Redundancy in Citrate and cis-Aconitate Transport in Pseudomonas aeruginosa

ABSTRACT

Tricarboxylates such as citrate are the preferred carbon sources for Pseudomonas aeruginosa, an opportunistic pathogen that causes chronic human infections. However, the membrane transport process for the tricarboxylic acid cycle intermediates citrate and cis-aconitate is poorly characterized. Transport is thought to be controlled by the TctDE two-component system, which mediates transcription of the putative major transporter OpdH. Here, we search for previously unidentified transporters of citrate and cis-aconitate using both protein homology and RNA sequencing approaches. We uncover new transporters and show that OpdH is not the major citrate importer; instead, citrate transport primarily relies on the tripartite TctCBA system, which is encoded in the opdH operon. Deletion of tctA causes a growth lag on citrate and loss of growth on cis-aconitate. Combinatorial deletion of newly discovered transporters can fully block citrate utilization. We then characterize transcriptional control of the opdH operon in tctDE mutants and show that loss of tctD blocks citrate utilization due to an inability to express opdH-tctCBA. However, tctE and tctDE mutants evolve heritable adaptations that restore growth on citrate as the sole carbon source.

IMPORTANCE Pseudomonas aeruginosa is a bacterium that infects hospitalized patients and is often highly resistant to antibiotic treatment. It preferentially uses small organic acids called tricarboxylates rather than sugars as a source of carbon for growth. The transport of many of these molecules from outside the cell to the interior occurs through unknown channels. Here, we examined how the tricarboxylates citrate and cis-aconitate are transported in P. aeruginosa. We then sought to understand how production of proteins that permit citrate and cis-aconitate transport is regulated by a signaling system called TctDE. We identified new transporters for these molecules, clarified the function of a known transport system, and directly tied transporter expression to the presence of an intact TctDE system.

INTRODUCTION

The Gram-negative bacterium Pseudomonas aeruginosa thrives on many carbon sources, but unlike Escherichia coli and many other model bacterial species prefers tricarboxylic acid (TCA) cycle intermediates such as succinate or citrate over glucose and other saccharides (1). The import, modification, and metabolism of glucose in P. aeruginosa have been extensively studied as a bacterium lacking the textbook Embden-Meyerhof-Parnas (EMP) glycolytic pathway (2). This organism also lacks sugar phosphotransferase systems (PTS) for many carbon sources, instead using a series of OprD family outer membrane porins that are typically induced by their respective substrates (3). The use of specific porins by P. aeruginosa has been linked to lower membrane permeability and, along with the presence of efflux pumps, bacterial tolerance to antibiotic treatment (4). Because P. aeruginosa is responsible for many hospital-acquired infections (5) and chronically infects the lungs of cystic fibrosis patients (6, 7), it is important to characterize membrane transport as a step toward learning how transport may impact antibiotic tolerance.

Despite citrate being a preferred carbon source for P. aeruginosa (8–10), its transport is not well understood in this organism. A transposon-insertion mutant of the opdH porin-encoding gene showed a growth defect on cis-aconitate (3), a molecular neighbor to citrate in the TCA cycle. The opdH gene was additionally induced on citrate but not glucose, yet a mutant showed no growth defect on citrate (11). OpdH is a predicted porin and showed channel-like activity in a planar bilayer apparatus, but transport of citrate or cis-aconitate could not be demonstrated in these experiments (11). The opdH gene is in an operon with the tctCBA genes, which encode a predicted tripartite tricarboxylate transport system. This system has not previously been tested for transport of citrate in P. aeruginosa (11), though the homologous TctC subunit is known to bind citrate, isocitrate, and cis-aconitate in Salmonella enterica (12). The TctA and TctB proteins in Salmonella enterica serovar Typhimurium are predicted transmembrane proteins that are putatively involved in the cross-membrane transport of the substrate bound by TctC (13). In P. aeruginosa, the opdH-tctCBA operon is divergently transcribed from the tctDE operon, which encodes a two-component system (TCS) that is thought to repress opdH-tctCBA expression via binding of the TctD response regulator to the promoter region (11, 14). A tctDE mutant reportedly failed to grow on citric acid as a sole carbon source (15), which raises the question of how the citrate-induced opdH-tctCBA operon behaves in citrate when its repressor is absent. If TctDE is a simple TCS that senses TCA cycle intermediates and responds by derepressing opdH-tctCBA, loss of the TctD repressor would result in constitutive expression of those genes rather than loss of growth in citrate. The opdH-tctCBA operon also appears to encode another protein, PA14_54580, which has not been previously examined. PA14_54580 is annotated as a conserved hypothetical protein of unknown function, though one report (11) labeled it as a monooxygenase.

In S. Typhimurium, a citrate transporter named CitA was characterized and could facilitate transport of citrate in E. coli cells (16), which do not normally grow on citrate, as they lack a functional transporter (17–19). A previously uncharacterized homolog of the S. Typhimurium citA gene exists in P. aeruginosa in both PAO1 and PA14 (PA5476 in PAO1; PA14_72280 in PA14). While it was suggested that opdH may be required for citrate transport (15), the presence of a citA gene that is highly similar to a predicted major facilitator superfamily (MFS) transporter of S. enterica (59.9% identity, 74% similarity by the EMBOSS Needle alignment tool [20]) and citA of S. Typhimurium (59% identity, 72.5% similarity) suggests that citrate transport in Pseudomonas is more redundant than currently believed.

In the present study, we examine redundancy in and control of citrate and cis-aconitate transport in P. aeruginosa by using a series of genetic deletions coupled with growth screens and by complementing nonfunctional citrate and cis-aconitate transport in E. coli using transporters from Pseudomonas. We clarify the role of OpdH, previously thought to be the major porin for citrate (15), and identify several new transport proteins using transcriptomic analysis. We then use deletions of the tctDE TCS coupled with reverse transcription-quantitative PCR (RT-qPCR) to probe the impact of TctDE on expression of citrate transporter-encoding genes in the presence and absence of citrate.

RESULTS

opdH and/or citA deletions do not affect growth on citrate but modestly impact cis-aconitate growth.

Since the citA gene putatively encodes a citrate transporter, we constructed strains to test whether deletion of citA would affect the ability of P. aeruginosa to grow on citrate. We obtained growth curves in M9 minimal medium supplemented with citrate as the sole carbon source. The ΔcitA and ΔopdH strains showed no obvious growth defects relative to the wild type (Fig. 1A), in contrast to previous results using an opdH gentamicin cassette insertion mutant in strain PAO1 that resulted in a modest citrate growth defect and a severe aconitate growth defect (11). A ΔcitA ΔopdH double mutant showed a marginally longer lag and a slightly slower initial growth rate that were barely noticeable (Fig. 1A and C). On cis-aconitate as the sole carbon source, we again observed no growth defects for either ΔcitA or ΔopdH strains (Fig. 1B). This result also stands in contrast to previous reports of severe cis-aconitate growth defects in a transposon insertion or a Gm^r insertion opdH mutant of strain PAO1 (3, 11). However, the ΔcitA ΔopdH double mutant displayed a substantial lag in growth on cis-aconitate (~500 min versus ~200 min for the parent and single deletions; Fig. 1B and D), suggesting that these transporters are more important for cis-aconitate transport than for citrate transport and are not the sole transporters of either molecule.

FIG 1.

Effects of citA and opdH on citrate and cis-aconitate transport. Growth curves of PA14, ΔopdH, ΔcitA, or ΔopdH ΔcitA as indicated in M9 with 7.5 mM citrate (A) or 7.5 mM cis-aconitate (B) as the sole carbon source. The data are representative curves of at least three independent experiments. (C and D) Mean lag times of ≥3 biological replicates corresponding to the strains from panels A and B, respectively. Error bars indicate the standard errors of the mean (SEM).

CitA, but not OpdH, enables citrate and cis-aconitate transport in E. coli.

E. coli MG1655 is a K-12 derivative that, like other K-12 derivatives, does not transport citrate and therefore does not grow on citrate as a sole carbon source. We took advantage of this property to test whether expression of citA or opdH under IPTG-inducible control would be sufficient to enable transport of citrate. As expected, MG1655 grew in M9 minimal medium supplemented with glucose but not in M9 with citrate (Fig. 2A). As a positive control, an MG1655 strain constitutively expressing citT, which encodes an E. coli citrate transporter, from a plasmid (gift of Profs. Zach Blount and Richard Lenski [21]) achieved substantial growth in citrate (Fig. 2A). Next, we individually expressed citA or opdH from an IPTG-inducible pTrc99A plasmid in MG1655. Neither plasmid permitted growth without IPTG (Fig. 2A). When induced, citA, but not opdH, enabled modest growth on citrate (Fig. 2A). We also performed the same complementation test in M9 containing cis-aconitate, which, like citrate, cannot be utilized by the MG1655 parent (Fig. 2B). Interestingly, expression of citA enabled substantial growth of E. coli on cis-aconitate, whereas expression of either opdH or citT did not allow growth (Fig. 2B). These results imply that CitA, unlike CitT, can transport both citrate and cis-aconitate but prefers cis-aconitate.

FIG 2.

Expression of citA, but not opdH, facilitates growth of E. coli on citrate and cis-aconitate. Growth curves of E. coli MG1655 bearing different transporter-expressing plasmids growing in M9 with 7.5 mM citrate or 7.5 mM cis-aconitate as the sole carbon source are shown. In panels A to D, MG1655 growing in 3 mM glucose as a positive control is shown by the dashed black curves. (A and B) Growth in citrate (A) or cis-aconitate (B) of MG1655 (WT), MG1655 expressing the CitT transporter as a positive control (black curves), or MG1655 carrying pTrc99A with citA or opdH with or without 100 μM IPTG, as indicated. (C and D) Growth in M9-citrate without (C) or with (D) 100 μM IPTG in strains bearing fragments of the opdH operon as noted. The data are representative curves of at least three independent experiments. (E) Mean lag times of ≥3 biological replicates corresponding to selected strains from panels C and D. Error bars indicate the SEM.

Given that previous insertional mutants of opdH in PAO1 (3, 11) caused citrate and cis-aconitate growth defects, and because OpdH alone did not enable citrate or cis-aconitate utilization (Fig. 2A and B), we considered the possibility that the downstream tctCBA genes (which might have been subject to polar effects by insertional mutation) have a role in tricarboxylate transport. Thus, we placed different sections of the opdH operon on pTrc99A for expression in E. coli and growth analyses. Strikingly, even in the absence of IPTG (isopropyl-β-d-thiogalactopyranoside) induction, a construct encoding the TctCBA tripartite transporter and OpdH porin grew slowly, whereas none of the other tested operon fragments grew (tctCBA, tctCBA-PA14_54580, and opdH-tctCBA-54580) (Fig. 2C and E). IPTG induction accelerated the growth of the strain bearing the opdH-tctCBA construct, and the strain encoding only the putative TctCBA transporter also grew, though not as quickly as with opdH (Fig. 2D and E). Interestingly, the constructs also bearing the downstream PA14_54580, a gene of unknown function, failed to permit growth in citrate irrespective of induction despite the presence of tctCBA and opdH (Fig. 2C and D). Evidently, expression of PA14_54580 prevents growth of E. coli by an unknown mechanism when coexpressed with these transporters.

Identification of new citrate and aconitate transporters in P. aeruginosa.

Our findings that citA and/or opdH deletion do not abolish growth on citrate or cis-aconitate (Fig. 1) indicate that there must be at least some redundancy in citrate and cis-aconitate transport in P. aeruginosa. Furthermore, complementation tests with E. coli (Fig. 2) suggest that OpdH, previously thought to be the major transporter of citrate and cis-aconitate in P. aeruginosa, does not act on its own but likely requires a partner to transport these solutes. To identify previously unknown citrate transporters, we performed transcriptomics to compare PA14 samples grown in either synthetic cystic fibrosis medium 2 (SCFM2) (22) or SCFM2 plus 7.5 mM citrate, reasoning that transporters are often induced by their substrate. We observed numerous differentially regulated genes (Fig. 3A). In Table 1, we show highly citrate-upregulated genes that were also marked as transporters; in principle, some of these may transport citrate and/or cis-aconitate in PA14. Indeed, this list of genes includes opdH and tctA.

FIG 3.

Effect of deletion of newly identified transporters on citrate growth. (A) Volcano plot of differentially regulated genes in the presence versus the absence of 7.5 mM citrate. A threshold false discovery rate (FDR) P value of 10⁻⁵ was set to denote significant regulation. Genes listed in Table 1 are shown as open circles, and genes subjected to genetic analyses are labeled. (B-C) Growth curves in M9-citrate of strains deleted for putative new citrate transport protein-encoding genes, singly (B) or in combination (C) as indicated. The data are representative curves of at least three independent experiments. (D) Mean lag times of ≥3 biological replicates corresponding to selected strains from panels B-C. Error bars indicate SEM.

TABLE 1.

Selected genes most upregulated on SCFM2 plus 7.5 mM citrate compared to SCFM2 without citrate and annotated as a type of transporter^a

Locus tag	Gene	Annotation	FDR	Log2 FC
PA14_18320	NA	EamA family transporter	1.63 × 10−16	5.73
PA14_54520	opdH	OprD family porin	2.34 × 10−17	4.96
PA14_72170	NA	CitMHS family transporter	4.83 × 10−25	4.49
PA14_54570	tctA	Tripartite tricarboxylate transporter permease	1.49 × 10−37	3.63
PA14_47560	NA	MFS transporter	6.01 × 10−13	2.27
PA14_17610	NA	Spermidine/putrescine ABC transporter substrate-binding protein	1.08 × 10−12	2.55
PA14_17620	NA	ABC transporter permease	2.84 × 10−14	3.16
PA14_17630	NA	ABC transporter permease	2.42 × 10−14	2.86
PA14_17640	NA	ABC transporter ATP-binding protein	1.78 × 10−16	2.52
PA14_33910	NA	ABC transporter substrate-binding protein	1.46 × 10−17	2.56
PA14_21150	NA	ABC transporter permease	2.27 × 10−9	1.84
PA14_72960	NA	MFS transporter	4.59 × 10−11	1.75
PA14_19160	rarD	EamA family transporter RarD	8.62 × 10−12	1.60

^aNA, not applicable; FDR, false discovery rate; FC, fold change.

Guided by our transcriptomic data, we deleted selected suspected transporters to assess their function in citrate or cis-aconitate utilization. Deletions of the tctA gene (PA14_54570), the two identified genes encoding EamA-family proteins (PA14_18320 and rarD), as well as PA14_72170, were all assayed for growth defects on citrate in liquid M9 culture. We found that deletion of 18320, 72170, or rarD alone had no significant impact on growth in citrate (Fig. 3B) or cis-aconitate (see Fig. S1A in the supplemental material). In contrast, the ΔtctA strain showed a substantially longer lag time in citrate (Fig. 3B and D) and completely failed to grow on cis-aconitate (see Fig. S1A). All of these mutants grew similarly to PA14 on M9 with 3 mM glucose (see Fig. S1B), implying that the growth changes are specific to tricarboxylate carbon sources.

We then examined combinatorial deletions of tctA with the other identified transporters to learn whether we could build a strain that completely failed to grow on citrate. Deletion of 72170 with tctA resulted in a greater lag (Fig. 3C and D), and further deletion of either 18320 or rarD completely abolished citrate growth (Fig. 3C). The similar impact of these deletions at least suggests that 18320 and RarD may act together, such that loss of either has the same effect on citrate transport.

The situation appeared more complicated on cis-aconitate. While a ΔtctA single mutant did not grow, the ΔtctA Δ72170 deletion did grow, but with a substantial lag of ~18 h relative to the wild type (see Fig. S1C and D). Further deletion of 18320 increased the lag to roughly 55 h but did not abolish growth, whereas the ΔtctA Δ72170 ΔrarD triple mutant grew very poorly on cis-aconitate, with a barely detectable increase in optical density at 600 nm (OD₆₀₀) after ~60 h (see Fig. S1D). The 18320 and rarD deletions yielded similar results in a ΔcitA ΔopdH background (see Fig. S2), agreeing with our conclusion from Fig. 2 that neither CitA nor OpdH is critical to transport.

To test whether any of the newly identified transporters was sufficient for citrate transport, we ectopically expressed each of the same four genes of interest from pTrc99A in E. coli MG1655. Irrespective of IPTG induction, only 72170 supported E. coli growth on M9-citrate (Fig. 4). None of the proteins enabled growth in cis-aconitate (see Fig. S3). This result suggests that the 72170 protein is a standalone citrate (but not cis-aconitate) transporter, in accord with the reduced growth of P. aeruginosa on citrate when 72170 was additionally deleted in a ΔtctA background (Fig. 3C). Moreover, the failure of either 18320 or rarD to support E. coli growth in tricarboxylate (Fig. 4; see also Fig. S3 in the supplemental material) is consistent with the above-hypothesized scenario in which their protein products work together (Fig. 3C).

FIG 4.

PA14_72170, but not other identified transporters, can facilitate E. coli growth on citrate. (A and B) Growth curves of E. coli MG1655 in M9-citrate carrying pTrc99A constructs encoding putative new citrate transporters as indicated. Genes were either uninduced (A) or induced with 100 μM IPTG (B). Data are representative curves of at least three independent experiments. (C) Mean lag times of ≥3 biological replicates corresponding to the indicated strains from panels A and B. Error bars indicate the SEM.

Deletion of tctDE causes a lag in citrate growth followed by heritable adaptation.

A previous study found that markerless deletion of the tctDE TCS from the genome of P. aeruginosa resulted in a growth defect on M63 medium using arginine and citric acid as combined carbon sources (15). TctD is thought to repress expression of the opdH operon in the absence of tricarboxylates (11); repression is presumably relieved by the TctE kinase via extracellular sensing of a tricarboxylate substrate. We deleted tctD and tctE independently and in combination to probe how growth in citrate and cis-aconitate is affected by each component. Using the previous model as guidance, we expected to lose repression of opdH when tctD was deleted, resulting in citrate growth. All three tctD/E mutants grew normally on glucose (see Fig. S4A) and similarly on a mixture of citrate and glucose, with PA14 growing to a higher OD₆₀₀ on the mixture (see Fig. S4B), consistent with failure of the tct mutants to use citrate. Surprisingly, we observed that ΔtctD mutants were unable to grow in citrate, whereas the ΔtctE and ΔtctDE mutants both grew but after a long lag (Fig. 5A and D; approximately 25 h for ΔtctE and 33 h for ΔtctDE). A control, in which 2 μL of the phosphate-buffered saline (PBS) used to wash the cells, was inoculated into a well of M9-citrate, ensured that this late growth was not due to contamination (Fig. 5A).

FIG 5.

Adaptation of tctDE mutants to growth on citrate. (A-C) Growth curves of P. aeruginosa strains deleted for genes encoding the TctDE two-component system. (A) Growth of PA14 (black), ΔtctD (light gray), ΔtctE (dark gray), and ΔtctDE (dashed) on M9 with 7.5 mM citrate as the sole carbon source. (B) The ΔtctE (black curve) and ΔtctDE (light gray curve) strains that grew in panel A were back-diluted from M9-citrate culture into fresh M9-citrate and tracked for growth. (C) Growth data for PA14 (black), ΔtctE (dark gray), and ΔtctDE (dashed) strains passaged on citrate and subsequently streaked on LB plates, grown in liquid LB, and then washed and diluted into M9-citrate. The data are representative curves of at least three independent experiments. (D) Mean lag times of ≥3 biological replicates corresponding to the indicated strains in panels A and C. R1 and R2 refer to adapted strains arising in separate cultures and independently passaged. Error bars indicate the SEM.

To learn whether the growth after the lag was due to acquisition of a heritable adaptation, we grew ΔtctE and ΔtctDE mutants to turbidity in shaking flasks in M9-citrate before two washes in PBS and back-dilution into M9-citrate. Both strains resumed exponential growth after just 3 to 4 h (Fig. 5B), a vast reduction compared to their initial lag times and more similar to wild-type cells (Fig. 5A). We then tested whether this adaptation was physiological or genetic by growing three replicates of each background in M9-citrate until they grew, then passaging them on Luria-Bertani (LB) agar plates to obtain single colonies and subsequently culturing them in liquid LB medium. These cells were frozen as stocks for later use, washed twice in PBS, diluted into M9-citrate, and tracked for growth. It was immediately evident that the adapted strains grew more quickly in citrate than the parental strains (Fig. 5C and D; we show one representative strain in Fig. 5C for simplicity). The adapted ΔtctDE strain showed the greatest degree of growth acceleration, with a lag of ~8 h compared to a parental lag of 33 h; the ΔtctE strain only accelerated from ~25 h to an ~17-h lag time.

Because cis-aconitate transport appears to overlap significantly with citrate transport and because the tctDE mutants showed the same growth pattern in cis-aconitate as in citrate (see Fig. S5A), we also grew the citrate-adapted strains in M9-cis-aconitate to determine whether they displayed accelerated growth. The growth curves were almost identical to those for M9-citrate in lag time (see Fig. S5B and C), with significant acceleration relative to the parental strains in cis-aconitate. Together, these data suggest that passaging the ΔtctE and ΔtctDE strains on citrate as the sole carbon source generates a heritable adaptation to growth on both citrate and cis-aconitate.

Adapted tctE and tctDE mutants growing on citrate regain opdH expression that is not mimicked by a glucose-citrate mixture.

The unexpected heritable adaptation of the tct mutants to growth in citrate and cis-aconitate led us to ask how tctDE might regulate the expression of the citrate transporters we were studying. We therefore examined transcript levels of the transporter-encoding genes, performing RT-qPCR on RNA extracted from mid-log-phase PA14 and each of our tct mutants growing in M9 supplemented with glucose, citrate, or a mixture of both. The mixture of glucose and citrate was important for probing expression in the ΔtctD mutant, which does not grow on citrate alone. We used the parental (i.e., unadapted) ΔtctE and ΔtctDE strains for this analysis.

The first genes we examined by qPCR were tctD and tctE themselves. This analysis showed that expression of tctD and tctE was not affected by the carbon source and confirmed that tctD and tctE were indeed deleted in the mutant strains (Fig. 6A and B). Our mutation of tctD appears to have potentially lowered the expression of tctE, albeit quite modestly (Fig. 6B). However, the ΔtctE strain showed normal or modestly elevated expression of tctD that, unlike in the wild type, appeared to be induced by the presence of citrate (Fig. 6A).

FIG 6.

tctDE system mutants do not express opdH operon genes highly in the presence of citrate. RT-qPCR data show the gene expression per 16S rRNA unit. In panels A to E, colors represent PA14 (white bars), ΔtctD (red bars), ΔtctE (blue bars), or ΔtctDE (yellow bars). In these panels, experiments were performed in citrate (cit), glucose (glc), or a mixture of citrate and glucose (mix). (A) tctD transcripts/16S rRNA; (B) tctE transcripts/16S rRNA; (C) opdH transcripts/16S rRNA; (D) tctA transcripts/16S rRNA; (E) citA transcripts/16S rRNA. (F) Expression of opdH in citrate-passaged ΔtctE mutants (replicate 1, white bars; replicate 2, red bars) or ΔtctDE mutants (replicate 1, blue bars; replicate 2, yellow bars). Error bars in panel F are light green because they are otherwise too small to be visible. Error bars represent the calculated measurement error (see Materials and Methods for more details). Significance between indicated pairs was evaluated by Student’s t test (*, P < 0.05; **, P < 0.01; NS, not significant [P > 0.05]).

We next assessed opdH expression, as our transcriptomic analysis identified it as a citrate-upregulated gene. Indeed, in PA14, expression of opdH is specifically induced by citrate, even in combination with glucose, since citrate is preferred over glucose by P. aeruginosa (1) (Fig. 6C). The ΔtctD mutant, however, failed to induce opdH in a citrate-glucose mixture, as expected given the failure of the ΔtctD strain to grow in citrate (Fig. 5A). Similarly, opdH was not induced by citrate/glucose mixtures in the ΔtctE and ΔtctDE backgrounds (Fig. 6C). However, when grown in citrate alone, the ΔtctE and ΔtctDE strains regained expression of opdH, consistent with their eventual ability to grow on citrate (Fig. 6C). The same patterns of induction for the citrate/glucose mixture versus citrate alone were observed for tctA expression (Fig. 6D), suggesting that the whole opdH-tctCBA operon is expressed in tct mutants that gain the ability to grow on citrate.

Next, we tested whether citrate or the tctDE system are able to affect expression of citA. We expected citA to be affected by both, since control of citA by citrate and TctDE might explain how P. aeruginosa can encode a functional CitA standalone citrate transporter but tctD mutants cannot grow on citrate. Surprisingly, however, we found that citA was not induced on citrate (Fig. 6E). Moreover, citA was unaffected by the presence of the TcdDE TCS, and indeed appears to be barely transcribed at all on citrate or glucose (Fig. 6E), explaining why citA deletion has no citrate growth phenotype (Fig. 1A).

Finally, we sought to formally test whether our citrate-adapted and LB medium-passaged ΔtctE and ΔtctDE strains had retained citrate-induced opdH expression or whether they now constitutively expressed the opdH-tctCBA operon. Hence, we examined transcription of opdH in two passaged strains of each parental background grown in either citrate or glucose. We clearly saw that opdH was expressed less in glucose than in citrate across all samples, though the transcript per 16S rRNA count was five times higher than in the parental strain in glucose (Fig. 6C). In fact, even in glucose the passaged strains showed opdH expression levels that were roughly equal to the levels in citrate-grown wild-type cells (Fig. 6F versus Fig. 6C). Collectively, these results suggest that the restoration of citrate growth in in the adapted ΔtctDE and ΔtctE strains is a product of much stronger basal opdH-tctCBA operon expression that nonetheless retains a degree of inducibility by citrate.

In an attempt to identify a genetic change conferring restored opdH operon expression and citrate growth, we sequenced the genomes of two independently passaged ΔtctDE strains and one ΔtctE strain and compared them to their respective parents. In all three passaged strains, single nucleotide polymorphisms (SNPs) not present in the parent were located in a region between a 16S rRNA gene and the annotated sbcD exonuclease gene. For both ΔtctDE passaged mutants, nucleotide 4957604 was changed from a T to a C; in the ΔtctE passaged mutant, there were two changes: 4957532 A to G and 4957549/4957550 GC to AT. In both of the ΔtctDE passaged mutants, there was also a frameshift in similar, C-terminal regions of the pilY1 fimbrial biogenesis gene. Since neither of these positions has any salient connection to citrate utilization, the significance of these SNPs with respect to the ability of the passaged strains to use tricarboxylates as carbon sources presently remains unclear.

A ΔtctD mutant can grow in citrate if the opdH operon is expressed.

We had observed that the ΔtctD and ΔtctDE mutants showed substantially reduced opdH transcript levels in medium containing both glucose and citrate and that the adapted strains showed elevated opdH expression. Thus, we hypothesized that loss of tctD blocks citrate utilization at least in part by abrogating expression of the opdH-tctCBA-PA14_54580 operon, further implying that ectopic expression of one or more genes in the opdH operon might compensate for deletion of tctD. To test this notion, we inserted different genes from this operon into pJN105, a replicative plasmid with arabinose-inducible control. We transformed the plasmids into a ΔtctD strain and tracked the growth of the resulting strains in M9-citrate or cis-aconitate with or without the addition of 100 μM arabinose. We immediately observed that induction was not necessary for growth on M9-citrate with nearly all the constructs (Fig. 7A). Importantly, a ΔtctD strain with empty pJN105 did not grow in citrate or cis-aconitate with or without arabinose but did grow in M9 with 3 mM glucose and gentamicin, indicating that the contents of the plasmid are important and that arabinose is not a growth-supporting carbon source (see Fig. S6). All strains expressing tctCBA grew, and expression of additional genes in the operon improved growth relative to tctCBA alone (Fig. 7A to C). All of the constructs generally grew better with arabinose induction than without (Fig. 7A to C), but the very similar growth curves obtained for the full operon (Fig. 7A and B, blue curves) suggest that its basal expression from the plasmid is not rate limiting for growth. We also found that, unlike in E. coli, expression of citA alone did not permit growth of ΔtctD in citrate (Fig. 7A and B), indicating that CitA alone cannot facilitate citrate transport in P. aeruginosa. We observed indistinguishable results in cis-aconitate (see Fig. S7A to C) except that citA induction induced barely detectable growth (see Fig. S7B), consistent with CitA favoring cis-aconitate (Fig. 2B). Collectively, our results argue that the citrate utilization defect in a ΔtctD mutant is caused by lack of opdH operon expression and that the core tctCBA genes are sufficient to restore citrate and cis-aconitate transport but are dramatically abetted by opdH and 54580 expression.

FIG 7.

Ectopic expression of the opdH operon allows growth of ΔtctD on citrate. (A and B) Growth curves of the ΔtctD mutant expressing portions of the opdH operon as indicated from an arabinose-inducible multiple cloning site on plasmid pJN105. Strains were grown without (A) or with (B) 100 μM arabinose as an inducer. The data are representative curves of at least three independent experiments. (C) Mean lag times of ≥3 biological replicates corresponding to the indicated strains from panels A and B. Error bars indicate SEM. (D) Working model summarizing the roles of different citrate (CIT) and cis-aconitate (ACO) transporters examined in this study.

DISCUSSION

Despite TCA cycle intermediates being preferred carbon sources for P. aeruginosa (8–10), the process of moving citrate molecules from the extracellular space to the cytoplasm has not been well understood in this organism. Previously, it was shown by Tamber et al. that transposon insertion in opdH blocks growth on cis-aconitate (3). However, further work by Tamber et al. failed to demonstrate transport of citrate or cis-aconitate by OpdH in a planar lipid bilayer (11), suggesting that transport does not rely on a single channel protein but may involve cooperation between outer membrane and inner membrane components. It has been thought that the opdH operon is repressed by TctD, the response regulator of the TctDE TCS (11, 14). However, given recently noted growth defects of tctDE mutants on citric acid (15), if OpdH is the major transporter of citrate across the cell membrane, TctD would likely not be acting as a repressor, leading to questions as to how TctDE regulates citrate transport.

We first queried the importance of OpdH in citrate and cis-aconitate transport by constructing a markerless in-frame deletion, in contrast to the previous insertional method (11) that might have impacted downstream genes in the operon. We found (Fig. 1) that deletion of opdH had no effect on growth in either carbon source, in agreement with the previous suggestion (11) that OpdH is not crucial to transport. Combined with our finding that tctA mutants lag in growth (Fig. 3B), we infer that earlier work (3) may have inadvertently disrupted expression of tctCBA in addition to opdH, resulting in a strong growth defect on citrate. Our finding that opdH expression in E. coli fails to allow growth on citrate or cis-aconitate (Fig. 2) confirms the notion that OpdH is not a standalone transport unit for these tricarboxylic acids.

We next identified and tested a CitA ortholog for its ability to transport citrate. Deletion of citA in combination with opdH, but not alone, resulted in a growth lag on cis-aconitate but not citrate (Fig. 1). Moreover, expression of citA in E. coli MG1655 permitted modest growth on citrate but substantially better growth on cis-aconitate (Fig. 2). We interpret these results as evidence that CitA is primarily a transporter of cis-aconitate and has an additional but weak ability to transport citrate. We also failed to observe significant transcription of citA in glucose, citrate, or a mix of both (Fig. 6E). This finding agrees with the idea that CitA is primarily a cis-aconitate transporter; however, it is not induced by cis-aconitate (see Fig. S8), raising the possibility that it has yet another substrate.

Since deletion of both opdH and citA did not cause growth defects in citrate, we turned to transcriptomic comparisons to identify novel citrate-upregulated transporters. Encouragingly, we identified opdH and its downstream operon member tctA, which are known to be citrate-induced (11). We selected tctA, encoding the predicted major transmembrane section of the TctCBA transporter, for markerless deletion and phenotypic screening. We also selected for deletion rarD, PA14_72170, and PA14_18320, which potentially encode transport systems (Fig. 7D). PA14_18320 and RarD are annotated as EamA family transporters, a group of proteins in the membrane drug/metabolite superfamily (23, 24). PA14_72170 is marked as a CitMHS family transporter (UniProt annotates it citM), a type of porin that transports divalent metal cations complexed with citrate (25).

In agreement with our hypothesis that these proteins are transporters for citrate, we found that a triple deletion of tctA, 72170, and either rarD or 18320 did not grow on citrate (Fig. 3C), suggesting that RarD and 18320 may work together (Fig. 7D). In addition, a ΔtctA mutant lagged in growth on citrate, and additional deletion of 72170 exacerbated the lag, both in PA14 (Fig. 3) and ΔcitA ΔopdH (see Fig. S2) backgrounds. To our knowledge, ours is the first study to use a markerless deletion strategy to generate a P. aeruginosa strain that does not grow on citrate due to transporter knockout. However, the above results were insufficient to conclude that the predicted porins are themselves transporters of citrate, and so we expressed them from pTrc99A in E. coli MG1655. Among tctA, 72170, rarD, and 18320, only 72170 transported citrate when individually expressed in E. coli (Fig. 4). A likely explanation for why TctA alone failed to allow E. coli to grow on citrate is that it likely needs TctB and TctC to function; similarly, RarD and PA14_18320 may need one another or other partner proteins that are not present in E. coli to achieve transport.

How do these putative citrate transporters relate to transport of cis-aconitate? Deletion of tctA appears to completely abolish growth on cis-aconitate (see Fig. S1), though, curiously, not in a ΔopdH ΔcitA background (see Fig. S2). More oddly, deleting 72170 or 18320 in the ΔtctA strain restored at least some cis-aconitate growth. A triple ΔtctA Δ72170 ΔrarD mutant, however, did not grow on cis-aconitate (see Fig. S1B). We are unsure how to interpret these results without further study but speculate that loss of TctA while keeping an intact OpdH could cause toxic periplasmic buildup of cis-aconitate if OpdH is the outer membrane porin and TctCBA the inner. None of the newly identified transporters was able to allow growth of E. coli MG1655 on cis-aconitate (see Fig. S3), implying that 72170 is specific for citrate. We were also surprised that 54580, the last member of the opdH operon, was inhibitory to E. coli growth when expressed in conjunction with the rest of the operon (Fig. 2C and D) but improved growth of a P. aeruginosa ΔtctD strain when expressed with the rest of the opdH operon. The 54580 protein was previously found to be 70% similar to AmoA of Pseudomonas putida (11), a membrane bound ammonia dioxygenase. It was suggested that because such enzymes can have fairly wide substrate specificity, 54580 may have a role in the metabolism of compounds transported by OpdH/TctCBA (11). If true, however, the data require that this function abet growth in P. aeruginosa but inhibit growth in E. coli; one possible mechanism for this difference is that E. coli does not possess a downstream metabolic pathway for a product of 54580 activity.

We next tried to understand the regulation of citrate transporter expression. It is known that ΔtctDE strains have a pH-independent growth defect when citric acid is added to M63-arginine medium (15) and that P. aeruginosa has elevated opdH expression in response to tricarboxylates (11), as confirmed by our transcriptomic analyses. We attempted to grow both a ΔtctDE strain and individual ΔtctD and ΔtctE mutants on citrate to understand the relative roles of these genes. We found that ΔtctDE and ΔtctE strains have severe growth lags, while the ΔtctD mutant does not grow at all. This result does not fit the previous proposition that TctD is acting as a repressor of expression. Instead, the available evidence supports a model in which TctD is a positive regulator, not a repressor, of the opdH operon, since the loss of TctD can be compensated for by opdH operon expression (Fig. 7).

Because tandem loss of tctE and tctD does allow eventual growth on citrate, it is possible that loss of previously documented interactions between TctD and the PhoB response regulator (14) causes a TctE-dependent toxic effect. This would make sense, as TctD was shown to repress genes that are activated by PhoB (14). If a toxic gene were also tricarboxylate-inducible and the interaction TctE-dependent, such a strain would not grow on citrate or cis-aconitate. The ΔtctD strain was able to grow in citrate or cis-aconitate when elements of the opdH operon were ectopically expressed (Fig. 7), suggesting that any such toxic effect can be remedied by allowing citrate transport into the cell. We speculate that without TctD there is a TctE-dependent way for citrate to enter the periplasm but not cross into the cytoplasm, resulting in a buildup that compromises cell growth. We also found that every element of the opdH operon contributes to improving the growth of this strain, indicating that while TctCBA may form the core transport machinery (its expression alone is sufficient to enable some growth), OpdH and PA14_54580 have supporting roles to play in transport or use of tricarboxylates. The fact that CitA functions alone as a transporter in E. coli but does not complement the loss of tctD suggests that CitA requires a partner protein in P. aeruginosa that ΔtctD does not express—perhaps OpdH. Taking our results together, we argue that the TctDE TCS likely positively regulates citrate transport in P. aeruginosa rather than TctD acting as a repressor and that loss of TctDE can be compensated for via heritable mutations. We have found candidate genetic differences in adapted strains whose role in citrate growth adaptation will be an important aspect of future studies.

In summary, we have probed the transport of citrate and cis-aconitate in P. aeruginosa and identified proteins that are of interest in this process. We have clarified that the entire opdH-tctCBA-PA14_54580 operon is impactful for tricarboxylate transport, with tctCBA as the key cross-membrane component (Fig. 7D). We have also shown that the TctDE TCS is important, but not strictly required, for growth on citrate and cis-aconitate and have supplied evidence for a possible mechanism of its control over transporter expression. Finally, we have shown that growth of tctDE mutants on citrate induces a heritable change that rescues citrate growth via enhanced expression of the opdH transport operon.

MATERIALS AND METHODS

Strains and growth conditions.

P. aeruginosa and E. coli strains (Tables 2 and 3) were grown in LB (Lennox) medium (10 g/L tryptone, 5 g/L yeast extract, and 5 g/L NaCl) for overnight cultures. Liquid cultures were grown with shaking in 14-mL round-bottomed tubes at 180 rpm and 37°C, unless otherwise specified. M9 medium was made according to the Cold Spring Harbor protocol (26) without glucose, which was replaced with indicated carbon sources. In all cases, before experiments, cells were pelleted and washed twice in phosphate-buffered saline (PBS; pH 7.4) to remove any residual LB medium from the overnight culture. Cells were then diluted as indicated into the new growth medium used for the experiment. E. coli strains bearing pEXG2 or JN105 constructs were grown in an additional 20 μg/mL gentamicin; P. aeruginosa strains bearing pJN105 constructs were grown in 60 μg/mL gentamicin. E. coli bearing pTrc99A constructs were grown in 50 μg/mL carbenicillin, whereas E. coli bearing pCitT was grown in 50 μg/mL kanamycin. Unless otherwise indicated, citrate and cis-aconitate were added to M9 at a concentration of 7.5 mM; glucose was added at a concentration of 3 mM. Mixtures of citrate and glucose were 7.5 mM citrate plus 3 mM glucose.

TABLE 2.

P. aeruginosa strains used in this study^a

Strain	Genotype	Characteristicsb	Source or reference
MTC1	PA14	Laboratory wild-type P. aeruginosa PA14	28
MTC2550	ΔcitA	PA14 with markerless citA deletion	This study
MTC2551	ΔopdH	PA14 with markerless opdH deletion	This study
MTC2552	ΔcitA ΔopdH	PA14 with markerless opdH and citA deletions	This study
MTC2553	ΔtctA	PA14 with markerless tctA deletion	This study
MTC2554	ΔrarD	PA14 with markerless rarD deletion	This study
MTC2555	ΔPA14_18320	PA14 with markerless PA14_18320 deletion	This study
MTC2556	ΔPA14_72170	PA14 with markerless PA14_72170 deletion	This study
MTC2557	ΔtctD	PA14 with markerless tctD deletion	This study
MTC2558	ΔtctE	PA14 with markerless tctE deletion	This study
MTC2559	ΔtctDE	PA14 with markerless tctDE deletion	This study
MTC2560	ΔtctE passaged	ΔtctE passaged in M9-citrate, streaked on LB medium, and then frozen from LB medium overnight culture	This study
MTC2561	ΔtctDE passaged R1	ΔtctDE passaged in M9-citrate, streaked on LB medium, and then frozen from LB medium overnight culture; replicate 1	This study
MTC2562	ΔtctDE passaged R2	ΔtctDE passaged in M9-citrate, streaked on LB medium, and then frozen from LB overnight culture; replicate 2	This study
MTC2563	ΔtctD pJN105 empty vector	PA14 ΔtctD carrying pJN105 empty vector; Gmr	This study
MTC2564	ΔtctD pJN105-tctCBA	PA14 ΔtctD carrying pJN105 encoding the tctCBA gene cluster under arabinose-inducible promoter; Gmr	This study
MTC2565	ΔtctD pJN105-opdH-tctCBA	PA14 ΔtctD carrying pJN105 encoding the opdH-tctCBA gene cluster under arabinose-inducible promoter; Gmr	This study
MTC2566	ΔtctD pJN105-tctCBA-PA14_54580	PA14 ΔtctD carrying pJN105 encoding the tctCBA-PA14_54580 gene cluster under arabinose inducible promoter; Gmr	This study
MTC2567	ΔtctD pJN105-opdH-tctCBA-PA14_54580	PA14 ΔtctD carrying pJN105 encoding the entire opdH-tctCBA-PA14_54580 operon under arabinose-inducible promoter; Gmr	This study

^aParental lineages and details of plasmids and E. coli cloning/conjugation strains are given in the supplemental material.

^bGm^r, gentamicin resistance.

TABLE 3.

E. coli strains used in this study

Strain	Genotype	Characteristicsa	Source or reference
MTC2568	MG1655	Wild type K-12 derivative E. coli	Yale Genetic Stock Culture Collection CGSC 6300
MTC2569	MG1655 pCitT	MG1655 bearing E. coli CitT citrate transporter; Kmr	This study
MTC2570	MG1655 pTrc99A-citA	MG1655 bearing pTrc99A plasmid with citA inserted at lac controlled multiple cloning site; Cbr	This study
MTC2571	MG1655 pTrc99A-opdH	MG1655 bearing pTrc99A plasmid with opdH inserted at lac controlled multiple cloning site; Cbr	This study
MTC2572	MG1655 pTrc99A-tctCBA	MG1655 bearing pTrc99A plasmid with tctCBA inserted at lac controlled multiple cloning site; Cbr	This study
MTC2573	MG1655 pTrc99A-opdH-tctCBA	MG1655 bearing pTrc99A plasmid with opdH-tctCBA inserted at lac controlled multiple cloning site; Cbr	This study
MTC2574	MG1655 pTrc99A-tctCBA-PA14_54580	MG1655 bearing pTrc99A plasmid with tctCBA-PA14_54580 inserted at lac controlled multiple cloning site; Cbr	This study
MTC2575	MG1655 pTrc99A-opdH-tctCBA-PA14_54580	MG1655 bearing pTrc99A plasmid with opdH-tctCBA-PA14_54580 inserted at lac controlled multiple cloning site; Cbr	This study
MTC2576	MG1655 pTrc99A-tctA	MG1655 bearing pTrc99A plasmid with tctA inserted at lac controlled multiple cloning site; Cbr	This study
MTC2577	MG1655 pTrc99A-PA14_72170	MG1655 bearing pTrc99A plasmid with PA14_72170 inserted at lac controlled multiple cloning site; Cbr	This study
MTC2578	MG1655 pTrc99A-PA14_18320	MG1655 bearing pTrc99A plasmid with PA14_18320 inserted at lac controlled multiple cloning site; Cbr	This study
MTC2579	MG1655 pTrc99A-rarD	MG1655 bearing pTrc99A plasmid with rarD inserted at lac controlled multiple cloning site; Cbr	This study

^aCb^r, carbenicillin resistance; Km^r, kanamycin resistance.

Strain construction.

E. coli strains bearing plasmids were generated by standard chemical competence transformation with Gibson-assembled plasmids (see the supplemental material). P. aeruginosa allelic replacement mutants were generated using plasmid pEXG2 (27) containing flanking homologous regions of the gene to be deleted, which were amplified using high-fidelity PCR. Plasmids were inserted into P. aeruginosa PA14 by conjugation with E. coli SM10 on LB agar (see supporting information in the supplemental material).

Growth curve experiments.

For the automated measurement of growth curves, cells were washed in PBS from overnight culture as described above. M9 was prepared with either 3 mM glucose, 7.5 mM citrate, or 7.5 mM cis-aconitate as indicated. Cells were diluted 100-fold into a clear 96-well polystyrene plate (Corning, Inc.) and covered with a lid. This was then incubated at 37°C in a BioTek Synergy H1 plate reader (BioTek, CA), and the OD₆₀₀ was measured every 10 min after a 2-s shake.

For the Δtct mutant back-dilution experiment, cells were first grown in M9 with 7.5 mM citrate until the culture was turbid. Then, 1 mL of the culture was transferred to a 1.5-mL microcentrifuge tube, and the cells washed twice in PBS by centrifugation and resuspension. The resulting cells were 100-fold back-diluted into shaking flasks containing M9 with 7.5 mM citrate. Growth was monitored by measurement of the OD₆₀₀ using cuvettes in a spectrophotometer.

To measure the lag time of a strain, the numerical derivative of the OD₆₀₀ versus time trace was calculated using Matlab. This was then scanned for a point at which the gradient increased for four time points in a row in the positive direction. Manual inspection of the growth curve was then conducted to ensure that the flagged point was not an anomalous OD₆₀₀ spike. If growth was sustained thereafter, the first time point at which this occurred was taken to be the end of the lag phase.

RT-qPCR experiments.

For RT-qPCR, cells in biological triplicate overnight cultures were diluted 50-fold from a 2× PBS wash into 50 mL of M9 with either 3 mM glucose, 7.5 mM citrate, or 3 mM glucose plus 7.5 mM citrate. Cells were allowed to grow to an OD of between 0.198 and 0.397 (mid-exponential-phase growth) before centrifugation. The pellet was then resuspended in 250 μL of TE buffer (10 mM Tris [pH 8.0], 1 mM EDTA) supplemented with 10 mg/mL lysozyme from chicken egg white and allowed to incubate at room temperature for 15 min. After incubation, the RNA extraction protocol from New England Biolabs’ Monarch RNA extraction kit (NEB, MA) was followed. The purified RNA was quantified using UV spectrophotometry, and 500 ng of RNA was reverse transcribed using the RevertAid RT Reverse transcription kit (Thermo Fisher, MA). The resulting cDNA was diluted 25-fold in water, and a 1-μL portion of this was used in each 10-μL qPCR. Reactions were carried out in Bio-Rad semiskirted qPCR plates covered with an adhesive film (Bio-Rad, CA) using the Promega GoTaq BRYT Green dye-based qPCR kit (Promega, WI) with a 200 nM concentration of the appropriate primers. The plate was centrifuged at 500 rpm in a Sorvall ST 40R centrifuge equipped with a TX-1000 rotor for 2 min. This was then processed in a CFX96 Touch real-time PCR cycler (Bio-Rad). Initial denaturation was performed for 2 min at 95°C, and cycles were performed using a 15-s denaturation step at 95°C, followed by 1 min of annealing and extension at 60°C.

The results of the qPCR were determined by using a standard curve method. Standards were produced by gel purifying an ordinary PCR made using the qPCR primers and PA14 genomic DNA. The result was quantified spectrophotometrically, and the average weight per base of the strand was used to dilute it to a concentration of 10⁹ copies/μL. This was then serially 10-fold diluted to generate a standard curve down to 10² copies/μL. These standards were included in the qPCR plate and a plot of log(copy number) versus C_q (the number of cycles needed to reach the fluorescence quantification threshold) used to determine the relationship between C_q and the copy number of DNA strands. The standard was then used to infer how many copies of a particular reverse-transcribed RNA were present in the cDNA samples. Each of these was then divided by the number of 16S rRNA copies to normalize to a housekeeping gene. Error was calculated by taking the standard error in the mean of two technical replicates for each biological replicate, both for the 16S measurement and the queried genes. These errors were then propagated forward into the quotient using standard error propagation methods.

RNA sequencing.

RNA for sequencing was extracted as for the RT-qPCR experiments. The RNA was then depleted of rRNA using the Illumina RiboZero kit (Illumina, CA). Depleted samples were submitted for bulk Illumina sequencing (15M reads/sample, 150-bp paired end) at the University of Oklahoma Health Sciences Center core facility in Oklahoma City, OK. Sequence mapping and analysis were performed at the Oklahoma University Health Sciences Center Laboratory for Molecular Biology and Cytometry Research using CLC software. A complete list of citrate-regulated genes is deposited as a supplemental data file.

Whole-genome sequencing.

Samples for whole-genome sequencing were grown overnight in LB medium as described above. Next, 1 mL of cells was processed using a Promega Wizard genomic DNA preparation kit (Promega) to obtain genomic DNA. The DNA was then shipped on dry ice to Novogene (Novogene Corporation, Inc., CA) for quality control, library preparation, and sequencing (Illumina NovaSeq, 150-bp paired-end reads). Resequenced genomes (with a typical mean coverage depth of ~300) were assembled to the PA14 reference genome (GenBank NC_008463) with Geneious 11.1.5 (Biomatters, Ltd.). The sequenced mutant genomes have been deposited in GenBank with accessions SAMN30953206 (PA14), SAMN30953207 (ΔtctE parental), SAMN30953208 (ΔtctDE parental), SAMN30953209 (ΔtctDE passaged R1), SAMN30953210 (ΔtctDE passaged R2), and SAMN30953211 (ΔtctE passaged R1).