The response to acidic stress conditions is understood best in enterobacteria, and Escherichia coli, S. enterica var. Typhimurium, P. mirabilis and Y. enterocolitica (Castanie-Cornet et al., 1999; Kieboom and Abee, 2006; De Biase and Pennacchietti, 2012), all have effective systems to contend with acid stress. Well-known are the decarboxylation systems, which are composed of two components: a decarboxylase and an antiporter. Protons are consumed in the cytoplasm through the decarboxylation of specific amino acids and the corresponding antiporter exports the decarboxylation product and imports more of the required amino acid (Foster, 2004). The best-studied example is the glutamate decarboxylase (Gad) system depending upon the concerted action of glutamate decarboxylase (GadA/GadB) and of the glutamate/GABA antiporter, GadC (Foster, 1999; Audia et al., 2001; Lund et al., 2014).

Rhizobium tropici CIAT899 is an α-proteobacterium capable of establishing a symbiosis with different leguminous plants including common bean (Phaseolous vulgaris) (Martínez-Romero et al., 1991). During this symbiosis root nodules are formed, which are specialized organs where biological nitrogen fixation (BNF) will take place (Suzaki et al., 2015). The efficiency of this symbiosis can be restrained by different environmental conditions, such as high temperature and low pH (Martínez-Romero et al., 1991; Graham et al., 1994; Hungría et al., 2000; Vinuesa et al., 2003). For example under acidic pH conditions, where survival and persistence of the bacteria are limited, nodulation and BNF can be severely affected. Acidic conditions can be found in the rhizosphere of plants, where the pH is lowered by plant exudates containing protons and organic acids, and inside symbiosomes (Udvardi and Day, 1997). Compared to most other nodule-forming bacteria R. tropici CIAT899 presents an increased resistance to acidic growth conditions.

A few studies trying to identify the genetic determinants of growth at acidic pH have been made in R. tropici (Riccillo et al., 2000; Vinuesa et al., 2003; Rojas-Jiménez et al., 2005; Vences-Guzmán et al., 2011). In 2003, Vinuesa et al. (2003), identified novel rhizobial genes required for acid tolerance. During a screen of a small Tn5 mutant library composed of 1,728 clones, they identified two mutants affected in acid tolerance. In one mutant, the Tn5 was inserted in the sycA-olsC gene cluster and in the second mutant, it was inserted in the lpiA-atvA operon. OlsC catalyzes the hydroxylation at the 2-position of the secondary fatty acid of ornithine lipids (OLs). The presence of this hydroxyl group has been correlated later to an increase tolerance to acidic conditions and high temperatures in CIAT899 (Vinuesa et al., 2003; Rojas-Jiménez et al., 2005; Vences-Guzmán et al., 2011). LpiA is a lysyl-phosphatidylglycerol synthase homologous to MprF from Staphylococcus aureus and the atvA gene is encoding a putative serine lipase homologous to the virulence proteins AcvB and VirJ from Agrobacterium tumefaciens. It was demonstrated that lpiA is induced under acidic conditions, and that LpiA participates in lysyl-phosphatidylglycerol (LPG) biosynthesis, which confers an increased resistance of R. tropici CIAT899 to the cationic peptide polymyxin under acidic growth conditions (Vinuesa et al., 2003; Sohlenkamp et al., 2007). Transcriptional induction of lpiA expression was also induced in Sinorhizobium medicae and Sinorhizobium meliloti as part of the response to low pH (Reeve et al., 2006; Hellweg et al., 2009). GshB participates in glutathione biosynthesis (Riccillo et al., 2000), which is necessary to grow in several environmental conditions like oxidative stress, osmotic stress and acid stress and transcription of gshB is induced under acidic stress conditions as shown by quantitative PCR (Muglia et al., 2007).

In addition, the CIAT899 genome encodes homologs to several proteins that are involved in acid stress response in other bacteria: (1) eptA encodes a putative lipid A phosphoethanolamine transferase, which confers acid resistance in E. coli, Salmonella typhimurium, and Shigella flexneri 2a (Martinić et al., 2011); (2) cfa encodes cyclopropane-fatty-acyl-phospholipid synthase which enhances acid tolerance in E. coli by reducing the permeability of the membrane to H⁺ (Shabala and Ross, 2008; Sohlenkamp, 2017). Five different copies of cfa genes are encoded in the genome of R. tropici CIAT899; (3) H⁺/Cl⁻ antiporters are responsible for protons extrusion when E. coli or Helicobacter pylori are exposed to acidic conditions (Inoue et al., 1999; Padan et al., 2004; Padan, 2008). CIAT899 possesses four genes encoding antiporters that possibly help to maintain pH homeostasis; (4) Exopolysaccharide (EPS) biosynthesis has been related to acid tolerance. In S. meliloti it has been shown that genes related to EPS production are expressed under acidic conditions (Cunningham and Munns, 1984; Hellweg et al., 2009) and the genome of CIAT899 encodes several exo genes and EPS production could help the bacteria to resist acid stress. It is not known, if the orthologues to these genes involved in acid stress response in other bacteria have a function during the acid stress response in R. tropici.

Vinuesa et al. (2003) had identified at least two novel genes involved in the acid stress response in R. tropici out of a screen using only 1,728 mutants. As the genome of R. tropici CIAT899 contains more than 6,280 genes (Ormeño-Orrillo et al., 2012), we think that it should be possible to discover further novel genes having a function in acid stress response in this organism. Here we present a study combining random transposon mutagenesis with transcriptomics to obtain a broad catalog of genes important or essential for the acid stress response in R. tropici CIAT899.

The effects of acid stress have been characterized best in Enterobacteriaceae that can be exposed to extreme changes in extracellular pH as they pass through the inhospitable environment of the stomach into the lower digestive tract (Foster, 2004; Lund et al., 2014). Rhizobia show significant variability in their ability to grow under different low pH conditions, and within this family R. tropici CIAT899 is among the most tolerant strains (Martínez-Romero et al., 1991; Graham et al., 1994). The capacity to grow and persist under environmental changes is essential for survival in the rhizosphere and within the nodules of leguminous plants. A Tn5-based transposon library of 18,300 insertional mutants was screened to identify mutants growing as the wildtype at neutral pH, but that did not grow or that showed a drastically reduced growth under low pH conditions. This strategy aimed at identifying genes specifically required for the acid stress response, but not for growth at neutral pH. We isolated 26 Tn5 mutants that were unable to grow under acidic conditions (see Table 3). For 17 of these we could identify the transposon insertion sites, which could be located mainly to the chromosome, but also to the plasmids pA and pC. None of the selected mutants presented an insertion site in the symbiotic plasmid pB (Figure 2).

Vinuesa et al. (2003) had made a similar screening, but on a much smaller scale. One of the mutants identified in the earlier study had an insertion in the sycA/olsC gene cluster (Vinuesa et al., 2003; Rojas-Jiménez et al., 2005). Interestingly, the mutant JG8656 identified in the present study also had an insertion in the gene olsC. This gene codes for the ornithine lipid (OLs) hydroxylase OlsC, which is responsible for the 2-hydroxylation of the secondary fatty acid in OLs. This hydroxylation has been correlated to an increased resistance of R. tropici to acid stress and high temperatures (Vences-Guzmán et al., 2011). One hypothesis is that the presence of the additional hydroxyl group allows the formation of hydrogen bonds between lipid head groups, thereby making the membrane less fluid and less permeable to protons (Nikaido, 2003; Rojas-Jiménez et al., 2005; Vences-Guzmán et al., 2011; Sohlenkamp and Geiger, 2016). Another gene found by Vinuesa et al. was lpiA (low pH inducible) encoding a putative lysyl-phosphatidylglycerol (LPG) synthase, implicated in LPG biosynthesis. LPG is a membrane lipid whose presence confers resistance to various cationic peptides to S. aureus (Peschel et al., 2001). The lpiA gene is found in an operon with atvA (acid tolerance and virulence), which is an orthologue of acvB from A. tumefaciens. Interestingly, in our study we found a mutant (JG163) affected in atvA, which is consistent with the results obtained by Vinuesa and co-workers (Vinuesa et al., 2003). Furthermore, three mutants were affected in a gene whose product is a hypothetical protein (JG1076, JG4655 and JG6057), in this case two of them have an insertion in the same ORF (JG1076, JG4655). Interestingly, two more mutants (JG283 and JG9867) also presented transposon insertions in an identical ORF. This could be indicating that the transposon mutagenesis is close to saturation.

Mutant JG241 presents an insertion in a gene encoding a putative triose phosphate isomerase enzyme (tpiA1). It has been reported that H. pylori, in addition to the genes involved in urea hydrolysis, induces genes at pH 5.5 that are related to carbohydrate metabolism (including the tpiA gene). An explanation suggested was that the cell's energy requirement is increased under this condition (Ang et al., 2001; Wen et al., 2003; Zanotti and Cendron, 2010). Besides, we found other mutants with insertions in genes related to anabolic and energy generation process like ribulose phosphate 3-epimerase (locus tag RTCIAT899_RS09635), prolyl aminopeptidase (RTCIAT899_RS05920) and acetyl-coenzyme A synthetase (Acs) (RTCIAT899_RS18500). Another mutant (JG283) presented an insertion in a gene encoding a polysaccharide biosynthesis protein. Polysaccharide and exopolysaccharide biosynthesis has been reported to help counteract high proton concentrations by preventing protons from passing into the cell (Aarons and Graham, 1991; Graham et al., 1994; Reeve et al., 1997, 1998; Hellweg et al., 2009). Mutant JG9587 has a Tn5 insertion in an ORF encoding a putative antiporter protein, whose function could be to expel protons toward the periplasm under acid stress conditions. In E. coli the Na⁺/H⁺ antiporter (NhaA) helps to maintain Na⁺ and H⁺ homeostasis and several reports indicate that NhaA activity is increased at high or neutral pH (Padan et al., 2004; Padan, 2008). In contrast, H. pylori has a Na⁺/H⁺ antiporter whose activity is high at acidic and neutral pH, indicating that H. pylori employs this antiporter under acidic conditions (Inoue et al., 1999).

In some bacteria an increase of the extracellular proton concentration is sensed through two-component systems (Lund et al., 2014), examples being the PhoPQ system from S. typhimurium which is induced in acidic conditions (Prost et al., 2007) and the ArsRS system from H. pylori, which senses increased proton concentrations and responds by regulating genes of biosynthesis and metabolism of the urease system (Pflock et al., 2006). Mutant JG2646 presented a transposon insertion in a gene encoding a response regulator. This gene is found in an operon with a gene encoding a histidine kinase, and so this two-component system may be sensing the external pH and may be responsible for the activation of a subset of genes under this condition. Two more mutants had transposon insertions in genes encoding transcription factors (JG9477 and JG8654). These genes code for proteins with homology to MarR and FtrA (locus tag RTCIAT899_ RS00995 and RTCIAT899_RS009190), respectively. MarR presents similarity to the transcriptional regulator Rv1404 from Mycobacterium tuberculosis, which regulates the transcription of the genes rv1403c and rv1405c encoding putative S-adenosyl methionine (SAM)-dependent methyltransferases. This activation under acid stress conditions (pH 5.5) also requires phoP (Healy et al., 2016). In the case of FtrA, there are no reports about its role acid stress conditions.

In mutant JG5373 the transposon insertion is probably affecting a gene whose product might be involved in catalyzing the condensation of formaldehyde and glutathione to S-hydroxymethylglutathion (gshA). In the genomic context of R. tropici CIAT899, a glutamate synthase (possibly gshB) is located downstream of gshA. Ricillo et al. reported that a mutant affected in a gene involved in the biosynthesis of tripeptide glutathione (gshB) in R. tropici was affected in growth under acidic conditions (Riccillo et al., 2000). It is possible that the transposon insertion in the mutant JG5373 creates a polar effect affecting glutathione biosynthesis and causing susceptibility to acid stress. Most often an polar effect occurs when the transposon is inserted into the first ORFs of an operon affecting the expression of the downstream genes of the same operon (Zipser, 1969). In our study, we identified three mutants (JG4882, JG163 and JG2646), whose Tn5 insertion occurred in putative operon, however, only the mutant JG2646 has the insertion in the first gene (Supplementary Figure S1) of the operon. The latter transposon insertion can be expected to cause a polar effect on the expression of the downstream genes. This hypothesis was verified by a complementation assay of mutant JG2646: the gene encoding for the response regulator alone was not able to complement the mutant phenotype when provided in trans whereas the complete operon (including the histidine kinase (HK), locus tag RT899_RS05455) complemented the mutant phenotype when provided in trans.

The results obtained by our library screening indicate that outside of certain conserved components of the acid stress response, such as response regulators or antiporters, the genes necessary for an acid stress response vary from species to species. To study the acid stress response from another angle we used RNA-seq to analyze the transcriptomes of R. tropici CIAT899 cells grown at neutral pH (6.8, control condition), of cells grown at pH 4.5 (acid-adapted cells), and of cells pre-grown at pH 6.8 and then acid-shocked for 45 min at pH 4.5. We chose such a short time (45 min) for the acid shock because we were interested in the very first transcriptional responses. Earlier transcriptome studies in S. meliloti 1021 applied a pH shift from pH 7.0 to pH 5.75 (the lowest pH at which S. meliloti grows) for varying times (from 3 to 63 min), observing the maximum induction between 33 and 63 min (Hellweg et al., 2009). Our RNA-seq results exhibit that gene expression of R. tropici CIAT899 was broadly changed by the acid stress. Hundreds of genes (351 up-regulated and 32 down-regulated in acidic conditions and 11 up-regulated in acid shock) were significantly up- or down-regulated under acidic conditions (Table 4 and Supplementary Table S1). These differently expressed genes (DEGs) can be functionally grouped according to the predicted functions of the encoded proteins. We identified several genes associated with metal transport (like RTCIAT899_RS04615, RTCIAT899_RS04620, RTCIAT899_RS04610, RTCIAT899_RS17535, RTCIAT899_RS10385, or RTCIAT899_RS14255). It is known that in tropical acid soils high concentrations of metals like Zn²⁺, Co²⁺, Cd²⁺, Ni²⁺, and often, Mn²⁺, Fe²⁺, Cu²⁺ and mercury ions are very common. Therefore, these metal transporters could be contributing to the efflux of these metals to avoid the toxic effects of metal ions (Montanini et al., 2007) when the cells encounter high proton and metal concentrations. ABC transporters transport solutes across the membrane utilizing the energy of ATP hydrolysis and they have been over-expressed after acid shock in other bacteria, including S. aureus (Bore et al., 2007). In our study, several of the identified DEGs encoded ABC transporters that were significantly over-expressed under acid stress (Jia et al., 2017) (Figure 5 and Supplementary Table S1). Furthermore, we identified a few putative transcription factors that might be implicated in the regulation of efflux pumps under acid stress (like DeoR, RTCIAT899_RS29995; TetR, RTCIAT899_RS10760, RTCIAT899_RS28830, or YebC/PmpR, RTCIAT899_RS14585). The TetR family of transcriptional regulators has been reported to be involved in the regulation of efflux pumps (Perrone et al., 2017) and drug efflux in mycobacteria (Betts et al., 2003; Wei et al., 2014). Among the DEGs that were up-regulated were also the hyc genes encoding hydrogenases proteins. In E. coli and Salmonella these hydrogenases can reduce acid stress by proton consumption and H₂ production. In this study, 10 genes encoding hydrogenases were over-expressed and the transcriptional induction of hydrogenase genes observed in this study is probably involved in the acid stress response, indicating that R. tropici uses this mechanism to resist high proton concentrations (Supplementary Table S1; Hayes et al., 2006; Zbell and Maier, 2009; Noguchi et al., 2010; Jia et al., 2017). F₀F₁ synthase normally catalyzes the synthesis of ATP from ADP using the energy derived from an electrochemical proton gradient. Under acid stress conditions, hydrolysis of ATP may be used to expel protons from the cytoplasm. This flow of protons through the F₀ subunit could contribute to pH homeostasis and therefore contributed to acid stress response like in E. coli or Corynebacterium glutamicum (Diez et al., 2004; Barriuso-Iglesias et al., 2013). In our study we identify two gene clusters encoding several subunits of F₀F₁ ATPase encompassing nine genes involved in the assembly of the two units of ATPase (F₀ and F₁), which are up-regulated more than 4 times at pH 4.5 compared to pH 6.8. Sigma factors are subunits of the bacterial RNA polymerase (RNAP) playing important roles in transcription initiation, especially during promoter recognition. Specific sigma factors have functions during the differential expression of genes during abiotic stress, during development or during specific growth phases (Paget, 2015). The sigma factor E (SigE) in M. tuberculosis is up-regulated during acid stress response, and Bansal et al. (2017) demonstrated that PhoP interacts with acid-inducible extra-cytoplasmic SigE to regulate a complex transcriptional of genes. In this study, we observed that three sigma factors (RTCIAT899_RS15835, RTCIAT899_RS13855, and RTCIAT899_RS12250) were induced during the acid stress response. These sigma factors are possibly responsible for regulating other genes important for growth under acid conditions. In addition, forty hypothetical proteins were differentially expressed in acidic conditions (see Table 4 and Supplementary Table S1). Further studies are required to explore the functions of these proteins under acid stress. We also specifically looked at the transcriptome data of orthologues of genes that had been reported to contribute to the acid stress response in other bacteria. For example, the gene glutathione synthase (RTCIAT899_RS01980) and the molecular chaperone DnaK (RTCIAT899_RS00760), have a log₂FC of 1 and 1.8 respectively. Genes such as lpiA (RTCIAT899_RS14615) and atvA (RTCIAT899_RS14610) also have a similar log₂FC, and most of the orthologues of genes involved in the acid response in other bacteria have very low log₂FC values or are even repressed (Table 5). To our surprise only a very small number of genes was induced in response to 45 min of acid shock. We had designed our experiment based on the article published by Hellweg et al. (2009). Maybe the transcription induction responding to acid stress is slower in R. tropici, but there is also the possibility that pH 4.5 is not sufficiently stressful for the bacteria to induce a major short-term response (Figure 6). Although R. tropici grows slower at pH 4.5 than at pH 6.8, there is no clear adaptation phase visible in the growth curve obtained at the lower pH.

As expected, both, the Tn5 mutagenesis and RNA-seq approach, produced complementary sets of results. But, the use of both approaches allows the identification of a wider range of genes involved in the acid stress response. Therefore, with this work, we make an important step toward comprehending the genetic bases to understanding the acid stress response in R. tropici CIAT899. Much works remains to be done to understand how all the identified genes contribute to the acid stress response. This is especially interesting in the case of genes where the correlation to acid stress is not completely obvious.

JG-C and CS designed the study. JG-C and CS carried out the experiments. JG-C and LL carried out the data analysis. JG-C and CS were involved in drafting the manuscript and all authors read and approved the final manuscript.