Significant gains have been made in our understanding of host–pathogen interactions during Salmonella infection
. Several studies, involving a variety of different species of farm and model animals, have investigated the host response to Salmonella infection and have successfully identified genes differentially expressed upon infection or gene variants and chromosomal loci associated with immune response traits during infection with Salmonella[7, 20, 27, 28]. A genetic basis for differences in resistance to Salmonella has also been shown with SLC11A1 (NRAMP1), the seminal example of a gene with genetic variants dramatically affecting resistance to bacterial infection
[29, 30]. While many studies have looked into the host response to Salmonella infection
[4, 20, 28], relatively few have focused on identifying the genes whose variable expression among different individuals may be associated with differences in Salmonella clearance and resistance.
Initial characterisation of the pigs used in an earlier Salmonella challenge study
 revealed a significant positive correlation between serum interferon-γ (IFN-γ) levels at day 2 p.i. and faecal Salmonella shedding levels at day 2 and day 7 p.i. The same study also demonstrated that the peak of both clinical symptoms (fever, diarrhea, decreased appetite) and Salmonella shedding occurs at day 2 p.i. and that substantial whole blood transcriptome changes occur at day 2 p.i. compared to day 0 in pigs belonging to both LS and PS groups. Therefore, here we chose to profile whole blood transcriptomes at the same time-points, day 0 and day 2, but belonging to a different set of pigs, using RNA-Seq instead of microarrays, for a different purpose: to identify genes whose expression prior to inoculation is correlated with Salmonella shedding levels observed p.i. The genes identified may serve as blood-based candidate biomarkers that could potentially be used to develop quick screening tests for determining the host’s resistance/susceptibility to Salmonella infection and predicting their shedding characteristics early into or even before infection.
The extent of DE and degree of change in expression between day 0 and day 2 were, in general, higher in PS than LS. This finding, as previously speculated
, may indicate that LS animals respond faster and more effectively against infection than PS so that by day 2, while the LS are returning back to normal, the PS are still actively fighting the infection. With this in mind we believe that the comparison between LS and PS at day 2 likely identifies DE due to differing levels of infection. A comparison at day 0, on the other hand, stands to better highlight genes responsible for differences in the efficacy of the initial response to the bacteria, assuming that some of these genes exhibit expression differences prior to infection. However, the DE analysis between LS and PS at day 0 did not yield any significant DE genes here as well as in an earlier Salmonella challenge study
. This failure to find DE genes could be due to a combination of the subtlety of the expression differences, the relatively small sample size, and the strict multiple testing corrections. Hence we used an alternative approach, WGCNA, to find genes associated with Salmonella shedding.
Remarkably, the genes whose pre-inoculation expression profiles were found associated with post-inoculation Salmonella shedding levels included major genes already reported in literature as DE during Salmonella infection and involved with host resistance against Salmonella such as SLC11A1, TLR4, CD14 and CCR1[4, 20]. Moreover, the majority of genes within the modules significantly associated with Salmonella shedding, following further refinement based on up-regulation and DE at day 2 (Additional file
9), were found to have an established or possible role in innate defense against bacterial/Salmonella infections. These include mainly the early innate immune response genes responsible for cytokine-cytokine receptor interactions (CXCL16, CCR1, CCR3, CNTF, CSF2RA, IFNGR1, TNFSF9, TNFSF12, TNFSF13, TNFSF13B, TNFRSF1A, TNFRSF1B, IL1A, IL15, IL18, IL1RAP, IL1R2, IL18RAP), genes involved in toll-like receptor pathway (TLR4, TLR2, MYD88, CD14, LY96), NF-Kappa B signalling pathway (NFKBIA, TNFRSF1A, TNFSF13B), NOD-like receptor signalling pathway (CASP1, PSTPIP1, IL18, NFKBIA) or otherwise linked to response to bacterial infections (SLC11A1, SERPINB1, S100A8, S100A9, S100A12, ARG2, CEBPB). Prioritisation of these genes based on gene significance, module membership and gene connectivity in LS at day 0 (Table
2) highlights some genes which have not been previously associated with Salmonella infection. For example, little is known about the role of SIGLEC5, a member of the Siglec family of sialic acid-binding lectins in host response to bacterial infection. However, it has been reported in humans that the absence of a functional SIGLEC14 (with which human SIGLEC5 shows extensive sequence similarity) results in attenuated cytokine response to some Gram-negative bacteria in null individuals
The mean expression levels of the Salmonella shedding associated genes at day 0 were generally higher in LS than PS and mostly up-regulated in both at day 2 compared to day 0. In most instances, the expression was higher in PS than LS at day 2. We showed, at least for the top candidate genes reported here, that the pattern of expression is consistent with that from a previous microarray based Salmonella challenge study involving a different set of LS and PS animals. Examining the connectivities of genes within the Salmonella shedding associated modules in LS and PS, it became apparent that the genes in general showed higher connectivity in LS than PS, indicative of higher correlation/connection strengths with other network genes. The differences in connectivity measures for a set of genes between different conditions may signify differences in the co-ordination or strength of transcriptional regulation of that set of genes. Highly connected genes (hub genes) have been shown to play central roles in the biological processes that are represented by the module
, and strong positive correlations have been reported between gene connectivity within the whole network and gene essentiality
. Here, the significantly higher connectivity despite the lack of significant DE between LS and PS may be considered analogous to the results in a study on breast cancers of different histological grades
. The authors of that study concluded that the differential connectivity patterns were not due to primary alterations of hub gene expression, but rather due to more subtle changes in expression of numerous genes interacting with those hubs. Further, they reported that complex epistatic interactions that underlie cellular functions might also be responsible for differences in network connectivity patterns as a function of a phenotypic trait. A study on aging in mice
 reported a decreasing correlation of gene expression within genetic modules and attributed this to changes in expression of certain transcription factors (TF) as well as deterioration of chromatin structure with age. It is possible that genetic differences at mutiple levels as discussed above may contribute to the differences in strengths of coexpression and connectivity between LS and PS. Exploring these contributions may be a direction for future research.
One of the limitations of our study is the absence of samples from time points post-inoculation but before day 2 p.i., which are crucial to capture the early immune response during which the LS pigs have effectively managed the Salmonella challenge. Secondly, this study would benefit from a larger sample size, which would provide more power to detect the subtle changes in expression expected between LS and PS animals. Further experiments are required to rank the relative functional importance of our suggested candidate genes as contributors to distinct responses to Salmonella challenge with respect to faecal Salmonella shedding. However, the use of multiple criteria and strict cut-offs to refine the set of candidate genes, the agreement with existing literature regarding the immune related functions of many candidate genes and the concordance of the expression patterns of top candidate genes reported here with the corresponding expression patterns from an independent dataset, all lend further support to our predictions.