Gene rdsA (RHE_PE00024) encodes the sensory component of an essential hybrid Histidine Kinase (HK) regulatory system. Canonical HK proteins constitute the sensor part of two-component regulatory systems. In these, the canonical HK contains an input domain able to sense regulatory stimuli and subsequently the catalytic ATP binding (CA) domain phosphorylates a His residue in the dimeric histidine phosphotransfer (DHp) domain. The phosphoryl group is then transferred to an Asp residue in the receiver (REC) domain of a separate response regulator protein (Gao and Stock, 2009; Rodríguez et al., 2020). In contrast, the predicted hybrid histidine kinase RHE_PE00024 is comprised of a putative N-terminal sensory region (containing two PAS sensor domains) and a kinase core containing both the CA and the DHp domains. Additionally, it contains a REC domain with a potentially phosphorylable Asp residue. In systems of this kind, transfer of the phosphoryl group to the corresponding response regulator protein frequently entails an intermediate protein with a phosphorylable His residue, designated as histidine phosphotransferase or Hpt (Gao and Stock, 2009). No genes encoding a plausible response element or a Hpt protein were found in its vicinity. To better understand the possible function of RdsA, we implemented a conditional knockdown system in which the rdsA gene is put under the control of a cumate inducible promoter (Chubiz et al., 2013). In the corresponding mutant (cKD24), expression of rdsA depends on the inducible cumate system (see section “Materials and Methods”). Interestingly, growth of the cKD24 mutant in rich medium without cumate was slower [growth rate constant (μ) = 0.23 h^–1, mean generation time (g) = 2.9 h] than the wild type cells (μ = 0.33 h^–1, g = 2 h, Supplementary Figure 1). The slower growth of cKD24 was observed even upon addition of cumate to the medium (μ = 0.22 h^–1, g = 3.1 h). No significant differences were observed in relative viability between strains (data not shown). Notably, microscopical analysis of the cKD24 knockdown mutant showed a striking change in shape compared to the wild-type cells. In the absence of cumate, cKD24 showed 63.6% of cells with a round shape, while 36.4% maintained a bacillary shape (Figure 1). DAPI staining revealed nucleoid material in both kinds of cells (Figure 1), suggesting that they are viable. In contrast, when expression of rdsA is increased by addition of cumate, the cKD24 mutant showed a reduced proportion of round cells (41.2%), and a correspondingly higher proportion of bacillary cells (58.8%, Figure 1). Even in the presence of induction there are a high percentage of cells with a round shape, suggesting that wild-type expression levels of rdsA were not achieved. To verify this, rdsA gene expression was monitored by RT-qPCR analysis. In the absence of induction, a low expression level of rdsA was seen, at a level representing just 4.6% of the wild-type levels. However, for unknown reasons, cumate addition did not lead to high expression levels of the gene, showing only 12% of the wild-type levels (Figure 1). To enhance the level of expression of rdsA in the cKD24 mutant, we introduced supernumerary copies of the rdsA gene (under the control of an IPTG inducible system), in a replicative plasmid. The complemented strain (cKD24 pSRKGm:rdsA) showed expression levels of rdsA 1.68-fold higher than in the wild-type cells, even in the absence of induction; in contrast, a fourfold increase in expression compared to the wild-type was seen upon IPTG addition (Figure 1). Microscopical analysis of cells in the complemented cKD24 pSRKGm:rdsA strain, revealed a wild-type phenotype, where 97% of the cells showed a bacillary shape (Figure 1), with a low proportion of cells with a round shape. As a control, a mock complementation experiment was done, introducing an empty plasmid in mutant cKD24. Both the expression levels of rdsA (Figure 1) and proportion of round cells (61.1 and 48.1% without and with induction, respectively, Figure 1) were indistinguishable from the ones seen in the cKD24 mutant. Together, these results indicate that RdsA depletion causes an anomalous round shape in R. etli CFN42. Since cell morphology in Gram-negative bacteria is determined by a proper structuring of the cell wall (Cava et al., 2013), rdsA may participate in the regulation of cell wall biogenesis.

Many HK proteins require homodimerization to carry out their phosphorylation function (Mann et al., 2016). To evaluate if the HK-hybrid protein RdsA requires homodimerization for its function, an Escherichia coli, LexA-based two-hybrid system (Dmitrova et al., 1998) was used (Supplementary Figure 2). In this system, repression of a target reporter gene is seen only if a fusion of LexA with the desired protein is able to dimerize (Dmitrova et al., 1998). Most of the clones obtained with rdsA revealed a strong repression, indicating the ability of the corresponding protein to dimerize. Interestingly, one of the clones revealed a reduced ability to repress the target gene, indicating a reduced dimerization (Supplementary Figure 2). Although the amplification protocol used to generate these clones was not designed to introduce mutations, this clone displayed a spontaneous single-base change provoking an aminoacid change (W169R), located in the second PAS domain of RdsA. PAS domains are frequently involved in protein-protein interactions (Möglich et al., 2009). To evaluate if this failure to dimerize affects function, the mutant version rdsAW169R was introduced in a replicative plasmid (see section “Materials and Methods”). The cKD24 pSRKGm:rdsAW169R strain, showed 56 and 54% of round cells both in the absence and presence of induction, respectively (Figure 1), indicating a failure to complement the changes in shape characteristic of the cKD24 mutant. Failure for proper complementation was not due to an absence of expression of the mutant gene, as evaluated by a RT-qPCR analysis (Figure 1). These results are consistent with the interpretation that homodimerization of RdsA is essential for its function.

To evaluate if depletion of RdsA lead to changes in cell division pattern, a time-lapse analysis of living cells was carried out (see section “Materials and Methods”). For this, the division pattern of both R. etli CE3 (wild-type, n = 217) and cKD24 (without cumate induction, n = 194) cells was followed over a time course of at most 10 h. To minimize damage to the cells during incubation, a system that reduces evaporation (FluoroDish™ plates) was used. Figure 2 shows the type of growth patterns detected, with a summary of the results of cell fate for each strain (Figure 3). As reported for A. tumefaciens (Cameron et al., 2015) and S. meliloti (Krol et al., 2021), R. etli CE3 displays an unipolar division pattern (Figure 2), where deposition of new material is concentrated in the new poles (Supplementary Movie 1). During these time-lapse experiments, no round cells were observed for the wild type strain. In contrast, cells from strain cKD24 showed alterations in several aspects of cell division, including cell division time, polarity and shape (Supplementary Movies 2–4). Since populations of cKD24 are nearly evenly split in two shape classes (round and bacillary) the progeny of each class was followed separately. For mother cells with bacillary shape, most of their progeny was composed of bacillary cells (70.9%), showing unipolar and asymmetric cell growth (Figure 2, row cKD24 rod-rod; Figure 3B), although their cell division time was slower (3 vs. 2 h) than in the wild type cells. Of these bacillary cells, 12.3% had a polarity change (Figure 2, see row cKD24 polarity change), namely, they begin to synthesize cell envelope by the old pole, while the new pole resulted inactivated for growth (Supplementary Movie 4). This change of polarity is a transient effect; further divisions in these cells resume growth from the new poles. A large fraction of mother cells of bacillary shape (28.1%) produced daughter cells with round shape (Figure 2, see row cKD24 rod-round) with a cell division time of 2.5 h. Finally, some of the mother cells of bacillary shape (1%) produced round daughter cells with multiple growth foci, giving a branched appearance.

On the other hand, cell division in mother cells with round shape produced daughter cells with round (62.6%), bacillary (34.1%) and branched (3.3%) shapes. Notably, the bacillary cells that came from mother cells with a round shape, showed polarity change in a higher proportion (29%) than the one seen in daughter bacillary cells from a mother with bacillary shape (12.3%) (Figures 3B,C). Even in these cases the change of polarity is a transient effect, with cells resuming growth from newly-generated poles. Most of the round cells that produced bacillary daughter cells did so solely from a single point in the round cell (Figure 2 see column cKD24 round-rod). Round cells were able to divide and produced new round cells, but they did so with a long division time, of about 8 h (Figure 2 see row cKD24 round-round, Supplementary Movie 2).

A significant proportion of both bacillary (1%) and round cells (3.3%) from cKD24 produced growth foci at several points in the cell, without producing novel cells, giving a branched shape (Figure 2, see row cKD24 round-branch, Supplementary Movie 3). These branched cells remained in this state for the whole course of the experiment, failing to produce new cells. We were unable to detect a growth pattern, either, for the type of shape generated by each mother cell, or in the position and number of growth sectors, an effect that is clearly seen in the round cells that grow from several points of the cell in a quasi-simultaneous manner (Supplementary Movie 3). There are reports in Rhizobiales of null mutants that generate round cells (Kobayashi et al., 2009; Calatrava-Morales et al., 2017; Zupan et al., 2019; Krol et al., 2020). What is unusual in this case is that depletion of RdsA produced a wealth of effects, modifying cell morphology, polarity, and time and number of growth foci. All these point toward a major disruption of the cell division program in these cells.

To identify which genes presented alterations in expression in strain cKD24, we undertook a RNAseq analysis using the Illumina system (see section “Materials and Methods”). For that, both R. etli strains CE3 and cKD24 were grown in minimal medium (without cumate induction) to mid-exponential phase. Three biological replicates were used for each strain. Details for this analysis are shown in Supplementary Datasets 1C–E and Supplementary Figure 3.

Using as criteria a FDR or adjusted p-value cutoff ≤ 0.02 and a log2 fold change of 1.5, a total of 1,105 genes were classified as differentially expressed genes (DEGs). Of these, 437 DEGs were downregulated, while 668 DEGs were upregulated (Supplementary Dataset 1E). Functional classification of the DEGs was achieved using the Gene Ontology (GO) database by biological processes. Assignation to GO groups belonging to clear biological processes was possible only for 352 DEGs, indicating that most of the changes in expression occur in genes without a known function. To evaluate if there are significant overrepresentation of particular biological processes among the functionally annotated DEGs, these were subjected to a hypergeometric analysis (Figure 4). From this analysis, 101 DEGs were identified as belonging to significantly overrepresented groups. The most overrepresented groups were cell division (including FtsZ-cytokinesis), oxidative respiration (comprised of electron transport chain, ATP synthesis-coupled proton and electron transport, proton transmembrane transport, aerobic electron transport chain, and aerobic respiration), and cell motility (including chemotaxis, flagellum-dependent cell motility, and organization and assembly of the flagellum). The DEGs belonging to these three biological processes were mostly downregulated. On the other hand, the most prominent, overrepresented biological processes that were found to be upregulated are the oxidation-reduction process and pathogenesis (Figure 4). The subset of DEGs comprising the oxidation-reduction process is comprised of proteins with a general prediction of oxidoreductases. No genes that participate in tolerance to oxidative stress were found among the DEGs. The pathogenesis group is composed of proteins belonging to the BA14K protein family, which are poorly characterized.

This analysis was complemented by classification of all the DEGs according to their function in the COG (Clusters of Orthologous Groups) database (Figure 5). This approach revealed groups of genes that correspond in function to those identified by hypergeometric analysis, such as cell division, oxidative phosphorylation and cell motility. Interestingly, 55 DEGs belonging to novel groups were easily detected, corresponding to the classes of replication/nucleotide metabolism; cell wall/membrane/envelope biogenesis; and translation, ribosomal structure, and biogenesis. This combined approach allowed the establishment of a collection of 156 DEGs to be analyzed (Figure 5). Many of these DEGs appear in classes that may be considered essential for bacterial growth. For that reason, we carefully reviewed the existence of orthologs to essential genes among the DEGs. Currently, there is no report about essential genes in R. etli, so the list of DEGs was examined looking for orthologs to essential genes of A. tumefaciens (Curtis and Brun, 2014). In this way, we were able to establish the existence of 70 potentially essential DEGs in cKD24 (Figure 5).

As expected from the phenotype exhibited by the cKD24 mutant, significant alterations were found in 21 genes participating in cell division. Most of these changes were corroborated by RT-qPCR analyses (Supplementary Figure 4). These are depicted in the KEGG map corresponding to cell division, based on the cell division pathway of C. crescentus (Laub et al., 2002), as shown in Figure 6. Most of these genes are conserved in Rhizobiales, such as S. meliloti (Pini et al., 2015). The first aspect to highlight is that the genes ftsA, E, I, K1, K2, Q, W, X, and Z, whose products comprise most of the divisome, as well as the minCDE operon, participating in the correct location of the septum (Rothfield et al., 2005), were downregulated. Significant reductions in expression were also seen for dnaA (responsible for genome replication in coordination with cell division), dipI [a periplasmic protein that interacts with the ftsQLB complex (Osorio et al., 2017)], gcrA (an activator of ctrA) and cckA (a cell-cycle kinase). Interestingly, the master cell division regulator CtrA (Mann et al., 2016) was not downregulated (Figure 6 and Supplementary Figure 4). On the other hand, genes such as pleC (Kim et al., 2013), podZ (Grangeon et al., 2017; Howell et al., 2017), popJ (Fields et al., 2012; Anderson-Furgeson et al., 2016), divK, divJ (Pini et al., 2013), and GPR (Zupan et al., 2019, 2021), participating in determination of asymmetric and unipolar growth, displayed normal expression levels in the cKD24 mutant. The downregulation in many genes for cell division suggests a major perturbation in cell division, which is consistent with the longer cell division times and cell division problems exhibited by the cKD24 mutant (Figures 2, 3 and Supplementary Movies 2–4). These data indicate that depletion of RdsA affects, either directly or indirectly, a sizable fraction of genes involved in cell division.

For the PG synthesis pathway (Figure 7), extensive downregulation, verified by RT-qPCR analysis (Supplementary Figure 4) was also detected. The uppS and upK genes, performing the synthesis of di-trans, poly-cis-undecaprenyl phosphate from farnesyl diphosphate, were shown to be downregulated. Downstream in the pathway, the genes murA, murE, and mraY (involved in the synthesis of precursors of PG) displayed reduced levels in the cKD24 mutant. Reductions in the expression levels of both ddl and alr genes, which participate in synthesis of D-amino acids, were undetectable in the RNAseq analysis, but showed a distinct downregulation in the RT-qPCR analysis. Low gene expression was also seen for genes, such as murJ, participating apparently in flipping of the PG precursors across the plasma membrane for subsequent polymerization (Egan et al., 2020) and ftsW, a peptidoglycan glycosyltransferase (Taguchi et al., 2019). These observations point toward a general decrease in the synthesis of PG precursors.

Furthermore, the penicillin-binding proteins (PBP) ftsI and PBP1b1, which are responsible for the transpeptidation and transglycolyzation of PG molecules during the cell division cycle (Cameron et al., 2014), were downregulated in the cKD24 mutant, although the reduction in expression of PBP1b1 was only detected in the RT-qPCR analysis (Supplementary Figure 4). Interestingly, similar levels of expression to the wild-type cells were seen for PBPs that participate in cell elongation, such as PBP1c, or essential for polar growth in A. tumefaciens and S. meliloti, as PBP1a (Williams et al., 2021). Down regulation was also observed for low molecular weight PBPs, such as PBP6B and dacF. The significance of these reductions is difficult to ascertain, given that its individual participation might be dispensable for the polymerization of PG, due to their redundant activities (Egan et al., 2020; Williams et al., 2021). However, cell branching and ectopic cell poles have been reported for E. coli cells lacking multiple low molecular weight PBPs (Nilsen et al., 2004; Potluri et al., 2012). For the LD-transpeptidases, participating in PG biosynthesis during cell elongation (Egan et al., 2020; Aliashkevich and Cava, 2021), enhanced expression was found for ldtE, but lower expression was seen for ldtA, ldtB, and ldtD. Interestingly, one of these (ldtB) localizes to the growth pole, at least in A. tumefaciens (Cameron et al., 2014; Grangeon et al., 2015). These data indicate that RdsA may function as an activator, either directly or indirectly, of the PG biosynthetic pathway. The complex phenotype displayed by mutant cKD24, characterized by a change in cell shape and altered cell division, may be the product of the reductions in expression of genes for cell division and PG biosynthesis.

Concerning the DEGs involved in oxidative respiration, most of the genes of the nuo clusters, which encode respiratory complex I, were shown to be downregulated. Similar behavior was observed for the succinate dehydrogenase operon (sdhBADC), comprising respiratory complex II. A large fraction of the genes that encode the ATP synthase complex in R. etli, affecting both the F₁ (atpA, C, D, G, H) and the F₀ (atpB) regions were also downregulated in cKD24. The marked reduction in expression of genes for respiratory complexes suggests that the mutant cKD24 could face difficulties in the generation of energy.

Surprisingly, reductions in expression of several genes that participate in ribosome structure and function, corroborated by RT-qPCR (Supplementary Figure 4), were also seen. For structural ribosomal proteins, S7 (rpsG), L5 (rplE), L13 (rplM) and L35 (rpmI) displayed reduced expression in cKD24. Significant reductions were also seen for infC that encodes for the translation initiation factor IF-3 (Singh et al., 2005) and proS encoding the proline-tRNA ligase (Zajac et al., 2020). These results suggest that cKD24 may present problems in the synthesis of proteins.

All the Rhizobium etli strains were grown in PY medium supplemented with calcium chloride (Noel et al., 1984) or in MMY minimal medium (Bravo and Mora, 1988) supplemented with biotin at 1 mg l^–1, at 30°C. When needed, cumate (10 μg ml^–1) or Isopropyl β-D-1-thiogalactopyranoside (IPTG, 1.0 mM) were added as inducers. Escherichia coli strains were grown in LB broth at 37°C. When needed, antibiotics were added at the following concentrations (in μg ml^–1): kanamycin (Kan), 30; nalidixic acid (Nal), 20; gentamicin (Gm), 30 and spectinomycin (Sp), 100, tetracycline (Tet), 10. For R. etli growth curves, overnight cultures of the appropriate strains in PY medium were centrifuged, washed in saline solution and used to inoculate 200 μl microplate cultures to an A₆₀₀ of 0.05. Microplates were incubated at 30°C, with hourly determinations of absorbance in a BioTek Synergy™HT reader. For viability determinations, the alamarBlue™ HS cell viability reagent (Invitrogen) was used, according to the instructions of the manufacturer.

The conditional knockdown (cKD) mutagenesis employed aimed to put gene rdsA (RHE_PE00024) under the control of a cumate-inducible promoter. To that end, a region encompassing the transcriptional repressor CymR in conjunction with the PR/cmtO promoter (Chubiz et al., 2013) was amplified by PCR and ligated into pK18mob (Schäfer et al., 1994), generating pMASQ plasmid. The specific knockdown vector was generated by amplification of a 306 pb segment of rdsA (encompassing the translation initiation site and 295 bp downstream) and ligated into pMASQ, giving rise to pMASQ:rdsA_300nt. This plasmid was transformed into E. coli S17-1 and used as donor in plate biparental matings with R. etli CE3, selecting for transconjugants resistant to Nal and Kan. A single-crossover recombination event with the homologous target cointegrates the whole plasmid, producing the displacement of the native rdsA promoter and the incorporation of the cumate-inducible promoter upstream of the full-length rdsA gene (strain cKD24). To minimize apparition of second-site suppressors of cKD24, all experiments were started either from newly-constructed cKD24 or from frozen stocks preserved in glycerol. For complementation, the full-length rdsA was cloned into the replicative plasmid pSRKGm (Khan et al., 2008), under the control of a lac-inducible promoter, with the restriction enzymes NdeI-BamHI, giving rise to pSRKGm:rdsA. Primers employed in this work are shown in Supplementary Datasets 1A,B. All primers were designed with Primer-BLAST software, and purchased from commercial providers. Taq DNA polymerase (Thermo Scientific) and Platinum Taq polymerase (Invitrogen), were used in PCRs.

For static visualization, samples of 9 h liquid cultures in PY medium (either with or without inducer) of R. etli strains were spotted on agarose beds (1% in PBS buffer, Grace Bio-Labs SecureSeal™ imaging spacer) and analyzed on a Nikon Eclipse Ti inverted microscope, using 100X/1.45 NA Plan Apochromat oil objective. Images were recorded with a Nikon DS-Qi2 camera. For DAPI (4′,6-diamidino-2-phenylindole) staining, cell pellets (washed thrice with PBS buffer), were resuspended in PBS and DAPI was added to a final concentration of 3 μg ml^–1, cells were incubated in the dark for 5–10 min at ambient temperature. Cell washed three times in the same PBS buffer were spotted on an agarose pad and observed. Fluorescence was detected using a maximum emission of 456 nm, excitation of 340 nm filters. For time-lapse microscopy, samples of 9 h liquid cultures in PY medium of R. etli strains were spotted on agarose pads (1% in MMY medium, FluoroDish™ plates). Images were captured every 10 min for 24 h with Nikon Eclipse Ti inverted stand, using 100X/1.45 NA Plan Apochromat oil objective, a Nikon DS-Qi2 camera and the Nikon’s Digital Sight DS-U3 camera controller. Control of imaging parameters was done with Nikon NIS-Elements AR 4.20 software and a perfect focus system. Temperature (30°C) and humidity control was maintained with a Lexan Enclosure Unit with Oko-touch temperature control. Time-lapse series were analyzed with Fiji (Schindelin et al., 2012).

A LexA-based two-hybrid system (Dmitrova et al., 1998) was used. Full-length rdsA gene was amplified by PCR from genomic DNA of R. etli CE3, ligated into the LexA plasmid (pSR658-A) and transformed into E. coli SU101. E. coli SU101 harboring different plasmid constructs were grown (A₆₀₀ = 1.0) in LB medium with Tet, Kan, and IPTG. The β-galactosidase and microplate protein concentration determinations were as previously described (Guadarrama et al., 2014). The extent in reduction of β-galactosidase activity was used as a proxy of homodimerization efficiency (Dmitrova et al., 1998). For complementation with rdsAW169R in R. etli, the wild type rdsA gene was excised from plasmid pSRKGm:rdsA and substituted by a BlpI-AatII segment harboring rdsAW169R (generated by PCR, Supplementary Dataset 1A). pSRKGm:rdsA and pSRKGm:rdsAW169R were transferred separately to cKD24 by biparental matings, selecting for transconjugants on PY medium with Kan, Nal, and Gm.

Samples were subjected to SDS-PAGE (12% polyacrylamide) and transferred to 0.45 μm PVDF Immobilon-P membranes (Millipore) in a Tank transfer system (Bio-Rad). Membranes were blocked with 5% non-fat milk, incubated with anti-LexA polyclonal antibody or anti-DnaK monoclonal antibody (Ab-cam) and washed with PBS 1X, 0.05% Tween 20. Immunodetection was performed with a 1:15,000 dilution of horseradish peroxidase-conjugated anti-rabbit (polyclonal) or anti-mouse (monoclonal) antibodies (MDL Millipore Corporation) and visualized with an Immobilon Western chemiluminescence reagent kit (Millipore).

Uninduced R. etli strains were grown in MMY at 30°C to an A₆₂₀ of 0.5 (9 h). Three biological replicates were set up for each strain analyzed. Cell pellets for each replicate (10⁸ cells) were stored on RNAlater (Thermo-Fisher). RNA extraction, rRNA depletion, library preparation and Illumina sequencing was done by GeneWiz (South Plainfield, NJ, United States). Quality checking was done using FastQC¹ and sequences were trimmed using Trimmomatic v.0.36 (Bolger et al., 2014). Paired-end reads were mapped toward R. etli CFN42 reference genome (GenBank accession number GCF_000092045.1) using Bowtie2 aligner v.2.2.6. (Langmead and Salzberg, 2012). Read counts for each CDS (6226) were generated using featureCounts software (Liao et al., 2014), allowing MultiMapping and MultiOverlap reads. For Principal Component Analysis (PCA), the NOISeq PCA function (Tarazona et al., 2015) was used, and clustering was done with Distance Matrix Computation package in R using the function dist², with normalized counts in natural logarithm. Heatmaps were visualized using pheatmap³ of R. Differential expression analysis was done with NOISeqBio (Tarazona et al., 2015). Reads were filtered with NOISeq using method 1 and cpm = 2. Data post filtering were used for differential gene expression in NOISeqBIO, using TMM normalization. For identification of significantly Differentially Expressed Genes (DEGs), False Discovery Rate (FDR) or adjusted p-value cutoff ≤ 0.02 and log2 fold change ≥ 1.5 were used. Categorization of DEGs was achieved using the Clusters of Orthologous Groups database (Galperin et al., 2021), and mapped to known metabolic pathways with KEGG Mapper (Kanehisa and Sato, 2020). Significant overrepresentation of gene ontology (GO) groups among DEGs was determined by hypergeometric testing with Gostats (Falcon and Gentleman, 2007), using a p-value cutoff ≤ 0.05. The universe used was GO terms in the R. etli genome, recovered from Uniprot (The UniProt Consortium, 2017), with the GO database of Gostasts. For evaluation the essential genes in the DEG dataset, entries corresponding to Agrobacterium tumefaciens str. C58 in the Database of Essential Genes [DEG (Luo et al., 2014)] were scanned. Matches in the DEG dataset were verified by BLASTP bidirectional best-hit analysis.

Transcriptional differences were validated by reverse-transcription quantitative PCR (RT-qPCR) analysis. All primers (Supplementary Dataset 1B) amplify 100–300 bp regions. RNA was isolated using Trizol Reagent (Zymo Research), and DNA traces were removed by DNAase I treatment. Absence of DNA from RNA samples was corroborated by PCR with rpoA gene primers, and RNA integrity was verified by gel electrophoresis. RNA Purity and concentration were quantified with Nanodrop 20000 Spectrophotometer (Thermo Scientific™). cDNA was generated with 2.5 μg of RNA, using RevertAid™ H Minus First Strand cDNA Synthesis kit (Thermo Scientific™). RT-qPCR analysis was carried out on a StepOnePlus (Applied Biosystems™) Real-Time PCR System using Maxima SYBR Green/ROX qPCR Master Mix (Thermo Scientific) according to the manufacturer instructions. Three biological replicates were performed for each strain/condition, using three technical repeats. Expression levels were normalized using rpoA as reference gene (Supplementary Dataset 1B). The relative expression (RQ) was determined as 2^ΔΔCT. To detect significant differences between samples, an unpaired Student’s t-test (t test function of the stats package in R) was used, employing the RQ values and a cutoff P ≤ 0.05. RQ values were converted to log2 fold changes, where RQ values ≥ 0.65 are shown as positive log2 fold change, and RQ values ≤ 0.65 as negative log2 fold change. To give a negative value of RQ values ≤ 0.65, we use the formula (-1)/RQ.