Intracellular and extracellular STM and SPA exhibit distinct gene expression profiles

Although STM and SPA belong to the same species, share about 89% of their genes and express many common virulence factors [26], the disease they cause in immunocompetent humans is very different. One possible reason, expected to contribute to their different pathogenicity is a distinct expression of genes involved in host-pathogen interactions. To test this hypothesis, we compared the transcriptomic architecture and gene activity of SPA or STM in a relevant context of epithelial cell infection. We chose to sample Salmonella infection at 8 h post infection (p.i.) when the pathogen is considered to be well engaged in replication and optimally adapted to the intracellular environment [27]. Thus, HeLa cells were infected with fluorescent STM and SPA for 8 h and sorted by flow cytometry to isolate Salmonella-infected cells. Overall, RNA was extracted from six independent HeLa cell cultures infected with SPA (N = 3) and STM (N = 3). In each experiment, 2–2.5×10⁶ GFP-positive HeLa cells, (infected with either STM or SPA) were collected. In parallel, RNA was extracted from six independent extracellular STM (N = 3) and SPA (N = 3) cultures that were grown in LB to stationary phase microaerobically, prior to cells infection. Isolated RNA from these 12 samples was deep-sequenced (RNA-Seq), aligned to the corresponding Salmonella genome and compared between all conditions. Principal Component Analysis (PCA) of normalized gene read counts of these RNA-Seq data indicated a unique expression profile for each condition and good reproducibility between independent biological triplicates (Fig 1A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The number of STM and SPA genes expressed in LB medium and during intracellular growth.

(A) Principal component analysis (PCA) of RNA-Seq data. The PCA was performed by using factomineR package on the normalized expression of 3851 shared genes of three independent SPA and STM cultures grown in LB medium (M1 to M3) and of three independent HeLa cell cultures (C1 to C3) infected with STM and SPA (as described in Materials and Methods). The percentages on each axis represent the percentages of variation explained by the principal components. Points that are closer together are more similar in gene expression patterns, and the barycenter of the triplicates used for each condition is shown by a larger symbol. (B and C) Salmonella RNA-Seq reads were aligned to S. Typhimurium 14028S and S. Paratyphi A 45157 genomes and normalized by transcripts per million (TPM) transformation. Genes with TPM ≥ 10 were considered as expressed. The number of expressed genes in STM (B) and SPA (C) growing in LB medium to stationary phase under microaerobic conditions and within HeLa cells at 8 h p.i. is shown.

https://doi.org/10.1371/journal.ppat.1010425.g001

The intracellular transcriptomic landscape of S. Typhimurium

RNA extracted from three independent Fluorescence-Activated Cell Sorting (FACS) HeLa cell cultures infected with STM have generated 496–575 million RNA-Seq reads (human and bacterial reads) per experiment. Of which, 4.9–6.6 million sequence reads, assigned to non-rRNA bacterial genes were used for intracellular bacterial transcriptome analysis. The three extracellular STM cultures grown in LB to stationary phase under microaerobic conditions, have generated 4.3–13.8 million RNA-Seq informative reads that were assigned to non-rRNA bacterial genes. To analyze the STM gene expression, we calculated for each gene the number of transcripts per kilobase million (TPM) from its feature read counts. The expression level threshold was set to a TPM value of 10 or above [28]. Altogether, we identified 1,018 distinct genes that were exclusively expressed by STM during intracellular infection of HeLa cells and 191 distinct genes that were expressed only when STM was grown in LB. 3,127 transcripts were commonly expressed at both conditions (Fig 1B and S1 Table).

RNA-Seq successfully identified 365 upregulated and 465 downregulated genes that were changed by at least twofold inside epithelial cells at 8 h p.i. (adjusted p-value ≤ 0.05) by STM, relative to their expression in LB (Fig 2A and S2 Table). In agreement with previous studies [14,29,30], during STM infection, at least 27 genes belonging to the flagella-motility regulon and several genes from the type I fimbriae (fimAWYZ) were significantly repressed. Interestingly, the expression of genes encoding the Saf fimbriae (safABCD) was upregulated during HeLa cells infection by STM.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. DEGs of intracellular S. Typhimurium compared to growth in LB medium.

(A) a Volcano plot showing the fold change (log2 ratio) in the expression of STM genes grown intracellularly in HeLa cells (8 h p.i.) vs. the growth in LB medium (X-axis) plotted against the -log₁₀ adjusted p-value (Y-axis). Each dot on the plot represents the mean value (from three independent cultures) of one gene. Genes that were changed by more than twofold are colored in red. (B) Heatmap based on RNA-Seq results showing the relative transcription of STM genes of interest during intracellular infection (C1 to C3) vs. their extracellular expression in LB medium (M1 to M3).

https://doi.org/10.1371/journal.ppat.1010425.g002

In contrast to the flagella and the fim genes, at least 31 genes from the PhoPQ regulon were highly induced intracellularly (Fig 2B). The two-component system PhoPQ, orchestrates Salmonella adaptation to the intracellular milieu and regulates a wide array of genes central for Salmonella virulence including those encoded on SPI-2 [31,32]. Correspondingly, we identified upregulation of the SPI-2 regulon including ssaBCDEGHIJKLMVNOPRSTU, sseABCDEFGIJL, sseK3, sspH2, pipB, pipB2, and steC genes (Fig 2B). These results are consistent with the known notion that SPI-2 expression is induced during intracellular replication of Salmonella [15,33] and with previous reports that studied STM expression during macrophages infection [34,35]. Other PhoP-regulated genes including pgtE (involved in bacterial resistance to antimicrobial peptides), pagN (encoding an adhesion/invasion protein), phoN (acid phosphatase) and pagCDJKO were also induced (Fig 2B).

Iron is an essential micronutrient required by nearly all bacterial species, including Salmonella. Limiting the availability of the nutrient metals to intracellular pathogens is one of the main mechanisms of nutritional immunity. In return, Salmonella uses a variety of high-affinity iron uptake systems to compete with the host for essential transition metals required for its intracellular growth [36]. Congruently, we identified elevated expression of multiple iron (entABEFH and iroNBCDE), iron/ manganese (sitABCD), zinc (zntR, zitT and znuA), and magnesium (mgtAB) acquisition system genes, highlighting the essentiality of these metal ions for STM intracellular replication (Fig 2B).

More than 100 suitable carbon substrates as well as various nitrogen, phosphorus and sulfur sources are available to invading pathogens, at different niches of the vertebrate hosts [37,38]. Interestingly, multiple genes belonging to the ethanolamine operon (eutQTMGAC) and genes involved in glycerol (glpABDK), and maltose (malEKM) metabolism were particularly induced during STM infection (Fig 2B). Nevertheless, the most highly induced metabolic gene was uhpT, encoding a hexose phosphate transporter that was upregulated by more than 40-fold at 8 h p.i. These results suggest that sugar phosphates are major substrates required for cytosolic intracellular proliferation of STM within HeLa cells, but also that ethanolamine, glycerol and maltose could be utilized as alternative carbon sources in epithelial cells.

Similarly, biotin (bioABCDF), purine (nrdEFHI and purDEHKLT), and sulfate/thiosulfate (cysACHITW and sufABCDES) biosynthetic metabolic pathways were highly transcribed intracellularly by STM. Moreover, genes involved in biosynthesis of the amino acids histidine (hisACDFH and hutG), tryptophan (trpABDE), and glutamine (glnAGKL), and genes that catalyze the cleavage of N-acetylneuraminic acid (sialic acid; Neu5Ac) to form pyruvate and N-acetylmannosamine (ManNAc) (nanAEKT and nagB) were also upregulated by STM during HeLa cell infections (Fig 2B).

A large number of genes involved in phosphate metabolism were identified to be induced within HeLa cells. For example, the transcriptional levels of pstABCS (encoding an ATP-dependent phosphate uptake system), which is responsible for inorganic phosphate uptake during phosphate starvation, ugpABCEQ, phoR and apeE were markedly increased. The ugp operon is involved in hydrolysis of diesters during their transport at the cytoplasmic side of the inner membrane, generating sn-glycerol-3-phosphate (G3P) which is used as a phosphate source [39]. The upg operon is regulated by a two-component system, PhoBR, which also regulates the outer membrane esterase, ApeE [40]. Additional genes belonging to the PhoB regulon, such as phnSVWU encoding the phosphonatase pathway, which is responsible for breaking down phosphonate and yielding cellular phosphate [41] were also significantly induced.

Collectively, these results suggest that STM likely experiences iron, biotin, purine and phosphate deprivation in host epithelial cells and induces the expression of genes involved in their transport and utilization. In addition, these results imply that ethanolamine, glycerol and maltose are available sources of carbon for STM in epithelial cells.

The intracellular transcriptomic landscape of S. Paratyphi A

RNA-Seq analysis was similarly applied for intracellular SPA. 313.8 to 805.9 million RNA-Seq reads were obtained from three independent FACS-sorted HeLa cell cultures infected with SPA. Out of these, 14.6 to 36.6 million sequence reads were assigned to non-rRNA SPA genes and used to determine the intracellular SPA transcriptome (S3 Table). As a reference control, RNA extracted from extracellular SPA cultures grown in LB, generated 7.3 to 23 million informative reads (assigned to non-rRNA SPA genes). Overall, we identified 322 genes that were exclusively expressed (TPM ≥10) by SPA during infection of HeLa cells compared to 576 genes that were expressed only by extracellular SPA in LB culture. 2,995 SPA genes were expressed at both conditions (Fig 1C and S3 Table).

At 8 h p.i. of HeLa cells, 322 SPA genes were upregulated and 579 genes were down-regulated (Fig 3A and S4 Table) by at least twofold (adjusted p-value ≤ 0.05). Like in intracellular STM, the genes for type I (Fim) fimbriae (fimCIWYZ) and additional fimbria clusters including Bcf (bcfDE), the Curli (csgBCDE), Stb (stbBE), Stc (stcABC), Stf (stfAD), Sth (sthABCD), and Stk (stkABC) were all strongly downregulated intracellularly (Fig 3B). Noteworthy, the downregulation of type I (fimACIZ), Curli (csgBDE) and Bcf (STM14_0037 and bcfB) fimbriae was more pronounced in SPA than in STM (S1 Fig), indicating that the SPA fimbriome is robustly repressed during intracellular replication, while some basal level of expression of these fimbrial genes is still maintained during STM infection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. DEGs of intracellular S. Paratyphi A compared to growth in LB medium.

(A) a volcano plot showing the fold change (log2 ratio) in the expression of SPA genes grown intracellularly in HeLa cells (8 h p.i.) vs. the growth in LB medium (X-axis) plotted against the -log₁₀ adjusted p-value (Y-axis). Each dot on the plot represents the mean value (from three independent cultures) of one gene. Genes that were changed by more than twofold are colored in red. (B) Heatmap based on RNA-Seq results showing the relative transcription of SPA genes of interest during intracellular infection (C1 to C3) vs. their extracellular expression in LB medium (M1 to M3).

https://doi.org/10.1371/journal.ppat.1010425.g003

In contrast, but consistent with our findings for intracellular STM, genes belonging to the PhoPQ (including polymyxins resistance) and SPI-2 regulons were significantly induced by intracellular SPA. Moreover, various transition metals import systems including manganese and iron transporter (sitABCD), enterobactin biosynthesis system (entABCDEF and fepAG) and the iron ABC transporter IroC were considerably upregulated during HeLa cell infection by SPA (Fig 3B).

In addition, the expression of several metabolic pathways that were significantly elevated during STM infection were induced by intracellular SPA as well, including biotin (bioABCDF) and purine (purCDEHKLT) biosynthesis genes. Further overlap in the intracellular transcriptome of these pathogens includes several amino acid biosynthesis pathways such as glutamine (glnAGKLPQ), tryptophan (trpABCDE) and histidine (hisAFGHI) that were upregulated in both intracellular STM and SPA.

Nevertheless, comparison of the intracellular transcriptome between STM and SPA, identified 234 differentially expressed genes (DEGs; S5 Table), indicating that some key pathways are distinctly expressed during the intracellular phase of STM vs. SPA. Such differences included expression of several amino acids biosynthesis genes, which was more pronounced in intracellular SPA than in intracellular STM. These include biosynthesis genes of glycine (glyASU), isoleucine (ilvCDEG), leucine (leuABCD), lysine (lysAC and sucBD), methionine (metCEFHKLR), serine (serAC), and threonine (thrBC). In contrast, the threonine degradation pathway (tdcABCDEG) was found to be repressed in intracellular SPA. These results may suggest increased protein synthesis in SPA vs. STM during epithelial cell infection. Additional pathways that were upregulated in SPA, but not in STM are the oxidative phosphorylation (nuoEFGHIJKLMN and atpCD), and peptidoglycan cell wall formation (murACDEG) pathway genes (Fig 3B). Remarkably, further differences between the intracellular transcriptome of SPA vs. STM included different expression profiles of the SPI-1, motility and chemotaxis, and carbon utilization regulons as detailed in the following sections.

Differences in carbon catabolism during intracellular infection of S. Paratyphi A and S. Typhimurium

In order to survive, persist and replicate in host cells, intracellular pathogens must adapt their metabolic pathways to the specific physical conditions and available nutrients found in the intracellular environment. Carbon catabolism provides bacteria with energy by means of reducing equivalents, ATP and essential biosynthetic precursors. Different studies have suggested that hexose monosaccharides such as glucose and glucose-6P are the main source of carbon during Salmonella infection [42–46]. Comparison of the metabolic gene expression between intracellular STM and SPA revealed significant differences in the expression profile of several carbon catabolic pathways. For example, the gene uhpT encoding an antiporter for external hexose 6-phosphate and internal inorganic phosphate [47], presented about 90-fold higher expression by intracellular STM than SPA (Fig 4A). This different expression might be because the sensor-regulator system encoded by uhpABC is defective by pseudogene formation in SPA [48]. Similarly, the gene encoding for galactonate permeases (dgoT) was upregulated in intracellular STM, but not in SPA, suggesting that STM utilizes more hexose 6-phosphate and galactonate than SPA as a carbon source during intracellular growth.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Differences in carbon utilization gene expression between STM and SPA.

(A) Heatmap of carbon catabolism gene expression by intracellular STM vs. intracellular SPA. Pseudogenes in the SPA genome are marked with an asterisk. (B) The fold change in the transcription of the glucose-6P (uhpT), citrate (citA and citX), and ethanolamine (eutP, eutQ and eutG) utilization pathways in intracellular STM relative to their expression in intracellular SPA was determined using qRT-PCR. RNA was extracted from FACS-sorted intracellular salmonellae 8 h p.i. The mean and the standard error of the mean (SEM) of 3–8 independent qRT-PCR reactions are shown. Student’s t-test was used to determine statistical significance. **, p < 0.01; ***, p < 0.001; ND, not detected. (C) Growth of STM and SPA in M9 in the presence (+) or absence (-) of casamino acids (AA) and ethanolamine (ETA), under anaerobic conditions. (D) Growth of STM and SPA in M9 supplemented with citrate as a sole carbon source.

https://doi.org/10.1371/journal.ppat.1010425.g004

Similarly, RNA-Seq identified that at least seven genes involved in citrate metabolism including citCDEFTX and citC2 exhibited higher transcription levels in intracellular STM than SPA (Fig 4A). Independent qRT-PCR further confirmed that the intracellular expression of citA and citX encoding citrate-proton symporter and apo-citrate lyase phosphoribosyl-dephospho-CoA transferase, respectively is lower by about 3- and 10-fold in intracellular SPA vs. STM (Fig 4B).

STM can utilize ethanolamine as a sole source of carbon, nitrogen, and energy in a cobalamin (vitamin B12)-dependent manner. Ethanolamine may be important to STM growth in the host, since it is derived from the membrane phospholipid phosphatidylethanolamine that is particularly prevalent in the gastrointestinal tract, in a process involving breakdown of the precursor molecule phosphatidylethanolamine by phosphodiesterases to glycerol and ethanolamine [49]. In STM, ethanolamine catabolism involves 17 genes organized in the eut operon, controlled by the transcriptional activator EutR [50]. At least six ethanolamine catabolism genes (eutSPQTGH) were expressed at higher levels by intracellular STM compared to SPA (Fig 4A). These differences were independently verified by qRT-PCR that showed significantly reduced expression of eutG, eutP and eutQ genes in intracellular SPA vs. STM (Fig 4B).

Collectively, these observations suggest that STM is able to better utilize intracellular citrate, ethanolamine and galactonate as intracellular carbon sources than SPA, and that SPA may use an alternative carbon source such as amino acids during intracellular growth.

To further test this hypothesis, we compared the growth of SPA and STM in a defined M9 medium supplemented with ethanolamine or citrate as a carbon source. Growth under anaerobic conditions in M9 medium supplemented with ethanolamine, without amino acids, did not allow bacterial growth of neither STM nor SPA. Adding casamino acids, without additional carbon source, resulted in a minimal and comparable growth of both STM and SPA, likely due to the utilization of amino acids as a restricted carbon source. Nevertheless, adding ethanolamine as a bona fide carbon source to the medium resulted in increased and rapid growth of STM, but only a minor improvement of SPA growth (relative to its growth on amino acids only; Fig 4C), indicating the ability of STM, but not SPA, to efficiently ferment ethanolamine. Replacing the carbon source with citrate under aerobic growth conditions allowed delayed, but substantial growth of STM, in comparison to restricted and slower growth of SPA (Fig 4D). These results are concurring with the transcriptomic data and indicate a better ability of STM than SPA to utilize both ethanolamine and citrate as a sole carbon source. These differences may also contribute to a higher intracellular replication pace of STM than SPA found in HeLa cells (S2 Fig).

Intracellular S. Paratyphi A expresses SPI-1 genes at higher levels than intracellular S. Typhimurium

Previously we showed that in LB medium, under aerobic growth conditions to the late logarithmic phase, SPI-1 gene expression, as well as secretion of SPI-1-T3SS effector proteins, occur at significantly lower levels in SPA compared to STM [23]. Here, in sharp contrast, we found that intracellular SPA transcribed significantly higher levels of SPI-1 and T3SS-1-effector genes than intracellular STM (Fig 5A). This was clearly evident (log 2 change ≥ 2 and p-value ≤ 0.05) for at least 14 SPI-1 genes including the SPI-1 regulators hilA and hilD, the effector genes sopBDF, the T3SS-1 structural genes invCHIJ and spaOPQ, sigE encoding a chaperon (a.k.a. pipC) and iagB (S5 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Intracellular SPA expresses higher levels of SPI-1 genes than intracellular STM.

(A) Heat map of RNA-Seq results showing the relative transcription of SPI-1 genes in three independent HeLa cell infections with STM and SPA. (B) The fold change in the transcription of six SPI-1 genes (hilA, hilD, invH, sopB, sopD and spaO) in intracellular SPA relative to their expression in intracellular STM was determined by qRT-PCR. RNA was extracted from FACS-sorted intracellular salmonellae 8 h p.i. The results show the mean of 3–6 qRT-PCR reactions and the error bars indicate the SEM. (C) SPA and STM strains expressing GFP and a 2HA-tagged version of the SPI-1 effectors SopE2 and SopB were used to infect HeLa cells. At 8 h p.i., cells were fixed and FACS-sorted for infected (GFP-positive cells). Equal number of infected HeLa cells (1×10⁶ for each sample) were collected and separated by 12% SDS-PAGE. Western blotting using anti-hemagglutinin (HA) tag antibodies was used to detect the intracellular expression of SopE2 and SopB. Anti-DnaK antibodies were used as a loading control.

https://doi.org/10.1371/journal.ppat.1010425.g005

Independent qRT-PCR was used to verify these observations and demonstrated that intracellular SPA transcribed hilA, hilD, invH, sopB, sopD and spaO at 4- to 27-fold higher levels than STM (Fig 5B). Moreover, Western blotting against a 2HA-tagged SopB and SopE2 further indicated elevated translation of these two SPI-1 effectors in intracellular SPA vs. STM (Fig 5C). Collectively, these results indicate that SPA expresses SPI-1 genes at higher levels during infection of non-phagocytic cells, in comparison to intracellular STM.

Intracellular S. Paratyphi A is motile and expresses flagella-chemotaxis genes at higher levels than S. Typhimurium

Functional flagella are necessary for Salmonella motility and invasion of host cells [24,51,52]. In S. enterica, more than 50 genes are involved in the assembly and motion of flagella, organized in one regulon that composes three ordered regulatory phases corresponding to classes I (early), II (middle), and III (late) [53].

Previous studies have shown that the genes coding for the flagellar machinery, as well as many genes involved in chemotaxis, are strongly repressed during STM intracellular lifestyle [30,34,35]. Here, RNA-Seq analysis has demonstrated that at least 38 out of 50 chemotaxis and flagella genes, belonging to class II and III of the flagella regulon are expressed at significantly higher levels in intracellular SPA compared to STM (Fig 6A and S5 Table). These include the flagellin gene fliC that was expressed at 11-fold higher levels by SPA vs. STM, and many other flagellar genes including flgC that demonstrated 64-fold difference in its expression between intracellular SPA and STM. In addition to the flagellar genes, SPA residing in the host cell presented much higher expression levels of chemotaxis genes (e.g. cheB, cheR, cheM, cheW, cheZ, and cheY) than intracellular STM.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Intracellular SPA expresses higher levels of motility-chemotaxis genes than intracellular STM.

(A) Heat map of RNA-Seq results showing the relative transcription of the motility-chemotaxis regulon in three independent HeLa cells infections with STM and SPA. (B) The fold change in the transcription of six flagellum genes (fliC, fliZ, glgM, flgF, cheA and cheB) in intracellular SPA relative to their expression in intracellular STM was analyzed by qRT-PCR. RNA was extracted from FACS-sorted intracellular salmonellae 8 h p.i. The results show the mean of 3–5 independent reactions and the error bars indicate the SEM. (C) SPA and STM strains expressing GFP were used to infect HeLa cells. At 8 h p.i., cells were fixed and FACS-sorted. Equal numbers of 1×10⁶ infected HeLa cells were collected, lysed, and proteins were separated on 12% SDS-PAGE. Western blotting using anti-FliC antibodies were used to detect synthesis of FliC by intracellular Salmonella. Anti-DnaK antibodies were used as a loading control.

https://doi.org/10.1371/journal.ppat.1010425.g006

Independent qRT-PCR analysis confirmed these results and showed that in sorted infected HeLa cells, the expression of fliC, fliZ, flgM, flgF, cheA and cheB was 8- to 130-fold higher in intracellular SPA than in intracellular STM (Fig 6B). In agreement with the RNA-Seq and qRT-PCR analyses, Western blotting using anti-FliC antibodies demonstrated significantly higher translation of FliC in intracellular SPA vs. STM (Fig 6C). Taken together, these findings uncover significantly higher expression levels of motility and chemotaxis genes by SPA vs. STM within infected HeLa cells.

To further analyze if the increased expression of motility genes by intracellular SPA has physiological consequences, we investigated the possible presence of flagella in intracellular SPA. Immunostaining of intracellular SPA in either HeLa (Fig 7A) or Caco-2 cells (Fig 7B) at 8 or 4 h p.i., respectively, indicated the presence of flagella filaments on intracellular SPA. Flagella staining was absent for a ΔfliC strain of SPA, and restored in the plasmid-complemented SPA ΔfliC. No immunolabeling of flagella was observed for intracellular STM WT or STM ΔfliC ΔfljB (Fig 7A), indicating that this phenotype is SPA-specific.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Formation of flagella by SPA within the host cell cytoplasm.

(A) HeLa cells stably expressing LAMP1-GFP (green) were infected with SPA WT, SPA fliC, SPA fliC [fliC], STM WT, or STM fliC fljB as indicated, each constitutively expressing mCherry (red). Infected cells were fixed 8 h p.i. and subsequently immunolabeled with serovar-specific antibody against FliC (blue). (B) Caco-2 well were infected with SPA WT constitutively expressing mCherry (red). Cells were fixed 4 h p.i. and immunolabeled for FliC (blue). Scale bars: 10 μm.

https://doi.org/10.1371/journal.ppat.1010425.g007

Differences in intracellular compartmentation and SIF-formation between STM and SPA

Phenotypic comparison by live-cell imaging between host cells infected with STM or SPA was performed. To distinguish Salmonella in the SCV from Salmonella in other compartments or in the cytosol, and to enable live-cell imaging of infected host cells, we used LAMP1-GFP expressing HeLa host cells (Fig 8, Movies 1, 4). Lifeact-GFP expressing HeLa cells were used to investigate potential interactions with the host cell actin cytoskeleton (Movie 1–3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Distinct phenotypes of intracellular SPA.

HeLa cells stably transfected for expression of LAMP1-GFP (green) were infected with STM WT (A, MOI 25) or SPA WT (B-E, MOI 75), each harboring pFPV-mCherry for constitutive expression of mCherry (red). Live-cell imaging was performed at 8 h p.i. Images are shown as maximum intensity projection. Arrowheads indicate subpopulations with distinct phenotypes in infected cells: SIF (green), thin SIF (green with asterisk), Salmonella residing in SCV (yellow), cytosolic Salmonella (red), hyper-replicating cytosolic Salmonella (red with asterisk). Scale bars: 10 μm (overview), 5 μm (inset).

https://doi.org/10.1371/journal.ppat.1010425.g008

We quantified the intracellular phenotype of SPA in comparison to STM after infection of HeLa LAMP1-GFP cells using live-cell imaging. Representative recordings of live-cell imaging sessions are shown in Movies 5 and 6 for SPA WT and STM WT, respectively. Quantification revealed that more than 80% of STM WT were present in SCV with associated Salmonella-induced filaments (SIFs; Figs 8A and 9). STM WT was observed residing in SCV and the majority of infected host cells showed formation of extensive SIF networks. The STM ssaV strain was typically present in SCV, but tubular endosomal filaments were almost absent. A smaller proportion of host cells (ca. 10%) harbored cytosolic STM, and for these cells, massive cytosolic proliferation or hyper-replication was evident (Fig 8A). For intracellular SPA WT, more heterogeneous phenotypes were observed (Fig 8B–8E). SPA was located in the SCV, or present in the cytosol without association with LAMP1-positive membranes. Intracellular SPA also induced SIF in a T3SS-2-dependent manner, but the frequency of SIF-positive cells was much lower with ca. 9% for SPA WT compared to more than 80% for STM WT (Figs 8B, 8C, and 9). SPA-infected host cells often contained thin SIF, that are likely composed of single-membrane tubules, in contrast to SIF with more intense LAMP1-GFP signal, indicating double-membrane tubules (Fig 8B and 8C). For SPA WT, SIF formation was significantly less frequent, yet dependent on T3SS-2. The frequency of SPA associated with SCV marker LAMP1 was highly reduced compared to STM. As intracellular SPA was observed to reside in tightly enclosing SCV [54], we considered flagella expression by vacuolar SPA less likely than by cytosolic bacteria. The transcriptional analyses showing expression of flagellar genes by intracellular SPA, detection of flagella filaments on intracellular SPA, and cytosolic SPA in a major proportion of infected HeLa cells (~ 40%) all pointed to functional role of flagella expression. Thus, we next investigated potential function of flagella expressed by intracellular SPA by live-cell imaging of Salmonella-infected host cells (Movies 1–4). Indeed, we observed actively motile intracellular SPA in about 40% of the infected host cells (Fig 9).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Quantification of intracellular phenotypes.

HeLa cells stably expressing LAMP1-GFP (green) were infected with Salmonella strains expressing mCherry (red). Quantification of intracellular phenotypes was performed with a Zeiss Cell Observer with 63x objective. From 6–8 h p.i. at least 100 infected cells per strain (for SPA fliC at least 50) were examined for intracellular phenotype. Data represent the means with standard deviations (SD) for three biological replicates for each strain and condition. Statistical analysis was performed with One-way ANOVA and is indicated as n.s., not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001.

https://doi.org/10.1371/journal.ppat.1010425.g009

As motility in intracellular microbes is typically mediated by polymerization of host cell actin by microbial surface proteins [55], we investigated F-actin dynamics in SPA-infected HeLa Lifeact-GFP cells (Movies 2 and 3). No association of F-actin with static or motile intracellular SPA was evident, and F-actin trails as observed for intracellular motile Listeria monocytogenes [56] were absent in SPA infected cells. No cytosolic motility was observed for intracellular STM. To control if cytosolic motility is linked to the observed flagella expression, we used a SPA fliC mutant strain. This non-motile mutant completely lacked cytosolic motility, but showed increased presence in SCV (Fig 9, Movie 1). Cytosolic motility of SPA was not affected in the T3SS-deficient ΔssaR strain (Fig 9, Movie 1). Live-cell imaging at high frame rates (Movies 2 and 3) demonstrated the dynamics of intracellular motility of SPA in real time, with characteristic alternations between straight swimming and stops with tumbling. Analyses of the vectors of individual SPA indicated collisions with the host cell plasma membrane, but there were no indications for formations of protrusions, showing no evidence for lateral cell-to-cell movement as observed for intracellular motile L. monocytogenes [56].

We next set out to analyze the broader relevance of intracellular motility of SPA and investigated a set of four additional clinical SPA isolates [25], and two SPA reference strains of SARB collection [57]. We observed that all seven SPA strains tested showed intracellular motility in HeLa cells (Movie 4). Expression of flagella genes and presence of flagella filaments was previously demonstrated for STM in dead host cells extruded from polarized epithelial monolayers [58]. As these experiments were performed with STM strain SL1344, we also analyzed this strain in our experimental setting. As for STM ATCC 14028, also STM SL1344 showed cytosolic hyper-replication in infected HeLa cells, but cytosolic motility was absent (Movie 4).

To investigate the intracellular motility of SPA in other host cell types, infection of human colonic epithelial cell line Caco-2 and human monocyte cell line U937 by SPA WT was analyzed. We observed intracellular motility of SPA WT in Caco-2 (Movie 7), a result in line with the detection of flagella (Fig 7B). Motile SPA WT were also detected in U937 (Movie 8). Due to the localized distribution of motile SPA in U937 and the frequent association with endosomal cargo Dextran-Alexa647, we conclude that motility of SPA in U937 macrophages occurs within spacious phagosomes rather than in the cytosol of host cells.

These data confirm that the expression of flagella by intracellular SPA and its localization outside of the SCV leads to flagella-mediated motility, and that these phenotypes are conserved traits specific to SPA and distinct from the cellular phenotypes characterizing intracellular STM.

Intracellular SPA are primed for reinfection of naïve host cells

We set out to investigate potential functional consequences of the increased expression of SPI-1 and motility genes by intracellular SPA. We hypothesized that increased expression of SPI-1, higher amounts of T3SS-1 effector proteins, and motility could affect the interaction of SPA released from infected cells with naïve host cells. As host cell decay and release of the intracellular Salmonella is asynchronous, and quantification of new infection events is difficult in gentamicin protection assays, we set up a reinvasion assay that provides experimental control. HeLa cells were infected by invasive Salmonella, at a defined time point of 8 h p.i., host cells were osmotically lysed, and released bacteria were used to reinfect new epithelial cells. The number of viable bacteria in inoculum, released from initially infected host cells, and the number of bacteria in the second round of invasion were quantified (Fig 10A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. SPA is more invasive after release from infected host cells.

(A) HeLa cells were infected with STM WT or SPA WT at MOI 5, and invasion by each strain was determined by gentamicin protection assays. At 8 h p.i., one portion of infected cells was lysed by hypotonic conditions, and the released bacteria were used to reinfect naive HeLa cells. (B) Intracellular CFU counts were determined 1 h p.i. after initial infection (invasion, black bars), or 1 h p.i. after infection with Salmonella released from host cells at 8 h p.i. (reinvasion, grey bars). The percentage of Salmonella in inoculum able to invade host cells was calculated. (C) Multiplicity of infection (MOI) calculated for invasion and reinvasion experiments shown in (B). For invasion, MOI was adjusted based on cell density of the inoculum and controlled by CFU determination, while MOI for reinvasion was calculated based on CFU determination of lysates of infected cells. Depicted are means and SD of three biological replicates, each performed in triplicate. For statistical analysis, Student’s t-test was performed. Significance is indicated as n.s., not significant; *, p < 0.05; **, p < 0.01.

https://doi.org/10.1371/journal.ppat.1010425.g010

For this reinvasion analysis we have used SPA culture that was grown to stationary phase microaerobically and STM culture that was grown to the late logarithmic phase under aerobic conditions. These growth conditions have resulted in a similar initial infection rate for SPA and STM (3.25% and 3.41% of inoculum, respectively). Interestingly, reinfection assays indicated that 6.88% of the SPA released from HeLa cells at 8 h p.i. were capable of reinvasion to naïve HeLa cells. In contrast, only 0.91% of released STM were able to reinvade, and reinvasion of STM after microaerobic culture was lower with 0.75% (Fig 10B). These data indicate that higher expression of SPI-1 genes and bacterial motility by intracellular SPA results in increased invasiveness of SPA following exit from host cells.

Bacterial strains, growth conditions and data sets

Bacterial strains utilized in this study are listed in S6 Table. S. Typhimurium 14028 S [92] and S. Paratyphi A strain 45157, an epidemic strain responsible for a paratyphoid outbreak in Nepal [9] were used as the wild-type strains. Bacterial cultures were routinely maintained in Luria-Bertani (LB; BD Difco) liquid medium at 37°C. An M9 medium [33.7 mM Na₂HPO₄, 22 mM KH₂PO₄, 8.5 mM NaCl, 18.7 mM NH₄Cl, 12.1% Trizma base, pH = 7.4] supplemented with 2 mM MgSO₄, 0.3 mM CaCl₂, 3.7 μM thiamine, 4.1 μM biotin, 0.134 mM EDTA, 31 μM FeCl₃, 6.2 μM ZnCl₂, 7.6 nM CuCl₂, 4.2 nM CoCl₂, 1.62 μM H₃BO₃, 8.1 nM MnCl₂, with or without 0.1% Casamino acids, and with one of the following carbon sources glucose (11 mM), ethanolamine hydrochloride (51 mM) or sodium citrate (34 mM) was used as a minimal growth medium.

For HeLa cell infections, Salmonella cultures were grown to the stationary phase under microaerobic conditions (see below). The final dataset for RNA-Seq included 12 samples resulting from three independent cultures in LB or infection experiments that were used for RNA extraction: (i) S. Typhimurium 14028 S bacteria grown in LB medium (hereafter named STM_m) (ii) S. Paratyphi A strain 45157 grown in LB medium (hereafter named SPA_m) (iii) Typhimurium 14028 S bacteria-infected HeLa cells (hereafter named STM_i) and (iv) S. Paratyphi A strain 45157 bacterium in HeLa cells (hereafter named SPA_i).

Infection of HeLa cells

Human epithelial HeLa (ATCC CCL-2) cells were purchased from the American Type Culture Collection and were cultured in a high-glucose (4.5 g/l) DMEM supplemented with 10% FBS, 1 mM pyruvate and 2 mM L-glutamine at 37°C in a humidified atmosphere with 5% CO₂. Cells were seeded at 6×10⁶ cells/ml in a onewell plate (Greiner) tissue culture dish 18 h prior to bacterial infection and infected at multiplicity of infection (MOI) of ~200 (bacteria per cell). For RNA extraction, five plates for S. Typhimurium and six plates for S. Paratyphi A were used. Salmonella strains carrying the pBR-GFP2 plasmid were grown in 2 ml LB containing 20 μg/ml tetracycline at 37°C with shaking (250 rpm) for 6–8 h, and then diluted 1:75 into 10 ml LB supplemented with 20 μg/ml tetracycline and grown without shaking at 37°C for 16 h. Bacterial cultures were centrifuged for 5 min at 4,629 × g, at room temperature (RT), and resuspended in 3 ml prewarmed complete DMEM. 3 ml of DMEM suspended salmonellae (~2.5×10⁹ CFU) were used to replace the medium of seeded cells and plates were centrifuged for 5 min at 142 × g at RT. After centrifugation, 7 ml of prewarmed DMEM were added and plates were incubated at 37°C under 5% CO₂ atmosphere. At 2 h p.i. the medium was aspirated, cells were washed three times with 10 ml PBS containing CaCl₂ and MgCl₂ (PBS +/+), and incubated with 10 ml prewarmed DMEM containing 100 μg/ml gentamicin. After 60 min, the media was replaced with DMEM containing 10 μg/ml gentamicin and cells were further incubated as before. At 8 h p.i., the cells were washed three times with 10 ml PBS +/+ and trypsinized with 2 ml prewarmed trypsin at 37°C. When the cells were detached, prewarmed DMEM was added to quench the trypsin. The cells were transferred to a 50 ml conical tube, pelleted by 5 min centrifugation at 247 x g and washed in 15 ml PBS+/+. Finally, the cells were resuspended in 1 ml / plate fixation reagent (RNAlater: RNAprotect 350:1), incubated at RT for 20 min, and then stored at 4°C until sorting.

HeLa cell lines stably transfected for expression of LAMP1-GFP or Lifeact-GFP have been described before [93] and were maintained and infected as described above.

Cell sorting and RNA extraction

RNA was extracted from three independent infection experiments of HeLa cells with STM strain ATCC 14028S or SPA 45157. As controls, RNA was also extracted from three independent cultures of uninfected cells and from bacterial cultures grown in LB to stationary phase under microaerobic conditions. HeLa cells (infected or uninfected) were fixed, centrifuged at 247 × g for 5 min and resuspended in 3 ml PBS+/+. The cells were filtered through 30 μm pre-separation filter into 5 ml FACS tubes (SARSTEDT) and sorted on a BD-FACSAria IIu (100 μm nozzle; ND filter 2) at a rate of approximately 300 cells/sec. 2–2.5×10⁶ GFP-positive (infected) HeLa cells were collected and centrifuged at 247 × g for 5 min. RNA extraction was performed using Hybrid-R RNA Purification System (GeneAll). RNA samples were treated with RNase-free DNase (Qiagen) according to manufacturer’s protocol, followed by ethanol precipitation and Ribosomal RNA depletion that was performed by Epidemiology Ribo-Zero Gold rRNA Removal Kit (Illumina).

RNA sequencing

RNAseq was performed at the GeT-PlaGe core facility, INRAE Toulouse. RNA-Seq libraries were prepared using the Illumina TruSeq Stranded mRNA Library Prep, without a poly-A mRNA selection, according to the manufacturer’s protocol. Briefly, RNAs were fragmented to generate double-stranded cDNA and adaptors were ligated. 11 cycles of PCR were applied to amplify the libraries and their quality was assessed using a Fragment Analyser. Libraries were quantified by qPCR using the Kapa Library Quantification Kit (Roche). To reach 5 million bacterial reads for each sample, as recommended in [94], RNA-Seq reads were generated from three lanes of an Illumina HiSeq3000 and two lanes of an Illumina NOVAseq6000 using a paired-end read length of 2×150 bp. Ribosomal RNA (rRNA) reads were filtered out using sortmerna-2.1b [95] against Silva and Rfam databases. The raw reads data for this study have been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB46495 and as specified in S7 Table.

Genome annotation and quality control

A flowchart summarizing the RNA-seq pipeline and bioinformatic analyses is shown in S3 Fig. Annotation of SPA 45157 and STM 14028S genomes was performed using EugenePP 1.2 [96] with general and targeted evidence to predict coding genes and noncoding features (S1 Fig). Functional prediction was produced with InterProscan v5.15–54.0 with GO terms, COG version COG2014 [97], REBASE v708 [98], IslandViewer4 [99] and PHAST [100].

Structural and regulatory non-coding RNA were annotated using RNAspace software v1.2.1 [101] together with the following tools: RNAmmer 1.2 for rRNA [102], tRNAScan-SE 1.23 for tRNA [103] and blastn 2.2.25 against RFAM 10.0 for other ncRNA [104]. The gff3 files produced by Eugene and RNA annotation were merged by a semi-manually procedure using BEDTools-2.26.0 [105].

FastQC v0.11.7 was used for quality control of the raw sequence data. Cutadapt 1.8.3 was used to remove Illumina Truseq adaptors and to filter out all reads with length shorter than 36 nucleotides.

Reads mapping and counting

The remaining, non rRNA reads were mapped with STAR-2.6.0c [106], with default parameters, against a “hybrid” genome composed by the Ensembl GRCh38 human primary assembly (release 92) and the SPA (accession number CP076727) or STM (accession number CP001362—CP001363) genomes. In addition, in order to analyze bacterial samples, we indexed each bacterial genome separately by decreasing the parameter “genomeSAindexNbases” to 10 for STM and to 9 for SPA, while producing the alignment (.bam) files sorted by coordinates. Samtools flagstat v1.8 was used to obtain mapping statistics. Then we added read groups to trace the origin of each read with AddOrReplaceReadGroups of Picard tools v2.18.2. Next, we merged the bam files that originated from resequencing of the same sample using the MergeSamFiles command of Picard tools. The featureCounts command from the Subread-1.6.0 tool [107] was used to compute fractional counting for all annotated genes.

Clustering of orthologous genes

Orthologous genes were computed using Roary v.3.11.2 [108] with default parameters (minimum of 95% identity between coding sequences of the same cluster) except for the “-s” option that avoids paralog splitting in different clusters. Based on the orthologous groups, we produced raw count matrices that compare the expression level between orthologous genes of SPA and STM. We also added groups including paralog genes on either SPA or STM genome into the raw count matrix to sum up the raw reads count of each putative paralog gene.

Differential gene expression analysis

Normalization and differential analysis were carried out according to the DESeq2 model and package [109] using adjusted p-value with FDR correction and the Wald test to infer the probability value. Genes that had too few reads (the sum of reads for all the samples was less than 10) were discarded. To compare DEGs between SPA and STM, we created a count matrix composed of the shared protein-coding and ncRNA genes between these two serovars. The heatmaps were created by applying DESeq2 variance stabilizing transformations (the vst command) on the counts followed by subtracting the mean of the gene across all samples. The heatmaps were drawn using the ComplexHeatmap R package [110].

Reverse transcription and quantitative real-time PCR (qRT-PCR)

RNA was extracted from infected and FACS-sorted HeLa cells, as detailed above. 500 ng of RNA were reverse-transcribed using qScript cDNA synthesis kit (Quanta-bio). Real-time PCR was performed as previously described [23] on a StepOnePlus Real-Time PCR System (Applied Biosystems). Relative quantity of transcripts was calculated as 2^-ΔΔCt, using the rpoD gene as an endogenous normalization control.

Western blotting

HeLa cells were infected with Salmonella strains carrying pBR-GFP2 plasmid and pWSK29 carrying HA-tagged alleles of sopE2 or sopB. The infection procedure was performed as above with Salmonella cultures that were grown overnight under microaerobic conditions. Cell fixation was performed by incubation with 5 ml of 4% formaldehyde for 15 min under gentle shaking. Five ml of PBS +/+ were added and the cells were centrifuged for 5 min at 2,700 × g. Subsequently, the cells were washed three times with PBS +/+ and finally resuspended in PBS +/+ containing 1% BSA. The cells were stored at 4°C until sorting. 1×10⁶ GFP-positive cells were centrifuged for 5 min at 247 × g, and resuspended in 100 μl sample buffer. Western blot of the infected cells was performed as previously described [111] using the anti-2HA (Abcam ab18181; 1:1,000), anti-FliC (Abcam ab93713; 1:200,000) and anti-DnaK (Abcam ab69617; 1:10,000) antibodies.

Host cell infection and live-cell imaging

HeLa cells stably transfected with LAMP1-GFP were seeded in surface-treated 8-well slides (ibidi) 24 or 48 h prior to infection to reach ~80% confluency (~80,000 cells) and were then used for infection. STM strains for infection experiments were subcultured from an overnight culture (1:31) in fresh LB medium and grown for 3.5 h at 37°C. For infection, SPA strains were grown for 8 h under aerobic conditions, subcultured (1:100) in fresh LB medium and stationary-phase subcultures were grown for 16 h under microaerophilic conditions as described in [24]. Bacteria were adjusted to an optical density of 0.2 at 600 nm in PBS and used for infection with the respective MOI (between 25 and 100, dependent on strain). Bacteria were centrifuged onto the cells for 5 min at 500 × g to synchronize infection and the infection was allowed to proceed for 25 min. After washing thrice with PBS, cells were incubated in medium containing 100 μg × ml^-1 of gentamicin to kill extracellular bacteria. Afterwards, the cells were maintained in medium supplemented with 10 μg × ml^-1 gentamicin. Live-cell imaging (LCI) was performed with Cell Observer microscope (Zeiss) equipped with a Yokogawa Spinning Disc Unit CSU-X1a5000, an incubation chamber, 63× objective (α-Plan-Apochromat, NA 1.4), two ORCA Flash 4.0 V3 cameras (Hamamatsu) and appropriate filters for the respective fluorescence proteins. For overview sequences, HeLa LAMP1-GFP cells in 8-well slides were infected at MOI 30 as described above. LCI was performed at 6 to 8 h p.i. as described above using a 40x objective (Plan-Apochromat, NA 1.4). A tile area of 641 x 664 x 2 μm was imaged for 180 s and stitched.

Caco-2 cells were seeded in 8-well slides 48 h prior to infection to reach ~80% confluency (ca. 70,000 cells) and then infected. If indicated, cells were pulse-chased with Dextran Alexa Fluor 488/568/647 (Thermo Fisher Scientific) 16 h prior infection. Infection was conducted as described before at MOI 50. Imaging was performed from 3 to 5 h p.i. as described above. Infected Caco 2 cells in 8-well slides were fixed and used for flagella staining as described below.

Human monocyte cell line U937 cells were seeded in 8-well slides 72 h prior infection to reach ~80% confluency (ca. 200,000 cells) and treated with 50 ng x μl-1 phorbol 12 myristate 13-acetate (PMA) for differentiation and cell attachment until medium was exchanged after 48 h. Cells were pulse-chased with Dextran Alexa Fluor and infected with overnight cultures at MOI 30 as described above. LCI was performed from 3 to 5 h p.i. as described above.

Immunostaining of infected cells

HeLa cells stably transfected with LAMP1-GFP were seeded in surface-treated 24-well plates (TPP) on coverslips 24 h or 48 h prior infection to reach ~80% confluency (~180,000 cells) and were used for infection. Cultivation of bacteria and infection of HeLa cells was carried out as described above. After the desired incubation time the cells were washed thrice with PBS and fixed with 3% PFA in PBS for 15 min at RT, followed by immunostaining. Fixed cells were washed thrice with PBS and incubated in blocking solution (2% goat serum, 2% bovine serum albumin, 0.1% saponin, 0.2% Triton X-100) for 30 min. Cells were stained 1 h at RT with primary antibodies against Salmonella (H:i, BD Difco 228241; H:a, Sifin TR1401; O:5, Sifin TR5303). After washing thrice with PBS, cells were incubated with the appropriate secondary antibodies for 1 h at RT. Coverslips were mounted with Fluoroshield (Sigma) and sealed with Entellan (Merck). Microscopy of fixed samples was either performed using a Cell Observer Spinning Disc microscope (Zeiss) as described above, or using a Leica SP5 confocal laser-scanning microscope using 100× objective (HCX PL APO CS, NA 1.4–0.7) and polychroic mirror TD 488/543/633.

Reinvasion assays

HeLa cells were seeded in surface-treated 6-well plates (TPP) 24 h or 48 h prior to infection to reach ~100% confluency (1×10⁶ cells) on the day of infection. Infection cultures of STM and SPA strains were grown as either 3.5 h aerobic or 16 h microaerobic subcultures and used for infection as described above. At 1 h p.i., cells were washed thrice with PBS and lysed using 0.1% Triton X-100. CFU were determined by plating serial dilutions of lysates and inoculum on Mueller-Hinton II (MH) agar and incubated overnight at 37°C. The percentage of internalized bacteria of the inoculum (from subcultures) was calculated. 8 h p.i., cells were washed twice with PBS, once with ice-cold ddH₂O, and lysed using ddH₂O. Bacteria liberated from host cells by hypotonic lysis were used to infect naïve HeLa cells, and 1 h p.i. cells were washed and lysed and CFU were determined as described above. The percentage of internalized bacteria of the inoculum (from liberated bacteria) was calculated.