Disruption of the OLE ribonucleoprotein complex causes magnesium toxicity in Bacillus halodurans

Summary

OLE RNAs represent an unusual class of bacterial noncoding RNAs common in Gram positive anaerobes. The OLE RNA of the alkaliphile Bacillus halodurans is highly expressed and naturally interacts with at least two RNA-binding proteins called OapA and OapB. The phenotypes of the corresponding knockouts include growth inhibition when exposed to ethanol or other short chain alcohols, or when incubated at modestly reduced temperatures (e.g. 20°C). Intriguingly, the OapA ‘PM1’ mutant, which carries two amino acid changes to a highly-conserved region, yields a dominant negative phenotype that causes more severe growth defects under these same stress conditions. Herein, we report that the PM1 strain also exhibits extreme sensitivity to elevated Mg²⁺ concentrations, beginning as low as 2 mM. Suppressor mutants predominantly map to genes for aconitate hydratase and isocitrate dehydrogenase, which are expected to alter cellular citrate concentrations. Citrate reduces the severity of the Mg²⁺ toxicity phenotype, but neither the genomic mutations nor the addition of citrate to the medium overcome ethanol toxicity or temperature sensitivity. These findings reveal that OLE RNA and its protein partners are involved in biochemical responses to several stress conditions, wherein the unusual sensitivity to Mg²⁺ can be independently suppressed by specific genomic mutations.

Graphical Abstract

Abbreviated summary

The large, noncoding OLE RNA from the bacterium B. halodurans forms a membrane-localized RNP complex whose precise biochemical function is unknown. Surprisingly, disruptions to the components of the complex cause extreme sensitivity to Mg²⁺, in addition to previously known sensitivities to ethanol and low temperatures. Suppressor mutations in citrate metabolism genes overcome Mg²⁺-toxicity only, which suggests that OLE RNP complexes might broadly monitor changes at the membrane to permit cells to respond to diverse stresses.

Introduction

With the exception of the ubiquitous ribosomal and RNase P ribozymes, and the widely distributed group I and group II self-splicing ribozymes, large and highly-structured noncoding RNAs (ncRNAs) are rare in the bacterial domain of life (Harris & Breaker, 2018). Another exception is the class of unusually long bacterial ncRNAs, called OLE (ornate, large, extremophilic), which are commonly found in extremophilic Gram-positive bacterial species that can grow under anaerobic conditions (Puerta-Fernandez et al., 2006; Block et al., 2011; Wallace et al., 2012; Harris et al., 2018). OLE RNAs exhibit several characteristics indicative of a role in complex biological and biochemical processes. For example, they average approximately 600 nucleotides in length, and form an intricate secondary structure including more than a dozen base-paired substructures (Fig. S1) (Puerta-Fernandez et al., 2006, Wallace et al., 2012). Also, the RNAs include nearly 100 nucleotide positions whose base identities are conserved in 97% or more of the 795 unique-sequence representatives examined (Harris et al., 2018). These sequence and structural characteristics are indicative of the formation of a complex tertiary architecture that is needed to perform sophisticated biochemical functions.

All previous studies investigating the cellular function of OLE RNAs have been performed using the bacterium Bacillus halodurans, which thrives in alkaline conditions (Takami and Horikoshi, 1999). This species is a conveniently culturable and genetically-feasible (Wallace and Breaker, 2011) facultative anaerobe with a fully-sequenced genome. B. halodurans produces large amounts of OLE RNA under optimal laboratory growth conditions, and further boosts its concentration under ethanol stress conditions to become one of the most abundant RNAs in the cell (Wallace et al., 2012). These observations suggest that OLE RNAs might be important for bacterial adaptation to challenging physicochemical environments. The B. halodurans OLE RNA representative (Fig. 1), produced by transcription of the ole gene, carries all the sequence and structural features common to most representatives of this ncRNA class (Fig. S1). Given the features of this species noted above, and given that its OLE RNA representative is typical of this ncRNA class, B. halodurans serves as an excellent model organism for exploring the characteristics of this large and unusual RNA.

Fig. 1.

Sequence and secondary structure model for the OLE RNA from B. halodurans, wherein red letters identify nucleotides that are conserved in 97% or more of the known representatives in bacteria (Fig. S1). The figure is adapted from the consensus model published previously (Harris et al., 2018).

Although the precise role that OLE RNAs play in cells has yet to be elucidated, additional characteristics of OLE RNAs have been uncovered since the first description of this ncRNA class (Puerta-Fernandez et al., 2006). Importantly, each OLE RNA forms a ribonucleoprotein (RNP) complex with an ~200-amino acid membrane-associated protein called OapA (Fig. 2A) (Harris et al., 2018). The resulting complex thus localizes the RNA to cell membranes (Block et al., 2011). A second protein, now called OapB (Fig. 2B, top), was identified in B. halodurans to be critical for forming the biologically active OLE RNP complex (Harris et al., 2018). OapB contains a widespread bacterial KOW domain. This protein motif often functions as an RNA-binding domain (Kyrpides et al., 1996). Indeed, we determined that OapB binds OLE RNA with high affinity and specificity, and that mutation of its most highly-conserved amino acids recapitulates the growth defect phenotypes observed with strains lacking OLE RNA or OapA protein (Harris et al., 2018). OLE RNAs might have the ability to interact with additional protein partners or other molecules, but this has yet to be determined.

Fig. 2.

Protein components of the OLE RNP complex.

A. Amino acid sequence, extent of conservation at each position, and predicted transmembrane architecture of the OapA protein from B. halodurans. Sequence conservation reflects both the conservation of the identity and physicochemical properties of the amino acid at each position (Harris et al., 2018). Invariant positions are 100% conserved. High conservation indicates that, although the amino acid can vary, the properties are conserved. Positions with Moderate conservation have JalView scores (Waterhouse et al., 2009) of 8 and 9, and positions with Low conservation have scores of 6 and 7. The positions of the D-to-A mutations that constitute the PM1 variant (gray box, right panel) are at positions 100 and 104 (green circles).

B. (Top) Amino acid sequence and extent of conservation for the OapB protein from B. halodurans. (Bottom) Position (57) and amino acid identity changed (H to Y) in the KH02 OapB mutant. Annotations are as described in A. The figures for OapA (see A) and OapB are adapted from the consensus models published previously (Harris et al., 2018).

Previous analyses of B. halodurans strains carrying genetic knockouts of OLE RNA (Δole), OapA (ΔoapA), or both (Δole-oapA) revealed that these genes contribute to ethanol and cold tolerance (Wallace et al., 2012). Furthermore, we determined the effects of mutations at several conserved sites in the RNA and protein sequences, and uncovered a dominant-negative OapA mutant (termed OapA protein mutant 1 or ‘PM1’) that causes a more severe growth defect under these same stress conditions (Harris et al., 2018). The specific mutations involve substitution of two strictly conserved aspartic acid residues each with a single alanine residue (D100A and D104A). B. halodurans cells carrying the PM1 mutant incubated in media containing 5% ethanol, or incubated at 20°C, exhibit severe reductions in growth and survival (Harris et al., 2018). Aspartic acid-rich motifs, like DxxxD (where ‘x’ is variable), have been shown to bind Mg²⁺ ions to coordinate the diphosphate moieties of certain prenyl pyrophosphates (Brandt et al., 2009) and presqualene diphosphate (Lin et al., 2010), and to bind Mn²⁺ in specific ion transporters identified in plants (Chen et al., 2016). Therefore, we speculated that OapA might possibly bind to and/or regulate the homeostasis of certain divalent ions.

In the current study, we sought to determine whether exposure to various divalent metal ions causes phenotypic differences between wild-type (WT) and various mutant strains of B. halodurans. We discovered that B. halodurans cells lacking the ole and/or oapA genes exhibit growth deficiency when exposed to elevated concentrations of Mg²⁺ specifically. Intriguingly, the PM1 strain exhibits a more severe growth defect at even modest concentrations of Mg²⁺. Cells stably adapt to this stress by mutation, and a genetic selection reveals that mutations in certain genes in the citric acid cycle confer Mg²⁺ tolerance on the RNP complex knockout and PM1 strains.

Results and Discussion

Mutant strains of B. halodurans are strongly sensitive to Mg²⁺

Previously, we noted that OapA (BH2780) proteins are strictly present in organisms that carry OLE RNAs (Puerta-Fernandez et al., 2006), and that the protein had no close homologs (Block et al., 2011). Thus, at that time, we could not infer additional biochemical functions for OapA by comparison with homologous protein classes whose functions were already known. In the current study, we conducted a new bioinformatics search for additional OapA representatives and for proteins that are most closely related to this unique protein class. Two notable observations were made. First, the oapA gene homolog in Aeribacillus pallidus is annotated as citMHS. The gene with this name in some species encodes a symporter protein for citrate and Mg²⁺ (Boorsma et al., 1996; Wakesman et al., 2014). Second, our searches identified a DUF21 domain (also termed PF01595) within the OapA sequence. Other protein hits with weak similarity to OapA were identified in our search, as discussed below, and these also include the DUF21 domain. This structural domain is a transmembrane feature present in various transporter proteins from all domains of life (Bateman et al., 2018).

Recently, two DUF21-containing proteins with highly similar amino acid sequences compared to each other were reported to function as Mg²⁺ exporters. These proteins, YhdP in Bacillus subtilis (Akanuma et al., 2014) and MpfA in Staphylococcus aureus (Armitano et al., 2016), have amino acid sequences that are vaguely similar to OapA (Fig. S2A). In addition to the transmembrane DUF21 domain, YhdP and MpfA each possess two CBS domains and a CorC/HlyC domain (Fig. S2B). CBS domains regulate enzymatic and/or transporter activity through the binding of certain molecules that carry an adenosyl moiety (Baykov et al., 2011), such as ATP (Kemp, 2004) and cyclic di-AMP (Sureka et al., 2014). CorC/HlyC domains (also termed PF03471) (Bateman et al., 2018) have less-well established functions, but are presumed to also play regulatory roles. Although ion transport has not yet been directly observed for these proteins, deletion of the gene for MpfA causes host cells to experience strong Mg²⁺ toxicity (Armitano et al., 2016). Also, disruption of the gene for YhdP suppresses a deleterious growth phenotype caused by problems with ribosome production in B. subtilis, presumably by increasing cellular Mg²⁺ concentrations (Akanuma et al., 2014).

If OapA is involved in transport, we speculated that it might be responsible for ligand export due to the large number of conserved residues residing on the predicted cytoplasmic side of the membrane-embedded protein. Moreover, given the presence of a DUF21 domain, and given the apparent similarity between OapA, YhdP, MpfA, and CitMHS proteins as noted above, it seems reasonable to speculate that OapA might function as a Mg²⁺ transporter. However, the A. pallidus oapA gene annotated as citMHS is located immediately downstream of the ole gene, and therefore is a strong oapA representative. Because many types of proteins carry DUF21 domains, assigning functions of these proteins based on homology assessments is challenging.

Despite these uncertainties, it is worth considering that B. halodurans lacks a close homolog of YhdP or MpfA, which might be expected if OapA is functionally similar. Conversely, OapA lacks CBS and CorC/HlyC domains that are major features of YhdP and MpfA, and thus is substantially smaller than many known Mg²⁺ transporters (Fig. S2). Although it would be unprecedented, we considered the possibility that OLE RNAs might be an essential component of a system that contributes to Mg²⁺ homeostasis. To assess this hypothesis, we sought to establish a link between the OLE RNP complex and Mg²⁺ homeostasis by evaluating the effects of Mg²⁺ on the growth of B. halodurans ole and oapA mutant strains.

Intriguingly, when the B. halodurans ∆ole-oapA knockout strain and the PM1 strain are grown on Lysogeny Broth (LB, adjusted to pH 10) agar supplemented with 10 mM MgCl₂, both confluent growth and colony formation is inhibited (Fig. 3). For comparison, the concentration of Mg²⁺ present in unaltered LB is expected to be approximately 200 μM (Wee and Wilkinson 1988) or less (Snavely et al., 1989). These same cells exhibit no inhibition of growth in the presence of this lower Mg²⁺ concentration. Strong inhibition of bacterial growth by modest (low mM) concentrations of Mg²⁺ is extremely unusual, and to our knowledge has only been reported with the knockout strain of the putative Mg²⁺ transporter MpfA as noted above (Armitano et al., 2016). ∆ole-oapA cells expressing WT ole-oapA on a plasmid (rescue) regain their normal growth phenotype when exposed to LB (pH 10) agar media supplemented with 10 mM MgCl₂.

Fig. 3.

Growth assays for various B. halodurans strains reveal Mg²⁺ sensitivity caused by mutations in the OLE RNP complex. Strains grow normally on LB agar (pH 10) in the absence of added Mg²⁺ (left), whereas Δole-oapA and PM1 cells exhibit strong inhibition when the medium is supplemented with 10 mM MgCl₂ (right). Bacterial strains noted are described in the text.

To better understand the effect of Mg²⁺ on the growth of these strains, we monitored their growth by recording absorbance at 600 nm (OD₆₀₀) over 30 h at 37ºC in liquid LB (pH 10) media encompassing a range of MgCl₂ concentrations (Fig. 4). Similar to the results on LB agar, WT and ole-oapA rescue cells grow normally under all concentrations tested, whereas the growth of various knock-out strains, including ∆ole-oapA, ∆ole, and ∆oapA, is negatively affected by MgCl₂. Notably, these strains require more time to transition from lag-phase growth to exponential growth with progressively higher concentrations of MgCl₂. The PM1 strain exhibits the greatest sensitivity to Mg²⁺, even at concentrations as low as 2 mM, and no growth was observed with higher concentrations over the 30-h time course.

Fig. 4.

Growth characteristics of various B. halodurans strains in the presence of different Mg²⁺ concentrations. Growth curves were generated with WT, ∆ole-oapA, rescue, ∆ole, ∆oapA, and PM1 strains incubated for 30 h in LB (pH 10) with MgCl₂ supplementation as noted. The OD₆₀₀ was measured in 30-min intervals. Each data point is the mean of three experimental replicates.

Cells experience the same inhibition effect when incubated in media containing MgSO₄ (Fig. S3), suggesting that the type of negatively charged counter ion is not important. Various other divalent ions were also examined in growth assays. The addition of NiCl₂, ZnCl₂, or CaCl₂ did not cause cell growth inhibition among the strains (Fig. S4). CoCl₂ appears to have a small reverse effect in which ∆oapA and ∆ole-oapA are more tolerant than WT and ∆ole. However, the difference in growth caused by CoCl₂ is modest (Fig. S5), and therefore was not further pursued in the current study. However, it is worth noting that a feature of many Mg²⁺ transporters is the ability of their mutation or deletion to confer Co²⁺ resistance (Gibson et al., 1991). This effect is proposed to occur because the transporter can mistakenly recognize and import divalent cations other than Mg²⁺ (Groisman et al., 2013, Snavely et al., 1989).

Restored Mg²⁺ tolerance in mutant strains is achieved through genetic adaptation

A key observation from growth curve assays demonstrating Mg²⁺ toxicity is that ole and/or oapA knockout cells appear to fully recover after 10–20 hours of incubation, and grow to the same density as cells not exposed to additional MgCl₂ (Fig. 4). Even the PM1 strain can eventually grow to full density. We sought to determine if the selected cells from the ∆ole-oapA and PM1 cultures were biochemically adapting (e.g. upregulating expression of certain genes) or genetically adapting (e.g. through suppressor mutations). After these strains were subjected to selection for Mg²⁺ tolerance, and subsequently grown for 48 h in in media supplemented with MgCl₂ (10 mM for ∆ole-oapA, 5 mM for PM1), an aliquot of each was transferred to fresh LB (pH 10) containing no additional MgCl₂. The cultures were incubated overnight and then inoculated into fresh LB (pH 10) for the next incubation. This serial passage culture process was repeated for 12 days to ensure that the cells had fully reacclimated to low-Mg²⁺ media.

After 12 serial passages, growth curves were again generated using media containing various concentrations of supplemental MgCl₂. If cells were biochemically adapting, they should grow similarly to how they did upon initial exposure to Mg²⁺ supplementation. If cells adapted genetically, the mutant ∆ole-oapA and PM1 strains should grow without sensitivity to added Mg²⁺. Both the selected ∆ole-oapA and PM1 populations grew without any unusual lag phase when the media was supplemented with Mg²⁺ (Fig. 5A). These findings indicate that both the ∆ole-oapA and the PM1 strains likely acquired suppressor mutations that enable them to grow normally when challenged with elevated Mg²⁺ concentrations. To confirm that these cultures had not been contaminated by WT cells (which would also explain the observed growth), or by another bacterial species, the sequences of the regions of the chromosome and plasmid where ole and oapA genes reside were confirmed by DNA sequencing of corresponding PCR amplification products.

Fig. 5.

Genetic selection of Mg²⁺-resistant populations of otherwise Mg²⁺-sensitive B. halodurans ∆ole-oapA and PM1 parental cells.

A. Growth curves generated for populations of ∆ole-oapA and PM1 cells selected for resistance to 8 mM Mg²⁺, and subjected to 12 serial passages without Mg²⁺. Plots were generated using LB (pH 10) media supplemented with various concentrations of MgCl₂ as noted. Additional details are as described in the legend to Fig. 3.

B. Plot of the occurrences of mutations in four genes on sequencing a total of 30 strains exhibiting Mg²⁺ resistance. See Table S1 for details.

Suppressor mutations for Mg²⁺-toxicity resistance are associated with citrate metabolism genes

Based on the ability of the ∆ole-oapA and PM1 B. halodurans cells to adapt to the toxic effects of Mg²⁺, we conducted a genetic selection to recover additional strains for further analysis. The selection was initiated by inoculating either the ∆ole-oapA or the PM1 strains of B. halodurans cells without intentional mutagenesis into multiple liquid LB (pH 10) cultures supplemented with 8 mM MgCl₂. Cultures were incubated at 37°C for 48 h with shaking, and an aliquot of each resulting mixture was streaked onto LB (pH 10) agar supplemented with 10 mM MgCl₂. One variant from each culture was isolated by picking single colonies, and genomic DNA was prepared for whole genome sequencing.

For the ∆ole-oapA B. halodurans selection, eight unique genomes were sequenced from cells that exhibited resistance to Mg²⁺ toxicity. Four genes were found to have mutated, including two related to the citric acid cycle (Fig. 5B). An additional 22 unique genome sequences were examined from the selection initiated with the PM1 B. halodurans strain, all of which carry mutations in the same two citric acid cycle genes. Specifically, 18 genomes each carry a single mutation in the gene encoding aconitate hydratase (acnA, also called citB in some species) (Fig. 5B). The encoded protein catalyzes the isomerization of citrate to isocitrate via cis-aconitate, and has been shown to bind RNA and regulate gene expression in B. subtilis (Alén & Sonenshein, 1999), as well as many other organisms (e.g. see Austin and Maier 2013). Also in B. subtilis, null mutants of citB resulted in the accumulation of citrate in cells and/or the culture medium (Craig et al., 1997). Citrate accumulation is believed to be the result of a complex autoregulatory loop in which mutant CitB can neither metabolize citrate nor regulate expression of the main citrate synthase, CitZ (Pechter et al., 2013).

The remaining four genomes from the PM1 selection carry mutations in the icd gene, which codes for the protein isocitrate dehydrogenase. Icd is an NADP⁺-dependent enzyme that converts isocitrate to 2-oxoglutarate. This step in the citric acid cycle immediately follows the reaction catalyzed by AcnA/CitB. Presumably, the mutations in icd block the degradation of isocitrate, and subsequently cause a build-up of citrate. As with the other genes identified in the genetic selection, different strains acquired distinct mutations (see Table S1 for details). Therefore, it appears that the parental B. halodurans strains carrying defects in the OLE RNP complex most easily acquire various mutations in genes that likely affect the concentration of citrate to overcome the toxic effects of Mg²⁺.

The discovery of suppressor mutations in citrate metabolism genes is particularly intriguing because, as noted above, an oapA gene has been previously annotated as citrate-Mg²⁺ symporter. We considered several hypotheses regarding the mechanism by which elevated citrate might protect cells against Mg²⁺ overabundance. Citrate has been used as a divalent cation chelation agent for industrial applications, so it seemed possible that cells simply use an abundant intermediate in the citric acid cycle to reduce the free concentration of Mg²⁺. To examine this possibility, we employed agar diffusion assays to determine the effects of citrate or isocitrate exposure to various B. halodurans strains. Citrate, and to a lesser extent isocitrate, promote the growth of ∆ole-oapA and PM1 cells when grown on LB (pH 10) agar media supplemented with 10 mM MgCl₂ (Fig. 6). However, the addition of the strong divalent metal-chelating agent ethylenediaminetetraacetic acid (EDTA) does not promote the growth of these strains, suggesting that Mg²⁺ sequestration in the medium is not the source of Mg²⁺ toxicity resistance. Therefore, the mechanism by which the acnA and icd mutations provide Mg²⁺ resistance to the B. halodurans strains carrying disrupted OLE RNP complexes appears to be more complicated than Mg²⁺ chelation outside of cells. To address if OapA functions as a citrate-specific symporter of Mg²⁺ ions further in-depth examination is required.

Fig. 6.

Agar diffusion assays with citrate, isocitrate, and EDTA (a strong Mg²⁺ chelator). Plates containing LB (pH 10) agar supplemented with 10 mM MgCl₂, 1 mM IPTG, and 3 µg mL⁻¹ chloramphenicol were uniformly inoculated with exponential phase B. halodurans cells (either WT, ∆ole-oapA, or PM1). Disks contain 2 M sodium citrate (pH 8), 2 M DL-isocitrate (pH 8), or 0.5 M EDTA (pH 8).

Mg²⁺-tolerant strains do not resist ethanol or cold stresses

We sought to determine if mutations that suppress Mg²⁺ toxicity in the B. halodurans PM1 cells also address the general biochemical defect caused by the PM1 alteration to OapA. To examine this possibility, representative Mg²⁺-resistant strain M01, which carries an icd mutation, and Mg²⁺-resistant strain M04, which carries an acnA mutation (Table S1) were subjected to culture conditions including either 5% ethanol or a temperature of 20ºC. These strains fail to exhibit growth improvements compared to the parental PM1 strain in LB (pH 10) when exposed to ethanol or cold (Fig. 7). Although the genomic mutations in these strains provide resistance to 8 mM MgCl₂, these mutations do not overcome the larger biochemical defect caused by mutation of OapA.

Fig. 7.

Growth characteristics of citrate-related Mg²⁺-toxicity suppressor mutants exposed to Mg²⁺, ethanol, and cold stresses. Optical density (OD₆₀₀) of cultures of various B. halodurans strains in LB (pH 10) media (i) after 24 h incubation with 8 mM MgCl₂ supplementation, (ii) after 48 h incubation at 20ºC, or (iii) after 48 h incubation with 5% ethanol supplementation. Data are the mean and standard deviation of three independent experiments with three technical replicates normalized to WT for each condition.

Previously, we used genetic selection techniques at 20°C to isolate B. halodurans PM1 variants that can overcome both the severe ethanol toxicity and the cold sensitive phenotypes (Harris et al., 2018). For example, a strain (KH02) isolated from this earlier cold-tolerance selection carries a deleterious mutation (H57Y) to an otherwise strictly conserved amino acid of OapB (Fig. 2B, bottom). This and other PM1 cells carrying certain oapB mutations exhibit a reduction in the severity of the ethanol- and cold-sensitive phenotypes, such that they closely approximate the more modest strength of the same phenotypes exhibited by ∆ole-oapA cells (Harris et al., 2018). This outcome can be rationalized if mutation of OapB disrupts formation of the OLE RNP complex, thus mimicking the phenotype observed with ∆ole, ∆oapA, and ∆ole-oapA strains.

Likewise, the KH02 cells overcome the strong Mg²⁺ toxicity phenotype that is characteristic of the parent B. halodurans PM1 cells, as observed by culturing in LB (pH 10) media supplemented with 8 mM MgCl₂ (Fig. S6). This observation demonstrates that mutations at key sites within OapB could have permitted PM1 cells to be isolated as a successful outcome of the Mg²⁺-tolerance genetic selection conducted in the current study. The reason why such OapB mutations were not isolated is uncertain, but could be due to the relative sizes of the genetic targets. For example, mutations only at the most conserved amino acid positions in OapB were isolated in the previous cold-tolerance selection (Harris et al., 2018), whereas numerous different mutations, mostly to two citrate metabolism genes, were identified via the Mg²⁺-tolerance selection. This indicates that there are more codons in the genome whose mutation overcomes Mg²⁺-tolerance alone, than those that simultaneously overcome all three stresses. This hypothesis is also supported by the fact that the recovery of cold-tolerant strains of PM1 required the application of a chemical mutagen to induce a high frequency of mutations (Harris et al., 2018), whereas the Mg²⁺-tolerant strains emerged quickly and without the need for a mutagen. Regardless of the reasons why OapB mutations were not observed, it seems certain that the genetic selection has not been saturated. Therefore, additional Mg²⁺-toxicity resistance genes might remain to be identified for the PM1 strain.

Presumably, if the Mg²⁺-tolerance selection with the PM1 strain were to be continued, strains that carry OapB mutations would eventually be identified. However, such mutations are not expected to emerge if the ∆ole-oapA strain is used for the selection. OapB is required for normal function of the OLE RNP complex (Harris et al., 2018). Therefore, the previously observed OapB mutants would not permit ∆ole-oapA cells to function like WT cells when Mg²⁺ concentrations are elevated. Importantly, we observe that only mutations to the acnA and icd genes allow ∆ole-oapA cells to fully resist Mg²⁺ toxicity (Fig. 5B). Although the Mg²⁺-tolerance selection is not fully saturated, it is possible that mutations to genes related to citrate metabolism might be the easiest and most effective way to overcome Mg²⁺ toxicity for cells with defects or mutations in components of the OLE RNP complex.

CONCLUSIONS

As the most abundant divalent cation in bacterial cells, Mg²⁺ has essential roles in almost all biological processes. To maintain Mg²⁺ homeostasis, which is estimated to include roughly 0.3 mM free divalent ion from a total of approximately 30 mM in cells (Maguire & Cowan, 2002), many bacterial species carry several classes of well-studied transporters (Groisman et al., 2013). Typically, cells are not affected by media containing high concentrations of Mg²⁺. However, we report the surprising finding that alterations to a large bacterial ncRNA, OLE, or to one of its two known protein partners, OapA, can cause B. halodurans cells to exhibit sensitivity to Mg²⁺ when present even at low millimolar concentrations. This inability to tolerate even modest levels of Mg²⁺ appears to be largely overcome by genetic mutations that increase the cellular concentration of citrate (Fig. 4, Table S1).

The mechanism by which Mg²⁺ sensitivity occurs is still unclear. The possibility of Mg²⁺ toxicity was explored because of the vague similarities between OapA and two recently-described proteins implicated in Mg²⁺ export (Akanuma et al., 2014, Armitano et al., 2016), and due to homology of A. pallidus oapA to the citMHS domain in the Pfam database. In addition, deletion of the gene for MpfA causes the unusual phenotype of Mg²⁺ toxicity in S. aureus (Armitano et al., 2016). Given those observations, we considered the possibility that OapA might have a similar function. However, the structural and functional characteristics associated with OapA are surprisingly complex. For example, OapA lacks the additional CBS and CorC/HlyC domains that are so common in known Mg²⁺ transporters. Thus, it is intriguing to consider the possibility that the OLE RNA might take on the biochemical functions of these missing protein domains to regulate OapA transport function, or to participate in a broader signaling process involved in cellular maintenance of membrane integrity.

As noted above, CBS domains are known to regulate protein activity by exploiting adenosyl-binding domains (Baykov et al., 2011). Similarly, numerous riboswitch classes have been discovered that selectively bind to metabolites carrying adenosyl moieties, wherein ligand binding regulates gene expression. This highlights the intriguing possibility that OLE RNA might sense and/or process signaling molecules to regulate the function of OapA. Specifically, riboswitches have been identified that recognize the enzyme cofactor adenosylcobalamin (Nahvi et al., 2002), the nucleobase adenine (Mandal et al., 2003), the enzyme cofactor S-adenosylmethionine (Epshtein et al., 2003; McDaniel et al., 2003; Winkler et al., 2003, Corbino et al., 2005; Fuchs et al., 2006; Weinberg et al., 2008; Poiata et al., 2009; Weinberg et al., 2010; Mirihana Arachchilage, et al., 2018) and its metabolic product S-adenosylhomocysteine (Wang et al., 2008), the nucleotide ADP (Sherlock et al., 2019), and c-di-AMP (Nelson et al., 2013). Of these, cyclic di-AMP is a proven regulator of cellular stress responses (Corrigan and Gründling, 2013) and is known to regulate several ion transporters (Gundlach et al., 2019). However, direct binding of OLE RNA to c-di-AMP or other adenosyl-containing biomolecules has not been observed using various biochemical assays (data not shown).

It is also interesting to consider how the OapA PM1 strain could produce the stronger phenotypes to three distinct stresses. The D100A and D104A mutations alter strictly conserved amino acid positions, which might most predictably disrupt the possible transport function of OapA. However, the PM1 strain exhibits stronger phenotypes to the same stresses as the ∆oapA strain. Perhaps OapA PM1 induces a functional defect in another, yet undetected, partner of the OLE RNP complex. Alternatively, the amino acid differences that constitute PM1 cause a change in signaling activity that propagates to other processes in the cell. Participation of WT OLE RNA is required for normal tolerance to Mg²⁺, as well as resistance to the other two stresses (Wallace et al., 2012, Harris et al., 2018). Intriguingly, OLE RNA is required for cells carrying the PM1 variant of OapA to exhibit stronger growth inhibition caused by ethanol and cold stress conditions. These findings suggest that OapA localization and/or function requires the proper function of OLE RNA.

Unfortunately it remains unknown how sensitivities to Mg²⁺, short chain alcohols such as ethanol, and mildly cold temperatures such as 20°C, are all biochemically related through the action of OLE RNA and its associated proteins OapA and OapB. Our current findings reveal that the Mg²⁺ toxicity phenotype can be readily overcome by mutations to genes involved in citrate metabolism. The fact that these specific adaptive mutations have no effect on the other two stresses suggests that they are unrelated to the source of the biochemical perturbation that causes these unusual phenotypes. Therefore, the most commonly-mutated genes we identified from the Mg²⁺-tolerance selection are not likely to provide clues regarding the larger biochemical defects caused by alteration of the OLE RNP complex. Genetic selections using cold or ethanol tolerance coupled with induced mutagenesis, or an expanded genetic selection for Mg²⁺ tolerance, could yield resistance mutations in other genes that are more directly related to the function of this unusual bacterial ncRNA.

Experimental Procedures

Bacterial strains, plasmids, and cultures

All B. halodurans strains and plasmids used in this study were constructed and described previously (Wallace et al., 2012, Harris et al., 2018). Unless otherwise specified, B. halodurans was grown in LB media (USB Corporation) that was prepared at 90% volume, autoclaved, and adjusted to full volume and to pH 10 with 10% (w/v) filter-sterilized Na₂CO₃ [1% (w/v) final concentration]. Solid media was prepared with 1.5% (w/v) agar. Liquid and solid media were supplemented with MgCl₂ as noted for each experiment. In addition, chloramphenicol and isopropyl-β-D-thiogalactopyranoside (IPTG) were added to culture media. The antibiotic favors cells harboring the plasmid pHCMC05, which carries the versions of ole and oapA genes used in the study. IPTG promotes expression of the various constructs, and is added to maintain uniformity in growth conditions even when an empty plasmid is present. Unless otherwise indicated, all liquid cultures (3 mL) were incubated aerobically at 37°C with shaking in 14-mL Falcon round-bottom polypropylene culture tubes.

Agar plates and agar diffusion assays

LB agar (pH 10) was prepared as described above with the addition of 1 mM IPTG, 3 µg mL⁻¹ chloramphenicol, and 10 mM MgCl₂ (where noted). Cells were streaked from cultures in exponential growth phase. For diffusion assays, 15 µL of either 2 M sodium citrate (pH 8), 2 M DL-isocitrate (pH 8), or 0.5 M EDTA (pH 8) was pipetted onto a filter disc (BD BBL Taxo 0.25-in blank discs), which was then placed on the agar. The cultures were incubated at 37ºC for 24 h.

Bacterial growth curve assays

Growth curves with B. halodurans strains were obtained using a Bioscreen C instrument (Growth Curves USA). Cells were first grown overnight in 3 mL of LB (pH 10) with 3 µg mL⁻¹ chloramphenicol, and 1 mM IPTG. OD₆₀₀ values were determined using a Cary 60 UV-Vis Spectrophotometer (Agilent Technologies). These cultures were then diluted to 0.01 OD₆₀₀ in 3 mL fresh media and incubated for 3 h. Stock solutions of media containing MgCl₂ were prepared at 1.25x the final concentrations (0, 2.5, 5, 7.5,10, 12.5 mM) using autoclaved 1 M MgCl₂ and LB (pH 10) containing 3 µg mL⁻¹ chloramphenicol and 1 mM IPTG. For each subsequent culture, 160 μL of these media were pipetted into the wells of a Bioscreen plate. Culture mixtures, in triplicate, were inoculated with 40 µL of a solution of cells diluted to 0.05 OD₆₀₀ in LB (pH 10) with 3 µg mL⁻¹ chloramphenicol and 1 mM IPTG. This process yields a final volume of 200 µL in each well, wherein cell density is ~0.01 OD₆₀₀. Cultures were incubated with continuous shaking at 37°C for 30 h. OD₆₀₀ readings were recorded every 30 min. Growth analyses conducted under cold and ethanol stress conditions were performed as previously described (Harris et al., 2018).

To rule out contamination of WT cells in the KO and PM1 cultures, cells were taken from the Bioscreen plate wells and the identity of each strain was confirmed by PCR and sequencing of the region on the chromosome and the plasmid where ole-oapA reside.

Genetic selection for suppressor mutations

B. halodurans cells were grown as described above with slight modifications. After the 3-h recovery growth period, cells were diluted in LB (pH 10) containing 3 µg mL⁻¹ chloramphenicol, 1 mM IPTG, and 8 mM MgCl₂ to 0.05 OD₆₀₀ and aliquoted into 96-well plates using 300 µL culture volumes. Cultures were incubated on a shaker at 37°C for 48 h. Cells from wells that contained turbid cultures were streaked onto LB (pH 10) agar plates containing 3 µg mL⁻¹ chloramphenicol, 1 mM IPTG, and 8 mM MgCl₂. These plates were incubated overnight at 37°C. A single colony from each population was selected, inoculated into media, grown overnight, and stored as a glycerol stock.

Genomic DNA was isolated from each strain using the ZR Fungal/Bacterial DNA Miniprep kit (Zymo Research) according to manufacturer’s instructions. If necessary, the DNA was further purified with the gDNA Clean & Concentrator kit (Zymo Research) according to manufacturer’s instructions. DNA library preparation and genome sequencing was performed by the Yale Center for Genome Analysis using an Illumina NovaSeq with paired-end, 150 base-pair reads. Reads were aligned and analyzed using breseq (Deatherage and Barrick, 2014) with NC_002570.2 as the reference genome.