Lifespan extension and paraquat resistance in a ubiquinone-deficient Escherichia coli mutant depend on transcription factors ArcA and TdcA

ArcA-dependent lifespan extension in the ubiG mutant

The excretion of amines, as a by-product of the metabolism of amino acids in LB, results in an alkaline pH (approximately 9) as a wt strain reaches stationary phase in this medium [12]. On the other hand, the ubiG strain reaches a stationary phase pH of approximately 7 (data not shown). Also, the saturating cell density of the wt strain is approximately 1.5 × 10⁹ colony forming units (CFU) per ml, while the ubiG mutant reaches a density of approximately 8 × 10⁸ CFU / ml (Figure 1E). The ubiG strain survives considerably longer than wt when no changes are made in the media of the two strains (Figure 1A). To dissect the effect of differences in stationary phase pH and cell density on the observed lifespan extension, we equalized stationary phase pH to 9 using the biological buffer AMPSO and stationary phase cell density to 1.5 × 10⁹ CFU / ml for both strains. Though diminished in comparison to the unbuffered experiment, lifespan extension in the ubiG strain under these conditions is still significant (Figure 1B). On the other hand, when stationary phase cell density is equalized to 1.5 × 10⁹ CFU / ml and stationary phase pH is equalized to 7.5 using the biological buffer HEPES, the lifespan of the ubiG strain is similar to that of wt (Figure 1C). Therefore, the ubiG mutation causes a lifespan extension only in an alkaline environment and we performed all subsequent experiments under these conditions. It is important to note that E. coli is exposed to alkaline pH in one of its secondary ecological habitats, seawater [13]. The growth curves of wt and the ubiG strain demonstrate that both strains enter stationary phase (cease proliferating) at approximately the same time (Figure 1E). Therefore, the observed lifespan extension in the ubiG mutant is not a result of delayed entry into stationary phase. Lack of ubiG also causes extended survival in the commonly used genetic background MG1655 (Figure 1D). It is worth noting that the stationary phase survival of the ubiE and ubiF mutants is intermediate between that of wt and the ubiG mutant (data not shown).

Figure 1. The ubiG mutant survives longer than wt. (A-C) Survival curves of the indicated strains under the shown conditions. (D) Fraction of surviving cells at 24 hours in stationary phase for wt and the ubiG mutant in the MG1655 genetic background. (E) Growth curves of the shown strains. (F) Survival of wt and the ubiG mutant under normoxia or hypoxia. (G) Survival of the arcA mutant under normoxia or hypoxia.

Prompted by our observations with the lipA and lpdA strains [5], we tested the effect of arcA deletion on the survival of the ubiG mutant. Lack of arcA has no effect on the survival of wt, but causes a significant shortening of the lifespan of the ubiG strain, rendering it closer to that of the wt strain (Figure 1B). Therefore, the hypoxia-inducible transcription factor ArcA is required for the fully extended lifespan of the ubiG mutant. Also, we have previously shown that a constitutively active arcB allele (the upstream activator of ArcA) results in an ArcA-dependent lifespan extension [5]. Based on these results, we tested the effect of hypoxia on the survival of the two strains, by incubating them in conical flasks with different culture:air ratios (see Methods). Confirming previous findings [14], we found that the survivorship of the wt strain under hypoxic conditions is approximately 2-fold higher compared to normoxia at both 18 and 24 hours (Figure 1F). On the other hand, hypoxia causes an approximately 50% increase in survivorship only at one time-point during the survival experiment of the ubiG strain (36 hours, Figure 1F). We also found that the survival of the arcA mutant is not different under hypoxic and normoxic conditions, based on four independent replicates per condition, analysed using the student's t test at a signifiance level of p<0.05 (Figure 1G). The dependence of the fully extended lifespan of the ubiG mutant on the hypoxia-inducible transcription factor ArcA, as well as the decreased magnitude of the hypoxia-induced lifespan extension in the ubiG mutant compared to wt are consistent with the hypothesis that the genetic simulation of hypoxic conditions in the ubiG strain is partially responsible for the observed lifespan extension.

Physiological changes in the ubiG mutant

Ubiquinone mediates the transport of electrons from several electron donors to oxygen and other electron acceptors, such as nitrate in the electron transport chain of E. coli [15]. Consistent with these observations, we found that the ubiG mutant consumes oxygen at a rate that is one order of magnitude lower than wt during stationary phase. In the absence of arcA, the rate of oxygen consumption of the mutant doubles (Figure 2A). Therefore, it appears that the reduced rate of oxygen consumption in the ubiG strain is mostly due to the impaired synthesis of ubiquinone and partly due to the action of ArcA, presumably on gene expression.

Figure 2. ArcA-regulated physiological changes in the ubiG strain. (A) Rate of oxygen consumption of the shown strains in stationary phase. (B) Extracellular acetate accumulation of the shown strains at stationary phase. (C) Survival curves of the shown strains.

Acetate is one of the four metabolic by-products of E. coli under fermentative conditions, along with lactate, ethanol and succinate [16]. Since the ubiG mutant appears to adopt a hypoxia-like metabolism and because we previously showed that acetate production and uptake are required for the extended lifespan of the lipA strain [5], we measured the concentration of extracellular acetate in the ubiG strain and found it to be higher than that of wt. Deletion of arcA reverts the extracellular acetate concentration of the mutant back to wt levels (Figure 2B). Acetyl-coA synthetase, encoded by acs, is responsible for the uptake of acetate at extracellular concentrations lower than 10 mM and its conversion to acetyl-CoA [17]. Deletion of acs does not affect the lifespan of the ubiG mutant (Figure 2C); hence acetate uptake is not required for the longevity of this strain.

Long-lived organisms in several model systems have been shown to be resistant to thermal and oxidative stress [18]. The ubiG mutant is more resistant than wt both to heat shock and to treatment with the superoxide-generating agent paraquat in stationary phase. The paraquat resistance of the mutant is entirely dependent on ArcA, whereas the heat shock resistance is independent of ArcA (Figure 3). Note that ArcA also contributes to the paraquat resistance of the wt strain, albeit to a much smaller extent compared to the ubiG mutant. Therefore, ArcA regulates the lifespan, rate of oxygen consumption, extracellular acetate accumulation and paraquat resistance in the ubiG mutant.

Figure 3. The ubiG mutant is stress-resistant. (A) Survival of the shown strains after a 5′ incubation at 54°C. (B) Survival of the shown strains after a 12-hour incubation with 300 μM paraquat.

Transcriptional regulation of lifespan

Since ArcA is a transcription factor, we performed a microarray analysis in order to explore its effect on stationary phase gene expression in the ubiG strain. The complete dataset is available as an online supplementary file. Hundreds of genes were found to be both significantly downregulated in the ubiG strain compared to wt and also significantly upregulated in the ubiG arcA strain compared to the ubiG strain. In order to communicate the function of these genes more effectively, we used the groups suggested in the Kyoto encyclopedia of genes and genomes [19]. As shown in Table 2, ArcA causes the downregulation of 16 different functional groups of genes in the ubiG mutant. Among them, the group of oxidative phosphorylation, an observation consistent with the aforementioned role of ArcA in suppressing oxygen consumption in the ubiG mutant (Figure 2A).

Although ArcA predominantly suppresses the expression of genes required for oxidative metabolism, a small number of genes / operons are known to be activated by this transcription factor in response to hypoxia [6]. We found no gene ontology group to be specifically upregulated by ArcA in the ubiG mutant.

However, several genes displayed both a higher-than-2-fold upregulation in the ubiG strain compared to wt and a higher-than-2-fold downregulation in the ubiGarcA strain compared to the ubiG strain. We chose to confirm the expression of seven of those genes by RT PCR, selecting these genes based on their involvement in well-described metabolic pathways. As shown in figure 4 and confirming the microarray findings, the expression of dcuS, fldB, iclR and tdcA is enhanced in the ubiG mutant by ArcA. The expression of gatC, feoB and cybB displayed a similar behavior (data not shown.)

Figure 4. ArcA enhances the expression of several genes in the ubiG mutant. (A-D) Expression of the shown genes relative to frr in the respective strains. The expression of the ribosome recycling factor encoded by frr was similar across the tested strains.

In order to explore any effects of ArcA-regulated gene expression on the lifespan of the ubiG strain, we tested the effect of the deletion of the ArcA-upregulated genes on the survival of the strain. Deletion of dcuS, fldB, gatC, feoB, cybB or iclR has no effect on the survival of the mutant (data not shown). On the other hand, deletion of tdcA causes a partial reversion of the extended lifespan of the ubiG mutant, while the tdcA strain has a stationary phase survival similar to wt (Figure 5A). TdcA is a transcription factor regulating the anaerobic metabolism of serine and threonine [20]. We next explored the genetic interaction between ArcA and TdcA in the regulation of the lifespan and paraquat resistance of the ubiG mutant. As shown in figure 5B, the ubiG arcA tdcA mutant has a similar survival to the ubiG arcA strain, consistent with the hypothesis that ArcA and TdcA function in the same pathway to extend the lifespan of the ubiG mutant. TdcA also contributes partially to the paraquat resistance of the ubiG strain, since the ubiG tdcA mutant is more resistant than wt, but less resistant than the ubiG strain to a stationary phase paraquat treatment; TdcA has no effect on the paraquat resistance of wt (Figure 5C). The ubiG arcA tdcA mutant displays similar paraquat resistance to the ubiG arcA strain. Therefore, the data presented in figures 5B and 5C are pointing towards a single genetic pathway formed by the transcription factors ArcA and TdcA enhancing both stationary phase survival and paraquat resistance in the ubiG mutant. Furthermore, the regulation of both paraquat resistance and lifespan by a common genetic pathway suggests that these two phenotypes are physiologically interdependent.

Figure 5. ArcA enhances lifespan and stress resistance through TdcA. (A, B) Survival curves of the shown strains. (C) Survival of the shown strains after a 12-hour incubation with 300 μM paraquat. (D) Expression of arcA relative to rpoA in the strains indicated. The expression of the RNA polymerase encoded by rpoA was similar across the tested strains.

To test whether the expression of arcA is under the control of TdcA, we measured the expression of arcA in the wt, ubiG and ubiG tdcA strains. The expression of arcA is elevated in the ubiG mutant compared to wt and this elevation is not affected by the absence of TdcA (Figure 5D); therefore, the expression of arcA does not seem to be under the control of TdcA. The elevated expression of arcA in the ubiG mutant might be due to the derepression of the ArcB-ArcA phosphorelay in the absence of fully-formed ubiquinone [21].

Strains and genetic manipulations

The wild-type Escherichia coli strain BW25113, a K-12 derivative, and its respective single-gene knock-outs were provided by the Keio collection [4]. Each gene is knocked out by the insertion of a kanamycin-resistance cassette. To create strains deleted for multiple genes, the kanamycin cassette was excised using FLP-mediated recombination, resulting in deletions carrying only a single FRT site (denoted, for example ubiG::FRT in Table 1), as previously described [37], with the only difference that the non-selective incubation took place at 37°C and not at 43°C. Kanamycin alleles were transduced by bacteriophage P1 using standard techniques and the correct insertion was verified by PCR using primer K1 described in [4], along with a locus-specific primer annealing to a sequence upstream of the disrupted locus. Genomic sequence information was obtained from the “Profiling of the E. coli Chromosome” web site (www.shigen.nig.ac.jp/ecoli/pec/index.jsp). All strains used in the study are shown in Table 1.

Survival experiments

LB medium consisted of 1% bacto tryptone, 0.5% yeast extract and 0.5% NaCl w/v. Cultures were inoculated 1:1000 (vol:vol) using an overnight culture created by inoculating 2-3 colonies from an LB plate to 1 ml of LB. Cultures were grown in 3 ml's of LB in 16-mm diameter test tubes rotating orbitally at 220 rpm for 12 hours, at which point cell density was adjusted to ~1.5 × 10⁹ CFU per ml by resuspending a pellet containing the desired number of cells in cell-free spent medium of the same strain for the long-lived strains with reduced saturation cell density. AMPSO was added to 100 mM to achieve a stationary phase pH of 9. Due to the differences in stationary phase pH reached spontaneously by each strain, the pH of the cultures had to be adjusted accordingly for each strain. Spontaneous and adjusted pH was quantified using a pH electrode and pH test strips. All cultures were grown and maintained at 37°C and 70% relative humidity and colony-forming units (CFU) were enumerated over time by removing an aliquot, serially diluting in 0.5% NaCl, followed by colony enumeration after plating on LB plates that were incubated at 37°C. For the hypoxia experiments, 30-ml cultures were incubated for 12 hours in orbitally shaking 125-ml Erlenmeyer flasks. The resulting stationary phase culture was split to two different cultures in 25-ml flasks, each with 100 mM AMPSO; one 5-ml culture (‘normoxic’) and one 28-ml culture (‘hypoxic’); note that the 25-ml notation is nominal, since the flask can actually hold approximately 30 ml. CFU's were enumerated over time as described above.

Stress resistance

For the heat shock experiment, cultures were processed as described above with AMPSO, returned to the incubator and 4 hours later (16 hours after inoculation) subjected to either a 5-minute incubation in a 54°C waterbath without shaking and CFU enumerated before and after the treatment.

For the paraquat experiment, cultures were processed as described above with AMPSO and paraquat (methyl viologen dichloride hydrate, Sigma-Aldrich) was added to 300 μM at the time of processing (12 hours after inoculation). CFU were enumerated before and after a 12-hour incubation at 37°C.

Oxygen consumption and acetate quantification

Oxygen consumption measurements were performed with 2-ml of culture stirred by a magnetic stir bar in a 37°C waterbath using a Clark-type electrode. Conversion to nanomoles of oxygen consumed was done by assuming that the liquid culture contains the same amount of oxygen as water equilibrated with 21% oxygen in 1 atmosphere pressure, which is 5.02 μl/ml (manufacturer's manual) and was further normalized by the number of CFU's present. Data were recorded until a straight line trace was obtained indicating that a steady state of oxygen consumption had been reached. Extracellular acetate concentration was quantified on cell-free samples obtained by centrifugation using the R-Biopharm acetic acid kit (catalog number 10148261035) according to the manufacturer's instructions.

Microarrays and RT PCR

Stationary phase cultures of wt and the ubiG and ubiG arcA mutants were prepared as mentioned in the ‘survival experiments’ section. Cultures were removed from the incubator 45 min. after the addition of AMPSO and 1-ml aliquots were added to a 95μl ethanol + 5μl water-equilibrated acidic phenol mixture and rapidly centrifuged for 45 seconds at 4°C. RNA was subsequently extracted using the MasterPure complete DNA and RNA purification kit (Epicentre Biotechnologies, catalog number MC85200) according to the manufacturer's instructions. The MicrobExpress kit (Ambion) was used to enrich the resulting RNA for mRNA, using the manufacturer's instructions. 200 ng of MicrobExpress RNA was used to generate double stranded cDNA using Ambion's MessageAmp II-Bacteria kit. This cDNA was subsequently used as a template for an in vitro transcription reaction to synthesize biotin-labeled amplified RNA in the presence of 10 mM biotin-11-CTP and 10 mM biotin-16-UTP. The in vitro transcription reaction volume was 20 μl and incubated for 14 hours. Fragmentation of the labeled, amplified RNA was performed according to the instructions found in the Affymetrix GeneChip Eukaryotic Expression Manual. 4 μg of the fragmented target was hybridized at 45°C with rotation for 16 hours at 60 rpm to probe sets present on an Affymetrix E. coli 2.0 array. The microarrays were washed, stained (streptavidin, phycoerythrin) and scanned on an Affymetrix fluidics station 450. Arrays were processed at the UC Irvine DNA & protein microarray facility.

2 microarrays were analysed per strain. After imaging process, the expression of each sample is represented by a CEL file, which includes the fluorescence intensities of all probes. We processed the microarray data using the invariant-set normalization [38] followed by median polish summarization [39], and obtained expression for each probe set. The computation was carried out in statistical software R using the Affy package [40]. The interpretations of gene expression changes were based on the gene set enrichment analysis. It determines whether a pre-defined set of genes shows significantly up- or down-regulation compared to other gene sets. We downloaded the KEGG pathway database from ftp://ftp.genome.jp/pub/kegg/pathway/organisms/eco/. It provides manually annotated metabolic and signaling pathways. We selected 102 well-established Escherichia coli pathways to define KEGG gene sets. We carried out the enrichment analysis followed by the Wilcoxon scoring scheme as previously described [41]. Briefly, based on the expression differentiation between two samples, we compared the fold changes of KEGG gene set Si against the background expression changes denoted by G-Si, where G is the union of all KEGG gene sets. We adopted the Wilcoxon rank test to calculate p-values (one-sided) and assigned the final significance scores after multiple testing correction using the “qvalue” package [42] in R.

To test the expression of dcuS, fldB, iclR, tdcA and arcA, RNA was extracted as described above. 3 μg of RNA were used per reverse transcription reaction using Superscript III reverse transcriptase (Invitrogen) and random hexamers as primers according to the manufacturer's instructions. 50 ng of reverse-transcribed RNA were then used as substrate for real-time PCR. The expression level of three different housekeeping genes was measured (rpoA, frr and dnaA) and the one showing similar expression levels among all tested strains was used for normalization. Standard curves were constructed for each assayed transcript and used for quantification. The primers used are shown in Table S2.