The transcriptional response to rhamnolipids
Genome-wide expression profiling is a powerful approach to characterize the response to a certain stimulus, such as the presence of antimicrobial compounds. It has also been used to gain insights into inhibitory mechanisms and to differentiate between different modes of action of novel antibiotics (Hutter et al., 2004; Fischer & Freiberg, 2007; Wecke et al., 2009). We used genome-wide DNA microarray analysis to investigate the response of the model organism B. subtilis to the presence of rhamnolipids, which have been shown to affect cell envelope integrity (Vasileva-Tonkova et al., 2011). B. subtilis was treated with sublethal concentrations (50 μg mL−1) of rhamnolipids, which is sufficient to induce a transcriptional response, but does not impair growth of the culture, as can be demonstrated by concentration-dependent lysis curve experiments (see below and Fig. 3). After 10 min of induction, total RNA was prepared and DNA microarray analysis performed. Expression of 40 loci was ≥fivefold increased by rhamnolipids compared with the mRNA levels of an uninduced culture (Table 3 and Fig. 1a). Almost half of these loci can be assigned to known regulons of TCS or ECF σ factors. The most strongly induced locus was the liaIHGFSR operon (c. 640-fold), which is autoregulated by the LiaRS TCS (Mascher et al., 2004). The first two genes of this locus, liaIH, represent the main targets of LiaRS-dependent signal transduction and liaH encodes a phage-shock protein homolog. The LiaRS TCS is activated by cell wall antibiotics, especially lipid II-interacting compounds, but it does not mediate resistance against most of its inducers (Mascher et al., 2004; Wolf et al., 2010). Strong expression of the lia locus also resulted in significant read-through transcription of the downstream located gerAAABAC operon, which has been observed previously for both B. subtilis and Bacillus licheniformis (Mascher et al., 2003; Wecke et al., 2006).
Figure 1. The transcriptional response to rhamnolipids. (a) Scatter plot of DNA microarray analysis. The average signal intensities for each gene are shown from cells induced with 50 μg mL−1 rhamnolipids for 10 min (y-axis) and uninduced control (x-axis). The pyr operon (▲), pstSCABABB (Δ), des (●) and genes regulated by LiaRS (□), CssRS (♦) and σM (■) are highlighted; all other genes are represented as gray squares. (b) Verification of the transcriptome data by real-time RT-PCR. Real-time RT-PCR was performed as described in 'Materials and methods' with the same RNA as used for DNA microarray analysis. Induction ratios for each gene were calculated based on the uninduced control, as described previously (Talaat et al., 2002). Each value is the average of two microarray hybridizations or real-time RT-PCR experiments, the error bar indicating the standard deviation.
Download figure to PowerPoint
Table 3. Genes significantly induced or repressed by rhamnolipids
|Gene(s)a||Fold changesb||Regulatorsc||Homology, (putative) function, remarks|
|Genes induced ≥fivefold|
|liaIHGFSR||640 ± 501||LiaRS||Phage-shock protein homolog, TCS, unknown|
|htrA||58 ± 15||CssRS||Serine protease|
|htrB||26 ± 6.6||CssRS||Serine protease|
|yuxN||13 ± 4.5|| ||Putative transcriptional regulator, TetR family|
|yqjL||11 ± 1.2||σM||Putative hydrolase|
|pbpE-racX||11 ± 5.2||σW||Penicillin binding protein 4, amino acid racemase|
|yhaSTU||10 ± 4.4|| ||Potassium efflux K+/H+ antiporter|
|yxeI||9.4 ± 2.2|| ||Similar to penicillin amidase|
|yraA||8.6 ± 3.5|| ||Similar to general stress protein|
|yuaE||8.2 ± 2.8|| ||Hypothetical protein with DUF1569 domain|
|yrhHIJ||8.0 ± 3.5||σM, σX, σV||Putative methyltransferase, transcriptional regulator, and reductase|
|sigM-yhdLK||7.8 ± 1.3||σM||ECF σ factor|
|yebC||7.8 ± 1.4||σM||Putative membrane protein|
|ybfO||7.7 ± 3.3||σW||Similar to erythromycin esterase|
|ylbP||7.4 ± 2.1|| ||Putative acetyltransferase, GNAT family|
|phoA||7.2 ± 3.8|| ||Alkaline phosphatase A|
|bcrC||7.2 ± 2.7||σV, σM, σW, σX||Undecaprenyl pyrophosphate phosphatase|
|gabD||7.1 ± 2.7|| ||Succinate-semialdehyde dehydrogenase|
|ywrO||7.0 ± 2.5|| ||Similar to NAD(P)H oxidoreductase|
|ydaH||6.9 ± 2.4||σM||Putative membrane protein with DUF2837 domain|
|opuCABCD||6.8 ± 4.8|| ||Osmoprotection|
|ypbGH||6.7 ± 1.4||σM||Putative phosphoesterase and MecA paralog|
|yceB||6.5 ± 2.8|| ||Putative monooxygenase|
|yvrD||6.5 ± 1.1|| ||Similar to ketoacyl-carrier protein reductase|
|ywaC||6.4 ± 2.8||σW, σM, σV||Similar to GTP-pyrophosphokinase|
|yqjG||6.2 ± 1.7|| ||Similar to lipoprotein SpoIIIJ-like|
|yheCDE||6.0 ± 1.1|| ||Spore coat proteins|
|yhjN||5.9 ± 0.8|| ||Putative membrane-anchored ammonia monooxygenase|
|dhaS||5.8 ± 1.4|| ||Aldehyde dehydrogenase|
|yfjR||5.8 ± 2.8|| ||Similar to 3-hydroxyisobutyrate dehydrogenase|
|radC||5.5 ± 1.8||σM||DNA repair protein|
|yvgP||5.5 ± 0.9|| ||Monovalent cation/H+ antiporter NhaK|
|trxA||5.4 ± 1.2|| ||Thioredoxin, putative monooxygenase|
|nfrA-ywcH||5.2 ± 1.2|| ||NADPH-linked nitro/flavin reductase, similar to monooxygenase|
|gerAABC||5.2 ± 2.2|| ||Germination, downstream of liaIHGFSR|
|ypuA||5.2 ± 0.9||σM, σV||Protein of unknown function with DUF1002 domain|
|ywnJ||5.2 ± 1.5||σM, σW, σX||Putative VanZ-like membrane protein|
|yrbC||5.1 ± 2.5|| ||Uncharacterized conserved protein with DUF28 domain|
|ycgJ||5.0 ± 1.6|| ||Putative methyltransferase|
|yfiBC||5.0 ± 1.4|| ||Similar to ABC transporter|
|Genes repressed ≥fivefold|
|cydABCD||0.19 ± 0.09|| ||Cytochrome bd ubiquinol oxidase|
|rbsRKDACB||0.19 ± 0.07|| ||Ribose transport|
|yuaJ||0.19 ± 0.07|| ||Putative thiamine transporter|
|yonPO||0.18 ± 0.07|| ||Hypothetical proteins (prophage SPβ)|
|narGHJI||0.18 ± 0.05|| ||Nitrate reductase|
|mtbP||0.17 ± 0.06|| ||Modification methylase|
|pur operon||0.17 ± 0.08|| ||Purine biosynthesis|
|yxaI||0.17 ± 0.08|| ||Putative membrane protein|
|yolJ||0.15 ± 0.06|| ||Similar to glycosyltransferase|
|xylAB||0.13 ± 0.02|| ||Xylose metabolism|
|sboAXablA-G||0.13 ± 0.03|| ||Bacteriocin subtilosin A|
|bdbA||0.12 ± 0.06|| ||Thiol-disulfide oxidoreductase|
|pyr operon||0.07 ± 0.06||PyrR||Pyrimidine biosynthesis|
|pstSCABABB||0.07 ± 0.03||PhoPR||Phosphate ABC transporter|
|des||0.06 ± 0.04||DesKR||Fatty acid desaturase|
The genes htrA (c. 60-fold) and htrB (c. 25-fold), both encoding serine proteases, were also strongly induced by rhamnolipids (Table 3 and Fig. 1a). Expression of both genes is controlled by the TCS CssRS, which is activated by heat and secretion stress. Expression of cssRS itself was not induced by rhamnolipids, similar to the effect of heat stress, although moderately increased expression of this operon can be observed under secretion stress conditions caused by overexpression of the secretory protein α-amylase (Darmon et al., 2002; Hyyryläinen et al., 2005).
Almost one-third of the remaining ≥fivefold induced loci represent target genes of ECF σ factors, predominantly σM, with its own autoregulated operon sigM-yhdLK being approximately eightfold induced (Table 3 and Fig. 1a). As a result of a previously described regulatory overlap between different ECF σ factors of B. subtilis (Qiu & Helmann, 2001; Mascher et al., 2007), expression of some genes, such as bcrC and ywaC, can be regulated by more than one ECF σ factor. But the autoregulated loci of the remaining six ECF σ factors of B. subtilis were not significantly induced (≤threefold), indicating that the ECF response to rhamnolipids is mediated mainly by σM. This ECF σ factor is activated by cell wall antibiotics like vancomycin, bacitracin, and phosphomycin, but also under acid, salt, and heat stress conditions (Cao et al., 2002a, b; Mascher et al., 2003; Thackray & Moir, 2003).
Other genes significantly induced by rhamnolipids cannot be assigned to known cell envelope stress regulons. They often encode proteins of unknown function or proteins presumably involved in metabolic and redox processes (e.g. gabD encoding a succinate-semialdehyde dehydrogenase or trxA encoding thioredoxin).
We verified the main findings of our DNA microarray analysis, in particular the activation of the TCS LiaRS and CssRS as well as σM, independently by real-time RT-PCR and basically obtained the same results, albeit with an overall higher induction ratio (Fig. 1b). Such discrepancy was observed in numerous studies before and is attributed to the overall lower dynamic range of DNA microarrays compared with other methods such as real-time RT-PCR (Conway & Schoolnik, 2003; Pappas et al., 2004).
Treatment with rhamnolipids also led to decreased expression of a certain set of genes (Fig. 1a and Table 3). Among the ≥fivefold repressed loci are genes encoding proteins involved in purine and pyrimidine biosynthesis (pyr and pur operon), phosphate transport (pstSCABABB) and sugar metabolism (rbsRKDACB, xylAB) (Table 2). Differential expression of the pyr operon in response to cell envelope stress has been observed previously for B. licheniformis (Wecke et al., 2006).
With almost 20-fold repression, the most strongly downregulated gene is des, which encodes a fatty acid desaturase (Aguilar et al., 1998). Expression of des is controlled by the TCS DesRK and induced by cold shock. The desaturase is important for maintaining membrane fluidity at low temperature by introducing double bonds in phospholipids (Aguilar et al., 2001), indicating that rhamnolipid treatment at sublethal concentrations could interfere with membrane fluidity.
Hierarchical clustering analysis of genes differentially expressed in response to rhamnolipids
Our DNA microarray analysis clearly indicates that rhamnolipids induce both the cell envelope and the secretion stress response. To further validate this novel induction pattern, we performed hierarchical clustering analysis using transcriptome data of B. subtilis induced with different cell wall antibiotics (vancomycin, bacitracin, daptomycin and friulimicin) and exposed to secretion stress. For reason of clarity, we limited our analysis to genes induced ≥threefold and repressed ≥fivefold by rhamnolipids.
Genes controlled by the same regulator form discrete clusters based on their expression pattern under different stress conditions (Fig. 2a). Genes belonging to the cell envelope stress response of B. subtilis are grouped in three clusters and can be assigned to two regulators, σM and the LiaRS TCS (Fig. 2b). They are induced by cell wall antibiotics and rhamnolipids, but not by secretion stress (with the exception of liaH). One of these three clusters contains the target operon of the LiaRS TCS as well as the downstream genes gerAAAB. The other two clusters include mostly target genes of σM. Noteworthy, within the σM regulon, there is a subset of genes, including the mreBCDminCD operon involved in cell division, that is not induced by vancomycin (upper part of σM1 cluster in Fig. 2b). Differences in the induction profiles of subsets of σM-dependent genes have been observed previously (Eiamphungporn & Helmann, 2008).
Figure 2. Hierarchical clustering analysis of genes differentially expressed in response to rhamnolipids. The clustering analysis was performed using the software Cluster 3.0 (de Hoon et al., 2004). Transcriptome data for Bacillus subtilis treated with friulimicin (fri), vancomycin (van), rhamnolipids (rha), bacitracin (bac), daptomycin (dap) and exposed to secretion stress (sec) caused by overexpression of α-amylase were analyzed (see 'Materials and methods' for details). Green indicates induction of the corresponding gene, red repression under the designated condition. Cluster analysis was limited to genes induced ≥threefold and repressed ≥fivefold by rhamnolipids (a). Cluster containing target genes of σM, LiaRS and CssRS (b) and genes repressed by rhamnolipids (c) are shown in detail. A schematic representation of the network orchestrating the response to rhamnolipids summarizes the results of the cluster analysis (d). The thickness of the arrows corresponds to the induction of the given regulators. Sec, secretion stress; Rha, rhamnolipids; Ces, cell envelope stress.
Download figure to PowerPoint
Figure 3. Growth of Bacillus subtilis wild-type and mutant strains exposed to different concentrations of rhamnolipids. Bacillus subtilis wild type (W168), TMB1070 (cssRS::kan), TMB589 (ΔliaR), TMB329 (ΔliaF), TMB1003 (sigM::kan), TMB149 (sigW::MLS), TMB1393 (sigM::kan sigW::MLS) and TMB1392 (ΔliaRsigM::kan sigW::MLS) were grown in LB medium to mid-logarithmic growth phase. The cultures were split into 1 mL samples and induced with increasing concentrations of rhamnolipids: 0 μg mL−1 (■), 50 μg mL−1 (□), 100 μg mL−1 (▲), 200 μg mL−1 (Δ) and 300 μg mL−1 (♦). Cell density was monitored by measuring OD600 nm over a time period of 7 h.
Download figure to PowerPoint
Genes mediating the secretion stress response also cluster together (Fig. 2b). The CssRS-dependent target genes htrA and htrB are not only induced by secretion stress and rhamnolipids, but also weakly by vancomycin and bacitracin.
Genes repressed by rhamnolipids show almost unchanged expression under the other conditions tested (Fig. 2c). One exception is the pyr operon, which is strongly repressed by rhamnolipids, but weakly induced by friulimicin and vancomycin.
Taken together, the hierarchical clustering analysis indicates that rhamnolipids induce a combination of two different stress responses: the cell envelope stress response represented by the LiaRS TCS and the ECF σ factor σM, and the heat and secretion stress response mediated by CssRS. Simultaneous induction of the LiaRS TCS and σM is common for cell wall antibiotics such as daptomycin, vancomycin, or bacitracin (Mascher et al., 2003; Hachmann et al., 2009; Wecke et al., 2009). But none of the σM-dependent target genes is induced by secretion stress, while both the CssRS and LiaRS TCS are induced by cell wall antibiotics, rhamnolipids, and secretion stress, but with different intensities (Fig. 2d).
The LiaRS TCS and σM protect cells from rhamnolipid-dependent lysis
Bacteria use signal transducing systems to detect harmful compounds and alter gene expression to protect the cell. We hypothesize that the signal transducing systems activated by rhamnolipids confer resistance and counteract cell damage caused by this antimicrobial compound. Therefore, we compared the growth behavior of B. subtilis wild-type cultures exposed to different rhamnolipid concentrations with strains carrying gene deletions leading to ‘ON’ or ‘OFF’ states of the induced signal transducing systems, which results either in no or constitutively high expression of the corresponding target genes. The strains were grown in LB medium to mid-logarithmic growth phase, the cultures were split and different concentrations of rhamnolipids were added. Subsequent lysis of each sample was monitored by measuring OD600 nm.
For the B. subtilis wild-type strain W168, a concentration of 50 μg mL−1 rhamnolipids did not affect growth (Fig. 3), but was sufficient to induce a transcriptional response as investigated using DNA microarray analysis (Fig. 1a and Table 3). Higher concentrations of rhamnolipids lead to rapid lysis of the culture within 1 h after addition (Fig. 3). Remarkably, even after severe lysis the cultures resumed growth.
To reveal a possible protective function of the LiaRS TCS, we compared the lysis in response to rhamnolipids of two strains carrying deletions in the lia locus: deletion of the response regulator LiaR results in a ‘Lia OFF’ mutant, while deletion of the inhibitory protein LiaF represents a ‘Lia ON’ strain with constitutive expression of the target genes liaIH (Jordan et al., 2006; Wolf et al., 2010). Behavior of the ΔliaR mutant was comparable to the wild-type strain, while the ΔliaF mutant clearly displayed recovery advantages and regained growth more quickly even after addition of high rhamnolipid concentrations (Fig. 3). We also investigated the effect of rhamnolipids on a mutant strain lacking the CssRS TCS that orchestrates the secretion stress response, but did not observe any differences compared with the wild type (Fig. 3).
As a large part of the induced genes are regulated by σM, we investigated how this ECF σ factor contributes to resistance against rhamnolipids. Compared with the wild type, a sigM::kan mutant strain showed an impaired growth phenotype (Fig. 3). While growth of the wild type was not affected at concentrations of 50 μg mL−1, growth of the sigM::kan mutant was clearly arrested. σM controls expression of at least 30 operons involved in cell division, DNA repair and cell envelope synthesis (Eiamphungporn & Helmann, 2008). Another ECF σ factor which controls a similar large regulon is σW (Helmann, 2006). Since expression of the sigW–rsiW operon was induced 2.8-fold by rhamnolipids (Table S1), we also included a sigW::MLS mutant strain in our lysis curve experiments. But this strain shows the same behavior as the wild type, indicating that σW is not responsible for resistance against rhamnolipids (Fig. 3). Therefore, the ECF response to rhamnolipids is mainly mediated by σM, which is in agreement with induction ratios of the sigM and sigW operons (eight- vs. threefold, respectively).
We also tested if a combined deletion of both σM and σW has an additive affect and leads to a more pronounced phenotype, as a functional overlap of ECF σ factors in response to different antimicrobial compounds has already been demonstrated (Mascher et al., 2007). Indeed, the double mutant shows an increased sensitivity compared with the sigM::kan strain, as it did not resume growth in the presence of 100 μg mL−1 rhamnolipid (Fig. 3). Additional deletion of liaR, resulting in inactivation of a third cell wall stress responsive system, did not lead to a stronger susceptibility phenotype (Fig. 3). Taken together, σM seems to play a central role in rhamnolipid resistance, while σW and the LiaRS TCS have only minor functions.