Induction of Rhodobacter capsulatus Gene Transfer Agent Gene Expression Is a Bistable Stochastic Process Repressed by an Extracellular Calcium-Binding RTX Protein Homologue

GTAs catalyze horizontal gene transfer (HGT), which is important for genomic evolution because the majority of genes found in bacterial genomes have undergone HGT at some point in their evolution. Therefore, it is important to determine how the production of GTAs is regulated to understand the factors that modulate the frequency of gene transfer and thereby specify the tempo of evolution. This work describes a new type of genetic regulation in which an extracellular calcium-binding protein homologue represses the induction of the Rhodobacter capsulatus GTA, RcGTA.

ABSTRACT

Bacteriophage-like gene transfer agents (GTAs) have been discovered in both of the prokaryotic branches of the three-domain phylogenetic tree of life. The production of a GTA (RcGTA) by the phototrophic alphaproteobacterium Rhodobacter capsulatus is regulated by quorum sensing and a phosphorelay homologous to systems in other species that control essential functions such as the initiation of chromosome replication and cell division. In wild-type strains, RcGTA is produced in <3% of cells in laboratory cultures. Mutants of R. capsulatus that exhibit greatly elevated production of RcGTA were created decades ago by chemical mutagenesis, but the nature and molecular consequences of the mutation were unknown. We show that the number of cells in a population that go on to express RcGTA genes is controlled by a stochastic process, in contrast to a genetic process. We used transposon mutagenesis along with a fluorescent protein reporter system and genome sequence data to identify a gene, rcc00280, that encodes an RTX family calcium-binding protein homologue. The Rc280 protein acts as an extracellular repressor of RcGTA gene expression by decreasing the percentage of cells that induce the production of RcGTA.

IMPORTANCE GTAs catalyze horizontal gene transfer (HGT), which is important for genomic evolution because the majority of genes found in bacterial genomes have undergone HGT at some point in their evolution. Therefore, it is important to determine how the production of GTAs is regulated to understand the factors that modulate the frequency of gene transfer and thereby specify the tempo of evolution. This work describes a new type of genetic regulation in which an extracellular calcium-binding protein homologue represses the induction of the Rhodobacter capsulatus GTA, RcGTA.

INTRODUCTION

The purple, nonsulfur, phototrophic, alphaproteobacterium Rhodobacter capsulatus is a model organism for studying an unusual horizontal gene transfer (HGT) process that is mediated by an extracellular particle known as a gene transfer agent (GTA) (in this case, RcGTA) (1, 2). R. capsulatus induces the production of RcGTA particles as cultures enter the stationary phase of growth. These small, bacteriophage-like particles were found to contain ∼4-kb-long random genome DNA fragments from the producing cell (3, 4), presumably packaged by a head-full mechanism. The release of these particles is through lysis of the host cell (4–6). An extracellular polysaccharide facilitates RcGTA adsorption to cells, resulting in injection of its DNA content into the periplasm of the recipient cell, after which the DNA is transferred to the cytoplasm via a natural competence-like pathway, allowing for RecA-dependent allelic exchange (7).

Most of the genes encoding the RcGTA particle are located in an ∼15-kb cluster (rcc01682 to rcc01698) (8), and some share an ancestor with tailed bacteriophage genes. Genes encoding head spikes, an endolysin and holin responsible for cell lysis, and a putative tail fiber protein are located ∼670 kb, ∼1,200 kb, and ∼1,600 kb, respectively, from the RcGTA main gene cluster (4, 5, 9, 10). The expression of the RcGTA gene cluster is controlled by cellular proteins, including the quorum-sensing proteins GtaI and GtaR (11, 12) and the proteins CckA, ChpT, and CtrA, which appear to be part of a phosphorelay pathway (8, 13, 14). The response regulator CtrA controls expression of the gafA gene (10) by binding to its promoter region to induce transcription, and the GafA protein induces transcription of the RcGTA structural gene cluster (15).

The amount of RcGTA produced by a wild-type (WT) strain such as SB1003 is very small, which was a major obstacle in early research on the molecular properties of RcGTA, but Yen et al. (3) isolated an overproducer mutant, which greatly facilitated characterization of RcGTA and improved its utility in strain construction for genetic analyses of a variety of R. capsulatus biological properties. However, the nature of the overproducer mutation was unknown. Here, we describe the gene mutated in the overproducer strain, as well as some properties of the protein product.

The expression of the RcGTA gene cluster in an isogenic population is heterogeneous; i.e., in the WT strain SB1003, <3% of cells express the RcGTA genes in an induced, stationary-phase culture. In overproducer strains derived from SB1003 by chemical mutagenesis, a larger percentage (>30%) of cells express RcGTA genes (4, 5). Several processes, including genetic mutation, phase variation, and “noise” in gene expression (stochastic gene expression), are known to cause heterogeneous gene expression within a clonal population. These processes can be differentiated by their switching frequencies.

Spontaneous genetic mutations in the Escherichia coli genome occur at about 2 × 10⁻¹⁰ mutations per nucleotide per generation (16), or approximately 10⁻⁷ mutations per 1-kb gene per generation. It follows that the switching frequency (i.e., reversion of a particular mutation) would be lower than the forward frequency, because a relatively wide variety of mutations throughout a gene could result in loss of function, compared to the number of mutations that would specifically restore mutant gene function versus having a neutral or further deleterious effect (17). Phase variation refers to reversible genetic changes that often lead to surface structure changes in pathogenic bacteria (18). Phase variation is reversible with a frequency that is higher than that of mutation and variable, depending on mechanism or species, but generally is found to be around 10⁻³ per cell per generation. For example, a study showed that the reverse frequency of DNA conversion in Neisseria gonorrhoeae was about 4 × 10⁻³ per cell per generation (19). The E. coli P-fimbrial phase variation off-to-on switching frequency was found to be 4.4 to 6.1 × 10⁻³ per cell per generation and the on-to-off switching frequency about 10-fold higher, at 3.3 × 10⁻² per cell per generation (20). Although this number could be as high as 1 in 10 cells per generation (18), it is a heritable trait that would be expected to result in enrichment in the frequency of the trait upon sorting and subculturing of individual cell types.

Noise in gene expression, also known as stochastic gene expression, comes from differences in the number of regulatory molecules per cell within an isogenic population. Through amplification and stabilization by feedback genetic circuits, this noise can lead to the bifurcation of an isogenic population into a “bistable” state of two distinct phenotypes (21, 22). In contrast to genetic changes, sorting and subculturing would not be expected to result in enrichment in the frequency of individual cell types resulting from a stochastic process, as has been indicated in studies of E. coli persister cells and natural competence and sporulation of Bacillus subtilis (23).

In this study, we used a fluorescent protein as a reporter in fluorescence-activated cell sorting (FACS) and time-lapse fluorescence microscopy to track the single-cell switching frequencies of “on” cells induced for expression of RcGTA genes and “off” cells repressed for expression of RcGTA genes. The results indicated that a stochastic process governs the frequency of RcGTA-expressing cells in a population, because there was no change in the frequency of on cells upon sorting and subculturing of either on or off cells. Time-lapse imaging of individual on and off cells as they divided and formed microcolonies confirmed the interpretation that switching between the on and off cell phenotypes is a stochastic process.

Furthermore, by transposon mutagenesis of a strain in which the RcGTA structural gene promoter drives transcription of a fluorescent protein reporter gene, we identified a gene that represses the frequency of RcGTA-expressing cells in a population. A point mutation in this gene was found to be responsible for the phenotype of the original RcGTA overproducer strain, which was first described in 1979 (3). This mutation is located in an open reading frame (ORF) (rcc00280) annotated as encoding a calcium-binding, extracellular, RTX family protein. We suggest that this protein, designated Rc280, acts as an extracellular signal to repress RcGTA gene expression.

RESULTS

Single-cell expression of RcGTA genes monitored by a fluorescent reporter.

To assess whether RcGTA-expressing cells arise from a genetic change such as mutation or phase variation or because of stochastic gene expression, we initially employed a single-copy, chromosome-integrated, fluorescent protein (mCherry) reporter gene in both the WT SB1003 strain and the overproducer DE442 strain (Fig. 1). The mCherry coding sequence uses a copy of the start codon and ribosome-binding site of the rcc01682 gene (ORF g1 in the RcGTA structural gene cluster) and the RcGTA promoter, as described previously (5). The promoter is absent from the sequence separating the stop codon of the mCherry gene and the start codon of the g1 gene (6), there is no intervening transcription terminator, and transcription of RcGTA genes thus occurs as readthrough from the mCherry gene.

FIG 1.

Location of the chromosomally integrated mCherry reporter gene used in this study, showing flanking regions. The bent arrow represents the RcGTA promoter, and the magenta rectangle represents the duplicated ribosome-binding site region of rcc01682 (ORF g1). The DNA sequence starts with the ATG start codon of ORF g1, which also is the start codon of the mCherry gene.

The resultant strains, i.e., SBpG derived from the WT SB1003 strain and DEpG derived from the overproducer mutant DE442, had similar frequencies of fluorescent cells, compared with the previously constructed plasmid-borne gene fusion strains (5), when viewed by fluorescence microscopy (Fig. 2). When the flow cytometry profiles of strains SBpG and DEpG were compared with that of the nonfluorescent parental strain SB1003, there was a shift of a small percentage of SBpG cells to greater fluorescence and a shift of a much larger percentage of DEpG cells, which resulted in a well-resolved peak (Fig. 2) using either a biexponential scale or a logarithmic scale (24). In addition to the increase in the number of fluorescent cells, the maximum fluorescence increased, indicating that greater levels of expression were reached by some individuals in the overproducer population. When the numbers of functional RcGTA particles were measured as gene transfer frequencies, strains SBpG and DEpG had the same levels as the respective parental strains SB1003 and DE442 (Fig. 3). These results show that this mCherry gene acts as a reliable indicator of RcGTA gene expression.

FIG 2.

Fluorescence of strains containing the chromosomally integrated mCherry gene transcribed from the RcGTA promoter in strain SBpG, derived from the WT strain SB1003, and in strain DEpG, derived from the overproducer strain DE442. (A and B) Fluorescence microscopy of strain SBpG (A) and strain DEpG (B). The cell images were captured using two different microscopes. (C and D) Flow cytometry histograms comparing fluorescent strains SBpG (red) and DEpG (blue) to the nonfluorescent WT strain (orange), using a a biexponential scale (C) or a logarithmic scale (D). Counts indicate the number of cell equivalents, and fluorescence is in arbitrary units (A.U.).

FIG 3.

Relative gene transfer activities of strains containing the chromosomally integrated mCherry gene transcribed from the RcGTA promoter (SBpG, derived from the WT strain SB1003, and DEpG, derived from the overproducer strain DE442). Error bars represent the standard error of the mean (n = 3).

However, because the RcGTA-expressing on cells release RcGTA particles and therefore die by lysis (4–6), such cells of these strains would not be capable of giving rise to progeny for subculture. Therefore, we deleted the RcGTA structural gene cluster (rcc01682 to rcc01698) and disrupted the endolysin gene rcc00555, so that the on cells of the resultant strains SBpGΔGTA555::KIXX (called SB/G⁻L⁻ here) and DEpGΔGTA555::KIXX (called DE/G⁻L⁻ here) would neither produce RcGTA particles nor lyse. This allowed us to follow the fates of cells in terms of the on and off phenotypes.

FACS of RcGTA on and off cells in liquid culture and flow cytometry monitoring of subcultured cells.

We used flow cytometry to monitor the percentages of on and off cells in populations, each of which was grown from a single on or off cell obtained by FACS. The rationale was that cells that differentiated as a result of a genetic process would have an increased proportion of their progeny with the mutant phenotype, whereas cells from a bistable population as a result of processes associated with the high concentrations of cells in the stationary phase would display the parental phenotype upon return to low cell concentrations that do not induce RcGTA production. Because the percentage of on cells in SB/G⁻L⁻ cultures was very low, we used the overproducer strain DE/G⁻L⁻ to monitor the phenotypic fate of single cells in liquid cultures that were grown to the stationary phase, using FACS to sort fluorescent on cells from nonfluorescent off cells.

An early stationary-phase culture of the DE/G⁻L⁻ strain was sorted into 96-well plates based on mCherry fluorescence, with each well containing an on or off cell in growth medium. The percentage of viable cells recovered was about 40% to 45%, and there was no difference (P > 0.9999, Mann-Whitney test) in the frequency of recovery of viability between sorted on and off cells, indicating that the loss of viability was due to the FACS process and applied equally to the two cell types (Fig. 4A). The percentages of on and off cells arising in subcultures from the cell in each well were determined using flow cytometry. As shown in Fig. 4B, on and off cells from 12 randomly selected wells grew into populations with approximately the same percentages of on and off cells after growth to stationary phase, as shown by cytometry histograms, which were almost identical in all cases. These results indicate that the expression of the RcGTA promoter is induced within a subpopulation of cells at the same frequency regardless of the expression state (on or off) of the progenitor cell, because the final populations that accumulated in 12 independent subcultures of on and off cells had essentially the same percentages of on and off cells.

FIG 4.

Recovery of viable cells and fluorescence profiles of sorted and subcultured strains in FACS analyses. (A) Percentage recovery of FACS-sorted single on and off cells. Each experiment sorted three 96-well plates of on cells and off cells. Bars indicate the standard deviation. (B) Flow cytometry histograms showing gene expression profiles of regrown populations from sorted cells overlaid on the parental culture histogram. Twelve on and 12 off sorted cells were grown to the stationary phase, and the resultant cells were counted on the basis of mCherry fluorescence intensity. Data are for 10⁴ counts each, ungated, displayed in a biexponential format, and fluorescence is in arbitrary units (A.U.).

RcGTA gene expression, monitored by fluorescence, of single on and off cells growing on solid medium.

The DE/G⁻L⁻ strain (with the mCherry gene transcribed from the RcGTA structural gene cluster promoter but lacking the structural genes and the endoglucanase lysis gene) was grown to the stationary phase, and cultures were diluted and used to inoculate solid culture medium. Individual off and on cells were viewed with a fluorescence microscope to follow cell division and fluorescence over time, to track the presence or absence of RcGTA gene expression.

Starting with off cells, fluorescence was undetectable until after 22 h of incubation, and microcolonies reached about 15 cell lengths (∼150 μm) in diameter (Fig. 5A). This result is consistent with previous studies of cell populations in liquid culture, where RcGTA gene expression was induced as cultures reached high cell concentrations upon entry into the stationary phase (8, 12, 25), showing that cells induce expression similarly under the two conditions.

FIG 5.

Bistable RcGTA gene expression is not genetically heritable. For time-lapse fluorescence microscopy analyses of strain DE/G⁻L⁻, stationary-phase cultures were diluted and spotted onto YPS medium solidified with agarose. Numbers indicate times, in hours. Both on and off cells showed repressed RcGTA gene expression at low cell concentrations (0 to 18 h); as cell concentrations passed a threshold, RcGTA gene expression was induced in a subpopulation of cells (24 to 32 h). (A) Starting with off cells. (B) Starting with on cells.

In the case of initially fluorescent on cells (Fig. 5B), fluorescence dimmed as cells divided and became undetectable after 4 to 6 h. Evidently, previously induced RcGTA gene expression ceases after cells are transferred from the stationary phase of a liquid culture to low population density on solid medium. As microcolonies grew on the solid medium, fluorescence appeared in a subpopulation of cells after 24 h.

These results show that RcGTA expression in on cells halted rapidly upon transfer from inducing conditions to noninducing conditions and the induction of RcGTA expression resumed in a subset of the population as cell concentrations increased, as in liquid cultures entering the stationary phase of growth (8, 12, 25). Therefore, the fluorescence microscopy data support a stochastic process, as opposed to a genetic process, for regulation of the frequency of RcGTA-producing cells, congruent with the interpretation of the flow cytometry experiments.

Disruption of rcc00280 increases the percentage of RcGTA-expressing cells within a clonal population.

Previous studies revealed that the RcGTA overproducer strain DE442 has a higher frequency of cells expressing the RcGTA structural genes than its kin WT strain SB1003 (4, 5), and we wished to elucidate the genetic basis of this difference. Unfortunately, the genesis of these strains is not well documented. It is thought that DE442 was obtained by mutating the crtD (carotenoid biosynthesis) gene of Y262, an RcGTA overproducer strain derived from strain BB103 (3), which is a streptomycin-resistant mutant of the environmental isolate strain B10, using 1-methyl-3-nitro-1-nitrosoguanidine (MNNG) mutagenesis (26). Strain SB1003 was constructed by RcGTA-mediated transfer of a rifampin resistance gene from strain BB101 to strain B100, a “cured of phage” derivative of strain B10 (27, 28).

The kin relationship among SB1003, Y262, and DE442 was confirmed by genome sequencing of these strains (29); it was not possible to identify the genetic basis of RcGTA overproduction, however, because there are more than 700 single-nucleotide polymorphisms (SNPs) between the DE442 and SB1003 genomes. Therefore, we opted to carry out transposon mutagenesis of the SB1003 derivative strain SBpG, in which the RcGTA structural gene promoter drives expression of a gene encoding the fluorescent protein mCherry, to identify an overproducer on the basis of increased fluorescence.

To enhance the robustness of the transposon mutagenesis screen, we constructed the TnFruOut transposon, in which a fructose-inducible promoter (pFru) (30) was placed within the hyperactive transposon Tn5 (31), facing outward from the transposon genes. Transposition of TnFruOut into a gene would lead to disruption of the gene; meanwhile, 3′-adjacent genes could be induced by the addition of fructose if the transposon was inserted with the orientation of pFru being the same as the adjacent genes. Therefore, TnFruOut was used in a screen for both positive and negative regulators.

A total of 4,608 SBpG transposon mutants were screened. One of the mutants exhibiting increased fluorescence in liquid culture, strain SBT2-C22, was found to have an increase in the percentage of fluorescent cells by flow cytometry, compared to SBpG. Culture supernatants of SBT2-C22 yielded ∼10⁴-fold more gene transfer recipients than the WT strain SBpG in an RcGTA gene transduction bioassay, similar to the overproducer strains DE442 and DEpG, which contain the chemically induced mutation. The insertion site of TnFruOut in strain SBT2-C22 was found to be between nucleotides 934 and 935 of the coding sequence of the 1,056-bp-long ORF rcc00280 (data not shown).

To confirm that the transposon insertion within rcc00280 was sufficient to yield an overproducer phenotype, we first used RcGTA to transduce the defective transposon-borne kanamycin resistance (Km^r) marker from strain SBT2-C22 into SBpG. The resultant Km^r transductant SBpG280::FruOut was found by PCR to carry the transposon insertion within rcc00280 (data not shown) and to have the overproducer phenotype (Fig. 6). When the plasmid pCM280, which carries rcc00280 with ∼700 bp of flanking sequences on each end, was introduced into SBpG280::FruOut, the percentage of on cells in flow cytometry profiles decreased greatly, as did the number of transductants in a gene transfer assay (Fig. 6).

FIG 6.

Mutations within rcc00280 result in increased RcGTA production by increasing the numbers of cells within the R. capsulatus population expressing the RcGTA genes and increasing the level of expression within expressing cells. Promoter activity is indicated by cell cytometry histograms (biexponential format). RcGTA production gives the relative numbers of rifampin-resistant colonies obtained, and error bars indicate the standard deviation (n = 3).

Furthermore, Western blots probed with antisera raised against the RcGTA capsid protein showed an increase in this RcGTA protein in rcc00280 mutants and a decrease upon trans complementation of these mutants (Fig. 7). Although the different mutant alleles of rc00280 did not yield exactly the same phenotypes regarding gene transfer frequencies and Western blot intensities, there were consistently >10-fold differences between strains carrying WT and mutant alleles and between mutant and trans-complemented strains (Fig. 6 and 7).

FIG 7.

Western blots of RcGTA capsid protein present within cells and in the culture medium of R. capsulatus strains. Loadings were normalized to the total number of cells, and samples of expected nonoverproducers were concentrated 50-fold. (A) Blot of cells. (B) Blot of extracellular culture medium. Vertical lines were added to facilitate differentiation between lanes.

RNA and bioinformatic analyses of rcc00280.

Because the genome annotation of rcc00280 was based solely on in silico predictions, we used strand-specific primers in reverse transcription (RT)-PCR amplifications to determine whether rcc00280 was transcribed as predicted. It was found that only the RT primer designed to detect RNA in the same transcriptional orientation as rcc00280 yielded a PCR-amplified product of cDNA, visualized as a band in gel electrophoresis (Fig. 8A). Differential RNA sequencing (dRNA-seq) (32) data were used to locate the transcription start site of rcc00280 and, in addition to a site located 5′ of the start codon, the rcc00280 gene appeared to have transcription start sites within the coding region (Fig. 8B). The positions of the three most abundant RNA 5′ ends were all preceded by −10/−35 sequences similar to an R. capsulatus consensus promoter (see Fig. S1 in the supplemental material). Therefore, only one strand of rcc00280 is transcribed, and a promoter is located 5′ of the predicted start codon.

FIG 8.

Analysis of rcc00280 transcripts and mapping of transcription start sites. (A) Gel electrophoresis of RT-PCR products, showing that the RNA has the same polarity as the gene. Lane gDNA, PCR product with genomic DNA as the template; lane RNA (no RT), negative control lacking reverse transcriptase; lane RNA (F), forward primer was used for the RT step; lane RNA (R), reverse primer was used for the RT step. (B) Diagram of transcription start sites (see File S2 in the supplemental material for sequences of reads). The region targeted for RT-PCR is indicated by the red line, with forward (F) and reverse (R) primer sites indicated. Reads from dRNA-seq mapping to rcc00280 are shown as black lines below the ORF (blue), with arrows showing 5′ ends of transcripts and locations of homologues of consensus −10/−35 sequences (see Fig. S1).

The rcc00280 gene is predicted to encode a 351-amino-acid, hemolysin-type, Ca²⁺-binding repeat family protein, an extracellular type 1 secretion system (T1SS) protein. The putative Rc280 protein is predicted to be extracellular (PSortB v.3.0.2) (33), consistent with the fact that T1SS target proteins are directly secreted from the cell in a single step. The secretion signal of T1SS-secreted proteins typically is located within 60 residues of the C terminus, with no sequence conservation (34).

The predicted Rc280 protein contains six Ca²⁺-binding repeat domains, with 9-residue loops (GGXGXDXUX, with X indicating any amino acid and U indicating a hydrophobic amino acid), located from amino acid position 121 to position 220 (https://prosite.expasy.org). Such glycine- and aspartic acid-rich nonapeptide motifs found in T1SS RTX (repeat in toxin) proteins form a parallel β-roll structure upon Ca²⁺ binding after secretion, and the binding of Ca²⁺ appears to promote folding into a functional protein after secretion. Apart from the conserved Ca²⁺-binding regions, such proteins have widely varying activities, such as protease, lipase, toxin, or transporter (34). Software such as SWISS-MODEL predicts an Rc280 monomer of largely β-roll structure except for the C-terminal region. The C-terminal 100 residues are predicted to begin at the end of the β-roll, yielding a structure of mixed β-strand and loops, with homologues of unknown function (Fig. S2).

A BLASTp search of the NCBI nonredundant protein sequence database yielded over 100 homologues with E values of e⁻⁴⁵ or lower from alphaproteobacteria, mainly in the Rhodobacteraceae family; therefore, Rc280-like proteins are widespread in this family of bacteria. BLASTp searches of the UniProtKB (Swiss) database yielded hits to Rc280 as low as 9.71 × e⁻⁰⁷, but the alignments are within conserved G/D repeats and not within domains that endow these proteins with their disparate functions. Databases such as Model Organism and RefSeq yielded similar results, with many very low E value hits to uncharacterized genes. Therefore, Rc280 homologues are widespread but have not been studied.

To decrease the number of homologues that share only the conserved Ca²⁺-binding regions, we used the Rc280 C-terminal 100 amino acids in BLASTp queries of the Swiss-Prot database, which yielded three poor hits (E values of 7 to 9). The PDB database yielded a weakly similar (E value of 0.09) C-terminal helicase domain of the human DNA-binding protein INO80 (PDBETX) (35). The Web tool MotifFinder gave the same hits as the Swiss-Prot database but with E values of 0.09 to 0.5. We conclude that Rc280, like most RTX homologues encoded in genomes, has an unknown biochemical activity that in this case functions in repression of RcGTA.

These bioinformatic analyses raised questions regarding the rcc00280 gene, such as the following. Does the overproducer mutant DE442 contain a mutant allele? Is the rcc00280 transcript translated to yield a protein? If it is, is the protein found in the extracellular milieu? These questions are addressed in the following sections.

A single point mutation within rcc00280 results in the overproducer phenotype of strains such as DE442.

Because the SBpG280::FruOut strain has a phenotype almost identical to that of the RcGTA overproducer strain DE442, it was of interest to determine whether the overproducer strain DE442 carried an rcc00280 mutant allele. We found that there were two SNPs in the DE442 rcc00280 coding region (29), compared to the WT SB1003 sequence (36), leading to amino acid changes from glycine (GGC codon) to aspartic acid (GAC codon) at residue 271 (G271D) and from lysine (AAG codon) to glutamic acid (GAG codon) at position 338 (K338E). However, sequencing of the rcc00280 gene of our WT SB1003 strain revealed that it also carried the K338E change; therefore, the overproducer strain DE442 contained only one change within the rcc00280 gene, compared to the WT SB1003 strain in our strain collection. When we examined the rcc00280 homologues in other R. capsulatus WT strains, we found that environmental isolates B6, YW1, and YW2 all had the same GAG codon encoding glutamic acid at residue 338 as our SB1003 and DE442 strains (29), indicating that glutamic acid is the WT consensus for amino acid residue 338. Furthermore, the three genome-sequenced overproducer strains, DE442, Y262, and R121, all had the same GAC codon for aspartic acid at residue 271D (29).

Although it appeared likely that the G271D mutation in DE442 was solely responsible for the overproducer phenotype, this strain contained hundreds of SNPs, compared to the WT strain SB1003. Therefore, a complete deletion of the rcc00280 coding region in the SBpG background was created, yielding strain SBpGΔ280, which had the overproducer phenotype (Fig. 6). The introduction of plasmid pCM280 into strains DEpG and SBpGΔ280 decreased the frequencies of fluorescent cells and gene transfer in both strains. When the plasmid pCM280G271D (differing from plasmid pCM280 by the G271D mutation) was introduced into SBpGΔ280, there was no complementation of the overproducer phenotype (Fig. 6). Analogous results were obtained in Western blots, where the intensity of the RcGTA capsid band increased when cells contained rcc00280 mutant alleles and decreased when the WT allele was present (Fig. 7). These results show that the G271D point mutation alone in rcc00280 is sufficient for the overproducer phenotype of DE442 and related strains, because this mutation resulted in increases in total numbers and maximal brightness of fluorescent cells expressing the mCherry reporter gene and parallel increases in RcGTA production.

The strain SBpG280::FruOut has the transposon insertion in rcc00280 after nucleotide 933, leaving the N-terminal 311 codons intact, and DE442 has a point mutation in the same vicinity; therefore, it might appear possible that the overproducer phenotype results from disruption of a C-terminal T1SS sequence and the intracellular accumulation of a mutant protein. However, because the rcc00280 deletion mutant SBpGΔ280 retains only the first 3 and last 4 codons but has the same phenotype as SBpG280::FruOut (Fig. 6 and 7), it appears that it is the loss of rcc00280-encoded activity, and not the accumulation of a mutant protein, that leads to the overproduction of RcGTA.

A 6×His-tagged Rc280 protein is present in the extracellular culture medium.

We created the plasmid pCM280C6His, which encodes a C-terminally 6×His-tagged Rc280 protein, and we found that this plasmid complements rcc00280 mutants (Fig. S3). This 6×His tag was used to concentrate the protein and detect it in SDS-PAGE and Western blots.

Culture supernatants of strain SBpGΔ280 containing plasmid pCM280 and strain SBpGΔ280 containing plasmid pCM280C6His were passed through a Ni²⁺-nitrilotriacetic acid (NTA) column, and bound proteins were eluted with 250 mM imidazole. The major SDS-PAGE band in the SBpGΔ280(pCM280C6His) sample migrated with an apparent molecular weight of ∼50 kDa, and this band was not present in the SBpGΔ280(pCM280) sample (Fig. 9A). The major band detected in a Western blot of the SBpGΔ280(pCM280C6His) sample probed with an anti-6×His antibody also migrated with an apparent molecular weight of ∼50 kDa, and this band was absent from the SBpGΔ280(pCM280) sample (Fig. 9A). The intracellular fraction yielded a large number of bands attributed to nonspecific binding of the antiserum, which appeared to be the same for both strains.

FIG 9.

Analyses of Ni²⁺-NTA-affinity-chromatography-purified His-tagged Rc280 from cell-free culture supernatants. (A) SDS-PAGE of strain SBpGΔ280 containing plasmid pCM280 (WT) and strain SBpGΔ280 containing plasmid pCM280C6His (280) proteins eluted from Ni²⁺-NTA columns (left) and Western blot of gel probed with anti-6×His serum (right). MW, protein ladder, with masses in kilodaltons. (B) Peptides detected by matrix-assisted laser desorption ionization–time of flight mass spectrometry after digestion of His-tagged Rc280 with the endoproteinase GluC. Lines beneath the sequence represent the residues and amounts of the peptides detected.

The predicted 6×His-tagged Rc280 protein has a molecular weight of ∼36 kDa but is extremely negatively charged (calculated pI of 2.9) and could migrate as a higher-molecular-weight band in SDS-PAGE, as seen with other proteins (37). Therefore, the presence of the ∼50-kDa band in both the Coomassie-stained gel and the Western blot of the 6×His-tagged protein from the SBpGΔ280(pCM280C6His) culture supernatant that bound to the Ni²⁺-NTA column and the absence of such a band from the SBpGΔ280(pCM280) culture supernatant may be explained by anomalously slow migration of the 6×His-tagged Rc280 protein in SDS-PAGE.

The protein sample that gave rise to the ∼50-kDa band in SDS-PAGE and Western blots was digested with the endoproteinase GluC, and the resultant peptides were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) to identify the protein. The peptides detected showed that the major protein in this sample was Rc280 (Fig. 9B). These results confirm that Rc280 is secreted from the cell and exhibits anomalously slow migration in SDS-PAGE.

DISCUSSION

Using a fluorescent reporter of RcGTA gene expression, we found that a stochastic process and not a genetic process appears to govern the percentage of cells that express RcGTA genes. Stochastic gene expression of a variety of genes has been observed in several species. In an E. coli K-12 strain, persister cells that survived antibiotic treatment were present at the frequency of 10⁻⁶ to 10⁻⁵ (38, 39). In B. subtilis genetic transformation, about 10% of cells were competent in the stationary phase (40, 41). In contrast to genetic changes, sorting and subculturing would not be expected to result in enrichment in the frequency of individual cell types resulting from a stochastic process, as was found in studies of E. coli persister cells and natural competence and sporulation of B. subtilis (23). Our suggestion that the induction of fluorescence in a small percentage of R. capsulatus cells is congruent with a stochastic process, as opposed to a genetic process, is in part because changes in RcGTA gene expression monitored by changes in the percentage of fluorescent cells were not inherited by progeny of on and off cells separated by FACS (Fig. 4B). Furthermore, individual on cells monitored by fluorescence microscopy reverted to off cells within one or two generations and then gave rise to on cells as cell concentrations increased in microcolonies (Fig. 5B).

The genetic basis of RcGTA overproducers has been a puzzle in the 40 years since the initial mutant strain Y262 was generated by chemical mutagenesis (3), followed by strains DE442 and R121, which contain an additional mutation resulting in a change in carotenoid pigmentation (26). It was shown previously that <3% of the cells in R. capsulatus WT strain SB1003 stationary-phase cultures expressed the RcGTA gene cluster, whereas >30% of cells of the overproducer strain DE442 expressed the RcGTA gene cluster (4, 5). By use of transposon mutagenesis, we identified a mutant in which the transposon had been inserted into rcc00280, exhibiting an overproducer phenotype, as confirmed by subsequent deletion of the rcc00280 ORF (Fig. 6 and 7). We found that Y262 and the related overproducer strains DE442 and R121 (26) all contained the same rcc00280 allele, in which the GGC codon 271 for glycine was changed to an aspartic acid codon (GAC), consistent with the most frequent type of base change resulting from the MNNG mutagenesis that was used to obtain Y262 (42). This allelic identity also is strong evidence that the poorly documented strains DE442 and R121 are both derived directly from the original overproducer mutant Y262.

Bioinformatic analyses of the predicted protein product of the rcc00280 gene indicated that it is an extracellular RTX-like protein that contains multiple Ca²⁺-binding motifs and is secreted by a T1SS process. Using an rcc00280 allele that encodes a 6×His-tagged Rc280 protein, SDS-PAGE and Western blots probed with anti-6×His serum (Fig. 9A) indicated that cell-free culture supernatants contained this Rc280 protein. Peptides identified in LC-MS/MS analysis after proteolytic digestion confirmed that the Rc280 protein was present (Fig. 9B). Surprisingly, the 6×His-tagged Rc280 protein appeared to migrate as an ∼50-kDa band in SDS-PAGE, although the 6×His-tagged Rc280 protein has a molecular weight of ∼36 kDa. However, the Rc280 protein would be extremely negatively charged (calculated pI of 2.9), which could result in migration as a higher-molecular-weight band in SDS-PAGE because of low binding of SDS (37).

The rcc00280 gene appears to be transcribed from a promoter located 5′ of the annotated start codon, although there were RNA 5′-triphosphate ends that mapped within the coding region, to the codons for Tyr108 and Asp237 (Fig. 8). All three RNA 5′ ends are preceded by plausible R. capsulatus −10 and −35 sequences (43) (see Fig. S1 in the supplemental material). Genome-wide transcriptome studies have revealed large numbers of promoters located inside genes (44); therefore, our results are not unprecedented. Regardless of the possibility of 5′-truncated rc00280 transcripts, mass spectrometric analysis of proteolytic fragments of the extracellular 6×His-tagged Rc280 protein showed that peptides representing four N-terminal regions were present (Fig. 9B). Whether rcc00280-encoded proteins of less than full length are produced remains to be determined, but the presence of a predominant high-molecular-weight band in the Western blots argues for a single protein as the major gene product (Fig. 9A).

This research solves the decades-old mystery of the mutation responsible for the overproducer phenotype of the R. capsulatus mutant Y262, which was isolated in 1979, and strains DE442 and R121 derived therefrom, because all of these strains share the same GGC to GAC mutation in rcc00280. Although the mechanism by which the Rc280 protein represses stochastic induction of RcGTA is unknown, Hynes et al. (10) showed that the RcGTA structural gene transcription activator encoded by the rcc01865 gene, recently named gafA (15), is upregulated 18.7-fold in the overproducer strain DE442. Therefore, it appears that the Rc280 protein is an extracellular signal at the top of a pathway that leads to the repression of transcription of the gafA gene. The GafA positive regulatory protein appears to activate transcription of the gafA gene (15) and, because such feedback loops have been proposed to underlie stochastic bistable processes (21, 22), we speculate that Rc280 dampens a gafA feedback loop, thereby inhibiting stochastic differentiation of off cells into on cells.

The data in this report describe a new process, namely, extracellular protein regulation of bistable differentiation of R. capsulatus cells into two populations, RcGTA producers and competent recipients of RcGTA-borne genes (45). A challenging task remaining is to identify the immediate target of Rc280 and the molecular pathway leading from this extracellular protein to repression of gafA transcription. We look forward to elucidation of the pathway linking the Rc280 protein to stochastic repression of RcGTA gene expression.

MATERIALS AND METHODS

Bacterial strains, growth conditions, and plasmids.

The bacteria and plasmids used in this study are listed in Table 1. R. capsulatus strains were grown in defined RCV (46) or complex YPS (47) medium at 30°C, either aerobically in the dark or anaerobically with incandescent lamp illumination. When required, media were supplemented with rifampin (80 μg/ml), gentamicin (3 μg/ml), kanamycin (20 μg/ml), or tetracycline (0.5 μg/ml). E. coli strains were grown at 37°C in LB medium, supplemented with ampicillin (100 μg/ml), gentamicin (10 μg/ml), kanamycin (50 μg/ml), or tetracycline (10 μg/ml) when required.

TABLE 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant genotype or description	Source or reference
E. coli strains
DH5α λpir	λpir lysogen of DH5α, allowing replication of plasmid with R6K-based origin of replication	56
S17-1 λpir	Broad-host-range plasmid mobilizer strain containing RP4 tra genes	57
R. capsulatus strains
B10	WT; sensitive to rifampicin	2
SB1003	Derivative of B10; resistant to rifampicin	36
DE442	RcGTA overproducer derived by chemical mutagenesis of crtD	B. Marrs, personal communication
R121	RcGTA overproducer derived from B10 with crtD	26
SBpG	Derived from SB1003; contains RcGTA promoter driving transcription of mCherry gene integrated into chromosome	This study
DEpG	Derived from DE442; contains RcGTA promoter driving transcription of mCherry gene integrated into chromosome	This study
SB/G−L−	Derived from SBpG; contains deletion of RcGTA structural gene cluster and disruption of rcc00555, same as SBpGΔGTA555::KIXX	This study
DE/G−L−	Derived from DEpG; contains deletion of RcGTA structural gene cluster and disruption of rcc00555, same as DEpGΔGTA555::KIXX	This study
SBT2-C22	SBpG with transposon TnFruOut inserted in rcc00280	This study
SBpG280::FruOUT	Derived from SBpG by RcGTA-mediated transduction of transposon TnFruOut from SBT2-C22	This study
DEpG280::FruOUT	Derived from DEpG by RcGTA-mediated transduction of transposon TnFruOut from SBT2-C22	This study
SBpGΔ280	Deletion of rcc00280 from SBpG	This study
DEpGΔ280	Deletion of rcc00280 from DEpG	This study
Plasmids
pRL27	Suicide plasmid delivering hyperactive Tn5-based transposon	31
pRL27::FruOut	Insertion of R. capsulatus fructose-inducible promoter within Tn5 on pRL27, with promoter facing outward	This study
pCM62	Broad-host-range plasmid	58
pZDJ	Suicide plasmid	11
pmCherry	Plasmid containing mCherry gene	Clontech
pZDJ::pGmCGTA	pZDJ-derived plasmid used to integrate pGTA-mCherry-GTA fusion into R. capsulatus genome	This study
pZDJ::pGmCg15down	pZDJ-derived plasmid used to delete GTA gene cluster	This study
p555::KIXX	Plasmid carrying KIXX-disrupted rcc00555 gene	4
pCM280	pCM62-derived plasmid containing rcc00280 and flanking regions	This study
pCM280G271D	pCM280-derived plasmid containing DE442 rcc00280 allele encoding G271D	This study
pCM280C6His	rcc00280 encoding 6×His tag at 3′ end	This study

Transposon mutagenesis.

The TnFruOut transposon was modified from a hyperactive Tn5 transposon construct, pRL27 (31). Briefly, the fructose promoter (30) was amplified and inserted into pRL27 as a PacI-SphI fragment. The resulting pRL27::FruOut was then conjugated from E. coli into R. capsulatus strain SBpG, which contains the mCherry fluorescent protein coding region transcribed from the RcGTA structural gene promoter. Transconjugants were plated on RCV agar plates with kanamycin. Kanamycin-resistant R. capsulatus colonies were individually picked into 384-well plates by using a QPiX-2 robotic colony picker (Molecular Devices). After growth to stationary phase (>40 h), each 384-well plate was scanned for the optical density at 660 nm and mCherry fluorescence (excitation, 591 nm; emission, 610 nm) using a Varioskan plate reader (Thermo Fisher Scientific).

Construction of mutants, fusion strains, and plasmids.

The construction of chromosome-integrated fusion strains and knockout mutants was performed by introducing the suicide plasmid pZDJ, carrying fusion or knockout sequences, into R. capsulatus for allelic exchange (11). The RcGTA structural gene promoter was fused to the mCherry coding region in-frame with the rcc01682 (ORF g1) start codon (RcGTAp::mCherry fusion) essentially as described previously (5), except that the RcGTA promoter fragment (625 bp) was PCR amplified using primers pGTA5SacI (5′-AGAGGAGCTCGATGCGGCTGCAGACCGATCC-3′; pGTA5 with a SacI restriction site replacing the PstI site in pGTA5 [5]) and pGTA2.6 (5′-GAACCGGATCCATCGCCAGGG-3′) (12); the mCherry coding region was excised from plasmid pmCherry (Clontech) with BamHI and EcoRI. The RcGTA promoter SacI-PstI fragment and the mCherry gene fragments were sequentially subcloned into plasmid pBlueScriptSK such that the rcc01682 start codon was used by the mCherry gene, generating plasmid pBSpGmC. The downstream RcGTA fragment (containing a 108-bp duplication of the rcc01682 5′ region) was generated by PCR using primers GTARarmF (5′-AGAGGAATTCCACGCAAGACATGGACAT) and GTARarmR (5′-GACATCCTCCAGCAGAAC), followed by insertion of the EcoRI- and SalI-digested fragments (an EcoRI-EcoRI fragment and an EcoRI-SalI fragment, sequentially) into pBSpGmC, resulting in plasmid pBSpGmCGTA. The construct was confirmed by sequencing, and the pGmCGTA fragment was subcloned into the suicide vector pZDJ, resulting in pZDJ::pGmCGTA, which was conjugated into R. capsulatus strains SB1003 and DE442 using E. coli S17-1 λpir, with the correct integration of mCherry being indicated by resistance to sucrose and sensitivity to gentamicin. The validity of strains SBpG and DEpG was confirmed by PCR amplification of the RcGTA promoter region, and the gene organization is shown in Fig. 1.

To delete the RcGTA gene cluster (rcc01682 to rcc01698), the 783-bp sequence starting at the stop codon of rcc01698 was amplified with g15downF (5′-AGAGGAATTCTGAGGTGTGCGGCCGATCC-3′) and g15downR (5′-TGAATGTCGACCCCGAACGC-3′) and was used to replace the EcoRI-SalI fragment in pBSpGmC, yielding plasmid pBSpGmCg15down, which contained the RcGTA promoter region, followed by the mCherry gene, followed by the region 5′ of rcc01698. The pGmCg15down fragment was then subcloned into the suicide vector pZDJ, giving rise to plasmid pZDJ::pGmCg15down, which was conjugated into strains SBpG and DEpG using E. coli S17-1 λpir; allele exchange was obtained, based on resistance to sucrose and sensitivity to gentamicin. The RcGTA gene cluster deletion in selected clones was confirmed by PCR. RcGTA-mediated gene transfer was used to transduce the lysis gene (rcc00555) knockout by using the overproducer strain R121 carrying plasmid p555::KIXX (4) as the RcGTA donor strain, as described previously (6).

Plasmid pCM280 was constructed by cloning the rcc00280 gene after PCR amplification of a chromosomal segment from the start of the rcc00279 coding region to the rcc00281 5′-flanking region, using the primers rcc00280A (5′-AGAGGTCGACTATCGCTATCTGCTGACGC-3′) and rcc00280D (5′-AGAGGAGCTCGAGGAATTCTGCGAGGTG-3′). The chromosomal deletion of rcc00280 in strain SBpGΔ280 extended from codon 5 to codon 347. The 603-bp upstream segment inserted into the suicide plasmid pZDJ was created by PCR amplification using primers rcc00280A (see above) and rcc00280B (5′-TCAGATGGCGCCGAAAAGCTTCCCAGACAGGAACAGCAT-3′), and the 620-bp downstream segment was amplified with primers rcc00280D (see above) and rcc00280C (5′-ATGCTGTTCCTGTCTGGGAAGCTTTTCGGCGCCATCTG-3′). Conjugation and subsequent steps were performed as described above for the RcGTA gene cluster deletion. Mutations in rcc00280 and the C-terminal 6×His tag were introduced using the FastCloning method (48). Briefly, PCR primer pairs containing the mutation with a 16-bp overlap, relative to each other, were designed to reverse-amplify pCM280. All PCRs were performed with Phusion polymerase (NEB), the linear PCR products were digested with DpnI, and E. coli S17-1 λpir was transformed with DNA. Mutations were confirmed by DNA sequencing.

Time-lapse fluorescence microscopy and flow cytometry.

Time-lapse fluorescence microscopy was carried out by minor modification of protocols described by Young et al. (49). Briefly, stationary-phase cultures of strain DE/G⁻L⁻ were diluted 100-fold and spotted on RCV medium-1.5% agarose pads. Each agarose pad was then covered with a glass cover slide, and the edges were sealed with paraffin wax to prevent drying during the course of time-lapse microscopy. Fluorescence images and differential interference contrast images were obtained with a DeltaVision microscope (Applied Precision).

Flow cytometry and FACS were performed at the ubcFLOW facilities (University of British Columbia). FACS was carried out on a BD influx cell sorter with a 70-μm nozzle and a sheath pressure of 30 lb/in². Cell populations with mCherry fluorescence were analyzed on a LSRII flow cytometer equipped with a 561-nm laser, as described previously (5) except that laser intensity was set using the WT SB1003 strain so that the negative peak tailed off before 10³ arbitrary units. A total of 50,000 events from each culture sample were recorded for analysis using FlowJo v.10 software. Both biexponential and logarithmic scales are used in histograms to present the data (24), to show that, regardless of which of these two popular methods is used, there are clear differences between the WT and overproducer strains.

Gene transfer assay and RcGTA capsid protein Western blotting.

RcGTA activities were measured using the gene transfer frequency of a rifampin resistance marker into the rifampin-sensitive strain B10, as described previously (25). RcGTA capsid Western blotting was performed as described previously (5), except that culture supernatants that were expected to have WT levels of RcGTA production were concentrated 50-fold in a centrifuge under vacuum.

Differential RNA sequencing and RT-PCR.

For RT-PCR, total RNA was isolated from R. capsulatus early-stationary-phase cultures using TRIzol reagent (Invitrogen), following the manufacturer’s suggestions except that cells were suspended in 1 ml of TRIzol and subjected three times to bead beating (0.4-mm Zircon beads; Ambion), using a FastPrep 24 cell (MP Biomedicals) at a power setting of 6.5 for 30 s, before organic extractions. Isolated total RNA was subjected to DNase I treatment using the DNA-free DNA removal kit (product no. AM1906; Thermo Fisher Scientific), followed by standard phenol-chloroform extraction and ethanol precipitation. RT was carried out using Superscript III reverse transcriptase (Thermo Fisher Scientific), following the manufacturer’s suggestions.

dRNA-seq data for rcc00280 were obtained from a larger data set (50) and are given in File S2 in the supplemental material. Briefly, cultures were grown under anaerobic phototrophic conditions at 35°C in YPS complex medium (47) until 4 h after reaching stationary phase. Cultures were mixed 5:1 with a mixture of 95% ethanol and 5% saturated phenol (51), cells were pelleted by centrifugation, and pellets were frozen in dry ice-ethanol and stored at –80°C. Total RNA isolation was performed with the DNA, RNA, and protein purification kit (Macherey-Nagel), following the manufacturer’s protocol for purification of total RNA. Seven micrograms of total RNA was used as the input for the dRNA-seq protocol (52) and treated with Terminator 5′-phosphate-dependent exonuclease (Epicentre). The original remaining 5′-PPP ends on the resulting purified RNAs were then reduced to 5′-P using RppH (NEB). Library preparation was performed using the Ion Torrent RNA Seq v.2 kit (Thermo Fisher Scientific) according to the manufacturer’s recommendations for total RNA sequencing, with the exception that the 5′ adaptor was first ligated onto the RNA, followed by RNA fragmentation and subsequent 3′ adaptor ligation.

His-tagged Rc280 protein concentration and LC-MS/MS analysis.

Cultures (1 liter, in RCV minimal medium) were grown to the stationary phase, and cells were removed by centrifugation, followed by filtration through a 0.2-μm-pore-diameter filter. Culture filtrates were passed over a 2-ml Ni²⁺-NTA column, which was subsequently washed with 20 ml of buffer A (10 mM potassium phosphate [pH 7.5], 50 mM NaCl) and eluted with 250 mM imidazole in buffer A (1-ml fractions). Gel electrophoresis and Western blotting were performed as described previously (5), except that the anti-6×His primary antibody was a rabbit polyclonal serum (Abcam) and the secondary antibody was an IRDye 800CW-labeled goat anti-rabbit purified immunoglobulin conjugate (Li-Cor). For mass spectrometry, 50 μl of the sample was reduced and alkylated as described previously (53) and 0.5 μg of endoproteinase GluC (NEB) was added, followed by incubation at 37°C for 8 h. The resultant peptides were desalted using a C₁₈ StageTip method (54), and the dried peptides were analyzed with a quadrupole time of flight mass spectrometer (Impact II; Bruker Daltonics), as described previously (55). Data analysis was performed with Byonic v.2.7.7 (Protein Metrics), searching the UniProt R. capsulatus SB1003 FASTA database (3,632 entries, retrieved on 6 April 2018) plus common contaminant sequences, with cysteine carbamidomethylation as a fixed modification and oxidized methionine and deamidated asparagine as variable modifications. The false discovery rate was set to 1% and was determined by searching a decoy database. Enzyme specificity was set to the C-terminal glutamic and aspartic acids, with a maximum of 2 missed cleavages.