Salmonella enterica serovar Typhimurium strain 14028s (S. Typhimurium 14028s) with spontaneous rifampicin resistance (50 μg/mL) used in this study was constructed by Fornefeld et al. (2017). S. Typhimurium 14028s single gene mutation strains (ΔaceA, ΔaceB, and ΔaceE) with kanamycin resistance (50 μg/mL) were provided by Michael Hensel (Universität Osnabrück, Germany). Since aceB and aceE are the first genes in corresponding operons, the complete sequences of aceBAK and aceEF were cloned and introduced to ΔaceB and ΔaceE, respectively. Complemented strains (ΔaceB::aceBAK and ΔaceE::aceEF) are both kanamycin (50 μg/mL) and carbenicillin (50 μg/mL) resistant. Briefly, sequences were cloned to the backbone of pWSK29. The vectors (PaceB aceBAK and PaceE aceEF) were thereafter transformed into Escherichia coli strain NEB5α (New England Biolabs, USA), respectively. Single colonies picked from medium-added carbenicillin (50 μg/mL) were verified by sequencing. Afterward, vectors of PaceB aceBAK and PaceE aceEF extracted from E.coli NEB5α were electroporated using the Ec2 program in the MicroPulser Electroporator (Bio-Rad, Germany) into ΔaceB and ΔaceE competent cells, respectively. After 1 h of recovery in SOC medium [SOB (Carl Roth GmbH + Co. KG, Germany) + 20 mM glucose (Carl Roth GmbH + Co. KG, Germany)] at 37°C with a shaking speed of 150 rpm, the bacterial suspension was spread on LB agar containing kanamycin (50 μg/mL) and carbenicillin (50 μg/mL) and incubated at 37°C overnight. Single colonies were picked, streaked onto novel plates, and verified by PCR. Primers for verification are listed in Supplementary Table S1.

Salmonella was cultured in LB medium (Carl Roth GmbH + Co. KG, Germany) with the required antibiotics for inoculums. Minimal medium (MM) was prepared from 12.3 mM glucose, 1 × M9 salts (5 × , Sigma-Aldrich Chemie GmbH, Germany), 2 mM MgSO₄ (Carl Roth GmbH + Co. KG, Germany), and sterile deionized water (Jechalke et al., 2019). Diluvial sand (DS) soil suspension is composed of DS soil (Rühlmann and Ruppel, 2005) and 10 mM MgCl₂ in a 1:2 (m/m) proportion (Schierstaedt et al., 2020). Tomato and lettuce leaf-based media (TM and LM) contain 25% (v/v) of leaf homogenates, 20% (v/v) M9 salts, and 55% (v/v) sterile deionized water (Fornefeld et al., 2017; Zarkani et al., 2019). Leaf homogenates were prepared from 45 g of blended leaves in 300 mL of sterile deionized water and afterward filtered through two layers of paper tissues. Tomato and lettuce root exudate-based media (TE and LE) were made by mixing root exudates with 10 mM MgCl₂ (Jechalke et al., 2019; Zarkani et al., 2019). The collection of root exudates was executed as previously published (Witzel et al., 2017). Briefly, plant roots were washed and immersed in sterile deionized water for 1 h, followed by another 4 h of immersion in fresh water. A suspension of every 25 plants was filtered (pore size of 0.22 μm) and concentrated using freeze-drying as one sample. XLD agar (Carl Roth GmbH + Co. KG, Germany) was used for the selective enumeration of Salmonella. Salmonella cultures were grown at 37°C in LB and on XLD and at 28°C in environment-related media.

A dialysis membrane system was used to collect Salmonella samples in vitro for metabolite measurement and RT-qPCR assays (Han et al., 2023). In brief, Salmonella cultures in the stationary phase were centrifuged, washed, and diluted in 10 mM MgCl₂ to 10⁹ CFU/mL. Three milliliters of the dilution were added into a cellulose ester dialysis membrane (MWCO 100,000, ϕ = 16 mm, Spectra/Por Biotech, USA) bag, which was then sealed and immersed in 30 mL of the respective medium in a 50 mL centrifuge tube. The tube was shaken at 28°C for 24 h. In the metabolism experiment, part of the samples was separated for bacterial enumeration, while the rest was pelleted at 4°C in a precooled centrifuge tube. After being rinsed on ice with 10 mM MgCl₂, the pellets were centrifuged again and instantly frozen in liquid nitrogen. Salmonella cultures designated for RNA extraction were immediately vortexed with double the volume of RNAprotect Bacteria Reagent (Qiagen, Germany). After 5 min of incubation, the bacteria were centrifuged at 5,000 × g for 10 min at room temperature. Bacterial pellets for both metabolism and RT-qPCR assays were stored at −80°C until further processing. Four biological replicates were prepared for each treatment.

Metabolites were extracted from wet biomass through the methanol–water–chloroform extraction process using 100 μL of methanol/¹³C-ribitol solution to re-suspend the cells. Thereafter, the sequential addition of 100 μL of H₂O and 150 μL of chloroform was performed with vortexing for 5 min in between. After centrifugation, 150 μL of the polar phase was dried in a vacuum concentrator. For extracellular metabolites, 10 μL of cell-free culture supernatant was mixed with 10 μL of methanol/¹³C-ribitol solution and then dried in a vacuum concentrator. Derivatization and GC-MS measurement (splitless and split 1:10) of metabolites were performed as described in a previous study (Will et al., 2019). Processing of raw data was performed using the MetaboliteDetector (Neumann-Schaal et al., 2015). The peak area data were normalized to the internal standard ³C-ribitol and—for intracellular metabolites—to the determined cell number of the harvested cells. Data are available on FAIRDOMHUB (seek ID: https://fairdomhub.org/projects/348). MetaboAnalyst 5.0 (https://www.metaboanalyst.ca/), a web-based tool, was used for analysis. The principle component analysis (PCA) was performed using the statistical analysis (one factor) module. In the heatmaps, the cluster method is Ward.D and the distance is Euclidean distance. The relative abundance of metabolites in S. Typhimurium 14028s and ΔaceB grown in TM was compared to that in MM using log₂ fold change. Four biological replicates were prepared for each treatment.

Salmonella total RNA was extracted using the RNeasy Mini Kit (QIAGEN, Germany), DNA was eliminated, and cDNA was synthesized using the Maxima H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, USA) according to the manufacturer's instructions. Thereafter, 20 μL of the cDNA generated from 1 μg DNA-digested RNA was diluted to 100 μL for qPCR. The qPCR reaction system was 20 μL, consisting of 10 μL of LUNA Master Mix (New England Biolabs, USA), 4 μL of nuclease-free water, 0.5 μL of forward or reverse primers (10 μM), and 5 μL of diluted cDNA. The Bio-Rad CFX Connect cycler (Bio-Rad, Germany) was programmed with an initial heating of 95°C for 5 min, followed by 39 cycles of 95°C for 15 s and 60°C for 30 s (+plate read). Melting curves were measured for quality control. The rfaH gene (Cardenal-Munoz and Ramos-Morales, 2013) was used as a calibrator for raw data normalization. The fold change was calculated using the 2^−ΔΔCt method (Livak and Schmittgen, 2001) compared to gene expression in MM. Unpaired Student's t-test or one-way ANOVA with the Tukey HSD test was used to assess differences. The primers used are listed in Supplementary Table S1. Two technical replicates of the qPCR reaction were set for each sample, and the error bars represented the standard deviation among the four biological replicates.

Three weeks after sowing, tomato (Solanum lycopersicum cultivar Moneymaker) seeds germinated in Substrate 1 (Klasmann-Deilmann GmbH, Germany) were transplanted to 9 cm diameter pots in the greenhouse with 18 h of daylight at 20°C. To avoid contamination of the leaves, plants were irrigated and fertilized from the bottom of the trays as needed. Plants were ready for treatment 3 weeks after transplantation.

To evaluate the persistence of S. Typhimurium 14028s, single gene mutant strains, and complemented strains in tomato leaves, 6-week-old plants cultivated as indicated above were gently infiltrated with bacterial suspension (10⁷ CFU/mL). The most marginal two leaflets of the plant's second and third leaves were chosen. When the inoculated leaflets recovered from their water-stained appearance 3 h post-infiltration, as well as 7 and 14 days post-inoculation (dpi), four 5-mm diameter leaf disks were sampled and pooled in a 2 mL centrifuge tube, homogenized with 200 μL of 10 mM MgCl₂ using a motor handpiece (Xenox, Germany), and then replenished to 1 mL. Two technical replicates of homogenized leaf slurry were serially diluted in a 96-well plate. Ten μL of the dilution was dropped on XLD agar supplemented with appropriate antibiotics in duplicate. After overnight incubation at 37°C, Salmonella colonies with easily identifiable black coloration were enumerated. The persistence was evaluated by calculating the log₁₀ CFU in each leaf disk. One-way ANOVA with the Tukey HSD test was used to verify differences. The error bars represented the standard deviation among the four biological replicates.

Salmonella enterica serovar Typhimurium strain 14028s (S. Typhimurium 14028s) was reported to persist in diluvial sand (DS) soil for several weeks (Schierstaedt et al., 2020), as well as in tomato and lettuce leaves (Jechalke et al., 2019; Zarkani et al., 2019). Carbon utilization is critical for microbial adaptation to agricultural environments. Salmonella's central carbon metabolism is based on glycolysis, pyruvate oxidation, and the tricarboxylic acid (TCA) cycle. To determine whether S. Typhimurium 14028s used a carbon metabolism adaptation strategy to persist in agricultural environments, the abundance of the central carbon metabolism intermediate products was evaluated both in the mimicking media [DS soil suspension, tomato leaf-based media (TM), and lettuce leaf-based media (LM)] to imply the availability for bacteria and in the correspondingly incubated S. Typhimurium 14028s cells.

As references, 10 mM MgCl₂ and M9 salt-based minimal medium (MM) were used. MgCl₂ solution, which was used to pre-suspend the bacteria before inoculation and to wash pellets during sampling, displayed a clean background for all metabolites discussed below (Figure 1A). The MM, a minimal environment provided for S. Typhimurium 14028s basal carbon metabolism, did not contain any of the evaluated compounds, except glucose, which was added as the only carbon source and was overloaded in the measurement. Except for pyruvate, DS soil, the sandy soil collected from fields (Rühlmann and Ruppel, 2005), was poor in the evaluated central carbon metabolism intermediate compounds. Glycerol, on the other hand, was of extraordinarily high abundance (overloaded) and therefore a possible carbon source. By contrast, TM and LM were rich in nutrients. Excess glucose and fructose were detected (overloaded) in both media, as well as several di/trisaccharides in LM. Furthermore, pyruvate, a glycolysis end product, and the TCA cycle intermediates, including 2-oxoglutarate, succinate, fumarate, and malate, were also present in TM and LM (Figure 1A).

Differences in the composition of carbon-related metabolites in different environments could influence bacterial metabolism. Therefore, in the next step, related compounds were evaluated in S. Typhimurium 14028s. No difference in the cell multiplication of S. Typhimurium 14028s grown in TM, LM, or MM was observed at 24 hpi (Supplementary Figure S1). However, S. Typhimurium 14028s inoculated into DS soil suspension presented no change in the cell number at the harvest time point compared to the inoculum (Supplementary Figure S1). A clear dispersion was observed between central carbon metabolism intermediates present in bacterial cells under different culturing conditions. In a principal component analysis (PCA), the first PC1 (65.1%) and second PC2 (20%) dimensions explained 85.1% of the data variance and separated the data into treatment-related populations (Supplementary Figure S2). All four biological replicates were included in 95% confidence intervals. Detailed analysis revealed that, in S. Typhimurium 14028s introduced to MM, glycolysis intermediates including glucose-6-phosphate, fructose-6-phosphate, 3-phoshoglycerate, 2-phoshoglycerate, phosphoenolpyruvate, and pyruvate were present, although fructose-1,6-diphosphate was not detectable (Figure 1B). The situations in bacterial cells inoculated with TM and LM were slightly different. Instead of fructose-1,6-diphosphate, fructose-6-phosphate was not detectable in Salmonella cells in TM or LM. In DS soil, where glycerol was available rather than glucose, neither fructose-6-phospahte nor fructose-1,6-diphopsphate was detectable in S. Typhimurium 14028s cells. Furthermore, no 2-phosphoglycerate was detected. Dihydroxyacetone, the direct downstream product of glycerol catabolism, was present and abundant instead, indicating glycerol consumption activity in cells introduced to DS soil. TCA metabolites such as citrate, succinate, fumarate, and malate were detected in all samples. Interestingly, 2-oxoglutarate was not detectable in S. Typhimurium 14028s grown in TM (Figure 1B).

The results presented above strongly suggest that the availability of carbon compounds varied in diverse agricultural environment-related media, as well as the metabolite content of S. Typhimurium 14028s grown in those conditions (Figure 1). Predicatably, TM and LM media clustered close together, whereas the composition of DS soil suspension was closer to 10 mM MgCl₂ solution and MM (Figure 1A). Interestingly, the metabolite pattern of S. Typhimurium 14028s exhibited a highly comparable clustering pattern (Figure 1B). It was, therefore, reasonable to speculate that S. Typhimurium 14028s metabolism is related to the ecological niche, for example, root or leaf.

Since the metabolism appeared dependent on a root- or leaf-related ecological niche, we speculated which Salmonella genes potentially contributed to the adaptation. The transcriptome of S. Typhimurium 14028s incubated in DS soil suspension (Schierstaedt et al., 2020), TM (Zarkani et al., 2019), and LM (Jechalke et al., 2019) for 24 h was indicated in previous studies. Additionally, root exudates are another important component of the root environment. The transcriptome of S. Typhimurium 14028s cells incubated in tomato or lettuce root exudate-based media (TE and LE, respectively) provided supplementary information on their transcriptional adjustment to root-related niches (Jechalke et al., 2019; Zarkani et al., 2019).

To limit the number of potential candidate genes, four steps were undertaken. First, DS soil, TE, and LE were defined as root-related environments, whereas TM and LM were classified as leaf-related environments. Second, commonly upregulated genes in S. Typhimurium 14028s identified from root-related environments and leaf-related environments were determined: 47 genes in root-related environments, 29 genes in leaf-related environments, and 48 genes in environments related to both (Figure 2A). Third, enriched biological processes of the defined environments were identified (Figure 2B), such as, translation and carboxylic acid metabolic process in root-related environments, as well as response to heat, arginine biosynthetic process, glycolytic process, and tRNA aminoacylation for protein translation in leaf-related environments. Fourth, genes obtained in step 2, which could also be mapped to the enriched biological processes, were selected as potential candidates. As a result, 3-oxoacyl-[acyl-carrier-protein] reductase encoding gene fabG, functional in fatty acid biosynthesis, and phosphoenolpyruvate synthase encoding gene pps, were chosen as root-related gene candidates. Leaf-related gene candidates included chaperone protein DnaJ encoding gene dnaJ, ATP-binding subunit of serine protease encoding gene clpA, pyruvate dehydrogenase subunit E1 encoding gene aceE, acetyltransferase component of pyruvate dehydrogenase complex encoding gene aceF, and phosphofructokinase encoding gene pfkB. As for gene candidates related to both ecological niches, we identified threonine-tRNA ligase encoding gene thrS, enolase encoding gene eno, fructose-bisphosphate aldolase encoding gene fba, and glyceraldehyde-3-phosphate dehydrogenase-encoding gene gapA. ClpA and DnaJ are chaperone proteins that participate in either ATP-binding or a range of other cellular processes. Threonine-tRNA ligase ThrS is essential for RNA modification. PfkB, Fba, GapA, and Eno are responsible for glycolysis, while AceE and AceF are vital subunits in the pyruvate oxidation process (Table 1).

To further validate the expression of the putative niche-related genes, an RT-qPCR approach was used. After 24 h incubation of S. Typhimurium 14028s in DS soil suspension, TE, LE, TM, and LM, gene expression levels in Salmonella cells grown in each medium were compared to those in bacterial cells incubated in MM. Of the five candidates chosen as leaf-related, aceE and aceF showed expression patterns consistent with RNA-Seq results and an upregulation in TM and LM media. The expression of eno demonstrated consistent upregulation in all environments (Figure 3). Given the consistency in results and the importance of Salmonella contamination in leafy greens, in the next steps, we focused the study on leaf-related environments.

Compared to the soil environment, Salmonella contamination of plant leaves poses a direct public health risk, which increases with the popularity of raw-consumed salad (EFSA, 2022a). Although tomato fruit is consumed, earlier research has shown that Salmonella can translocate from contaminated leaves to uncontaminated parts, including fruits (Gu et al., 2011; Zarkani et al., 2019). Consequently, tomato leaves were used for further studies.

On the one hand, the upregulation of aceE and aceF, confirmed by RNA-Seq and RT-qPCR, suggested an important role in the pyruvate dehydrogenase complex for S. Typhimurium 14028s growth in TM. We expected, therefore, that a disruption of the complex function by the mutation of aceE, which is located upstream of aceF, would affect Salmonella's persistence ability. On the other hand, we speculated whether the glyoxylate pathway, which has been linked to Salmonella–tomato root interactions (Zarkani et al., 2019), would be involved in tomato leaf persistence as well. In the glyoxylate pathway, malate synthase A (encoded by aceB) catabolizes glyoxylate into malate.

To verify those assumptions, S. Typhimurium 14028s, aceE and aceB mutant strains (ΔaceE and ΔaceB), and genetically complemented strains (ΔaceE::aceEF and ΔaceB::aceBAK) were used. Gene expression of aceE or aceB in the complemented strains was similar to the level in S. Typhimurium 14028s cultured in TM, while no signal was detected in the corresponding mutant strains (Supplementary Figure S3). For further verification in vivo, a bacterial suspension was infiltrated into tomato leaves. Compared to S. Typhimurium 14028s, ΔaceE showed deficient persistence both 7 and 14 days post-inoculation (dpi), while complementation recovered that deficiency (Figure 4A). On the contrary, ΔaceB displayed enhanced persistence abilities. At 7 dpi, ΔaceB was present in a higher CFU number than S. Typhimurium 14028s, while no difference in persistence was observed between ΔaceB::aceBAK and S. Typhimurium 14028s (Figure 4B). This phenomenon may indicate probable contradictory adaptation strategies employed by both mutants during tomato leaf persistence.

Even though the expression of aceB was upregulated in S. Typhimurium 14028s grown in TM compared to MM (Supplementary Figure S4A), ΔaceB exhibited enhanced persistence ability in tomato leaves. We speculated that the mutation of aceB induced a compensatory or redundant process or metabolite flux, which assisted in ΔaceB persistence in tomato leaves. Since aceB plays a major role in carbon metabolism, the abundances of main metabolites in glycolysis and the TCA cycle of S. Typhimurium 14028s and ΔaceB incubated in TM were evaluated by GC-MS. In comparison to the control (S. Typhimurium 14028s grown in MM), abundances of multiple metabolites were altered, albeit to variable degrees (Figure 5A), indicating a metabolism adjustment of both strains to the TM environment. Notably, fumarate was the sole metabolite that reversed the changing direction, from a decreased abundance in S. Typhimurium 14028s to an increased abundance in ΔaceB. Moreover, the fumarate abundance in the mutant in pyruvate dehydrogenase subunit E1 (ΔaceE), which had lower persistence ability in tomato leaves, was significantly lower than that in S. Typhimurium 14028s when grown in TM (Supplementary Figure S5). We hypothesized, therefore, that fumarate abundance in Salmonella may affect, to some extent, its persistence in leaves.

Fumarate can be accumulated through upregulated biosynthesis or economical catabolism. In the catabolic pathway, fumarate can be catabolized by fumarate reductase (frdDCBA) to succinate. To investigate the possible origin of the fumarate accumulated in ΔaceB, we evaluated the expression of frdD and frdA in ΔaceB and S. Typhimurium 14028s using RT-qPCR. Compared to the control (S. Typhimurium 14028s grown in MM), frdD and frdA expressions in S. Typhimurium 14028s were upregulated in TM (Figure 5B). However, expressions of frdD and frdA in ΔaceB decreased in TM compared to S. Typhimurium 14028s (Figure 5B). These results implied a possible accumulation of fumarate in ΔaceB grown in TM because of a diminished catabolism ratio.