consistent with one or two copies of pJS3 inserted at the dif site. The free forms of the plasmid seen in this blot are discussed below. No hybridization of the plasmid to ND500 chromosomal DNA was detected, demonstrating that the hybridization specifically detected pJS3.
Interestingly, the dif-containing plasmid insertion was rapidly lost from the chromosome; without selection only 40% of colonies maintained erythromycin resistance and the insertion after overnight growth. This observed frequency of loss is much higher than that expected from homologous recombination. For example, we were unable to detect the loss in 105 colonies of a similar inserted plasmid that carried 900 bp of homologous DNA (Hamilton et al., 2001). Southern analysis of undigested ND518 chromosomal DNA clearly showed bands for free forms of the plasmid (Fig. 6). The most prominent bands had identical mobility to the circular monomer and linear monomer of pJS3. Because the pJS3 origin allows replication in E. coli and does not allow replication in N. gonorrhoeae, the free forms seen in ND518 must represent inserts that have been excised from the chromosome. If XerCD excised pJS3 from the chromosome, the circular monomer would be the expected form. Linear or nicked monomer may represent an intermediate form of the plasmid that occurs upon action of XerCD. However, it is also possible that this form of the monomer is generated during purification of the DNA for the Southern blot. Faint bands for additional forms of pJS3 are also seen, showing identical mobilities to circular dimers and tetramers. These could result from excision of a multimer insertion from the chromosome or multimerization of pJS3 monomers after excision. The ability to easily detect free forms of the model GGI and the high-frequency loss of the erythromycin resistance marker suggest that XerCD acts to remove the insertion from the chromosome.
Mutation of T4SS genes affects DNA secretion
To address the role of GGI-encoded genes in secretion, we made mutations in several genes predicted to be required for type IV secretion. Using a fluorescence-based DNA secretion assay as previously described (Hamilton et al., 2001), the mutants were assessed for their ability to secrete DNA during normal growth. We previously demonstrated that mutations in three genes, traH, traG and atlA, eliminate DNA secretion by N. gonorrhoeae (Dillard and Seifert, 2001; Hamilton et al., 2001).
In this study, initial mutations were designed to be polar on downstream genes to increase the chances of eliminating secretion. Polar mutations were made in traC, traN and traF and each of these mutations eliminated DNA secretion (Fig. 7). traF is predicted to be at the end of a transcript and therefore, may not have polar effects on downstream genes; complementation confirmed this prediction. Insertion mutations designed to be effectively non-polar were made in traN, traI and dsbC. These insertions contain a promoter that will drive transcription of genes downstream of the insertion (Hamilton et al., 2001). TraN is similar to mating-pair stabilization proteins of conjugation systems and may form part of the secretion channel (Table 1) (Frost et al., 1994). TraI is predicted to act in nicking the DNA prior to transport. A disulphide bond isomerase similar to the GGI DsbC was recently shown to be required for secretion of pertussis toxin (Stenson and Weiss, 2002). Mutations in traN, traI and dsbC diminished DNA secretion to background levels. Complementation of the non-polar mutants of traN, dsbC and a previously described mutant of traH restored DNA secretion (Fig. 7).
An insertion mutation in traD, the gene encoding the docking protein homologue, resulted in an intermediate level of DNA secretion. Docking proteins of T4SSs, also known as coupling proteins, are thought to recognize substrates for secretion. The insertion mutation constructed truncates traD, eliminating coding sequence for 317 of the 820 amino acids (38%). Thus it was possible that the truncated TraD was partially functional. A second possibility is that TraD is not required for DNA secretion and that the interruption of traD might instead have affected the stability of its transcript and expression of the adjacent gene traI, which we have shown to be necessary for type IV secretion of DNA. Therefore we made a 33 bp in-frame deletion in traD, eliminating the predicted nucleotide binding motif (Walker box). Walker-box mutations have been shown to eliminate function of multiple T4SS docking proteins (Balzer et al., 1994; Moncalian et al., 1999). However, the traD deletion mutation did not diminish DNA secretion. These results suggest that either the gonococcal TraD functions differently from other docking proteins or that it is not required for DNA secretion.
Although the parA gene is not found in the T4SS region of the GGI, a partitioning protein homologue, VirC1, was shown to be necessary for T-DNA transport in A. tumefaciens (Yanofsky et al., 1985). A point mutation constructed in the Walker box sequence of parA resulted in a loss of DNA secretion, and complementation confirmed its role in gonococcal DNA secretion.
Mutations were also made in genes not predicted to be required for secretion. The GGI gene, ydbA, which shows no homology to known sequences, was disrupted with a polar insertion. Similarly, a gene found just outside the GGI, ung, encoding the gonococcal uracil-N-DNA glycosylase, was disrupted with a polar insertion. Mutation of ydbA or ung had no effect on DNA secretion.
These data indicate that the encoded proteins TraF, TraN, DsbC, TraI and ParA are necessary for type IV secretion of DNA, while TraD is not. TraC may also be necessary, but the polarity of the mutation may have effects on other genes as well. Furthermore, YdbA and Ung are dispensable for DNA secretion.
Determination of strain viability
Neisseria gonorrhoeae is an autolytic organism; that is, it encodes a number of peptidoglycan hydrolysing enzymes that are capable of lysing the cell. Autolysis normally occurs during stationary phase or during non-growth conditions in vitro, such as non-optimal temperature, pH, or osmolarity (Morse and Bartenstein, 1974; Hebeler and Young, 1975). This proclivity and the previous findings that a GGI-encoded peptidoglycan hydrolase, AtlA, contributes to cell death in late stationary phase culture (Dillard and Seifert, 1997) compelled us to examine the viability of gonococcal cultures to ascertain whether the deficiency in DNA secretion by the mutants is due to a reduction in autolysis. Previous studies had shown that release of the cytoplasmic protein CAT was not different between the wild-type strain and an AtlA-defective T4SS mutant during log phase or early stationary phase culture (Dillard and Seifert, 2001). However, it was possible that a small reduction in autolysis in the mutants might be sufficient to give the observed difference in coculture transformation, but would be missed by the CAT assay. It has been shown that lysed gonococci maintain morphology similar to living cells in both growth medium and in magnesium-containing buffers (Wegener et al., 1977). This fact allowed us to directly examine the viability of gonococci in culture using live/dead (green/red) staining and fluorescence microscopy. Stained cultures of wild-type strain MS11A appeared similar to those of the mutants (Fig. 8A). To further validate these findings, quantification of the fluorescence of these strains revealed that the viability of MS11A is similar to that of the T4SS mutants, or slightly better (Fig. 8B). In log-phase cultures of MS11A, an average of 1.4% of gonococci stained red, while in mutant cultures, 2.2% of bacteria stained red. These data indicate that the difference in levels of DNA in the medium is not due to a difference in lysis or death of gonococci, but rather due to a functional secretion system.
We have shown that the GGI encodes a T4SS whose components are necessary for DNA secretion by N. gonorrhoeae. Eight separate mutations in T4SS genes eliminated DNA secretion into the medium. We have previously shown that secreted DNA is active in transformation of recipient gonococci (Dillard and Seifert, 2001; Hamilton et al., 2001). This is the first example of a mechanism for donation of DNA for natural transformation that does not require cell lysis and death. One apparent advantage of this process is that a cell can spread its genetic material horizontally through the population without sacrificing vertical transmission of genes to its progeny.
The similarities of the gonococcal T4SS to conjugation systems and other T4SSs, together with our mutation data, allow us to make predictions about the functions and locations of the T4SS proteins. Mutations in the GGI traH, traF, traN and traC presumably disrupt the function or the assembly of a secretion apparatus as the encoded proteins are homologous to the structural proteins or pilus biogenesis proteins necessary for transfer of E. coli F-plasmid. The polar insertions in traN and traC may affect co-transcribed genes by the nature of the mutation. However, polar mutation of traF may have little effect on any co-transcribed genes because a near consensus promoter is found between traF and traH and complementation restored DNA secretion. Although the disulphide bond isomerase DsbC is not encoded in all T4SSs, DsbC of the gonococcal T4SS performs some function that is necessary for secretion of DNA by the apparatus. It may be that DsbC catalyses the formation of disulphide bonds for proper conformation of one or more gonococcal T4SS proteins, without which the apparatus fails to function. TraI and ParA may be performing DNA processing functions for type IV secretion. In E. coli, TraI nicks and unwinds the F-plasmid beginning at oriT and covalently attaches to the 5′ end of the transferred DNA strand. Sequence homology would predict that the GGI TraI is a nicking enzyme. Gonococcal ParA is homologous to chromosomal partitioning proteins, and though it is not encoded in the same region of the GGI as the other necessary T4SS components, it also is necessary for DNA secretion. Chromosomal partitioning proteins are involved in localizing duplicated bacterial chromosomes to the one-quarter and three-quarters positions in the dividing cell, ensuring that each daughter cell receives a chromosome (Gerdes et al., 2000). It may be that gonococcal ParA escorts chromosomal DNA to the apparatus for secretion. Homologues of TraD in other T4SSs are predicted to be inner-membrane docking proteins that recognize type IV secretion substrates, or as is the case for F-plasmid, may pump the transferred DNA from the donor to the recipient cell (Frost et al., 1994). If TraD is performing a similar function in the gonococcal T4SS, mutation of the encoding gene would presumably eliminate DNA secretion. However, in our assay, a traD insertion mutant displayed an intermediate secretion phenotype and an in-frame deletion of the Walker box region did not decrease DNA secretion. These results indicate that TraD may not be necessary for DNA secretion. These data may suggest that gonococci encode an additional coupling protein that can interact with the T4SS, just as some plasmids encode coupling proteins for mobilization in the presence of heterologous conjugative plasmids (Hamilton et al., 2000). Although no such protein is easily identifiable in the FA1090 genome sequence, the chromosome of strain MS11A is some 110 kb larger than that of FA1090 and has the potential to encode many additional factors (Bihlmaier et al., 1991; Dempsey et al., 1991).
Although it has several similarities to conjugation systems, the gonococcal system also has some important differences. The gonococcal system secretes DNA into the medium. This is the first identified DNA secretion system that does not require cell contact for secretion. As gonococci efficiently take up DNA, direct cell-to-cell contact would not be necessary for DNA transfer. The only other T4SS that secretes in a non-contact dependent manner is the B. pertussis Ptl system which secretes the multisubunit protein pertussis toxin (Burns, 1999). Second, the DNA secreted is gonococcal chromosomal DNA. This might simply be due to the presence of the GGI in the chromosome rather than on a plasmid, and nicking may occur at an origin of transfer within the GGI, similar to the processing that occurs in an E. coli Hfr. However, it is also possible that the gonococcal chromosome has one or more origins of transfer dispersed around it.
The GGI appears to have been horizontally acquired and may be a mobile genetic element. The GGI has the characteristics of a genomic island (Hacker and Kaper, 2000), in that it is large, has a significantly different G + C content compared to the rest of the gonococcal chromosome, and it is flanked by direct repeats. The paucity of uptake sequences, as well as the low percentage G + C, suggests that the GGI is a relatively new element introduced into the genome of N. gonorrhoeae that has not yet adapted to match the rest of the genome. Though the GGI appears to only exist as a genetic element integrated into the chromosome, its similarity to conjugative plasmids suggests that it may have once been a plasmid. The similarity in the gene order of the GGI and E. coli F-plasmid (Fig. 2) supports this idea, suggesting that the two systems have a common ancestor. Furthermore, several of the ORFs in the non-T4SS region of the GGI are similar to proteins encoded on conjugative plasmids. Partitioning proteins similar to GGI ParA and ParB are often encoded on plasmids to ensure stable maintenance. The function of most of the non-T4SS genes of the GGI is unknown, but several show similarity to genes of virulence plasmids from different bacterial species. These may be remnants of plasmid replication or maintenance genes, or they might function in virulence in gonococci.
As approximately 20% of strains do not contain the GGI, these strains have either never acquired the GGI or have had the GGI excised from the chromosome. Because the repeated sequence at the ends of the GGI is similar to the sequence recognized by the site-specific recombinase XerCD, we speculate that the GGI is, or was once, a mobile element. It likely inserted into the dif site of N. gonorrhoeae by site-specific recombination, as we demonstrated for the dif-containing plasmid pJS3. The use of the dif site as an insertion point for horizontally acquired DNA has been noted for filamentous phages including Vibrio cholerae CTX-phi and Xanthomonas campestris Cf16-v1 (Dai et al., 1988; Huber and Waldor, 2002). Also, dif-like sequences have been found on plasmids (Hill, 1996). One of the initial descriptions of a dif-like sequence was as a short region of plasmid R1 that allowed maintenance of a small plasmid without a replication origin, using integration and excision from the host chromosome (Clerget, 1984; 1991). Similarly, we were able to detect high-frequency loss of the pJS3 insertion. We did not detect loss of the GGI, suggesting that the GGI's presence in the chromosome has been stabilized. This may have occurred by mutation of the left-flanking dif into an imperfect site for XerCD binding, making site-specific recombination events less likely to occur. We cannot rule out the possibility that the GGI is excised from the chromosome by site-specific recombination at low frequency, but we were unable to detect this loss by our methods. Spread of the GGI to the majority of gonococcal isolates may simply have been achieved by vertical transmission; however, it is also possible that the natural competence of N. gonorrhoeae has played a role in its distribution. We have demonstrated that the GGI can be lost from a strain by transformation; it is possible that the GGI could also be gained by this mechanism.
The consequences of horizontal transfer of genetic material in the gonococcal population are well known. Horizontal transfer is so frequent that the entire genome is in flux, each gene undergoing recombination with imported DNA (Smith et al., 1993). This high degree of horizontal transfer results in rapid spread of antibiotic resistance genes; thus, not only are antibiotic resistance genes quickly spread on conjugative and mobilizable plasmids (Eisenstein et al., 1977), but chromosomal markers can also be transmitted throughout the population by transformation. A major mechanism by which bacteria acquire antibiotic resistance is by recombination between a gene encoding an antibiotic sensitive target molecule and related genes resulting in novel, mosaic genes encoding resistant proteins. This mechanism is believed to be responsible for the development of sulphonamide resistance in Neisseria meningitidis and some forms of penicillin resistance in N. gonorrhoeae (Maiden, 1998). It is evident that transformation also contributes to antigenic variation. For example, the gonococcal porin gene exists in only one copy in each cell, but multiple alleles have been found throughout the population including mosaics produced by recombination (Cooke et al., 1998; Fudyk et al., 1999). Genetic recombination in the por gene produces variation in the many surface-exposed loops of porin and may allow the gonococcus to avoid elimination by the host. Additionally, pilin variation is known to occur via transformation (Seifert et al., 1988), as well as intracellular recombination (Gibbs et al., 1989). The ability of gonococci to secrete DNA for transformation may enhance the ability of genes to be spread through the population increasing antigenic variation and the spread and creation of antibiotic resistance genes.
Because N. gonorrhoeae is an obligate human pathogen, the DNA secreted by gonococci will interact with human cells and tissues and may influence pathogen–host interactions. DNA has been implicated as a requirement for the formation of Pseudomonas aeruginosa biofilms, communities of bacteria that are held together by a matrix consisting of macromolecules including polysaccharides, proteins and DNA (Whitchurch et al., 2002). Although it is unclear whether or not gonococci form biofilms, they have been observed to form aggregates or microcolonies on infected cells (Griffiss et al., 1999) and secreted DNA may be involved in this process. Active secretion of DNA may be more advantageous for the bacterium than autolysis in that it may allow for release of DNA for transformation or possible biofilm formation without stimulation of the host immune system. By contrast, autolysis releases not only DNA but also highly inflammatory products including lipooligosaccharide and peptidoglycan. Although bacterial DNA is also inflammatory, this inflammatory reaction can be blocked by DNA methylation (Krieg, 2002); the conserved gonococcal chromosome carries at least 14 DNA methylases (Stein et al., 1995). The GGI also encodes a putative DNA methylase, YdhA (Table 1). It is possible that either because of DNA methylation or because of low amounts of inflammatory molecules, type IV secretion of DNA allows DNA transfer without stimulation of the innate immune response. If, however, secreted gonococcal DNA is unmethylated, its affect on the immune response may be very different. Unmethylated CpG motifs in bacterial DNA are known to activate cell signalling pathways through the pattern recognition receptor toll-like receptor 9 (TLR-9). Because CpG motifs are common in bacterial and viral DNA and are suppressed in vertebrate DNA, they are recognized as foreign by the host. Engagement of TLR-9 by CpG DNA leads to activation of pathways including the mitogen-activated protein kinases and NF-κΒ. This activation leads to the production of primarily Th1-like cytokines, interferons and chemokines (Krieg, 2002). Thus DNA secreted by N. gonorrhoeae may contribute to inflammation associated with symptomatic gonorrhea, pelvic inflammatory disease, or disseminated gonococcal infection.
In summary, we have shown that most strains of N. gonorrhoeae carry a large, horizontally acquired genetic island inserted into the chromosome at the dif site. The GGI encodes a novel T4SS that secretes chromosomal DNA in a non-contact dependent manner. The secreted DNA can participate in natural transformation and may contribute to the spread of antibiotic resistance in N. gonorrhoeae, provide alleles for antigenic variation, or enhance the inflammatory nature of gonococcal disease. Future studies will provide greater insight to the advantages that the GGI may confer upon N. gonorrhoeae.
Bacterial strains and plasmids
Bacterial strains are described in Table 2. Insertions for mutagenesis and cloning of GGI genes were constructed using N. gonorrhoeae strain MS11A (Swanson et al., 1971; Segal et al., 1985). Gonococci were grown on GCB (Difco) plates containing Kellogg's supplements (Kellogg et al., 1963) or in GCB liquid medium (GCBL) containing 0.042% NaHCO3 (Morse and Bartenstein, 1974) and Kellogg's supplements. For blue/white screening, N. gonorrhoeae were grown on GCB-Tris-XP plates (Boyle-Vavra and Seifert, 1995). E. coli strains were grown in Luria–Bertani (LB) broth or on LB agar plates. For gonococci, chloramphenicol was used at 10 µg ml−1 and spectinomycin at 75 µg ml−1 (except when in combination with chloramphenicol where 50 µg ml−1 was used). Erythromycin was used at 2 µg ml−1 and 10 µg ml−1 for gonococci and 500 µg ml−1 for E. coli.
Cloning and sequencing of the GGI
The GGI was cloned from the chromosome of N. gonorrhoeae MS11A by chromosome walking as previously described (Hamilton et al., 2001). A fragment of GGI DNA from the most distal end of the known sequence was cloned into one of the pIDN plasmids (Hamilton et al., 2001) and used to direct insertion into the gonococcal chromosome. Southern blotting was used to identify restriction sites for cloning the flanking DNA. DNA fragments were gel-purified, self-ligated, and transformed into E. coli. The resulting plasmid clones were sequenced at the DNA Sequence Laboratory of University of Wisconsin Biotechnology Center using the BigDye fluorescent method as described by the manufacturer. The locations of ORFs in the GGI were determined using blast programs (Altschul et al., 1997) as well as GenMark. The DNA sequence was submitted to GenBank under accession number AY803022.
Generation of gonococcal mutants
Gonococcal mutants were constructed as described in Table 2. Briefly, insertion–duplication mutations (traH, traF, traN, traC, traI, ydbA and ung) were generated by cloning an internal fragment of the desired gene into one of the pIDN plasmids described previously (Hamilton et al., 2001). These plasmids were then used to transform the wild-type gonococcal strain MS11A. Mutation of dsbC was achieved by subcloning an ermC cassette from pKS65 into its coding sequence. Mutation of parA was generated by incorporating a point mutation in the sequence encoding a putative Walker box A motif, changing a lysine residue to a glutamine. Mutation of the traD was generated by a 33 bp deletion that removed six of the eight amino acids of the Walker box A motif. Complementation of gonococcal mutants was generated by cloning the full-length gene of interest into a vector (pKH9 or pKH35), which upon transformation into gonococci, directs its insertion between aspC and lctP in the chromosome. Plasmid pKH9 was generated from pGCC6 by a series of deletions, DraIII–NheI, NotI–SgrAI, HindIII–SacI, generating ligatable blunt ends using T4 DNA polymerase at each step, and then by replacing the truncated lacI gene with that of pGCC4 (Stohl et al., 2003). Plasmid pKH35 was generated from pKH9 by filling in the HindIII site, deleting the Ecl136II-BseRI fragment, and inserting the polylinker from pIDN1 at the KpnI site.
Determining the presence of the GGI
Gonococcal genetic island (GGI) positive-specific and GGI negative-specific PCR amplification was performed on all strains examined. Chromosomal DNA of each strain was amplified using primers 77F (5′-TAACAGCAGACGCTCC ATTC-3′) and 86R (5′-CAAGCGCATGGTACATGAAT-3′) (Tm 57°C; extension time 90 s) and using primers 73F (5′-AGCC ATCAGGGAGGCGGATA-3′) and hlh-ggiR (5′-CAGGCAAAC AGCTATTTGAG-3′) (Tm 54°C; extension time 90 s). The dif site and the imperfect dif site were sequenced from the following strains: PID2076, JC1, IN644, IN113831, LT38089, LT38093, DGI14, DGI20, LT37971, PID2004 (Dillard and Seifert, 2001), RUN5290, RUN5287 (Campbell et al., 1985).
Southern blots to confirm pIDN insertions were performed according to standard procedures (Sambrook et al., 1989). Chromosomal DNA from N. gonorrhoeae was prepared as described by Boyle-Vavra and Seifert (Boyle-Vavra and Seifert, 1993). DNA was separated by gel electrophoresis in a 0.8% agarose TBE gel. DNA was transferred to Stratagene Duralon-UV membrane using a vacuum blotter. Following UV cross-linking, the chromosomal DNA was probed with digoxygenin-labelled plasmid DNA (see below). Blots were washed at high stringency, and chemiluminescent detection was performed as suggested by the manufacturer (Roche).
For confirmation of ND500 (MS11AΔGGI) construction, MS11A, ND500 and FA1090 chromosomal DNA was digested with BsrGI, which cuts within the GGI, and then probed with pARM3. For site-specific recombination experiments, chromosomal DNA from ND500 (negative control) and ND518 (ND500 dif::pJS3) was digested with BsrGI. In addition to digested DNA, undigested ND518 chromosomal DNA and undigested pJS3 plasmid DNA (positive control) were also analysed. This blot was probed with the pIDN1. Dimerization of plasmid pJS3 was generated by transformation of the plasmid into LE392, a RecA-proficient E. coli strain.
Coculture transformation assay
Coculture transformation assays were performed as previously described (Hamilton et al., 2001). Briefly, P+ ND518 (ND500-SpR) and ND516 (HH522 recA6, CmR, TcR) were grown together in the same tube with and without aeration for 5 h at 37°C. The coculture was plated for cfu ml−1 (t = 0 h and t = 5 h) on GCB, GCB-Cm, GCB-Sp, and GCB-Cm, Sp plates.
DNA secretion assays
DNA secretion assays were performed as previously described (Hamilton et al., 2001). Briefly, P– transparent gonococcal strains were grown overnight on GCB agar plates and inoculated with a sterile Dacron swab into 3 ml of GCBL medium with Kellogg's supplements (Kellogg et al., 1963) and 0.042% NaHCO3 (Morse and Bartenstein, 1974). These cultures were grown for ∼ 2–2.5 h. Cultures were then diluted in 3 ml of Cellgro Complete tissue culture medium (Mediatech) supplemented with cysteine, cystine, pyruvate, Kellogg's supplements, starch and NaHCO3 (t = 0). These cultures were grown for an additional 5 h and culture supernatants were collected at 0 h and 5 h. Supernatants were assayed for DNA using PicoGreen (Molecular Probes), which in our hands binds both single-stranded and double-stranded DNA. Fluorescence of supernatants was measured and the amount of DNA secreted was calculated by comparison to a standard of HindIII-cut λ DNA (New England Biolabs) in tissue culture growth medium. DNA secretion was normalized to total cell protein as determined by Bio-Rad Protein Assay. Background fluorescence levels were determined by treating the supernatants of MS11A with DNase I for 30 min; this background was subtracted from the fluorescence of each strain examined. Only gonococcal cultures that had grown ≥ 1 log unit in 5 h were used for analysis.
Live/Dead staining of gonococci
Gonococci were stained using the BacLight Live/Dead staining kit (Molecular Probes). For this staining, gonococci were grown in the same manner as described for DNA secretion assays. After 5 h of growth, 1 ml of wild-type and mutant gonococcal cultures were pelleted and bacteria were washed once with prewarmed buffer (0.1 M MOPS, 1 mM MgCl2, pH 7.2). In addition, an aliquot of MS11A (wt) was washed once with isopropanol and again once with buffer; this aliquot constituted the ‘Dead’ bacteria for use in standardization. Bacterial cultures were diluted in buffer to 107 bacteria ml−1 and 100 µl aliquots added to a 96-well opaque plate. To each aliquot 100 µl of 2× Dye solution was added. Immediately, bacteria were examined by fluorescence at Ex. 485 nm and Em. 535 (green) and Em. 635 (red). Percentage of live bacteria for each culture was calculated upon comparison to a standard of different live : dead ratios of wild-type gonococci. For fluorescence microscopy, bacteria were prepared as above. To 1 ml of bacterial suspension, 3 µl of 1:1 dye mixture (Syto 9: Propidium iodide) was added. Immediately, 5 µl of stained bacterial suspension was examined by fluorescence microscopy on a Zeiss fluorescence microscope.
We thank Wilmara Salgado-Pabón and Jonathan Skarie for technical assistance. We acknowledge the Gonococcal Genome Sequencing Project supported by USPHS/NIH grant ♯AI38399, and B.A. Roe, L. Song, S.P. Lin, X. Yuan, S. Clifton, Tom Ducey, Lisa Lewis and D.W. Dyer at the University of Oklahoma. We thank the Cremer Fellowship in the Basic Sciences for financial support of H.L.H. This investigation was supported by NIH grant AI47958 to J.P.D. and traineeships on NIH 5 T32 G08349 to N.M.D. and NIH T32 AI055397 to H.L.H.