Salmonella Persistence in Tomatoes
ABSTRACT Human enteric pathogens , such as Salmonella spp . 
and verotoxigenic Escherichia coli , are increasingly recognized as causes of gastroenteritis outbreaks associated with the consumption of fruits and vegetables . 
Persistence in plants represents an important part of the life cycle of these pathogens . 
The identification of the full complement of Salmonella genes involved in the colonization of the model plant ( tomato ) was carried out using transposon insertion sequencing analysis . 
With this approach , 230,000 transposon insertions were screened in tomato pericarps to identify loci with reduction in fitness , followed by validation of the screen results using competition assays of the isogenic mutants against the wild type . 
A comparison with studies in animals revealed a distinct plant-associated set of genes , which only partially overlaps with the genes required to elicit disease in animals . 
De novo bio-synthesis of amino acids was critical to persistence within tomatoes , while amino acid scavenging was prevalent in animal infections . 
Fitness reduction of the Salmonella amino acid synthesis mutants was generally more severe in the tomato rin mutant , which hyperaccumulates certain amino acids , suggesting that these nutrients remain unavailable to Salmonella spp . 
within plants . 
Salmonella lipopolysaccharide ( LPS ) was required for persistence in both animals and plants , exemplifying some shared path-ogenesis-related mechanisms in animal and plant hosts . 
Similarly to phytopathogens , Salmonella spp . 
required biosynthesis of amino acids , LPS , and nucleotides to colonize tomatoes . 
Overall , however , it appears that while Salmonella shares some strategies with phytopathogens and taps into its animal virulence-related functions , colonization of to-matoes represents a distinct strategy , highlighting this pathogen 's flexible metabolism . 
IMPORTANCE Outbreaks of gastroenteritis caused by human pathogens have been increasingly associated with foods of plant origin , with tomatoes being one of the common culprits . 
Recent studies also suggest that these human pathogens can use plants as alternate hosts as a part of their life cycle . 
While dual ( animal/plant ) life-styles of other members of the Enterobacteriaceae family are well known , the strategies with which Salmonella colonizes plants are only partially understood . 
Therefore , we undertook a high-throughput characterization of the functions required for Salmo-nella persistence within tomatoes . 
The results of this study were compared with what is known about genes required for Salmonella virulence in animals and interactions of plant pathogens with their hosts to determine whether Salmonella repurposes its virulence repertoire inside plants or whether it behaves more as a phytopathogen during plant colonization . 
Even though Salmonella utilized some of its virulence-related genes in tomatoes , plant colonization required a distinct set of functions . 
Approximately 15 % of all cases of human salmonellosis are thought to be associated with the consumption of fruits and vegetables ( 1 ) . 
The ability of nontyphoidal strains of Salmonella enterica to colonize plants , including sprouts , alfalfa , lettuce , melons , and tomatoes , is highlighted by the rise in the number and severity of salmonellosis outbreaks linked to produce . 
The capacity of S. enterica ( and other enteric pathogens ) to readily multiply within plant tissues led to the hypothesis that persistence on plants is a part of the Salmonella life cycle , serving as reservoirs prior to reinfection of the preferred animal hosts ( 2 -- 5 ) . 
Several studies have dissected the molecular basis of plant-Salmonella interactions ( 6 -- 16 ) . 
However , the entire complement of genetic functions required for plant colonization by Salmonella is not yet known . 
Previous studies established that colonization of plants by Salmonella is a complex process , dependent on both host and bacterial factors . 
It involves major changes in the bacterial transcriptome and requires genes involved in amino acid biosynthesis and transport , cellulose production , fimbriae , regulators , and surface structures ( 6 -- 17 ) . 
Some of the same mechanisms were involved in the colonization of both vegetative and reproductive tissues on different plant species . 
Colonization of the tomato pericarp involves differential regulation of at least 50 genes ( 9 ) , some of which were also required for proliferation , including the gene cysB , encoding a cysteine metabolism regulator ( 9 ) . 
Surface structures , like the O-antigen capsule and curli fimbriae , were also required for the colonization of tomatoes , as indicated by reduced fitness of the corresponding Salmonella mutants ( 11 , 17 ) . 
The O-antigen capsule also plays a role in the colonization of alfalfa sprouts by Salmonella spp. , as do curli fimbriae , suggesting that these structures may be broadly required for the colonization and/or persistence within plants ( 7 , 14 ) . 
Iron acquisition was shown to play an important role during Salmonella colonization of vegetative and reproductive tissues . 
The siderophore transporter fepDGC was required for Salmonella proliferation in tomato pericarps ( 16 ) , and siderophore synthesis was necessary for the colonization of lettuce and alfalfa ( 10 ) . 
Biosynthesis of amino acids was also a likely factor required for Salmonella growth in alfalfa root exudates ( 15 ) . 
Plant colonization by Salmonella is affected by host factors ; the genetic background can influence Salmonella colonization of sprouts ( 18 ) , proliferation in the tomato phyllosphere ( 19 , 20 ) , and pericarps ( 21 ) . 
Some of the previous work focused on the roles of Salmonella virulence factors during its interaction with plants . 
The term `` virulence factor '' was sometimes used to broadly define any gene ( e.g. , rpoS ) required for full virulence in animal models . 
While the contribution of the Salmonella genes that generally impact animal virulence was characterized during the attachment and colonization of alfalfa sprouts ( 6 ) , the involvement of the more dedicated virulence-related functions , such as type III secretion systems ( T3SS ) , and their regulators encoded on the Salmonella pathogenicity islands ( SPIs ) remains controversial . 
Strains lacking SPI-1 and SPI-2 T3SSs had reduced proliferation in Arabidopsis thaliana leaves , failed to suppress the plant immune response against bacteria , and were required for the chlorotic appearance of plants ( 22 ) . 
The effector spvC carried by the virulence plasmid suppressed inducible plant defenses when it was directly transformed into Arabidopsis protoplasts ( 22 ) . 
Salmonella strains lacking SPI-1 T3SS induced a stronger response from plants , suggesting that this pathogenicity island represses plant immunity ( 23 ) . 
However , studies linking the T3SSs encoded by SPIs to plant colonization by Salmonella required high bacterial titers , which are unlikely to happen naturally . 
Moreover , Salmonella spp . 
induced plant immune responses at lower levels than specialized plant pathogens , like Pseudomonas syringae ( 24 ) . 
Some studies reported that Salmonella strains lacking SPI-1 , SPI-2 , SPI-3 , SPI-4 , and SPI-5 are not defective for the colonization of tomato and cantaloupe fruits ( 9 , 25 ) . 
Furthermore , there is no direct evidence that Salmonella can translocate effectors into plant cells via SPI-encoded T3SS , calling into question the requirement for the corresponding genes during plant colonization ( unless the T3SS apparatus performs other noncanonical functions during plant colonization , a hypothesis that has not yet been ruled out ) . 
Reverse genetics was previously employed to identify Salmonella virulence genes in animals . 
Mutants were created and individually screened to identify those defective in invasion and intracellular proliferation in cell cultures , and those with impaired virulence in intraperitoneal or oral infection models ( 26 -- 33 ) . 
Recently , the development of high-throughput technologies allowed coverage of the entire genome in these screens ( 34 ) . 
Here , we used a similar approach to test the hypothesis that Salmonella genes required for the colonization of plants are different from those required for virulence in animals , and , furthermore , that the strategies that Salmonella uses to colonize plants are distinct from those used by phytopathogens . 
We used a transposon insertion sequencing approach to identify Salmonella enterica serovar Typhimurium loci required for persistence in tomatoes and subsequently confirmed the phenotypes using competition assays with isogenic mutants . 
We also compared the results obtained in this study with similar screens in animals , as well as with the data obtained using mutant screens of phytopathogens . 
RESULTS
Transposon insertion library screening and sequencing . 
The tested transposon mutant library consisted of about 280,000 independent mutants with Tn5 insertions , of which 230,000 insertion sites were mapped onto the genome . 
The mapped insertions disrupted 88 % of all features of the genome . 
The remaining 12 % of the loci ( 1,263 features ) are likely essential for Salmonella survival and proliferation under the culture conditions used . 
We defined 10,632 features in the Salmonella enterica serovar Typhi-murium ATCC 14028 genome , including coding sequences , intergenic regions , and noncoding RNA . 
Salmonella spp . 
can proliferate within red ripe tomato pericarps , reaching 107 CFU per fruit ( 9 , 19 , 21 , 25 , 35 ) . 
To represent the entire transposon insertion library , the screening needed to be done using an inoculum titer very close to the carrying capacity of tomatoes . 
Consequentially , bacterial growth appeared constrained , potentially imposing a limit on the possible changes in frequency of transposon insertions with differential fitness . 
We tested whether screening the mutant library under this condition would allow us to detect selection , using the rcsA : : kan mutant strain , whose fitness is known to be modestly reduced in tomatoes ( 36 ) . 
Because following the screen , recovered mutants were briefly grown out in LB to increase DNA output , we also confirmed in preliminary experiments with the rcsA : : kan mutant that this outgrowth does not affect the recovery ratio of the mutant or its DNA ( data not shown ) . 
These optimization experiments established the feasibility of a high-inoculum titer to identify Salmonella transposon insertions with differential fitness . 
Moreover , it suggested that even when the population is stagnant ( death and reproduction were present at the same rate ) , the turnover of bacterial cells is sufficient to lead to changes in the abundance of genotypes with differential fitness . 
The mutant library was screened in ripe Campari tomato pericarps using 12 biological replicates and three technical replicates each ; the biological replicates were pooled , resulting in six final samples for library sequencing , with three input samples and three output samples . 
The screen identified 1,245 features with differential fitness ( false-discovery rate [ FDR ] 0.1 ) ; 1,112 of them were under negative selection , and only 132 features were under positive selection ( Fig. 1 and Data Set S1 ) . 
This represents 11.8 % and 1.4 % of all tested features , respectively . 
Among the loci with differential abundance , 886 loci were coding sequences . 
Loci with reduced fitness often appeared adjacent to each other , mostly due to operon organization . 
Salmonella metabolic requirements during its interaction with tomatoes . 
A high number of features identified in this screen were involved in metabolic pathways . 
To investigate the metabolic pathways required by Salmonella spp . 
during their interaction with the plant , we projected the KEGG Orthology terms for each locus against the Salmonella metabolic map using KEGG Mapper . 
Overall , the screen revealed that Salmonella proliferation within tomatoes is a complex process , requiring the combination of several catabolic and anabolic pathways , including biosynthesis of amino acids , nucleotides , and lipids , as well as glycolysis ( Fig. 2 ) . 
The results indicate that carbohydrates that are metabolized to pyruvate via glycolysis are a major carbon source for Salmonella within tomatoes . 
In Salmonella spp . 
colonizing the tomato , pyruvate is likely fermented to acetate , as indicated by the reduced fitness of mutants in pta and ackA and the absence of phenotypes for the mutants of the genes involved in the tricarboxylic acid ( TCA ) cycle . 
The TCA cycle oxidizes acetyl-coenzyme A ( acetyl-CoA ) originating from pyruvate under aerobic conditions . 
The use of fermentation instead of the TCA cycle is consistent with the fact that the environment inside tomato fruits is microaerophilic or anaerobic . 
The Gene Ontology enrichment analysis indicated that the methionine , arginine , and branched-chain amino acid synthesis pathways were overrepresented among the Salmonella functions required for tomato colonization . 
This is an indication that these nitrogen compounds are not readily available for the bacterium within tomatoes and that Salmonella has to synthesize them de novo . 
Since the data indicate that carbohydrates may provide the major carbon source for Salmonella within tomatoes , it is possible that these are providing the carbon backbone for bacterial amino acid and nucleotide synthesis in this environment . 
Four genes related to nitrogen uptake were identified in the screen : lysP , encoding the lysine-specific permease ; STM14_1095 , encoding a putative amino acid transporter ; potB , encoding a high-affinity spermidine / putrescine ABC transporter ; and STM14_1979 , encoding a putative ABC polar amino acid transporter . 
None of these genes are involved in the transport of the most abundant nitrogen compounds available in tomatoes ( glutamine , gamma-amino butyric acid [ GABA ] , and inorganic nitrogen ) . 
Salmonella spp . 
may be exploiting several nitrogen sources and then synthesize the complement of the needed amino acids . 
Fatty acid biosynthesis and degradation were required by Salmonella spp . 
within tomatoes , perhaps because functional cellular membrane synthesis is necessary for bacterial proliferation and maintenance . 
Genes in fatty acid metabolism that contrib-uted to Salmonella fitness included the 3-oxoacyl - [ acyl-carrier-protein ] synthase gene fabF , and the acyl-CoA dehydrogenase gene fadE . 
Besides these two well-known genes , we also identified STM14_1005 , another putative acyl-CoA dehydrogenase . 
Of interest is that fadH ( 2,4-dienoyl-CoA reductase ) was among the genes differentially regulated in tomatoes , and its expression was shown to be dependent on the availability of linoleic acid , which in turn differentially accumulates in tomatoes as they mature ( 9 ) . 
Salmonella metabolic requirements for proliferation within tomatoes differ from the requirements for systemic infection of mice . 
We explored whether Salmo-nella metabolic requirements for the systemic infection of mice were comparable with the metabolic requirements for fitness within tomatoes . 
We retrieved data from previous work using transposon insertion libraries in Salmonella enterica serovar Typhimu-rium ( 37 , 38 ) that had been applied in a mouse model for systemic infection and compared them with our results from the Salmonella-tomato interaction . 
A total of 327 KEGG Orthology ( KO ) identifiers were mapped for the genes required for persistence in tomatoes , and 505 KO identifiers required for systemic infection of mice were mapped . 
Only 125 KO identifiers were shared by the two conditions , showing that the Salmonella metabolic requirements under these conditions are different ( Fig. 2A ) . 
While amino acid biosynthesis was required for Salmonella spp . 
in the interaction with tomatoes , scavenging of amino acids was required in the systemic mouse infection . 
Transposon disruptions in just a few pathways ( e.g. , the urea cycle ) impaired Salmonella fitness during systemic infection ( 37 , 38 ) . 
Besides being involved in the synthesis of arginine , the urea cycle is also involved in the incorporation of excess nitrogen . 
Carbohydrates were likely a major carbon source for Salmonella spp . 
during systemic animal infection and during tomato colonization . 
However , Salmonella spp . 
may use the less-energy-efficient pathway of acetate fermentation in tomatoes . 
During systemic infection , Salmonella oxidizes carbohydrates using the TCA cycle , as indicated by phenotypes of the mutants that represent this pathway . 
Phenotypes observed for the genes encoding 2-ketoglutarate dehydrogenase and succinate dehydrogenase , which are only fully expressed under aerobic conditions , support the role of oxidation in the bacterial metabolism in animals . 
Nucleotide biosynthesis is required by Salmonella spp . 
in murine systemic infection and in proliferation within tomatoes , as evidenced by the reduction in fitness when mutations occurred in the genes involved in the biosynthesis of pyrimidines and purines in either model of infection . 
While most Salmonella metabolic requirements diverge for colonization of plant and animal hosts , nucleotide biosynthesis is a point of convergence . 
Enrichment analysis also showed that genes related to pathogenesis ( GO : 0009405 ) contribute to fitness during systemic infection and proliferation within tomatoes ( Fig. 2B and C ) . 
This GO term is heterogeneous and includes regulators , effectors , and surface proteins . 
The global regulator phoPQ , for example , belongs to this group of genes and displayed a phenotype in both tomatoes and mice . 
The phoPQ regulator activates the expression of other pathogenesis-associated genes that code for effectors and have a role in adaptation to low Mg2 and resistance to antibacterial peptides . 
Several effectors and the mtgC virulence factor required for survival in low Mg2 had phenotypes under both conditions , indicating that some Salmonella adaptations for virulence in animals are also at play during colonization of tomato fruits . 
Few metabolic parallels between fitness of Salmonella in tomatoes and behavior of phytopathogens . 
After finding that Salmonella metabolic requirements for proliferation within tomatoes differed from the requirements for the systemic infection of mice , we explored whether these requirements are similar to the metabolic requirements of bacterial phytopathogens during plant infection . 
To test the hypothesis that Salmonella behavior in tomatoes is similar to the behavior of plant pathogens , we extracted the information identified in mutant screens of Pectobacterium carotovorum in Chinese cabbage ( 39 ) , Ralstonia solanacearum in Arabidopsis thaliana and tomatoes ( 40 ) , Xanthomonas citri in grapefruit ( 41 ) , Xanthomonas oryzae pv . 
oryzicola in rice and tobacco ( 42 ) , Pseudomonas tolaasii in A. thaliana ( 43 ) , Pseudomonas syringae pv . 
macu-licola in A. thaliana , and Xanthomonas campestris pv . 
campestris in cabbage ( 44 ) . 
We adopted the strategy of grouping the results of different screens to increase the coverage of KO identifiers , since most of these studies did not have the same depth used in transposon sequencing methods , and to obtain a broader comparison with the different lifestyles that phytopathogens may assume . 
There appears to be a limited overlap of metabolic functions between Salmonella requirements for colonization of tomatoes and those of phytopathogens ( Fig. 3 ) , indicating that the mechanisms employed by Salmonella to colonize tomatoes require a distinct combination of functions . 
Lipopolysaccharide ( LPS ) and nucleotide biosynthesis were among the functions shared by phytopathogens and Salmonella in tomatoes and were also involved in systemic infection of mice , suggesting that there is a small set of genes required for pathogenesis , regardless of the host . 
These functions are likely involved in resistance to host defense and metabolic limitations that bacteria experience within eukaryotes . 
Amino acid biosynthesis was a requirement shared by Salmonella during colonization of tomatoes and by phytopathogens but absent from Salmonella during systemic infection of mice . 
This divergence between colonization of animal and plant hosts could be associated with the nutrient availability . 
Fatty acid biosynthesis and degradation were required for Salmonella infection of various hosts , but this requirement was absent from phytopathogens . 
Salmonella requires biosynthesis of arginine , glutamine , glutamate , branched amino acids , methionine , tryptophan , and threonine to colonize tomatoes . 
Nitrogen metabolism , including amino acid biosynthesis and the global nitrogen regulator glnG , contributed to Salmonella fitness during its colonization of tomatoes . 
We explored this role further using competition assays with isogenic mutants . 
A lower bacterial inoculum was used in these assays to more realistically approximate natural interactions . 
We identified the amino acid biosynthetic pathways required by the bacterium for proliferation within tomatoes . 
Transposon insertion sequencing revealed the synthesis of tryptophan , arginine , branched amino acids ( leucine , isoleucine , and valine ) , glutamine , glutamate , threonine , methionine , and proline to be required for complete bacterial fitness ( Fig. 4A ) . 
Mutants were created in the genes responsible for the key steps of each biosynthetic pathway , and their auxotrophic phenotype was confirmed in minimal medium . 
However , because there are at least two enzymes that can catalyze the last step in the formation of methionine and glutamate , metA and gltB were not auxotrophs . 
The competition assays confirmed the transposon insertion sequencing results for all but one of these 10 metabolic pathways ( Fig. 4B ) . 
Only the auxotrophic mutant for proline ( proA : : kan ) deviated from the transposon insertion sequencing results and did not appear to have a reduced fitness compared to the wild type . 
A deletion of the global nitrogen regulator glnG also led to a reduction in fitness of Salmonella in tomatoes , consistent with its function in bacterial responses to nitrogen assimilation . 
To verify that the observed reduction in fitness caused by the disruption in trpC , argA , ilvD , glnA , serA , thrC , gltB , metA , and glnG was not due to a reduction in overall fitness , we compared the growth kinetics of these isogenic mutants in LB . 
We observed that the growth kinetics of these mutants did not differ from the wild-type strain ; consequently , the reduced fitness observed in tomatoes is not caused by a reduction in overall fitness ( Fig . 
S1 ) . 
We tested whether trpC , argA , ilvD , glnA , serA , thrC , gltB , and metA are expressed within tomato fruits using resolvase-in vivo expression technology ( RIVET ) reporters that allow quantification of gene expression of the targets in vivo . 
One day after inoculation , the RIVET resolution for the reporters of trpC , argA , ilvD , glnA , serA , thrC , gltB , and metA was between 80 and 100 % , and 3 and 7 days after inoculation , the resolution reached almost 100 % for all reporters ( Fig. 5 ) . 
When tested in LB , the reporters were not resolved , while in M9 medium with low concentration of amino acids , the resolution reached high levels ( Fig. 5 ) . 
The results of the RIVET experiments are consistent with the fitness phenotypes of the corresponding mutants : amino acid biosynthesis genes are strongly expressed in tomatoes and are required for proliferation within the fruit . 
Tomato genotype affects Salmonella fitness . 
Tomatoes carrying a ripening inhib-itor ( rin ) mutation do not present the characteristic ripening phenotype , they do not accumulate lycopene , and their fruits do not soften . 
Fruits of rin mutant tomatoes accumulate many amino acids in different ( generally larger ) amounts during their development compared to the wild type ( 45 ) . 
Because amino acids accumulate at higher levels within rin mutant tomatoes , we tested the hypothesis that the phenotypes of the amino acid biosynthesis mutants will be less severe than in the wild-type tomatoes . 
To this end , we performed competition assays using isogenic mutants in trpC , argA , ilvD , glnA , serA , thrC , gltB , and metA in the tomato cultivar Ailsa Craig and its rin homozygous mutant . 
The competitive indices of argA , gltB , thrC , and trpC mutants were not affected by the tomato genotype , while isogenic mutants for glnA , ilvD , metA , and serA presented a reduction in their fitness within rin mutant tomatoes compared to wild-type tomatoes . 
The metA mutant exhibited the highest difference : in wild-type tomatoes , the CI indicated a 2-fold reduction compared to wild-type Salmonella spp. , whereas in rin mutant tomatoes , the reduction was 32-fold ( Fig. 6 ) . 
LPS biosynthesis is required for Salmonella proliferation in tomatoes . 
Most of the data presented so far underscore the differences in Salmonella requirements for animal and plant colonization . 
However , biosynthesis of LPS was needed under both conditions . 
Genes involved in LPS core and O-antigen synthesis , contained by the rfa and rfb clusters , respectively , were identified as required for tomato colonization in our screen , as well as in transposon insertion sequencing analyses of mouse colonization ( Fig. 2 and 7A ) . 
The competition assays using isogenic mutants for rfaB , rfaI , rfbN , and rfbP con-firmed the transposon insertion sequencing data ( Fig. 7B ) . 
All mutants had an 4-fold fitness reduction . 
The similar CI for the mutants involved in the same biological process strongly suggests that the observed phenotype is related to the defective LPS biosynthesis and not to a secondary mutation . 
The role of LPS during disease in animals has been extensively studied for several pathogens . 
LPS protects bacterial cells from antibacterial peptides produced by host organisms ( reviewed in reference 46 ) . 
Salmonella LPS mutants are severely attenuated during systemic infection and are incapable of developing disease . 
In bacterial plant pathogens , LPS is also a major virulence factor , protecting the pathogen against plant defenses . 
We hypothesize that the immunity mechanisms in plants and animals target the same cellular structures in bacterial cells , and Salmonella 's countermeasures may be similar in animal and plant hosts . 
Salmonella inoculation titer affects fitness of SPI and nucleotide synthesis mutants . 
The transposon insertion sequencing experiments suggested that genes involved in nucleotide biosynthesis and those that are a part of the Salmonella pathogenicity islands ( SPI ) 1 , 2 , and 3 were required for Salmonella full fitness during proliferation in tomatoes ( Data Set S1 ) . 
Auxotrophic mutants for purines and pyrimi-dines exhibited strong attenuation ( Data Set S1 ) . 
SPI-1 , -2 , and -3 are the major Salmonella virulence determinants required for the invasion of epithelial cells , macro-phages , and survival during the intracellular phase ; strains lacking any of these SPIs are avirulent in vertebrate animal models . 
However , the observed phenotypes of the SPI mutants were in direct contradiction to the previous study ( 9 ) . 
It is of note that the inoculum doses used in the transposon insertion sequencing screen were 4 to 5 orders of magnitude higher than in the study of infections in tomatoes performed by 
Noel et al. ( 9 ) . 
Therefore , we tested a hypothesis that the phenotypes of the SPI ( and nucleotide synthesis ) mutants were dependent on the inoculum dose . 
Competition assay for the pyrB , purH , SPI-1 , SPI-2 , and SPI-3 mutants and the wild type were carried out with an inoculum of 103 and 107 CFU per tomato . 
The results of these competition assays showed that the reduction in fitness for all mutants tested was inoculum dependent , with phenotypes apparent only at the high inoculum doses ( Fig. 8 ) . 
For pyrB and purH mutants , which are purine and pyrimidine auxotrophs , respectively , the competition to scavenge the nucleotides available in the environment will increase at higher titers , and the mutants unable to synthesize purines or pyrimidines de novo will have a stronger deficit for these compounds that will negatively affect their growth . 
Proteins encoded on SPI-1 , -2 , and -3 are not known to contribute to population density-dependent functions , and we do not have a ready explanation for the mutants ' phenotype that was dependent on the inoculum density . 
We hypothesize that for the SPI-3 mutant , an increasing Mg2 deficit could lead to the observed phenotype at higher concentrations . 
One of the genes present in SPI-3 , mgtC , encodes a virulence factor required for intracellular proliferation under low-Mg2 conditions . 
The screening data suggest that Salmonella in tomatoes has a deficit in Mg2 , as indicated by the loss of fitness for mutants in the phoPQ system that is activated by low Mg2 , and by the loss of fitness for corA mutants , which lack a Mg2 channel . 
DISCUSSION
The ability of human enteric pathogens , such as Salmonella , to robustly colonize plant tissues likely helps this pathogen build up populations that are high enough to then reinfect herbivorous hosts . 
This ability of Salmonella to exploit plants as a part of its life cycle and use them as vehicles to reach preferred animal hosts was recently supported by multitrophic experiments ( 47 ) . 
Salmonella and enteropathogenic E. coli are not the only members of the Enterobacteriaceae family capable of forming successful relationships with both plants and animals . 
Klebsiella pneumoniae is both a nitrogen-fixing plant endophyte and an opportunistic pathogen in animals and humans ( 48 ) . 
Pantoea stewartii , the causal organism of Stewart 's wilt in sweet corn , spends a portion of its life within the gut of the corn beetle ( 49 ) . 
Dickeya dadantii is a pathogen causing soft rots in plants and is an aphid pathogen ( 50 ) . 
It is possible that these bacteria `` repurpose '' their metabolic and virulence genes to adapt to different hosts , or they may have evolved two distinct strategies for the infections of plants and animals . 
The hypothesis of distinct strategies is supported by the discoveries of two distinct type III secretion systems in P. stewartii , one used in plants and another in beetles ( 49 ) . 
To gain insight into the interactions of Salmonella spp . 
with their different hosts and to compare Salmonella behavior with that of the phytopathogens , we carried out high-throughput screens , followed by the comparative analyses . 
It appears that 40 % of the genes required for tomato colonization by Salmonella are also involved in a systemic infection in mice , and 35 % of the genes required for murine systemic infection are required for colonization of tomato fruits . 
The same pattern was found in the metabolic requirements for phytopathogens , where 42 % of the mapped KO identifiers were shared with Salmonella colonization of tomatoes . 
This partial overlap suggests a certain degree of repurposing of the Salmonella metabolism to adapt to different hosts but also reveals metabolic versatility of this bacterium . 
Salmonella mutants in amino acid biosynthesis genes had the most pronounced phenotype in tomatoes but were not required for murine systemic infection , as determined in high-throughput screenings ( 37 , 38 , 51 ) and competition assays with isogenic mutants ( 52 ) . 
We focused on this group of genes to explore the differences between the requirements for interactions with animals and plants . 
Competition assays confirmed that the biosynthesis of tryptophan , arginine , branched amino acids , glutamine , serine , threonine , glutamate , and threonine is required for colonization of tomatoes . 
In this regard , Salmonella behavior closely paralleled that of phytobacteria : amino acid biosynthesis was required for plant colonization by the pathogens Pseudomonas tolaasii ( 43 ) and X. campestris pv . 
campestris ( 44 ) and by the rhizosphere colonizer Pseudomonas fluorescens WCS365 ( 53 ) . 
The amino acid metabolism requirements for plant-associated bacteria could be a result of unbalanced amino acid concentrations in this environment . 
Amino acids appear to be a relatively minor component of the exudates of mature fruits , leaves , and roots of tomatoes ( 20 ) , which is consistent with the phenotypes of the auxotrophic mutants . 
Because tomatoes carrying the rin mutation are known to accumulate higher levels of certain amino acids ( 45 ) , we tested the hypothesis that the auxotrophy of the Salmonella mutants will be at least partially complemented within rin mutant tomatoes . 
Surprisingly , this did not prove to be the case , and Salmonella auxotrophic glnA , ilvD , metA , and serA mutants were even less fit within rin mutant tomatoes than in the wild type . 
Differences between Salmonella proliferation in Campari and Ailsa Craig cultivars or between the wild type and the rin mutant could influence these results . 
Levels of Salmonella growth in Ailsa Craig ( wild-type ) and Campari tomatoes are not significantly different ( 35 ) , but Salmonella growth is reduced in Ailsa Craig rin mutants ( 35 ) . 
This reduction in growth might indicate that the reduction in fitness of amino acid biosynthesis mutants in rin mutant tomatoes could be even bigger and was masked by a limited number of cell doubling . 
Our observations suggest that either plant amino acids are not accessible to Salmonella spp . 
( due to their compartmentalization or chemical modification ) or that Salmonella spp . 
can make better use of inorganic nitrogen sources during plant colonization . 
LPS biosynthesis was a point of convergence between the requirements for Salmo-nella colonization of plants and animals , as indicated by our screening and confirmed for the competition assays with rfaB , rfaI , rfbN , and rfbP mutants . 
The LPS role for Salmonella virulence in animals is well established ; mutants with a truncated LPS are attenuated , and Salmonella uses phoPQ and pmrAB to modify its LPS to evade the innate immunity through resistance against antibacterial peptides ( 46 ) . 
Plants have antibacterial peptides that are not homologous to the ones found in animals , but they share the same mechanisms of action ( reviewed in references 54 and 55 ) . 
These plant-based antibacterial peptides are small cationic molecules that target the cell membranes , leading to destabilization and cell death . 
LPS is required by plant bacterial pathogens to avoid being targeted by the plants ' antibacterial peptides . 
Salmonella LPS may similarly support bacterial proliferation in this environment . 
Another point of convergence was the biosynthesis of nucleotides and the requirement for the Salmonella pathogenicity island 1 ( SPI-1 ) , SPI-2 , and SPI-3 . 
Competition assays confirmed phenotypes of these mutants only for high , but not low , titers . 
SPI-1 and SPI-2 T3SSs were shown to be important factors for plant colonization from high-inoculum ( 108 CFU/plant ) experiments with Arabidopsis ( 24 ) . 
It is possible that the phenotypes observed are related to increased intraspecies competition due to a relative scarcity of available resources within the plant . 
SPI-1 and SPI-2 may also be needed to subvert immunity , which may become more important as more bacterial cell numbers are invading and therefore activating the plant 's immune response . 
While Salmonella effectors were shown to be functional when directly transformed into cell wall-less plant protoplasts ( 22 ) , it is not clear how they are delivered across the plant cell wall . 
This study demonstrated that while Salmonella is capable of repurposing some of the genes it uses during animal infections to colonize alternate hosts ( such as plants ) , the overlap between the sets of genes required for animal and plant infection is only 40 to 50 % . 
Conversely , only similarly partial overlap was observed when Salmonella genes required for tomato colonization were compared with the genes required for virulence of phytopathogens . 
Considering that 520 features required for fitness of Salmonella in tomatoes have no known functions , there remains much to be learned about the ecology of this human pathogen inside alternate hosts , such as plants . 
MATERIALS AND METHODS
Bacterial strains and DNA manipulation . 
Bacterial strains were grown in LB broth at either 30 °C or 37 °C ( as indicated in the text ) , and antibiotics were added at the following concentrations : 100 g/ml ampicillin , 50 g/ml kanamycin , and 20 g/ml tetracycline , unless otherwise indicated . 
A full list of the strains and plasmids used in this study is in Table 1 . 
Salmonella isogenic mutants used to confirm screening data were constructed using Datsenko-Wanner mutagenesis ( 56 ) by replacing a specific open reading frame ( ORF ) with the kanamycin resistance cassette from pKD4 , followed by P22-mediated transduction into S. enterica serovar Typhimurium ATCC 14028 . 
RIVET reporters to evaluate in planta gene expression were constructed as described previously ( 57 ) ; the kanamycin cassette from isogenic mutants was removed with FLP recombinase using pCP20 , followed by the insertion of tnpR-lacZ , from pCE70 or PCE71 in the remaining FLP recombination target ( FRT ) site . 
The tnpR-lacZ cassette , which replaced almost the entire ORF of the deleted gene , was inserted immediately after the start codon . 
The tetracycline resistance gene tetA , flanked by the res1 sites in a neutral site of the genome , was transduced using P22 to the final construct from the donor strain JS246 . 
Primers used for cassette construction and deletions confirmation are listed in Table S1 in the supplemental material . 
Plant material . 
Tomatoes of the cultivar ( cv . ) 
Campari were purchased at the local supermarket . 
Tomatoes of the cv . 
Ailsa Craig and its isogenic derivative rin mutant line were grown in a rooftop greenhouse . 
Tomatoes were tagged at anthesis and were harvested 34 days later to ensure that all tomatoes were in the same developmental stage , as described before ( 35 ) . 
Transposon insertion library construction . 
A library of S. enterica serovar Typhimurium 14028 Tn5 insertion mutants was constructed using the Epicentre EZ-Tn5 T7/Kan2 promoter insertion kit . 
Briefly , primers Right_Tn_T7_Kan2 and Kan2_right_code were used in a standard PCR to add N18 barcodes to both sides of the EZ-Tn5 T7/KAN -2 transposon . 
The PCR product was gel purified using the QIAquick gel extraction kit ( Qiagen ) , according to the manufacturer 's recommendations . 
About 200 ng of the purified PCR product was used in an 8 - l transposase reaction mixture in 0.17 % glycerol that also contained 1 l of TypeOne restriction inhibitor and 2 U of EZ-Tn5 transposase . 
The reaction mixture was incubated at room temperature for 3 h and subsequently dialyzed against water for 30 min before electroporation into fresh electrocompetent cells of S. enterica serovar Typhimurium 14028 ( MZ1597 ) . 
Transformed cells were isolated on LB agar plates with kanamycin ( 60 g/ml ) after overnight growth at 37 °C , enumerated , and collected . 
The final library consisted of a pool of about 280,000 independent colonies . 
Mapping of library barcodes to the S. enterica serovar Typhimurium ATCC 14028 genome . 
Genomic DNA of the obtained library of S. enterica serovar Typhimurium ATCC 14028 Tn5 insertion mutants was prepared using the GenElute bacterial genomic DNA kit ( Sigma-Aldrich ) . 
The DNA was fragmented and ligated to Illumina primers using standard methods . 
Approximately 150 ng of this material was then used to PCR amplify the regions flanking the transposon insertion sites , in a stepwise PCR regimen . 
In the first PCR , primer Illumina_P5_Read1 was paired with a primer that aligned with a region close to either the left end ( primer Tn5_Left_CGTACA_Read2 ) or the right end ( primer Tn5_Right_ACATGC_Read2 ) of the inserted transposon . 
PCR proceeded for 20 cycles , using 1.25 U of Taq polymerase and under standard conditions . 
Exactly 1/10 of the 1st PCR product was subsequently used in a 2nd PCR that engaged Illumina primers Illumina_P5_Read1 and Illumina_P7 and proceeded for 10 cycles . 
Products were subjected to QIAquick PCR product purification ( Qiagen ) , according to the manufacturer 's recommendations . 
The material was subsequently Illumina sequenced for 150-bp reads at both ends . 
Sequence analysis to map the barcode to the exact location on the genome is described further below . 
Screen optimization . 
To ensure that the inoculum representative of the transposon insertion library was seeded into tomatoes and that a mutant 's fitness could be assessed by changes in its relative abundance , competitive fitness assays with mutants known to have a phenotype in tomatoes ( hns : : kan , ΔbcsA ΔlpfA ΔfadH cysB : : kan , phoN : : kan , and rcsA : : kan ) were carried out . 
Although not all combinations of these mutants were tested in all preliminary experiments , we compared the impact of the different inoculum doses ( 104 and 107 ) and various ratios of the mutants to the wild type . 
Prior to the infections into tomatoes , strains were grown for 16 h in LB broth at 37 °C 250 rpm . 
The cultures were pelleted , washed in phosphate-buffered saline ( PBS ) twice , and diluted 1:10 , reaching a final density of approximately 108 CFU/ml . 
When 107 CFU of Salmonella was inoculated into tomatoes , we were able to reliably observe even modest decreases in fitness . 
Therefore , this inoculation dose was selected for the library screen . 
Library screening . 
The MZ1597 library was screened in Campari tomatoes using the tomato wound model to identify loci that affect Salmonella fitness in this environment . 
We chose this model based on the FDA assessment that wounds and punctures are a potential route for internalization of human pathogens ( such as Salmonella ) in fresh fruits and vegetables , and specifically in tomatoes 
( http://www.fda.gov/Food/GuidanceRegulation/HACCP/ucm082063.htm ) . 
MZ1597 cultures were grown ( with shaking at 250 rpm ) for 16 h in LB broth supplemented with kanamycin at 37 °C . 
The cultures were pelleted , washed in PBS twice , and diluted 1:10 , reaching a final density of approximately 108 CFU/ml ; 3 l of this suspension was inoculated into three shallow ( 2 - to 3-mm deep , 1 mm in diameter ) wounds in tomato pericarps ( 106 CFU per tomato ) using a small pipette tip insertion under the epidermis of the tomato fruits . 
Tomatoes were incubated at 22 °C and a relative humidity of 60 % for 7 days , a period of time that allows Salmonella to reach the carrying capacity of tomato fruits and approximates the time period between tomato harvest and consumption . 
Salmonella was recovered by collecting 1-g samples of the pericarp around the inoculation site ; samples from the same fruit were combined and homogenized in a stomacher ( Sevard ) . 
Salmonella cells were recovered by centrifugation and were then resuspended in 50 ml of LB broth , followed by 6 h growth at 37 °C and 250 rpm , reaching 108 CFU/ml . 
One milliliter of culture was recovered and used for library preparation . 
Library preparation for sequencing . 
Aliquots of around 5 107 CFU from input and output libraries were subjected to three washes in water , followed by proteinase K ( 100 g ) digestion for 1 h at 55 °C in lysis buffer ( 10 mM Tris [ pH 8.0 ] , 1 mM EDTA , 0.1 % Triton X-100 ) . 
After inactivation of the enzyme for 10 min at 95 °C , a nested PCR regimen was performed to specifically amplify the DNA regions adjacent to the inserted transposons . 
Primers 1st_PCR_Tn5_EZ_left_reverse / Left_Forward_201 and Right _ Reverse_fixed / Right_Forward_983 were used to amplify the left and right flanks , respectively , in a 20-cycle PCR in 50 l of 1 Kapa HiFi reaction mixture . 
One microliter of the successful PCR product was then utilized as the template in a second rate-limited PCR amplification using 1.25 U of Taq ( Invitrogen ) and 0.1 mM dinucleoside triphosphates ( dNTPs ) , with primers 2nd_PCR_Tn5_EZ_Right_Forward and Tn _ SetYY_XXXXXXXX . 
The second PCR amplified the ( already enriched ) right flank of the transposon insertion site . 
Products of the second PCR were pooled and subjected to QIAquick PCR product purification ( Qiagen ) , according to the manufacturer 's recommendation . 
Illumina sequencing proceeded with custom primers Tn5_EZ_Right_Seq_fixed and Tn5_EZ_Index_Seq_new for a single indexed run , with a read length of 25 bases . 
Sequence analysis . 
For mapping of the barcodes introduced into the mutants of the transposon insertion library , the Tn5 primers and the corresponding N18 sequence flanking the Tn5 priming sites were trimmed from the raw sequence reads using custom Perl scripts . 
The trimmed reads were mapped to the S. Typhimurium 14028 reference genome using Bowtie2 . 
PCR duplicate reads were removed using Picard tools ( https://github.com/broadinstitute/picard ) . 
Only reads mapped in proper pairs were considered for further downstream analysis . 
The left-most position of read 2 for each mapping was extracted from the SAM alignments using custom scripts , and this was identified as the Tn5 insertion site in a strand-specific manner . 
The N18 barcode tag for each mapped read was then identified from the raw untrimmed reads using custom Perl scripts . 
The above-mentioned analysis identified the N18 barcode tag flanked by conserved priming sites for each Tn5 insertion mutant . 
For the identification of barcoded mutants , raw sequencing data consisted of single-end 25-bp reads . 
The first 18 bases , which represented the unique N18 tag for each Tn5 mutant , were extracted , and the abundance of all unique 18-mers was calculated using custom Perl scripts . 
The abundances of all N18 barcodes mapped within each annotated genome feature were summed in a strand-specific manner . 
This represented the aggregated abundance for each feature in the coding strand and the noncoding strand . 
The aggregated abundances for the input and output libraries were statistically analyzed using edgeR , and the log2 fold changes and FDRs were reported . 
Competition assays . 
Competition assays were used to estimate the fitness of specific mutants . 
Overnight cultures of the wild type and an isogenic mutant were adjusted based on optical density at 600 nm ( OD600 ) to the same population density and combined in a 1:1 ratio . 
The combined cultures were washed three times in PBS and diluted 10,000-fold , and 3 l of the diluted mixture was inoculated into the tomato pericarp in three separate shallow wounds ( 103 CFU per tomato ) . 
An aliquot of the inoculum was plated onto xylose lysine deoxycholate ( XLD ) plates to enumerate CFU . 
The inoculated tomatoes were incubated at 22 °C and relative humidity 60 % for 7 days , and Salmonella cells were recovered by inserting a loop into the wound and streaking the material onto XLD plates . 
The competition index was calculated as previously described ( 9 ) using the formula log2 ( [ MUTout/WTout ] / [ MUTin/WTin ] ) , where MUT is the number of mutant colonies recovered , WT is the number of wild-type colonies recovered , `` in '' represents the cultures inoculated in tomatoes , and `` out '' represents the colonies recovered after the experiment . 
Competition index ( CI ) statistical significance values of isogenic mutants were compared against the CI for the neutral mutant CEC1000 using analysis of variance ( ANOVA ) . 
The CI significance for the rcsA : : kan mutant against CEC1000 , the CI for isogenic mutants in the wild type against rin mutant tomatoes , and the comparison of different inoculation titers were done using a pairwise t test . 
The software JMP version 12 was used for all CI analyses . 
RIVET assays . 
RIVET reporters were used to evaluate Salmonella gene expression in planta . 
In this system , gene expression is quantified as the percentage of cells that lost tetracycline resistance [ which ultimately results from the activation of the promoter of interest that leads to the transcription of recombinase tnpR , and TnpR-catalyzed excision of tetA flanked by res1 sites , causing tetracycline sensitivity ( 57 ) ] . 
For RIVET assays in tomatoes , Salmonella strains were grown in LB broth with tetracycline for 16 h at 37 °C . 
Cultures were then washed extensively in PBS to remove traces of the medium and the antibiotic , and inoculum suspensions in PBS were seeded into tomato pericarps ( 103 cells per tomato ) in three separate shallow wounds . 
Cells were recovered 1 , 3 , and 7 days after inoculation using a sterile wire loop , and the material was streaked onto XLD plates . 
For the positive control of RIVET activation , cultures were grown in M9 agar ( 0.7 % ) supplemented with the appropriate amino acids ( leucine , isoleucine , and valine for ilvD , arginine for argA , serine for serA , glutamine for glnA , glutamate for gltB , methionine for metA , tryptophan for trpC , and threonine for thrC ) . 
The tested amino acid concentrations ranged from 2 mg/ml to 8 g/ml in a 2-fold dilution series , and cells were recovered from the lowest concentration that allowed for growth of the reporter . 
Cultures were then streaked onto XLD . 
The percentage of Salmonella cells sensitive to tetracycline recovered from tomatoes and M9 with the amino acid was estimated by plating the colonies from XLD plates into LB tetracycline plates . 
Similar control experiments were done in LB agar ( 0.7 % ) , only without the supplementation with amino acids . 
Metabolic mapping and functional characterization . 
Genes required for Salmonella systemic infection in mice were retrieved from data sets deposited with the original publications using parameters established by the authors of the original papers . 
For the study by Chaudhuri et al. ( 37 ) , a P value of 0.05 and log2 ( fold change ) of less than 1 were used ; for the study by Santiviago et al. ( 38 ) , a P value of 0.0005 and log2 ( fold change ) of less than 0.75 were used ; and for the study by Silva et al. ( 51 ) , a P value of 0.001 and log2 ( fold change ) of less than 0.75 were used . 
BlastKOALA ( 58 ) was used to assign KEGG Orthology ( KO ) terms for Salmonella enterica ATCC 14028 coding sequences , and the KEGG Mapper Web interface was used to visualize metabolic pathways . 
Gene Ontology ( GO ) term enrichment was performed with Panther ( 59 ) , with Bonferroni 's correction . 
Genes required for bacterial phytopathogens to elicit disease in their preferred plant host were extracted from previously published studies ( 39 -- 42 , 44 , 60 ) , and the KO terms for them were also retrieved with BlastKOALA . 
It should be noted that the results of mutant screens in phytopathogens were less comprehensive than the high-density transposon screens . 
We also note that , generally , screens in phytopathogens identified fewer ( 10 to 200 ) functions required for virulence in plants , which may also represent more stringent screening conditions than those of the studies with high-density transposon insertion libraries . 
Growth in LB . 
To evaluate potential growth defects of isogenic mutants , we compared their growth kinetics with the wild-type strain . 
Cultures grown for 16 h at 250 rpm and 37 °C in LB broth were diluted to an OD600 of 0.02 in 50 ml of LB broth . 
Three replicate cultures of each strain were incubated at 37 °C and 250 rpm . 
Samples were withdrawn hourly , and the OD600 was measured with a BioSpec-mini spectrophotometer ( Shimadzu ) . 
SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/ AEM.03028-16 . 
ACKNOWLEDGMENTS
We are grateful to Alex Gannon for technical assistance . 
We thank J. Giovannoni for sharing tomato seeds . 
This study was supported by funding from USDA-NIFA and the Center for Produce Safety .