Novel Role of mfd: Effects on Stationary-Phase Mutagenesis in Bacillus subtilis

Abstract

Previously, using a chromosomal reversion assay system, we established that an adaptive mutagenic process occurs in nongrowing Bacillus subtilis cells under stress, and we demonstrated that multiple mechanisms are involved in generating these mutations (41, 43). In an attempt to delineate how these mutations are generated, we began an investigation into whether or not transcription and transcription-associated proteins influence adaptive mutagenesis. In B. subtilis, the Mfd protein (transcription repair coupling factor) facilitates removal of RNA polymerase stalled at transcriptional blockages and recruitment of repair proteins to DNA lesions on the transcribed strand. Here we demonstrate that the loss of Mfd has a depressive effect on stationary-phase mutagenesis. An association between Mfd mutagenesis and aspects of transcription is discussed.

Since the mid-1950s, microbiologists have been aware of mutations occurring in nondividing populations of cells (22, 29). The formation of these mutants was alternatively termed “starvation-associated mutagenesis” (29), “adaptive mutation” (8), or “stationary-phase mutagenesis” (13). Recently, variations of this phenomenon have been investigated with Escherichia coli (8, 29, 38), Pseudomonas (33), Bacillus subtilis (41), and the eukaryotic yeast Saccharomyces cerevisiae (15, 40). These phenomena reveal that starving populations of cells can acquire mutations favoring growth after the application of selection.

While the phenomenon of stationary-phase mutation is widespread, it is clear that the mechanism(s) by which it arises varies from organism to organism. To date, the most favored system for studying adaptive or stationary-phase mutagenesis is the RecA-dependent E. coli FC40 system investigated by, among others, the laboratories of Cairns, Foster, Rosenberg, and Roth (7, 8, 11, 18). Recent results strongly suggest that this mutagenesis is the result of gene amplification followed by mutation in a transiently growing population of cells (18). In light of this, we chose to investigate the possibility that transiently growing cells may play a role in a B. subtilis system containing three chromosomal point mutations. The phenomenon of transcriptional mutagenesis, or retromutagenesis, whereby RNA polymerase bypasses an unrepaired DNA lesion or otherwise produces an altered mRNA, which is then translated into a protein of altered function, could provide a transient growth advantage for the cell. This mechanism has been proposed for other model systems, including eukaryotes (6, 9).

We have previously shown the existence of one such mutagenic phenomenon occurring during stationary phase in B. subtilis cells starved for amino acids (41). This mutagenic process appears to enhance the survivability of cell populations undergoing nutritional stress. In brief, isogenic strains of B. subtilis carrying three amino acid auxotrophies conferred by hisC952 (amber), metB5 (ochre), and leuC427 (missense) are incubated on medium lacking one of the required amino acids. After several days of incubation, revertant colonies prototrophic for the missing amino acid begin to appear. We propose that this adaptive process is a developmentally regulated response. In B. subtilis, it is influenced by the ComA and ComK transcription factors, which are normally required for cell competence, but occurs in the absence of a functional RecA protein (41). We have also shown the involvement of a member of the Y superfamily of DNA polymerases, YqjH (43), in this adaptive process. Also, we have demonstrated that increasing the amount of mismatch repair proteins MutS and MutL significantly reduces the frequency of reversion to prototrophy (25) in this system while not affecting the mutation frequency in growing cells. Decreased synthesis and/or titration of the MutS and MutL proteins due to an increased level of DNA damage in stationary-phase cells may contribute to a higher mutation frequency in the stressed population, as may the presence of error-prone DNA polymerases.

Transcription repair coupling is the preferential repair of the transcribed strand of DNA at a rate higher than that for the nontranscribed strand or untranscribed double-stranded DNA. In E. coli, the product of the mfd gene is the transcription repair coupling factor protein (TRCF). TRCF (the Mfd protein) has been linked to both to the phenomena of mutation frequency decline (MFD), which is a rapid decrease in DNA damage-induced mutations following a transient stoppage of protein synthesis (48), and transcription-coupled repair (23, 37). E. coli TRCF/Mfd is a 130-kDa protein with a potential UvrA-binding domain, homology to the E. coli RecG helicase, and a potential leucine zipper motif (36). E. coli TRCF interacts specifically with RNA polymerase (RNAP) stalled at a UV-induced lesion or other distortion of the DNA and dissociates the RNA polymerase ternary complex (35). These observations have suggested a model in which TRCF recruits the nucleotide excision repair machinery, via its interaction with UvrA, to sites of damage on the transcribed DNA strand after displacing the stalled RNAP and the nascent RNA chain (36).

For B. subtilis, the mfd gene has been previously described as producing a 135-kDa protein with homology to E. coli TRCF and as involved in strand-specific DNA repair, lesion-stalled RNA polymerase displacement, and recombination functions (1). It has also been implicated in the mediation of catabolite repression (53, 54). Furthermore, Mfd may act as a translocase, realigning backtracked RNAP complexes (24). The work presented here demonstrates the previously unknown involvement of TRCF in the process of adaptive mutagenesis in B. subtilis. The involvement of mfd in this system is of considerable interest, since mfd-like genes, protein homologs, and transcription repair coupling phenomena have been observed in a wide range of organisms, including humans and other eukaryotes (20, 45).

To this end, we have constructed a TRCF-deficient (mfd-deficient) mutant of B. subtilis isogenic to our research strain YB955. One might expect that in the absence of TRCF, mutation rates would increase as DNA lesions go unrepaired. We show here that the Mfd-deficient strain is actually significantly diminished in its capacity to generate prototrophic revertants in stationary phase.

MATERIALS AND METHODS

Bacterial media and strains.

Strains and plasmids used in this study are given in Table 1. B. subtilis strains were maintained on tryptose blood agar base (TBAB) (Difco Laboratories, Detroit Mich.). Liquid cultures were routinely grown in Penassay broth (PAB) (antibiotic medium 3; Difco) supplemented with Ho-Le trace elements (12). Tetracycline and kanamycin were added, as appropriate, to final concentrations of 5 to 10 μg/ml and 10 μg/ml, respectively. E. coli DH5α cultures were grown in Luria-Bertani broth supplemented with ampicillin to a final concentration of 100 μg/ml.

TABLE 1.

Bacillus subtilis strains and plasmids

Strain or plasmid	Relevant genotype	Reference
YB955	hisC952 metB5 leuC 427 xin-1 spβSENS	12a
YB9800	YB955 Δmfd::tet	This study
YB9801	YB955 Δmfd::tet	This study
MPRYB509	YB9801 carrying pMPRYB503	This study
MPRYB511	YB9801 carrying pDG 148-Stu	This study
MPRYB513	YB9801 carrying pMPRYB512	This study
pMPRYB512	pMPRYB503 carrying an internal deletion in mfd	This study
pMPRYB503	pDG148-Stu carrying full-length mfd	This study
pDG148-Stu	bla kan carrying spac promoter	19
pDG1515	tet	13a

Reversion experiments were performed on Spizizen minimal medium (SMM) (39) supplemented with 0.5% glucose, 50 μg of two of the required amino acids per ml, 200 ng of the third required amino acid, insufficient for sustained growth (approximately two doublings) and included to improve plating efficiency and survival, and 50 μg (each) of isoleucine and glutamic acid per ml to enhance viability of the cells as previously demonstrated (42). For example, medium selective for His⁺ revertants contains 50 μg methionine and leucine and 200 ng histidine. The trace amount of the amino acid for which the reversion is being selected is added in order to mimic the depletion of nutrients that would be occurring in the soil.

YB955 is a prophage-cured (SPβ^SENS xin-1) derivative of B. subtilis 168 carrying the hisC952, metB5, and leuC427 alleles (51). YB9800 and YB9801 are isogenic mfd-deficient derivatives of YB955 constructed by integrating a tetracycline antibiotic cassette into the mfd structural gene (described below). For each experiment, YB955 was run as a control alongside the isogenic mutant strain.

Strain construction.

DNA containing the mfd gene (GenBank accession no. NC_000964) from YB955 was obtained by PCR (30) using the oligonucleotide primers 5′-CGGGATCCGCCGCCATGACAGACAGCAAAAAAG-3′ and 5′-AACTGCAGATCCGCTTTCCGCTCAATGTCCTC-3′. Compatible cohesive ends were generated by digestion of the PCR product with BamHI and PstI (underlined), and the product was cloned into the similarly digested plasmid vector pUC18. A tetracycline cassette excised from pDG1515 (Bacillus Genetic Stock Center) was inserted into the cloned mfd gene using an internal SwaI restriction site. The resulting vector was linearized by restriction digestion and used to transform YB955 by natural competence (52). The resultant mfd-disrupted strain was selected on TBAB containing 5 μg/ml tetracycline and termed YB9800. To ensure that only a single recombination event occurred, limiting amounts of genomic DNA isolated from YB9800 were used to transform YB955 to Tet⁺, forming strain YB9801. Disruption of the chromosomal mfd gene was confirmed each time by PCR.

Design of a plasmid to overexpress mfd.

A plasmid to overexpress mfd from the isopropyl-β-d-thiogalactopyranoside-inducible spac promoter of pDG148-Stu (19) was constructed as follows: the mfd gene was first amplified by PCR using 0.1 μg of chromosomal DNA from B. subtilis YB955 and the oligonucleotide primers 5′-AAGGAGGAAGCAGGTATGGACAACATTCAAACC-3′ (forward) and 5′-AAGGAGGAAGCAGGTCTTGAAATTAGTATCCGTC-3′ (reverse). Amplification was performed with Vent DNA polymerase (New England BioLabs, Beverly, MA). The PCR product purified from a low-melting-point agarose gel was ligated into StuI-treated pDG148-Stu to generate pMPRYB503, which was replicated in E. coli DH5α. The correct orientation of mfd cloned into PDG148-Stu was verified by restriction analysis with EcoRI and SmaI/BamHI enzymes. The plasmid pMPRYB512 was generated by treating pMPRYB503 with BsiWI to release a 2,276-bp fragment of the mfd open reading frame downstream of nucleotide 1255. Plasmids pMPRYB503, pDG148-Stu, and pMPRYB512 were introduced by transformation into competent cells of B. subtilis strain YB9801 (mfd) to generate strains MPRYB509, MPRYB511, and MPRYB613, respectively.

Stationary-phase mutagenesis assay.

The stationary-phase mutagenesis assay was conducted as described by Sung et al. (41). Ten milliliters of cells were grown in PAB medium supplemented with tetracycline in a 50-ml nephloflask at 37°C with aeration to 90 min after the cessation of exponential growth (T-90). Cultures were harvested by centrifugation at 10,000 × g for 10 min at room temperature and then resuspended in 10 ml of Spizizen minimal salts (SMS). Then, 100 μl of cells were plated in quintuplicate on SMM supplemented with 200 ng/ml of histidine, methionine, or leucine (depending on the reversion selected) and 50 μg/ml of the other two required amino acids. Plates were incubated at 37°C for the duration of the experiment. Each experiment was repeated at least three times. In addition, each time YB9801 (mfd) was examined for the ability to perform stationary-phase mutagenesis, a YB955 control was tested simultaneously. The number of revertants appearing as discrete colonies was scored daily. The initial number of bacteria plated for each experiment was determined by titration on SMM supplemented with all the essential amino acids.

The data for numbers of CFU per day were normalized to the initial number of bacteria plated and are represented as the total accumulation of revertants over time.

A viable count of nonrevertant B. subtilis background on the selective media was determined as follows: three agar plugs were removed using sterile Pasteur pipettes from areas of each selection plate where no revertant colonies were observed, beginning on day 2 and done every other day thereafter. The three plugs from each plate were suspended in 400 μl of 1× SMS, mixed, serially diluted, and spread plated on SMM containing all of the essential amino acids (50 μg/ml). Resultant colonies were counted after 48 h of growth at 37°C (41).

Fluctuation test.

Reversion rates during exponential growth were determined by fluctuation testing as previously described (41). In brief, a saturated culture of the strain in question and the corresponding isogenic parental strain were diluted 1:10⁵ into 38 (or more) 28-mm culture tubes containing 2 ml of PAB medium. Cultures were grown to saturation overnight, and a 1-ml sample was extracted from each culture, resuspended in SMS, and plated on selective media. Colonies arising on each plate were counted after 48 h of growth at 37°C. Exponential reversion rates were then determined for each strain by the Ma-Sandri-Sarkar maximum-likelihood method, instead of the Lea-Coulson formula (28), as previously described. An Excel spreadsheet was constructed that iteratively refines the value of m, the number of mutations per culture, based on the fit of the observed data to a mathematically derived Luria-Delbrük distribution model generated from a previous estimate of m. This model was generated using the following equations:

where p_r represents the probability of r mutants arising in a given culture. The final value of m is that which maximizes the product of the individual probabilities that describe the data set. The 95% confidence intervals around ln(m) were estimated by the following method, where C is the number of cultures:

As before, the mutation rate per cell per doubling was defined to be m/2N_t, where N_t is the plated count. To get sufficient revertants to be statistically meaningful, 76 cultures were tested for reversion to leucine prototrophy. For histidine and methionine reversion, 38 cultures were sufficient. As previously described, three cultures were run in tandem with each experiment and titrated to determine the plated population.

Reconstruction test.

To demonstrate that the revertant colonies arose due to mutations occurring during starvation conditions and do not represent slow-growing variants arising during the initial exponential outgrowth, reconstructions were performed as previously described (43). In brief, stationary-phase revertants were picked on days 6 and 7 of the assay, and the isolated colonies were streaked on SMM lacking the appropriate amino acid. A random sampling of revertants were individually grown to saturation in PAB plus trace elements. Approximately 500 to 1,000 CFU/ml of the revertant culture was diluted into 3 × 10⁸ CFU/ml of the corresponding parental strain. Then, 100 μl of the mixed population and of the nonmixed revertant control was spread-plated on selective SMM appropriate to the reversion being tested. Revertant colonies were recorded after 48 h of growth and compared to the 50 to 100 colonies expected for inoculation with a true prototrophic revertant as opposed to the few growth-dependent reversions expected to occur in the parental population. Selective SMM plates with 100 μl of parental culture were tested simultaneously to observe reversions in the background. Reconstruction tests were repeated for several of the YB955 and YB9801 revertants to confirm the consistency of the results.

Sequencing of DNA.

Genes of interest were amplified in individual PCRs using primers and conditions described previously (41). Sequencing was performed by the Nevada Genomics Center on a Prism 3730 DNA analyzer using the ABI BigDye Terminator v1.1 cycle sequencing kit (Applied Biosystems). Sequencing primers for each allele are as previously described (41).

Genetic and molecular biology techniques.

Preparation of competent E. coli or B. subtilis cells and their transformation with DNA were performed as previously described (3, 31). Chromosomal DNA from B. subtilis was purified according to the protocol of Cutting and VanderHorn (16). Small-scale preparation of plasmid DNA from E. coli cells, enzymatic manipulations, and agarose gel electrophoresis were performed by using standard techniques (31). Medium-scale preparation and purification of plasmid DNA were accomplished by using commercial ion-exchange columns according to the instructions of the supplier (QIAGEN, Inc., Valencia, Calif.).

RESULTS

Loss of Mfd reduces the prototrophic reversion rates.

To establish the impact of the loss of transcription repair coupling factor (TRCF) in B. subtilis, we first constructed a disruption in the mfd gene of YB955 as described in Materials and Methods, resulting in strain YB9801. The loss of mfd (YB9801) resulted in a reduction in revertants arising on selective media during starvation. Selection for Met⁺ and His⁺ prototrophy (Fig. 1A and B) shows a significant reduction in the number of mutants arising in YB9801 after the second day relative to numbers for the YB955 control. However, there was an even more striking, nearly complete, absence of revertants to leucine prototrophy in this strain after the third day of incubation. Figure 1C (leucine reversion rate) illustrates the frequency with which Leu⁺ revertant colonies appeared on the leucine-deficient medium (selective SMM) compared to the YB955 control results. These results were reproduced by different laboratory personnel. After several attempts to obtain Leu⁺ YB9801 revertants with more than 10¹⁰ cells plated in total, 6 revertants were finally recovered. Similar reductions in all three reversion frequencies were also seen with YB9800 (data not shown).

FIG. 1.

Stationary-phase reversion frequency. Number of revertants to histidine (A), methionine (B), or leucine (C) prototrophy occurring during prolonged incubation at 37°C on minimal selective media is shown for strain YB955 (closed circles) or YB9801 (mfd deficient) (open circles). Numbers of revertants have been normalized to the number of cells plated on day 0 as determined by titration onto complete minimal medium. Error bars represent ±1 standard error from three independent trials. The experiment was repeated at least three times.

It could be argued that the reduction in mutants arising in the prototrophy assay is a result of a possible loss of viability of the Mfd-deficient strain in the absence of required amino acids. Survival of YB9801 on selective media is shown in Fig. 2. After the second day of incubation, no loss of viability or net growth was observed for the duration of the experiment and titers were consistent between YB9801 and YB955. These results are comparable to those obtained from tandemYB955 controls and previous experiments (41, 43), which also demonstrated no net increase or decline in viable cell counts for the Bacillus strains that have been tested.

FIG. 2.

Survival of strain YB9801 on selective media. Plot shows survival of YB9801 obtained by titrating cells from agar plugs taken at 2-day intervals from histidine-deficient (squares), leucine-deficient (circles), or methionine-deficient (triangles) medium as described in Materials and Methods. Closed symbols represent YB955; open symbols represent YB9801. The experiment was repeated at least twice.

The cloned mfd⁺ gene restores the stationary-phase reversion rates.

Complementation of the disruption in YB9801 with Mfd expressed from an inducible plasmid (strain MPRYB503) alleviated the deficiency in all three reversion categories (Fig. 3A to C). Due to some inherent leakiness in the spac promoter of the pDG148-Stu vector, no isopropyl-β-d-thiogalactopyranoside was added to restore the Mfd deficiency, suggesting that Mfd is required only in low amounts in stationary-phase cells to support adaptive mutagenesis. As a control, rates of reversion for YB9801 containing pDG148-Stu with (pMPRYB503) or without (pMPRYB512) the mfd gene were compared. Only inclusion of pMPRYB503 with mfd⁺ restored the accumulation rate of the revertants. Cell survival was unaffected by the addition of either plasmid (Fig. 4). Partial disruption of the mfd gene cloned in pMPRYB503 (resulting in pMPRYB512; see Materials and Methods) resulted in loss of the Mfd⁺ phenotype (data not shown), demonstrating that the complementation seen was due to the transcription of the cloned mfd gene.

FIG. 3.

Stationary-phase reversions of mfd-complemented strain. Accumulated revertants to histidine (A), methionine (B), or leucine (C) prototrophy in strain MPRY509 (pDG148-mfd; closed circles) or MPRYB511 (pDG148-Stu; open circles) on appropriate selective media. Error bars represent ±1 standard error from three independent trials. Each experiment was performed at least three times.

FIG. 4.

Survival of YB9801 (mfd deficient) when complemented in trans. Plot shows survival of strains MPRY509 (YB9801 pDG148-mfd) (A) and MPRYB511 (YB9801 pDG148-Stu) (B) during histidine (black), methionine (light gray), and leucine (dark gray) starvation. Three plugs of bacteria containing agar were taken from selection plates, and titers were determined on media containing all essential amino acids (as described in Materials and Methods) every other day for testing of the viability of nonrevertant background cells. The experiments were repeated at least twice.

Impact on exponential mutation rate.

It is possible that disrupting the mfd gene has an impact on the spontaneous mutation frequency during growth in B. subtilis. An increase in mutation during exponential growth could complicate attempts to calculate its impact on stationary-phase phenomena, possibly by increasing the appearance of viable but slow-growing prototrophic revertants during exponential growth that would appear after several days of incubation. To assess the impact of the loss of Mfd on the exponential mutation frequency, fluctuation testing was performed for each biosynthetic allele as described in Materials and Methods. The results of this are shown in Fig. 5 (exponential reversion rates). Loss of Mfd function actually lowers the exponential spontaneous mutation rate for each allele. Reversions to histidine and methionine prototrophy declined by 22% and 8%, respectively. The rate of reversion to leucine prototrophy dropped by 51% during exponential growth. Direct comparisons of these rates to the reversion rates in stationary phase cannot be done because of the difference in growth rates of the populations in question.

FIG. 5.

Exponential mutation rates. Exponential reversion rates of the three biosynthetic markers in YB955 (dark bars) or YB9801 (mfd deficient; light bars) were determined by fluctuation testing as described in Materials and Methods. Error bars represent 95% confidence intervals determined by the method of Rosche and Foster (28). The experiment was replicated a minimum of three times.

Reconstruction results.

Reconstruction tests provide a further demonstration that the majority of the revertants are not slow growing and are not inhibited in the presence of nonrevertant parental background. The results are listed in Table 2. These revertants showed variation in colony morphology and were from different parental cultures. Only six YB9801 Leu⁺ revertants were available for testing, but they were clearly the result of stationary-phase mutation. The Met⁺ revertants and the majority of the His+ revertants were robust, with colonies arising 48 h after plating, indicating that they were not the result of slow-growing exponential-phase mutations that appeared as colonies on day 6 of the initial assay. However, several of the YB9801 His⁺ revertants appeared between 48 and 72 h. These would presumably have mutated on or around day 4 in the original assay. It is possible that they contain another nonbeneficial mutation or that the parental background inhibits their growth. One YB9801 His⁺ revertant did not form colonies within the time frame of the reconstruction experiment and was most likely slow growing. Also, one YB955 His⁺ revertant could not be tested, since it repeatedly lysed in PAB. This was most likely the result of the induction of a defective prophage or possibly of a mutation resulting in the upregulation of lytic enzyme genes.

TABLE 2.

Reconstruction of revertant colony growth from mixed cultures

Strain and revertant test no.a	No. of revertant coloniesb
	Leu+		Met+		His+
	WB	BG	WB	BG	WB	BG
YB955
1	124	154	88	134	49	28
2	34	52	95	141	87	86
3	33	29	97	165	99	124
4	63	53	117	145	127	73
5	68	50	130	150	38	25
6	54	54	30	35	97	122
7	62	74	46	46	31	87
8	36	66	34	64	107	110
9	155	218	96	125	128	89
10	111	87	94	96	—c	—
YB9801 (mfd)
1	106	91	51	42	115	71
2	85	50	23	28	96	53
3	68	83	58	41	80	86
4	50d	56	40	30	56f	78f
5	16d	48	85	96	N/G	N/G
6	68d	65	76	38	60	64
7	NDe	ND	40	61	23	12
8	ND	ND	48	91	64f	32f
9	ND	ND	123	219	18	37
10	ND	ND	70	59	118	21f

^aLeu⁺, Met⁺, and His⁺ revertant colonies arising on days 6 and day 7 were serially diluted to obtain fewer than 100 colonies and mixed with approximately 3 × 10⁸ CFU of the parental strain from which they arose. After spread plating on minimal medium, the plates were incubated at 37°C. Numbers of revertants were determined by 48 h of incubation.

^bDilutions of revertants without background (WB) were compared to revertants mixed with their parental strain as background (BG).

^c—, revertant was repeatedly lysed in Penassay broth and was not spread plated.

^dDue to the low number of Leu⁺ reversions for YB9801, tested Leu⁺ revertants from days 8, 10, and 3 respectively.

^eND, not tested.

^fColonies arising by 72 h of incubation.

Tests for the presence of potential suppressors.

Isolated revertants were tested on SMM containing two required amino acids and missing one amino acid. For example, if a revertant was tested on a medium lacking leucine, then only histidine and methionine were present. Growth was scored at 48 h. This provided a screen for any potential slow-growing revertants, as mentioned above, and for identifying incidents of suppression. Results are given in Table 3 for revertants of YB955 and of YB9801 (mfd deficient). Some colonies arising under histidine selection (20.6% in YB955; 35.1% in YB9801) do not arise after 48 h of growth and may be slow-growing mutants or phenocopies. We have not observed these types of colonies under either methionine or leucine selection. Fifty-one percent of revertants to methionine prototrophy from YB955 recovered after day 6 are also capable of growth on histidine selective medium, whereas 100% of recovered YB9801 Met⁺ revertants also were no longer auxotrophic for histidine. Simultaneous reversion of both auxotrophic phenotypes suggests that these methionine revertants have acquired an ocher suppressor tRNA mutation. Conversely, the rate of recovery of histidine revertants capable of growing on methionine selective medium (indicative of amber tRNA suppression) appears similar between the two strains.

TABLE 3.

Growth of stationary-phase revertants on alternative selective media^a

Revertant	No. of revertants that grew/no. tested (% revertants that grew)
	Leu− medium	Met− medium	His− medium
YB955 (mfd+)
Leu+	33/33 (100)	0/33 (0)	0/33 (0)
Met+	0/92 (0)	92/92 (100)	47/92 (51.1)
His+	0/113 (0)	7/113 (6.2)	90/113 (79.6)
YB9801 (mfd deficient)
Leu+	1/1 (100)	0/1 (0)	0/1 (0)
Met+	0/65 (0)	65/65 (100)	65/65 (100)
His+	0/97 (0)	3/97 (3.1)	63/97 (64.9)

^aLeu⁺, Met⁺, and His⁺ revertant colonies from day 6 were tested on 1× SMM missing one required amino acid (His, Met, or Leu) to screen for suppressor mutations. Plates were scored after 48 h of incubation.

Sequencing.

Ten revertants per strain for each prototrophic phenotype were analyzed by sequencing the leuC, metB, and hisC alleles to determine the type of mutation that led to the reversion. The sequencing of these alleles permitted us to support, but not prove, the data from Table 3 that indicate the generation of suppressor mutations (Table 4). Results with the strain lacking a functional Mfd protein are similar to those with the control with the exception of reversions to methionine prototrophy. In the mfd-deficient strain, all Met⁺ reversions sequenced are due to amber suppression, as established by growth in the absence of histidine, whereas only half of the Met⁺ reversions are the result of suppression in the isogenic parental strain.

TABLE 4.

Base changes in revertants of mutant alleles

Strain	Revertant allelea	Position of mutation (bp)	No. of revertants sequenced	Type of mutation	DNA change	Result of reversiond
YB955	hisC	952	3	Transition	TAG→CAG	Stop→Gln
	hisC	952	1	Transversion	TAG→AAG	Stop→Lys
	hisC	No reversion	6
	metB	346	7	Transition	TAA→CAA	Stop→Gln
	metB	346	1	Transversion	TAA→AAA	Stop→Lys
	metB	No reversionb	2
	leuC	427	1	Transition	AGA→GGA	Arg→Gly
	leuC	No reversion	9
YB9801 (mfd deficient)	hisC	953	1	Transition	TAG→TGG	Stop→Trp
	hisC	No reversion	9
	metB	No reversion	10
	leuCc	No reversion	5

^aRevertants isolated from day 6 after plating.

^bColonies were also His⁺.

^cRevertants isolated from days 3 to 10 after plating.

^dCodon changes from the wild type present in YB955 are as follows: for hisC, CAG (Gln) to TAG (amber); for metB, GAA (Glu) to TAA (ochre); for leuC, GGA (Gly) to AGA (Arg).

DISCUSSION

Mfd is a component of stationary-phase mutagenesis in B. subtilis.

Our laboratory examines mechanisms that increase genetic diversity in growth-limiting conditions for a model gram-positive organism, B. subtilis. In this study, we show that inactivation of TRCF or the Mfd protein results in a decrease or abrogation of stationary-phase mutagenesis, as measured by the generation of reversions or suppression of alleles that render B. subtilis auxotrophic for the synthesis of three amino acids (Fig. 1). In addition, our reconstruction experiments suggest that the majority of the mutants that arise in the absence of amino acids 3 to 6 days after plating are indeed the result of mutagenesis that takes place under the growth-limiting conditions and not the product of an early mutation that confers slow growth (Table 2). Also, in this study we again demonstrate that the deficiency in stationary-phase mutagenesis is not due to differences in cell viability (Fig. 2) (41-43). Since the disruption of a gene might lead to detrimental effects on other genes in the vicinity of the loci that are inactivated, we cloned the mfd⁺ allele and showed that the presence of this gene restored, in trans, the stationary-phase mutation frequency seen in the isogenic parental strain (Fig. 3).

In contrast to our results, inactivation of the gene encoding TRCF in Escherichia coli showed no effects on the ability to generate reversions in a strain deficient for tyrosine biosynthesis (5). This suggests that either TRCF's role in DNA repair or increasing genetic diversity differs between gram-positive and gram-negative bacteria or that the effects are somewhat dependent on the mutagenesis assay used. In the F′-lac plasmid system used with E. coli, the generation of mutants under conditions of starvation for carbon requires a RecA protein (8), possibly to mediate the formation of duplications of a leaky β-galactosidase gene (18) or to promote error-prone double-strand-break repair in stressed cells (17). In the case of B. subtilis cells placed under conditions of amino acid starvation, a functional recA gene is not required for the generation of mutations that confer prototrophy (41). In addition to the requirement of Mfd (41), the transcription factors ComA and ComK, necessary for the development of genetic competence (14, 46), have been reported to influence this mechanism. Other components that do not directly affect transcription but affect stationary-phase mutagenesis in B. subtilis are a Y-family DNA polymerase and factors involved in DNA mismatch repair (25, 43).

Unlike results in previous B. subtilis studies in which the strain deficient in stationary-phase mutagenesis had no change in mutation rates during exponential growth (41, 43), the loss of Mfd also impacted the exponential leucine reversion frequency. This indicates a role for Mfd in exponential mutagenesis. Since Mfd is involved in the process of transcription, which spans all portions of the cell cycle, it is not surprising to find effects during exponential growth. Mutation rates affected by the loss of Mfd may be dependent on the physiological state of the cell and the transcription rate of the allele in question; more-transcribed genes would be more vulnerable to accumulating mutations and would require Mfd's activity for repair. In the absence of Mfd, cells that suffer a bulky lesion in an expressed gene, giving rise to a stalled RNA polymerase at that site, have an obvious transient difficulty to overcome before cell division can occur (44). This obstacle to replication could lower the measured mutation rate for that allele. One of the assumptions of the Lea-Coulson estimator is that mutants and the wild type have the same growth rate (28), and this may not be the case for the prerevertant cells, at least until the blockage to replication is corrected for the aforementioned reason.

The mfd disruption alone is responsible for the decrease in stationary-phase mutagenesis.

In creating the YB9801 constructs, we noted that the next downstream transcriptional unit from the mfd disruption is spoVT and that there is no obvious transcriptional terminator following mfd. This raises the possibility that a knockout of mfd may have polar effects on spoVT. spoVT transcripts produced by mfd read-through are seen during exponential growth, but no translation of spoVT-lacZ fusions was observed (2). Furthermore, spoVT has a promoter that utilizes σ^G-containing RNA polymerase, which becomes active late in stationary-phase growth (2). In fact, SpoVT is required for forespore-specific gene expression (2). We have used glucose as a carbon source; therefore, our cells are in a catabolite-repressed condition and should never reach this stage of sporulation. More significantly, the complementation of YB9801 (mfd) by pMPRY503 establishes that the loss of mfd alone is responsible for the decline in stationary-phase revertants. The subsequent loss of the phenotype, once again, by disrupting mfd on pMRYB503 (resulting in pMPRYB512) ensures that there is no plasmid-based system that is responsible for the observed complementation (Fig. 3).

Mfd mediates stationary-phase mutagenesis by a yet-to-be-identified mechanism.

Mfd is a well-characterized DNA repair factor that targets lesions in transcribed regions (32, 35-37) and is involved in genetic recombination (1). In addition to the role in DNA repair and recombination, Mfd has also been shown to influence carbon catabolite repression in B. subtilis (53). It has been proposed that Mfd mediates carbon catabolite repression by dissociating transcription elongation complexes stalled at cre sites located within coding sequences (53). While our results are counterintuitive to the DNA repair concept presented above, they also highlight the possibility that Mfd mediates mutagenic processes through aspects of transcription, DNA repair, and cell physiology.

In general, there are three salient features to Mfd-mediated stationary-phase mutagenesis. One, there is a sizable reduction in the number of revertants in the three alleles tested in the mfd-deficient strain (Fig. 1). Two, the mfd⁺-mediated stationary-phase mutagenesis process does not appear to target tRNA genes. Mfd appears to be required for the generation of true revertants to methionine prototrophy by mutation of the metB5 nonsense codon. This notion is supported by the increase in the proportion of tRNA suppression to the total number of accumulated revertants and the decrease in the overall reversion rate under mfd-deficient conditions. The proportion of revertants to methionine protorophy by tRNA suppression is 51% for mfd-replete strains and 100% for mfd-deficient strains (Table 3). Three, the changes in the nonsense and missense alleles that led to prototrophy could be attained by depurination, deamination, or oxidative damage in the template strand. Sequencing results with the mfd⁺ background tentatively suggest a bias towards the formation of T→C transitions over T→A transversions in the nonsense alleles (metB5 and hisC952). Interestingly, the only occurrence of true reversion to leucine prototrophy, an A→G transition, took place in the mfd-proficient strain (Table 4). These DNA changes have also been observed in previous studies (43). The events leading to these mutations are, in general, nondistortive and likely to be bypassed during transcription (34, 47). Taken as a whole, these three features support the possibility that the role of Mfd is to potentiate mutation at the alleles under direct selection and perhaps associate transcriptional derepression with the ability to accumulate mutations. Such association has been the subject of studies with E. coli, where it was demonstrated that increases in transcription levels of a defective leuB gene correlated positively with increases in the rate of reversion at the mutated site (26). It has been proposed that transcription leaves coding DNA unprotected, supercoiled, and prone to form secondary structures and base deaminations and that these features result in introduction of permanent DNA changes by subsequent replication (49). Our experiments were conducted under conditions that lead to derepression of transcription. Microarray analyses under conditions that lead to the stringent response and catabolite repression also suggest derepression of the test alleles used in this study (10, 21). It is conceivable that the differences in the number of mutations that accumulated in the test alleles are due to differences in the rates of transcription and the susceptibility to deamination and/or the formation of secondary structures.

Another interesting possibility is that Mfd may facilitate transcriptional bypass. This would imply that during transcription, in the presence of an existing DNA lesion, Mfd may allow the elongation complex to proceed at the expense of generating an altered mRNA (transcriptional bypass). An altered mRNA may then give rise to a temporary growth phenotype that would trigger DNA replication under conditions of deficient repair. Hastened replication, outpacing repair, would introduce a heritable change at the sites of the lesion. Mfd, as a facilitator of transcriptional bypass, has been discussed elsewhere (34) and relies on three observations: (i) in vivo RNAP efficiently bypasses different types of DNA lesions (9); (ii) Mfd is recruited to nondistortive DNA lesions (e.g., 8-oxoguanine) (4); and (iii) in vitro Mfd rescues paused elongation complexes to active transcription without dissociating RNAP (24). One would then expect that any factors involved in transcription elongation, such as Mfd, will influence the rate of the accumulation of mutations. Transcription-associated mutagenesis appears to extend to eukaryotic organisms, since this type of mechanism has been implicated in the generation of p53 mutations and cancer (27, 50).

The role of Mfd in mutagenesis and its association with transcription would be confirmed by detecting a decrease in levels of transcriptional bypass in an mfd-deficient background, and a decrease in the ability to generate mutations in stationary phase in cells that lack factors involved in rescue of stalled polymerases (but not involved in DNA repair). Additionally, the lack of Mfd would be expected to result in a decrease in the accumulation of mutations in the nonselected allele in a transcriptional fusion. Finally, whether the transient phenotype is required or the increased levels of transcription are sufficient to generate mutants in stationary phase could be discerned by conducting assays that allow transcription but prevent protein synthesis or function. All of the above scenarios are testable, and experiments are under way to determine the validity of the model that we have presented.