RNA Sequencing Reveals Differences between the Global Transcriptomes of Salmonella enterica Serovar Enteritidis Strains with High and Low Pathogenicities

Abstract

Salmonella enterica serovar Enteritidis is one of the important causes of bacterial food-borne gastroenteritis worldwide. Field strains of S. Enteritidis are relatively genetically homogeneous; however, they show extensive phenotypic diversity and differences in virulence potential. RNA sequencing (RNA-Seq) was used to characterize differences in the global transcriptome between several genetically similar but phenotypically diverse poultry-associated field strains of S. Enteritidis grown in laboratory medium at avian body temperature (42°C). These S. Enteritidis strains were previously characterized as high-pathogenicity (HP; n = 3) and low-pathogenicity (LP; n = 3) strains based on both in vitro and in vivo virulence assays. Using the negative binomial distribution-based statistical tools edgeR and DESeq, 252 genes were identified as differentially expressed in LP strains compared with their expression in the HP strains (P < 0.05). A majority of genes (235, or 93.2%) showed significantly reduced expression, whereas a few genes (17, or 6.8%) showed increased expression in all LP strains compared with HP strains. LP strains showed a unique transcriptional profile that is characterized by significantly reduced expression of several transcriptional regulators and reduced expression of genes involved in virulence (e.g., Salmonella pathogenicity island 1 [SPI-1], SPI-5, and fimbrial and motility genes) and protection against osmotic, oxidative, and other stresses, such as iron-limiting conditions commonly encountered within the host. Several functionally uncharacterized genes also showed reduced expression. This study provides a first concise view of the global transcriptional differences between field strains of S. Enteritidis with various levels of pathogenicity, providing the basis for future functional characterization of several genes with potential roles in virulence or stress regulation of S. Enteritidis.

INTRODUCTION

Salmonellosis is one of the leading causes of diarrheal illness in humans, with an estimated 1 million cases of food-borne illnesses resulting in 19,336 hospitalizations and 378 reported deaths annually in the United States (1). Salmonella enterica serovar Enteritidis (S. Enteritidis) is the most common nontyphoidal Salmonella serovar responsible for food-borne gastroenteritis and is usually one of the top two serovars reported, along with S. enterica serovar Typhimurium, in surveys from various nations around the world (2–4). S. Enteritidis primarily causes food-borne gastroenteritis, which is characterized by diarrhea, fever, headache, abdominal pain, nausea, and vomiting (5). In addition, S. Enteritidis has been recently reported to cause increased incidence of invasive, recurrent, and multiple-site infections in African countries (4, 6, 7). Poultry is the major reservoir of S. Enteritidis, and live poultry or poultry products, such as eggs and meat, are the primary sources of human infection worldwide (3, 8–12). It has been demonstrated that, irrespective of the phage type (PT) or source or geographical location of isolation, field strains of S. Enteritidis are relatively genetically homogeneous (13–16). Recent high-resolution genetic studies using individual-gene or whole-genome sequencing have revealed that S. Enteritidis is relatively genetically homogeneous, with the major genetic differences between field strains of S. Enteritidis occurring at the level of single nucleotide polymorphisms (SNPs) (14, 17, 18). However, field strains show remarkable differences in their phenotypes and virulence potentials (15, 16, 18–24). For instance, it has been reported that naturally occurring strains of S. Enteritidis vary in their virulence potentials in murine and chicken models of infection (18, 20, 22, 23, 25–29). In addition, field strains of S. Enteritidis show remarkable differences in terms of their epithelial cell invasiveness, motility, biofilm production, resistance to acidic and oxidative stress, and the ability to survive within avian macrophages and egg albumen (15, 18–20, 22, 24, 30–32). The role that these SNPs might play in differential phenotypic characteristics, virulence regulation, or persistence of strains within the host or environment remains elusive.

It was recently reported that certain poultry-associated field strains of S. Enteritidis show impaired virulence in orally infected BALB/c mice and day-old chickens while most other strains are naturally virulent (18, 20). For the purpose of this study, naturally virulent strains were designated high-pathogenicity (HP) strains, whereas strains with impaired virulence were designated low-pathogenicity (LP) strains. LP strains were previously reported as having low invasive capabilities in cultured human and avian epithelial cells and showed impaired survival within avian macrophages and reduced resistance to acidic and oxidative stress (18, 20). Comparative genomic hybridization microarray did not reveal genetic differences between LP and HP strains that could be attributed to their phenotypes (18). Consequently, a central hypothesis for this study was that the differences in virulence and phenotypic properties of LP strains of S. Enteritidis are due to the underlying differences in their global transcriptional signatures. To test this hypothesis, next-generation mRNA sequence analysis (RNA-Seq) was performed to identify genes differentially expressed by multiple HP (n = 3) and LP (n = 3) strains of S. Enteritidis in response to growth in laboratory medium at avian body temperature (42°C). Using multiple strains to obtain a comprehensive image of their global transcriptional signatures, this study demonstrates that the LP strains have a distinct signature that is characterized by significantly reduced expression of virulence and stress-associated genes and that these differences correlate with their respective phenotypes. To the best of my knowledge, this is the first study that compared transcriptional signatures of multiple well-characterized S. Enteritidis strains using RNA-Seq. The deduced functions of differentially regulated genes in relation to virulence and stress are also discussed.

MATERIALS AND METHODS

Bacterial strains.

Six representative wild-type S. Enteritidis strains (UK, G1, BC8, C19, C45, and G45) isolated from poultry were analyzed in this study (Table 1). In previous studies, these strains showed consistent differences in their phenotypic characteristics and virulence potentials in avian and murine models of infection (19, 20). Based on these characteristics, the selected strains were classified in two groups: high-pathogenicity (HP) and low-pathogenicity (LP) groups. The HP group included the UK (corresponds to the sequenced phage type 4 [PT4] P125109 strain), G1 (phage type 4), and BC8 (phage type 8) strains, whereas the LP group included C19 (phage type 13), C45 (phage type 13), and G45 (phage type 13a) strains. Frozen stocks of cultures, stored in 15% (vol/vol) glycerol at −80°C, were grown on Luria-Bertani (LB) agar for 16 h at 37°C.

TABLE 1.

Phenotypic characteristics of high-pathogenicity and low-pathogenicity strains of S. Enteritidis used in this study^a

Group	Strain	Phage type	MLVA typeb	Caco-2 cell invasiveness	Survival or growth characteristic by condition				Motility	Biofilm
					Avian macrophages	Acidic stress (pH 2.6)	Oxidative stress	Egg albumen
LP	C19	PT13	11	Low	Low	Low	Low	No growth	Low	Negative
	C45	PT13	11	Low	Low	Low	Low	No growth	Low	Negative
	G45	PT13a	4a	Low	Low	Low	Low	Growth	Low	Negative
HP	G1	PT4	13a	Medium	High	High	High	Growth	High	Positive
	UK	PT4	13a	High	High	High	High	Growth	High	Positive
	BC8	PT8	9a	High	High	High	High	Growth	High	Mixedc

^aSee references 19 and 21. For each category, “low” indicates that the difference is statistically significant (P < 0.05) in comparison with “high.”

^bMLVA, multilocus variable-number tandem-repeat analysis.

^cMixed, this strain has a mixed subpopulation of both biofilm-positive and -negative cells.

Preparation of RNA samples.

For all experiments, a single colony from the overnight culture was inoculated into LB broth and grown at 42°C (normal body temperature of chicken) for 16 h with shaking at 200 rpm. The overnight cultures were diluted 1:100 in fresh LB broth and incubated at 42°C for 4 h (exponential phase) with shaking at 180 rpm. Approximately 1 × 10⁹ CFU of each strain was pelleted (at 7,500 rpm and 20°C for 15 min). There were no differences in the in vitro growth kinetics of LP and HP strains at 42°C (data not shown). The cell pellets were processed for total RNA extraction and DNase treatment using a RiboPure Bacterial Kit (Ambion, USA) according to the protocol supplied by the manufacturer. Total RNA was extracted from three replicates of each strain. Subsequently, equal concentrations of the total-RNA samples obtained from three replicates of each strain were mixed and subjected to a second treatment with DNase to remove any residual genomic DNA, according to the protocols supplied with the RiboPure Bacterial Kit. The pooled total-RNA sample from each strain was tested for genomic DNA (gDNA) contamination by real-time quantitative PCR (qPCR) amplification of two genes, rpoD and sipA, using the following primers: rpod_F, 5′-ACATGGGTATTCAGGTAATGGAAGA-3′; rpo_R, 5′-CGGTGCTGGTTGGTATTTTCA-3′; sipA_F, 5′-TTTTAACGCCTCAGCGTCTT-3′; sipA_R, 5′-CAGAGAAAGTGCCACAACGA-3′. The qPCR was performed using SsoFast EvaGreen Supermix as per the manufacturer's instructions (Bio-Rad) using an iQ5 iCycler (Bio-Rad, USA). Each RNA sample was tested in duplicate. All samples showed threshold cycle (C_T) values of >30 for both rpoD and sipA, indicating negligible levels of DNA contamination. As a final step, RNA integrity was assessed using a 2100 Bioanalzyer (Agilent, Foster City, CA). The total-RNA samples were stored at −80°C until further use.

RNA-Seq.

The total-RNA samples were submitted to Illumina, Inc., San Diego, CA, for mRNA enrichment and subsequent RNA-Seq. Removal of 16S and 23S rRNAs from total RNA was performed using a duplex-specific nuclease (DSN) treatment (33). The mRNA was used to prepare individually bar-coded (indexed) RNA-Seq libraries with a TruSeq RNA Sample Prep Kit (Illumina, USA). Six RNA-Seq libraries prepared from S. Enteritidis strains were sequenced in one lane on a HiSeq2000 instrument (Illumina Inc., USA) along with the six other unrelated RNA-Seq libraries, resulting in a total of 12 libraries per lane using version 3 chemistry, and reads were base called and quality filtered with the CASAVA (Consensus Assessment of Sequence and Variance) version 1.8, pipeline (Illumina, Inc.) to generate 75-bp reads. The genome sequence and functional annotation information of S. Enteritidis were obtained from the NCBI database (GenBank accession number AM933172). CASAVA (version 1.8) quality-filtered reads were aligned to the reference genome sequence using CLC Genomics Workbench, version 5.0 (CLC Bio, USA). Mapping was based on the minimal length of 75 bp with an allowance of up to two mismatches, and >90% of the each read's length had to map to the reference sequence for it to be considered a mapped read. After reads were mapped with the CLC Workbench, total raw read counts for each gene were generated. These read counts were used for further statistical analysis to determine differentially expressed genes as described below.

Experimental design and data analysis.

It is recommended that for RNA-Seq-based comparative transcriptome analysis, the experimental design should include at least three biological replicates per treatment group (34). The primary objective of this study was to identify genes that are consistently differentially expressed in all LP strains compared with all HP strains. The experimental design included three wild-type S. Enteritidis strains in each treatment group: for the HP group, strains UK, G1, and BC8; for the LP group, strains C19, C45, and G45. Therefore, for the purpose of this study, the individual strains within each group served as independent biological replicates for that group. Biological replicates within the HP and LP groups demonstrated >0.86 and >0.92 correlations (Spearman rank, P < 0.01), respectively, when RPKM (reads mapping to the genome per kilobase of transcript per million reads sequenced)-normalized values were compared, indicating high reproducibility of replicates. For the subsequent statistical analysis, the total raw read count data for the six S. Enteritidis strains were used. Because of the high variability between methods for determining differential expression using RNA-Seq data, it has been recommended that more than one bioinformatics tool be used to ensure that a conservative list of differentially expressed genes is obtained (35). Therefore, two well-established statistical methods (DESeq and edgeR) were employed to analyze RNA-Seq data using the R software package (version 2.1.5.2) (36). As a first step, any gene that had zero mapped reads for all six samples was removed, resulting in 4,418 genes mapped by the CLC Workbench out of the 4,420 genes comprising the S. Enteritidis transcriptome (see File S1 in the supplemental material). For DESeq analysis, differential expression testing was performed by using negative binomial distribution and a shrinkage estimator for the distribution's variance and size factor-normalized data (37). Differential expression analysis was also performed using edgeR, which also employs a negative binomial distribution-based method (38–40). For edgeR analysis, the trimmed mean of the M values (TMM; where M = log₂ fold change) method was used to calculate the normalization factor, and the quantile-adjusted conditional maximum likelihood (qCML) method for estimating dispersions was used to calculate expression differences using an exact test with a negative binomial distribution (38–40). These analyses were performed independently using the same mapping file (see File S1). Finally, the lists of genes obtained by the DESeq and edgeR methods were integrated to identify genes determined to be differentially expressed by both methods. The genes that were determined to be significantly regulated by only one statistical method were not considered further (35).

Nucleotide sequence accession number.

All raw data have been submitted to the Gene Expression Omnibus (GEO) under accession number GSE46391.

RESULTS AND DISCUSSION

The wild-type strains of S. Enteritidis differ in their virulence potentials and other phenotypic characteristics, such as motility, biofilm production, survival in egg albumen, and tolerance to acidic and oxidative stress (15, 18–26, 29–32, 41). To explore if a genetic underpinning for the strain differences that were previously characterized as high-pathogenicity (HP) and low-pathogenicity (LP) strains resides within their global transcriptional signatures, a comparative transcriptomic analysis of multiple well-characterized HP and LP strains of S. Enteritidis was performed using RNA-Seq. In RNA-Seq analysis, sequence depth is one of the important factors that influence downstream differential gene expression data analysis. For instance, Hass et al. (42) showed that a sequencing depth of 5 to 10 million non-rRNA fragments is needed for optimal profiling of the vast majority of transcriptional activities in diverse bacterial species, such as Escherichia coli, Mycobacterium tuberculosis, and Vibrio cholerae, grown under diverse culture conditions (42). In this study, alignment of the sequence reads to the S. Enteritidis P125109 genome (43) yielded 17.3 million (BC8), 18.1 million (G1), 18.5 million (UK), 18.7 million (G45), 20.9 million (C19), and 21.4 million (C45) total mapped reads. The total numbers of non-rRNA and non-tRNA reads that mapped uniquely to the reference genome were 13.8 million (G1), 15.8 million (UK), 16.4 million (BC8), 17.5 million (G45), and 20.4 million (C45 and C19). Therefore, the sequencing depth obtained in this study should provide optimal coverage of the S. Enteritidis transcriptome.

As a first step in the data analyses, any genes that had zero mapped reads for all six strains were removed, resulting in mapping of 4,418 out of the 4,420 genes comprising the S. Enteritidis reference genome (43), indicating >99% coverage of the whole transcriptome of S. Enteritidis (see File S1 in the supplemental material). For differential expression analysis between HP and LP strains, two statistical methods, edgeR (38–40, 44) and DESeq (37), were employed. Both methods are based on the negative binomial distribution and have been widely accepted for modeling the variation inherent between biological replicates. The DESeq method identified a total of 298 differentially expressed genes that showed ≥2-fold differences in the transcript abundances between the HP and LP groups, with a P value of <0.05 after adjusting for multiple testing for each gene (P_adj). In contrast, the edgeR method identified 498 differentially expressed genes with ≥2-fold differences in the transcript abundances between the HP and LP groups, with a false discovery rate (FDR) of ≤5% and a P value of <0.05. From the above list of genes, all the genes from ϕSE20 (SEN1919A-SEN1966) were removed because this phage is present in S. Enteritidis PT4 strains but absent from non-PT4 strains. This resulted in the total of 253 (DESeq) and 448 (edgeR) genes that were differentially expressed between HP and LP strains. Because the edgeR method identified 196 additional genes representing approximately 43.5% of the significant genes, a single optimized data set was constructed by integrating the results analyzed separately by DESeq and edgeR (35). This resulted in selection of a list of overlapping genes, with a total of 252 genes that were identified as differentially expressed between the HP and LP groups by both methods (see Files S2 and S3 in the supplemental material). As expected, this approach resulted in elimination of 197 genes that were identified as differentially expressed by only one method (see File S4). For this set of 197 genes, the log₂ fold change values and the levels of significance (P values) obtained by both edgeR and DESeq were compared. Interestingly, 96% (189/197) of the genes that were originally eliminated from the analysis showed ≥2-fold differences in transcript abundances between HP and LP groups when analyzed by both methods (Fig. 1A; see also File S5). The edgeR method revealed that these differences were statistically significant, with a ≤5% FDR and a P value of <0.05 (Fig. 1B; see also File S5). In contrast, none of these differences achieved statistical significance (P ≥ 0.05) when analyzed using the DESeq method (Fig. 1B; see also File S5). Because of the discrepancy between the two statistical methods, this set of 197 genes was not considered for further analysis. However, it is possible that some of these genes are likely to be potentially differentially expressed, and therefore this gene list is included as File S4 in the supplemental material.

FIG 1.

Summary of functional classes of genes downregulated in low-pathogenicity strains compared with the high-pathogenicity strains of Salmonella Enteritidis.

Differentially expressed genes between LP and HP strains.

The majority of differentially expressed genes (93.2%, or 235 out of 252 genes) showed significantly reduced expression in all LP strains compared with all HP strains (see File S2 in the supplemental material). According to their cluster of orthologous groups (COGs), genes expressed at low levels were classified into seven broad functional categories including genes encoding Salmonella pathogenicity islands (SPIs) (n = 32), cellular processes and signaling (n = 65), metabolism (n = 37), information storage and processing (n = 13), fimbriae (n = 4), pseudogenes (n = 5), and several functionally uncharacterized or poorly characterized genes (n = 79) (Fig. 1). In contrast, only 17 (6.8%) genes showed significantly increased expression in all LP strains compared with all HP strains (see File S3), including genes involved in metabolism (n = 10) and cellular processes and signaling (n = 4) and genes with unknown functions (n = 3).

Virulence-associated genes.

SPI-1 encodes a type 3 secretion system (T3SS) that plays an important role in invasion of intestinal epithelial cells, translocation of effector proteins in the host cells, and induction of enteropathogenesis caused by Salmonella (45, 46). RNA-Seq analysis revealed that all of the 38 genes located on SPI-1 showed reduced expression in LP strains (Fig. 2A; see also File S5 in the supplemental material). Of these, 28 showed a ≥2-fold reduction in transcript abundance with a ≤5% FDR and a P value of <0.05 (see File S2). The remaining 10 genes also showed reduced expression, but the differences were not statistically significant. In addition, two genes (spoE and spoE2) that encode effector proteins translocated by the SPI-1 T3SS but are located elsewhere in the genome showed reduced expression. Finally, three genes expressed at low levels (pipBC and sopB) were located on SPI-5, which reportedly contributes to intestinal pathogenesis in murine, bovine, and avian models (47–49).

Among the non-SPI pathogenic factors that showed reduced expression in LP strains, two genes (cydA and cydB) encode a putative cytochrome bd oxidase and ggt encoding γ-glutamyl transpeptidase (GGT). Cytochrome bd oxidase performs a variety of physiological functions in prokaryotes, such as energy-transducing respiration, aerotolerant nitrogen fixation, and protection against metal toxicity and oxidative stress (50–54). This enzyme has also been implicated in the virulence of Shigella flexneri and several S. enterica serovars such as Typhimurium, Gallinarum, and Dublin (55), suggesting that cytochrome bd may be particularly important for the growth and survival of pathogens that encounter environments in which O₂ is progressively limited. GGT (EC 2.3.2.2) is reported to contribute to the virulence of Helicobacter pylori in mice (56, 57) and has been implicated in inhibition of T-cell proliferation and mediation of cell apoptosis (58, 59). Another gene, yncD, encodes TonB-dependent transporters and was recently identified as an in vivo-induced antigen contributing to S. enterica serovar Typhi virulence in a murine model (60). It is important to note that all of the LP strains used in this study exhibited reduced invasiveness in cultured human intestinal cells and reduced virulence in murine and avian models of infection (18, 20). Therefore, reduced expression of SPI-1, SPI-5, and other non-SPI-related genes in LP strains likely explains the mechanisms underlying their virulence attenuation.

Motility-associated genes.

The flagellar regulation of Salmonella includes more than 50 genes divided into three classes, class 1 (early genes), class 2 (middle genes), and class 3 (late genes), according to their temporal expression after induction of the flagellar regulon (61, 62). Motility and flagella are also important to gastrointestinal disease caused by Salmonella in avian, rodent, and bovine models (63–65). Additionally, in vitro studies using cultured epithelial cells show that motility-impaired S. Enteritidis mutants are defective in entering intestinal epithelial cells (66–68). Interestingly, RNA-Seq analysis revealed that, irrespective of the class, the vast majority of flagellar genes (n = 51) showed significantly reduced expression in all LP strains tested (see File S2 in the supplemental material). These included the majority of genes involved in flagellar assembly (Fig. 2B; see also File S5) and several genes encoding known or putative methyl chemotaxis proteins, such as SEN2995, SEN3058, SEN1374, and SEN2296 (see also File S2). In addition, other genes with ties to motility included yhjH and ycgR, which may encode novel flagellar components, given their ability to rescue motility defects of an H-NS (histone-like nucleoid structuring protein) mutant (69–71). Recent studies have revealed that several field strains of S. Enteritidis either express paralyzed or unipolar flagella or may fail to express flagella, resulting in motility-impaired strains that may also be virulence attenuated (20–22, 41). All of the LP strains of S. Enteritidis used in this study represent a population of such naturally occurring, motility-impaired phenotypes (18, 20). The motility impairment might be due to the impaired ability to secrete flagellar proteins such as FljB, FlgK, and FlgL (20). In addition, other investigators have shown that a point mutation (T551 → G) in motA, a gene essential for flagellar rotation, may also result in naturally induced motility impairment in field isolates (21). Therefore, reduced expression of entire flagellar gene clusters in LP strains correlates with the previously reported motility-impaired phenotype among these strains and suggests that flagellar regulation is significantly impaired in LP strains.

Fimbrial genes.

S. Enteritidis harbors 13 fimbrial gene clusters (43). In this study, three fimbrial genes (bcfA, safA, and csgC) showed significantly reduced expression in LP strains. In S. Typhimurium, Saf fimbriae encoded on SPI-6 have been implicated in porcine ileal colonization and virulence in mice (72, 73). The role of Saf fimbriae in S. Enteritidis pathogenesis is not known, but the safA gene in S. Enteritidis shows 81% nucleotide identity with S. Typhimurium (43), suggesting that it may have similar functions. S. Typhimurium csg is important for biofilm formation on chicken intestinal mucosa cultured ex vivo (74). In contrast, S. Enteritidis Csg (curli) fimbriae have been implicated in egg contamination (75) and may also contribute to human and avian epithelial cell invasiveness (68). In addition to these known fimbrial genes, the hopD gene encoding the putative bifunctional prepilin peptidase HopD showed reduced expression in LP strains. The amino acid sequence of HopD showed 60% similarity with the PilD peptidase protein of Pseudomonas stutzeri DSM 4166. Functional PilD is required for extracellular secretion (excretion) of several virulence-associated proteins such as exotoxin A, phospholipase C, and elastase production in P. aeruginosa (76, 77). While the exact function of hopD in S. Enteritidis has not been characterized, it would be of interest to determine whether hopD also impacts secretion of virulence factors in Salmonella.

Transcriptional regulators.

Several known or putative transcriptional regulators showed reduced expression in LP strains. These included ygaE, yncC, sdiA, ydcI, ecnR, SEN4085, SEN4086, yiaG, and SEN1787 (see File S2 in the supplemental material). YgaE (also known as GabC) and YncC belong to the GntR family of transcriptional regulators. This family of regulators has a conserved N-terminal DNA binding domain and a diverse C-terminal domain involved in the effector binding and/or oligomerization (78, 79). In both S. Enteritidis and E. coli, ygaE is located in the gabDTPC operon. In E. coli the products of the gab operon are involved in degradation of γ-aminobutyrate (GABA) and contribute to polyamine (putrescine, spermidine, and spermine) homeostasis during nitrogen-limited growth (80) and also maintain high internal glutamate concentrations under stress conditions (81). Expression of the gab operon is enhanced at high pH (82) and at high cell density in nitrogen-rich environments (83–85). While the exact role of ygaE in S. Enteritidis has not been investigated, the simultaneous reduced expression of two additional genes such as gabP (a GABA-specific permease) and patA (a putrescine-2-oxoglutarate aminotransferase) raises a possibility that the polyamine homeostasis in LP strains may be disregulated, potentially making these strains inherently more susceptible to environmental stressors.

Four genes with significantly reduced expression in LP strains (sdiA, ydcI, ecnR, and SEN4085) belong to LysR family of transcriptional regulators (LTTRs). Despite considerable structural and functional conservation, LTTRs are known to regulate a diverse set of genes, including conjugation, bioluminescence, metabolism, motility, and virulence genes (86, 87). For instance, sdiA is a positive activator of the ftsQAZ genes that are essential for septation, and its amplification results in diverse genetic effects in E. coli including mitomycin C resistance, downregulation of several motility- and chemotaxis-related genes, and upregulation of genes involved in DNA repair and replication (88). Similarly, when S. Typhimurium is grown in motility medium, sdiA regulates expression of virulence plasmid-associated genes (89). Interestingly, sdiA is also dually controlled by iron concentration and culture density-derived signals, and the deletion of the helix-turn-helix (HTH) domain of sdiA results in increased virulence of S. Typhimurium in a murine model of infection (90), suggesting that sdiA contributes to diverse genetic effects with possible links to motility, bacterial cell septation, response to DNA-damaging agents, and virulence. Similar to sdiA, ecnR is linked to motility because it negatively regulates flhDC transcription and affects bacterial motility; however, the role of ydcI and SEN4085 in S. Enteritidis has not been established.

Apart from LTTRs, SEN4086, which belongs to the family of AraC transcriptional regulators, showed significantly reduced expression in LP strains. AraC transcriptional regulators control diverse bacterial functions including sugar catabolism, responses to stress, and virulence (91). For example, certain well-characterized AraC transcriptional regulators such as rstA and hilD control expression of the invasion genes in S. Typhimurium (92, 93). Finally, YiaG is a putative HTH type of transcriptional regulator that binds to DNA and regulates gene expression, whereas SEN1787 is a putative LuxR-type DNA-binding HTH domain that is activated by several different mechanisms including a two-component sensory transduction system. The role of SEN4086, yiaG, and SEN1787 in the transcriptional regulation of S. Enteritidis is currently unknown although their reduced expression in all LP strains raises a possibility that these genes may play a role in the regulation of virulence or that they are stress associated.

Stress-associated and iron-regulated genes.

Among other genes with significantly reduced expression in LP strains, many genes (yeaG, yncC, rpsV, osmE, yiaG, otsA, fbaB, talA, poxB, yehY, dps, sodC, katE, bfr, wraB, yddX, ygaU, yibJ, psiF, yciF, and osmY) are regulated by the alternative sigma factor RpoS, which is not only required for survival of bacteria under starvation or other cellular stresses but also essential for Salmonella virulence (94–96). It is also suggested that Salmonella lacks an ectoine biosynthetic pathway and therefore cannot tolerate environments with high osmolarity (97). While the functions of several of these genes have not been fully characterized, a few genes such as osmY, osmC, osmE, dps, katE, talA, yciF, ygaU, yjbJ, poxB, otsB, tktB, yciE, acnA, and yhbO are known to be induced when E. coli is subjected to osmotic stress conditions in an aerobic milieu (98–100). Additionally, two genes, SEN1557 and yehZ, belong to ABC superfamily of binding proteins potentially involved in glycine betaine/choline transport, which is required for osmoprotection. Under stress conditions, if glycine betaine cannot be imported, Salmonella produces the disaccharide trehalose, a highly effective compatible solute. It has been reported that mutants defective in trehalose synthesis display impaired osmotic tolerance in minimal growth medium without glycine betaine and impaired stationary-phase-induced heat tolerance (101). Trehalose accumulation also increases bacterial resistance to stresses such as high salt, low pH, and hydrogen peroxide, conditions that mimic aspects of innate immunity (101, 102). Interestingly, treA, which is needed for trehalose utilization at high osmolarity, showed significantly reduced expression in LP strains. In addition, the genes encoding trehalose-6-phosphate synthase (otsA) and anabolic trehalose-6-phosphate phosphatase (otsB) also showed reduced expression. The expression of these genes is induced by both osmotic stress and growth into the stationary phase (84). Similarly, proP, which encodes a proline/glycine betaine transporter and protects Salmonella from inhibitory effects of high salinity, was downregulated (97). Finally, two RpoS regulated genes, rpsV and yddX, appear to be a part of a seven-gene operon (rpsV-yddX-osmC-SEN1494-SEN1497) in which all of the genes are transcribed in the same direction. Six genes in this operon showed reduced expression in LP strains (see File S2 in the supplemental material). While the function of this operon is not well characterized, osmC appears to be an acid and osmotic stress response gene (98), suggesting a possible link to osmotic stress. Overall, significantly reduced expression of several osmotically inducible genes suggests that the LP strains may have reduced tolerance to high osmolarity.

Another host defense mechanism that Salmonella encounters during infection is the production of reactive oxygen species such as hydrogen peroxide (H₂O₂) by the phagosome NADPH oxidase. Hydrogen peroxide can diffuse across bacterial membranes and damage biomolecules. Several genes that play an important role in protection of microbial pathogens against oxidative stress showed reduced expression in LP strains. These included katE, which encodes a catalase enzyme required for degradation of H₂O₂, and sodC, which encodes a periplasmic copper and zinc superoxide dismutase that protects bacteria from phagocytic oxidative burst (103). Another gene expressed at low levels, gth (encoding glutathione S-transferase [GST]), not only affords protection against oxidative stress but also plays a role in correct folding, synthesis, regulation, and degradation of enzymes and multienzyme complexes in a large number of metabolic processes (104–106).

Several genes involved in iron regulation also showed reduced expression in LP strains. For instance, the entire sufABCDSE operon, which plays a vital role in Fe-S cluster biogenesis, showed reduced expression. The iron-sulfur (Fe-S) clusters are key metal cofactors of metabolic, regulatory, and stress response proteins in most organisms, and their assembly and maturation in vivo require complex machineries. It has been reported that in E. coli, the sulfur utilization factor (SUF) system is induced under adverse stress conditions such as oxidative stress or iron deprivation (107), whereas in S. Typhimurium, chlorine-based oxidative stress induces expression of the SUF operon (108). S. Typhimurium also possesses four ferritins: bacterioferritin (Bfr), ferritin A (FtnA), ferritin B (FtnB), and DNA starvation/stationary-phase protection protein (Dps). We found that two ferritin genes (bfr and dps) were downregulated in LP strains. The heme containing Bfr is maximally expressed when Fe is abundant and accounts for the majority of stored Fe (109). While inactivation of bfr elevates the intracellular free Fe concentration and enhances susceptibility to H₂O₂ stress, the DNA-binding protein Dps provides protection from oxidative damage without affecting the steady-state intracellular free Fe concentration, and it is also known to be induced during oxidative stress in Burkholderia pseudomallei, thereby protecting this pathogen from organic hydroperoxides and aiding its intramacrophage survival (110, 111). Taken together, the reduced expression of several genes involved in resistance of Salmonella to osmotic and oxidative stresses and iron-limiting conditions correlates with the impaired survival of LP strains under oxidative and acidic stress as well as reduced intramacrophage survival (18). These results suggest that LP strains are impaired in their ability to resist host innate immune responses or propagate under the iron-limiting conditions encountered early during infection. For example, the host intestinal milieu represents an environment where Salmonella is exposed to high osmolarity, whereas intramacrophage survival of Salmonella is partly dependent on the ability of this bacterium to resist oxidative stress. Further studies involving targeted mutations of genes with possible ties to osmotic or oxidative stress and their effects on the kinetics of intestinal colonization or intramacrophage survival of S. Enteritidis may provide a better understanding of the colonization potential and persistence of these strains within the host.

Genes involved in biofilm production and egg contamination.

Biofilm formation is important for the survival of Salmonella on surfaces, for increased resistance to disinfectants, and for survival in the avian reproductive tract (112, 113). Several genes in the S. Enteritidis genome provide this bacterium a unique ability to infect reproductive organs of chickens and contaminate forming eggs (114–117). Two genes (wrbA and yshA) with possible involvement in biofilm formation (112, 118) and two genes (ygdl and SEN2997) that were previously implicated in the survival of S. Enteritidis in egg albumen (115) showed significantly reduced expression in LP strains. It is important to note that irrespective of their virulence potential, not all S. Enteritidis strains produce biofilm or are able to survive within egg albumen (18, 20). This is consistent with the relatively fewer differences in the expression of genes associated with these phenotypes.

Functionally uncharacterized genes.

Several genes (n = 79) encoding functionally uncharacterized proteins or proteins with putatively assigned functions showed significantly reduced expression in LP strains (Fig. 1). While few of these genes have been previously identified as stress regulated (see File S2 in the supplemental material), the functions of most others remain elusive. Of particular interest are genes encoding putative lipoproteins (SEN0081, yfbK, and ygdI). Bacterial lipoproteins perform various roles, including nutrient uptake, signal transduction, adhesion, conjugation, and sporulation, and participate in antibiotic resistance, transport (such as ABC transporter systems), and extracytoplasmic folding of proteins (119). In the case of pathogens, lipoproteins play a direct role in virulence-associated functions, such as colonization, invasion, evasion of host defense, and immunomodulation (119). Therefore, it is possible that the putative lipoprotein-encoding genes identified here may be important in the virulence or survival of S. Enteritidis. Interestingly, five pseudogenes (SEN0912, SEN1154, SEN1171A, SEN1171B, and SEN1809) showed reduced expression in LP strains. The role of these pseudogenes in the pathogenesis of Salmonella is not clearly understood although a transposon-mediated mutation in pseudogenes such as SEN1154 and siiE results in attenuation in the cell invasiveness of S. Enteritidis (68). In addition, other investigators have shown that several pseudogenes are transcriptionally active in S. Typhi (120), suggesting their potential misclassification as these genes in fact play a role in the pathobiology of Salmonella.

Genes involved in nutrient metabolism show increased expression in LP strains.

Seventeen genes showed significantly increased expression in LP strains (see File S3 in the supplemental material). These included all of the genes encoded on the polycistronic tdcABCDEFG operon which, under anaerobic conditions, are implicated in degradation of l-serine and l-threonine to acetate and propionate, respectively (121). The regulatory gene tdcA, which encodes a protein homologous to the LTTRs, is required for maximal induction of the tdc operon (122) and also regulates expression of several genes, including 50S ribosomal subunit proteins and a lipoprotein that links inner and outer membranes (123) and is involved in the virulence of S. Typhimurium (124, 125). Interestingly, two genes (rpsC and rplV) encoding ribosomal subunit proteins and one gene (SEN1797) encoding a lipoprotein with potential ties to tdcA expression showed increased expression in LP strains. Finally, several genes encoding enzymes of the d-glucarate (gudP, ygcY, and gudD) and d-glycerate (garDLRK) pathways showed increased expression in LP strains. Interestingly, genes of the garDLRK operon are located immediately downstream of the tdcABCDEG operon in the S. Enteritidis genome (43). Exposure to H₂ in S. Typhimurium leads to significant upregulation of the gar and gud genes and downregulation of several virulence-associated genes (126). Hydrogen can be an important energy source for bacteria growing in an environment where high-energy organic substrates are limiting (127). For instance, addition of H₂ significantly augments the growth of S. Typhimurium in a culture medium containing amino acids as the only carbon source, mainly due to the enhanced ability of the bacteria to acquire amino acids from the medium (126). Therefore, the availability of H₂ under a nutrient-limited condition such as the competitive environment within the host intestinal tract could be crucial to the survival of Salmonella. While the LP strains in this study were not grown under a nutrient-limited condition or under anaerobic conditions, the fact that LP strains show increased expression of the tdc, gar, and gud operons and reduced expression of several virulence-associated genes suggests that the primary focus of these strains could be survival and cell growth through enhanced nutrient acquisition rather than invasion and proliferation.

Conclusions.

Using comparative RNA-Seq analysis of multiple field strains of S. Enteritidis with various phenotypic and virulence properties, this study demonstrates that the LP strains have significantly impaired expression of virulence, motility, and stress-associated genes. The study design included a subset of low-pathogenicity strains of S. Enteritidis with naturally impaired motility, reduced resistance to oxidative and acidic stresses, and virulence attenuation. Although the strains were grown in synthetic culture medium within a laboratory setting, overall the transcriptional signatures of LP and HP strains grown at avian body temperature (42°C) correlated well with their phenotypic characteristics, suggesting that RNA-Seq provided a better understanding of the potential mechanisms underlying their differential phenotypic characteristics and virulence potential. The reduced expression of several genes with uncharacterized functions and of transcriptional regulators demonstrates the complexity of the regulatory network in S. Enteritidis and may be further studied using diverse conditions that better simulate in vivo infection kinetics. Differential regulation of several known and putatively identified transcriptional regulators is intriguing because these genes could be involved in virulence regulation or stress response and/or environmental persistence of this important food-borne pathogen. For instance, LTTRs represent the most abundant type of transcriptional regulator in the prokaryotic kingdom (86), and because of their abundance in diverse bacterial species and relatively conserved characteristics, detailed transcriptomic studies of strains with a defined genetic background of different LTTRs under culture conditions that mimic in vivo infection processes may facilitate identification of novel virulence-attenuated strains with a potential for use as vaccine candidates.

While field strains of S. Enteritidis with naturally impaired motility, reduced resistance to oxidative and acidic stresses, and virulence attenuation have been commonly isolated (18, 20–22), their significance to the epidemiology of S. Enteritidis in poultry and people remains elusive. It is possible that these strains may have an impaired ability to infect chickens or contaminate eggs or may have a reduced ability to persist in the poultry environment. Nevertheless, occurrence of such strains in poultry or a poultry-associated environment raises a possibility that these strains may represent a unique subpopulation that is adapted to a unique niche, leading to intermittent infection of the flock or transmission to humans with a relatively mild disease that remains largely unnoticed or undiagnosed. The whole-genome sequences of the majority of strains used in this study are not available, and therefore it is currently unknown if these strains also carry other genomic differences such as SNPs which may potentially alter some of the phenotypes or expression differences observed in this study. In the future, comprehensive phenotypic and genotypic characterization of a larger set of wild-type strains of S. Enteritidis may provide some clues to the epidemiological significance of these low-pathogenicity strains of S. Enteritidis.