Insertion Mutations in pilE Differentially Alter Gonococcal Pilin Antigenic Variation

Abstract

Pilus antigenic variation in Neisseria gonorrhoeae occurs by the high-frequency, unidirectional transfer of DNA sequences from one of several silent pilin loci (pilS) into the expressed pilin gene (pilE), resulting in a change in the primary pilin protein sequence. Previously, we investigated the effects of large or small heterologous insertions in conserved and variable portions of a pilS copy on antigenic variation. We observed differential effects on pilin recombination by the various insertions, and the severity of the defect correlated with the disruption or displacement of a conserved pilin DNA sequence called cys2. In this study, we show that disruption or displacement of the pilE cys2 sequence by the same insertions or a deletion also affects pilin recombination. However, in contrast to the insertions in pilS, the analogous insertions in pilE impaired, but did not block, recombination of the flanking pilin sequences. These results, the change in the spectrum of donor silent copies used during variation, and our previous results with pilS mutations show that the donor pilS and recipient pilE play different roles in antigenic variation. We conclude that when high-frequency recombination mechanisms are blocked, alternative mechanisms are operative.

Neisseria gonorrhoeae (the gonococcus [Gc]) is the causative agent of the sexually transmitted disease gonorrhea. Gc pili, filamentous appendages that emanate from the cell surface, are comprised mainly of the pilin protein (5). Pili undergo antigenic variation through alteration of pilin primary amino acid sequences (10, 26). Pili are critical for the initial attachment of the bacterium to the host mucosal epithelia (19, 33) and are one of several factors required for natural DNA transformation competency (7, 29, 31).

Pilin is encoded by the pilE gene present at each of two pilE expression loci in laboratory strain MS11 variant A (pilE1 and pilE2) and at a single expression locus in all other isolates (34). In addition to the pilE gene, four to seven transcriptionally inactive, silent loci (pilS) contain from one to six partial copies of potential pilin coding sequences (8). Present at the 3′ end of the pilE locus and at each pilS locus is a conserved DNA sequence called the Sma/Cla repeat (SCR) (21, 23). The pilE SCR is required for efficient pilin antigenic variation (34) and is specifically bound by at least three proteins detected in partially fractionated Gc lysates (36, 37). There are one or two silent pilin copies immediately upstream of pilE (upstream silent copies), and each upstream silent pilin copy carries a partial SCR (37).

The silent pilin copies lack a promoter, a ribosome binding site, and the 5′ portion of conserved pilin sequences (21). The variable pilin sequences, present in the expressed and silent copies, consist of short lengths of conserved pilin sequences interspersed between regions of semivariable (SV) and hypervariable (HV) sequences (8, 10, 26). Near the 3′ end of the coding region are two conserved sequences called cys1 and cys2 (26) (see Fig. 1). Each encodes approximately 10 amino acids, including a cysteine residue, and both are conserved at the protein and DNA levels (8, 23). The cysteine residues form a disulfide bond and surround the 45- to 66-bp HV loop sequence (HV_L) (see Fig. 1) (also designated mc2 [8]). The HV_L region comprises the major portion of the HV domain, which is also the most divergent region of the pilin gene family (8, 9, 26). The portion of the HV region that is downstream of cys2 is also highly divergent and encodes the C terminus of the protein we have designated the HV tail (HV_T) (11) (also called mc1 [8]).

FIG. 1.

pilE genes from pilE mutant strains. Each cartoon represents the pilE locus of the indicated strain. Above the BHAC5 cartoon are the five regions of the pilE coding region. Const., constant. The triangles below the cartoon show the sites of insertion in the HV_L and cys2 regions. Dotted boxes represent the initial pilE sequence. Hatched boxes show the conserved cys1 or cys2 sequence. The black boxes show the NotI linker. Vertically striped boxes represent the ′cat insertion. The boxes with the wave design show the pilS1 copy 3 HV_L::NotI sequence. Only relevant restriction sites are shown. N, NotI; S, SmaI; C, ClaI; P, pilin promoter.

Pilin antigenic variation occurs when a portion of a silent copy replaces the respective sequence in the expressed gene. The transfer of a portion of a pilS copy into pilE is gene conversion, since the donor pilS sequence remains unchanged (8, 26). This RecA-dependent (16), RecO-dependent and RecQ-dependent (20) unidirectional flow of information can be achieved through two routes, intracellular recombination (1, 8, 10, 26) and recombination with extrachromosomal DNA taken up from autolyzed neighbors during transformation (7, 22, 28). The majority of in vitro pilin recombination reactions occur through intracellular reactions (30a, 35, 39).

Recombination of pilS sequences into pilE can also alter the level of pilus expression via introduction of nonsense codons (1), codon combinations that result in the secretion of truncated, soluble pilin proteins (9) or duplicated pilin sequences creating nonfunctional over-long pilin (L-pilin) (7, 9). Change from a highly piliated state (P⁺) to an underpiliated or nonpiliated state (P⁻) can be visually observed under a stereomicroscope as a change in the colony morphology. Some P⁻ colony variants are able to revert to a more piliated phenotype. Revertible and nonrevertible P⁻ colony variants are also generated by several mechanisms that are different from the RecA-dependent unidirectional recombination that produces pilin antigenic variants (7, 9, 12, 15, 18, 21, 25).

We previously attempted to measure the frequency of antigenic variation using a promoterless chloramphenicol acetyltransferase gene (′cat) inserted into pilS1 copy 3 (a copy that is unlinked to the SCR) as a surrogate marker for pilin variation. The transfer of ′cat from pilS1 copy 3 into pilE was never observed (11). Instead, chloramphenicol-resistant (Cm^r) variants each contained a new hybrid pilin locus consisting of pilE sequences linked to one of three pilS1 copies upstream of ′cat, with recombination junctions occurring in small regions of homology between the pilE gene and the target pilS copy. The structure of the hybrid pilin loci and surrounding sequences allowed the formulation of models for the movement of DNA during antigenic variation (11). Some of the predictions of these models have been supported experimentally (reference 10a and see Discussion).

Analysis of antigenic variation and pilin hybrid locus formation in other gonococcal mutants containing either the large 780-bp ′cat or a small 10-bp NotI linker in the conserved cys2 or HV_L regions of pilS1 copy 3 allowed us to conclude that both the size and the position of heterologous insertions in a pilS copy differentially affected recombination of the mutated pilS copy with pilE. Specifically, the small 10-bp NotI linker efficiently transferred from pilS1 copy 3 into pilE when it was inserted in the HV_L region of copy 3 but not when it was present in the cys2 region of copy 3. Moreover, the ′cat gene present in the HV_L or cys2 region of pilS1 copy 3 was never found to transfer into pilE. The effect of these insertions on hybrid-locus formation was somewhat different. Hybrid loci were formed by recombination between pilE and pilS1 copy 3 when ′cat was in the cys2 region of copy 3 but were never found when ′cat was in the HV_L region of pilS1 copy 3.

In this study, we have extended this analysis of the effect of ′cat and NotI linker mutations in the recipient pilE gene. These matched pilE insertion mutations exhibited effects on pilin recombination that were different from the effects of the previously described pilS mutations. The divergent effects show that donor and recipient pilin sequences act differently during pilin antigenic variation. Moreover, the altered spectrum of silent copies used when antigenic variation is inhibited by these mutations indicates that more than one recombination mechanism can mediate pilin variation. These data further confirm that insertion in or displacement of cys2 impairs or blocks normal pilin recombination. We report the first data showing that recombination through the pilE SCR can occur during a subset of pilin recombination reactions. Finally, we present updated models that explain how pilin sequences move during pilin variation.

MATERIALS AND METHODS

Bacterial strains, plasmids, and media.

All Gc strains were derived from VD300recA6 (27), a variant of strain MS11 (14). Table 1 describes the plasmids used in this study. Gc organisms were grown on GC Medium Base (GCB) (Difco) with Kellogg Supplements (13) at 37°C in 5% CO₂. For antibiotic selection of Gc, chloramphenicol (10 μg ml⁻¹), nalidixic acid (1.5 μg ml⁻¹), and erythromycin (8.0 μg ml⁻¹) (all from Sigma) were used.

TABLE 1.

Plasmids used in this study

Plasmid	Relevant characteristic(s)	Reference
pHSS25	NotI ′cat fragment in the vector pHSS6	11
pNG1312	pilS1 locus from MS11 in pBR322	21
pNG1312cat4	pNG1312 with ′cat inserted in the HincII site of the copy 1 cys2	This work
pNG1100-1	pilE1 from MS11 in pBA	4
pNG1100cat1	′cat from pHSS25 in the HincII site (cys2) of pNG1100-1	This work
pNG1100-2	pNG1100-1, ′cat with GCU in the StuI site 3′ of pilE	This work
pNG1100-3	pNG1100-2, 60-bp deletion of the pilE HVL and a portion of cys1, insertion of GGCGGC (encoding Gly-Gly) in place of the deletion	This work
pNG1100-1::NotI	NotI linker in the HincII site in the cys2 region of pilE1	This work
pUP1	Synthetic transformation uptake sequence in pHSS6	6
pNG1100-4	Transformation uptake sequences of pUP1 inserted into the HindIII site of pNG1100-1::NotI	This work
pNG1711	MS11 pilE1 opaE1 clone in the vector pBR322	21
pNG3005	pNG1711 with a PacI linker replacing the SCR ClaI site and ermC in SalI of opaE1	37
pNG1100-5	pNG1100-4 with the opaE1::ermC SmaI fragment of pNG3005 inserted at SmaI of pilE	This work
pSY6	Chromosomal fragment carrying a gyrB gene conferring nalidixic acid resistance	32

Generation of pilE and pilS mutations.

Restriction enzymes and linkers were obtained from New England Biolabs and were used under the manufacturer’s recommended conditions. DNA extraction from agarose gels was performed with GeneClean (Bio 101). PCR products used in cloning were generated with Native Pfu DNA polymerase (Stratagene). Unless specified, PCR primers are as previously described (11).

To create pNG1100-2, pNG1100-1 was digested with StuI and ligated to a ′cat PCR product [CATF (ATCGAGATTTTCAGGAGCTAAG)-GCUCATREV (GCCGTCTGAATTTCTGCCATTCATCCGC)]. Escherichia coli transformants were selected on chloramphenicol to ensure that only clones containing ′cat expressed from the pilin promoter were isolated. To create pNG1100-3, the pilE::′cat of pNG1100-2 was amplified in two PCRs with the oligonucleotide probes PILSTART and HVPEDELREV (GACGGCAGGTGCTTGGTGTCGATGCCGCCCTGT) for the 5′ end and HVPEDELF (GTTCGGTAAAATGGTTCTGCGGACAGGGCGGCATCGACACCA) and SP3A for the 3′ end. The two PCR products were mixed, and a third PCR with PILSTART and SP3A was performed. A PCR product of approximately 1,500 bp was gel isolated, cut with Bsu361 and SmaI, and ligated into Bsu361- and SmaI-cut pNG1100-1. Transformants were selected on chloramphenicol, and their pilE genes were sequenced. A clone containing the desired 60-bp deletion of the HV_L was identified. The pilE nucleotide sequence of pNG1100-1 (beginning with amino acid Cys 121 and ending with amino acid Thr 144) is TGCGGACAGCCGGTTACGCGCACCGGCGACAACGACGACACCGTTGCCGACGCCAAAGACGCCAAAGAAA TCGAC. The pNG1100-3 clone had a deletion (indicated by the underline) and a 6-bp replacement (GGCGGC) encoding a Gly-Gly linker in the HV_L. This pilE deletion was introduced into VD300recA6 by selection for a ′cat cassette located between the stop codon and the SCR of pilE to make strain BHA-HV (see Fig. 1). A similar cat insertion in the 3′ untranslated region of pilE was previously shown to support wild-type levels of pilin variation in MS11 (39).

Generation of Gc mutants and a Gc pilE variant.

BHAcat3 and BHANot5, a P⁻ antigenic variant of BHANot2 containing the pilS1 copy 3 HV_L::NotI sequence in pilE, and BHAC5, a P⁻ antigenic variant of BHANot3 with the entire pilS1 copy 5 HV_L in pilE, were described previously (11). DNA transformation generated the following Gc strains: BHAcat4 (generated with pNG1312cat4), BHA-HV (generated with pNG1100-3), BHANot4 (generated with pNG1100-5). PCR and Southern blot analyses were used to verify the presence of the desired insertions and to confirm that all other pilin loci were normal. DNA sequence analysis confirmed the desired deletion in the pilE gene of BHA-HV.

To create BHAcat5, the 5′ and 3′ halves of the pilE-opaE1 region were amplified from pNG1711 (21). PILSTART and 1711HVNHEREV (AAGCTAGCTGGTGTCGTCGTTGTCGCC) were used to amplify the 5′ end. 1711HVNHEREV anneals to the HV_L of pNG1711 and contains a 5′ tail encoding an NheI site. 1711HVNHEFOR (AAGCTAGCTGTTGCCGACGCCAAAGACG) and OPAEREV2 (TGTAAACGGCGGGCGCATCA) were used to amplify the 3′ end. 1711HVNHEFOR anneals to the HV_L of pNG1711 and contains a 5′ tail encoding an NheI site. OPAEREV2 anneals in the 3′ end of opaE1. Each PCR product was digested with NheI and ligated individually to NheI-digested ′cat from pHSS25. From the 5′-end PCR product–′cat ligation, a second round of PCR was performed with PILSTART and GCUCATREV. From the 3′-end PCR product–′cat ligation, a second round of PCR was performed with CATF and OPAEREV2. The two separate PCR products were gel isolated, cut with EcoRI, and ligated together. Ligated DNA was used directly in transformation. A Cm^r P⁻ transformant was isolated and checked by PCR and Southern blot analyses (BHAcat5).

Isolation and quantitation of Cm^r variants.

BHAcat1 and BHAcat4 were induced for 18 to 24 h on 2 mM IPTG (isopropyl-β-d-thiogalactopyranoside) plates. Gc organisms were swabbed into GCB liquid medium (GCBL), and dilutions were plated onto GCB plates to determine the total number of CFU per ml and onto GCB with chloramphenicol to determine the number of Cm^r CFU per ml.

Southern hybridization analysis.

Chromosomal DNA was extracted as described previously (2). ClaI- and ClaI- plus HpaI-digested DNAs were separated by agarose gel electrophoresis and blotted to nylon membrane as described by Sambrook et al. (24). The blots were probed with the oligonucleotides PILSTART (to identify the pilE locus) and CYS2R (to identify all pilin loci). Blots were often stripped and probed with the appropriate oligonucleotides used in PCR amplification (listed above). All oligonucleotide probes were end labeled with [γ-³²P]ATP (Amersham) with T4 polynucleotide kinase (New England Biolabs). Blots were hybridized and were washed at 15 to 20°C below the melting temperature of the oligonucleotide probe as recommended previously (24). Blots were exposed to X-Omat AR film (Kodak), either with or without an intensifying screen.

DNA sequence analysis.

For DNA sequence analysis of the pilE variants, 3 to 5 μl of the PCR product was treated with 1 μl of shrimp alkaline phosphatase (U.S. Biochemicals) and 1 μl of exonuclease I at 37°C for 15 min followed by heat inactivation at 85°C for 15 min. Eight milliliters of sequencing mix (from either an ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction Kit or an ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kit [Perkin-Elmer]), 3.2 pmol of primer, and H₂O to equal a total reaction mixture of 20 μl were added. All cycle sequencing was performed in an MJ Research PTC-1000 thermocycler retrofitted with a gold block and a hot bonnet. Reaction mixtures were run on either an ABI 373 or an ABI 377 sequencing apparatus as specified by the manufacturer. Single-stranded DNA sequences of the SV and HV regions with primer CONSTF2 (38) was sufficient to determine the pilS copy(s) that contributed to the observed sequence changes.

Transformation assay to enrich for P⁺ variants from P⁻ mutant strains.

Plasmid pSY6 DNA was used at a subsaturating level (60 ng) to minimize transformation of P⁻ Gc organisms via the pilus-independent transformation route (3). Each P⁻ strain was induced for 18 to 22 h before Gc colonies were collected with a sterile Dacron swab into GCBL–5 mM MgSO₄–2 mM IPTG at a density of approximately 2 × 10⁹ cells per ml. Cells (20 μl) were added to 200 μl of GCBL-MgSO₄-IPTG that contained 60 ng of pSY6. After 30 min at 37°C in a CO₂ incubator, the transformation mix was diluted into 800 μl of GCBL and incubated another 4 to 5 h. Serial dilutions were performed, and 20 μl from each dilution was spotted on plain GCB and on GCB containing nalidixic acid to determine numbers of total and nalidixic acid-resistant (Nal^r) CFU.

Analysis of Nal^r transformants to determine the proportion of pilE variants.

Nal^r transformants from BHAcat3 and BHAcat5 were plated onto GCB containing chloramphenicol to screen for loss of ′cat from pilE. Amplification of a wild-type-sized pilE PCR product with PILSTART-CYS2R and the failure to amplify with PILSTART-CATREV confirmed the loss of ′cat from pilE. The pilE genes from Nal^r transformants of BHANot4 and BHANot5 were amplified by the oligonucleotides PILSTART and SP3A, and the products were digested with NotI to determine whether the NotI linker was lost from pilE. The extents of sequence changes in Nal^r pilE variants of BHANot4 were determined by Southern blot analysis of chromosomal DNAs and by PacI restriction digests of PILSTART-OPAE2REV PCR products. The pilE genes from BHA-HV were amplified with PILSTART-HVPEDELREV to determine whether the 60-bp deletion was retained or lost. Amplification with PILSTART and CYS2R served as a positive control. Growth on medium containing chloramphenicol and PCR with PILSTART and CATREV detected the presence of ′cat in the 3′ untranslated region of pilE. Nal^r transformants of BHAC5 were tested for retention of the pilS1 copy 5 HV_L in pilE by amplification with PILSTART and PS1C5HVR, a primer specific for the pilS1 copy 5 HV_L region. Control amplifications of these same variants were performed with PILSTART and SP3A.

Analysis of Cm^r variants of BHAcat4.

To determine whether variants contained a hybrid locus or whether they were created by insertion of pilS1::′cat sequences into pilE, two separate PCRs were used. PCR with the primers PILSTART and CATREV established the linkage of pilE to ′cat. PCR with CATF and OPAEREV2 was diagnostic for ′cat sequences in the pilE locus. Variants that had ′cat 3′ of the pilin promoter but failed to amplify with CATF and OPAEREV2 contained a new hybrid locus where pilE sequence was linked to the pilS1::′cat sequence, which lacks the downstream OPAEREV2 sequence.

RESULTS

Description of strains.

In our previous study, we examined strains containing a large 780-bp ′cat gene or a small 10-bp NotI linker in either the conserved cys2 region or the HV_L region of pilS1 copy 3. Only the NotI linker in the HV_L region of copy 3 was efficiently transferred into pilE (11). In this study, the equivalent ′cat and NotI linker mutations were introduced into pilE to determine whether a large or a small insertion in the conserved cys2 or the HV_L region of pilE could be replaced by any pilS copy. The relevant features of the four pilE mutant strains corresponding to the previously described pilS1 mutants are shown Fig. 1. All four of the strains used in this work also contain the IPTG-regulatable recA6 allele to control when homologous recombination, including pilin antigenic variation, occurs (27). Strain BHAcat3 (11) has the ′cat fragment inserted into the cys2 region of pilE (pilE-cys2::′cat). Strain BHANot4 has a NotI linker in the cys2 region of pilE (pilE-cys2::NotI). BHANot4 also has a PacI linker in place of the ClaI site of the SCR. It was previously shown that replacement of the ClaI site by a PacI linker did not affect antigenic variation as measured by a reverse transcriptase-PCR assay (37). Strain BHAcat5 has the ′cat fragment in the HV_L region of pilE (pilE HV_L::′cat). Finally, we isolated a P⁻ variant of BHANot2 (pilS1 copy 3 HV_L::NotI) that had transferred the NotI linker from the HV_L region of copy 3 into pilE (BHANot5 [11]). All four of these strains expressed a P⁻ colony morphology and were over 1,000-fold less competent for DNA transformation than a P⁺ VD300recA6 variant after transient RecA induction (data not shown).

A fifth pilE mutant strain, BHA-HV (Fig. 1), contained a mutation different than those previously studied. BHA-HV carried a deletion of the pilE HV_L region, providing the opportunity of testing whether the spacing between cys1 and cys2 and/or the presence of the HV_L region was important for antigenic variation. BHA-HV expressed a P⁻ colony morphology, did not express detectable pili by transmission electron microscopy, and was incompetent for DNA transformation (data not shown).

In addition to the five pilE mutant strains described above, a control strain that expressed a P⁻ colony morphology due to normal sequence changes in pilE (BHAC5) was included. This colony variant of VD300recA6 carries pilS1 copy 5 sequences in pilE. We have previously shown that pilS1 copy 5 produces a predominance of P⁻ colony variants of this strain (11). BHAC5 was over 100-fold less competent for DNA transformation than the P⁺ VD300recA6 parental strain during transient RecA induction (data not shown). BHAC5 served as a control strain for the rate of P⁻-to-P⁺ colony morphology variation when wild-type pilE was present.

Quantifying pilE variation by a transformation enrichment assay.

pilE variation, producing a P⁻-to-P⁺ colony phase transition, was detectable by visible inspection of colony morphology changes in two of the six variants, BHANot5 and BHAC5 (data not shown). Therefore, a more sensitive assay to detect pilE variants of each pilE mutant strain was used. Fully piliated (P⁺) organisms are 100- to 10,000-fold more competent for DNA transformation than nonpiliated organisms (7, 29, 31, 39). Moreover, P⁺ organisms are 5- to 1,000-fold more competent for DNA transformation than P⁻ colony variants (7, 17). Therefore, the higher transformation competency of P⁺ variants than that of P⁻ variants was used to enrich for P⁺ pilE variants. After induction of RecA with IPTG, the strains were transformed with pSY6 DNA, which confers Nal^r. Many of the transformants expressed a P⁺ colony phase morphology, showing that transformation enrichment is effective. This result is similar to that of a previous study (39) where 60 to 80% of transformants arising from pilE point mutants had a P⁺ colony morphology. While the transformation frequency could be used to estimate the P⁻-to-P⁺ variation frequency, the different levels of residual transformation competence of the P⁻ mutants prevent an accurate comparison between mutants. Therefore, sequence analysis of the pilE gene in randomly selected transformants from each strain was performed. Thirty to 90% of all transformants from each P⁻ strain had sequence changes in pilE. Since variants with sequence changes in pilE that do not increase transformation competence are not detected by this assay, the frequency of production of Nal^r pilE variants does not directly measure the total frequency of pilE variation. However, by correcting the transformation frequency for the proportion of transformants that have a changed pilE, a relative measure of pilE variation can be obtained. We used this relative variation frequency to compare the effects of pilE mutations on pilin variation.

Analysis of pilin variation in mutant strains.

Analysis of the cys2 mutant stains BHAcat3 and BHANot4 showed that insertions in the pilE cys2 region did not block recombination from all pilS copies (Table 2). This finding is in contrast to the finding that with the same insertions in the pilS1 copy 3 cys2 region, recombination of ′cat or NotI into pilE was never detected (frequency, <10⁻⁹ or <10⁻⁴, respectively [11]). However, the frequency of pilE variation in BHAcat3 and BHANot4 was reduced 26- and 15-fold (3.8 and 6.7%, respectively) (Table 2) relative to that of the P⁻ control strain BHAC5. Therefore, insertions in the cys2 region of pilE interfere with recombination that can remove the insertion but do not block it.

TABLE 2.

Relative effects of pilE mutations in transformation and pilE variation

Strain	Nature of pilE	% Transformationa	% pilE variationb
BHAC5	Wild type	100	100
BHAcat3	cys2::′cat	5.6	3.8
BHANot4	cys2::NotI	7.7	6.7
BHAcat5	HVL::′cat	0.42	0.20
BHANot5	HVL::NotI	100	100
BHA-HV	ΔHVL	2.0	1.9

^aThe level of transformation detected in the wild-type control strain BHAC5 was assigned a value of 100%. The percent transformation in each pilE mutant strain relative to the level of transformation in BHAC5 was determined by the following formula: (the frequency of transformation in pilE mutant strain/frequency of transformation in BHAC5) × 100. 

^bThe level of pilE variation detected in the wild-type control strain BHAC5 was assigned a value of 100%. The percentage of pilE variation in each pilE mutant strain relative to the level of variation in the control was determined by the following formula: [fraction of relative transformation of pilE mutant × (fraction of pilE variants among pilE mutant transformants/fraction of pilE variants among BHAC5 transformants)] × 100. 

pilE variants were generated when a NotI linker was present in the HV_L region of pilE (strain BHANot5) at the same frequency as with the P⁻ control strain BHAC5 (pilS1 copy 5 HV_L in pilE) (Table 2). Therefore, a small insertion in the pilE HV_L does not affect the frequency of antigenic variation. This result confirms the observation that a pilS1 copy 3 HV_L::NotI mutation is efficiently transferred into pilE (11) and shows that the HV_L region can accept minor changes in sequence without disrupting antigenic variation.

The ′cat insertion into the HV_L region of pilE (BHAcat5) demonstrated the most drastic reduction in pilE variation: 0.2% or an average 511-fold decrease relative to the level of variation in BHAC5 (Table 2). This effect is consistent with the inability of the strain with ′cat in the HV_L region of pilS1 copy 3 to create hybrid loci between pilE and pilS1 copy 3 (11). Southern blot and PCR analyses showed that 50 to 66% of the P⁺-enriched transformants of BHAcat5 retained ′cat in pilE. These variants had either a pilE::′cat locus that was larger than the parental pilE::′cat locus or a wild-type-sized pilE locus in addition to the original pilE::′cat locus (data not shown). Induction of RecA expression with IPTG resulted in a high-frequency loss of the P⁺ phenotype (20 to 22% P⁻ per total CFU), suggesting that these enriched variants carried tandem pilE sequences. The remaining 44 to 50% of P⁺ transformants of BHAcat5 were pilE variants that had lost the ′cat insertion by recombination with a silent copy (data not shown).

Finally, BHA-HV (pilE-ΔHV_L) showed a decrease in transformation-enriched pilE variation of 54-fold relative to that of BHAC5 (01.9%) (Table 2), showing that deletion of the HV_L interferes with pilE variation.

Sequence analyses of independent P⁺ variants derived from P⁻ mutants.

The transformation efficiencies of P⁺-enriched pilE variants showed that all mutations in pilE, except for the HV_L::NotI mutation, reduced the frequency of pilE variation relative to that of BHAC5. To examine whether the mutations in pilE also altered the spectrum of donor pilS copies available for recombination, the pilE genes of 10 independently derived P⁺ variants from each P⁻ strain were sequenced (Table 3). Analysis of the pilE gene sequences from the P⁺ revertants of the P⁻ control strain BHAC5 showed that six different pilE variants were isolated, with five different donor pilS copies. Two of the P⁺ variants of BHAC5 had extensive sequence changes in pilE, where most of the SV, HV_L, and HV_T regions were replaced by pilS sequences. Three variants had changes in the HV_L and HV_T regions, and three variants had changes solely in the HV_L region. We conclude that replacement of the pilE copy 5 HV_L sequences is sufficient to produce a P⁺ variant in this strain and that the different amounts of pilS sequence brought into pilE represent the inherent variability of this system. Four of the five donor pilS copies used to produce the P⁺ variants were located adjacent to an SCR in the donor silent locus (SCR linked). Therefore, while both SCR-linked and non-SCR-linked copies were used, there was a preference for SCR-linked copies. We cannot determine whether this preference was due to the proximity of these silent copies to the SCR, a high affinity of these copies for the resident pilE sequence during recombination, or an enhanced ability of these particular HV_Ls to produce piliated, highly competent variants of BHAC5.

TABLE 3.

Sequence analysis of independent P⁺ variants enriched from P⁻ strains

Initial strain	Source of pilin sequence changesa in:			nb
	SV	HVL	HVT
BHAC5 (S1c5HV)	S2c1 (177 [2])	S2c1	S2c1	2
	S2c1 (177 [2])	S2c1	—	2
	—	S6c1	—	2
	—	S7	S7	2
	—	S5	S5	1
	—	S1c4	—	1
BHAcat3 (cys2::′cat)	S2c1 (138, 177 [5])	S2c1	S2c1	6
	S6c1 (66 [3], 173)	S6c1	S6c1	4
BHANot4 (cys2::NotI)	S2c1 (138, 177 [4])	S2c1	S2c1	5
	upsS1 (161 [2], 176)	upsS1	upsS1	3
	S1c1 (179 [2])	S1c1	S1c1	2
BHAcat5 (HVL::′cat)	S2c1 (177 [3])	S2c1	S2c1	3
	upsS1 (176 [7])	upsS1	upsS1	7
BHANot5 (HVL::NotI)	S2c1 (138, 177)	S2c1	S2c1	2
	S2c1 (177 [2])	S2c1	—	2
	S1c1 (178 [2])	S1c1	S1c1	2
	S7 (168)	S7	S7	1
	S5 (173 [2])	S5	S1c1	2
	S1c1 (135)	S1c1	S2c1 or S7	1
BHA-HV (ΔHVL)	S2c1 (177 [2])	S2c1	S2c1	2
	S2c1 (177 [2])	S2c1	—	2
	—	S1c5	—	2
	Slc2 (199)	Slc2	—	1
	Slc5 (177)	S7 or S2c1	—	1
	S6c1 (173)	Slc5	—	1
	S1c4 (130)	S1c4	—	1

^aThe sources of changed pilin sequences in the SV, HV_L, and HV_T regions of pilE. The source of the pilin sequence is indicated by the pilS locus from which it came (e.g., S5 is pilS5) and by the number of the copy in the silent locus if more than one copy exists (e.g., S1c1 is pilS1 copy 1). upsS indicates the pilE-associated upstream silent locus. Silent copies with identical sequences in both the HV_L and HV_T regions are listed. Changes in the HV_L and HV_T always encompassed these entire regions. Each change in the SV region begins from a defined 5′ endpoint and extends through the end of the SV region at cys1. The 5′ endpoint(s) is indicated in parentheses as the number of bases it is upstream from the first conserved base of cys1. In brackets are the numbers of variants with the indicated 5′ endpoint. —, no change in pilin sequences. 

^bNumber of independently derived P⁺ variants with pilE sequences from the indicated silent copies. 

The pilE genes of the sequenced BHAcat3 (cys2::′cat), BHANot4 (cys2::NotI), and BHAcat5 (HV_L::′cat) P⁺ transformants had extensive sequence changes starting near the beginning or middle of the SV region and extending through the end of the HV_T. Replacement of just the mutated cys2, which should also restore piliation, was not observed. Only SCR-linked pilS sequences were found in the P⁺ transformants of these strains, and the donors were limited to two or three donor pilS copies. The limited repertoire of copies used as donors to remove these mutations and the extensive regions of sequences replaced in each variant suggest that these mutants use alternate mechanisms to effect pilin variation.

The pilE sequences of 10 independent P⁺ variants of BHANot5 (HV_L::NotI) were analyzed. Two similarities were noticed between P⁺ variants of BHAC5 and P⁺ variants of BHANot5. First, sequences from several different donor pilS copies were found among the P⁺ variants of each strain. Second, the HV_Ts of some P⁺ variants of each strain remained unchanged, showing that recombination had occurred at cys2. Interestingly, all 10 P⁺ variants of BHANot5 (HV_L::NotI) were created by recombination with an SCR-linked pilS copy.

Although the BHA-HV (ΔHV_L) strain was severely impaired in its ability to undergo variation, both the spectrum of pilS copies used and the different amounts of sequence transferred during variation suggest that the mechanism(s) used for variation in this strain is similar to that of BHAC5. However, 6 of 10 P⁺ variants were generated by recombination with a pilS copy not linked to an SCR, which is in contrast to what occurred with the other strains. It is notable that a majority of the sequence changes in these variants ended at cys2, but the basis for this is not known.

An SCR-linked silent copy can transfer a ′cat gene into pilE.

Since the ′cat in the cys2 of pilE could be replaced by sequence from an SCR-linked pilS copy, albeit at a frequency 26-fold less than that of the P⁻ control, it was possible that the presence of the SCR homology 3′ of ′cat facilitated the recombination of ′cat from a pilS copy into pilE. To test whether cys2::′cat could recombine from an SCR-linked pilS copy into pilE, a strain (BHAcat4) with ′cat in the cys2 region of an SCR-linked copy was constructed (pilS1 copy 1::′cat). Cm^r variants arose at a frequency of about 10⁻⁵ per CFU, about 10-fold more frequently than Cm^r variants of BHAcat1 (pilS1 copy 3 cys2::′cat) arose in a direct comparison (data not shown). Cm^r variants from four trials were analyzed by both PCR and Southern blot analyses (data not shown). A majority of these Cm^r variants of BHAcat4 (66 to 83%) carried a pilE-pilS1 copy 2 hybrid locus in addition to all other normal pilin loci (class II) (Fig. 2). Zero to 17% had a pilE-copy 1 hybrid locus as well as all other normal pilin loci (class I) (Fig. 2). Interestingly, 0 to 8.3% had copy 1::′cat sequences in pilE (they were analogous to normal pilE antigenic variants) and 0 to 25% had copy 2-copy 1::′cat sequences in pilE (they were analogous to long pilin [L-pilin] variants). These results show that ′cat can recombine from pilS into pilE as long as it is in an SCR-linked pilS copy, but even with the more extensive homology provided by the SCR, recombination of pilS copy 1::′cat with pilE usually resulted in hybrid locus formation.

FIG. 2.

Cm^r variants of BHAcat4 have either a pilin hybrid locus or ′cat in pilE. Cartoons I and II represent new hybrid loci found in some Cm^r variants of BHAcat4. Class I (cartoon I) hybrid loci were created by recombination between pilE and pilS1 copy 1 (copy 1) upstream of ′cat. Class II (cartoon II) hybrid loci were created by recombination between pilE copy 2 and pilS1 copy 2. Cartoons A and B depict the pilE locus in Cm^r variants of BHAcat4 in which ′cat transferred into pilE. Cartoon A shows a pilE antigenic variant containing copy 1::′cat sequences. Cartoon B shows an L-pilin-like variant containing copy 2-copy 1::′cat sequences in pilE. upsS1 and upsS2 are upstream silent copies in the pilE locus. Dotted boxes, pilE sequences; black boxes, recombination points; white boxes, pilS sequences; striped boxes, ′cat; cross-hatched boxes, the SCR; boxes with the modified herringbone pattern, opaE1. The average distribution (Ave.) of each type of Cm^r varaint was determined from the results of four independent trials of 12 randomly isolated variants per trial. P, pilin promoter. H, HpaI; S, SmaI; C, ClaI.

Recombination between pilin sequences can occur beyond the SCR.

During construction of strain BHANot4 (cys2::NotI), a PacI linker replaced the ClaI site in the pilE SCR. This replacement increased the size of the pilE ClaI fragment. Southern blot analysis of Nal^r transformants derived from BHANot4 revealed two differently sized fragments carrying pilE (Fig. 3). Several P⁺ variants of BHANot4 had the pilE locus on a larger ClaI fragment, identical in size to that of the parental strain, BHANot4. In these variants, recombination at pilE occurred upstream of the PacI site. The remaining 44% of P⁺ variants now had a pilE ClaI fragment that was the same size as that of VD300recA6. In these variants, recombination occurred through the SCR, restoring the ClaI site. This result is the first evidence that recombination during pilin variation can extend into the SCR but also shows that it does not have to extend past the ClaI site of the SCR.

FIG. 3.

Recombination at pilE can occur in the SCR. (A) Southern blots of ClaI-digested Gc DNA probed with either the pilin-promoter-specific oligonucleotide probe or a cys2 NotI-specific oligonucleotide probe. Variant 2 is mixed. (B) Representations of the pilE gene. Cartoon i indicates that variants 4 and 5 retained the NotI and PacI linkers in pilE. Cartoon ii indicates that variants 1, 2, 9, 10, and 11 lost the NotI linker but retained the PacI linker. Cartoon iii indicates that variants 3, 6, 7, and 8 lost both the NotI linker and the PacI linker. Dotted boxes, original pilE sequence; white boxes, new pilS sequence; gray boxes, cys1 and cys2; black box, sequence with NotI linker (N). S, SmaI; P, pilin promoter.

DISCUSSION

Our previous study of mutations in pilS provided insights into which pilin sequences are required for antigenic variation (11). The present study of analogous mutations in pilE provides the first direct evidence that the donor and recipient genes act differently during variation, shows that correct spacing of conserved elements is required for efficient recombination, clarifies the role of conserved and variable pilin sequences during recombination, and supports the hypothesis that there are multiple recombination mechanisms that allow recombination between silent copies and the expression locus.

pilE variation, resulting in a P⁻-to-P⁺ colony morphology switch, occurs in wild-type Gc organisms at a frequency of about 1% and can be quantified by observing changes in colony morphology in a stereomicroscope. Since this method cannot detect lower frequencies of colony variation, DNA transformation was used to enrich for P⁺ pilE variants from P⁻ strains. The relative frequency of pilE variants detected by this method from the P⁻ mutants does not directly measure variation, since not all pilE variants provide a selectable increase in transformation competence. Moreover, this method does not correct for the differences in transformation efficiencies among the spectrum of pilE variants generated from each strain (17). However, this method does enrich for P⁺ pilE variants from each preinduced P⁻ strain, since 50 to 90% of all Nal transformants expressed a P⁺ colony morphology and/or sequence changes in pilE. It should be noted that, in the absence of recA preinduction, transformation frequencies were low in each strain and that the few transformants that were obtained did not show sequence changes in pilE (data not shown). These findings show that these mutants retain residual transformation competence (29) but that the residual transformation competence is too low to influence the selection for variants by transformation. Since pilE variants made up similar percentages of the total numbers of transformants from individual strains and increased transformation efficiency was dependent on pilE variation, the transformation enrichment method provides a reasonably accurate assessment of the relative level of pilE variation in each strain.

Similar to results of our earlier studies, disruption of the pilE cys2 by either ′cat or a NotI linker did not prevent recombination upstream of cys2 but prevented recombination within the 3′ end of cys2 or the HV_T. In contrast to results of our previous studies, insertions in cys2 of pilE could be replaced by pilS sequences, albeit at a reduced frequency. However, all P⁺ variants derived from pilE cys2 mutants contained extensive sequence changes originating from a subset of SCR-linked pilS copies. The reduced frequency of recombination, extended recombination tracts, and limited repertoire of donor pilS sequences all point to the inhibition of a high-frequency recombination mechanism and an uncovering of a lower-frequency mechanism(s).

When ′cat was in the cys2 gene of pilS1 copy 3, it was never found to transfer into pilE even with chloramphenicol selection (11). This result raises the question of whether the only difference between the abilities of pilE and pilS1 copy 3 cys2 insertions to recombine is the presence of a linked SCR. When ′cat was inserted into cys2 of an SCR-linked pilS gene (BHAcat4 pilS1 copy 1::′cat), transfer of ′cat into pilE was observed. However, the vast majority of Cm^r variants of BHAcat4 contained new pilin hybrid loci comprised of sequences duplicated from portions of the pilE and pilS1 copy 1::′cat loci. The low-frequency transfer of ′cat from pilS1 copy 1 into pilE resembles L-pilin formation. First, both the frequency of L-pilin formation (1 to 2% of P⁻ variants [11]) and the frequency of variants with a copy 1 cys2::′cat sequence in pilE are much lower than the frequency of normal pilE variation as directly measured by Serkin and Seifert (30). Second, both the transfer of cys2::′cat and the transfer of extra pilin sequences to create an L-pilin are dependent on SCR-linked pilS sequence (18). Both the reduced efficiency of transfer of a large region of heterology from pilS into pilE and the SCR dependence of this reaction indicate that this reaction is mechanistically different from recombination of short pilS segments into pilE during normal pilin antigenic variation. This putative alternate recombination mechanism which generates L-pilin variants may also be operative in the replacement of pilE cys2 insertions, which also occur with reduced efficiency and involve extensive changes from an SCR-linked copy.

In our previous work, the presence of ′cat in the HV_L of pilS1 copy 3 prevented recombination between pilE and copy 3 during hybrid-locus formation. In strain BHAcat5, the ′cat in the HV_L of pilE did not block recombination. However, recombination at pilE in the pilE HV_L::′cat mutant was reduced over 500-fold. In addition, the only two pilS copies that recombined with the mutant pilE were the upstream silent copy linked to pilE and pilS2 copy 1. Both of these copies are adjacent to a partial or full SCR, and both map just upstream of pilE (9). Interestingly, using a pilEΔSCR mutant impaired in antigenic variation, Wainwright et al. (37) found that only one of five P⁺-to-P⁻ colony phase variants had sequence changes in pilE and that these variant sequences originated from the pilS2 locus. These findings led to the hypotheses that the recombination of pilS sequences in proximity to pilE was not blocked by the pilEΔSCR mutation and that it occurs by an alternative mechanism. Similarly, it appears that ′cat in the HV_L blocks most pilin antigenic variation but that recombination with SCR-linked pilS copies just upstream of pilE can still occur. We predict that recombination of the silent copies located immediately upstream of pilE is mechanistically distinct from recombination of silent copies at distant loci.

In strain BHA-HV (ΔHV_L), pilin variation was reduced 54-fold. Sequence analysis of the P⁺ transformants showed that, in each case, short tracts of pilS sequence had replaced the deletion. P⁺ variants had sequences from a varied repertoire of pilS genes, including pilS copies not linked to an SCR. Since the types of recombinants were similar to those with wild-type pilE, we conclude that the HV_L is dispensable for antigenic variation. However, since recombination was drastically reduced, it is clear that the alteration in spacing between cys1 and cys2 and/or the loss of conserved sequence at the end of cys1 affects the efficiency of recombination. Why the deletion of the HV_L sequences shifts the preference to a non-SCR-linked copy is not known.

Since recombination during pilin variation is predominantly a gene conversion event, we have assumed that the recipient gene (pilE) must be treated differently from the donor pilS copy. The data presented here provide the first experimental support for this assumption. We also conclude from the effects of both insertions in pilE and different pilS copies and the deletion in pilE that the spacing of conserved elements in the interacting DNAs is crucial for efficient recombination. We have previously proposed a model for the movement of DNA during pilin variation that invokes an initiating recombination reaction between a donor pilS copy and a donor pilE at a region of identity (8 to 40 bp) to produce a circular hybrid intermediate and that circular pilE-pilS hybrid intermediates can recombine with a recipient pilE to mediate pilin variation (10a, 11). This model is consistent with the data presented herein, since it incorporates different roles for pilS copies, which act only as donors, and the pilE locus, which acts as both a donor and a recipient. However, this model does not explain why the correct spacing of conserved elements would be required for efficient variation. Moreover, we have found that the proposed covalently closed, circular intermediates carrying hybrid loci (11) are not formed frequently enough in Gc organisms to be the predominate intermediate of pilin variation (10a). We have indirect evidence that a majority of hybrid intermediates formed are in a recombinogenic form and that when a pilE-pilS hybrid intermediate is produced, it is very efficient at mediating recombination with a recipient pilE (10a). Therefore, pilE-pilS hybrid intermediates remain a central feature of our working model. Based on all our data, we postulate that the majority of variation intermediates are partially single stranded and that multiple conserved portions of the recombining genes need to base pair to enable recombination (Fig. 4). This base pairing would require a synaptase and/or the action of an exonuclease. The effect of ′cat insertions in the HV_L show that the relative spacing between cys1 and cys2 in the donor pilE gene and pilS copy is required for hybrid-locus formation (11). We have not examined whether this spacing is also required for recombination of a pilE-pilS hybrid intermediate into the recipient pilE. This revised model accounts for the requirement for maintaining spacing between conserved elements for efficient pilin variation. Proof of this model awaits the isolation of the true intermediates of pilin variation but takes into account all of the observed features of antigenic variation.

FIG. 4.

Working model for the movement of DNA during pilin variation that relies on a pilE-pilS hybrid intermediate. (A) Recombination between pilS1 and pilE to form a pilE-pilS hybrid intermediate. (Left) Recombination (shaded diamond) between pilS1 (horizontally lined boxes) copy 4 and pilE (checkered box) to form a hybrid intermediate. The dotted line of the intermediate shows that it is not a covalently closed circle, but the exact form is unknown. (Right) More-detailed interactions between the donor pilS copy and pilE. The black boxes show regions of conserved nucleotides. The thin X shows the crossover site at regions of homology that form the hybrid intermediate. The shaded double-headed arrows with box centers show the proposed interactions between the conserved cys1 and cys2 sequences. The open double-headed arrows with box centers show putative interactions with other conserved sequences. The four-point stars represent the SCRs when present. (B) Recombination of a hybrid intermediate with a recipient pilE to introduce a portion of the pilS copy into pilE. (Left) Recombination between a hybrid intermediate and the recipient pilE. The open diamond shows recombination between the pilS portion of the hybrid and similar portions of the recipient pilE. This recombination is similar to that which formed the hybrid in panel A. The crossover (thick X) upstream of the pilE sequences represents a homologous recombination event required to insert pilS sequences into pilE. (Right) More-detailed interactions between a hybrid intermediate and the recipient pilE. The thin X shows the second crossover between regions of homology. The open double-headed arrows with box centers show putative interactions with other conserved sequences. The double-headed arrow without a box center represents proposed interactions and recombination between the SCRs.