Investigation of Burkholderia cepacia Complex Methylomes via Single-Molecule, Real-Time Sequencing and Mutant Analysis

ABSTRACT

Bacterial genomes can be methylated at particular motifs by methyltransferases (MTs). This DNA modification allows restriction endonucleases (REs) to discriminate between self and foreign DNA. While the accepted primary function of such restriction modification (RM) systems is to degrade incoming foreign DNA, other roles of RM systems and lone RE or MT components have been found in genome protection, stability, and the regulation of various phenotypes. The Burkholderia cepacia complex (Bcc) is a group of closely related opportunistic pathogens with biotechnological potential. Here, we constructed and analyzed mutants lacking various RM components in the clinical Bcc isolate Burkholderia cenocepacia H111 and used single-molecule, real-time (SMRT) sequencing of single mutants to assign the B. cenocepacia H111 MTs to their cognate motifs. DNA methylation is shown to affect biofilm formation, cell shape, motility, siderophore production, and membrane vesicle production. Moreover, DNA methylation had a large effect on the maintenance of the Bcc virulence megaplasmid pC3. Our data also suggest that the gp51 MT-encoding gene, which is essential in H111 and is located within a prophage, is required for maintaining the bacteriophage in a lysogenic state, thereby ensuring a constant, low level of phage production within the bacterial population.

IMPORTANCE While the genome sequence determines an organism’s proteins, methylation of the nucleotides themselves can confer additional properties. In bacteria, MTs modify specific nucleotide motifs to allow discrimination of “self” from “nonself” DNA, e.g., from bacteriophages. Restriction enzymes detect “nonself” methylation patterns and cut foreign DNA. Furthermore, methylation of promoter regions can influence gene expression and hence affect various phenotypes. In this study, we determined the methylated motifs of four strains from the Burkholderia cepacia complex of opportunistic pathogens. We deleted all genes encoding the restriction and modification components in one of these strains, Burkholderia cenocepacia H111. It is shown that DNA methylation affects various phenotypic traits, the most noteworthy being lysogenicity of a bacteriophage and maintenance of a virulence megaplasmid.

INTRODUCTION

The genus Burkholderia is metabolically very diverse and consists of bacteria that can thrive in a wide range of environments, including soil, water, the rhizosphere, humans, and animals (1, 2). The genus Burkholderia was recently divided into several clades. One clade is the Burkholderia sensu stricto clade, which contains pathogenic species. Other clades are the Burkholderia cepacia complex (Bcc), the Burkholderia pseudomallei group, and the plant-pathogenic species Burkholderia plantarii, Burkholderia glumae, and Burkholderia gladioli. Additionally, new genera were introduced, namely, Paraburkholderia, Caballeronia, and Robbsia (3, 4). Together, these clades are referred to as the Burkholderia sensu lato (Burkholderia in the broad sense).

The ability of the Burkholderia to thrive in highly varied niches is attributed to their unusually large multireplicon genomes, with sizes ranging from 6.2 to 11.5 Mbp (5, 6). All Burkholderia sensu stricto species harbor a primary replicon (chromosome 1 [C1]) encoding genes with essential housekeeping functions, such as DNA replication, cell division, and gene transcription, and a secondary chromosome (chromosome 2 [C2]), which also carries essential genes. Genes carried on C2 are less conserved throughout the Burkholderia but are important for niche adaptation (7). In addition, Burkholderia members often carry further highly variable accessory replicons with unique metabolic capabilities. Within the Burkholderia sensu stricto is a group of closely related bacteria known as the Burkholderia cepacia complex (Bcc). The Bcc was originally considered to have three chromosomes, but recent work has shown that the third replicon is actually a nonessential megaplasmid, which was consequently renamed pC3 (8). Since the pC3 plasmids of various Bcc strains encode antifungal metabolites, we thought to use a “replicon shuffling” approach to construct a plant-beneficial biocontrol strain with negligible pathogenicity. However, the transfer of this replicon was achieved only between some B. cenocepacia strains, and between some B. cenocepacia strains and B. lata 383. Many other transfer attempts between Bcc species were unsuccessful (9). We considered defense against incoming foreign DNA by restriction modification (RM) systems the most likely barrier to transfer.

Bacteria utilize DNA methylation by RM systems as a means of discriminating their own genome from invading DNA such as incoming viral or plasmid DNA. Bacteria use methyltransferases (MTs) to methylate their own genomes, while corresponding restriction endonucleases (REs) cleave differently methylated or unmethylated (incoming) DNA (10, 11). Four different types of RM systems have been described based on their subunit composition, cofactor requirements, sequence recognition, and cleavage position, known as Types I to IV (abbreviated here as TI-TIV) (12). In addition to the basic function of RM systems in genome defense, methylation of bacterial genomic DNA is known to have important roles in chromosome replication, DNA mismatch repair, and DNA-protein interactions, as well as generating phenotypic heterogeneity through phase variation. Epigenetic phase variation is a reversible process by which changes in DNA methylation can regulate gene expression (13–16). Even though both adenine (m6A) and cytosine (m4C and m5C) methylations are present in bacteria, the methylation of adenine bases has been suggested to have a greater effect on bacterial gene regulation, whereas cytosine modification is more commonly used by higher eukaryotes (13).

While most MTs are part of RM systems, solitary or “orphan” MTs (lacking a cognate restriction enzyme) have been described. For example, the Escherichia coli DNA adenine MT Dam is involved in the regulation of DNA replication, DNA repair, and gene expression (17–19). Orthologs of the orphan MT CcrM are widely distributed among the Alphaproteobacteria. This enzyme dictates the timing of DNA replication and is essential for Caulobacter viability (20, 21).

Recent advances in DNA sequencing technology, such as single-molecule, real-time (SMRT) sequencing, have provided new opportunities to detect and analyze the frequency and distribution of methylated bases (22, 23). Here, we used SMRT technology and whole-genome bisulfite sequencing to detect motifs and modifications in several Bcc members, as well as to investigate methylation patterns in B. cenocepacia strain H111. We also deleted the genes encoding putative H111 RM components to allow us to draw conclusions on the role of RM in gene regulation and its effect on bacterial phenotypes.

RESULTS

Distribution of RM components within the genus Burkholderia.

Inspection of REBASE, a comprehensive database for DNA restriction and modification curated by New England BioLabs (NEB) (24), revealed that seven putative RM loci are present on C1 and C2 (two RM system loci, three orphan methylase-encoding loci, and two restriction endonuclease loci), while none was identified on megaplasmid pC3 (Fig. 1). Inspection of our transcriptome sequencing (RNA-seq) data showed that all of the predicted RE and MT genes were expressed in B. cenocepacia strain H111 under standard culture conditions (see Table S1 in the supplemental material). We next compared the RM loci of strain H111 to those of all other Burkholderia sensu latu members entered in REBASE. This comparison revealed that homologues of one H111 RM system (the C1-encoded TI system [TIRM]) and two orphan methylases (the TII orphan MT on C2 [TIIMc2] and gp10) were present in numerous Burkholderia sensu latu genomes, while the remaining five RM components present in H111 were very rare (Table S2). Inspection of the strains bearing each RM component revealed that H111 TIIMc2 was present in the majority of Burkholderia sensu lato strains, while the H111 TIII RM system (TIIIRM) was broadly present across the Bcc and the B. pseudomallei group. The prevalence of these RM components in the Bcc led us to consider them as “core” RM components in this complex. These two methylases were previously identified in the B. pseudomallei 982 genome, and predictions had been made for their recognition motifs (25, 26). To find homologues of the H111 RM components in the Burkholderia sensu lato, tBLASTN was used to screen the genomes of one representative strain from each Burkholderia sensu lato species present in REBASE in addition to two commonly studied B. cenocepacia strains (J2315 and HI2424). The results are illustrated as a heatmap in Fig. 2, next to a phylogenetic tree of the strains generated from their concatenated gyrB and rpoD genes. This analysis highlighted the homology of TIIIRM, TIIMc2, and Gp10 within the Burkholderia sensu stricto (Fig. 2). A homologue of TIIMc2 was found in each Burkholderia sensu lato representative and within the Ralstonia pickettii outgroup, demonstrating the presence of this MT in the broader Burkholderiales order. The phage-encoded gp10 MT gene was present within bacteriophage regions in multiple Burkholderia sensu lato species in a pattern consistent with acquisition by phage transduction. Interestingly, no close homologue of the H111 gp51 MT gene, which is part of the same bacteriophage (ϕ1), was found. The TIV RE encoded on C1 (TIVRc1) and TIRM were specific to B. cenocepacia H111, and the remaining TIV RE (TIVRc2) was found in only two other strains. To allow us to identify and compare genomic methylation patterns, the methylomes of Burkholderia cenocepacia H111, B. lata 383, B. ambifaria AMMD, and B. multivorans ATCC 17616 were sequenced using single-molecule, real-time (SMRT) sequencing.

FIG 1.

RM components present in the B. cenocepacia H111 genome. R, restriction endonuclease gene; M, methyltransferase gene. Recognition motifs confirmed in this study have been shown in green. The locations of prophages have also been shown.

FIG 2.

Distribution of homologous B. cenocepacia H111 RM components throughout the Burkholderia sensu lato. The phylogenetic tree was generated from concatenated gyrB and rpoD genes, which are essential and highly conserved, with the Ralstonia pickettii 12D genes used to give an outgroup. The amino acid sequence of each restriction endonuclease (R) or methyltransferase (M) gene was used to query a BLAST database including the genomes of each strain shown. Numbers on the heatmap represent percentage identity, calculated as the number of identical residues between the query and the match, as a percentage of the number of residues present in the H111 query sequence (excluding the stop codon). Matches with less than 20% identity were excluded.

Our sequencing data confirmed the predicted methylated motifs in B. cenocepacia H111 (CACAG and GTWWAC) and also found methylated motifs for B. ambifaria AMMD and B. multivorans ATCC 17616, for which no motif predictions had yet been made (Table 1). We found a further methylated motif (5′-CAG-N₆-TTYG-3′) in B. cenocepacia H111 of the type methylated by type I MTs (27, 28). Such a system was identified by REBASE on H111 C1. In addition to the two core motifs, a TI motif was found in our B. multivorans ATCC 17616 genome sequence. No such TI RM system was present in the publicly available ATCC 17616 genome; however, inspection of our sequencing data revealed the presence of a TI RM system encoded on C1, which is probably responsible for methylating the 5′-CCA-N₆-RTTC-3′ motif. This suggests that our B. multivorans ATCC 17616 differs from the previously sequenced strain.

TABLE 1.

Methylated motifs in Bcc strains, detected by PacBio SMRT sequencing^a

Bcc strain	Gene coding for methylase	Motif	Modified position	Type	No. of motifs detected	No. of motifs in genome	Mean modification QVb	Mean coverage	Partner motif
B. cenocepacia H111	I35_3274	CACAG	4	m6A	6,289	6,347	95.05	56.09
I35_3254	CAGN6TTYG	2	m6A	3,249	3,295	97.39	57.21	CRAAN6CTG
I35_3254	CRAAN6CTG	4	m6A	3,243	3,295	86.11	55.21	CAGN6TTYG
I35_4914	GTWWAC	5	m6A	1,688	1,758	86.13	56.34	GTWWAC
B. lata 383	Bcep18194_A3213	CACAG	4	m6A	7,216	7,823	66.3	40.1
Bcep18194_B2124	GTWWAC	5	m6A	1,585	1,792	63.05	40.6	GTWWAC
B. ambifaria AMMD	Bamb_0020	CACAG	4	m6A	6,204	6,323	80.4	47.41
Bamb_1850 or Bamb_6619	RGATCY	3	m6A	11,063	11,860	74.64	48.5	RGATCY
Bamb_3349	GTWWAC	5	m6A	1,531	1,670	68.76	47.81	GTWWAC
B. multivorans ATCC 17616	GAAYN6TGG	3	m6A	841	1,198	94.29	62.67	CCAN6RTTC
CCAN6RTTC	3	m6A	835	1,198	98.14	64.39	GAAYN6TGG
BMUL_RS00145	CACAG	4	m6A	3,543	5,303	106.18	62.82
BMUL_RS23375	GTWWAC	5	m6A	1,250	1,880	92.77	62.41	GTWWAC

^aSMRT sequencing was performed using the PacBio RSII system. Secondary analysis was performed using the SMRT portal.

^bThe mean modification quality value (QV) for all instances where this motif was detected as modified motifs with quality value (QV) of >40 are shown.

An additional motif (RGATCY) typical for TII RM systems was found in B. ambifaria AMMD (29, 30). Two TII systems are encoded in the AMMD genome. The B. lata genome was methylated only at the core motifs, consistent with REBASE predictions.

In B. cenocepacia H111, 60,867 modifications were detected, of which 14,585 were adenine base modifications (m6A) and 2,804 were cytosines (m4C). Of the detected adenine modifications, 98% were assigned to a specific motif, whereas no specific motifs could be identified for the modified m4C bases detected (Fig. 3, Table 1; see also Fig. S1 in the supplemental material). It should be noted that SMRT sequencing detects m6A and m4C modifications much more efficiently than m5C modification and that the two phage-encoded methylases, gp10 and gp51 (dealt with in more detail later on), were predicted to be m6A/m4C and m5C methylases, respectively. A further 43,478 “modified bases’” were detected, to which the PacBio SMRT Portal could not assign a precise type of modification (Fig. S1). These unspecific “modified bases” could represent methylated bases or could result from DNA damage occurring before or during DNA extraction and purification. Such damage also results in modified bases, such as 8-oxoguanine and 8-oxoadenine.

FIG 3.

Distribution of m6A base modifications. Patterns in purple represent m6A modifications that were assigned to specific motifs: inner circle motif, CAGN₆TTYG/CRAAN₆CTG TIRM; second circle motif, CACAG (TIIIRM); outer circle motif, GTWWAC (TIIMc2). Every hatchline in each circle is a representation of a 10,000-bp window. The numbers of modified bases or motifs per window is represented by the color range. The darkness of the color corresponds to the number of modifications (maximum number of modifications found in a window: C1, 24 modifications; C2, 28 modifications; pC3, 22 modifications). Circles were constructed using the circlize package for R. Position 0 (where the break in each circle occurs) is the origin of replication as determined by reversal of GC skew.

To analyze the H111 methylome for hot spots, we divided each replicon into 10,000-bp windows and calculated the abundance of modifications (Fig. 3). The occurrence of each type of modification was mostly evenly distributed; however, some windows contained an increased number of motifs (Fig. 3). Most noticeable was an increase in m6A methylations at the origin of replication of each replicon (Fig. 3). These modifications were mainly at CACAG motifs, recognized by the core TIIIRM system identified within the Burkholderia.

Construction of an RM null mutant of B. cenocepacia H111.

To investigate the biological functions of the B. cenocepacia H111 RM components, we sequentially deleted the seven RM-encoding loci present in this strain (TIIIRM, TIRM, TIVRc1, gp10, gp51, TIVRc2, and TIIMc2 [Fig. 1]) using an I-SceI-dependent, markerless gene deletion approach (31). This approach resulted in various intermediate mutants in addition to the final RM null mutant, and two additional single RM mutants were also constructed (Table 2). After four rounds of deletion mutagenesis, it became apparent that the TIIM encoded by the prophage III gp51 gene was essential, in full agreement with our recent mapping of essential genes required for growth of B. cenocepacia H111 (32). For further analysis, we constructed a conditional mutant (strain CM51), in which gp51 expression was controlled by a rhamnose-inducible promoter (Table 2). This mutant could grow only in the presence of rhamnose, demonstrating that gp51 is an essential gene (Fig. S2). In the course of constructing the gp51 deletion mutant, we isolated a spontaneous mutant which had lost the entire prophage III region from its genome, including the methylase genes gp10 and gp51. This demonstrated that gp51 was only essential as part of the prophage III region, probably for maintaining the phage in its lysogenic state. The remaining methylase gene (I35_2582, encoding TIIMc2) was then deleted to give an RM null mutant of B. cenocepacia H111. During phenotypic and sequence analysis of this mutant, it was determined that the ∼1-Mb megaplasmid pC3 had been lost. A mobilized pC3 was therefore introduced into the null mutant using previously developed techniques (7), to give strain NullpC3⁺. An analogous version of strain H111 was constructed as a control and will be referred to as H111pC3⁺.

TABLE 2.

Bacterial strains and plasmids used in this study

Species and strain	Genotype or characteristic(s)	Reference or source
Burkholderia cenocepacia H111 and derivatives
H111	CF isolate from Germany, R-6282, Ampr	70
H111Δc3	pC3 deletion mutant	7
TI	TIRM (I35_3251–I35_3254) deletion mutant on C1, markerless	This study
DM	Double mutant: derived from the TI mutant with additional deletion of TIVRc1 (I35_3250)	This study
TM	Triple mutant, derived from DM strain with additional deletion of TIIIRM (I35_1826, I35_1825) on C1	This study
SM-TIII	Single TIIIRM (I35_1826, I35_1825) on C1–pC3 deletion mutant	This study
QM	Quadruple mutant, derived from TM strain with additional deletion of the TIVRc2 (I35_ 1041). Prophage region III on C1 including type II MT gp10 (I35_2397) and gp51 (I35_2438) spontaneously lost	This study
Null	Complete RM system mutant, derived from QM with additional deletion of the TIIMc2 (I35_2582)	This study
SM-OMC2	Single mutant of the TIIMc2 (I35_2582)	This study
NullpC3-	Complete RM system mutant, pC3 lost, phage region III lost	This study
NullpC3+	Derived from NullpC3-, pC3 reintroduced by conjugation into a spontaneous Rifr mutant, gabD::pSHAFT on pC3, Cmr	This study
H111pC3+	H111 gabD::pSHAFT on pC3, Cmr Rifr	This study
Null-tag	Strain derived from NullpC3+, C1 bears dhfrII under the control of the LacI-controlled pA1-04/03 promoter, pC3 bears lacI under the control of promoter pJ23109. Tpr Gmr Cmr	This study
H111-tag	Strain derived from H111-hfe (9), gabD::pSHAFT on pC3, C1 bears dhfrII under the control of the LacI-controlled pA1-04/03 promoter, pC3 bearing lacI under the control of promoter pJ23109. Tpr Gmr Cmr	This study
CM51	Conditional mutant (CM): gp51::pSC200 in H111	This study
CM10	CM: gp10::pSC200 in H111	This study
Other Burkholderia strains
K56-2	B. cenocepacia K56-2, CF isolate, prototroph, BCESM + cblA+	71
HI2424	B. cenocepacia HI2424, LMG24507, soil isolate, USA	8
MCO-3	B. cenocepacia MC0-3, LMG24308, soil isolate, maize rhizosphere, USA	8
383	B. lata 383, LMG22485, ATCC 17660, R18194, wild type, isolated from forest soil in Trinidad and Tobago	LMG strain collection, laboratory collection
17616	B. multivorans ATCC 17616, LMG17588, type strain	71; laboratory collection
AMMD	B. ambifaria AMMD, LMG19182, CF isolate	72
LMG10929	B. vietnamiensis LMG10929. soil isolate, prototroph	LMG strain collection, laboratory collection
Escherichia coli strains
MC1061	hsdR araD139 Δ(ara-leu)7697 ΔlacX74 galU galK rpsL (Smr)	73
CC118λpir	Δ(ara leu)7697 araD139 ΔlacX74 galE galK phoA20 thi-1 rpsE rpoB(Rifr) argE(Am) recA1 λpir+	57
DH5α	F− ϕ80dlacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rK− mK+) supE44 thi-1 relA1 gyrA96	74
SY327λpir	Wild-type strain; araD Δ(lac pro) argE(Am) recA56 Rifr nalA λpir	75
S17-1λpir	thi proA hsdR recA RP4-2-tet::Mu-1 kan::Tn7 integrant (Tpr Smr) λpir	76; laboratory collection

Verification of methylation loss in the RM null mutant and identification of motifs recognized by the MTs.

We used SMRT sequencing to assign the methylated motifs present in the H111 genome to their cognate RM components. Three m6A-type methylated motifs were detected. The genomes of strains carrying single mutations in I35_3254 (encoding TIRM), I35_2582 (encoding TIIMc2), and I35_1825 (TIIIRM) were subjected to SMRT sequencing. Each mutant was not methylated at one of the motifs that was methylated in strain H111, allowing the predicted motifs 5′-CAG-N₆-TTYG-3′ for TIRM, GTWWAC for TIIMc2, and CACAG for TIIIRM to be confirmed.

The genome of the RM null mutant was also subjected to SMRT sequencing. No m6A or m4C modifications were identified; however, 179,975 unassignable modifications were detected by SMRT portal analysis (Fig. S1B). This figure is approximately fourfold higher than in the wild type, in which 43,478 modified bases could not be assigned. While the detected modifications could result from base methylation, it is likely that this represents damage to the DNA, which gives rise to modified bases such as 8-oxoguanine. SMRT sequencing also verified the complete and clean loss of phage region III from C1, and the loss of pC3. Additionally, the sequence was scrutinized for unintended deletions that might have occurred during construction, and none were found.

Whole-genome bisulfite sequencing reveals the target of the Gp10 methylase.

To investigate cytosine methylation in strain H111, we carried out whole-genome bisulfite sequencing (WGBS) for strains H111pC3⁺ and NullpC3⁺ and a single deletion mutant of the gp10 MT gene, SMgp10. Overall, cytosine methylation was found to be relatively low, with approximately 10% of the cytosine read positions analyzed showing methylation, but generally in only one or two reads for a single replicate. For our analysis of phage region III, we considered only methylated positions that were found in at least one read in all three biological replicates. The 41,500-bp phage region III contained 458 positions that showed methylation in all three H111pC3⁺ replicates (a mean of one methylated cytosine per 91 bases). Certain genes contained very few methylated positions, such as the holin- and excisionase-encoding genes (Table S3). Several genes encoding phage structural proteins showed higher than average methylation, as did gene gp42, which codes for a putative transcriptional regulator. Methylation was sparse within intronic regions (26 methylated positions out of a total of 3,465 bp, a mean of one per 133 bp).

Only one position, which was methylated in 50%, 50%, and 25% of reads in the three H111pC3⁺ replicates, was not methylated in any of the SMgp10 reads. The position of the probable target of the Gp10 methylase is the final C in the sequence 5′-CTCGACGGGCTGCAGCGATGGGTGGCC-3′, which is within the gene encoding the tape measure protein (responsible for determining tail length). No conclusions could be drawn about the Gp51 MT recognition sequence, as a single mutant could not be constructed (due to its essentiality) and since the phage region was absent from the RM null mutant.

Transcriptional profiling suggests effects of DNA methylation on cell motility, iron uptake, and genome integrity.

RNA-seq was carried out on the H111 RM null mutant NullpC3⁺ and the control strain H111pC3⁺. The top 500 genes showing significant changes in their expression (P value ≤ 0.01 and absolute log₂ fold change ≥ 0.5) were further analyzed. Of these genes, 240 were upregulated and 260 were downregulated in the null mutant compared to the control (Table S4). As expected, the 61 spontaneously lost prophage genes, and the deleted RM component genes were among the 260 genes showing decreased expression in the null mutant. Several downregulated genes are involved in motility, including fliI, flgD, fliL, motB_2, and motA_1. Also notable was murA, which codes for a cell wall hydrolase that plays a role in cell wall formation and cell separation, and I35_1589, encoding the putative fimbria usher protein StfC. Another gene that showed decreased expression in the RM null mutant was trpB. This gene is part of the tryptophan cluster and is located upstream of the core type II methylase gene. The TIIMc2 gene was cleanly deleted, leaving the intergenic regions on either side intact, and therefore, the decreased trpB expression is unlikely to be an artifact resulting from the deletion process.

Several genes involved in replication, recombination, and repair, especially in the SOS response, were upregulated in the RM null mutant compared to strain H111. These genes include the repressor-encoding gene lexA, recA, and I35_2899 (which putatively encodes the RecA/RadA recombinase), and the genes I35_2143 and I35_2898 (which code for DNA polymerase IV and a homologue of DNA polymerase-like protein PA0670 from Pseudomonas aeruginosa [33]). Other genes important for DNA replication and recombination (I35_1669, coding for a homolog of eukaryotic DNA ligase III and dnaE_2, encoding a DNA polymerase III alpha subunit), as well as DNA repair (I35_7256, coding for an exonuclease subunit A, which is part of the UvrABC DNA repair system) also showed an increase in transcription in the null mutant compared to strain H111 (34). In addition, we observed higher read counts of several genes coding for assembly proteins of the two remaining prophages in the null mutant (Table S4), likely reflecting SOS-dependent prophage expression, as has been reported for Burkholderia thailandensis (35). Furthermore, genes involved in secondary metabolite biosynthesis, transport, and metabolism, especially in pyochelin biosynthesis and utilization (e.g., pchB, pchE, pchR, ftpA, ftpB) were found to be upregulated in the null mutant compared to H111.

Phenotypic analysis of the RM mutant.

To further investigate the importance of RM components for the characteristics suggested by transcriptomic analysis (cell replication, cell morphology, DNA repair, cell motility, iron uptake, and maintenance of lysogeny), and in other possible traits, phenotypic assays were carried out. As a basis for these assays, growth of strains H111pC3⁺ and NullpC3⁺ in liquid culture was examined spectrophotometrically and found to be comparable (Fig. S3).

DNA transfer into the RM null mutant is more efficient than into the wild-type strain.

Given that the main purpose of RM systems is to limit the uptake of foreign DNA, we determined the transfer frequencies of the null mutant and the wild-type strain for a mobilizable derivative of pC3 (9). We used B. cenocepacia K56-2, which possesses the two core methylases but lacks the TIRM and the two phage methylases (Table S2), as the donor and either the RM null mutant (NullΔpC3) or H111Δc3 (7) as the recipient. A 24-fold-higher transfer rate was observed into the RM null mutant compared to the H111 control strain, confirming that RM plays an important role in protecting the genome against incoming DNA (Fig. 4A). We attempted to use this increase in conjugational efficiency to achieve our original aim of replicon shuffling within the Burkholderia genus. Transfer of pC3 from Burkholderia vietnamiensis LMG 10929 into NullΔpC3 was therefore attempted, but without success. B. vietnamiensis LMG 10929 carries the core TIIIRM and TIIMc2 components, as well as a plasmid bearing four MT genes and a TIVR. While these should not affect conjugation into the RM null recipient, they could restrict the uptake of the helper plasmid used in the triparental mating procedure. To exclude this possibility, we introduced the broad-host-range plasmid pBBR1MCS into B. vietnamiensis LMG 10929 and B. cenocepacia K56-2 and used these as donors to determine conjugation efficiency using strains NullΔpC3 and H111ΔC3 as the recipients (Table S5). The RM null recipient was found to be 60-fold more efficient in uptake than the control when LMG 10929 was used as the donor, although only 7-fold more efficient when the donor was K56-2 (Fig. S4).

FIG 4.

Phenotypic changes observed in RM null mutant. (A) Conjugal uptake was increased 24-fold in the absence of RM. Rifampicin-resistant derivatives of pC3-cured strains of NullpC3⁻ and H111Δc3 were used as recipients for conjugal transfer. K56-2 pC3 was mobilized by the integration of pSHAFT2, which carries an oriT and a chloramphenicol resistance marker. Transfer frequency was calculated by dividing CFU on media containing rifampicin plus chloramphenicol (total exconjugants) by CFU on media containing rifampicin (total recipient cells). Error bars represent the standard deviations (SD). (B) Fluorescence microscopy reveals filamentous cells. Exponential-phase cells were subjected to microscopic analysis. Cells were observed with an epifluorescence Leica DM6000 B research microscope at ×1,000 magnification. (C) Swarming motility is decreased in the absence of RM. The means of three biological replicates are shown, error bars represent the standard deviations (SD). Significance was determined using a two-tailed t test (P value of 0.0333). (D) Swimming motility is decreased in the absence of RM. The means of three biological replicates are shown; error bars represent the SD. Significance was determined using a two-tailed t test (P value of 0.0435). (E) Proteolytic activity is decreased in the RM null mutant. Bars represent the means of three biological replicates. Error bars represent the SD. The absorbance at OD₄₄₂ was measured and normalized against the cell density OD₆₀₀. Significance was determined using a two-tailed t test (P value of 0.00931). (F) RM is important for biofilm formation. Photographs illustrate differences in pellicle formation. The images are representative of a data set of at least three replicates. Graph shows biofilm formation using the crystal violet assay. Error bars indicate SD (n = 3). Significance was determined using a two-tailed t test (P value of 0.0187).

Increased pC3 loss in the RM null mutant confirms the importance of RM in replicon stability.

To determine the frequency of pC3 loss, a modified version of a previously described experiment was carried out (8). The RM null mutant and H111 control strains were modified to allow positive selection of cells that had lost pC3. To achieve this, the trimethoprim (Tp) resistance gene (dhfrII) was placed under the regulation of a synthetic lac promoter and integrated onto C1. The gene coding for the Lac repressor and a gentamicin resistance marker (for selection) were inserted into pC3. As a result, cells bearing pC3 were Tp sensitive, due to repression of dhfrII by the Lac repressor. This repression was relieved upon loss of pC3, allowing positive selection. Loss of pC3 in the RM null mutant strain was 173-fold higher compared to the H111 control strain, demonstrating that RM components play a central role in replicon maintenance (Fig. 5A).

FIG 5.

pC3 loss was more frequent in the absence of RM. (A) To assess pC3 stability, strains H111-tag and Null-tag, which were designed to become Tp resistant on loss of the pC3 replicon, were grown in rich medium for 24 h at 37°C. The frequency of pC3-deficient cells was plotted. Error bars indicate SD (n = 3). (B) Increased pC3 loss in the absence of RM was phenotypically visible on NYG medium. (C) Inoculum from the places indicated by the numbers in panel B was streaked on M9 medium with uracil (ura) (7) to confirm the presence/absence of pC3. Strong growth indicates pC3 presence.

The loss of pC3 from H111 alters the phenotype on NYG plates, as a result of reduced exopolysaccharide (EPS) production (7). Loss of pC3 in the RM null mutant occurred so frequently it could be observed through the formation of wedge-shaped areas with a more transparent appearance (Fig. 5B and C). This phenotypic change probably occurred due to the location of the shvR regulator gene on pC3, which is known to influence colony morphology (36). When analyzed for the presence of pC3 by PCR, the absence of pC3 from the more transparent wedges was confirmed.

Loss of DNA methylation leads to filamentous cell growth.

Comparison of the RM null mutant (NullpC3⁺) and the control strain by fluorescence microscopy revealed that the RM null strain formed chains of cells during the exponential growth phase, while the control did not (Fig. 4B). In cultures grown to stationary phase, these chains were no longer apparent (Fig. S5A). A filamentous phenotype can occur due to the replication arrest caused by the SOS response (37). To investigate the importance of the SOS response in chain formation, conditional recA mutant versions of strains H111pC3⁺ and NullpC3⁺ were constructed (named H111pC3⁺recA and NullpC3⁺recA, respectively). This allowed us to maintain repression of the SOS response by adding glucose to the culture medium. Microscopic analysis of these mutants revealed the presence of chains of cells in strain NullpC3⁺recA but not in H111pC3⁺recA (Fig. S5), suggesting that the chain phenotype was not related to the SOS response.

RM components influence cell motility, biofilm formation, and proteolytic activity.

As suggested by our RNA-seq data, we observed a reduction in swimming and swarming motility in the RM null mutant compared to the control strain (Fig. 4C and D). We investigated other phenotypes known to be associated with the RpoN sigma factor, since this was found to be less expressed in the RM null mutant in our RNA-seq analysis. RpoN is known to repress multiple phenotypes, including the production and secretion of extracellular proteases, EPS, and biofilm formation (38–40). There was no difference in EPS production between the two strains (Fig. S5A); however, proteolytic activity and biofilm formation were reduced in the null mutant compared to strain H111 (Fig. 4E and F). To investigate whether the observed differences occurred due to activation of the SOS response by DNA damage resulting from deletion of the RM components, we also tested the recA mutants for proteolytic activity and biofilm formation. The H111 strain and RM null recA mutants showed similar proteolytic activities to their parent strains, suggesting that the difference between the H111 control and RM null strains was independent of the SOS response (Fig. 4E). However, the H111 control strain produced significantly more biofilm than its recA mutant, which produced biofilm in an amount similar to those of the RM null and RM null recA mutants (Fig. 4F), indicating that the observed difference in biofilm formation may be a consequence of the induction of the SOS response in the null strain.

DNA methylation is not involved in oxidative, heat, membrane damage, ciprofloxacin or osmotic stress tolerance, or in antifungal activity.

Tests of the RM null mutant versus the H111 control strain for persistence under oxidative, osmotic, membrane damage, and heat stresses showed no significant difference between the two strains (Fig. S5). Tests for antifungal activity and pathogenicity against wax moth larvae likewise showed no differences (Fig. S5E). No difference was observed in ciprofloxacin resistance between the two strains (Fig. S5A). Ciprofloxacin acts by causing double-strand breaks, and resistance to it requires upregulation of the SOS response.

Deletion of RM components leads to an increase in phage and membrane vesicle production.

Our RNA-seq analysis had suggested that phage production might be increased in the RM null mutant. Transmission electron microscopy (TEM) was therefore used to investigate the presence of phages and phage-like structures in the supernatants of the wild-type H111 strain and the clean RM null mutant. We observed the presence of phages and phage tails, either from partially assembled phages or tailocins (Fig. 6), as well as membrane vesicles of various sizes. To compare membrane vesicle production between the wild type and the null strain, they were collected and quantified by staining with the fluorescent membrane dye FM 1-43. We observed a 2.2-fold increase in membrane vesicle (MV) production in the RM null mutant compared to the H111 wild type (Fig. 6C), probably as a result of phage-triggered cell lysis (Turnbull et al. [41]).

FIG 6.

Membrane vesicles and phage particles were more abundant in cultures of the RM null mutant. Concentrated culture supernatants were visualized at ×180,000 magnification by TEM. White arrows indicate phage tails of the Myoviridae or Siphoviridae family. (A) H111pC3⁺; (B) RM null mutant, NullpC3⁺. (C) Membrane vesicles were more abundant in the absence of RM. Concentrated supernatants from the H111 control strain (H111pC3⁺) and the RM null mutant, NullpC3⁺, were stained with FM 1-43, to quantify membrane vesicles (MVs). Error bars indicate SD (n = 3).

Chemical and phenotypic assays for pyochelin production confirm the RNA-seq results.

Genes involved in the biosynthesis and utilization of the siderophore pyochelin showed a transcriptional increase in the RM null mutant compared to strain H111 (Table S4). To verify this result, we extracted the siderophores from culture supernatants of strains NullpC3⁺ and H111pC3⁺ and analyzed them by thin-layer chromatography. The band observed for the RM null mutant sample was clearly darker than for the H111 control strain, suggesting that this strain produced more pyochelin than the control. Liquid chromatography-mass spectrometry (LC-MS) was used to identify and quantify the pyochelin fraction. The relative amount of pyochelin was estimated to be around twofold higher in the RM null strain relative to the wild-type strain (Fig. 7).

FIG 7.

The absence of RM results in an increase in pyochelin production. (A) Thin-layer chromatography performed on the RM null mutant and H111 control suggests that pyochelin production is increased in the null mutant. Siderophores were extracted and separated by TLC. Chloroform-acetic acid-ethanol at 90:5:10 (vol/vol/vol) was used as the developing solvent. Ferric chloride was used to visualize the siderophores. Liquid chromatography coupled to tandem mass spectrometry (LC-MC/MS) confirmed the location of pyochelin and salicylic acid on the TLC plate. LC-MC/MS was performed with Δppm < 3 ppm. (B) Estimation of pyochelin content. Replicates marked with the letter b in panel A were used. Differences in pyochelin production were approximately quantified by determining the area under the LC-MS curve.

Inspection of the upstream regions of the top 500 differentially regulated genes between strains H111pC3⁺ and NullpC3⁺ revealed only one candidate for direct involvement of methylation on promoter activity, the upstream region of the pchE gene, encoding the nonribosomal peptide synthetase for assembly of pyochelin. This contains a methylated CACAG motif, with the methylated base at position −54 relative to the translational start codon. Two lacZ reporter vectors were constructed bearing this region, one containing the wild-type (WT) PpchE sequence, and one in which the methylated motif was altered to CACAC to abrogate methylation. These reporters were analyzed in H111pC3⁺ and NullpC3⁺, and the methylation status of the promoters was confirmed. Activity of the nonmethylated PpchE promoter region was found to be approximately 25% higher than the WT sequence in strain H111pC3⁺. The activities of both promoters were similar in the null mutant, indicating that the single base change made to abrogate methylation had little effect on activity (Fig. 8).

FIG 8.

Methylation directly affects PpchE promoter activity. The activity of lacZ reporter vectors bearing WT PpchE and a derivative in which the methylated motif was altered were assayed in strains H111pC3⁺, NullpC3⁺ and their recA mutants. Bars represent the means of three biological replicates, and error bars indicate the standard deviations. Statistical significance was evaluated by Student’s t test, and the P values were as follows: 0.0077 (**) for H111pC3⁺ with the WT versus mutant (mut) PpchE reporters and 0.04 (*) for the H111 recA mutant with WT versus mutant PpchE reporters.

DISCUSSION

DNA methylation is important for various bacterial cell functions, including host defense, genome integrity, and regulation of cellular processes (20, 42). In this study, we aimed to investigate the methylome of B. cenocepacia H111, to allow us to identify specific methylation patterns, and to study the effects of epigenetics on a broad range of biological processes. We found that three RM loci on H111 C1 are very specific to this strain, while TIIIRM and the orphan methylase TIIRc2 were conserved within the entire genus. Homologues of TIM were found in six strains, while homologs of TIVRc1 and TIVRc2 were present in one and six strains, respectively. We made use of single-molecule, real-time (SMRT) sequencing technology to identify the methylated bases within the genomes of four Bcc members, each from separate species. Our sequencing data confirmed the two core methylated motifs predicted by REBASE (CACAG and GTWWAC) in the strains sequenced. In addition to the core motifs, further motifs were found in three of the four sequenced strains, while B. lata 383 showed only the core methylated motifs, consistent with REBASE predictions. The methylated motifs identified in B. cenocepacia H111 were later experimentally assigned to their cognate methylases by SMRT sequencing of single mutants. Whole-genome bisulfite sequencing was used to determine a target site of the Gp10 methylase, although since just one target site was found, the precise motif could not be determined.

We observed that the methylated CACAG motif occurred more frequently in and around the origins of replication of the three H111 replicons. Studies in other Proteobacteria have shown that DNA methylation is important for regulation of chromosomal replication and that m6A modification, for example of the GATC motif in E. coli, is densest at the origin of replication (oriC) (20, 43).

The core TIIM of strain H111 (TIIMc2) recognizes the motif GTWWAC. We speculated that the presence of this methylase might be involved in the regulation of the trp genes, required for tryptophan production, and indeed, the RM null mutant showed a decrease in expression of the trpB gene in our RNA-seq analysis. A single gene deletion mutant of TIIMc2 (SM-OMC2) did not show significant differences in growth in the presence and absence of tryptophan (data not shown). However, we observed increased production of a brown/orange pigment, presumably melanin, when growing the mutant strain in nutrient-rich IST medium. Melanin can act to quench reactive oxygen species (44, 45), and therefore, its production could be an indication that the cells are stressed. Interestingly, TIIMc2 was found to be highly conserved throughout the Burkholderia sensu lato, and even beyond, suggesting that it was acquired early in evolution. While the high level of conservation might reflect selective pressure for maintenance of the methylase gene, it could also be a consequence of it being within a gene cluster (I35_2579 to I35_2673) which contains the majority of the essential genes present on C2 (32).

To investigate the effect of RM on H111 phenotypes, we sequentially deleted the RM systems and components present in its genome. Attempts to delete the orphan methylase encoded by gp51, followed by the construction of a conditional mutant, revealed that this methylase is essential. This suggests that the encoded methylase is important in maintaining the lysogenic state of the phage. This has been previously demonstrated for the Dam MT in enterohemorrhagic E. coli, which carries the Shiga toxin-encoding bacteriophage 933W (46). Although gp51 was the only essential gene apparent in phage region III, inspection of transposon sequencing (Tn-seq) data revealed that gp44, whose product shows homology to the λ repressor protein, was also essential (32). This was initially overlooked due to its small size. The neighboring, divergently transcribed gp43 gene shows homology at the amino acid level to the λ Cro protein. Gene expression can be altered by promoter methylation, which generally prevents expression of a gene. The briefly occurring hemimethylation of replicons following replication can allow expression of such genes. The essential role played by gp51 in lysogeny leads us to speculate that the Gp51 methylase might affect the competition between Gp43 and Gp44 to control a genetic switch determining progression to the lytic or lysogenic lifestyle. By linking induction with the cell cycle via methylation, ϕ1 might ensure a constant, low level of induction by an epigenetically triggered stochastic switch. Analysis of cytosine methylation by WGBS revealed higher levels of methylation within some phage region III genes than others. Since methylation tends to inhibit gene expression, cytosine methylation might lower the expression of certain structural ϕ1 genes and the phage endolysin but increase levels of expression of the phage holin and excisionase proteins in the lytic cycle. Our WGBS analysis of cytosine methylation did not reveal any specific methylated motifs. This is reminiscent of a previous study which showed that the Gp56 cytosine methylase of B. thailandensis, which is also part of a phage region, specifically modifies the phage episome but not the prophage (47). We speculate that gp51 could modify a specific sequence within the phage region.

Our null mutant exhibited increased transcription of genes involved in the SOS response, which triggers phage induction (35, 48). We used TEM to investigate the presence of phages and phage-like structures in the supernatants of strain H111 and the RM null mutant. We detected an increase in phage-like structures in null mutant supernatants, confirming our RNA-seq data. It should be noted that this occurred despite the loss of phage region III, which encodes ϕH111-1, the only confirmed active bacteriophage of strain H111. In addition, we found a 2.2-fold increase in MV production in the RM system null mutant compared to the H111 wild type. Toyofuku and colleagues recently showed that an increase in prophage-encoded endolysin triggers MV formation in P. aeruginosa and Bacillus subtilis (41, 49).

The RM null mutant was sequenced to verify loss of all methylated motifs. We confirmed the loss of all m6A and m4C modifications previously detected in the wild type. However, compared to the wild-type strain, the abundance of unassigned modified bases was fourfold higher in the null mutant. DNA methylation slows base incorporation in SMRT sequencing, but so does DNA damage. The increase in unassigned modifications is likely to represent increased nicks in the genome sequence due to DNA damage. Various genes involved in replication, recombination, and repair, especially in the SOS response, such as lexA, recA, genes coding for DNA polymerase IV (which acts during the SOS response) and an exonuclease subunit A (part of the UvrABC DNA repair system), were found to be upregulated in the null mutant compared to strain H111, suggesting that the RM null mutant might be subject to a higher level of DNA damage. In E. coli, Dam⁻ mutants are subject to increased transcription of the SOS regulon. This effect is thought to occur indirectly; in the absence of Dam methylase-mediated strand discrimination, the mismatch repair system (MutHLS) causes double-stranded DNA (dsDNA) breaks, leading to SOS regulon induction (50, 51).

We observed several phenotypic changes in the RM null mutant that are known to be associated with sigma factor RpoN, and the gene encoding RpoN showed reduced expression in our RNA-seq analysis. Pyochelin production was increased, consistent with other studies which have shown that an increase in RpoN leads to a reduction in pyochelin production (38, 40). The RM null mutant was impaired in biofilm formation. In B. cenocepacia K56-2, the RpoN sigma factor is required for bacterial motility and biofilm formation (52). However, analysis of recA mutant strains suggested that the reduction in biofilm production observed in the null mutant resulted from an increase in the SOS response.

Given that methylation is a key determining factor of phenotype in eukaryotes, it is perhaps surprising that so few of the phenotypic tests carried out here were influenced by RM. Where phenotypic changes were observed, they were generally moderate. The more extreme differences were associated with the known key roles of RM systems in genome stability and protection against incoming DNA. Inspection of the upstream regions of the most differentially regulated genes in our RNA-seq analysis revealed only one promising candidate for direct alteration of promoter activity by methylation. Using a transcriptional reporter gene fusion, we observed that the methylated PpchE promoter was less active, suggesting that methylation can be involved in fine-tuning of gene expression.

A recent study by Vandenbussche and colleagues aimed to investigate the effects of methylation in B. cenocepacia strains K56-2 and J2315 (26). In this study, the two core methylases were individually deleted, and the resultant mutants were examined for changes in motility and biofilm formation. The authors observed a decrease in motility and biofilm formation in the TIIMc2 mutants. This suggests that the core TIIM is responsible for the reductions in these phenotypes observed in our RM null mutant.

RM system acquisition occurred early in bacterial evolution (53). The first investigations of RM systems demonstrated their important role in defense against foreign DNA by allowing self/nonself discrimination (reviewed in reference 54). Conjugative transfer experiments to move the small broad-host-range plasmid pBBR1MCS and megaplasmid pC3 between B. cenocepacia K56-2 and the RM system null mutant verified that RM systems do indeed play an important role in protecting the B. cenocepacia H111 genome against incoming DNA, since in an RM null recipient, transfer efficiency increased 7- and 24-fold, respectively, while transfer of pBBR1MCS from B. vietnamiensis LMG10929 was 60-fold more efficient into the RM null derivative than the control (see Fig. S6 in the supplemental material).

The spontaneous loss of pC3 that occurred during the construction of the RM null mutant suggested that the deletion of RM components might have reduced genome stability. To quantitatively evaluate pC3 stability, the frequency of pC3 loss was determined. This confirmed that RM components play a central role in the maintenance of genome integrity in B. cenocepacia H111. The loss of pC3 occurred so frequently in the RM null mutant that separation of the null strain into pC3-deficient and -positive strains was also observed through colony morphology. In E. coli Dam methylase mutants, the timing of chromosome replication is disturbed, resulting in various numbers of replicons in daughter cells (55). Since there was little effect of the deletion of RM components and systems on H111 growth, however, we conclude that the effect on the stability of the essential replicons must be slight.

Our work confirms the role of RM in genome protection and stability and suggests involvement in phenotypes such as biofilm formation, siderophore production, motility, and prophage induction.

MATERIALS AND METHODS

Bioinformatic analysis of RM components.

A file containing the amino acid sequences of all RM components, both putative and experimentally proven, logged within the REBASE database was kindly provided by REBASE. An initial analysis was carried out on all Burkholderia and Paraburkholderia members within this file (due to the strain nomenclature used within REBASE, this meant that all Burkholderia sensu lato strains in the database were included). The translated sequences of the H111 RM genes (gp10, I35_2397; gp51, I35_2438; TIVRc1, I35_3250; TIR, I35_3252; TIM, I35_3254; TIIIR, I35_1826; TIIIM, I35_1825; TIIMc2, I35_2582; TIVRc2, I35_ 1041) were used as BLASTP queries to find homologous components within the REBASE file using CLC Main Workbench v8. The percentage identity (ID) was calculated as the number of identical residues between the query and the match, as a percentage of the number of residues present in the H111 query sequence (excluding stop codon). This is shown in Table S4 in the supplemental material.

A further analysis was carried out by downloading the genomes of a representative strain from each of the Burkholderia sensu lato species represented in REBASE. Where possible, a commonly studied type strain was chosen. Three B. cenocepacia strains were chosen to illustrate the diversity among some of the RM components within the species. For Burkholderia fungorum, no strain designation was listed in REBASE and strain ATCC BAA-463 was chosen. The same query sequences from the first analysis were used in a tBLASTN search against these genome sequences. Percentage identity was calculated as described above. The phylogenetic tree was generated by concatenating the essential and highly conserved gyrB and rpoD genes from each species, aligning using CLC Main Workbench v8, trimming where less than 50% of the sequences aligned with TrimaAl (Phylemon2), and then generating a phylogenetic tree using the neighbor joining method with CLC Main Workbench v8. The genome of Ralstonia pickettii 12D was also included as an outgroup for the phylogenetic analysis.

Bacterial strains and media.

All bacterial strains and plasmids used in this study are listed in Tables 2 and 3, respectively. Unless otherwise stated, strains were grown aerobically in Luria-Bertani (Lennox) broth (Difco) at 37°C. When required, media were supplemented with antibiotics at appropriate concentrations as follows: chloramphenicol, 25 μg ml⁻¹ (E. coli) and 50 μg ml⁻¹ (Bcc); trimethoprim, 25 μg ml⁻¹ (E. coli) and 50 μg ml⁻¹ (Bcc); gentamicin, 20 μg ml⁻¹ (E. coli and Bcc); rifampicin, 50 μg ml⁻¹ (Bcc). M9 medium containing uracil as the nitrogen source, as described previously (7), was used for differentiation between H111 and pC3 cured derivatives.

TABLE 3.

Plasmids used in this study

Plasmid	Characteristics	Reference or source
pRK2013	Helper plasmid; RK2 derivative, mob+ tra+ ori ColE1 Kmr	77
pSC200	PrhaB rhamnose-inducible promoter, rhaR, rhaS and dhfr cassette; Tpr	60
pSHAFT2	Broad-host-range suicide plasmid, mobilizable for conjugation; Cmr	78
pDAIGm-SceI	pDA17 plasmid carrying the I-sceI nuclease gene; Gmr	31
pGPI-SceI	Suicide plasmid with oriR6K, mob+, I-SceI restriction site; Tpr	Elisabeth Steiner, laboratory collection
pGPI-FR1-3	pGPI-SceI plasmid with fused regions flanking the type I RM system on C1	This study
pGPI-FR1-4	pGPI-SceI plasmid with fused regions flanking the type I RM system and type IV RE on C1	This study
pGPI-III	pGPI-SceI plasmid with fused regions flanking the type III RM system C1	This study
pGPI-IVC2	pGPI-SceI plasmid with fused regions flanking the type IV RE system C2	This study
pGPI-OMC2	pGPI-SceI plasmid with fused regions flanking the type II MT on C2	This study
pSHAFT2-gabD	pSHAFT2 bearing an amplified fragment of H111 pC3	8
pSHAFT2-araJ	pSHAFT2 bearing an amplified fragment of H111 pC3	8
pEX18Gm-pMT- TpqueF	pEX18 bearing dhfrII under the control of the LacI-controlled pMT promoter	8
pSHAFT2-nonconp J23109-lacI-aacI	pSHAFT2 bearing lacI under the control of promoter pJ23109, flanked by regions amplified from pC3	8
pSU11	lacZ transcriptional fusion reporter vector; Gmr	79

Molecular techniques.

Chromosomal DNA isolation was performed using the Wizard Genomic DNA purification kit from Promega, with minor modifications to the manufacturer’s protocol as follows. According to how much genomic DNA (gDNA) was required, a different amount of bacterial overnight culture was collected. After the cells were harvested by centrifugation, the pellet was resuspended in TNE buffer (10 mM Tris-HCl, 200 mM NaCl, 100 mM EDTA [pH 8]) and incubated on ice for 20 to 30 min. The cell suspension was collected by centrifugation, and the isolation protocol was carried out per the manufacturer’s instructions. Plasmid preparation was routinely carried out using the Qiagen miniprep kit. DNA prepared by PCR amplification or restriction digestion was purified using the Qiagen PCR purification kit. Molecular methods were carried out as described by Sambrook et al. (56). DNA fragments were amplified using either GoTaq DNA polymerase (Promega) for diagnostic purposes or the proofreading Phusion high-fidelity DNA polymerase (New England BioLabs [NEB]) to amplify fragments for use in cloning.

Conjugal transfer of plasmids.

Bacterial conjugations were used to introduce plasmids into Bcc strains, using a filter mating technique (57). A helper strain (MC1061/pRK2013) was used to provide the tra genes. Conjugations were carried out on LB plates for approximately 16 h using saturated overnight cultures. Pseudomonas isolation agar (PIA; Difco), supplemented with antibiotics as appropriate, was used for selection.

PacBio SMRT methylome sequencing.

Genomic DNA was extracted using the Wizard kit (Promega), as stated above, and sequenced using single-molecule, real-time (SMRT) sequencing on the PacBio RS II, by the Functional Genomics Center Zürich (FGCZ) (University of Zurich). The raw data were analyzed using the PacBio SMRT Portal. The sequenced reads were mapped to the reference sequence to allow detection of specific methylation patterns using the “Base Modification and Motif Analysis” protocol.

Whole-genome bisulfite sequencing (WGBS).

The genomic DNA from overnight cultures of Burkholderia cenocepacia H111pC3⁺ and its mutant derivatives NullpC3⁺ and SMgp10 was extracted using the GenElute bacterial genomic DNA kit (catalog no. NA2100-1KT). The gDNA was first fragmented with the Covaris LE220 system (Covaris, Woburn, MA, USA) per the manufacturer’s instructions to achieve an average fragment size of 350 bp. The sizes of the fragments were validated with the Agilent 4200 TapeStation system D5000 ScreenTape (Agilent, Santa Clara, CA, USA). Unmethylated lambda DNA (dam⁻ dcm⁻) (catalog no. SD0021; Thermo Scientific) was also sheared to ∼350-bp fragments and spiked into each sample at 0.1% (wt/wt). The EZ DNA Methylation-Gold kit (catalog no. D5005; Zymo) was used for the bisulfite conversion. The library was analyzed with the Accel-NGS Methyl-Seq DNA library kit with associated indexing kit A (catalog no. 30024 and 36024; Swift Biosciences) according to the manufacturer’s instructions. The bead-based clean-ups of the library were performed with AMPure XP beads (catalog no. A63880; Beckman Coulter). The libraries were further quantified and sequenced with Illumina Novaseq 6000 with a paired-end 150-bp read configuration at the FGCZ.

Methylation visualization.

Prior to visualization, the abundance of modifications within each 10,000-bp length of DNA (window size) was calculated using ad hoc Python scripts. The visualization of the detected modifications per window across the chromosomes was performed using the circlize package in R within the Rstudio interface version 1.1.463 (58). We screened the genome for the presence of motifs in and around each replicon’s oriC, identified using the DoriC database, oriFinder, and DNAplotter (59).

Verification of prophage region III loss in the RM component QM mutant.

To investigate whether prophage region III was absent from the QM mutant, PCR was performed using primers designed to amplify genes within prophage region III encoding endolysin (gp12, I35_2399), holin (gp13, I35_4480), and the tail sheath protein (gp20, I35_2407). These genes could not be amplified from QM, but amplification was achieved from H111, and from each of the intermediate mutants leading up to QM. This suggested that phage region III had been lost in the construction of QM, and not in a previous step. We were able to amplify the genes flanking phage region III, suggesting that the phage had excised cleanly from the genome. Clean loss of the prophage region III was later confirmed by sequence analysis of the null mutant (see Fig. S1B in the supplemental material).

pC3 mobilization and curing.

The mobilization of pC3 was enabled through insertion of an oriT via single crossover insertion of a suicide vector bearing an oriT (either pSHAFT2-gabD or pSHAFT2-araJ), allowing conjugative transfer, as previously described (7). To delete pC3 from several Bcc strains, a straightforward replicon curing approach was performed using a constructed pC3 minireplicon called pMiniC3, bearing the single-copy pC3 origin of replication, as described previously (7).

pC3 stability assay.

Assessment of pC3 stability was carried out as previously described (8), with modifications to the protocol. Briefly, two specifically generated suicide vectors were used to construct strains to determine the pC3 stability. pEX18Gm-pMT-TpqueF carrying trimethoprim (Tp) resistance (dhfrII) under the regulation of a modified lac promoter was integrated into C1 via double homologous recombination. The gene encoding the repressor LacI was introduced into pC3 through a double crossover using pSHAFT2-nonconpJ23109-lacI-aacI. Strains were grown in IST medium for 24 h at 37°C, and the cell count was determined by plating dilutions at intervals. Where pC3 was present in the cell, the expression of the Tp resistance gene was repressed by LacI. When pC3 was lost, Tp was expressed, resulting in colonies on IST plates containing Tp (25 μg ml⁻¹). To test for spontaneous Tp resistance, colonies were replica plated on the pC3-selective medium M9ura (7) supplemented with Tp at 25 μg ml⁻¹.

Transfer efficiency test.

To test for pC3 transfer efficiency between Bcc members, pC3 was mobilized by integration of pSHAFT2, which carries an oriT allowing conjugative transfer and a chloramphenicol resistance marker, in the donor strains (B. cenocepacia K56-2 and B. vietnamiensis LMG 10929). The pC3 megaplasmid was cured from the recipient strains (wild-type B. cenocepacia H111 and RM null mutant NullpC3⁺) by the method of Agnoli and colleagues (7), and spontaneous rifampicin derivatives of the strains were selected by spreading 100 μl of the overnight culture on LB plates supplemented with 100 μg ml⁻¹ rifampicin (Rif). Resistant colonies were restreaked on LB plates supplemented with rifampicin. Since the donor strains do not carry the tra genes required for formation of the sex pilus, a helper strain was used (MC1061/pRK2013), in a triparental mating. Dilution series were plated on PIA plates supplemented with Rif to calculate the total number of recipients and exconjugants. Depending on the strain, either 100 μl of an undiluted suspension or a dilution was plated on PIA plates supplemented with 200 μg ml⁻¹chloramphenicol (Cm) and 50 μg ml⁻¹ Rif to calculate the total number of exconjugants. Transfer efficiency was defined as the total number of exconjugants/total number of recipients and exconjugants.

Construction of conditional mutants.

Conditional mutants were generated using the vector pSC200, which upon single crossover recombination with the genome separates a target gene from its native promoter, putting it under the control of the rhamnose-inducible PrhaB promoter (60). Primers and restriction enzymes used have been detailed in Table S6. Conditional mutants were selected on PIA plates supplemented with 0.2% rhamnose and trimethoprim (50 μg ml⁻¹). To test for essentiality, conditional mutants were grown overnight in LB medium supplemented with 0.2% rhamnose. Five microliters from each sample of a dilution series was spotted on PIA medium supplemented with either 0.5% glucose or rhamnose for each strain. The plates were grown for 24 h at 37°C.

Construction of targeted unmarked gene deletions.

To construct markerless gene deletions, a protocol modified from that previously described by Flannagan was used (31). Briefly, regions of homology of approximately 500 bp in size flanking the gene to be deleted were amplified using Phusion DNA polymerase. Both fragments, as well as the vector pGPI-SceI, were digested with the chosen restriction enzymes. A tripartite ligation was performed, the plasmid was transformed into electrocompetent E. coli SY327λpir and spread on selective LB plates. Positive clones were confirmed by colony PCR and sequence analysis using appropriate primers (Table S5), and the plasmid was introduced into B. cenocepacia H111 by triparental mating. Exconjugants were selected on PIA containing Tp and confirmed by PCR. A second homologous recombination was instigated by introducing vector pDAIGm-SceI into the recipient. Positive clones were selected on PIA plates containing gentamicin, verified by colony PCR, and later colony purified by streaking on PIA plates without antibiotics. Primers used for the amplification and for the final deletion verification are stated in Table S6 and were designed using the H111 GenBank files (accession numbers HG938370, HG938371, and HG938372).

The RM null mutant was constructed by the sequential deletion of each RM locus, resulting in a series of intermediate mutants, in addition to the final RM null mutant. The order of construction was as follows: (i) deletion of the TIRM genes (I35_3251-I35_3254, 7,054 bp), to give mutant TI; (ii) deletion of theTIVRc1 genes (I35_3250, 963 bp), to give strain DM; (iii) deletion of the TIIIRM genes (I35_3273, I35_3274, 5,051 bp), to give TM; (iv) deletion of the TIVRc2 gene (I35_ 5374, 729 bp), to give QM. Investigation of QM showed that prophage III had been lost from the genome, leaving only one RM component gene, I35_4914 (encoding TIIMc2). This was deleted to give NullpC3⁻. Upon discovery of the spontaneous loss of pC3 that occurred during the construction of NullpC3⁻, pC3 was moved back into the strain, as described previously (9), to give NullpC3⁺. Finally, an unmarked RM null mutant strain was constructed by repeating the deletion of the TIIMc2 gene (I35_4914) of QM, with care taken to select a pC3-containing clone. This strain was designated “Null.” The primers and restriction enzymes used for each deletion stage are indicated in Table S6.

Preparation of samples for RNA-seq analysis.

Overnight cultures of the strains of interest were used to inoculate 50 ml LB broth with a starting optical density at 600 nm (OD₆₀₀) of 0.01 and shaken at 220 rpm under aerobic conditions at 37°C until an OD₆₀₀ of 1 was reached. The culture was prepared, and total RNA extraction was carried out as detailed in reference 61. To remove the remaining DNA, samples were treated with RQ1 RNase-free DNase I (Promega) and purified using the RNAeasy minikit from Qiagen, according to the manufacturer’s guidelines. RNA quality was then checked with the RNA Nano Chip (Agilent 2100 Bioanalyzer; RNA integrity number of >8), and 150 ng of total RNA were used for cDNA library construction. The Ovation Complete Prokaryotic RNA-Seq DR multiplex system from NuGEN (NuGEN, San Carlos, CA, USA) was used to construct a strand-specific RNA-Seq library. This system uses insert-dependent adaptor cleavage (InDAC) technology to remove rRNA. The cDNA library was analyzed by capillary electrophoresis using a DNA chip from Agilent (Agilent High Sensitivity D1000 ScreenTape system). The prepared libraries were sequenced with the Illumina platform (single-end, HiSeq2500 instrument), by the Functional Genomics Center Zürich (FGCZ) (University of Zurich). Between 6.2 and 9.5 million unique reads were obtained and mapped to the B. cenocepacia H111 genome using CLC Genomics Workbench v7.0 (Qiagen CLC bio). The top 500 genes that showed the most significant changes in their expression (P value of ≤0.01 and absolute log₂ fold change of ≥0.5) were taken for further analysis, and statistical analysis was performed using the DESeq R-package v1.26 (62). The RNA-seq raw data files of wild type and mutant are accessible through the GEO Series accession number GSE147038.

Motility assays.

Swarming motility was determined on a semisolid nutrient medium (0.4% agar, 0.5% peptone, and 0.3% yeast extract) supplemented with 0.5% glucose. Overnight cultures were normalized to an OD₆₀₀ of 1, and 5 μl of the bacterial culture was spotted at the center of the plate. After 40 h of incubation at 30°C, plates were documented photographically. For analysis of the recA conditional mutants, the diameter of the region of growth was measured. Swimming motility was measured on a semisolid nutrient medium (0.3% agar, 0.3% peptone, and 0.3% yeast extract) supplemented with 0.5% glucose. The plates were inoculated by touching the agar surface with a toothpick dipped into a bacterial suspension with an OD₆₀₀ of 1 and incubated for 40 h at 30°C. When measuring the diameter of swimming or swarming, the mean diameter of two cross sections through the point of inoculation was used. On swimming plates in which localized swarming had occurred, such areas were not measured.

Colony morphology.

Colony morphology was observed on NYG agar plates (1.5% agar, 0.5% peptone, 0.3% yeast extract, and 2.0% [wt/vol] glycerol). Five microliters of an overnight bacterial culture was spotted on the plates and incubated for 3 days at 37°C, followed by a minimum of 2 days at room temperature (RT).

Biofilm formation assay.

Biofilm formation was quantified in 96-well microtiter plates as described previously (63), with the following modifications: the medium used in the 96-well plates was AB with glucose as the carbon source, supplemented with Cm to select against cells that had lost pC3 (which bore the cat gene). Incubation of the plates was carried out at 28°C for 48 h. The biofilm index (BI) was calculated as followed: BI = OD₅₇₀/OD₅₅₀ × 100 (64).

Pellicle formation assay.

Pellicle formation was tested in NYG broth (0.5% peptone, 0.3% yeast extract, and 2.0% [wt/vol] glycerol), supplemented with 0.5% glucose and Cm. The medium was inoculated 1:100 from a bacterial overnight culture and incubated at RT for a minimum of 5 days without shaking, in a capped tube to avoid evaporation.

Protease activity.

Bacteria were assayed for proteolytic activity using the method of Safarik (65) with modifications to the protocol as described by Schmid and colleagues (66). Cultures were grown at 37°C in NYG medium containing 0.5% glucose and supplemented with Cm.

Use of fluorescence microscopy to examine cell morphology.

Flasks with 20 ml LB broth were inoculated in duplicate with bacterial overnight cultures of strains NullpC3⁺ and H111pC3⁺ to a starting OD₆₀₀ of 0.01 and incubated at 37°C with shaking. Samples were taken at exponential phase (after 4-h growth) and stationary phase (after 16-h growth), and the plasma membrane was stained with FM 4-64 from Life Technologies (100 μg ml⁻¹). Cells were observed with an epifluorescence Leica DM6000 B or a DCM 5500 Q research microscope with a ×1,000 magnification.

Phage visualization using transmission electron microscopy.

Burkholderia cenocepacia H111 and the RM null mutant (newNull) was cultured in 10 ml LB medium at 37°C overnight. After centrifugation for 10 min at 5,000 rpm, the supernatant was collected and filter sterilized using a 0.22-μm-pore-size hydrophilic polyethersulfone filter (Merck Millipore, Germany). The cell-free supernatant was then ultracentrifuged at 150,000 × g for 1 h. The pellet was resuspended in 50 μl phosphate-buffered saline (PBS) for visual phage detection. Phages or phage-like structures were absorbed on glow-discharged Formvar-coated 300-mesh copper grids and negatively stained with 1% uranyl acetate for visualization using transmission electron microscopy (TEM).

MV isolation and quantification.

The wild-type H111 strain and the RM null mutant, Null, were grown overnight and used to inoculate flasks containing 20 ml LB broth to a starting OD₆₀₀ of 0.02. Cultures were shaken for 24 h at 37°C. The isolation and quantification of the membrane vesicles were performed as described by Turnbull and colleagues (41). Briefly, 10 ml of each bacterial suspension was spun for 10 min at 5,000 rpm (4,472 relative centrifugal force [rcf]) at 4°C, and the supernatant was collected and filter sterilized using a 0.22-μm filter. After the supernatants were ultracentrifuged at 150,000 × g for 1 h, each pellet containing membrane vesicles (MVs) was resuspended in 100 μl PBS buffer. For quantification, MVs were stained with the membrane-binding dye FM1-43 (Life Technologies, USA), and the fluorescence intensity (510-nm excitation/626-nm emission) was measured using a MWGt Sirius HT microplate reader from BioTek Instruments GmbH.

Reporter construction and assay.

Reporter vectors were constructed as follows. The region upstream of pchE was amplified using primers ppchEXhoF and ppchEHindR (to obtain the WT region) and ppchEXhoF and Rmethmut1 (to obtain the altered version). These products were cloned into pSU11 using XhoI and HindIII. The insert sequences were confirmed by sequencing with primers pSU11checkF and pSU11checkR (sequencing carried out by Microsynth AG).

β-Galactosidase assay was carried out as described by Miller (67), with alterations to the method as described in reference 68. Strains were assayed at an OD of 0.7 to 0.8 in M9 minimal salts medium containing 100 μg ml⁻¹ 2,2-dipyridyl, 25 μg ml⁻¹Cm, and 20 μg ml⁻¹ gentamicin (Gm). Activities were calculated in Miller units, using the formula (1,000 × OD₄₂₀)/(time × V × OD₆₀₀) where V is volume (in milliliters) and time is shown in minutes.

The expected presence or absence of the methylated base in the reporter vectors was confirmed by Sanger sequencing of the reporter plasmids isolated from strain H111pC3⁺, bearing in mind the effects of methylation on electropherogram peak height, as reported in reference 69.

Pyochelin extraction and analysis.

Strains were grown to saturation in 100 ml iron-free succinate (IFS) medium at 37°C for 40 to 43 h. The OD₆₀₀ of the two strains was determined and found to be equal. Cells were then collected by centrifugation for 20 min at 5,000 rpm at 4°C, and the supernatant was sterile filtered, followed by acidification with 1 M HCl to a pH of 1.5 to 2 and extraction by adding 0.4 volumes of ethyl acetate. The upper ethyl acetate phase was then collected and vacuum dried using the Rotavapor RE (Büchi) until the amount was concentrated to a total volume of about 5 ml. This concentrate was then distributed into Eppendorf tubes and completely desiccated using the Eppendorf Concentrator 5301. The residue was then resuspended in 100 μl methanol, and 1 μl of each sample was analyzed by thin-layer chromatography (TLC) using silica gel 60 F254 (Merck Millipore, Germany) with chloroform-acetic acid-ethanol (90:5:10 [vol/vol/vol]) as the developing solvent. The TLC plate was then quickly dipped into 100 mM FeCl₃ to visualize the purified pyochelins, which were visible as brown areas on the TLC plate.

Data availability.

The genomic data are available under NCBI BioProject number PRJNA609037 and can be accessed at https://www.ncbi.nlm.nih.gov/bioproject?term=PRJNA609037&cmd=DetailsSearch. The raw reads of the sequenced genomic DNA, the motifs, and modification supplemental files are deposited in the SRA and can be downloaded through the following link https://www.ncbi.nlm.nih.gov/bioproject/PRJNA609037. The WGBS data are available under the NCBI BioProject number PRJNA663836.