RNA-seq Brings New Insights to the Intra-Macrophage Transcriptome of Salmonella Typhimurium

1Salmonella enterica (S. enterica) is a food and water-borne pathogen responsible for widespread disease in humans and other animals. The serovars responsible for typhoid fever kill more than 250,000 people per year, while an estimated 94 million cases of Salmonella-mediated gastroenteritis cause 155,000 deaths each year. Recently, it has been discovered that non-typhoidal serovars are causing an epidemic of invasive disease that is killing 680,000 people each year.

2Decades of intense research have revealed intricate details of Salmonella pathogenicity. S. enterica initiates infection in the small intestine by penetrating the mucus layer that protects the gut epithelium. During the infection process, S. enterica endures a series of hostile environments within the host, including the acidity of the stomach, antimicrobial peptides and bile in the intestine, and the toxicity of intracellular vacuoles. These challenges are met by physiological and metabolic adaptations that allow the bacterium to resist the innate host defences. Salmonella pathogenicity island (SPI) 1 and SPI4-encoded proteins, and other virulence determinants, mediate the entry into epithelial cells. The bacteria subsequently exit from epithelial cells and are taken up by the phagocytic cells of the innate immune system such as macrophages.

3S. enterica responds to the phagosomal environment within macrophages by secreting effector proteins that generate a specialized intracellular compartment, the Salmonella-containing-vacuole (SCV). The SCV allows S. enterica to evade macrophage killing, and infected macrophages become a vehicle for systemic bacterial spread. Physiological, metabolic and effector protein-mediated adaptation strategies allow the bacteria to replicate within the SCV, and to form persister cells; many of these adaptive processes are regulated at the transcriptional level.

4Bacterial gene regulation is mediated by a combination of transcription factors, nucleoid-associated proteins and regulatory small non-coding RNAs (sRNAs). Following the publication of the first S. enterica genome, microarray-based transcriptomic approaches were used to define regulons and stimulons of the model pathogen S. enterica serovar Typhimurium (S. Typhimurium). Because the microarray-derived data only provided a limited view of Salmonella gene expression inside macrophages, an RNA-seq-based approach was required to gain the information for understanding mechanisms of gene regulation. RNA-seq analysis generates high-resolution transcriptomic data and accurate information on gene expression levels, and provides extensive information concerning the location of Transcriptional Start Sites (TSS), the 5′ and 3′ un-translated regions of genes, antisense transcription, and sRNAs. We recently used this approach to reveal the complete transcriptional network of S. Typhimurium during growth in 22 laboratory conditions.

5Here, we present the primary transcriptome of intra-macrophage S. Typhimurium strain 4/74. All intra-macrophage gene expression and transcriptional organisation data are presented in our online resource, SalComMac [http://tinyurl.com/SalComMac].

6The intra-macrophage transcriptome of S. enterica was determined with S. Typhimurium strain 4/74 (Dataset 1 in S1 Table) within cultured murine RAW 264.7 macrophage-like cells that do not express the Nramp1 (Slc11a1) host resistance cation-efflux pump. Because earlier transcriptomic analyses showed that more than 90% of S. Typhimurium genes were expressed at similar levels during early, middle and late stages of macrophage infection, we focused on a single time point. We used eight hours post-infection to coincide with the nitrosative burst in Salmonella-infected murine macrophages. Total bacterial RNA was isolated and analysed by RNA-seq (Materials and Methods) (Fig 1). Overall, 136 million sequence reads were generated from seven cDNA libraries. These represent two biological replicates of intra-macrophage Salmonella RNA-seq, two biological replicates of intra-macrophage Salmonella differential RNA-seq (dRNA-seq) and RNA-seq of the ΔssrA mutant and two biological replicates of wild-type 4/74 grown under in vitro SPI2-inducing conditions. Between 5 and 10 million uniquely-mapped reads were obtained from each library (Dataset 2 in S1 Table), providing sufficient coverage for robust transcriptomic analysis. Gene expression values were calculated by the Transcripts Per Million (TPM) approach. A threshold TPM value of 10 was used as a cut-off to define gene expression (Materials and Methods). The intra-macrophage transcriptome was compared to our published RNA-seq-based transcriptome for early stationary phase (ESP), an infection-relevant in vitro growth condition that is associated with high expression of S. Typhimurium SPI1 genes.

7The precise nucleotide position of individual TSS was identified on a genome-wide scale by dRNA-seq. In total, 3583 TSS were expressed by S. Typhimurium during infection of macrophages (Dataset 3 in S1 Table). This included 3538 TSS expressed in the ESP condition and 45 TSS which were newly identified in this study. To assign a relative strength to each TSS we determined the expression levels of the first 10 bases of each transcript, designated the promoter usage value (PUV). Because >99% of S. Typhimurium protein coding genes have a 5’ untranslated region (UTR) and 15% of protein-coding genes possess multiple TSS, the PUV allows promoter strength to be quantified independently of gene expression. We used the relative PUV to compare the expression of S. Typhimurium TSS between the intra-macrophage and the ESP in vitro condition, and categorised the TSS as either ‘Macrophage up-regulated’, ‘Macrophage down-regulated’ or ‘Macrophage-independent’ (Materials and Methods). Of the 3583 TSS expressed in macrophages, 883 were macrophage up-regulated and 834 were macrophage down-regulated, compared with ESP (Fig 2A; Dataset 3 in S1 Table). The TSS of the lgl-ripABC (STM3117-3120) SPI13 operon was the most highly up-regulated, with a relative PUV of >500 -fold. Other highly up-regulated TSS controlled the expression of genes such as trpE and sseJ. We found that 72% of the promoters reported to be highly expressed in the murine spleen were up-regulated in RAW macrophages.

8Forty five new TSS were identified in this study, including the TSS of STM0854 that controls intra-macrophage expression of the major polycistronic transcript of SPI14 (Fig 2B). Other novel TSS controlled the expression of genes involved in several core cellular processes including bglA, entB, fliN and nrdE, and a TSS that initiated a transcript antisense to the stfD coding gene.

9Of the 834 macrophage down-regulated TSS, the biggest reduction in promoter expression between macrophages and the ESP condition was more than 200-fold and associated with SPI1 genes and the flagellin-encoding fliC gene (Dataset 3 in S1 Table).

10Salmonella pathogenicity island 2 (SPI2) is required for Salmonella replication within eukaryotic cells and for systemic infection of mammalian hosts. SPI2 encodes the type III secretion system (T3SS) that delivers many effector proteins responsible for the function of the SCV within macrophages. The transcriptional organization of SPI2 is shown in Fig 3. We recently used dRNA-seq to discover a TSS upstream of ssaR, which we now confirm by 5’ RACE (S1 Fig). SPI2 is therefore transcribed as six operons inside macrophages (Fig 3).

11All SPI2 genes were up-regulated within macrophages, reflecting the phosphate/magnesium starvation and the acidity of the SCV. The RNA-seq data were used to calculate promoter usage values for the different SPI2 promoters, identifying PssaM as the most up-regulated SPI2 promoter, followed by PssaR and PssaB (Dataset 3 in S1 Table). We note that each of the six SPI2 promoters was also transcribed in the “InSPI2” growth condition, confirming that expression of all SPI2 operons occurs in vitro when stimulated by growth in an acidic low-phosphate environment.

12The SPI2 island and genes that encode SPI2-translocated effectors are activated by the SsrAB two component system. The SsrA sensor kinase phosphorylates the SsrB response regulator to activate gene expression. To investigate the role of SsrA in the regulation of macrophage up-regulated TSS, we used RNA-seq to analyse the transcriptome of a ΔssrA mutant and wild-type 4/74 grown in InSPI2 medium. Of the 883 macrophage up-regulated TSS, 221 showed reduced (>2-fold) expression in the absence of SsrA and we infer that these are SsrA-activated (Fig 2A; Dataset 3 in S1 Table). All the genes that encode SPI2-translocated effector proteins were controlled by SsrA-activated promoters. There are 662 macrophage up-regulated TSS that appear to have SsrA-independent regulatory mechanisms, and these merit further study.

13S. Typhimurium carries 12 pathogenicity islands on the chromosome of strain 4/74. Expression profiles of S. Typhimurium pathogenicity islands (Fig 4; Datasets 4 and 5 in S1 Table) reveal that SPI2 and SPI13 were the most highly up-regulated during infection of macrophages, by an average of 44 and 82-fold, respectively (Dataset 5 in S1 Table). The SPI3, SPI5, SPI11, SPI12 and SPI14 islands showed moderate intra-macrophage up-regulation. SPI6 and SPI9 show macrophage-independent expression, and both SPI1 and SPI4 were significantly down-regulated inside macrophages.

14Effector proteins of S. Typhimurium are secreted via the SPI1 T3SS, the SPI2 T3SS or through both translocation systems. We reported that the genes encoding SPI1-translocated effectors showed a SPI1-like expression pattern, and genes encoding SPI2-translocated effectors showed a SPI2-like expression pattern. Our data show that the genes encoding all SPI2-translocated effectors were highly macrophage up-regulated (Dataset 4 in S1 Table) (up to 70-fold), and the genes that encode the 7 effectors that are secreted by both the SPI1 and SPI2 T3SS were all expressed inside macrophages; the TPM values range from 50 to 230 (Fig 5). In contrast, genes encoding the 9 SPI1-translocated effectors were all macrophage down-regulated, by up to 160-fold, and were not significantly expressed within macrophages. Clearly, the actual intra-macrophage expression level of genes that encode candidate effector proteins has biological relevance.

15The transcriptomic data identified two specific SPI13 and SPI14-encoded operons that were highly up-regulated in macrophages (Datasets 4 and 5 in S1 Table) but were not significantly expressed in 20 in vitro conditions. First, the SPI13-associated lgl-ripABC (STM3117-STM3120) operon was >250 fold up-regulated within macrophage. The lgl-ripABC operon is required for Salmonella infection, encoding enzymes that catabolise itaconate, an anti-microbial metabolite that is synthesised by infected macrophages. Second, the SPI14-located STM0854-0857 operon is also required for Salmonella virulence, showed moderate (3 to 20-fold) intra-macrophage up-regulation, and was not expressed in in vitro growth conditions. The TSS of the STM0854 and STM0859 transcripts were only expressed in macrophages, and not in any in vitro conditions. Taken together, these data suggest that the STM0854-0857 and lgl-ripABC operons respond to an intra-cellular signal that remains to be identified in macrophages. For ripABC, this signal may be itaconate.

16For SPI3, the PhoP-activated mgtCBR operon was up-regulated >15 fold within macrophages, while other SPI3 genes (slsA, marT and rhuM) were moderately up-regulated. The role of mgtCBR in virulence involves the long leader of the mgtC transcript that encodes MgtP. The mgtC leader is responsive to ATP levels and inhibits F1Fo ATP synthase to maintain ATP homeostasis in the acidic intra-macrophage environment.

17SPI5 encodes effectors translocated by both SPI1 and SPI2 T3SS. The sopB gene encodes a SPI1-translocated effector and is macrophage down-regulated by 50-fold. In contrast, the gene encoding the SPI2 effector pipB is up-regulated. PipB localizes to the SCV membrane and brings about the formation of tubular extensions, the Salmonella induced filaments (SIFs).

18The SPI6-encoded Type 6 secretion system, is important for the colonization and systemic infections of chickens and mice. None of the SPI6 genes were expressed in macrophages or in various in vitro conditions. This is consistent with the reported repression of SPI6 genes by H-NS.

19During infection of the gastrointestinal tract, the SPI1-encoded T3SS of S. Typhimurium is responsible for inflammatory diarrhoea and the invasion of non-phagocytic epithelial cells. Thirty-three SPI1 genes were down-regulated within macrophages (Dataset 4 in S1 Table), and were highly expressed at ESP, confirming earlier reports. HilA, the transcriptional activator of SPI1, is controlled by the co-ordinated action of HilC/HilD/RtsA, and consequently up-regulates the SPI1 island & SPI1-translocated genes. The transcription of hilA is regulated by HilD, an important activator that controls cross-talk between SPI1 and SPI2 expression. The hilA, hilC, hilD and rtsA regulatory genes are down-regulated more than 100-fold within macrophages, consistent with the down-regulation of the SPI1 island.

20The siiABCDEF operon of SPI4 encodes a Type 1 secretion system, and was down-regulated within macrophages. SiiE is a non-fimbrial adhesin responsible for the adhesion of Salmonella to epithelial cells and is expressed during the extra-cellular phase of infection. Cross talk between SPI1 and 4 can promote tight binding of the bacterium to the epithelial membrane, and facilitate efficient SPI1 translocation.

21Intracellular expression of individual bacterial genes or entire regulons can be used to investigate the microenvironment inside the host cell vacuole. Direct comparison between this RNA-seq-based dataset (Dataset 4 in S1 Table) and previous microarray-based transcriptomic results confirm and extend key findings from Eriksson et al. (2003) and Hautefort et al. (2008). The datasets all show that the most highly macrophage up-regulated Salmonella gene is asr (STM1485), required for the intra-cellular replication of Salmonella. The 890-fold up-regulation of asr reflects the acidic conditions within the SCV (Dataset 4 in S1 Table).

22To investigate the gene expression network of intra-macrophage Salmonella, we focused on 157 transcriptional regulators (Dataset 6 in S1 Table). The levels of 34 transcription factors were >3-fold macrophage up-regulated, and 7 transcription factors were >3-fold macrophage down-regulated. To determine whether the differential expression of individual regulators was reflected by up- or down-regulation of the associated regulons, we compared the expression of several genes controlled by each transcription factor in ESP and macrophages. We observed that the up-regulation of SPI2 regulators ssrB, ompR and phoP and down-regulation of SPI1 regulators hilD, hilA, hilC, invF and sprB correlates with the expression of their respective regulons (Fig 6A and 6B). The macrophage up-regulation of regulons that detoxify peroxide, detoxify nitric oxide and relieve envelope stress and protein misfolding (soxS, oxyR, marA, marS, rpoE, rpoH and nsrR regulons and genes hmpA, msrA, ycfR, sbp, sodC, katG), reflects the bacterial response to the oxidative and nitrosative bursts that occurred during the infection process.

23Bacterial genes were assigned to functional groups to investigate the metabolic resources of macrophages. The most up-regulated functional categories of S. Typhimurium genes within macrophages are involved in carbohydrate and amino acid metabolism (S2 Fig). The “nutritional immunity” hypothesis posits that the innate immune response of the host reduces the availability of important nutrients for intracellular bacteria, which may explain why S. Typhimurium has evolved the ability to utilise a diverse range of host nutrients, including some sugars and amino acids that accumulate in murine macrophages during intracellular infection. It is known that the major carbon sources utilised by S. Typhimurium in macrophages of the mouse spleen are deoxyribonucleotides, fatty acids, glucose, gluconate, glycerol, lactate and N-acetyl-glucosamine. In our study, we observe the concerted up-regulation of multiple metabolic regulons in RAW macrophages that are consistent with the simultaneous degradation of deoxyribonucleotides, fatty acids, galactose, glucose, gluconate, glycerol, lactate, N-acetyl-glucosamine and sialic acid, while regulons controlling gluconate, maltose, myo-inositol and xylose metabolism showed significant macrophage down-regulation (Fig 6). Our current understanding of the intracellular metabolism of Salmonella in cultured macrophages coupled with the comprehensive data available for S. Typhimurium during infection of the murine spleen suggest that cultured macrophages represent a good model for the study of the intracellular metabolism of Salmonella.

24Mammalian macrophages reduce intracellular levels of metals such as iron as part of their strategy to limit bacterial replication, and S. Typhimurium responds by switching on the expression of metal-uptake systems. These include the intra-macrophage up-regulation of the sitABCD operon, responsible for manganese and iron transport and of genes responsible for iron transport and biogenesis of iron-sulfur cluster containing proteins (ent, fep, fhu, iro, sfb, sit and suf genes, as well as the yhfP (iscR), and rstA regulons), magnesium (mgtCBR) transport and zinc (zur) uptake. We suggest that these expression patterns reflect the relatively low levels of magnesium, manganese, iron and zinc metals within the SCV. Genes encoding the flagella and chemotaxis systems were significantly down-regulated in macrophages (between 50 to 100-fold), consistent with previous reports for both the Typhimurium and Typhi serovars (Dataset 4 in S1 Table; S2 Fig). Specifically flh, flg, fli, flj, mot, che and aer genes were down-regulated. The flhDC-mediated regulation of flagellar transcription is complex, and cross-talk between SPI1 and flagellar genes was recently reported. The flagellar regulator FliZ is a post-transcriptional activator of flhDC that positively regulates SPI1 by activating the hilD-rtsAB cascade. In turn, RtsB represses the flhDC promoter. These regulatory mechanisms probably account for the down-regulation of flagellar genes within macrophages, consistent with the shut-down of flagellar synthesis associated with the non-motile bacteria found in the SCV. This contrasts with the reported up-regulation of SPI1 and flagella that occurs when S. Typhimurium encounters the cytosol of epithelial cells.

25To find genes that were up-regulated in the intra-macrophage environment but not in standard laboratory conditions, we used a comparative transcriptomic approach to identify genes that showed significantly higher expression in macrophages than in any of 20 in vitro conditions (Materials and Methods). Our analysis identified 31 genes that were specifically up-regulated within macrophages (Dataset 7 in S1 Table; Fig 7A and 7B). These represent an interesting class of bacterial genes that are up-regulated in macrophages due to a factor encountered within macrophages and not in the in vitro growth conditions. The STM3117-STM3120 (lgl-ripABC) genes are a good example, of highly macrophage-induced genes (Fig 7C) that are involved in the detoxification of two SCV-specific metabolites, methylglyoxal and itaconate. We propose that comparative transcriptomics will be a useful approach for identifying genes that respond to specific components of the SCV environment. The majority of the genes in Fig 7A have a STM or a yxx prefix and are designated as “FUN” genes, for “function unknown”. Overall, 18 of the 31 genes that were specifically up-regulated within macrophages have previously been shown to be required for virulence (Dataset 7 in S1 Table). We speculate that these FUN genes respond to a specific component of the intra-vacuolar environment of the macrophage and could play important roles in the process of infection.

26The identification of a discrete set of promoters that are up-regulated in macrophages could have therapeutic applications. Attenuated strains of S. Typhimurium have been used extensively as vaccines, and for expressing anti-cancer proteins within tumours. These technologies require specific Salmonella gene promoters to drive the production of foreign antigens. For example, the ssaG promoter of SPI2 has been used to express E. coli heat labile toxin in S. Typhimurium. However, the ssaG promoter is active in the gut, and so may not be the ideal antigen delivery system. We sought to identify candidate promoters with the characteristics required to deliver antigens from attenuated live vaccine strains of S. Typhimurium during intracellular infection.

27We screened our intra-macrophage promoter expression data to identify primary TSS that were highly expressed within macrophages, and driving a downstream gene that was highly macrophage up-regulated. Eleven promoters were identified as suitable for antigen delivery during infection (Dataset 8 in S1 Table), controlling the asr, bioB, iroB, sseJ, STM0854 (SPI14) and ripC (SPI13) genes. Of these, sseJ is highly expressed within mouse organs. The ripC promoter may be ideal for antigen delivery as it is highly and specifically induced inside macrophages (Dataset 3 in S1 Table; Fig 7C). However, high-level expression of heterologous antigens does not always generate the optimal stimulation of immune responses, and over-expression of certain proteins could compromise bacterial fitness. For this reason, we categorized the macrophage-up-regulated genes from Fig 7A, based on their levels of intra-macrophage expression and identified the promoter of STM0854 as a promising candidate for moderate but specific induction of gene expression within macrophages (Fig 7B). The 11 promoter candidates have the potential to deliver different levels of heterologous antigens and could be used to improve Salmonella-based intracellular vaccine delivery systems.

28Bacterial gene expression is controlled by transcription factors, nucleoid-associated proteins and sRNAs. Bacterial sRNAs are roughly 50–300 nucleotides in length, and play regulatory roles in key physiological activities like iron homeostasis, carbon metabolism, anaerobic adaptation, envelope stress and pathogenesis. To date, 280 sRNAs have been identified in S. Typhimurium 4/74, but little is known about their role in virulence. The fact that 246 of 280 sRNAs were expressed within macrophages (TPM value >10; Dataset 9 in S1 Table) suggests that many could potentially play a regulatory role during infection. In terms of relative expression, we found that 34 sRNAs were macrophage up-regulated and 119 sRNAs were macrophage down-regulated, compared to ESP (Dataset 9 in S1 Table; Fig 8A). The Hfq chaperone protein mediates sRNA-mRNA interactions and binds to at least 115 S. Typhimurium sRNAs, of which 19 were up-regulated within macrophages (including RyhB-1/2, OxyS, MicF and RybB) and 56 were down-regulated (including ArcZ, DsrA and DapZ), compared to ESP (Dataset 9 in S1 Table; Fig 8B).

29The expression patterns of well-characterised sRNAs provide insight into the conditions experienced by S. Typhimurium bacteria in the SCV. For instance, up-regulation of the RpoE-dependent sRNAs MicA and RybB inside macrophages likely reflects envelope stress of S. Typhimurium during intracellular proliferation. Another sRNA that is RpoE-dependent in E. coli, MicL (RyeF) is up-regulated 30-fold within macrophages, but it is not yet known whether this sRNA is controlled by RpoE in Salmonella. The iron-regulated homologs RyhB-1 and RyhB-2 were the most highly up-regulated sRNAs within macrophages compared to ESP (Dataset 9 in S1 Table, Fig 8C), reflecting the iron-limited intra-macrophage environment. RyhB-1 and RyhB-2 (named RfrA and RfrB in S. Typhi) are also known to be important for replication of S. Typhi within macrophages. Our data confirm that the IsrH, RyhB-1 and RyhB-2 (IsrE) sRNAs are up-regulated, as originally reported within J774 macrophages. We analysed the expression of six sRNAs that were up-regulated within fibroblasts, a cell type that does not support the replication of Salmonella. Two of these sRNAs, RyhB-1 and RyhB-2, were also up-regulated in macrophages (Dataset 9 in S1 Table). We identified several uncharacterized Hfq-associated sRNAs that were up-regulated within macrophages, including STnc440, STnc470 and STnc3750 which have an expression pattern consistent with a role in virulence. The function of these sRNAs is currently under investigation.

30To determine whether macrophage-regulated sRNAs were phylogenetically conserved between fourteen serovars that represent much of the diversity of the Salmonella genus, we analysed 29 enterobacterial genomes (Dataset 10 in S1 Table). We found that 176 sRNAs were conserved (>90% sequence identity) within the Salmonella genus, but not in other members of the Enterobacteriaceae (<70% sequence identity), and were designated Salmonella-specific. About 10% (17) of the Salmonella-specific sRNAs were up-regulated within macrophages (including STnc440 and IsrH) while 74 were down-regulated in macrophages (including DapZ and InvR), compared to ESP (Fig 8C, Dataset 10 in S1 Table). We propose that some of these 91 macrophage-regulated sRNAs could play important roles in the regulation of gene expression during the intracellular phase of Salmonella infection.

31Salmonella bacteria are exposed to multiple stressors within the vacuolar compartment of macrophages, including acid pH, reactive oxygen and reactive nitrogen species. Adaptation to this hostile environment has a profound impact upon the transcriptome of S. Typhimurium, and we have now defined the TSS and sRNAs that react to the intra-vacuolar environment during the intracellular phase of the Salmonella infection cycle. Our data provide an overall view of sRNA expression within macrophages, and represent a resource for the investigation of post-transcriptional regulation during the intracellular life of Salmonella.

32This study offers new insights into the interaction of Salmonella with mammalian cells, and brings us a step closer to understanding the gene regulatory mechanisms that facilitate the success of this dangerous pathogen. The SalComMac online resource [http://tinyurl.com/SalComMac] is intended to simplify the comparison of the transcriptome of intra-macrophage and in vitro grown S. Typhimurium.

33Salmonella enterica subspecies enterica serovar Typhimurium strain 4/74 was used for all experiments; 4/74 is the prototrophic parent of strain SL1344; the two strains differ by just eight single nucleotide polymorphisms. For in vitro RNA isolation, bacterial cells were grown overnight in 5 mL Lennox (L-) Broth (Dataset 1 in S1 Table), diluted 1:1000 into 25 mL L-broth, grown at 220 rpm and 37°C in a 250 mL flask until early stationary phase (ESP, OD600 2.0). InSPI2 minimal media was used to induce expression of SPI2 in vitro. For all intracellular studies, RAW 264.7 (ATCC) murine macrophage cells were maintained in Dulbecco’s Minimal Essential Medium (DMEM) supplemented with 5% fetal bovine serum & L-glutamine (2 mM final concentration) and MEM non-essential amino acids without antibiotics, incubated at 37°C in 5% CO2. All tissue culture reagents were supplied by Lonza.

34Approximately 109 RAW 264.7 macrophage cells were seeded in 175 cm2 tissue culture flasks and infected with complement-opsonized 4/74 cells at a multiplicity of infection (MOI) of 100:1 (bacteria:macrophages). Mouse serum (Charles River Laboratories) was used for opsonisation, and was stored at −80°C prior to use. After 30 minutes of infection, extracellular bacteria were killed by media containing 100 μg mL−1 gentamicin and incubated for a further 1h. The medium was then changed to ‘maintenance media’ containing 10 μg mL−1 gentamicin for the rest of the experiment. At 8 hours post infection, the infected macrophages were lysed in ice cold ‘RNA stabilisation solution’ [0.2% SDS, 19% ethanol, 1% acidic phenol in water] and incubated on ice for 30 minutes to prevent RNA degradation. The lysates containing intracellular Salmonella were collected, centrifuged and RNA was isolated from the bacterial pellets by a TRIzol-based method that yields both mRNA and sRNA. Briefly, the supernatant was discarded, the pellet was washed three times in 19% ethanol, 1% acidic phenol, re-suspended in the remaining liquid, transferred to a clean 1.5 mL Eppendorf tube and centrifuged at 20,000 × g at 4°C. The cell pellet was dissolved in 1 mL TRIzol (Invitrogen) on ice and transferred into a 2 mL heavy phase lock tube (5 Prime) into which 400 μL of chloroform was added and immediately mixed for 10 seconds. After incubation at room temperature for 2 minutes, the mixture was centrifuged at 20,000 × g for 15 minutes. The RNA present in the upper phase was transferred to a fresh tube, and precipitated by adding 450 μL of isopropanol and incubated at room temperature for 30 minutes. The precipitated RNA was then pelleted by centrifugation at 20,000 × g for 30 minutes. The pellet was washed in 350 μL ethanol (70%) and centrifuged at 20,000 × g for 10 minutes. The washed pellet was air-dried, re-suspended in RNase-free water by shaking (900 rpm) for 5 min in a heating block (65°C) (Peqlab Thriller) and stored at −80°C until cDNA library construction.

35The integrity of RNA was verified using an Agilent Bioanalyzer 2100 and RNA concentrations were measured using the nanodrop spectrophotometer (Thermo Scientific) and the Qubit fluorometer (Invitrogen). Control RNA was isolated from bacterial cells grown in L- broth in vitro until ESP (see above). The infection process, RNA preparation, sequencing and analysis were conducted in duplicate to provide independent data from biological replicates.

36The cDNA library preparation and Illumina sequencing was done by Vertis Biotechnologie AG (Freising, Germany). The total RNA obtained from the biological replicates of intra-macrophage was digested for 45 minutes with DNase I (Thermo Scientific) according to the manufacturer’s instructions. Ribosomal RNA was not depleted. RNA samples were fragmented with ultrasound (4 pulses of 30 sec at 4°C). The 3’ ends of RNA were then subjected to poly (A)-tailing using poly (A) polymerase. The RNA was then treated with TAP (Tobacco acid pyrophosphatase) to remove the pyrophosphate group from the 5’ end, prior to ligation with an RNA adapter. First strand cDNA synthesis was done with an oligo (dT) adapter and M-MLV-RNaseH-reverse transcriptase (Invitrogen), following PCR amplification of cDNA using high-fidelity DNA polymerase to a final concentration of approximately 20–30 ng μL-1. The cDNAs were purified using the Agencourt AMPure XP kit (Beckman Coulter Genomics), and analysed by capillary electrophoresis. The cDNA libraries were sequenced on an Illumina HiSeq 2000 system. For dRNA-seq, prior to cDNA preparation, an aliquot of the RNA samples were enriched for primary transcripts by treating with Terminator 5’-monophosphate dependent exonuclease (Epicentre; TEX).

37The sequence reads obtained from the different cDNA libraries were mapped against the 4/74 reference genome using the Segemehl software, with accuracy set to 100%. The mapping coverage was increased by an iterative process that involved the sequential removal of any mismatched nucleotides from the 3’ end, and mapping the read against the 4/74 genome. This process was repeated until the individual sequence reads were accurately mapped to a single location on the chromosome, or until the length dropped below a minimum value of 20 nucleotides. These uniquely-mapped reads were visualised with the Integrated Genome Browser (IGB) and Jbrowse. In total, 6 cDNA libraries (including the biological replicates of RNA-seq, dRNA-seq and RNA-seq of InSPI2 grown ΔssrA & wild type S. Typhimurium 4/74) were generated.

38The expression values of each gene were calculated from the uniquely-mapped reads using the Transcript per Million (TPM) approach. TPM considers the transcripts to represent a mixture of two distributions of expressed and non-expressed genes, and so is ideal for the analysis of bacterial transcriptomic data. As this approach involves normalization to gene size and the total amount of genome-wide transcription, TPM values can be compared between genes and between growth conditions.

39The threshold for expression of a gene was TPM value 10. Genes with TPM value ≤10 were considered to be “not expressed”. The differential expression of each gene or sRNA within macrophages was calculated against the ESP comparator as a fold change (macrophage versus ESP).

40The average and standard deviation of RNA-seq data (TPM values) was calculated for each gene from the 20 in vitro growth conditions reported earlier. For each gene, the standard deviation was multiplied by five-fold to define a broad expression range that captured all but the most extreme expression levels across the 20 conditions. To identify genes that were specifically up-regulated in macrophage, we selected a strict cut-off of 3-fold more highly expressed than five standard deviations above the mean expression value from the 20 conditions. In other words, macrophage specific gene = TPM > 3 x (average TPM in 20 conditions + 5σ). The genes that passed this cut-off are ‘not significantly expressed’ in any of the 20 in vitro conditions, are up-regulated within macrophages, and are listed in Fig 7A.

41A strict criterion was used to identify TSS, after visualization with the IGB browser. Novel TSS were defined when a peak was enriched in the dRNA-seq data compared with the RNA-seq data in two biological replicates, and was located at the beginning of an expressed transcript.

42The Promoter Usage Value (PUV) for each TSS was quantified by calculating the TPM for the first 10 nucleotides from the TSS towards the direction of transcription (from +1 to +10). The PUV values were classified as follows: (a) ‘Macrophage independent’ TSS have similar PUV in macrophages and at ESP (less than 2-fold up- or down-regulated); (b) ‘Macrophage up-regulated’ TSS are expressed at least 2-fold higher in macrophages relative to ESP; and (c) ‘Macrophage down-regulated’ TSS are expressed at least 2-fold less in macrophages relative to ESP.

43The 5’ RACE (rapid amplification of cDNA ends) was carried out with or without treatment by TAP using DNase I-digested total RNA isolated from the InSPI2 condition. Gene specific amplification was done with the linker-specific primer JVO-0367 and gene specific reverse primers (Dataset 1 in S1 Table). TAP-enriched fragments were excised from an agarose gel, subcloned into a pTOPO vector (Invitrogen) and at least three clones were sequenced to validate individual TSS.

44The sRNA nucleotide sequences from 4/74 were aligned against a set of bacterial genomes belonging to Enterobacteriaceae using GLSEARCH, and identical hits were extracted.

45The RNA-seq data generated from this study are deposited at the NCBI GEO under the accession numbers GSM1462575 to GSM1462579, GSM1914919 and can be accessed at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE59945.