Analysis and construction of pathogenicity island regulatory pathways in Salmonella enterica serovar Typhi

Summary Signal transduction through protein-protein interactions and protein modifications are the main mechanisms controlling many biological processes. Here we described the implementation of MedScan information extraction technology and Pathway Studio software (Ariadne Genomics Inc.) to create a Salmonella specific molecular interaction database. Using the database, we have constructed several signal transduction pathways in Salmonella enterica serovar Typhi which causes Typhoid Fever, a major health threat especially in developing countries. S. Typhi has several pathogenicity islands that control rapid switching between different phenotypes including adhesion and colonization, invasion, intracellular survival, proliferation, and biofilm formation in response to environmental changes. Understanding of the detailed mechanism for S. Typhi survival in host cells is necessary for development of efficient detection and treatment of this pathogen. The constructed pathways were validated using publically available gene expression microarray data for Salmonella.


Introduction
S. Typhi is able to survive a variety of harsh conditions and defense mechanisms existing in the human gastrointestinal tract. Multiple survival strategies allow S. Typhi to cause epidemic outbreaks of typhoid fever in many developing countries. Therefore, Salmonella represents a major health risk according to the World Health Organization (WHO) [1]. Propagation of S. Typhi infection is due to its ability to enter a dormant state by forming biofilm in the human gallbladder (typhoid carriers), enabling it to evade the immune system [2]. Typhoid carriers do not show any symptoms, and are the only reservoir for S. Typhi which is transmitted via contaminated food or water. Existing diagnostic tools cannot detect S. Typhi in typhoid carriers.
reciprocal hit method from full length protein sequence similarities calculated from BLAST alignments as described previously [9]. The Bacterial database contains molecular interactions extracted by MedScan for all bacterial species from over 1,000,000 PubMed abstracts annotated with Medical Subject Headings (MeSH) term "Bacteria" and from more than 74,000 full-length articles from 22 microbiology journals. Proteins in the Bacterial database are annotated with Entrez Gene and GenBank identifiers from 32 bacterial genomes, including S. Typhi Ty2, S. Typhi CT18, and S. typhimurium LT2. Additional annotation from 716 partial genomes was obtained from the NCBI Protein Clusters database. The database allows quick identification of interactions reported for different bacterial organisms that can be relevant for pathway construction in S. Typhi. All interactions extracted for Salmonella orthologs were imported into the Pathway Studio Salmonella database for pathway building and network analysis of the gene expression data. All interologs used for pathway reconstruction in S. Typhi were manually curated. Only validated interactions were included in pathways.

Construction of pathways controlling expression of SPIs
The first step in pathway building was identification of proteins in the database encoded by each SPI in S. Typhi Ty2 or S. Typhi CT18. A simple search for the proteins with corresponding Entrez Gene ID was performed in Pathway Studio database. Entrez Gene IDs for SPI proteins were obtained by exploring the S. Typhi CT18 genome (GenBank accession number NC_003198) in NCBI sequence viewer. Once SPI proteins were identified, we connected them with either physical interactions or expression regulatory relations found in the Bacterial and Salmonella databases. We then expanded the pathways by adding all known transcriptional regulators for SPI proteins. We also added autophosphokinases that regulate the activity of transcriptional factors in two-component relay signaling system. Next, we added environmental signals that are sensed by two-component regulatory systems. Finally, we added human proteins that are known to interact with S. Typhi effectors.
We manually verified each interaction used for pathway construction by reading the original article and making substantiative assertion to validate the interaction. MedScan classifies extracted relations using only the information available in the sentence describing the extracted fact. Therefore, we manually converted all regulatory relations classified as Expresssion by MedScan into PromoterBinding if the regulation has been described as a direct interaction elsewhere in the text. Some indirect regulatory interactions were explained by connecting several intermediate proteins into a path consisting of consecutive direct physical interactions. Lastly, we excluded the redundant interactions that were extracted by MedScan from our pathways.

Network analysis of gene expression microarray data
Gene expression omnibus (GEO, NCBI) dataset GSE3096 was used for network analysis. GSE3096 measures S. Typhi gene expression during the infection of human macrophages (THP-1) [10]. We used Sub-network Enrichment Analysis (SNEA) algorithm [11] with option "Expression targets" available in Pathway Studio to identify significant transcription factors regulating most differentially expressed genes. If a gene was measured by multiple probes on the array only probe with best p-value was used for SNEA. All relationships used to identify major regulators were manually verified after the initial analysis, false positives were removed from the database and SNEA was run for second time to verify again the significance of transcription factors.

Salmonella Pathogenicity Island 2 (SPI-2)
SPI-2 consists of 45 genes that are required for survival of S. Typhi inside phagosomes ( Figure 3). OmpR activates SPI-2 genes by binding to the promoter of the ssrAB operon to induce expression of SsrA and SsrB proteins [16]. The OmpB-OmpR two-component system is activated by low osmolarity while the PhoQ-PhoP system, which also regulates SPI-2 genes, senses the acidity of the environment inside the phagosomes. Expression of SPI-2 genes is mainly regulated through the SsrA-SsrB two-component system. Many secreted effector proteins are located at different Salmonella loci but are translocated via the T3SS system encoded by SPI-2 (eg: PipA and PipB from SPI-5). SPI-2 also contains the ttrRSBCA operon which encodes tetrathionate reductase. Although TtrB, TtrC, and TtrA are not involved in virulence, they are essential for anaerobic respiration [12,17,18]. According to [17], the ability to respire tetrathionate is likely to be significant within the life cycle of Salmonella. This ability is a characteristic of only certain genera of Enterobacteriacea including Salmonella, Citrobacter, and Proteus [19]. Further in the text we demonstrate that low oxygen serves as a main trigger for activation of SPI-1 invasion genes during macrophage invasion. Hence, expression of tetrathionate reductase during SPI-2 activation may be used to promote Salmonella survival inside the host cell.
Ten SPI-2 genes were reblasted to confirm their identity and function. Analysis of BLAST results shows that the major portion of pseudogene sty1739 is highly similar to DeoR family transcriptional regulator and pseudogene sty1742 is similar to proline iminopeptidase, suggesting that these genes are functional as both were expressed in the microarray experiment.

Salmonella Pathogenicity Island 3 (SPI-3)
SPI-3 genes were shown to be important for S. Typhi survival inside the host cells. Existing literature indicates that SPI-3 consists of fourteen genes including six pseudogenes ( Figure  4). We found only five proteins in SPI-3 annotated with known functions: RmbA, SlsA, MgtA, STY4022 (MgtB), and STY4023 (MgtC). Expression of MgtA is dependent on RpoS, RcsC-YojN-RcsB, and PhoP. MgtA, MgtC, and MgtB function in high-affinity Mg 2+ uptake. The ability to survive in Mg 2+ limitation is necessary for S. Typhi virulence [20]. Nine other proteins encoded by SPI-3 were reblasted to refine their functional annotation available in S. Typhi CT18 genome. We found that sty4030 encodes a full length homolog of S. typhimurium MisL (an autotransporter) which serves as an intestinal colonization factor that binds to human fibronectin [21]. sty4024 was similar to CigR from S. typhimurium, and sty4027 was similar to S. typhimurium putative transcriptional regulator MarT. Surprisingly, sty4030, sty4024, and sty4027 are annotated as pseudogenes in the S. Typhi CT18 genome [22] but the microarray data shows that these genes are expressed during macrophage infection.

Salmonella Pathogenicity Island 4 (SPI-4)
SPI-4 has 7 genes, regulated by the same regulatory network as SPI-1 genes ( Figure 5). SPI-1 was shown to be required for the activation of SPI-4 [23], which further supports our SPI-4 pathway. In addition to the SPI-1 regulators SirA, HilA, and H-NS, expression of SPI-4 genes is also regulated by RfaH. RfaH is an anti-termination factor preventing premature termination of transcription in SPI-4 [23]. The organization of SPI-4 genes in S. Typhi is similar to the siiABCDEF operon in S. typhimurium. sty4456 (siiC), sty4457 (siiD), and sty4460 (siiF) encode a type I secretion system (T1SS) necessary for the secretion of siiE [24]. SiiE is a large repetitive protein that functions as a nonfimbrial adhesin in binding to epithelial cell surfaces [25]. Unlike in S. typhimurium, siiE in S. Typhi is encoded by two orfs, sty4458 and sty4459. Our sequence analysis suggested that sty4458 and sty4459 were not pseudogenes as reported previously [22]. Besides both genes being similar to siiE from S. typhimurium, microarray data also confirms that siiE is expressed in S. Typhi [10].

Salmonella Pathogenicity Island 5 (SPI-5)
SPI-5 is a 7.6 kb region encoding 8 genes: pipD, sigD/sopB, sigE, pipA, pipB, and three transposases (sty1124, tnpA, and sty1125) ( Figure 6). The genes are controlled by the SPI-1 and SPI-2 regulatory circuits and are known to contribute to Salmonella enteropathogenesis [12,26]. SopB is secreted through the T3SS encoded by SPI-1, while PipA and PipB are secreted through the T3SS encoded by SPI-2. Expression of PipA and PipB is regulated by the EnvZ/OmpR two-component regulatory system. SigE is a molecular chaperone which is important for the stabilization and secretion of SopB/SigD [27]. SigD/SopB is a secreted inositol phosphatase that triggers fluid secretion responsible for diarrhea [26]. It activates mammalian protooncogene Akt, a serine threonine kinase responsible for inhibition of apoptosis in normal intestinal epithelial cells during the infection [28]. pipD encodes a cysteine protease homolog which is crucial in contributing to long-term systemic infection [29]. 3.1.6 Salmonella Pathogenicity Island 6 (SPI-6) SPI-6 encodes 59 genes ( Figure 7). The function and regulation of SPI-6 genes is still largely unknown and they are not annotated in GenBank. Therefore, we performed additional sequence analysis for SPI-6 genes. We found that SciN, SciP, SciS, SciK, VapD, VgrS, SciF/ImpF, and SciQ are homologous to the type VI secretion system (T6SS) machinery identified in V. cholerae [30]. The Saf operon (safA, safB, safC, and safD) and tcf operon (tcfA, tcfB, tcfC, and tcfD) are fimbrial usher proteins. Twenty proteins are identified as cytoplasmic proteins, two proteins as integral membrane proteins, two proteins as periplasmic proteins, and four proteins as transposases. After our sequence analysis there are still fifteen genes left as hypothetical with no homology to proteins with known function. The complete results of our analysis are shown in Table 1.

Salmonella Pathogenicity Island 7 (SPI-7)
The SPI-7 region is unique to S. Typhi. It consists of 148 genes (Figure 8), encoding a prophage and genes for virulence factors such as Vi antigen (ten genes), SopE effector, and type IV pili (fifteen genes) [31]. The production of Vi antigen is governed by the twocomponent systems EnvZ-OmpR and RcsC-RcsB ( Figure 8). The TviA regulator encoded by SPI-7 interacts with transcription factor RcsB to promote transcription of Vi antigen genes [32]. Interestingly, the same system also controls the pil operon (type IV pili) [32]. Meanwhile, effector protein SopE is translocated through the T3SS of SPI-1. 80 out of 148 proteins were classified as either hypothetical proteins or proteins with unknown function. We performed an extensive sequence analysis using bioinformatics tools to assign predicted functions to these proteins. We found that thirteen are related to prophage, another thirteen are related to DNA recombination, and three are similar to transporters. The remaining proteins are assigned with different functions associated with prophage biology (Table 2).  SPI-8 encodes 16 genes. No interactions among proteins encoded by SPI-8 are described in the published literature. We found by sequence analysis that sty3280-sty3283 encode colicin/pyocin, and sty3274 and sty3277 encode for type VI secretion system (T6SS). The functions of the remaining ten proteins remain unknown. At the early stage of infection, S. Typhi may use pyocin to kill other bacteria in the intestine in order to compete for nutrients. T6SS is used by S. Typhi as a secretion machine to deliver proteins and toxins into the eukaryotic target cell. This is crucial for virulence and survival within the host cells [30].
3.1.9 Salmonella Pathogenicity Island 9 (SPI-9) SPI-9 has 4 genes, oprJ, prtC, prtB, and amyH (Figure 9), which are involved in type I secretion systems (T1SS) [22]. Our sequence alignment analysis found that OprJ (STY2876) has high similarity with TolC, a component of AcrAB, which pumps out bile acids, antibiotics, dyes, and disinfectants [33]. PrtC (STY2878) and PrtB (STY2877) have high similarity with HlyD and HlyB respectively, which are exporters for repeats in toxin (RTX toxin) proteins [34]. AmyH is a homolog of the BapA protein necessary to mediate bacterial recruitment into the biofilm pellicle [35]. BaeR regulates multidrug and metal efflux resistance systems [36] and is a component of the SPI-9 pathway. We show below that BaeR is the major regulator of gene expression in S. Typhi after 8 hours of macrophage infection according to the network analysis of microarray data. It can also synergize with the PhoR/PhoP signaling in E. coli [37]. Our results suggest that in Salmonella, BaeR may synergize with PhoP in response to the acidification of the environment in phagosomes during the infection. SPI-10 has 29 genes that encode a Sef/Pef fimbrial islet, transposases, helicases, IS element, and P4 like-phage proteins [38]. The overview of SPI-10 is illustrated in Figure 10. Three genes of the sef operon (sefA, sefD, and sefR) contain multiple frame-shift mutations. Indeed, microarray data showed that the sef genes are not expressed in S. Typhi [39]. SPI-10 has a truncated pefI gene and lacks pefA, pefB, pefC, and pefD in comparison to the pef operon of S. typhimurium [38]. The presence of P4-like phage, transposase , helicases, IS element, and integrase suggest that this is a hot spot for the insertion of transposable elements which played a major role in driving the variability of this region [38].   Table 4 shows the major transcriptional regulators identified by sub-network enrichment analysis (SNEA) at 2, 8, and 24 hours of macrophage invasion. SNEA is described in the Methods section. We only report and discuss transcriptional factors with p-values smaller than 0.05 as calculated by SNEA. We found that the PhoP transcription factor is active in the beginning of invasion. PhoP is a component of our SPI-1/2/3/4/5 pathways. Interestingly, the period of PhoP activity coincides with the down-regulation of Lrp targets. Lrp is a component of our SPI-1 pathway and is a major expression regulator in Table 1. SNEA identifies major regulators that are either activated or inhibited according to the expression data. The analysis of expression changes for Lrp targets revealed that this global regulator is repressed in the beginning of infection because most of its targets are down-regulated (data not shown). Genes inhibited by Lrp apparently become de-repressed during infection because Lrp is no longer significant after 8 hours. After Lrp targets are de-repressed, PhoP is no longer active. Thus, SNEA results suggest that PhoP appears to initiate the transcriptional program necessary for survival inside macrophage phagosomes together with SlyA (STY1678) transcription factor. SlyA is a component of our SPI-2 pathway.

SPI pathway validation by network enrichment analysis of Salmonella gene expression time-course during macrophage infection
SsrB transcriptional factor is encoded by SPI-2 and remains significant during the entire infection time-course. SNEA results also suggest that integration host factor (IHF) and BaeR transcriptional factor appear to drive up the expression of most differentially expressed genes after 8 hours of invasion (Table 1). IHF is a component of our SPI-7 regulation pathway and BaeR is a component of SPI-9 pathway. RpoN (sigma 54) targets are significantly downregulated throughout the entire time-course. RpoN is a component of our SPI-1/4/5 pathways. We further explain these results in the Discussion section.

Table 4: Most significant transcription factors identified by sub-network enrichment analysis (SNEA) from the time course of Salmonella invasion of human macrophages. pValue of SNEA indicates statistical significance of differential expressed downstream genes targeted by the transcription factors. This in turn signifies the activity of the transcription factor in the experiment.
Expression data was obtained from Gene Expression Omnibus at NCBI (GEO accession number GSE3096). Expression conformity shows how many targets are up-or down-regulated in the right direction relative to the reported activity of the transcription factor (which can be activator or repressor) towards the target.

Validation of pathways by clustering analysis of Salmonella SPI genes during macrophage infection
Co-expressed genes tend to participate in common biological processes [40,41]. Therefore, to further validate our SPI regulatory pathways we have investigated the correlation among expression profiles of genes in our SPI pathways. We have identified a significant number of genes in each SPI pathway with expression correlated during the time-course of Salmonella invasion of macrophages. SPI-1 genes form two distinct gene expression clusters during the time-course of Salmonella infection of macrophages. Expression profile of the biggest cluster SPI-1 is shown in Figure 11a. Gene clusters for other SPI pathways are reported in Figure  12 and 13 respectively. In the figure legend we show how the combination of gene expression clustering and pathway analysis allows the identification of principal transcriptional factors controlling expression of genes co-regulated in the cluster.            Table 9: Description for the genes in Figure 13. The colour of gene corresponds to the colour of the line in the gene expression graph in SPI-7.

Construction and applications of SPI regulatory pathways for Salmonella
To date, 17 SPIs have been discovered in S. enterica [12,43,44,45]. Nine of these SPIs are present in the genome of S.Typhi CT18 and were chosen for pathway reconstruction because experimental data is available to validate them. Our pathways are readily available for the analysis of future experimental data and for comparison of different Salmonella species. In total, our SPI pathways have 463 interactions with 157 of them classified as direct physical interactions. Pathways are consistent with previously published literature on Salmonella infection since only interactions reported in the literature were used for construction. We showed how to use new SPI pathways for analysis of gene expression microarray data inside Pathway Studio software. Because proteins in our SPI pathways are annotated with identifiers from multiple Salmonella species, the pathways can also be used for comparison of invasion mechanisms between different Salmonella strains. Our SPI pathways also provide a list of candidate biomarkers for Salmonella infection. The most suitable biomarkers for clinical diagnostics are proteins secreted and exposed outside the Salmonella cell and induced during the infection. The list of such proteins is readily available from our SPI pathways and can be used for development of diagnostics using ELISA assay. Further challenges associated with the development of diagnostics kit which must be specific to Salmonella species and at the same time provide comprehensive coverage of all enteric species can be addressed by comparison of SPI pathways between invasive S.enterica and other Salmonella species.
While literature suggests that SPI pathways can be activated by different environmental factors such as osmolarity, oxygen level, temperature acidic pH and cation concentration we found that the major factors activating Salmonella infection in macrophages are changes in starvation and osmolarity.
Our pathways also revealed that there is a lack of literature knowledge about SPI-6, SPI-8, SPI-9, and SPI-10 regulation. This knowledge gap does not allow complete reconstruction of the regulatory pathways for these regions and point to the areas for further experimental research, thus helping to develop most efficient research strategy for full understanding of Salmonella invasion mechanism which leads to typhoid fever outbreaks.

Experimental validation of SPI regulatory pathways
We have validated our SPI pathways by comparing them with the publically available microarray data. For comparison, we used statistical methods that have never been used for analysis of GSE3096 dataset. Therefore our network analysis provides novel findings never previously reported. The GSE3096 dataset measures the expression of the entire Salmonella genome and therefore represents an unbiased and independent sample that can be used for cross-validation of any pathways and networks constructed for Salmonella based on the information from other sources. Our only source of data for construction of SPI pathways was Salmonella protein interactions reported in peer-reviewed scientific literature. Most of these interactions were measured either prior to publication of the GSE3096 dataset or were determined by different methods and in different experiments unrelated to GSE3096.
Comparison with GSE3096 showed that the behavior of genes in our SPI pathways is consistent with the current view on Salmonella infection. The SPI-1 pathway is turned on during the first hours of host cell invasion, while the SPI-2 and SPI-3 pathways are necessary for survival inside host cell phagosomes and are activated at later stages of the infection.
Sub-network enrichment analyses of the expression time-course of Salmonella genes during human macrophage invasion identified several transcription factors (PhoP, IHF, SlyA, and Lrp) that were previously shown in the literature to be significant for infection and survival in phagosomes and therefore were components of our SPI pathways. SNEA also found novel significant transcription factors (RpoN and BaeR) that have never been reported playing a role during infection. RpoN is significantly down-regulated during the infection. This is evident from the levels of its mRNA expression as well as from the expression of its targets. One possible biological function of RpoN down-regulation is activation of PhoP transcription factor. PhoP acts upstream of SsrB and is essential for intra-macrophage control of T3SS [46]. PhoP was reported to bind to the ssrB promoter when Salmonella are inside macrophages [46]. It has been shown that RpoN opposes PhoP activation in vivo: the deletion of rpoN attenuates S. Typhi virulence and increases resistance to the cationic antimicrobial peptide polymyxin B [47]. Polymyxin B resistance is mediated by the PhoQ-PhoP system and rpoN deletion appears to act independently from PhoP by providing an alternative mechanism to develop Polymyxin B resistance [47]. Thus, down-regulation of rpoN during macrophage invasion may provide additional boost to PhoP activation.
Identification of major regulators by SNEA combined with analysis of SPI regulatory pathways allows identification of major environmental stimuli used by Salmonella to initiate program of macrophage host invasion. For example, PhoQ-PhoP system can be activated either by acidic pH or by lower concentration of divalent cations (Ca 2+ or Mg 2+ ) according to our SPI-1 pathway [48,49]. Salmonella forms a capsule in the macrophage lysosome to escape host intracellular defense mechanism. The intra-lysosomal environment is very acidic. The link between Mg 2+ concentration, PhoQ-PhoP, and transcriptional regulation of Salmonella invasion genes was reported previously [50]. It has been further suggested that PhoP-activated genes are highly expressed within the host cells due to the low intraphagosomal Mg 2+ concentration and these genes are necessary for intramacrophage survival [51]. The inactivation of Leucine-responsive regulatory protein (Lrp) appears to be noteworthy at the first 2 hours after invasion. It was reported that Lrp is a master regulatory protein that coordinates expression of most bacterial operons in response to nutrient availability [52,53]. It has been reported recently that lrp deletion promotes Salmonella virulence [54]. This is consistent with our findings that lrp is down-regulated after the first 2 hours of infection.
IHF (IhfA) and SlyA are also known SPI-2 regulators [16] and are included in our SPI-2 pathway. Expression of IhfA appears to be significant during 8 hours and 24 hours after invasion. According to [17], IHF was found to be essential for SPI-1 expression at early to late exponential growth phase and IHF levels possibly coordinate the expression of SPI-1 and SPI-2 genes. This is further supported by the previous work by [57] that shows IHF integrates stationary-phase and virulence gene expression and plays a critical role in the co-regulatory process. Expression of SlyA is significant during the first 2 hours after invasion. This is in accordance with the findings by [58] which reported that SlyA regulon is activated during infection of the host and at least 2 proteins expressed in macrophages were found to be SlyAdependent.
The involvement of transcriptional factor BaeR in the invasion process has not been reported previously. BaeR is identified in this work as the major regulator of gene expression in Salmonella after 8 hours of infection. BaeR was shown to regulate multidrug and metal efflux resistance systems [36] and is a component of our SPI-9 pathway. In E.coli, the BaeRS system was shown to influence indirectly the expression of PhoR-PhoB system which is part of our SPI-1 pathway [55]. PhoB is downstream of PhoP and necessary for PhoP regulation of HilA expression according to our SPI-1 pathway. Thus, our results suggest that BaeR can synergize with PhoP in response to the acidification and low cation concentration inside host phagosomes during the infection. It also suggests that Salmonella needs the increased production of multidrug efflux resistance pump in order to survive inside lysosomes, which is a convenient target for anti-typhoid drug development.
SNEA also found other transcription factors from our SPI pathways such as HilA, FlgM, InvF, MarA, and RfaH with p-values higher than 0.05. The p-value range calculated by SNEA depends greatly on the size of the microarray chip, which defines the size of reference distribution of expression values. Smaller chips tend to produce larger SNEA p-values due to the smaller statistical power provided by reference distribution. Therefore, SNEA p-values for smaller chips such as Salmonella genome chip can be used only as a relative rather than absolute measure of transcription factors activity. We reported and discussed only transcription factors with SNEA p-values below the conventional 0.05 cut off emphasizing those that previously were not reported to play a role during macrophage infection. Other transcription factors in our SPI pathways should be active during the infection, suggesting that the 0.05 cutoff was too stringent for the Salmonella chip.

Overview of pathogenicity islands' interaction
The construction of pathogenicity island pathways enables us to identify the higher level interdependencies between SPIs which are regulated by the common global regulators. Understanding of these interdependencies is necessary to predict pathogenicity of different Salmonella strains carrying various combinations of SPI regions in the genome. We found that SPI-1 is interconnected with SPI-4, SPI-5, and SPI-7. Activation of SPI-4 proteins is dependent on the regulators in SPI-1, secretion of SigD/SopB encoded by SPI-5 is via T3SS in SPI-1, and SopE encoded by SPI-7 is also secreted through SPI-1 T3SS. Similarly, SPI-2 is interconnected with SPI-5, whereby PipA and PipB from SPI-5 are secreted through T3SS encoded by SPI-2. STY3274 and STY3277 which are encoded in SPI-8 are related to SPI-6; STY3274 is secreted via T6SS and STY3277 is a T6SS Vgr family protein. SPI-6 and SPI-10 both have chaperon-usher fimbrial operon; saf and sef operon respectively. It was also shown that both SPI-4 and SPI-9 encodes for T1SS [22,25]. Interestingly, genes in SPI-3 are not connected to other pathways but this SPI is controlled by PhoQ-PhoP system which is found in SPI-1, 2, 4, 5, and 7. SPI-3 is very important for the ability of S.Typhi to survive in the macrophage with Mg 2+ limiting conditions. A summary of the interactions between the different SPIs is shown in figure 14.

Conclusion
We have built the collection of nine pathways regulating different stages of S. Typhi infection including host invasion, intracellular host survival, and drug resistance. Our collection shows that nine of the SPIs are interconnected and play an important role for Typhoid Fever. In general, S.Typhi is capable of responding to various environmental challenges such as acidic pH, low temperature, high osmolarity, and in response to divalent cations (Ca 2+ , Mg 2+ , Zn 2+ ). The pathways were validated by analysis of gene expression data. Sub-network enrichment analysis of gene expression has confirmed several major regulators crucial for SPI regulation and identified one novel transcription factor activated during macrophage infection. We have identified several clusters of genes co-expressed during macrophage infection in our SPI pathways.