BfmR regulates A. baumannii pellicle formation
The pellicle is characterized as a form of biofilm that is floating on the surface of culture media and allows bacteria to acquire favorable ecological niche and directly access high concentrations of oxygen and nutrients from the air and liquid, respectively [13]. A. baumannii BfmR is responsible for the up-regulation of the csuA/BABCDE operon leading to the biofilm formation on solid surfaces [3]. It has also been determined that the CsuA/B pilin is the most abundant component of A. baumannii pellicle [14]. Therefore, we were interested whether the BfmR is involved in the manifestation of this phenotype.
From the collection of clinical A. baumannii isolates, characterized previously [15], the isolate V15, which showed a pellicle forming phenotype, has been selected. The deletion of ΔbfmRS operon was generated as described in the Methods. The decision to obtain a mutant with the deletion of the whole bfmRS operon was based on the previously published results indicating that the sensor kinase BfmS acts as a negative regulator of BfmR, and that only bfmR can fully complement whole ΔbfmRS mutant to WT levels [9]. The deletion was confirmed by sequencing and by performing qPCR analysis of bfmR and bfmS genes using total RNA. For complementation experiments, the plasmids carrying the genes encoding BfmR (pbfmR), BfmS (pbfmS), or both proteins (pbfmRS) were constructed and introduced into A. baumannii V15 as described in the Methods.
We then performed the pellicle formation assay by growing the strains in TSB media under the stationary conditions at 30 °C for 30 h as described in the Methods. These growth conditions were previously suggested to generally promote pellicle formation [13, 16]. Compared to the WT strain, which formed a thick and uniform pellicle morphology (Fig. 1a), the ability of ΔbfmRS mutant to develop a pellicle was impaired as only some biomass (white structures) was located near the walls of the wells (Fig. 1a and b). The inability of the mutant to form a pellicle was fully complemented with the bfmR allele or the whole bfmRS operon, when supplemented in trans (Fig. 1a and b). It must be noted that due to the toxicity of bfmRS and bfmR constructs with the native upstream sequences, we used plasmids with a leaky Ptac promoter (transcripts were observed without additional supplementation of IPTG), which allowed the basal expression of bfmR at the level comparable to that of WT (approximately 2.86 ± 0.89 fold up-regulation compared to WT) and did not interfere with the growth of the strains. Finally, we observed no effect of the wild-type bfmS allele on the restoration of the pellicle phenotype in ΔbfmRS mutant, even using induction with IPTG up to 0.1 mM (Additional file 3: Figure S1a). These results show that BfmR is responsible for the pellicle phenotype manifestation, also they are consistent with the previously published data indicating that ΔbfmRS mutants may be complemented solely by the bfmR construct [9].
Hcp secretion into the culture media is regulated by the BfmR
Pellicle formation requires various secreted proteins, polysacharides, and/or DNA to stabilize the structure [13]. Currently, only a few A. baumannii proteins have been linked to a pellicle formation [14]. However, there was an observation that during this process, the expression of multiple virulence factors changes [17]. To investigate, whether BfmR is involved in the regulation of pellicle phenotype-associated components, we first compared the electrophoretic profiles of precipitated total protein fractions from culture media of the WT and ΔbfmRS strains.
As can be seen in Fig. 1c, SDS-PAGE analysis of proteins precipitated from the culture media, showed a single band, migrating as a ~ 18 kDa entity, which was significantly reduced in the ΔbfmRS mutant. The secretion of the protein was restored after complementation with either bfmRS or bfmR alleles (Fig. 1c), correlating with the restoration of the pellicle phenotype. The band was identified by mass spectrometry as A. baumannii Hcp protein. The protein is a structural component of the bacterial Type VI secretion system (T6SS), which together with the additional proteins assembles into a needle-like apparatus that is used to puncture adjacent cells and to deliver effectors (toxins) into target cell [18, 19].
To confirm that the secreted protein was indeed Hcp, we generated Δhcp deletion in A. baumannii V15 strain, which resulted in the loss of Hcp secretion into culture media (Fig. 1d). The secretion was readily complemented with a copy of hcp gene cloned under the inducible promoter in the plasmid phcp, when induced with 0.1 mM IPTG (Fig. 1d). The complementation under inducing conditions only, could be explained by the fact that before being secreted, the Hcp must assemble into tubular structure made from multiple copies of Hcp monomers [19].
The examination of the hcp-specific mRNA levels in the ΔbfmRS mutant showed an approximately five-fold reduction, when compared to the WT strain (Fig. 1e). The introduction of the plasmid pbfmR resulted in the restoration of transcription level. In parallel to the examination of hcp gene transcription, we assessed the transcription of tssM gene. The latter codes for the subunit of the membrane-anchoring complex, which is also essential for the assembly of T6SS apparatus [20]. As can be seen from the results presented in Fig. 1f, tssM mRNA levels in the ΔbfmRS were decreased four-fold, when compared to the WT. This indicates that the loss of bfmR might lead to the down-regulation of the whole T6SS system, resulting in the reduced secretion of Hcp into culture media.
The abundance of Hcp in culture media and the correlation between the Hcp secretion and the formation of pellicle exhibited by the WT strain, prompted us to test whether Hcp is required for the pellicle formation. The Hcp, secreted into media could be embedded into pellicle matrix, potentially reinforcing the structure. However, our results showed the deletion of hcp did not interfere with the manifestation of the phenotype (Additional file 3: Figure S1b). This shows that Hcp is not required for the pellicle formation of A. baumannii.
Loss of bfmRS does not affect T6SS-mediated inter-genus killing
The findings above suggest that the down-regulation of T6SS might impact the killing phenotype of A. baumannii as it is known that Hcp secretion is the indication of a functional T6SS [18]. Previously, it has been demonstrated that A. baumannii is able to eliminate competing bacteria in a T6SS-dependent manner [21,22,23,24]. Therefore, we performed competition assays using E. coli MC4100 strain as a prey. Remarkably, while A. baumannii V15 was able to significantly reduce E. coli MC4100 numbers by 50–250-fold, the ΔbfmRS strain did not display any impairment in the killing phenotype (Fig. 2a). The phenotype remained mainly unchanged in the ΔbfmRS strain complemented with either bfmRS, or bfmR alleles (Fig. 2a). To confirm that the observed killing phenotype against E. coli was due to the function of T6SS, we investigated killing capacity of V15 Δhcp and the double mutant ΔbfmRSΔhcp. The results showed that both mutants exhibited reduction in killing of E. coli MC4100 (Fig. 2b and Additional file 4: Figure S2a). The phenotype of Δhcp and ΔbfmRSΔhcp mutants was readily complemented with the wild-type hcp allele (phcp) under inducing (0.1 mM IPTG) conditions (Fig. 2b). We also evaluated whether there is a difference in killing efficiency against clinical strains of Pseudomonas aeruginosa (P16) and Klebsiella pneumoniae (K39). As can be seen in Additional file 5: Figure S3a-d, both strains can be killed via T6SS of A. baumannii V15. However, the results also show that the loss of bfmRS, apparently, does not influence the reduction of the killing efficiency of A. baumannii. These data indicate that despite clearly affecting the expression of the T6SS apparatus and significantly impairing the secretion of Hcp into the supernatant, the BfmRS system does not affect A. baumannii T6SS-mediated killing of E. coli, P. aeruginosa or K. pneumoniae.
BfmR regulates T6SS-independent killing mechanism against closely related species
Next, we evaluated A. baumannii aggressiveness against more closely related species. For this purpose, we used Acinetobacter baylyi ADP1 strain as a prey. As can be seen in Fig. 2c, ADP1 strain was significantly (approximately 70–200-fold) out-competed by A. baumannii V15 and the killing was T6SS-dependent as the Δhcp mutant displayed significantly reduced killing efficiency (Fig. 2c). The killing phenotype of the mutant was readily restored to the WT level with the hcp allele under inducing conditions (Fig. 2c). Remarkably, in contrast to the results obtained with E. coli MC4100 strain, the double mutant ΔbfmRSΔhcp, lacking an active T6SS, still significantly reduced ADP1 numbers at the efficiency comparable to that of WT strain (Fig. 2c). Interestingly, we observed that the ΔbfmRS mutant was able to significantly reduce ADP1 numbers as well, and displayed even more aggressive killing phenotype than the WT (approximately 10-fold) (Fig. 2d). The observed killing phenotype of ΔbfmRS and ΔbfmRSΔhcp mutants could not be complemented with either pbfmR or pbfmRS (Fig. 2d and Additional file 4: Figure S2b). These results suggest, that bfmRS deletion leads to the activation of T6SS-independent killing mechanism that is effective against A. baylyi ADP1 but not E. coli MC4100.
BfmR negatively regulates contact-dependent inhibition system of A. baumannii
We hypothesized that the observed killing of A. baylyi ADP1 but not E. coli MC4100 could be explained by the activation of currently poorly defined A. baumannii contact-dependent inhibition mechanism, which requires its receptor on the target cell and is classified as a type of Type V Secretion System (T5SS) and was shown to be functional in A. baumannii [25, 26].
CDI systems are composed of three components belonging to the cdiBAI gene cluster. The first two genes (cdiB and cdiA) encode a two-partner secretion system, which allows a large CdiA hemagglutinin-repeat protein to be displayed on the surface of bacterial cell. The last gene (cdiI) encodes an immunity protein, which binds and neutralizes the cognate toxin [25]. Previous work indicated that Acinetobacter sp. might contain up to two functional CDI systems [26, 27].
Therefore, based on the previous classification of A. baumannii CDI systems [28], we created a set of 3 primer pairs targeting a rather conserved cdiB genes and managed to identify and sequence the CDI locus of A. baumannii V15 strain (CDIV15). The results indicated that CDIV15 is a type-I CDI system with a CdiA protein, most identical to bau-A1/pit-A3 type CdiA proteins (90%) (Fig. 3a). Other CDIV15 proteins, namely CdiB and CdiI, were 98% and 66% identical to their counterparts in bau-A1 system, respectively (Fig. 3a). It is worth to note that some A. baumannii strains, namely AR_0037 (GenBank accession: MPBX01000005.1/bau-D9 CdiA type [28]), 1295549 (JFXB01000002.1/bau-B2), 426863 (JFYF01000002.1/bau-B2), ATCC19606 (JMRY01000015.1/bau-B2) contained nearly identical immunity proteins in the genome regions, which did not have a CDI system nearby, indicating that these strains are potentially immune to CDIV15.
Having identified that A. baumannii V15 contains an intact cdi locus, we further investigated, whether the observed T6SS-independent killing mechanism is indeed caused by the CDIV15 system. By comparing the mRNA levels of the cdiB gene between A. baumannii V15 ΔbfmRS mutant and WT we observed that the mutant displayed increased transcript levels of cdiB by ~ 6.4-fold. Additionally, complementation of ΔbfmRS mutant with the bfmR allele displayed a ~ 2.5-fold increase in the transcript levels of cdiB, when compared to the WT (Fig. 3b). This indicates that the complemented strain displayed an intermediate transcription level of the CDIV15. This result was consistent with the observed killing phenotype displayed against A. baylyi ADP1 strain, where the aggressiveness level of ΔbfmRS mutant could not be complemented with the bfmR allele to that of WT.
A. baumannii uses CDI to out-compete A. baylyi ADP1
To confirm our observation that the CDIV15 system is activated in the bfmRS mutant, a partial deletion of A. baumannii cdiV15 locus was generated in the WT and ΔbfmRS strains. In the resulting mutants, the cdiB gene was left intact but approximately 90% of the cdiAV15 along with cdiIV15 were deleted. This was due to the size of the genomic region that prevented us from obtaining the whole CDIV15 operon deletion. It is interesting to note, that during the characterization of A. baumannii ΔbfmRSΔcdiV15 and ΔcdiV15 mutants, we have observed the loss of ~ 200 kDa band in the SDS-PAGE gel of precipitated total protein fraction from culture media (Fig. 3c). The band corresponds to the predicted molecular weight of CdiA protein (~ 231 kDa), and its identity was subsequently confirmed by mass spectrometry. We have also noticed that, consistently with the mRNA data (Fig. 3b), the predicted CdiA band in WT was of lower intensity when compared to the ΔbfmRS strain (Fig. 3c). Additionally, ΔbfmRS strain complemented with the pbfmR displayed the predicted CdiA band of intermediate intensity, compared to the WT and ΔbfmRS (Fig. 3d).
We then investigated the aggressiveness of the mutants against A. baylyi ADP1. In addition, a plasmid containing the cdiIV15 allele under the inducible promoter was created and introduced into ADP1 strain. The cdiIV15 codes for the putative immunity gene, which should protect the host from the aggressor, if the latter uses the CDIV15 system for the killing. By performing the killing assays, we found that the deletion of cdiV15 in the ΔbfmRS and ΔbfmRSΔhcp mutants resulted in a low (~ 10-fold) but significant reduction of aggressiveness, when compared to the parent mutant (Fig. 3e). Interestingly, ΔbfmRSΔhcpΔcdiV15 mutant still displayed a killing phenotype (Fig. 3e). The reasons that caused this are currently unknown. However, it could be explained by the activation of a secondary CDI system that we were unable to detect with our PCR screen, as it is known that some Acinetobacter sp. strains contain two active CDI systems [26, 27].
The introduction of the cdiIV15 allele into A. baylyi ADP1 strain reduced the susceptibility to the killing by A. baumannii ΔbfmRSΔhcp and ΔbfmRS, but not by the ΔbfmRSΔcdiV15 and ΔbfmRSΔhcpΔcdiV15 strains (Fig. 3e), further confirming our findings that the ΔbfmRS and ΔbfmRSΔhcp mutants activate T6SS-independent killing mechanism, and show that this phenotype could be attributed to the activation of CDIV15 locus.
CDI-mediated A. baumannii intra-species competition
Bacterial inter-species killing via CDI system was observed only in a few cases and showed a very low efficiency, therefore it was suggested that the primary role of CDI is to differentiate sibling cells from other closely related bacteria from the same species [30, 31]. This could explain a rather modest changes in the killing efficiency that we observed towards A. baylyi ADP1. This prompted us to investigate the CDI-mediated killing phenomenon within the A. baumannii species. For this purpose, using primer pairs targeting cdiB genes of all known A. baumannii CDI systems, we have screened clinical A. baumannii isolates, representing different genotypically related groups (pulsotypes) of strains (n = 15) belonging to international clonal lineage II (IC II) by PCR [32]. The results showed that two clinical A. baumannii strains II-g and II-h did not contain a known cdiB gene. Therefore, these strains were selected for further competition experiments.
As can be seen in Fig. 4a and b, II-h strain was highly susceptible to the killing by A. baumannii V15, while II-g strain displayed a low susceptibility to this phenotype (Fig. 4a and Additional file 6: Figure S4a). The total numbers of the susceptible strain II-h were reduced ~ 107-fold by the ΔbfmRS and ΔbfmRSΔhcp mutants (Fig. 4b). When the competition was performed with the triple mutant ΔbfmRSΔhcpΔcdiV15, the recovery of II-h increased by a factor of 104, compared to ΔbfmRS. The strain containing only functional T6SS (ΔbfmRSΔcdiV15), displayed an intermediate phenotype (Fig. 4b). It is worth to mention that the WT displayed mostly T6SS-dependent killing phenotype (Additional file 6: Figure S4b). These results indicate, that the ΔbfmRS mutant kills II-h strain via both mechanisms – T6SS and CDI, while WT strain uses only T6SS-dependent killing.
Additionally, we investigated, whether the complementation with the bfmR allele resulted in an inhibition of the CDI-mediated killing against II-h. Notably, the results showed that when ΔbfmRS mutant was complemented with either pbfmR or pbfmRS, the recovery numbers of II-h increased by ~ 30-fold (Fig. 4c). Interestingly, when the ΔbfmRSΔhcp mutant was complemented with the same alleles, the reduction of killing phenotype was ~ 500-fold (Fig. 4c). Lastly, as expected, the complementation of either ΔbfmRSΔcdiV15 or ΔbfmRSΔhcpΔcdiV15 with the bfmR or bfmRS alleles did not influence any change in the killing phenotype of the mutants (Fig. 4c). These results were consistent with the observed intermediate transcriptional up-regulation of CDIV15 system in the ΔbfmRS strain complemented with the bfmR, when compared to the WT and ΔbfmRS strains (Fig. 3b). Additionally, bfmR or bfmRS complemented strains ΔbfmRSΔhcp, ΔbfmRSΔcdiV15, containing either functional CDI or T6SS, respectively, displayed an intermediate killing phenotype, compared to the ΔbfmRS and ΔbfmRSΔhcpΔcdiV15 (Fig. 4c). These results confirm observation that BfmR acts as a negative regulator of the CDIV15 system of A. baumannii V15.