The Spectrum of Spontaneous Rifampin Resistance Mutations in the rpoB Gene of Bacillus subtilis 168 Spores Differs from That of Vegetative Cells and Resembles That of Mycobacterium tuberculosis

Abstract

Mutations causing rifampin resistance in vegetative cells of Bacillus subtilis 168 have thus far been mapped to a rather restricted set of alterations at either Q469 or H482 within cluster I of the rpoB gene encoding the β subunit of RNA polymerase. In this study, we demonstrated that spores of B. subtilis 168 exhibit a spectrum of spontaneous rifampin resistance mutations distinct from that of vegetative cells. In addition to the rpoB mutations Q469K, Q469R, and H482Y previously characterized in vegetative cells, we isolated a new mutation of rpoB, H482R, from vegetative cells. Additional new rifampin resistance mutations arising from spores were detected at A478N and most frequently at S487L. The S487L change is the predominant change found in rpoB mutations sequenced from rifampin-resistant clinical isolates of Mycobacterium tuberculosis. The observations are discussed in terms of the underlying differences of the DNA environment within dormant cells and vegetatively growing cells.

The antibiotic rifampin, a potent inhibitor of prokaryotic transcription initiation (23), has long been used to study transcription in bacteria and has also been a highly clinically effective drug, particularly in the treatment of tuberculosis. However, mycobacterial resistance to rifampin can arise from mutations in the rpoB gene encoding the β subunit of RNA polymerase; the vast majority (96%) of these mutations occur within a short 69-bp stretch within rpoB (2, 14, 24) which corresponds to rifampin resistance (Rif^r) cluster I in the well-characterized Escherichia coli rpoB gene (5, 19). In Bacillus subtilis, a single Rif^r mutation, called rfm2103, was historically also found to reside in rpoB (1, 15), and more recently, a larger collection of 18 Rif^r mutations, both spontaneous and generated by ethyl methanesulfonate or N-methyl-N′-nitro-N-nitrosoguanidine mutagenesis, was characterized for B. subtilis (4). To date, all of the Rif^r mutations isolated in B. subtilis have been found to be single nucleotide substitutions resulting in specific amino acid changes located at only two positions, Q469R or Q469K (12 mutations) and H482Y (8 mutations), within the portion of rpoB which also corresponds to cluster I (1, 4) (Fig. 1). In contrast, the Rif^r rpoB mutations found in clinical Mycobacterium tuberculosis isolates (2) exhibited a very different spectrum of nucleotide and amino acid substitutions within cluster I (Fig. 1). Out of a total of 47 Rif^r mutations sequenced from clinical isolates of M. tuberculosis, the most abundant Rif^r mutation occurred at amino acids (using the B. subtilis rpoB coordinates) S487L (18 isolates) and S487W (6 isolates). Additional mutations were found at L467P (one isolate), D472V (four isolates), D472G (three isolates), and L489P (one isolate) (Fig. 1). No examples were found in M. tuberculosis of the Q469K mutation, which was the predominant mutation in B. subtilis, and only one example of the Q469R mutation was found in M. tuberculosis (Fig. 1). At H482, the rpoB mutations found in M. tuberculosis isolates were H482D (seven isolates) and H482C and H482Y (two isolates each), whereas H482Y was the only Rif^r rpoB mutation found in B. subtilis (Fig. 1). Notably, the single C-to-T transition resulting in S487L has been shown by several groups to be the predominant mutation in Rif^r M. tuberculosis isolates, accounting for approximately 40 to 60% of all clinical isolates tested worldwide (8, 20, 22, 26), but this mutation is absent in B. subtilis.

FIG. 1.

Comparison of the rifampin resistance regions of rpoB in B. subtilis and M. tuberculosis. The wild-type rpoB nucleotide and amino acid sequences of the rifampin resistance regions of rpoB in B. subtilis (Bsu) (top two lines) and M. tuberculosis (Mtb) (bottom two lines) are shown. The double dashed line indicates the region corresponding to Rif^r cluster I in the E. coli rpoB gene (5, 19). Nucleotide mismatches which are silent (vertical lines) or lead to amino acid differences (asterisks) are denoted between the sequences. Mutations leading to Rif^r in B. subtilis (1, 4) and M. tuberculosis (2) are denoted above and below the wild-type sequences. The number beside each mutation refers to the number of independent isolations of that mutation. Underlined nucleotides in the B. subtilis sequence denote the position of the PCR primers used to amplify the region in this study.

What might be the underlying reason for such a difference in the mutational spectrum between these two bacteria? We reasoned that clinical tuberculosis is often a reemergence of active M. tuberculosis infection after a period of dormancy in the host which can last for years (7, 13). DNA in dormant mycobacteria may therefore exist either in a different conformation or in a different cytoplasmic environment which alters its chemical reactivity, and thus, the spectrum of spontaneous mutation to Rif^r. There is ample evidence supporting such a notion in the case of dormant endospores of B. subtilis and other spore-forming Bacillus spp., in which dormant spore DNA (i) is associated with α/β-type small, acid-soluble spore proteins, (ii) is packaged in an A-like conformation, and (iii) exhibits a dramatic difference in its chemical reactivity and photochemistry compared to those of DNA from vegetative cells (reviewed in references 12, 17, and 18).

We reasoned that this hypothesis could be tested directly by characterizing spontaneous Rif^r mutants isolated from the same strain of B. subtilis in either the dormant spore phase or the vegetatively growing phase of its life cycle. We first plated 10⁸ spores of B. subtilis 168 (trpC2) from a purified spore stock, which had been stored in distilled water at 4°C for 8 years, onto solid Schaeffer's sporulation medium (SSM) (16) containing 50 μg of rifampin per ml. Strain 168 cannot grow in the presence of 0.5 μg of rifampin per ml (data not shown). Six Rif^r mutant colonies arose after overnight incubation at 37°C. Each colony was picked and restreaked on SSM containing rifampin, and then an isolated Rif^r colony was picked and streaked again on solid Luria-Bertani (LB) medium (11). A single colony from each of the second streaks was then resuspended in 0.2 ml of TE buffer (10 mM Tris-HCl, 1 mM Na₂EDTA [pH 8.0]), template chromosomal DNA was prepared by heating the cell suspension at 95°C for 40 min, debris was removed from the cell suspension by centrifugation for 1 min in a microcentrifuge (6), and the region of rpoB corresponding to cluster I was amplified by PCR using the primer pair shown in Fig. 1. Each 143-bp PCR product was prepared for sequencing (Wizard PCR Prep; Promega) and sequenced on both strands using the PCR primers as sequencing primers at the nucleotide sequencing facility at the Laboratory for Molecular Systematics and Evolution, University of Arizona. Sequencing revealed that each of the six Rif^r mutants from spores contained a single missense mutation in cluster I as follows: Q469R (two of six), H482R (two of six), and S487L (two of six). In addition to the previously observed Q469R mutation (4), spores exhibited two new mutations, H482R and S487L. The S487L mutations were particularly intriguing, as they had not been isolated before in B. subtilis, but were the most prevalent Rif^r mutations found in M. tuberculosis clinical isolates (2, 8, 20, 22, 26) (Fig. 1).

Because all six Rif^r mutants discussed above were isolated from the same batch of spores, and hence cannot be assured to have arisen independently, we decided to further investigate this observation by isolating parallel sets of independent Rif^r mutants from vegetative cells or spores. Strain 168 was streaked for single isolated colonies on solid LB medium, and 36 colonies were picked in duplicate and transferred onto either solid LB medium, on which B. subtilis does not sporulate efficiently, or onto solid SSM, which promotes high-efficiency sporulation. After 16 h of incubation at 37°C, the 36 colonies on LB agar were each removed completely from the plate and spread separately onto SSM plates containing 50 μg of rifampin per ml to select for Rif^r mutants. Meanwhile, the set of 36 duplicate colonies on SSM were incubated at 37°C for 2 days, removed completely from the plate, heat shocked (80°C, 10 min) to select for spores, and plated onto SSM containing 50 μg of rifampin per ml to select for Rif^r mutants. A total of 23 and 20 Rif^r mutant colonies were obtained from the 36 separate platings of vegetative cells and spores, respectively. It was separately determined that individual colonies picked and transferred onto LB and SSM contained on average ∼5.7 × 10⁷ and ∼9.6 × 10⁷ cells per colony, respectively (data not shown), resulting in frequencies of spontaneous mutation to Rif^r of ∼1.1 × 10⁻⁸ for vegetative cells and ∼5.8 × 10⁻⁹ for spores. The distribution of spontaneous Rif^r mutants observed from the 36 pairs of colonies (Fig. 2) strongly resembled that of two independent fluctuation tests (10). Furthermore, Rif^r mutant colonies arose in two different patterns among the plated colonies of vegetative cells and spores (Fig. 2), indicating that the Rif^r mutations had arisen independently in the duplicate sets of colonies. One Rif^r mutant arising from each separate selective plate was chosen, streak purified, PCR amplified, and sequenced as described above. As a control, chromosomal template DNA prepared in parallel from a colony of B. subtilis strain 168 grown overnight on LB agar was amplified and sequenced; its rpoB sequence in cluster I was identical to those previously published (1, 4, 9).

FIG. 2.

Distribution of Rif^r mutants among 36 duplicate selective platings of B. subtilis vegetative cells and spores. See text for details.

Sequence analysis of the cluster I region of rpoB in Rif^r isolates obtained showed a clear difference in the spectrum of spontaneous Rif^r mutations obtained from spores and vegetative cells (Fig. 3). Out of the 10 Rif^r mutations sequenced from vegetative cells, 7 were found to consist of the same A-to-G transition (underlined) changing codon Q469 (CAA) to R469 (CGA), and 3 were found to consist of the same A-to-G transition changing codon H482 (CAC) to R482 (CGC) (Fig. 3). The Q469R mutation has been observed previously in B. subtilis rpoB (4), but the H482R mutation is a new mutation uncovered in this study. In sharp contrast, out of a total of nine spontaneous Rif^r mutations isolated from spores, the majority of Rif^r mutations (five) were found to be located in two completely new codons; most of these new mutations (four) consisted of the same C-to-T transition changing codon S487 (TCA) to L487 (TTA), and a single C-to-A transversion which changed codon A478 (GCT) to N478 (GAT) was noted (Fig. 3). In addition, single examples of the following mutations were observed: a C-to-A transversion changing codon Q469 (CAA) to K469 (AAA), an A-to-G transition changing codon Q469 (CAA) to R469 (CGA), a C-to-T transition changing codon H482 (CAC) to Y482 (TAC), and an A-to-G transition changing codon H482 (CAC) to R482 (CGC) (Fig. 3). A comprehensive comparison of all rpoB mutations isolated in B. subtilis thus far is presented in Table 1.

FIG. 3.

Spectrum of Rif^r mutations in the B. subtilis rpoB gene isolated in this study. The wild-type rpoB nucleotide and amino acid sequences are shown from positions Q469 to S487. Positions of nucleotide changes leading to the indicated amino acid substitutions in vegetative cells and spores are denoted above and below the wild-type sequence, respectively, and the exact nucleotide change is underlined. Amino acid changes resulting from each mutation are in parentheses. The number of independent isolates containing each mutation is denoted to the right of the amino acid change.

TABLE 1.

Summary of mutational changes in B. subtilis and M. tuberculosis rpoB genes leading to Rif^ra

Codon positionb	Codon change	Amino acid change	No. (% of total) isolated in:
			B. subtilis vegetative cells	B. subtilis spores	M. tuberculosis clinical isolates
467	CTG to CCG	L to P	0	0	1 (2)
469	CAA to AAA	Q to K	9 (30)	1 (11)	0
469	CAA to CGA	Q to R	10 (33)	1 (11)	1 (2)
472	GAC to GGC	D to G	0	0	3 (7)
472	GAC to GTC	D to V	0	0	4 (9)
478	GCT to GAT	A to N	0	1 (11)	0
482	CAC to TAC	H to Y	8 (27)	1 (11)	2 (4)
482	CAC to CGC	H to R	3 (10)	1 (11)	0
482	CAC to TGC	H to C	0	0	2 (4)
482	CAC to GAC	H to D	0	0	7 (16)
487	TCR to TTRc	S to L	0	4 (44)	18 (40)
487	TCR to TGR	S to W	0	0	6 (13)
489	CTG to CCG	L to P	0	0	1 (2)

^aData are summarized from references 1, 2, and 4 and this study.

^bCodon positions using B. subtilis coordinates.

^cR, purine (A or G).

Thus, the spectrum of independent spontaneous Rif^r mutations differed considerably in vegetative cells and spores of B. subtilis which were derived from a common set of 36 colonies. This observation is very likely a reflection of the different conformations and cytoplasmic environments of DNA in actively growing cells and dormant spores (12). The major class of spontaneous Rif^r mutations observed in B. subtilis spores consisted of a unique C-to-T transition resulting in S487L (four of nine, or 44% of isolates); this identical nucleotide and amino acid substitution is also the major class of mutations found in clinical Rif^r M. tuberculosis isolates and is recovered at approximately the same frequency (2, 8, 20, 22, 26) (Table 1).

To verify that the new rpoB mutations identified in B. subtilis by sequencing were indeed sufficient to cause a Rif^r phenotype, the following experiments were performed. First, by standard techniques (3), chromosomal DNA (500 ng) prepared from mutants containing the new mutations Q469R, H482R, A478N, and S487L was introduced into competent cells of strain 168 simultaneously with 50 ng of DNA from plasmid pWN162, which carries the wild-type allele of the trpC2 mutation (25). Tryptophan-prototrophic (Trp⁺) transformants were selected on Spizizen’s minimal medium (SMM) (21) and replica plated onto SMM containing rifampin (50 μg/ml) to select for Trp⁺ Rif^r congressants. DNA was prepared from individual colonies of congressants and PCR amplified with the primers indicated in Fig. 1, and the resulting 143-bp PCR products were sequenced. In all cases, the identical nucleotide substitution from the donor strain was detected in the resulting Rif^r congressant (data not shown).

However, because only a 143-bp PCR product spanning cluster I was sequenced, it was formally possible that additional nucleotide changes responsible for Rif^r and residing outside the cluster I region could be present in the original mutants, hence transferred to the congressants. Because the 143-bp PCR product from Rif^r mutants was too small to be efficiently transferred by transformation (data not shown), we designed two primers, 5′-GCGAAAAGCTTGCTTGATTC-3′ and 5′-CCAACAAGAAGATCTCCGTC-3′, which amplified a 1,781-bp region of rpoB from a HindIII site at nucleotide 756 to a BglII site at nucleotide 2517 (underlined portions of the primers) and which included codons 252 to 840, including cluster I. Using the above primers, we amplified chromosomal DNA from the Rif^r mutant carrying the S487L and A478N mutations. Sequencing of the 1,781-bp PCR products confirmed that the only nucleotide changes present consisted of the S487 (TCA) to L487 (TTA) mutation and the A478 (GCT) to N478 (GAT) mutation, respectively (data not shown). We introduced 500 ng of the 1,781-bp PCR products and 50 ng of plasmid pWN162 into competent B. subtilis 168 cells by transformation as described above. In the experiment using S487L as the donor DNA, from a total of 960 Trp⁺ transformants, we obtained four Trp⁺ Rif^r congressants, one of which was shown by PCR amplification of the 1,781-bp segment of rpoB from chromosomal DNA and sequencing to contain only the S487L mutation (data not shown). In the experiment using A478N as the donor DNA, from a total of 1,640 Trp⁺ transformants, we obtained six Trp⁺ Rif^r congressants, one of which was shown by PCR amplification of the 1,781-bp segment of rpoB from chromosomal DNA and sequencing to contain only the A478N mutation (data not shown). Thus, the single C-to-T transition resulting in S487L or the single C-to-A transversion resulting in A478N on the 1,781-bp rpoB fragment was sufficient to cause the Rif^r phenotype in B. subtilis.

Our results therefore lend support to the hypothesis that DNA in dormant M. tuberculosis cells may exist either in a conformation or cellular environment analogous to that seen in dormant bacterial endospores. This structural analogy between DNA in Bacillus spores and dormant mycobacteria could be extended to imply that reactivation of latent M. tuberculosis infection may share features in common with spore germination, which in turn could yield important new insights into the treatment and prevention of tuberculosis.