A Key Regulator of the Glycolytic and Gluconeogenic Central Metabolic Pathways in Sinorhizobium meliloti

Abstract

The order Rhizobiales contains numerous agriculturally, biotechnologically, and medically important bacteria, including the rhizobia, and the genera Agrobacterium, Brucella, and Methylobacterium, among others. These organisms tend to be metabolically versatile, but there has been relatively little investigation into the regulation of their central carbon metabolic pathways. Here, RNA-sequencing and promoter fusion data are presented to show that the PckR protein is a key regulator of central carbon metabolism in Sinorhizobium meliloti; during growth with gluconeogenic substrates, PckR represses expression of the complete Entner–Doudoroff glycolytic pathway and induces expression of the pckA and fbaB gluconeogenic genes. Electrophoretic mobility shift assays indicate that PckR binds an imperfect palindromic sequence that overlaps the promoter or transcriptional start site in the negatively regulated promoters, or is present in tandem upstream the promoter motifs in the positively regulated promoters. Genetic and in vitro electrophoretic mobility shift assay experiments suggest that elevated concentrations of a PckR effector ligand results in the dissociation of PckR from its target binding site, and evidence is presented that suggests phosphoenolpyruvate may function as the effector. Characterization of missense pckR alleles identified three conserved residues important for increasing the affinity of PckR for its cognate effector molecule. Bioinformatics analyses illustrates that PckR is limited to a narrow phylogenetic range consisting of the Rhizobiaceae, Phyllobacteriaceae, Brucellaceae, and Bartonellaceae families. These data provide novel insights into the regulation of the core carbon metabolic pathways of this pertinent group of α-proteobacteria.

BACTERIA are commonly present in nutritionally complex environments and may move between niches. The fitness of a species is therefore influenced by the ability to utilize diverse nutrient sources, as well as the capability to induce and repress central and peripheral metabolic pathways. Escherichia coli and Bacillus subtilis have historically served as the primary model species for the study of these topics. However, many species do not follow the rules established in these organisms, including many economically important species such as the photosynthetic and biohydrogen-producing Rhodobacter sphaeroides and the nitrogen-fixing legume symbiont Sinorhizobium meliloti. It is therefore important to study regulation and metabolism in diverse model species in order to develop a broad understanding and to optimize success in the biotechnological manipulation of these organisms.

S. meliloti is a metabolically versatile organism (Biondi et al. 2009; Geddes and Oresnik 2014) of the Rhizobiales order of the α-proteobacteria, which includes N₂-fixing plant symbionts (rhizobia), plant pathogens (Agrobacterium, Lieberbacter), livestock and human pathogens (Brucella, Ochrobactrum), and methylotrophs (Methylobacterium), among others. The preferred carbon source for S. meliloti appears to be C₄-dicarboxylic acids (succinate, malate, fumarate), which also serve as the primary carbon source available to rhizobia during symbiotic nitrogen fixation (Ronson et al. 1981; Finan et al. 1983; Bolton et al. 1986). There are several differences between the metabolism of S. meliloti and its regulation compared to E. coli and B. subtilis (Geddes and Oresnik 2014). Notable differences include the use of the Entner-Doudoroff (ED) pathway for glycolysis (Stowers 1985), and the presence of succinate-mediated catabolite repression resulting in diauxic growth and the preferential use of succinate before glucose when both are present (Ucker and Signer 1978; Bringhurst and Gage 2002; Zhang et al. 2016).

Glucose is predominately metabolized through the ED pathway in S. meliloti, with a small portion flowing through the pentose phosphate (PP) pathway (Figure 1) (Fuhrer et al. 2005). The ED pathway produces one pyruvate and one glyceraldehyde-3-phosphate (G3P). The pyruvate can be directly converted into acetyl-CoA for entry into the tricarboxylic acid (TCA) cycle. The G3P can be further metabolized via the lower half of the Embden–Meyerhoff–Parnas (EMP) pathway to a second pyruvate (Figure 1) (Geddes and Oresnik 2014). When growing with substrates that enter central carbon metabolism directly through the TCA cycle, some carbon is diverted from the TCA cycle to pyruvate for synthesis of acetyl-CoA and continued production of citrate and other TCA cycle intermediates. Additional carbon is converted into phosphoenolpyruvate (PEP) that is metabolized to glucose and other sugars required for biosynthesis of essential polysaccharides, nucleotides, and amino acids. This occurs through reversal of the lower and upper halves of the EMP pathway in a process known as gluconeogenesis (Figure 1) (Finan et al. 1988; Geddes and Oresnik 2014). Synthesis of pyruvate from malate is catalyzed by the DME and TME malic enzymes (Driscoll and Finan 1993, 1996, 1997), and synthesis of PEP from oxaloacetate is catalyzed by phosphoenolpyruvate carboxykinase (PCK) (Østerås et al. 1995).

Figure 1.

Schematic of central carbon metabolism in Sinorhizobium meliloti. The pathways and associated genes for glycolysis, gluconeogenesis, and the TCA cycle are shown. Incomplete representations of the pentose phosphate pathway (PPP), the methylglyoxal pathway (MGP), as well as glycerol and galactose metabolism are also shown in gray. Blue indicates genes repressed by PckR during growth with succinate, and red highlights genes that are induced by PckR during growth with succinate. Light blue is used for glk as the level of induction (∼2.7-fold) was below the threefold cutoff used in this study. The primary carbon sources used in this study are indicated in magenta. The schematic was produced with the assistance of iPath v2 (Yamada et al. 2011), and the gene associations are based on the annotations of an S. meliloti metabolic model (diCenzo et al. 2016a).

The interconnectedness and cyclic nature of the ED, EMP, PP, and TCA pathways (Figure 1) suggest a necessity for strict regulation of the key metabolic crossroads. In S. meliloti, this is accomplished in part with the LacI-type transcriptional regulator PckR, which regulates expression of the pckA gene encoding for the PCK enzyme (Østerås et al. 1997). Expression of pckA is low during growth with glucose or lactose and high during growth with succinate or arabinose (Østerås et al. 1995, 1997). The pckR gene was identified as the locus to which three independent spontaneous mutations that influenced pckA expression mapped (Østerås et al. 1997). These pckR alleles resulted in elevated pckA expression and were recessive to the wild-type allele, but the actual nucleotide changes that resulted in the mutant alleles were not determined (Østerås et al. 1997). In silico studies predicted that PckR may regulate as many as 81 genes in S. meliloti (Galardini et al. 2015), suggesting that PckR may be a global regulator of carbon metabolism. In R. sphaeroides, the CceR regulator fulfils the role of a global regulator of carbon and energy metabolism; RNA-sequencing (RNA-seq) showed that CceR was required for the correct regulation of 225 genes, 31 of which are involved in central carbon and energy metabolism (Imam et al. 2015). 6-Phosphogluconate appeared to be the effector molecule of CceR, with elevated levels of 6-phosphogluconate resulting in the dissociation of CceR from its target promoters in vitro (Imam et al. 2015). However, despite potentially serving similar biological roles, PckR and CceR are nonorthologous proteins and the similarity in their properties is unclear.

Here, we provide RNA-seq data that shows PckR is a key regulator of central carbon metabolism in S. meliloti. The biological role and binding activity of PckR is examined with a combination of genetic and in vitro biochemical characterizations, and its phylogenetic distribution is presented.

Materials and Methods

Bacterial growth conditions, media, and genetic manipulations

All media [LB (Luria broth], LBmc (LB supplemented with magnesium and calcium), M9 minimal medium, and MM9 (M9 with the phosphate buffer replaced with a 3-[N-morpholino]propanesulfonic acid [MOPS] - potassium hydroxide [KOH] buffer with 2 mM phosphate) minimal medium), antibiotic concentrations, and growth conditions for S. meliloti and E. coli were as previously described (diCenzo et al. 2014). Cultures for RNA isolation were grown in M9 minimal media. Subsequent growth and expression experiments were performed using MM9 minimal media, as we found growth was more rapid and there was somewhat less background GFP fluorescence. Unless stated otherwise, all carbon sources in the minimal media were added to a final concentration of 15 mM. DNA manipulations and recombinant techniques, bacterial matings, and ΦM12 transductions were performed as described before (Finan et al. 1984; Sambrook et al. 1989; Cowie et al. 2006; Milunovic et al. 2014). Recombinant plasmids were produced through either ligations with T4 DNA ligase or through sequence and ligation independent cloning (Sambrook et al. 1989; Jeong et al. 2012). All strains used in this study are listed in Supplemental Material, Table S1 in File S1, and oligonucleotide primer sequences are provided in Table S2 in File S1.

Construction of the pckR::ΩSp^R allele

The ΩSp^R cassette from pHP45Ω was ligated into HindIII-digested pTH296, introducing the ΩSp^R cassette 23-bp downstream of the start of the pckR open reading frame. (Blondelet-Rouault et al. 1997; Østerås et al. 1997). The 4.5-kb region from the resulting plasmid was then subcloned into pRK7813 as an EcoRI fragment, producing pTH476. The pckR::ΩSp^R allele was recombined into the S. meliloti Rm5000 (Finan et al. 1984) genome and then transduced into S. meliloti RmG212 to produce RmK124, the pckR null mutant used in this study.

Construction of the gfp-lacZ fusion strains

Plasmids for construction of the pckA⁺::gfp-lacZ and zwf⁺::gfp-lacZ alleles were identified in the S. meliloti fusion library (Cowie et al. 2006). For construction of the fbaB⁺::gfp-lacZ, eda2⁺::gfp-lacZ, and mgsA⁺::gfp-lacZ strains, each promoter region was PCR amplified (oligos: fbaB – DF106/DF107; eda2 – DF108/DF109; mgsA – DF110/DF111), cloned into XhoI-digested pTH1522 via SLIC, and transferred via conjugation into the desired S. meliloti strain (Cowie et al. 2006).

Construction of pckR expression vectors

The coding sequence of the S. meliloti Rm1021 pckR gene (smc02975) and the putative pckR homolog (azc_4468) of Azorhizobium caulinodans ORS 571 were PCR amplified (oligos: smc02975 – DF112/DF113; azc_4468 – DF114/DF115). In each case, the forward primer contained a ribosome binding site active in S. meliloti (Poysti and Oresnik 2007) placed 7-nt upstream of the start codon, as well as a stop codon overlapping the 5′ end of the ribosome binding site by 1 nt. Each PCR product was cloned into HindIII-digested pRK7813 such that the stop codon was in frame with the N-terminal fragment of the lacZα gene. The constructs were conjugated into S. meliloti RmK124 (pckR::ΩSp^R), and the gene of interest expressed from the lac promoter without IPTG induction.

Whole-genome sequencing and analysis

S. meliloti total genomic DNA was isolated as described previously (Cowie et al. 2006) from S. meliloti RmG212 (pckR⁺), RmH166 (rpk-9), RmH167 (rpk-10), and RmH172 (rpk-15). Genome sequences were obtained using the Illumina MiSeq V2 technology, with 250 bp paired-end reads, at the Farncombe Family Digestive Health Research Institute located at McMaster University (Ontario, Canada). The location of mutations were determined by mapping the sequencing reads to the published S. meliloti Rm1021 reference genome (Galibert et al. 2001) using Geneious R8, followed by the identification of polymorphisms as described elsewhere (diCenzo et al. 2016b).

Growth experiments and GFP measurements

Strains were pregrown overnight in LBmc with the appropriate antibiotics. Growth experiments were set up in 96-well microtiter plates as described previously (diCenzo et al. 2017), and as described further in the Supplemental Methods in File S1. The microtiter plate was incubated in a Biotek Cytation 3 plate reader with shaking at 30°, and OD₆₀₀ and GFP fluorescence (excitation/emission: 485/514 nm) were measured every 15 min.

β-Galactosidase assays

Reactions were performed in 96-well microtiter plates largely as described previously (Cowie et al. 2006), and as described further in the Supplemental Methods in File S1.

RNA isolation

Duplicate cultures of S. meliloti RmG212 and RmK124 were grown overnight in LBmc. Cultures were pelleted, washed once, and resuspended with carbon-free M9 medium. Each cell suspension was used to inoculate 50 ml cultures of M9 with 15 mM glucose and M9 with 15 mM succinate to a starting OD₆₀₀ ∼0.025. Cultures were grown with shaking at 30° and cells harvested when cultures reached an OD₆₀₀ between 0.35 and 0.45. Cultures were harvested by immediately mixing 45 ml culture with 5 ml ice-cold cell stop solution (5% unbuffered phenol, 95% ethanol), centrifuging at 3700 × g for 10 min, and flash-freezing the pellet with liquid nitrogen. Frozen pellets were stored at −80° until use. Total RNA was isolated largely as described previously (MacLellan et al. 2005), and is described in detail in the Supplemental Methods in File S1.

RNA-seq and analysis

Depletion of ribosomal RNA was performed using Ribo-Zero rRNA Removal kits (Illumina), and sequencing was performed using an Illumina HiSeq 1500 with 60 nt single reads at the Farncombe Family Digestive Health Research Institute located at McMaster University. Sequencing data were analyzed using the Rockhopper v2.03 software (Tjaden 2015), with default settings and the following modifications: reverse complement reads and verbose output were selected. Reads were mapped to the reference S. meliloti Rm1021 genome sequence, using a file in which the three DNA replicons (NC_003037, NC_003047, and NC_003078) were combined as one. Differentially regulated genes were identified as those with an adjusted P-value <0.01, as determined by the Rockhopper v2.03 software, a fold change of at least three based on the average of the normalized read counts, and less than a twofold difference between replicates of the same sample. A de novo prediction of noncoding RNAs (ncRNAs) was performed by Rockhopper, and the same thresholds were applied as for the identification of differentially regulated coding regions. Additionally, the ncRNAs were only considered to be true ncRNAs if they overlapped those predicted previously (Sallet et al. 2013). Heatmaps were generated with the heatmap.2 function in the gplots package of R (Warnes et al. 2016), and the clustering performed using average linkage with a Pearson correlation distance. The output of the Rockhopper analysis is provided as File S2, and the raw and processed RNA-seq data are available through the Gene Expression Omnibus (accession number GSE100765).

Purification of PckR

The S. meliloti pckR (smc02975) open reading frame of wild-type S. meliloti RmP110 was PCR amplified (oligos: DF116/DF117) and cloned into NdeI-digested pTH2976, producing pTH3150, for expression of an N-terminal, H₆-tagged PckR protein. Plasmid pTH3150 was transformed into E. coli BL21 (DE3) pLysS for overexpression of the PckR protein. The PckR protein was overexpressed and purified using standard protocols, and these are described in detail in the Supplemental Methods in File S1.

Electrophoretic mobility shift assays

Promoter regions were PCR amplified from S. meliloti genomic DNA for use as probes in electrophoretic mobility shift assays (EMSAs; oligos: zwf – DF118/DF119, pckA – DF120/DF121, fbaB – DF122/DF123, eda2 – DF124/ DF125, mgsA – DF126/DF127). Each PCR product was between 135 and 200 bp in length, and were centered around the predicted PckR binding motifs. Purified PCR products were 5′ end-labeled with T4 polynucleotide kinase (New England Biolabs) using 0.5 pmol of probe and 1 pmol of [γ-³³P] ATP (Perkin Elmer), and purified. Binding reactions and electrophoresis were performed largely as described before (MacLean et al. 2008), and described in detail in the Supplemental Methods in File S1. Where applicable, PEP or pyruvate was added to the binding reactions to a final concentration as indicated, or double-stranded oligonucleotides were added to a molar ratio (relative to probe) of 1000:1.

Identification of the PckR binding motif

The 200-nt upstream of the translational start codon of the first gene of each PckR regulated operon were collected, and motifs were identified using the MEME motif discovery webserver (Bailey and Elkan 1994; Bailey et al. 2009) with default settings and the following selections: normal mode, standard DNA alphabet, any number of motif repetitions, and the identification of five motifs. The top hit was 15 nt in length, and manual inspection of the surrounding region led to the enlargement of this motif to 16 nt. The logo of this motif was created with skylign (Wheeler et al. 2014) based on the sense strand sequence of the seven copies of the motif, using the “observed counts,” “full length alignment,” and “information content – above background” options.

Modeling the structure of PckR

The tertiary structure of the PckR protein was modeled using the PckR amino acid sequence from S. meliloti Rm1021, and the Phyre2 webserver (Kelley et al. 2015) with the intensive mode. Phyre2 chose the following six crystal structures (identified according to their Protein Data Base accession) as templates for the modeling, all of which have 20–30% amino acid identity with the S. meliloti PckR protein: 1BDH, 3KJX, 1EFA, 3H5T, 2LBG, and 1ZVV. The output from Phyre2 was submitted to the 3DLigandSite webserver (Wass et al. 2010) for prediction of ligand binding domains. The tertiary structure of PckR was visualized with Chimera (Pettersen et al. 2004).

Blast bidirectional best hit analysis

The Blast bidirectional best hit (Blast-BBH) analysis was performed using an automated pipeline based upon the workflow described previously (Zamani et al. 2017), and this process is detailed further in the Supplemental Methods in File S1. All genomes of the order Rhizobiales available through the NCBI Genome database (accessed March 21, 2016) that were annotated as complete were included in this analysis. The PckR Blast-BBHs were aligned with MAFFT-linsi (Katoh and Standley 2013), trimmed with trimAl (Capella-Gutiérrez et al. 2009), and a maximum likelihood phylogeny built with the RAxML BlackBox mirror (Stamatakis 2014) on the CIPRES Science Gateway webserver (Miller et al. 2010), with a CAT rate heterogeneity and LG amino acid substitution model.

Multilocus sequence analysis

The multilocus sequence analysis was performed using an automated pipeline centered on the use of AMPHORA2 (Wu and Scott 2012), based upon the workflow described previously (Zamani et al. 2017). One representative genome for each Rhizobiales species, as well as several outgroup taxa from the α-proteobacteria, were used in this analysis. The phylogeny was built based on 20 proteins present in all proteomes (Frr, RplA, RplB, RplE, RplF, RplK, RplM, RplN, RplP, RplS, RplT, RpoB, RpsB, RpsC, RpsE, RpsI, RpsJ, RpsK, RpsM, and RpsS). Each set of proteins were individually aligned with MAFFT-linsi (Katoh and Standley 2013) and trimmed with trimAl (Capella-Gutiérrez et al. 2009). The trimmed alignments were concatenated, and a phylogeny built with the RAxML BlackBox mirror (Stamatakis 2014) on the CIPRES Science Gateway webserver (Miller et al. 2010), with a CAT rate heterogeneity and LG amino acid substitution model.

LacI phylogenetic analysis

All proteins of the ENOG4105ETE and ENOG410XPSF orthologous protein groups of the eggNOG database (Huerta-Cepas et al. 2016) were downloaded, yielding 4607 proteins. The proteins were aligned with MUSCLE (Edgar 2004) on the CIPRES Science Gateway webserver (Miller et al. 2010), with MUSCLE limited to two iterations. The alignment was trimmed with trimAl (Capella-Gutiérrez et al. 2009), and an approximate maximum likelihood tree build with FastTree v2 (Price et al. 2010). All phylogenetic trees produced in this study were visualized with FigTree v1.4.2.

Data availability

File S1 contains the Supplemental Methods, Figures S1–S7, Tables S1–S3, and the Supplemental References. File S2 contains the output of the Rockhopper analysis of the RNA-seq data. Raw and processed RNA-seq data are available through the Gene Expression Omnibus (accession number GSE100765).

Results

PckR regulates expression of five central carbon metabolism operons

To investigate the role of pckR in gene regulation, we examined the transcriptional profile of pckR⁺ and pckR⁻ cells grown with a glycolytic substrate, glucose, or with a gluconeogenic substrate (succinate). Based on RNA-seq analysis, 40 genes are differentially expressed greater than threefold in the wild-type pckR⁺ strain RmG212; 11 genes display higher expression with succinate whereas 29 genes display lower expression with succinate (Figure 2A). Of these 40 genes, the differential expression of five operons is not observed in the pckR null mutant strain, RmK124. Specifically, the increased transcription of pckA and fbaB, and the decreased transcription of zwf-pgl-edd, eda2, and mgsA, during growth with succinate appears to be pckR dependent (Figure 2A). The phenotypes of the pckR::ΩSp^R mutant are not due to polar effects of the insertion, as expression of the downstream gene smc04662 appeared unaffected (RNA-seq data in File S2), and a smc04662 mutant has none of the phenotypes of the pckR mutant (data not shown).

Figure 2.

Identification of the PckR regulon. Wild-type S. meliloti (pckR⁺) and a pckR insertion mutant (pckR⁻) were grown with either glucose (Glu.) or succinate (Suc.) as the sole carbon source. (A) A heatmap showing the expression of all genes that displayed a statistically significant (adjusted P-value <0.01, fold change >3) difference in expression level in wild-type cells grown with succinate vs. glucose, as determined through RNA-seq. Expression values are shown as the log₂ of the relative change compared to the average of the pckR⁺ strain grown with glucose, and the results for each biological duplicate are presented. Hierarchical clustering analysis, shown on the left, identified four primary groupings, as indicated on the right. (B) The expression of the five genes identified as putatively regulated by PckR through RNA-seq were independently validated using gfp-lacZ reporter gene constructs by measuring the β-galactosidase (LacZ) activity of stationary phase cultures. Data points represent the averages of triplicate samples, with error bars indicating the standard deviation.

To independently confirm the role of PckR in the regulation of these five transcripts, the transcription of each was monitored in both pckR⁺ and pckR⁻ strains grown with succinate or glucose using chromosomal gfp-lacZ reporter constructs. The β-galactosidase measurements confirm the transcriptional profiles seen in the RNA-seq data (Figure 2B). Consistent with the transcriptional data, the pckR null mutant displays impaired growth with gluconeogenic substrates (succinate, arabinose, and pyruvate), but not with glycolytic substrates (glucose, fructose, glycerol, and galactose) or those entering the PP pathway (xylose and ribose) (Table S3 in File S1). Together, these data suggest that PckR functions as a positive regulator of the pckA and fbaB genes involved in gluconeogenesis, and as a negative regulator of the zwf-pgl-edd, eda2, and mgsA genes involved in glycolysis (Figure 1).

Interestingly, three ncRNAs were identified as differentially expressed between growth with glucose and succinate (smc06497, smc06705 and sma6570) in a PckR-independent manner. In the pckR⁺ strain RmG212, the ncRNAs smc06497, smc06705, and sma6570 were upregulated ∼3.8-, 5.2-, and 4.5-fold during growth with succinate relative to glucose, respectively. These changes were not independently confirmed or pursued further.

PckR directly modulates expression through binding an imperfect palindromic sequence

PckR is homologous to the GalR/LacI family of transcriptional regulators, and we previously noted (Østerås et al. 1997) that the pckA promoter contains a region from 83 to 70 nt upstream of the transcriptional start site (5′-TTAAATCGATTAAT-3′) that fits well with the GalR/LacI consensus palindromic binding site (5′-NNNAANCGNTTNNN-3′) (Weickert and Adhya 1992) (Figure 3A). To identify a putative PckR binding site, the five promoter regions whose expression was influenced by PckR were searched for conserved motifs using the MEME software (Bailey and Elkan 1994; Bailey et al. 2009). This analysis identified a conserved 16-nt motif (Figure 3, A and B) that matches the previously identified motif in the pckA promoter (Østerås et al. 1997), and has a consensus sequence of TTTMAATCGATTWAAA (Figure 3B). One copy of this motif is present in each of the negatively regulated promoters (zwf, eda2, and mgsA), whereas two copies of this motif are present in the positively regulated promoters (pckA and fbaB) (Figure 3A).

Figure 3.

Binding of PckR to the promoter regions of PckR regulated genes. (A) Sequences of the DNA regions upstream of each operon belonging to the PckR regulon. Predicted transcriptional start sites (TSS) are indicated by boldface font and “+1.” The location of the putative PckR binding sites, as identified with the help of MEME (Bailey and Elkan 1994; Bailey et al. 2009), are in blue font and underlined, and the position of the 3′ end relative to the TSS is indicated. The location of the ZW and NS probes used in D are indicated by the red underlines. The location of the −10 and −35 regions of the zwf promoter are indicated by the gray lines above the sequence, and were identified based on the S. meliloti consensus promoter sequence (MacLellan et al. 2006). (B) A logo and the consensus sequence of the PckR DNA binding sequence, based on the seven predicted motifs, is shown. The logo includes the two positions upstream (−1, −2) and downstream (+1, +2) of the binding site, and the location of the mutations in the oligonucleotides used in (D) are shown. (C) Electrophoretic mobility shift assays (EMSAs) were performed using purified PckR protein and ∼135–200-bp ³²P-labeled probes that each included the promoter region of one of the PckR regulated promoters. (D) EMSAs are shown using a probe spanning the zwf promoter region and with an additional 25-bp oligonucleotide as indicated. The molar ratio of the oligonucleotide relative to the probe is 1000:1. Probes M1–M4 are ZW derivatives containing T to G, A to C, C to A, or T to C mutations, respectively, at the positions indicated in (B).

To investigate if PckR could bind the identified motif, EMSAs were used to detect binding of purified PckR protein to labeled DNA fragments. PckR bound all five of the promoter fragments in vitro (Figure 3C). In competitive binding assays, the addition of a 25-bp oligonucleotide (termed ZW), which included the predicted PckR binding site of the zwf promoter, abolished the shift of the zwf promoter probe (Figure 3D). Similar results were obtained if the oligonucleotide included either of the predicted PckR binding sites of the pckA promoter (data not shown). In contrast, a control oligonucleotide (termed NS), lacking the predicted PckR binding site, had no effect on the shift of the zwf promoter probe (Figure 3D). Four variants of the ZW oligonucleotide (termed M1–M4) that each had a single residue change were tested. The three oligonucleotides (M2, M3, and M4; Figure 3B) with mutations of conserved residues of the predicted PckR binding site had little effect on the binding of PckR to the zwf probe, whereas the M1 oligonucleotide with a mutation outside of the conserved region abolished the shift with the zwf promoter probe (Figure 3D). Together, these results strongly suggest that PckR directly regulates expression of the five operons identified above by binding to a palindromic sequence (consensus: TTTMAATCGATTWAAA) in the promoter regions.

Comparison of the location of the PckR binding sites with the predicted location of the promoter motifs, based on transcriptional start site mapping studies (Østerås et al. 1995; Sallet et al. 2013; Schlüter et al. 2013), indicates that the locations of the binding sites differ for positively and negatively regulated promoters. Both of the PckR binding sites in the positively regulated genes are located upstream of the promoter motifs. In contrast, the PckR binding sites of the negatively regulated genes overlap the promoter motif or the transcriptional start site. Together, these data indicate that PckR directly modulates expression of the PckR regulon, with positive vs. negative regulation dependent on the position and number of the PckR binding motifs.

PEP may function as a PckR effector metabolite

As is the case for other LacI/GalR regulatory proteins, we expected an effector molecule to interact with the C-terminal domain of PckR to modulate its DNA binding activity. As we hypothesized the effector molecule was a metabolite of central carbon metabolism, the effect of disrupting central carbon metabolism on the activity of PckR was examined by monitoring pckA expression in various carbon metabolic mutants. Mutation of pckA results in a strong induction of pckA expression in medium containing both glucose and succinate, whereas expression of pckA is low in strains carrying wild-type pckA when grown in the same medium (Figure 4, A and C). The induction of pckA in the pckA mutant is dependent on the growth medium, as pckA expression in the mutant is relatively low when provided with galactose as a carbon source (Figure 4, B and C). The catabolism of galactose by S. meliloti results in the production of G3P (glycolytic intermediate) and pyruvate (gluconeogenic), and is therefore similar to growth with glucose plus succinate, but without the effects of catabolite repression (Figure 1) (Geddes and Oresnik 2012).

Figure 4.

Effect of mutation of central carbon metabolic genes on pckA expression. Growth (closed symbols) and pckA expression (open symbols) of S. meliloti strains containing a pckA⁺::gfp-lacZ fusion as well as the indicated null mutation(s). The GFP fluorescence units (RFU) shown in the graphs are not standardized by OD₆₀₀, and data points represent the mean of triplicate samples. Strains were grown in minimal medium with either (A) 15 mM glucose and 15 mM succinate or (B) 15 mM galactose as carbon sources. The diauxic growth in glucose and succinate is not well represented due to the plotting of data points only once every 2 hr. (C) A key indicating which genotypes are represented by each symbol. The key also shows the standardized GFP fluorescence (RFU/OD₆₀₀) in the stationary phase, calculated as the mean of the final two time points, when grown in either glucose and succinate (Glu/Suc) or in galactose (Gal).

One hypothesis for the induction of pckA in a pckA mutant is that it is a consequence of the accumulation of TCA cycle intermediates during the catabolism of succinate (Figure 1). It was previously shown that TCA cycle intermediates build up in a S. meliloti dme tme malic enzyme double mutant when grown in the presence of succinate (Zhang et al. 2016). However, pckA transcription is not induced in a dme tme mutant grown with both glucose and succinate (Figure 4, A and C), inconsistent with pckA induction being dependent on a TCA cycle intermediate.

A second hypothesis is that the observed pckA induction is due to a reduced concentration of a metabolite whose synthesis depends on the activity of PckA. We therefore examined pckA expression in several mutants with lesions of the EMP pathway (Figure 1). Mutations of eno (enolase), pgk (phosphoglycerate kinase), and gap (glyceraldehyde-3-phosphate dehydrogenase) do not induce strong pckA expression when grown with glucose and succinate (Figure 4, A and C). This suggests that the compound of interest is located between PckA and Eno in the gluconeogenic pathway (Figure 1) As PEP is the only metabolite matching this description, we tested whether PEP would inhibit the binding of PckR to the zwf promoter using EMSAs. Inclusion of 2 mM PEP in the EMSA binding reactions results in a reduction in the amount of zwf probe that is shifted, and an 80% reduction occurs in the presence of 4 mM PEP relative to 0 mM PEP (Figure 5). This indicates that PEP inhibits the DNA binding activity of PckR in vitro. Because intracellular PEP concentrations in the millimolar range are detected in E. coli (Hogema et al. 1998; Hoque et al. 2005; Xu et al. 2012) and S. meliloti (A. Checcucci and A. Mengoni, personal communication), our data are consistent with PEP influencing the ability of PckR to regulate its target promoters at physiologically relevant concentrations. In contrast, pyruvate at concentrations up to 20 mM has no effect on the ability of PckR to bind the zwf promoter (Figure 5). Although this suggests that the effect of PEP is specific, consistent with PEP being the PckR effector molecule, it is possible that the effects of PEP occur through nonspecific ionic interactions via the phosphate group. Further in vitro experiments are required to conclusively demonstrate direct binding of PEP or another effector to the sensory domain of purified PckR.

Figure 5.

Effect of PEP and pyruvate on the DNA binding activity of PckR. Electrophoretic mobility shift assays were performed using purified PckR protein and a ³²P-labeled probe that included the zwf promoter region. Phosphoenolpyruvate or pyruvate were added to each reaction mixture at the indicated concentrations (in millimolar).

The F230, A289, and A304 residues of PckR influence effector binding

Modeling of PckR against known LacI/GalR family crystal structures using Phyre2 (Kelley et al. 2015) reveals prototypical N-terminal DNA binding and C-terminal sensory domains (Figure 6). The C-terminal tetramerization domain present in a subset of LacI/GalR family members, such as LacI (Lewis et al. 1996), is not seen. Østerås et al. (1997) identified three spontaneous pckR mutations, rpk-9, rpk-10, and rpk-15, which altered the regulation of pckA expression. We renamed these alleles as pckR9, pckR10, and pckR15, respectively, and genome sequencing identified single mutations within pckR in each mutant that resulted in F230C, A289T, and A304V amino acid substitutions, respectively. All three of these residues are located within the PckR C-terminal sensory domain (Figure 6), which suggests that these substitutions do not directly impact the DNA binding of PckR, but instead reduce its affinity toward an effector.

Figure 6.

Predicted structure of the S. meliloti PckR protein. The tertiary structure of PckR was predicted using the Phyre2 webserver (Kelley et al. 2015) and visualized with Chimera (Pettersen et al. 2004). The N-terminal DNA binding domain and the C-terminal sensory domain are indicated, as is the predicted ligand binding domain as determined via the 3DLigandSite webserver (Wass et al. 2010). The three amino acids (F230, A289, and A304) changed in the pckR mutants are indicated in red, and are all located in the sensory domain away from the ligand binding site.

The growth and pckA and zwf expression phenotypes of the pckR mutant strains are consistent with the above conclusion. Strains carrying the pckR9, pckR10, and pckR15 alleles grow slowly in minimal medium with glucose, whereas the pckR::ΩSp null mutant grows like wild type (Figure 7, A and D). The opposite was observed during growth with succinate: strains carrying the pckR9, pckR10, and pckR15 alleles grow like wild type, whereas the pckR::ΩSp null mutant displays reduced growth (Figure 7, B and E). Similarly, the pckA and zwf expression profiles of the null mutant are opposite those of the other alleles. In the null pckR::ΩSp mutant, pckA expression is constitutively low, whereas zwf expression is constitutively high (Figure 7). In contrast, zwf expression is constitutively low in strains carrying the pckR9, pckR10, or pckR15 alleles, whereas pckA expression is strongly induced during growth with succinate and during the stationary phase after growth with glucose (Figure 7). These data suggest that the F230C, A289T, and A304V substitutions result in elevated PckR activity, as opposed to the loss of activity of the pckR::ΩSp allele. As the activity of PckR appears inhibited by the binding of an effector, the observed phenotypes are likely a consequence of these substitutions reducing the ability of PckR to associate with its effector molecule.

Figure 7.

Growth and gene expression phenotypes of various pckR alleles. Growth (closed symbols) and gene expression (open symbols) of S. meliloti strains with various pckR alleles grown in minimal medium with either (A and D) 15 mM glucose or (B and E) 15 mM succinate as the sole carbon source. The GFP fluorescence units (RFU) shown in the graphs are not standardized by OD₆₀₀, and data points represent the mean of triplicate samples. (A and B) All strains carried a pckA⁺::gfp-lacZ fusion to monitor expression of pckA through GFP fluorescence. (D and E) All strains carried a zwf⁺::gfp-lacZ fusion to monitor expression of zwf through GFP fluorescence. (C and F) Keys indicating which genotype is represented by each symbol; (C) corresponds to graphs (A and B), whereas (F) corresponds to graphs (D and E). The keys also show the standardized GFP fluorescence (RFU/OD₆₀₀) in the stationary phase, calculated as the mean of the final two time points, when grown in either glucose (Glu) or succinate (Suc). The extent of the expression differences between samples are partially obscured due to the high background in the GFP fluorescence readings (Figure S6 and Figure S7 in File S1).

PckR is restricted to one clade of the Rhizobiales order

A Blast-BBH approach, starting with the PckR amino acid sequence of S. meliloti Rm1021, was used to identify putative PckR orthologs in species of the order Rhizobiales. Note that this approach simply identifies whether two proteins are more similar to each other than to any other protein in the two proteomes, which means that homologous proteins may have been identified as Blast-BBHs even if they do not share functional orthology.

PckR Blast-BBHs are found in all species within the monophyletic group that includes the families Rhizobiaceae, Phyllobacteriaceae, Brucellaceae, and Bartonellaceae, except for the genera Liberibacter and Bartonella that are obligate intracellular pathogens with reduced genomes (Alsmark et al. 2004; Leonard et al. 2012) (Figure 8A and Figure S1 in File S1). The three PckR residues (F230, A289, and A305) identified above as important in effector binding affinity are conserved in all of the proteins, and all shared >75% amino acid identity with the S. meliloti PckR. These proteins therefore likely represent true PckR orthologs. In contrast, only 7 of the 39 Rhizobiales species outside of the above-mentioned monophyletic group contain a PckR Blast-BBH (Figure 8A). These seven proteins form a monophyletic outgroup in a phylogeny of the PckR Blast-BBHs (Figure S2 in File S1), they share only 30–50% identity with the S. meliloti PckR protein, they are quite diverse even from each other, and none contain equivalents of all three of the F230, A289, and A305 residues. Moreover, a plasmid carrying the PckR Blast-BBH from A. caulinodans ORS 571 (azc_4468) failed to complement the pckA and zwf expression phenotypes or the succinate growth phenotype of the S. meliloti pckR::ΩSp^R null mutant (Figure S3 in File S1). We therefore conclude that none of these seven proteins represented true orthologs of PckR, and that PckR is specific to, and highly conserved within, one multifamily clade of the Rhizobiales.

Figure 8.

Phylogenetic analysis of the PckR protein. (A) A RAxML (Stamatakis 2014) maximum likelihood phylogeny based on the concatenated alignments of 20 conserved proteins is shown. The tree was rooted with several α-proteobacteria outgroups, containing one representative genome per species, and was produced as described in the Materials and Methods. Taxa are colored based on the presence (blue) or absence (red) of a PckR Blast-BBH. A compete phylogeny containing the outgroups and bootstrap values is provided as Figure S1 in File S1. (B) A subtree of a 4607-member LacI family protein phylogeny produced with FastTree (Price et al. 2010). Proteins in which the F230, A289, and A304 residues are conserved are indicated in blue. The complete LacI family protein phylogeny is provided as Figure S4 in File S1.

In a phylogeny of 4607 LacI family proteins (Figure 8B and Figure S4 in File S1), all of the predicted PckR proteins from the order Rhizobiales form a monophyletic group, which indicates that pckR has not been transferred and maintained outside of this clade. If PckR originated in a distinct bacterial lineage and was then transferred to an ancestral Rhizobiales species, the Rhizobiales PckR group should be nested within a group of proteins from a distinct bacterial taxon; however, this is not observed. Moreover, the two proteins most closely related to the predicted PckR proteins are from the species Fulvimarina pelagi and Aurantimonas manganoxydans. These species form a deep branching outgroup lineage within the Rhizobiales clade containing the Rhizobiaceae, Phyllobacteriaceae, Brucellaceae, and Bartonellaceae families (Cho and Giovannoni 2003; Denner et al. 2003), but only one of the three residues substituted in the pckR mutants is conserved in these proteins. These observations are consistent with an early Rhizobiales organism obtaining a gene encoding an LacI family protein, and that this protein subsequently gained the PckR function.

Discussion

The PckR transcriptional regulator is found in a subset of the Rhizobiales (Figure 8). PckR regulates expression of central carbon metabolic genes whose products include the entire ED pathway, the committed step of gluconeogenesis (PckA), and FbaB that functions at a crossroads of gluconeogenesis and glycolysis (Figure 1 and Figure 2). We propose a model of PckR-based regulation in which catabolism of glycolytic compounds results in elevated concentrations of an effector molecule that associates with PckR and reduces the affinity of PckR for its DNA binding motif (Figure 9). When concentrations of the effector drop sufficiently, such as during catabolism of gluconeogenic substrates, PckR dissociates from its effector and binds its cognate DNA motif. This results in repression of transcription if the binding site overlaps the promoter motif or transcriptional start site, or induction of transcription if there are two binding sites located upstream of the promoter motifs. The pckR9, pckR10, and pckR15 alleles displayed elevated, but not constitutive, expression of pckA (Figure 7). This suggests that the PckR proteins in these mutants have reduced affinity for its effector and that the F230, A289, and A305 amino acid residues of PckR are important in increasing the affinity of PckR to its effector. The combined genetic and in vitro data presented here (Figure 2, Figure 3, Figure 4, and Figure 5) is consistent with the PckR effector molecule being PEP, although additional studies are required.

Figure 9.

Proposed model of PckR activity. A schematic illustrating the proposed model of PckR activity is shown using pckA and zwf as example genes positively and negatively regulated by PckR, respectively. (A) When intracellular concentrations of the effector are high, such as during growth with glucose, PckR is predominately associated with the effector and disassociated from its target DNA binding sites. This results in high transcription of the negatively regulated promoters, and low expression of the positively regulated promoters. (B) When intracellular concentrations of the effector are low, such as during growth with succinate, PckR is predominately not associated with the effector and is bound to its target DNA binding sites. This results in low transcription of the negatively regulated promoters, and high expression of the positively regulated promoters.

We also attempted to identify additional PckR binding sites in promoters that were not identified as part of the PckR regulon. Using both HMMER (Eddy 2009) and PatScan (Dsouza et al. 1997) searches, we found only one additional sequence that was almost certainly a PckR binding motif; this motif was upstream of the feuNPQ locus, a signal transduction system responsive to osmotic conditions and involved in regulating export of cyclic β-glucans (Griffitts et al. 2008; Carlyon et al. 2010). However, the location of the motif relative to the transcriptional start site suggests that PckR binding would not impact transcription. Several other sequences that shared resemblance with the PckR consensus binding motif are present in the genome; however, they all contained enough mismatches that it was likely that they were nonfunctional. Nevertheless, it is notable that several of the genes with these sequences are related to carbon metabolism, including glgP, thuE, pckR, and sdhC. We hypothesize that at least some of these genes may eventually become integrated into the PckR regulon.

Given the location of the binding sites in the negatively regulated promoters (zwf-pgl-edd, eda2, and mgsA genes), PckR binding presumably impairs the binding or progression of the RNA polymerase, thereby decreasing expression. However, it is less clear how PckR induces expression of the positively regulated promoters of pckA and fbaB (Figure 2). We note that in the absence of PckR, there is sufficient pckA expression to allow growth with succinate as the carbon source (Figure 7B). Although this growth is reduced relative to the wild type, a pckA null mutant fails to grow with succinate as the sole carbon source (Østerås et al. 1995, 1997) (Figure S5 in File S1). One possibility is that PckR directly recruits the RNA polymerase to the target promoters. A second explanation is that PckR indirectly promotes RNA polymerase recruitment by displacing a negative transcriptional regulator. In support of the latter, a pckA promoter truncated at an EcoRI cut site (Figure 3A) that lacks the PckR binding sites is constitutively active (Østerås et al. 1997), even though induction of the full-length pckA promoter is dependent on PckR binding (Figure 2). These contradicting observations can be reconciled by theorizing that PckR does not recruit the RNA polymerase to the promoter, but instead outcompetes a negative transcriptional regulator for promoter occupancy, thereby indirectly providing RNA polymerase access to the promoter. Although most LacI/GalR family regulators repress transcription, several are known to act as repressors and activators and these include the carbon catabolite protein A (CcpA) of low G + C content Gram-positive bacteria (Schumacher et al. 2007), and the cytidine repressor protein (CytR) that regulates transcription of genes in E. coli and Vibrio cholerae (Rasmussen et al. 1996; Swint-Kruse and Matthews 2009; Watve et al. 2015). Both CcpA and CytR interact with effector ligands, DNA, and other regulatory proteins, and further studies are necessary to determine whether PckR interacts with other regulatory proteins at the pckA promoter.

The possibility of PEP modulating PckR activity is interesting for several reasons. PEP is the product of the PCK reaction, which is the committed step of gluconeogenesis. PEP allosterically modulates the activity of several proteins that function early in glycolysis, including phosphofructokinase (Ogawa et al. 2007), the committed and rate-limiting step of glycolysis in E. coli; however, S. meliloti lacks the phosphofructokinase enzyme. Phosphorylation of the proteins of the phosphotransferase system, a global regulatory pathway modulating secondary carbon metabolism in many prokaryotic species, is also influenced by the PEP/pyruvate ratio (Hogema et al. 1998). This is noteworthy as genetic evidence has implicated proteins of the S. meliloti phosphotransferase system in succinate-mediated catabolite repression (Pinedo et al. 2008; Pinedo and Gage 2009). PEP may therefore provide a link between succinate-mediated catabolite repression–based and PckR-based transcriptional regulation that could help coordinate these responses.

Although α-proteobacteria are a metabolically versatile group of biotechnologically important bacteria (Biondi et al. 2009; Imam et al. 2013), there has been little study of the regulation of central carbon metabolism in these organisms (Imam et al. 2015). Recent works suggest that the regulatory pathways in these organisms may be diverse and evolutionarily independent. PckR is specific to a clade of the order Rhizobiales (Figure 8), CceR regulates central carbon and energy metabolism in the order Rhodobacterales (Imam et al. 2015), and in silico predictions suggest that GluR regulates central carbon metabolism in the order Caulobacterales (Novichkov et al. 2013). PckR, CceR, and GluR are LacI-type transcriptional regulators, but all appear to have evolved independently based on a phylogenetic reconstruction of 4607 LacI proteins (Figure S4 in File S1). Experimental analyses indicate that the PckR and the CceR regulons overlap, yet these proteins recognize distinct DNA binding motifs and different metabolites appear to function as their effector (Figure 3 and Figure 7; Imam et al. 2015). How and whether the differences between these transcription factors influence their response changes in nutrient availability is currently unclear, but we hypothesize that differences would be most evident during growth in the presence of multiple carbon substrates.

Supplementary Material

Supplemental material is available online at www.genetics.org/lookup/suppl/doi:10.1534/genetics.117.300212/-/DC1.