Molecular Nature of Spontaneous Modifications in gacS Which Cause Colony Phase Variation in Pseudomonas sp. Strain PCL1171

ABSTRACT Pseudomonas sp. strain PCL1171 displays colony phase variation between opaque phase I and translucent phase II colonies, thereby regulating the production of secondary metabolites and exoenzymes. Complementation and sequence analysis of 26 phase II mutants and of 13 wild-type phase II sectors growing out of phase I colonies showed that in all these cases the phase II phenotype is caused by spontaneous mutations in gacA or/and gacS. Mutation of gac reduced both the length of the lag phase and the generation time. Isolation and sequencing of the gacS genes from the phase II bacteria revealed one insertion as well as several random point mutations, deletions, and DNA rearrangements. Most phase II colonies reverted with a high frequency, resulting in wild-type gacA and gacS genes and a phase I phenotype. Some phase II bacteria retained the phase II phenotype but changed genotypically as a result of (re)introduction of mutations in either gacA or gacS. The reversion of gacA or gacS to the wild type was not affected by mutation of recA and recB. We conclude that in Pseudomonas sp. strain PCL1171, mutations in gacA and gacS are the basis for phase variation from phase I to phase II colonies and that, since these mutations are efficiently removed, mutations in gac result in dynamic switches between the “wild-type” population and the subpopulations harboring spontaneous mutations in gacA and or gacS, thereby enabling both populations to be maintained.

Phase variation is a process of reversible, high-frequency phenotypic switching that is mediated by mutation, reorganization, or modification of DNA. This process is used by several bacterial species to generate population diversity that increases bacterial fitness and is important in niche adaptation (33). Phase variation can sometimes be observed by the appearance of morphologically distinct colonies or sectors within a colony (8,12). In contrast to spontaneous mutations, which occur at a frequency of approximately 10 Ϫ7 mutations per cell per generation, phase variation occurs at frequencies higher than 10 Ϫ5 switches per cell per generation (12). Four mechanisms of phase variation are known (12): (i) slipped-strand mispairing, dependent on variations in the length of a repeat tract, switching a gene on or off as a result of frameshifts, or regulating the level of expression by altering promoter spacing; (ii) genomic rearrangements, based on invertible elements or recombination events resulting in gene conversions; (iii) differential methylation, based on the presence of methylation sites within a promoter, which can regulate the binding of regulatory proteins; and (iv) random unprogrammed variation, which can switch traits on and off via random reversible mutations (3).
Phase variation has been reported to regulate the production of pili (22), outer membrane proteins (22), flagella (13), fimbriae (1), surface lipoproteins and other surface-exposed structures (8,12,29), secondary metabolites (5,38), and secreted enzymes such as proteases, lipases, and chitinases (5,38). In a previous paper (38), we reported that out of 46 Pseudomonas strains antagonistic against the wheat-patho-genic fungus Geaumannomyces graminis pv. tritici R3-11A (27), 43 displayed colony phase variation. One of these strains, PCL1171, was selected for study of the molecular basis of phase variation. In this strain, antagonistic activity, morphology, and expression of secondary metabolites (such as a lipopeptide with antifungal activity), biosurfactant activity, and the exoenzymes chitinase, lipase, and protease are regulated via phase variation and expressed only in the phase I phenotype. Estimation of the phase variation frequencies showed approximately 5.0 ϫ 10 Ϫ5 and 9.0 ϫ 10 Ϫ2 switches per cell per generation for phase I to II and for phase II to I, respectively (38).
A preliminary genetic analysis of the phase variation mechanism in PCL1171 showed that transposon mutation of the gacS gene resulted in the phase II phenotype, whereas mutation of mutS strongly increased the frequency of switching between phase I and phase II (38). The gacA/gacS two-component regulatory system consists of a sensor kinase, GacS, and a reponse regulator, GacA, belonging to the FixJ family of transcriptional regulators (20); this system regulates secondary metabolism and production of exoenzymes (2,15,20,28,38). Phase variation via a homolog of gacS, pheN, which regulates pathogenicity and colony morphology in Pseudomonas tolaasii, is dependent on a spontaneous duplication in this gene (9). Furthermore, it has been reported that the gacA/gacS system is subject to the accumulation of mutations in several Pseudomonas spp. (4,7,20,28,34). However, neither reversibility of these gac mutations nor the molecular nature of mutations in the gacS gene has been reported. In this paper we analyze the role of gacA/gacS in phase switching and give a detailed description of the molecular nature of mutations accumulating in the gacS gene responsible for reversibility of colony phase variation.

MATERIALS AND METHODS
Bacterial strains and plasmids. Bacterial strains and plasmids are listed in Table 1. Pseudomonas strains were grown in King's medium B (KB) (14) at 28°C. Escherichia coli strains were grown in Luria-Bertani medium (31) at 37°C. Solid growth media contained 1.8% (wt/vol) agar (Difco Laboratories, Detroit, Mich.). Kanamycin, tetracycline, and carbenicillin (Sigma, St. Louis, Mo.) were added for antibiotic selection in final concentrations of 50, 40, and 50 g/ml, respectively. Growth in liquid medium was analyzed in triplicate by measurement of the optical density at 620 nm over time. The numbers of viable cells in the starting culture were counted. Growth experiments were carried out three times in triplicate.
Measuring frequencies of phase variation. Bacteria from a colony with phase I or phase II morphology were inoculated from a plate in 20 ml of KB and were grown overnight at 28°C. The morphology and initial number of bacteria were determined after dilution plating of an average of 500 cells per plate. At least 1,500 colonies were counted for the estimation of frequencies expressed as the number of switches per cell per generation.
Isolation and characterization of phase II Tn5luxAB transposon mutants. A mutant library of strain PCL1171 was constructed by using plasmid pRL1063a (38), which harbors a Tn5 transposon with promoterless luxAB genes and a kanamycin resistance marker (40). Mutants with a phase II morphology were selected, replated several times, and grown in liquid medium to select those mutants locked in the phase II phenotype. General DNA recombinant techniques were performed according to the work of Sambrook and Russell (31).
Complementation of Tn5luxAB transposon mutants and wild-type phase II sectors. Phase II bacteria were complemented with pMP6562 (gacS) and pMP5565 (gacA) by parental mating. Primers oMP716 (5Ј-GGAATTCAGGAT GTCCATCAACACCA-3Ј) and oMP717 (5Ј-GGAATTCATCGTTGATGAAG GCACACA-3Ј), each containing a HindIII site (GGAATTCA), were used to produce a 3.2-kb PCR fragment harboring a gacS homolog from Pseudomonas sp. strain PCL1171. The PCR product was cloned into pMP5565 by using Hin-dIII. The resulting construct, pMP6603, harboring both the gacA and the gacS gene under the control of their own promoters, was electroporated into E. coli S17-1 (36) and used to transform phase II bacteria by parental mating. The morphology of the transformants was judged after 2 days of growth at 28°C on King's medium B (14) supplemented with tetracycline.
Isolation and construction of mutants. To isolate genes from PCL1171, a plasmid library was constructed. Chromosomal fragments of PCL1171 phase I bacteria, 3 to 5 kb long, were obtained after partial digestion using EcoRI. After ligation of the fragments into pBluescript, the clones were pooled and isolated. The partial recA gene was isolated from this library via PCR using two primers specific for pBluescript, T7 (5Ј-TAATACGACTCACTATAGGG-3Ј) and oMP61 (5Ј-GTTTTCCCAGTCACGAC-3Ј), and two internal primers for recA, oMP529 (5Ј-GGGGTACCAGCGCACCAGCAT-3Ј) and oMP530 (5Ј-GGAAT TCCAGATCGAACGCCAATTCGG-3Ј). A second PCR was performed using the internal primers to select the correct PCR fragment. A recA mutant was constructed by using the PCR product of oMP61 and oMP530 to obtain a 1.2-kb EcoRI fragment of the recA gene which was cloned into the suicide plasmid pMP5285, resulting in pMP6604. This construct was electroporated into S17-1 (36) and transferred to PCL1171 by parental mating. After selection on kanamycin, the mutation of recA was confirmed via Southern hybridization (31) by using the AlkPhos direct labeling kit for detection (Amersham Biosciences, Little Chalfont, Buckinghamshire, United Kingdom) according to the manufacturer's recommendations. The same approach was used to isolate the recB gene from PCL1171 by using internal primers oMP760 (5Ј-TTCGACAGCGGCAGCCTG TTCACCCAGA-3Ј) and oMP761 (5Ј-AAGGCGTA(CT)AT(GC)GC(AG)TGC TTGGGGTCGCCGAT-3Ј). To construct a recB mutation, a 1.2-kb PCR product obtained by using oMP840 (5Ј-CAGGAATTCGAAGGTTGGTG-3Ј) and oMP761 was cloned into the pGEM-TEasy vector (Promega Crop., Madison, Wis.) and subsequently transferred to the suicide plasmid pMP5285 by using EcoRI, resulting in pMP6605. pMP6605 was electroporated into S17-1 (36) and transferred to PCL1171 via parental mating. After selection on kanamycin, the mutation of recB was confirmed by Southern hybridization (31). To determine the effects of recA and recB mutations on phase variation, phase II colonies were replated on King's medium B and the colony morphology was determined after at least 2 days of growth at 28°C.

RESULTS
Isolation and preliminary characterization of stable phase II Tn5luxAB transposon mutants. The frequency of switching between phase I and phase II was determined for PCL1171 at 6.0 ϫ 10 Ϫ5 and 8.2 ϫ 10 Ϫ2 switches per cell per generation for phase I to phase II and for phase II to phase I, respectively. To identify genes involved in phase variation, we constructed a Tn5luxAB transposon library from PCL1171, which has a strong preference for the phase I phenotype, by using pRL1063a (40). From this library, 26 Tn5luxAB transposon mutants expressing a stable phase II phenotype were selected out of approximately 1,200 transformants. No mutants expressing a stable phase I phenotype were isolated. Observations from both the mutant library and growth of the wild type on solid medium indicated that the majority of the phase II bacteria switch back with a high frequency to the phase I phenotype, while occasionally phase II bacteria which are more stable can be isolated. In contrast to the 8.2 ϫ 10 Ϫ2 switches per cell per generation observed for the wild type, no switch back to the phase I phenotype was observed in these phase II mutants. These mutants are referred to below as stable phase II mutants. Sequencing of the flanking regions of several transposon insertions of these stable mutants showed homology to genes with widely different functions (data not shown). Since we had previously observed that a Tn5 insertion in gacS could lock PCL1171 in the phase II phenotype (38), we tested the possibility that, in addition to the Tn5luxAB insertion, at least some of the mutants had accumulated mutations in gacS or gacA. For this test we used plasmids pMP6562 (harboring a gacS homolog of PCL1171), pMP5565 (harboring a gacA homolog of Pseudomonas sp. strain PCL1446), and pMP6603 (harboring both a gacA and a gacS homolog). It appeared that out of 26 mutants, 13 could be complemented to the phase I phenotype by using pMP6562 (gacS) and the other 13 could be complemented to the phase I phenotype by using pMP5565 (gacA). Complementation using pMP6603, harboring both the gacA and the gacS gene, restored the phase I phenotype in all 26 mutants. Introduction of the empty parental vector pMP6010 did not alter the phase II phenotype. Subsequent restreaking of the complemented mutants resulted in phase I colonies from which phase II sectors appeared as in the wild type.
To study the effect of a gacS mutation on growth, we compared the growth curve of the previously identified gacS Tn5 mutant PCL1572 and those of three other phase II mutants, PCL1574, PCL1575, and PCL1576, with that of wild-type phase I bacteria (Fig. 1). It appeared that mutation of the gacS gene reduced the length of the lag phase and reduced the generation time compared to that for the wild type (50 Ϯ 4.2 versus 60 Ϯ 4.2 min, respectively) ( Fig. 1). Counting the numbers of CFU present at the start of growth showed no differences between the gacS mutants and the wild type, indicating that the reduced lag phase was not due to a higher number of dead cells in the wild-type culture.
Molecular basis of phase II phenotype of phase II mutants. Since only mutations accumulating in the gacA gene have been analyzed (4), we decided to focus on the role of gacS in phase II variation of Pseudomonas sp. strain PCL1171. We isolated the gacS genes from 10 phase II mutants. Nine of the isolated gacS genes, cloned into pMP5565 (gacA), could not restore the phase I phenotype in PCL1572, showing, as a control, that the mutations indeed interfered with GacS function. The gacS gene isolated from PCL1574 did restore the phase I phenotype in a small number of transformants. The gacA gene was included in the screening to eliminate the effect of possible mutations in the gacA gene of PCL1572. Complementation using a wild-type gacS gene in pMP5562 always restored PCL1572 to the phase I phenotype. Three of the mutated gacS genes were sequenced and aligned with the wild-type gacS sequence. The gacS genes from strains PCL1574 and PCL1575 were found to harbor a point mutation ( Table 2). The gacS gene from PCL1576 harbored a 115-bp inversion, which was flanked by an inverted repeat (5Ј-C 327 AGCAG-3Ј and 5Ј-C 442 TGCTG-3Ј). This repeat was created due to a point mutation at nucleotide position 329 (aA 329 c3aG 329 c) ( Table 2 and Fig. 2C). The inversion of the 115-bp fragment introduced a stop codon at amino acid position 134 of the predicted GacS protein.
Instability of mutations in gacA and gacS. The stable phase II mutant PCL1574 could be complemented to the phase I phenotype by using pMP5562 (gacS) ( Table 2); complementation of PCL1572 (gacS::Tn5luxAB) by using the gacS gene isolated from PCL1574 resulted in a mixture of phase I and phase II colonies. Since this gacS gene did not restore the phase I phenotype in all PCL1572 transformants, the construct harboring the gacS gene from PCL1574 was reisolated from colonies in which gacS did complement PCL1572. Sequencing of the gacS gene from these phase I colonies showed that the point mutation was no longer present.
Similar observations were made for the stable phase II mutant PCL1575. This strain harbors a point mutation in its gacS gene resulting in a premature stop codon (Table 2). Whereas initially introduction of the wild-type gacS gene completely restored the phase I phenotype in this mutant, the introduction of pMP6562 (gacS) in subsequent complementations also resulted in a number of phase II colonies. These could be complemented by using pMP5565 (gacA). This changed PCL1575 derivative was designated PCL1575a (Table 2). In addition, PCL1575 phase II colonies were found which could be complemented neither by using pMP5565 nor by using pMP6562. One such derivative was designated PCL1575b (Table 2). Sequencing of the gacS gene of PCL1575a after PCR amplification showed that the point mutation identified in PCL1575 had been removed (T 154 aa3C 154 aa) in PCL1575a while a phase II phenotype was maintained. This point mutation was still present in PCL1575b (T 154 aa) ( Table 2).
Molecular basis of phase II phenotype in the wild-type strain PCL1171. To test whether mutations in the gac system also form the molecular basis for the highly unstable phase II VOL. 187, 2005 gacS AND PSEUDOMONAS COLONY PHASE VARIATION 595 phenotype in the wild-type strain PCL1171, the gacS gene from phase II sectors was amplified by PCR. Since analysis of the stable phase II mutants showed that a phase II phenotype could be the result of mutations in either gacA or gacS (Table  2), and since wild-type phase II sectors cannot be maintained due to the instability of the phase II phenotype, we needed a method to distinguish between gacA and gacS mutations. For this purpose the amplified gacS gene from phase II sectors was cloned into pMP5565 (containing a wild-type gacA gene) and the resulting construct was used to complement PCL1572 (gacS::Tn5luxAB). Of 13 gacS genes isolated from these phase II sectors, 5 were unable to complement PCL1572 (Table 2). One, termed sector 16, resulted in a mixture of phase I and phase II colonies, whereas the other seven gacS genes complemented PCL1572 to the phase I phenotype, showing that in the latter seven sectors GacS function was not affected and that the phase II phenotype is most likely caused by a mutation in the gacA gene. As controls, the gacS genes from sectors 2 and 5, which could restore the phase I phenotype in PCL1572, were sequenced. Sector 2 harbored one point mutation, while no changes were identified in the gacS gene isolated from sector 5 ( Table 2).
Analysis of five out of the six noncomplementing gacS genes revealed various mutations ( Table 2). The gacS gene from sector 1 harbored a total of three mutations, a 12-bp deletion in combination with two point mutations, including one silent mutation ( Table 2 and Fig. 2A). The gacS genes from sectors 4 and 18 were found to harbor one and two point mutations, respectively (Table 2), while the gacS gene from sector 9 harbored a ϩ1 frameshift due to a 1-nucleotide insertion ( Table 2 and Fig. 2B), resulting in the introduction of stop codons downstream of the insertion. The gacS gene isolated from sector 16, reisolated from PCL1572 colonies not complemented by this gene, harbored a 307-bp deletion from nucleotide positions 2236 to 2542. A 10-bp tandem repeat (5Ј-G 2226 CCATCACCG-3Ј and 5Ј-G 2523 CCATCACCG-3Ј) was identified at the borders of the deletion as the result of a single point mutation (tT 2231 a3tC 2231 a) ( Table 2 and Fig. 2D). The gacS gene from sector 16, isolated from PCL1572 colonies complemented to the phase I phenotype, did not harbor any mutations. In addition, a silent point mutation was identified ( Table 2).
Phase II sectors from PCL1171 are highly unstable, and restreaking coincides with a high frequency (8.2 ϫ 10 Ϫ2 switches per cell per generation) of switching back to the phase I phenotype. This high frequency enabled us to analyze the gacS gene from phase II sectors after a switch back to the phase I phenotype. The three mutations previously identified in phase II sector 1 (a 12-bp deletion and two point mutations) were absent in phase I bacteria growing from sector 1 phase II cells. Also, all gacS mutations present in cells of sector 4 (single point mutation), sector 9 (1-bp insertion), sector 16 (307-bp deletion), and sector 18 (two point mutations) had been restored to the wild-type gacS sequence after switching to the phase I phenotype.
Role of recA and recB in phase variation. Based on the observation that a wide variety of mutations in the gacS gene, in both stable and unstable phase II bacteria, could be restored efficiently to the wild-type gacS sequence, we hypothesized that large DNA rearrangements could play a role in the switch from phase II to phase I. To test this hypothesis, a recA mutation was introduced into PCL1171, by using the suicide construct pMP6604, resulting in strain PCL1586. Strain PCL1586 (recA::Km r ) showed a phase I phenotype from which phase II sectors originated with a high frequency. After replating of these phase II sectors, phase I colonies and phase II colonies with phase I sectors were found.
The suicide construct pMP6605 harboring a recB fragment was used to introduce a recB mutation in PCL1171. Mutation of recB reduced the viability of the cells, which, however, still showed phase variation resulting in single phase I and phase II colonies. Replating of the phase II colonies resulted in phase I colonies and phase II colonies.

Molecular nature of mutations in gacS.
Pseudomonas sp. strain PCL1171 displays a high frequency of reversible phase variation between a thick, opaque phase I and a thin, translucent phase II colony, thereby regulating secondary metabolism and production of exoenzymes (38). Since we observed that mutation of the gacS gene locked the bacteria in a phase II phenotype (38), we studied the role of gacA/gacS in phase variation. Complementation analysis of a total of 39 phase II derivatives showed that mutations in gacA (20 mutants) and gacS (19 mutants) are responsible for the phase II phenotype ( Table 2). Introduction of pMP6603 (gacA gacS) restored all mutants to a phase I phenotype, showing that spontaneous mutation of the gacA/gacS two-component regulatory system forms the basis of the switch from phase I to phase II in phase variation of Pseudomonas sp. strain PCL1171.
The gacS genes isolated from phase II bacteria harbored random point mutations, deletions, DNA rearrangements, and an insertion ( Table 2 and Fig. 2). In addition to the mutations identified in the gacA gene of Pseudomonas fluorescens CHAO by Bull et al. (4) (point mutations, a 3-bp deletion, and a 1 insertion), the gacS gene of PCL1171 also harbored a 12-and a 307-bp deletion and a 115-bp inversion ( Table 2 and Fig. 2), showing that a large diversity of mutations, both in nature and in location, can accumulate in the gacA and gacS genes ( Table  2 and Fig. 2). We observed that not all mutations in the gacS gene affected GacS function (for example, single mutations in the periplasmic loop [e.g., sector 2] did not), a finding consistent with the observation that deletion of this domain did not affect GacS function in P. fluorescens CHAO (41). Accumulation of mutations in gacA and gacS has been reported previously for Pseudomonas syringae pv. syringae strain B728a (28), P. fluorescens CHAO (4,7,20), Pseudomonas aureofaciens 30-84 (6), and Pseudomonas chlororaphis isolate SPR044 (34). However, for none of these was a reversion to the wild-type phenotype reported. In contrast, in our strain the introduction and removal of mutations in gac is the basis of the observed phase variation, and the majority of the phase II cells switch back to the phase I phenotype at a high frequency, which correlates with the complete removal of the mutations from the gacS gene. Interestingly, in sector 16 we found that the 307-bp deletion was restored upon the switch to the phase I phenotype. But, since complementation of PCL1572, by using the gacS gene from sector 16, resulted in a mixture of phase I and II colonies, it is likely that a mixture of gacS genes was already present before PCR amplification. Therefore, we cannot exclude the possibility that the deletion identified cannot be repaired. This deletion mutant could be part of a subpopulation of stable phase II bacteria, whereas the majority of the bacteria in this sector harbor other mutations (e.g., gacA) and, as observed on plates, can switch back to the phase I phenotype.
Occasionally, more-stable phase II bacteria can be isolated. The isolation of stable phase II derivatives whose genetic basis switched between mutations in gacA, mutations in gacS, and mutations in both (Table 2) shows that even within such a "stable" phase II phenotype, mutations still can be removed and (re)introduced. Our results indicate that all mutations in the gacS and gacA genes of PCL1171 can be restored spontaneously. This may occur completely, resulting in a switch to the phase I phenotype, but, in the case of a stable phase II phenotype, mutations may not be removable, or the removal may coincide with the introduction of new mutations.
Growth experiments using gac mutants (Fig. 1) showed that mutation of gac decreases the length of the lag phase and increases the growth rate. Therefore, one explanation for the accumulation of gac mutants at a high frequency is increased growth competitiveness, especially under conditions where stationary and exponential growth often switch, as may be ex- The functional gacA/gacS component was determined by the introduction of pMP6562 (gacS) and pMP5565 (gacA). A plus sign indicates which component is functional; a minus sign indicates which component harbors mutations and is nonfunctional. In addition, strains PCL1565, PCL1568, and another eight phase II mutants had a functional gacA gene (gacS mutants), while PCL1564, PCL1573, PCL1566, PCL1568, and another eight phase II mutants had a functional gacA gene (gacS mutants), while PCL1564, PCL1573, PCL1566, PCL1567, and another nine phase II mutants had a functional gacS gene (gacA mutants).
c The functional gacA/gacS component was determined by complementation of PCL1572 (gacS::Tn5luxAB) by the gacS gene isolated from a phase II sector. A plus sign indicates that a phase I morphology was restored in PCL1572; a minus sign indicates that a phase II morphology was maintained after introduction of the gacS gene. In addition, the gacS genes isolated from another five sectors did complement PCL1572 (gacS::Tn5luxAB), suggesting that mutation of the gacA gene was responsible for the phase II phenotype in these sectors. d See Fig. 1 pected in colonies and in the rhizosphere. This explanation is consistent with the suggestion by Schmidt-Eisenlohr et al. (34) for P. chlororaphis isolate SPR044 that having a mixed population could, based on the growth characteristics of the gac mutants, provide a competitive advantage, especially in changing and heterogeneous environments. The observation that the gac subpopulation does not replace the wild-type population in culture, as found for PCL1171, shows the effect of the reversibility of the gac phenotype and indicates that, at least under laboratory conditions, there is no direct environmental selection in favor of these mutants. What is the mechanism of phase variation in PCL1171? According to the literature, phase variation can regulate gene expression via four mechanisms: (i) slipped-strand mispairing, (ii) differential methylation, (iii) genomic rearrangements, and (iv) random unprogrammed variation (3,12). To our knowledge, only two molecular mechanisms of phase variation in Pseudomonas spp. have been described. In Pseudomonas putida DOT-T1E, the expression of flagella is controlled via slipped-strand mispairing (35), and in P. tolaasii, a 661-bp spontaneous, reversible duplication in pheN controls colony morphology and pathogenicity (9). In addition, although no mechanism is described, the sss gene, encoding a site-specific recombinase, was suggested to play a major role in the phenotypic variation of P. fluorescens F113 during root colonization. The latter variation is combined with selection for gac mutants (32).
Considering the mechanism of phase variation via gac, none of the sequenced gacA/gacS genes harbors repeat tracts. Since phase variation in PCL1171 is dependent on mutation of gac, it is not epigenetic. Therefore, slipped-strand mispairing or differential methylation can be excluded as the major basis of phase variation via gacA/gacS in PCL1171. Genomic rearrangements were present in some gacS genes isolated from phase II bacteria and were based on the presence of inverted or tandem repeats (Table 2 and Fig. 2). This suggests a role for site-specific recombinases (30) in mutation of these gac genes. But, since these rearrangements could take place only after the occurrence of point mutations, introducing repeats (Fig. 2) which can be recognized by site-specific recombinases, we con- clude that genomic rearrangements play only a minor, secondary role in the phase variation of PCL1171. This conclusion differs from what has been reported for P. fluorescens F113, in which a site-specific recombinase is responsible for the majority of the phenotypic variation (32). In PCL1171, the accumulation of mutations in gacA/gacS was previously suggested to be suppressed by MutS-dependent mismatch repair, since mutation of the mutS gene resulted in a dramatic increase in the frequency of phase variation (38). Furthermore, based on the mechanism of repair (23), all point mutations identified in gacS (Table 2) could, upon introduction, have been recognized by MutS. This suggests that inefficient repair via MutS contributes to the mutations we found in gacA/gacS. In addition, the spontaneous mutation rate in PCL1171, measured by the frequency of spontaneous rifampin-resistant mutants, increases upon mutation of mutS from 5 ϫ 10 Ϫ7 to 3 ϫ 10 Ϫ5 mutants per cell per generation (S. de Weert, personal communication). This frequency is comparable to the frequency of phase variation in the wild type (6 ϫ 10 Ϫ5 switches per cell per generation), supporting the notion that inefficient repair of replication-related mismatches via MutS is responsible for the accumulation of random mutations in gacA/gacS. This observation supports the hypothesis that the gacA and gacS genes are not hot spots for mutations. Under growth-limiting conditions, microorganisms can increase their mutation frequency (16,25). Under these conditions, gac mutants could be selected for, either on the basis of their growth characteristics or by preventing restoration of the mutations. Our observations suggest that the mechanism of phase variation in PCL1171 is random and unprogrammed variation based on the introduction and restoration of random point mutations, insertions, deletions, and rearrangements in gacA/gacS. Unprogrammed variation is not dependent on specific DNA features, and spontaneous duplications and deletions have been reported to control, for example, capsule genes in Streptococcus pneumoniae (39), pathogenicity and colony morphology in P. tolaasii and Ralstonia solanacearum (9,26), and the virulence of Bordetella bronchiseptica (24).
Unprogrammed phase variation, as observed in PCL1171, is reversible in the sense that the original sequence of the wildtype gene is restored. At this moment no molecular mechanism for the restoration of the mutations in the strains mentioned has been elucidated. A preliminary study on the reversion of the phase II phenotype in PCL1171 focused on the possible role of homologous recombination. To test this idea, we constructed a RecA and a RecB mutant. Both genes are involved in homologous recombination (18) and, in some cases, in phase variation (12,17,21). Neither recA mutation nor recB mutation affected the ability of phase II bacteria to switch back to the phase I phenotype. Although this result cannot exclude the possibility that recombinational processes are involved in this phase variation, it shows that RecA and RecB are not essential for the repair of mutations in gacS and gacA of PCL1171. A similar observation was made for P. tolaasii (37), where RecA was not needed for the removal of the duplication in pheN (37).