The In Vitro Redundant Enzymes PurN and PurT Are Both Essential for Systemic Infection of Mice in Salmonella enterica Serovar Typhimurium

Abstract

Metabolic enzymes show a high degree of redundancy, and for that reason they are generally ignored in searches for novel targets for anti-infective substances. The enzymes PurN and PurT are redundant in vitro in Salmonella enterica serovar Typhimurium, in which they perform the third step of purine synthesis. Surprisingly, the results of the current study demonstrated that single-gene deletions of each of the genes encoding these enzymes caused attenuation (competitive infection indexes [CI] of <0.03) in mouse infections. While the ΔpurT mutant multiplied as fast as the wild-type strain in cultured J774A.1 macrophages, net multiplication of the ΔpurN mutant was reduced approximately 50% in 20 h. The attenuation of the ΔpurT mutant was abolished by simultaneous removal of the enzyme PurU, responsible for the formation of formate, indicating that the attenuation was related to formate accumulation or wasteful consumption of formyl tetrahydrofolate by PurU. In the process of further characterization, we disclosed that the glycine cleavage system (GCV) was the most important for formation of C₁ units in vivo (CI = 0.03 ± 0.03). In contrast, GlyA was the only important enzyme for the formation of C₁ units in vitro. The results with the ΔgcvT mutant further revealed that formation of serine by SerA and further conversion of serine into C₁ units and glycine by GlyA were not sufficient to ensure C₁ formation in S. Typhimurium in vivo. The results of the present study call for reinvestigations of the concept of metabolic redundancy in S. Typhimurium in vivo.

INTRODUCTION

Salmonella enterica is a common cause of foodborne disease worldwide. Annually, more than 93 million people have been estimated to suffer from nontyphoid salmonellosis, and more than 155,000 people succumb to the disease (1). Infection of mice with S. enterica serovar Typhimurium does not cause diarrhea but results in a systemic, life-threatening condition in which bacteria predominantly localize to cells of the immune system in the liver and spleen. For this reason, S. Typhimurium infection of mice is used as a model for systemic salmonellosis, including infection with the host-specific serovar S. Typhi (2).

The diversity of intracellular and extracellular host niches occupied by S. Typhimurium is reflected in a high degree of metabolic flexibility (3). This flexibility is achieved through component and system-level redundancy in the metabolic network (4), and recent years have seen several studies of S. Typhimurium in order to understand the importance of metabolic redundancy for its pathogenic lifestyle (5–8). These studies have used genome-scale metabolic modeling to predict essential and combined lethal metabolic reactions; the latter group consists of two or more nonessential reactions which, when considered as one unit, are found to be essential. Such combinations are referred to as minimal cut sets in metabolic modeling (9).

A list of 102 cut sets of metabolic reactions in S. Typhimurium was recently produced using a novel genome-scale metabolic model. Each cut set was predicted to be essential for growth in a modified M9 minimal medium, and the underlying assumption was that blocking such combinations of reactions would attenuate S. Typhimurium during infection. One cut set was the combination of reactions carried out by the enzymes PurN (glycinamide-RNase-transformylase N) and PurT (glycinamide-RNase-transformylase T) in the purine biosynthesis pathway (8). A study by Thiele et al. (5), using a similar approach, also predicted that combined blocking of PurN and PurT would be detrimental to growth.

De novo purine synthesis consists of 10 steps that convert phosphoribosyl-pyrophosphate (PRPP) to IMP (10). In proteobacteria, the third step, which converts 5′-phosphoribosyl-glycinamide (GAR) to formyl-phosphoribosyl-glycinamide (fGAR), is carried out by the two enzymes in the predicted cut set, PurN and PurT, using different formyl donors. Blockage of this step results in accumulation of GAR and depletion of all downstream products, which is why it is considered to be essential for purine synthesis (11). The reactions carried out by PurN (GART-RXN) and PurT (GARTRASFORMYL2-RXN) are thus textbook examples of functional redundancy in metabolic reactions, and their prediction as a cut set in S. Typhimurium was not surprising.

PurN and PurT are not isoenzymes, because PurN, which is found in both prokaryotes and eukaryotes (10, 12), obtains the formyl group for generation of fGAR from formyl tetrahydrofolate (fTHF), while the prokaryote-specific enzyme PurT uses formate (13) (Fig. 1A). As shown in Fig. 1B, PurN and PurT create a link between purine synthesis and folate carbon-1 metabolism, by which essential C₁ units are formed (14). Aside from delivering carbon-2 and carbon-8 for purine synthesis, the folate metabolism delivers methyl groups for the amino acid methionine and the deoxynucleotide dTMP. PurT is unique in that it uses formate as a C₁ donor. Under anaerobic conditions, formate is formed from pyruvate in the oxygen-sensitive pyruvate formate lyase (PFL) reaction (15), but under aerobic conditions, formate is derived solely from fTHF in a reaction catalyzed by the PurU enzyme (16) (Fig. 1A). If the PurT enzyme does not use formate, it is a dead-end product in aerobic metabolism, and it cannot be fueled back into the C₁ metabolism (17). As in Escherichia coli (14, 18), C₁ units for amino acid and purine synthesis are produced by the GlyA reaction during conversion of serine into glycine and from degradation of glycine to ammonia and carbon dioxide by the glycine cleavage system (GCV) (Fig. 1B).

FIG 1.

Third reaction step in de novo purine synthesis, carried out by the in vitro redundant enzymes PurN and PurT (A), and interconnection between purine synthesis and C₁ metabolism, by which essential C₁ units are produced (B). (A) PurN converts GAR to fGAR by using fTHF as a formyl donor, while PurT converts GAR to fGAR by using formate as a formyl donor. Formate is produced from fTHF by the enzyme PurU. (B) The production of C₁ units for amino acid and purine synthesis in S. Typhimurium happens when GlyA converts serine into glycine and when glycine is converted into ammonia and carbon dioxide by the glycine cleavage system. The pools of formyl THF (fTHF), THF, and methylene THF (mTHF) are shared between the purine and C₁ synthesis pathways.

Mutants that lack purN, purT, or both genes have been constructed and characterized in E. coli (11). purN and purT mutants grew slightly slower than their parental strains in minimal medium under both aerobic and anaerobic conditions. Interestingly, growth of the purN mutants was inhibited by addition of glycine under aerobic but not anaerobic conditions, suggesting that the effect of glycine was due to limiting formate production. A kinetic analysis of the purified PurU enzyme offered an explanation for this phenomenon, as its hydrolase activity was severely inhibited by glycine (18). In the purN mutant, glycine inhibition of formate production by PurU thus prevented fGAR synthesis by PurT. When the authors found the PurU activity to also be activated by histidine, they proposed that the PurU enzyme functions as a regulator that balances the folate intermediates tetrahydrofolate (THF), methylene tetrahydrofolate (mTHF), and formyl tetrahydrofolate (fTHF) as a function of the glycine and methionine concentrations.

We currently lack adequate knowledge on the in vivo metabolism (in the host) of pathogenic bacteria, even though this is arguably just as important as virulence factors for the ability of a pathogenic bacterium to carry out an infection (19). This includes a lack of knowledge on which nutrients bacteria can scavenge from the host, which metabolites different bacteria need for synthesis, and how these differ between different bacteria, different hosts, and even different places the bacterium occupies in the same host during infection. In the absence of experimental data, we rely on deductions based on in vitro (laboratory) growth phenotypes and model simulations. In some situations, we observe good correlations between in vitro and in vivo phenotypes with respect to prediction of the ability to carry out infection (8), but sometimes this is not the case, probably because functional auxotrophy may arise when metabolites are present at levels below an important threshold during infection of the host, resulting in unforeseen regulatory effects. Under such conditions, predicted redundant reactions may turn out to be nonredundant. In the current study, we show that the universally acknowledged redundant enzyme pair PurN/PurT does not show functional redundancy in S. Typhimurium during infection of mice. Rather, each enzyme itself is essential for infection. Likewise, the redundant enzyme pair MetE/MetH was shown to be nonredundant during mouse infection. An important message from this study is that well-established pairs of redundant enzymes may be functionally nonredundant in vivo and cannot be classified as redundant a priori based upon metabolic modeling. Thus, there is a dire need to reinvestigate the concept of metabolic redundancy in Salmonella and other pathogenic bacteria in the in vivo situation.

MATERIALS AND METHODS

Bacterial strains.

S. Typhimurium 4/74 was used as the wild-type (WT) strain and for the construction of mutant strains by lambda Red-mediated recombination (20), essentially as previously described (8) (Table 1). Mutated alleles were transformed to a clean wild-type S. Typhimurium 4/74 background by P22HT105/int⁻20-mediated transduction as described previously (8). Transduction was also used to construct double and triple mutants. For construction of triple mutants, the resistance marker, normally a kanamycin resistance gene, was flipped out by using the FLT system as described previously (20), while each gene in the double mutants contained a different antibiotic resistance cassette. Primers for mutant construction are listed in Table S1 in the supplemental material, together with primers used to control the mutations, generally by two PCRs: one targeting the inserted antibiotic resistance gene and the flanking regions of the desired genes and one targeting both flanking regions.

TABLE 1.

Strains and plasmids used in the study^a

Strain no. or plasmid	Strain name	Genotype and marker or relevant characteristics	Source or reference
Jeo3774	S. Typhimurium 4/74	Wild type	47
Jeo1473	S. Typhimurium ΔpurN::Cm	4/47 ΔSTM474_2603 Cmr	This study
Jeo1516	S. Typhimurium ΔpurN::Cm-comp	4/47 ΔSTM474_2603 + STM474_2603comp Cmr Ampr	This study
Jeo1496	S. Typhimurium ΔpurT::Kan	4/47 ΔSTM474_1915 Kanr	This study
Jeo1517	S. Typhimurium ΔpurT::Kan-comp	4/47 ΔSTM474_1915 + STM474_1915comp Kanr Ampr	This study
Jeo1509	S. Typhimurium ΔpurN::Cm ΔpurT::Kan	4/74 ΔSTM474_2603 ΔSTM474_1915 Kanr Cmr	This study
Jeo1512	S. Typhimurium ΔglyA::Kan	4/47 ΔSTM474_2659 Kanr	8
Jeo1528	S. Typhimurium ΔglyA::Cm	4/47 ΔSTM474_2659 Cmr	This study
Jeo1570	S. Typhimurium ΔglyA::Kan-comp	4/47 ΔSTM474_2659 + STM474_2659comp Kanr Ampr	This study
Jeo1526	S. Typhimurium ΔserA::Kan	4/47 ΔSTM474_3209 Kanr	This study
Jeo1514	S. Typhimurium ΔserA::Kan ΔglyA::Cm	4/47 ΔSTM474_3209 ΔSTM474_2659 Kanr Cmr	This study
Jeo1527	S. Typhimurium ΔpurT::Kan ΔglyA::Cm	4/74 ΔSTM474_1915 ΔSTM474_2659 Kanr Cmr	This study
Jeo1522	S. Typhimurium ΔpurT::Kan ΔserA::Cm	4/74 ΔSTM474_1915 ΔSTM474_3209 Kanr Cmr	This study
Jeo1521	S. Typhimurium ΔpurN::Kan ΔglyA::Cm	4/74 ΔSTM474_2603 ΔSTM474_2659 Kanr Cmr	This study
Jeo1525	S. Typhimurium ΔpurN::Kan ΔserA::Cm	4/74 ΔSTM474_2603 ΔSTM474_3209 Kanr Cmr	This study
Jeo1523	S. Typhimurium ΔpurU::Kan	4/47 ΔSTM474_1773 Kanr	This study
Jeo1577	S. Typhimurium ΔpurU-comp	4/47 ΔSTM474_1773 + STM474_1773comp Ampr	This study
Jeo1572	S. Typhimurium ΔpurU ΔpurT::Kan	4/47 ΔSTM474_1773 ΔSTM474_1915 Kanr	This study
Jeo1529	S. Typhimurium ΔgcvT::Cm	4/47 ΔSTM474_3202 Cmr	This study
Jeo1531	S. Typhimurium ΔserA::Kan ΔgcvT::Cm	4/74 ΔSTM474_3209 ΔSTM474_3202 Kanr Cmr	This study
Jeo1530	S. Typhimurium ΔglyA::Kan ΔgcvT::Cm	4/74 ΔSTM474_2659 ΔSTM474_3202 Kanr Cmr	This study
Jeo1574	S. Typhimurium ΔmetE::Kan	4/47 ΔSTM474_4143 Kanr	This study
Jeo1590	S. Typhimurium ΔmetH::Cm	4/47 ΔSTM474_4378 Cmr	This study
Jeo1593	S. Typhimurium ΔmetE::Kan ΔmetH::Cm	4/74 ΔSTM474_4143 ΔSTM474_4378 Kanr Cmr	This study
Jeo1599	S. Typhimurium ΔmetE::Kan ΔmetH::Cm-comp	4/74 ΔSTM474_4143 ΔSTM474_4378 + STM474_4143comp	This study
Jeo1594	S. Typhimurium ΔmetE::Kan ΔmetH::Cm ΔpurT::Kan	4/74 ΔSTM474_4143 ΔSTM474_4378 ΔSTM474_1915 Kanr Cmr	This study
	S. Typhimurium ΔssaV::Kan	4/47 ΔSTM474_1420 Kanr	21
pKD3		Template plasmid for amplification of chloramphenicol resistance gene cassette; Ampr Cmr	20
pKD4		Template plasmid for amplification of kanamycin resistance gene cassette; Ampr Kanr	20
pKD46		Lambda Red-mediated recombination system; Ampr	20
pACYC177		Low-copy-number plasmid used for complementation; Ampr	48

^aGene numbers refer to the 4/74 genome annotation in BioCyC (www.biocyc.org). Cm^r, chloramphenicol resistant through insertion of the chloramphenicol gene cassette from the plasmid pKD3; Kan^r, kanamycin resistant through the insertion of the kanamycin resistance gene cassette from the plasmid pKD4; comp, complemented in trans by cloning of the deleted gene into the plasmid pACY177.

Genetic complementation was obtained by PCR amplification of the relevant gene and subsequent cloning into the low-copy-number plasmid pACYC177, followed by transformation of the resulting plasmid into the mutant strain, essentially as described previously (21). Primers are listed in Table S1 in the supplemental material. The XhoI and BamHI restriction enzymes were used for cloning according to the manufacturer's recommendations (Thermo Scientific, Denmark), and constructs were verified by restriction analysis and sequence analysis.

Culture and growth conditions.

Growth of bacteria in rich media was done in Difco Lennox lysogeny broth (LB; Becton, Dickinson and Company, Albertslund, Denmark) and on LB agar plates (Becton, Dickinson and Company, Albertslund, Denmark), and growth in minimal medium was carried out in M9 broth (2 mM MgSO₄, 0.1 mM CaCl₂, 0.4% glucose, 8.5 mM NaCl, 42 mM Na₂HPO₄, 22 mM KH₂PO₄, 18.6 mM NH₄Cl). Chloramphenicol (10 μg/ml), kanamycin (50 μg/ml), glycine (0.55 mg/ml), serine (0.55 mg/ml), and methionine (0.55 mg/ml) (Sigma, Denmark) were added when appropriate.

Growth phenotypes of mutants were determined by first growing bacteria overnight in LB flasks at 37°C with shaking (200 rpm). Each culture was diluted 40-fold with 0.2 ml M9 broth, and growth was monitored for 24 h at 37°C with shaking (250 rpm) by measuring the optical density at 600 nm (OD₆₀₀) every 15 min in a BioScreen C format with biological triplicates and technical replicates. The wild-type strain and blank wells were included as controls. Growth curves were extracted using Excel (Microsoft, San Diego, CA), and OD₆₀₀ values were corrected based on the blank control values.

Microscopic appearance of bacteria.

The microscopic appearance of bacteria was determined by phase-contrast microscopy at fixed time points and under fixed conditions, using an AxioCamHR4 phase-contrast microscope. Three hundred individual cells were observed to determine the most common cell morphology. Continuous observations of cell morphology during growth in LB medium at 37°C were done using an oCelloScope bright-field camera (magnification, approximately ×200; resolution, 1.3 μm) as reported previously (22).

Macrophage survival experiments.

Survival and multiplication of bacteria inside J774 macrophages were measured as described previously (21). S. Typhimurium 4/74 was used as the reference wild type, and a ΔssaV mutant in the 4/74 background (21) was used as a negative control. Deletion of the ssaV gene renders S. Typhimurium incapable of intracellular replication (23). Briefly, the multiplicity of infection (MOI) was 5, and a 25-min incubation was allowed for the initial uptake of bacteria, after which gentamicin (100 μg/ml) was added for 1 h and then replaced with 25 μg/ml gentamicin for the rest of the experiments. CFU counts for three or four biological repeats with technical duplicates were obtained at the point of the first addition of gentamicin and after 1 h, 2 h, and 21 h of incubation with the drug.

Mouse infections.

Measurement of infection efficacy was performed using a systemic model of infection in C57/BL6 female mice (Taconic, Denmark). A competitive challenge model (24) in which wild-type and mutant strains were given together to the same mouse was used as described previously (8). Briefly, mice were inoculated by the intraperitoneal (i.p.) route with 0.1 ml of an approximately 1:1 mixture of wild-type and mutant strains in phosphate-buffered saline (PBS). The inoculum was standardized to contain a challenge dose of 5 × 10³ bacteria of each strain by using CFU measurements. The exact amount of each strain in the inoculum was determined by plating serial dilutions on LB plates. The ratio of wild-type to mutant bacteria in the spleen was determined at 6 days postinoculation by plating a dilution series on LB agar and subsequently determining the resistance (chloramphenicol or kanamycin resistance) of 100 colonies. Sensitive bacteria corresponded to the wild-type strain, and resistant bacteria corresponded to mutants. The competitive index (CI) was calculated as the mutant/wild-type ratio of the spleen count versus the mutant/wild-type ratio of the inoculum. Severely affected mice were humanely killed. If the spleens of such mice contained >10⁵ CFU Salmonella, this was expected to be the cause of the disease, and colony counts for such mice were included in the competitive index scoring.

Ethical statement.

Mouse challenge experiments were conducted with permission granted to the senior author from the Danish Animal Experiments Inspectorate according to Danish by-law 474 of 15 May 2014 (license number 2009/561-1675).

Statistical analysis.

Statistical differences between wild-type and mutant strains in CFU counts and in virulence measured in mice were determined using GraphPad Prism, version 5.0 (GraphPad software), with one-sample t test analysis. Grubb's outlier test was performed to exclude outliers, with a significance threshold of a P value of 0.05.

RESULTS

PurN and PurT are redundant in vitro but not in vivo.

PurN and PurT have been reported to be redundant for growth of E. coli, although purN and purT mutants grow somewhat slower than the wild type (11). Single mutants of S. Typhimurium 4/74, created by deletion of the purN and purT genes, grew as well as the parental wild-type strain in minimal medium (for the WT, μ = 0.37 ± 0.017; for the ΔpurN strain, μ = 0.35 ± 0.01; and for the ΔpurT strain, μ = 0.36 ± 0.02), while the double mutant did not grow. The method of growth determination for the ΔpurN mutant is shown in Fig. S1 in the supplemental material to demonstrate the way that specific growth rates were determined. When the double mutant was grown in rich medium (LB), it had no growth defect (data not shown). This corresponds to PurN and PurT being redundant for growth in minimal medium in vitro.

When the mutants were analyzed for virulence in mice in competitive challenge experiments, both single and double mutants were severely attenuated, showing competitive indexes (CIs) below 0.03. The virulence of single mutants could be raised to normal levels by complementation with the wild-type genes in trans (Table 2), showing that the attenuation was due to the lack of PurN or PurT. Thus, in the infection situation, one or more factors are limiting for growth, in an enzyme-specific manner, for the presumed redundant enzyme pair PurN/PurT, clearly showing that these enzymes are not redundant in vivo during infection of mice.

TABLE 2.

Growth and virulence of S. Typhimurium 4/74 pur mutants^a

Strain	Mutation(s)	Growth rateb (μ [h−1])	Virulence in mice (CI)	Growth in J774A.1 macrophagesc (multiplication/20 h)
Jeo3774	WT 4/74	0.37 ± 0.017	1.00	9.7 ± 2.5
Jeo1473	ΔpurN	0.35 ± 0.01	0.03 ± 0.0***	4.4 ± 3.5*
Jeo1516	ΔpurN/pACYC177purN	0.31 ± 0.08	0.60 ± 0.2	ND
Jeo1496	ΔpurT	0.36 ± 0.02	0.02 0.01***	8.0 ± 5.6
Jeo1517	ΔpurT/pACYC177purT	0.37 ± 0.01	0.90 ± 0.3	ND
Jeo1509	ΔpurN ΔpurT	NG	0.00 ± 0.0NT	0.3 ± 0.2****
Jeo1523	ΔpurU	0.25 ± 0.002***	0.60 ± 0.8	4.6 ± 2.5*
Jeo1575	ΔpurU/pACYC177purU	0.15 ± 0.002***	ND	2.1 ± 0.1*
Jeo1572	ΔpurU ΔpurT	0.38 ± 0.01	1.63 ± 0.6**	ND

^aData are means ± standard deviations. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NT, could not be tested because no mutant colonies were recovered from challenged mice; ND, not determined; NG, no growth in M9 medium plus glucose.

^bSpecific growth rate in vitro in M9 medium plus glucose (ln₂/doubling time).

^cFor the control strain, 4/74 ΔssaV, the multiplication/20 h was 2.1 ± 0.6***.

Mutation of purN but not purT attenuates the strain during interaction with cultured macrophages.

Interaction with macrophages is an important step in the development of systemic salmonellosis in mice, and the majority of mutants that fail to grow in cultured macrophages have turned out to be attenuated during mouse infection (25). To investigate the ability of bacteria to grow in macrophages, we challenged cultured J774A.1 macrophages with the ΔpurN and ΔpurT mutants. As shown in Table 2, the WT strain was found to multiply 9.7 ± 2.5 times in 20 h. This corresponds to an approximately 7-h generation time and to estimates for multiplication of S. Typhimurium in the mouse spleen in vivo (26). In contrast, the ΔpurN mutant multiplied only 4.4 ± 3.5 times, which was significantly less than the wild-type strain's growth (P = 0.03). The ΔpurT mutant resembled the wild-type strain and multiplied 8.0 ± 5.6 times, suggesting that the attenuation in virulence caused by the deletion of purT was not related to interaction at the macrophage level, while attenuation of the ΔpurN mutant might be related to replication in cells of this type. The double mutant was severely attenuated, and the number of bacteria after 20 h was reduced 70% compared to the number of cells taken up by the macrophages (Table 2). The control strain, S. Typhimurium 4/74 lacking ssaV, encoding an effector protein of the type three secretion system (SPI-2), associated with intracellular multiplication (23), multiplied 2.1 ± 0.6 times, which was within the expected range, showing that the assay was performing as expected. The differences in multiplication were not caused by different starting concentrations, since all strains were taken up by the macrophages to the same degree (data not shown).

Cooperation between PurT and PurU in vitro, in macrophages, and during mouse infection.

Conversion of GAR to fGAR by the PurT enzyme requires free formate, provided by PurU (16). If the attenuation caused by deletion of purT was related to an accumulation of formate, then deletion of purU in a ΔpurT background should eliminate the attenuation.

In contrast to the ΔpurT mutant, a ΔpurU mutant showed no growth defect in E. coli (16). Interestingly, the opposite was observed for S. Typhimurium. The ΔpurT mutant was not growth arrested, whereas a purU deletion in S. Typhimurium strain 4/74 resulted in reduced growth in M9 minimal medium (μ = 0.25 ± 0.002; P < 0.05 compared to the WT strain). Complementation by providing the purU gene in trans did not restore wild-type growth; on the contrary, it reduced the growth rate further (μ = 0.15 ± 0.002) (Table 2). A possible explanation is that provision of PurU in trans on a plasmid results in too high enzyme levels and excessive conversion of fTHF to THF and formate. A purT deletion in the ΔpurU mutant restored normal growth (μ = 0.38 ± 0.01). This showed that the low growth rate of the ΔpurU mutant was not related to a lack of substrate for PurT and that the growth attenuation of the ΔpurT mutant probably was caused by formate accumulation or wasteful conversion of fTHF to THF, which disappears when PurT and PurU are absent in the same bacterium.

In macrophage experiments, the ΔpurU mutant was significantly attenuated in multiplication over 20 h (4.6 ± 2.5; P < 0.05). The ΔpurU mutant was also significantly attenuated in mouse virulence, albeit not to a level that resembled those of the ΔpurN and ΔpurT mutants (Table 2). In accordance with the phenotype seen during growth in M9 medium, a completely restored virulence of the ΔpurT ΔpurU double mutant was seen in the mouse assay (CI = 1.6) (Table 2).

Glycine but not serine is available for C₁ production during mouse infection.

Purine synthesis and C₁ metabolism are closely linked, and the reason that PurN and PurT are not functionally redundant in vivo during infection of mice could be related to the need to secure sufficient purine synthesis and C₁ metabolism for these to occur concurrently. Detailed investigations of the role of C₁ metabolism of S. Typhimurium in virulence have not been reported, and we therefore undertook a series of characterizations of mutants of S. Typhimurium serine and glycine metabolism.

We previously predicted and validated by challenge experiments that a ΔserA mutant, which is unable to synthesize serine and therefore relies on uptake of serine or production of serine from glycine for serine production (Fig. 1B), is fully virulent in mice (8). Others have likewise predicted by in silico modeling that SerA is not needed for growth in vivo (5). From this, we previously concluded that serine was most likely taken up from the host environment but that further studies were needed to understand the relative contributions of GlyA and the glycine cleavage enzyme, GCV, to the phenotype of the ΔserA mutant (8). In the present study, we analyzed a ΔgcvT mutant that cannot convert glycine to C₁ units and CO₂ (Fig. 1B). Whether the ΔgcvT mutation was present alone or together with the ΔserA mutation, virulence was reduced significantly. The strain multiplied like the wild-type strain in cultured macrophages, suggesting that the limiting step in infection was not intracellular multiplication in this cell type (Table 3). The results show that exogenous glycine, not serine as previously believed, was available during infection. If serine had been available in sufficient amounts for synthesis of glycine and C₁ units, then the ΔgcvT mutation would have had no effect. This appears to be the situation in vitro, since the ΔgcvT mutant grew as well as the wild-type strain in M9 medium (μ = 0.31 ± 0.002), while the ΔglyA mutant was severely growth attenuated. This phenotype could be reversed by addition of the wild-type glyA gene in trans, and addition of glycine to the minimal medium also partially restored growth (Table 3). The wild-type 4/74 strain was not affected by addition of glycine to the medium (data not shown). These observations suggest that GlyA is the enzyme that is most important for production of C₁ units in vitro, while the glycine cleavage system has this role in vivo, which underscores the difficulty in predicting in vivo importance from in vitro growth phenotypes. We can also conclude from the ΔgcvT single mutant data that serine production through the SerA enzyme (present in the ΔgcvT single mutant) and subsequent conversion to glycine and C₁ units by the GlyA enzyme (also present) are not sufficient to support the virulence of Salmonella Typhimurium 4/74 in the mouse model.

TABLE 3.

Growth and virulence phenotypes of mutants in glycine and serine metabolism in S. Typhimurium 4/74^a

Strain	Genotype	Growth rateb (μ [h−1])	Virulence in mice (CI)	Multiplication in macrophages (1 h to 21 h)c
Jeo3774	WT 4/74	0.37 ± 0.017	1.00	6.9 ± 3.4
Jeo1529	ΔgcvT	0.36 ± 0.002	0.03 ± 0.03***e	7.2 ± 3.9
Jeo1531	ΔserA ΔgcvT	ND	0.05 ± 0.04***	ND
Jeo1522	ΔserA ΔpurT	NG	0.15 ± 0.15***	ND
Jeo1512	ΔglyA	0.03 ± 0.003	0.3 ± 0.4*	0.3 ± 0.03**
Jeo1512	ΔglyA + Glyd	0.20 ± 0.01	ND	ND
Jeo1570	ΔglyA/pACYC177glyA	0.35 ± 0.01	ND	5.6 ± 1.50e
Jeo1514	ΔglyA ΔserA	NG	0.01 ± 0.01***	ND
Jeo1530	ΔglyA ΔgcvT	NG	0.0 ± 0.0NT	ND
Jeo1521	ΔglyA ΔpurN	0.03 ± 0.003	0.02 ± 0.02***	1.2 ± 1.0***

^aND, not done; NG, no growth in M9 medium plus glucose; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NT, the strain could not be tested statistically because no mutant colonies were recovered from challenged mice. If, for statistical purposes, one assumed one colony to be obtained from each mouse of this double mutant, the results would not have been significantly different from the competitive index of the ΔgcvT mutant on its own (P = 0.06) but would have been significantly different from that of the glyA mutant (P = 0.03).

^bSpecific growth rate in vitro in M9 medium plus glucose (ln₂/doubling time).

^cFor the control strain, 4/74 ΔssaV, the multiplication/20 h was 2.1 ± 0.6***.

^dGrowth in M9 minimal medium with supplementation of glycine (Gly).

^eThe multiplication rate is significantly different from that of Jeo1512 (ΔglyA) but not from that of Jeo3774 (wild-type strain).

These conclusions were corroborated by the analysis of the ΔglyA mutant, which had only partly reduced virulence (Table 3). If glycine were normally synthesized from serine through the GlyA enzyme, then the mutation would have had severe consequences. As expected, the introduction of a ΔserA mutation (in itself a dispensable gene) in the ΔglyA strain (partly avirulent) resulted in total avirulence, because now the cell had no means of obtaining serine for protein synthesis. For the same reason, Thiele et al. (5) predicted this combination to be lethal in S. Typhimurium. Deletion of the gcvT gene in the ΔglyA background increased attenuation even further than that with the two single mutations, and no colonies were obtained from mice challenged with this strain. If one assumed, for statistical purposes, the presence of one colony per mouse, this attenuation would not be significantly different from that of the ΔgcvT mutant on its own but significantly more attenuated than the ΔglyA mutant (Table 3). This observation was expected, because now the production of C₁ units for purine and methionine synthesis was totally blocked. In macrophages, growth of the ΔglyA mutant was highly impaired, and a wild-type glyA gene provided in trans complemented this phenotype (Table 3), suggesting that either glycine biosynthesis or formation of serine from glycine is a limiting factor for S. Typhimurium growth in macrophages, but also that the lower multiplication rate does not lead to total avirulence in mice (CI = 0.3).

The results obtained with the glyA mutant to some extent contradict our previous report (8), in which we concluded that the ΔglyA deletion mutant was more attenuated in mice than the case reported here. Upon microscopic analysis of the different mutants, we discovered a possible reason for this discrepancy. The ΔglyA deletion or deletion of glyA in combination with gcvT resulted in a mixture of normally shaped and elongated bacteria (see Fig. S2 in the supplemental material). The altered relationship between the optical density and the number of CFU for the ΔglyA mutant cultures may have led us to overestimate the importance in virulence in the previous study, in which we used OD values to prepare the inoculum, because the elongated cells resulted in a lower mutant-to-wild-type ratio in the input pool. In the present study, the infections were performed at a 1:1 ratio based upon the number of CFU, eliminating this problem. Elongation of cells could be eliminated by complementation in trans (data not shown), showing that cell division was somehow affected by elimination of the GlyA reaction. We did not enquire further into this interesting phenotype in the present study.

We showed above that virulence attenuation of the ΔpurT mutant was probably associated with an imbalance in formate or fTHF conversion. To investigate whether a lack of THF production affected the phenotype of mutants in C₁ metabolism, we characterized a number of combined purine synthesis and C₁ metabolism mutants. All four combinations of ΔserA or ΔglyA mutations with mutations in purN and purT were growth attenuated to the same extent as the single ΔserA or ΔglyA mutants in vitro (for combinations with the ΔpurN mutation, μ = 0.00 [ΔserA mutant] and 0.04 ± 0.003 [ΔglyA mutant]; for combinations with the ΔpurT mutation, μ = 0.00 [ΔserA mutant] and 0.03 ± 0.00 [ΔglyA mutant]). From this, we concluded that there was no significant effect of extra deletions of purN and purT in the ΔserA and ΔglyA mutants in vitro, even though the competing use of mTHF for purine biosynthesis was prevented. We also tested the ΔpurN ΔglyA strain in the mouse model, and this strain was not significantly different from the ΔpurN mutant on its own (CI = 0.02 ± 0.02), while the ΔpurT ΔserA mutant apparently was less attenuated than the ΔpurT strain (CI = 0.15 ± 0.15); however, the difference was not statistically significant (P > 0.05). Together, these experiments indicated little or no overlap between the two systems.

Synthesis of methionine from homocysteine and C₁-THF in vivo.

Methionine synthesis is also interlinked with purine synthesis through the common pool of mTHF and THF (Fig. 1B), and we therefore also considered its importance for the results obtained with S. Typhimurium lacking PurN or PurT. Methionine is synthesized by two homocysteine-methylating enzymes, MetE and MetH (27). The apparent redundancy of the reaction should enable the strain to synthesize methionine even when either the metE or metH gene is mutated, and previous predictions of redundancy in S. Typhimurium predicted that the two enzymes form a cut set (6). However, while the MetE enzyme is functional under aerobic growth conditions, the MetH enzyme requires vitamin B₁₂, which in S. Typhimurium is synthesized only under anaerobic conditions (28). The ΔmetE mutant was found to be highly growth retarded (μ = 0.09 ± 0.03), most probably because the bacterium then relies on the anaerobic enzyme, MetH. Interestingly, however, a ΔmetH mutant was also found to grow at a lower growth rate than that of the wild type in M9 medium (μ = 0.20 ± 0.01), suggesting that the MetH enzyme is important under the growth conditions tested (Table 4). As expected, a ΔmetE ΔmetH double mutant did not grow in M9 minimal medium, since no methionine was available to this strain for protein synthesis. This phenotype could be complemented by addition of methionine to the medium (μ = 0.31 ± 0.02), a step that did not affect growth of the wild-type strain (data not shown) and that shows that methionine biosynthesis is dispensable when methionine can be taken up from the environment. The same was the case for the growth defects of the ΔmetH single mutant (μ = 0.35 ± 0.03 with methionine supplied). When the metE gene was supplied in trans to the ΔmetE ΔmetH double mutant, the growth was fully restored (μ = 0.24 ± 0.03) to the level of the ΔmetH single mutant, showing the importance of the MetH enzyme during aerobic growth, even with the MetE protein expressed from a multicopy plasmid. Addition of methionine to the ΔmetE mutant in M9 medium likewise raised the growth rate (μ = 0.20 ± 0.01). A triple mutant in which the double metE metH mutations were combined with mutation of purT did not grow in M9 medium, corresponding to the phenotype of the ΔmetE ΔmetH mutant on its own. Addition of methionine restored growth completely for this mutant (μ = 0.35 ± 0.01), which also corresponded to the phenotype of the ΔmetE ΔmetH mutant on its own. Taken together, the results were interpreted as showing no or very little interaction between PurT and MetH/MetE, and vice versa.

TABLE 4.

Virulence phenotypes of metE and metH mutants in S. Typhimurium 4/74^a

Strain	Genotype	Growth rateb (μ [h−1])	Virulence in mice (CI)	Multiplication rate in macrophages (1 h to 21 h)d
Jeo3774	WT 4/74c	0.37 ± 0.02	1.00	6.9 ± 3.4
Jeo1590	ΔmetH	0.20 ± 0.01	0.4 ± 0.2	ND
Jeo1590	ΔmetH + Metc	0.33 ± 0.03	ND	ND
Jeo1574	ΔmetE	0.06 ± 0.01	0.0 ± 0.0	ND
Jeo1574	ΔmetE + Metc	0.23 ± 0.05	ND	ND
Jeo1593	ΔmetH ΔmetE	NG	0.0 ± 0.0NT	2.9 ± 0.4
Jeo1593	ΔmetH ΔmetE + Metc	0.26 ± 0.005	ND	ND
Jeo1599	ΔmetH ΔmetE/pAYCY177metE	0.24 ± 0.002	0.00 ± 0.0NT	ND

^aND, not done; NG, no growth in M9 medium plus glucose; NT, could not be tested because no mutant colonies were recovered from challenged mice.

^bGrowth rate in vitro in M9 medium plus glucose.

^cGrowth in M9 minimal medium with supplementation of methionine.

^dFor the control strain, 4/74 ΔssaV, the multiplication/20 h was 2.1 ± 0.6***.

The ΔmetH mutant was slightly but significantly attenuated during mouse infection, with a competitive index of 0.4 (P < 0.001), while the ΔmetE single mutant and the ΔmetE ΔmetH double mutant were totally avirulent (Table 3). Although the metE gene in trans fully complemented the growth phenotype of the double mutant to the level of the ΔmetH mutant in M9 medium, the presence of the metE gene on a multicopy plasmid did not render the double mutant virulent at all.

Methionine appears to be available in macrophages to some extent, as the ΔmetE ΔmetH double mutant could multiply 3-fold within 20 h in cultured macrophages (Table 4). While methionine is available in macrophages, we may conclude from the infection studies that it is not available during other phases of mouse infection. Elimination of purT in the double mutant background significantly reduced the net multiplication rate in macrophages (0.73 ± 0.29), suggesting a synergistic effect of the mutations at this level.

DISCUSSION

The main aims of this study were to determine the contributions of the in vivo redundant enzymes PurN and PurT to virulence in S. Typhimurium and, when we realized that they were not redundant in vivo during infection of mice, to elucidate possible reasons for this. In the broader perspective, the study illustrates that one cannot safely assume that in vitro redundancy between enzymes is followed by a similar in vivo redundancy during infection of mice. Another important observation, based on the in vitro growth experiments, is that even though E. coli and S. Typhimurium share the basic architecture of the metabolic systems we investigated, the growth phenotypes associated with knocking out a gene cannot always be assumed to be the same. This notion was recently highlighted in another publication, dealing with thiamine biosynthesis in E. coli (29).

The concept of redundancy has been understood mainly from studies of E. coli K-12. When this strain grows on glucose as the sole carbon source, more than 80 of the 227 metabolic enzymes are nonessential (30). Redundancy may be a trade-off between efficiency and robustness, and organisms with a broad niche repertoire show the highest degree of redundancy. Therefore, redundancy has been interpreted as a mechanism that supports niche adaptation (31). Others, however, consider redundancy to be a way to withstand detrimental mutations (32). Our observations with PurN and PurT, and also with MetE and MetH, suggest that some pairs of redundant enzymes may be artifacts of studying bacteria in test tubes and that the two enzymes are maintained because they are both essential, possibly at different steps in the normal life cycle (infection steps in the case of pathogenic bacteria). This corresponds mostly to the niche adaptation theory, since the likely explanation for both enzymes being essential is that Salmonella goes through a series of different environments (niches) in the infection process. However, in the niche adaptation theory, the need for both enzymes is not absolute.

The total avirulence of both PurN and PurT mutants could be complemented by addition of the respective enzymes encoded in trans from a plasmid. This proved that the attenuation was related to the lack of the specific enzymes. It is well known that purine biosynthesis is required for intracellular multiplication, and purine biosynthesis mutants have been employed as live vaccines against S. Typhi (33), S. Typhimurium (34), Mycobacterium tuberculosis (35), Brucella melitensis (36), and Francisella tularensis (37). In light of this, it was not surprising that the ΔpurN ΔpurT double mutant was attenuated and unable to multiply inside macrophages. However, previous studies on putative target genes for attenuated live vaccines ignored the 3rd step in the biosynthesis of purines, catalyzed by PurN and PurT, due to the recognized redundancy between the enzymes in vitro. Based on our results, deletion mutants of either PurN or PurT are likely to also be good vaccine candidates against Salmonella.

Studies of multiplication ability in cultured macrophages are a sensitive method to identify virulence attenuation in Salmonella, and Leung and Finlay (38) showed that mutants that could not multiply in cultured macrophages were also avirulent in mice. The opposite, however, is not necessarily the case. The avirulent PurT mutant was not critically affected in propagation in the intracellular environment of J774A.1 macrophages. This suggested that the interaction with this cell type was not the limiting point for the PurT mutant. In contrast to the ΔpurT mutant, the ΔpurN mutant grew poorly or was killed more quickly than the wild type inside macrophages (these two things cannot be separated by the assay we used), suggesting that macrophage survival/growth could be one of the critical points in progression in the host for this mutant. Recent studies showed that one should be careful not to overinterpret results from experiments with cultured macrophages, and especially not to draw conclusions on mechanisms of virulence from such studies (39). Also, it should be noted that J774 cells have a different genetic background (BALB/c) from that of the mouse strain used, and even though they are both slr^−/− (N-ramp⁻), this may make it difficult to compare the in vivo and in vitro situations directly. Our results with macrophages should thus be interpreted with caution.

Studies using a mutant in the essential S. Typhimurium purG gene (also termed purL) showed that this purine auxotroph strain failed to repair DNA damage caused by reactive oxygen species in the phagosome environment of the macrophage, which can explain the inability of the mutant to multiply inside macrophages (40). Our results suggest that PurN has sufficient activity within macrophages to ensure that DNA damage can be repaired in the absence of PurT, while PurT cannot ensure wild-type propagation in macrophages in the absence of PurN. The reduction in net growth of the PurN mutant of approximately 50% corresponds to previous observations on the net growth rate of a ΔpurH mutant in mouse spleens (26). The results indicated that intracellular propagation was the rate-limiting step causing attenuation of the purN mutant, and also that purines cannot be supplied in sufficient amounts from external sources during macrophage infection. A recent study showed that the ΔpurH mutant grows as two distinct populations with respect to location and growth rate in the spleens of mice (26), probably reflecting the fact that growth happens in two compartments with two different demands for purine synthesis, and it would be interesting to investigate whether the ΔpurN mutant also grows as two separate populations inside macrophages.

Purine synthesis has been found to be marginally (down)regulated when S. Typhimurium grows inside cultured macrophages, and PurN and PurT are regulated to the same degree, although only PurN was significantly different from the control by the statistical analysis used in the study (41). The reference condition was growth of opsonized bacteria in cell culture media, and the observation is therefore difficult to compare to the present study, where bacteria were grown in LB medium prior to macrophage challenge; however, it indicates that S. Typhimurium requires as much or slightly more de novo purine biosynthesis to grow in cell culture medium than it requires for growth inside macrophages.

With current techniques, it is not possible to determine the exact amounts of important intermediates in purine biosynthesis, such as THF species, in bacteria during infection. In order to understand why both the PurN and PurT enzymes were essential and whether this was related to an imbalance in the consumption and production of THF species, we chose a mutant-based approach where relevant genes were deleted in different combinations. Our studies with ΔpurU and ΔpurT strains strongly suggested that a main reason for attenuation of the purT mutant was an accumulation of formate or wasteful use of fTHF by PurU, since the ΔpurT ΔpurU double mutant was fully virulent. For this to make sense, the metabolism should be aerobic, since this is the condition under which formate production by PurU is a dead-end product in the absence of PurT (16). In this sense, our observation supports a recent study showing that the environment perceived by Salmonella residing in cells of the monocyte line in the spleen is aerobic, because Salmonella resides exclusively in the red pulp, in close proximity to erythrocytes (26). It still needs to be determined whether formate accumulation, wasteful fTHF use, or both are the important factors and to detect the exact point in infection when this may be the case. The fact that the mutant without PurT could propagate in cultured macrophages suggested that formate accumulation and/or wasteful fTHF consumption is not a problem inside such cells.

The current study also made important observations with regard to the role of C₁ metabolism in virulence. Based on in vitro experiments, the glycine cleavage enzyme, GCV, has been estimated to contribute less to formation of C₁ units in E. coli than the hydroxyl-methyltransferase enzyme encoded by glyA (14). The growth experiments in the current study showed that S. Typhimurium had the same balance between the two enzymes in vitro, since the ΔgcvT mutant was not growth attenuated, while the ΔglyA mutant did not grow. In vivo, however, the situation was totally opposite. The ΔglyA strain was only partly attenuated, while the ΔgcvT mutant was highly attenuated, and its colonies were rarely isolated from any infected animals. Interestingly, concurrent mutation of serA did not change this, which showed that in contrast to what was previously expected (8), exogenous glycine, not serine, must be available to Salmonella during infection. Otherwise, glycine and C₁ units could have been formed from serine by GlyA, and then the ΔgcvT mutation would have had no effect (Fig. 1B). When the ΔpurN or ΔpurT mutation was combined with mutation of serA or glyA, the in vitro phenotypes indicated that the purine synthesis and C₁ metabolism systems did not interfere significantly with each other. Double mutants caused the same phenotype as the single mutants with the most influence on the performance of S. Typhimurium. Unexpectedly, we discovered that ΔglyA mutants formed elongated cells in vitro. Filament formation in Salmonella is well known as a response to osmotic and cold stress (42, 43), but this phenotype of glyA mutation, to the best of our knowledge, has not previously been described for this genus or any of its close relatives. The phenotype could be reversed by genetic complementation in trans, ruling out the possibility that it was caused by secondary mutations in genes of relevance for Z-ring formation.

In the process of detailing the link between the purine and methionine metabolism pathways, we discovered that the MetE and MetH enzymes, which are considered redundant (6), also did not show redundancy during infection. We confirmed that methionine is not taken up from the host during mouse infection, since the ΔmetE ΔmetH double mutant was avirulent. It was previously shown that metC, whose product mobilizes sulfur for methionine synthesis, is essential in S. Typhimurium (44), and S. Typhimurium is only one among several pathogens for which methionine biosynthesis appears to be essential during growth in the animal host (45, 46). In vitro, the double mutant did not grow either. This phenotype could be complemented by addition of methionine to the medium, showing that the methionine uptake system can compensate fully for lack of the biosynthesis system during growth in vitro. Provision of metE in trans restored growth to the level of the ΔmetH mutant. This demonstrated that the metE mutation can be complemented in trans and that MetE and MetH are not fully redundant; the MetH enzyme, which is associated with anaerobic growth (26), must play a role during this growth, even though the cultures were shaken. Further studies are needed to fully understand the reasons for this observation. As with the case for C₁ metabolism, we did not find any indication that purine mutation significantly affected the phenotypes obtained after mutation of methionine synthesis genes.

Funding Statement

This work was funded by Danish Council for Independent Research - Technology and Production (09-064052) through a grant to John E. Olsen.