Genome-wide transcriptional profiling of peripheral blood leukocytes from cattle infected with Mycobacterium bovis reveals suppression of host immune genes

Background Mycobacterium bovis is the causative agent of bovine tuberculosis (BTB), a pathological infection with significant economic impact. Recent studies have highlighted the role of functional genomics to better understand the molecular mechanisms governing the host immune response to M. bovis infection. Furthermore, these studies may enable the identification of novel transcriptional markers of BTB that can augment current diagnostic tests and surveillance programmes. In the present study, we have analysed the transcriptome of peripheral blood leukocytes (PBL) from eight M. bovis-infected and eight control non-infected age-matched and sex-matched Holstein-Friesian cattle using the Affymetrix® GeneChip® Bovine Genome Array with 24,072 gene probe sets representing more than 23,000 gene transcripts. Results Control and infected animals had similar mean white blood cell counts. However, the mean number of lymphocytes was significantly increased in the infected group relative to the control group (P = 0.001), while the mean number of monocytes was significantly decreased in the BTB group (P = 0.002). Hierarchical clustering analysis using gene expression data from all 5,388 detectable mRNA transcripts unambiguously partitioned the animals according to their disease status. In total, 2,960 gene transcripts were differentially expressed (DE) between the infected and control animal groups (adjusted P-value threshold ≤ 0.05); with the number of gene transcripts showing decreased relative expression (1,563) exceeding those displaying increased relative expression (1,397). Systems analysis using the Ingenuity® Systems Pathway Analysis (IPA) Knowledge Base revealed an over-representation of DE genes involved in the immune response functional category. More specifically, 64.5% of genes in the affects immune response subcategory displayed decreased relative expression levels in the infected animals compared to the control group. Conclusions This study demonstrates that genome-wide transcriptional profiling of PBL can distinguish active M. bovis-infected animals from control non-infected animals. Furthermore, the results obtained support previous investigations demonstrating that mycobacterial infection is associated with host transcriptional suppression. These data support the use of transcriptomic technologies to enable the identification of robust, reliable transcriptional markers of active M. bovis infection.


Background
Bovine tuberculosis (BTB) poses a serious threat to the health of domestic cattle herds worldwide. Infection is caused by the bacterium Mycobacterium bovis, an intracellular pathogen closely related to Mycobacterium tuberculosis-the causative agent of human tuberculosis. M. bovis infection is often slow and progressive with limited clinical symptoms. Although improved diagnostic tests and slaughter policies have done much to control and reduce the incidence of infection, BTB has remained recalcitrant to eradication in many countries where control programmes have been implemented [1][2][3].
Failure to detect and remove all infected animals from herds is partly due to limitations in the sensitivity of the current diagnostic tests, which often comprise an in vivo single intradermal comparative tuberculin test (SICTT) performed alone, or in combination with an in vitro enzyme-linked immunosorbent assay (ELISA)-based test for interferon gamma (IFN-γ)-an established biomarker of mycobacterial infection [4][5][6]. Diagnoses can be further confounded by exposure to environmental nonpathogenic mycobacterial antigens, which can generate false SICTT-positive signals in cattle [7]. Protection from natural M. bovis infection in cattle may be achieved through vaccination with M. bovis bacillus Calmette-Guérin (BCG); however, the level of protection attained is variable. In addition, current diagnostics cannot effectively differentiate between M. bovis-infected and BCG-vaccinated animals, thus compromising management strategies [8]. Consequently, there is a pressing need for novel M. bovis diagnostic methods with increased sensitivity and specificity.
The host immune response to mycobacterial infection is a complex process that involves interaction between the innate and adaptive immune systems. Upon initial exposure (generally via inhalation), bacilli are phagocytosed by host alveolar macrophages, which recognise mycobacteria using a diverse range of pathogen recognition receptors (PRRs), such as the Toll-like receptors (TLRs) and the nucleotide-binding oligomerisation domain (NOD)-like receptors (NLRs) [9][10][11][12][13]. Activation of macrophage PRR-mediated signalling pathways result in the release of endogenous cytokines, which initiate an adaptive immune response characterised by the secretion of proinflammatory cytokines, such as IFN-γ and tumour necrosis factor (TNF-α), by activated T cells [14]. In particular, IFN-γ activates infected macrophages and enables the formation of granulomas-collections of inflammatory cells comprising T cells, B cells, noninfected macrophages and neutrophils, which surround infected macrophages and act as barriers to contain and prevent dissemination of the infection [15]. In most cases, the host innate and adaptive immune systems successfully control mycobacterial growth within granulomas resulting in asymptomatic latent infection [16,17]. However, in some cases impairment of immune function can result in the development of active tuberculosis leading to disease progression [3,[17][18][19][20].
Recently, functional genomic technologies have been used to investigate the molecular mechanisms and cellular pathways underlying the host immune response to mycobacterial infection [ [21,22], for reviews see [23,24]]. Furthermore, results from these studies have the potential to identify molecules that are critical for host/pathogen survival during infection, and which may serve as robust, reliable transcriptional markers of mycobacterial infection [22].
Previously, we investigated the transcriptional profiles of peripheral blood mononuclear cells (PBMC) from M. bovis-infected and non-infected control animals using the immuno-specific BOTL-5 microarray (containing 1,391 gene probe sets; Gene Expression Omnibus [GEO] accession: GPL5751) and showed that suppression of innate immune genes was associated with BTB [25]. In the current study, we extend this earlier work by investigating the transcriptional profile of peripheral blood leukocytes (PBL) isolated from eight M. bovis-infected and eight non-infected control animals using the genomewide high-density Affymetrix ® GeneChip ® Bovine Genome Array. These 16 animals were sampled specifically for the present study and have not been used for any previous research work. The Affymetrix ® GeneChip ® Bovine Genome Array contains 24,072 gene probe sets representing more than 23,000 gene transcripts http:// www.affymetrix.com. In addition, we have adopted a systems biology approach using the Ingenuity ® Systems Pathway Analysis (IPA) Knowledge Base http://www. ingenuity.com for analysis of both over-represented cellular functions and known molecular canonical pathways from the resulting gene expression data.
The results presented in the current study contribute a novel layer of information regarding the gene expression profile of PBL from M. bovis-infected animals and highlight the value of high-throughput genomic technologies in understanding the host immune response to BTB. Furthermore, these results may facilitate the development of novel diagnostics for the detection of M. bovis infection in domestic herds.

Experimental animals
Sixteen age-matched female Holstein-Friesian animals from cattle herds that had not been analysed previously were used for this study. Eight infected individuals were selected from a panel of naturally M. bovis-infected animals maintained for on-going disease surveillance at the Irish Department of Agriculture, Fisheries and Food, Backweston Laboratory Campus (Celbridge, Co. Kildare, Ireland). These animals had a positive single intradermal comparative tuberculin test (SICTT) result where the skin-fold thickness response to purified protein derivative (PPD)-bovine exceeded that of PPD-avian by at least 12 mm. All of these animals were also positive for the whole blood IFN-γ-based BoviGAM ® assay [Prionics AG, Switzerland] (data not shown). In addition, these cattle were confirmed for BTB following detailed postmortem pathological examination and/or culture. Briefly, bronchial, mediastinal, submandibular, retropharyngeal, mesenteric and hepatic lymph nodes and lungs were examined macroscopically for tuberculosis lesions. Suspected lesions were cultured on Stonebrinks and Lowenstein-Jensen media at 37°C for eight weeks to detect M. bovis [26]. Non-infected control animals were selected from a herd with no recent history of M. bovis infection. The control animals were shown to be negative for both the SICTT and IFN-γ tests (data not shown). All animal procedures detailed in this study were carried out according to the provisions of the Cruelty to Animals Act (licenses issued by the Department of Health and Children) and ethics approval for the study was obtained from the UCD Animal Ethics Committee.

Blood collection
Two 8 ml vacutainers ® (Becton-Dickinson Ltd., Dublin, Ireland) of heparinised blood were collected from each animal, approximately 12 months after positive SICTT testing. One vacutainer ® was retained for haematological analysis using a Cell-Dyn 3500 haematology analyser (Abbott Laboratories Ireland Ltd., Dublin, Ireland); all haematological analysis was performed using 1 ml of blood. The other vacutainer ® was used for RNA isolation from peripheral blood leukocytes (PBL); the whole white blood cell fraction consisting of T and B lymphocytes, NK cells, monocytes, neutrophils, basophils and eosinophils. The count data from the leukocyte cell populations of infected and non-infected animals were assessed using the two-sample, two-tailed Student's ttest, following Kolmogorov-Smirnov tests of normality and Levene's F-test for equality of variance using the Minitab statistical package version 16 (Minitab Ltd., Coventry, UK).

RNA extraction and microarray analysis
All RNA extractions were performed within two hours of blood collection. Briefly, 7.5 ml of whole heparinised blood was mixed with 42.5 ml of erythrocyte blood lysis buffer (10 mM KHCO 3 , 150 mM NH 4 Cl, 1 mM EDTA pH 8.0), and incubated for 5 min at room temperature with gentle agitation. Following centrifugation (750 g for 10 min) the pelleted cells were washed once with 1× phosphate buffered saline (Invitrogen Ltd., Paisley, UK). The cell pellet was then fully resuspended in 2 ml Trizol ® reagent (Invitrogen Ltd., Paisley, UK) and RNA was extracted as per the manufacturer's instructions. The RNA was further purified using an RNeasy ® kit with on-column DNase treatment (Qiagen Ltd., Crawley, UK) according to the manufacturer's instructions. RNA quantity and quality was assessed using both the Nano-Drop™ 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and the Agilent 2100 Bioanalyzer using an RNA 6000 Nano LabChip kit (Agilent Technologies, Cork, Ireland). All samples displayed a 260/280 ratio greater than 1.8 and RNA integrity numbers (RIN) greater than 8.0. cDNA labelling, hybridisation and scanning for the microarray experiments were performed by Almac Diagnostics (Craigavon, Co. Armagh, Northern Ireland) using a one-cycle amplification/labelling protocol on the Affymetrix ® GeneChip ® Bovine Genome Array (Affymetrix UK Ltd., High Wycombe, UK).

Statistical analysis of microarray data
Affymetrix ® GeneChip ® Bovine Genome Array data were analysed using Bioconductor [ [27]; http://www. bioconductor.org] contained within the R statistical package http://www.r-project.org. Normalisation of raw data was performed using the Factor Analysis for Robust Microarray Summarization (FARMS) algorithm. The FARMS algorithm uses only perfect match (PM) probes and a quantile normalization procedure, providing both P-values and signal intensities [28]. In addition, gene expression profiles for each animal were clustered using the Hierarchical Ordered Partitioning and Collapsing Hybrid (HOPACH) clustering algorithm in Bioconductor with Euclidean distance as the distance metric [29].
Normalised data were then further subjected to filtering for informative probes sets using the R package I/ NI-calls [30]. This defines a probe set as being informative when many of its probes reflect the same change in mRNA concentration across arrays. Differentially expressed genes were extracted using the Linear Models for Microarray Data (LIMMA) package http://bioconductor.org/packages/release/bioc/html/limma.html contained within the R statistical package. Genes displaying differential expression patterns between control and infected groups were annotated using the Affymetrix ® bovine gene annotation http://www.affymetrix.com. The Benjamini-Hochberg multiple-testing correction method [31] was applied to all differentially expressed genes to minimise the false discovery rate (FDR) and adjusted Pvalues for differentially expressed (DE) genes were calculated. For genes represented by multiple probe sets the mean expression value is reported. Bootstrapping for the hierarchical clustering analysis was performed using a custom Perl script (see Additional file 1).

Systems biology analyses
Ingenuity ® Systems Pathway Analysis (IPA, Ingenuity Systems, Redwood City, CA, USA; http://www.ingenuity. com) was used to identify canonical pathways and functional processes of biological importance within the list of DE genes. The Ingenuity ® Knowledge Base contains the largest database of manually-curated and experimentally-validated physical, transcriptional and enzymatic molecular interactions. Furthermore, each interaction in the Ingenuity ® Knowledge Base is supported by previously published information.
For the IPA analyses, the Affymetrix ® GeneChip ® Bovine Genome Array was used as a reference gene set. All DE genes with an adjusted P value ≤ 0.05 were included. For duplicate probe IDs, the average log 2 expression fold change was used. Only DE genes mapping to molecules in the Ingenuity ® Knowledge Base were used for systems analysis. Functional analysis of genes was performed using IPA to characterise the biological functions of the DE genes between the BTB and control groups. For this, IPA performed an over-representation analysis that categorises the DE genes within the uploaded list into functional groups using the Ingenuity ® Knowledge Base. Each category in IPA is ranked based on the number of DE genes falling into each functional group. Right-tailed Fisher's exact tests were used to calculate a P-value for each of the biological function assigned to list of DE genes.
IPA contains a large library of known canonical pathways that were overlaid with the DE genes to identify major biological pathways associated with M. bovis infection in PBL. The significance of the association between DE genes and the canonical pathway was assessed using two methods: (1) a ratio of the number of molecules from the DE gene data set that map to the pathway, compared to the total number of molecules that map to the canonical pathway based on the reference gene list; and (2) a Fisher's exact test that generates a P-value for the assignment of the DE genes to a particular canonical pathway compared to the reference gene list. Canonical pathways were then overlaid with the expression values of the DE genes.
Real time quantitative reverse transcription PCR (qRT-PCR) validation of microarray results cDNA was prepared using 500 ng of total RNA from each sample from the microarray study using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Warrington, UK) in a 20 μl reaction using random primers according to the manufacturer's instructions. cDNA was diluted 1:50 and stored at -20°C prior to performing real time quantitative reverse transcription PCR (qRT-PCR).
Real time qRT-PCR reactions were performed using Fast SYBR ® Green Master mix (Applied Biosystems, Warrington, UK) on a 7500 Fast Real-Time PCR System apparatus (Applied Biosystems, Warrington, UK). Each reaction (20 μl) contained 5 μl of the diluted cDNA (equivalent to 2.5 ng of total RNA) and 300 nM final concentration each of forward and reverse primer. Additional file 2, Table S1 provides a complete list of primer sequences and the target accession numbers for each real time qRT-PCR amplicon analysed. Real time qRT-PCR primers were designed using the Primer3 software [32,33] and where possible intron-spanning primers were selected (see Additional file 2, Table S1). Negative real time qRT-PCR controls and a six-point, four-fold dilution series from pooled cDNA from all animals were included on every real time qRT-PCR plate and individual PCR efficiencies were determined from the standard curves using the qbase PLUS software package [ [34]; Biogazelle NV, Zwijnaarde, Belgium].
The PCR thermal cycling program consisted of one cycle at 50°C for 2 min, one cycle at 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min. A dissociation step was included to confirm amplification specificity and real time qRT-PCR products were analysed on a 2% agarose gel to confirm the presence of a single discrete amplicon of the correct size. All real time qRT-PCR data was analysed using the qbase PLUS software package with efficiency correction and normalization was performed using two reference genes: the 60S ribosomal protein L19 gene (RPL19) and peptidylprolyl isomerase A (cyclophilin A) gene (PPIA). The two reference genes were selected using the geNorm algorithm in the qbase PLUS package from a panel of eight genes tested (geNorm M > 0.15).
The qbase PLUS package generated a calibrated normalized relative quantity (CNRQ) of gene expression for each of the analysed samples. Log 2 CNRQ values for both the control and the M. bovis-infected animals were used for statistical analysis for all genes. One sample Kolmogorov-Smirnov tests, performed using the SPSS ® version 18 software package (SPSS ® Inc., Chicago, IL, USA) were applied to the residuals of the log 2 CNRQ values for each sample prior to statistical analysis to ensure the data conformed to a normal distribution-no significant departures from normality were observed for any of the genes analysed (P ≥ 0.05). Two-tailed, twosample Student t-tests were used to assess differences between infected and control groups based on log 2 CNQR values using SPSS and the statistical package contained within Microsoft ® Excel 2010 (Microsoft Corp., Redmond, WA, USA). In addition, Levene's Ftest was applied to the infected and control group log 2 CNRQ values to assess equality of variance within the two groups using SPSS prior to performing the twosample t-tests; equality of variance was observed for each sample group.
Geometric mean fold-changes in gene expression for the BTB group were calculated by dividing the geometric mean CNRQ value for the M. bovis-infected group by the geometric mean CNRQ value for the control group. The negative reciprocal in fold-change is reported where decreased gene expression was observed in the BTB group relative to the control group.

Analysis of leukocyte populations in control and M. bovisinfected blood samples
The infected animals used in this study were chosen on the basis of their large responses to the comparative tuberculin skin test. The IFN-γ levels measured in whole blood of the infected animals were at least 25-fold greater than in the healthy control cattle (P < 0.001, data not shown), demonstrating that the infected animals were generating strong cell-mediated immune responses. At post-mortem, each of the infected animals displayed gross tuberculosis lesions in the lungs and thoracic lymph nodes and were classified as being in the advanced stage of clinical disease.
To assess potential changes in leukocyte composition between control and M. bovis-infected samples, whole blood samples were subjected to haematological analysis ( Figure 1). No significant difference was observed in the total white blood cell (WBC) count between control and M. bovis-infected animals (P = 0.18). However, significant increases in the mean number of lymphocytes (P = 0.001) and significant decreases in the mean number of monocytes (P = 0.002) were observed in M. bovis-infected animals relative to the control animals. No significant differences in the mean number of eosinophils and neutrophils were observed between the two sample groups (P = 0.51 and P = 0.37, respectively).
Summary of differentially expressed genes between control and M. bovis-infected animals identified from Affymetrix ® GeneChip ® analysis The expression data generated for the current study are MIAME-compliant [35] and were deposited in the NCBI Gene Expression Omnibus (GEO) repository [36] with experiment series accession GSE33359. All array transcripts that passed informative probe filtering were used for cluster analysis to examine the grouping of samples based on infection status ( Figure 2). In total, 5,388 transcripts passed the filtering process and were used for the cluster analysis, which showed a clear partitioning of samples based on their disease status, indicating a distinct difference in expression profile between the two sample groups. The division of the two sample groups was supported by a 100% bootstrap value after 1,000 permutations.
Genome-wide transcriptional profiles generated from the PBL of eight M. bovis-infected and eight control animals were compared to assess differential gene expression between the two sample groups. The microarray analysis revealed a total of 2,960 transcripts, representing 2,757 unique genes, as differentially expressed between the M. bovis-infected and non-infected control animals (adjusted P value ≤ 0.05). In summary, 1,397 Figure 2 Hierarchical clustering dendrogram generated using the complete gene expression profiles for the M. bovis-infected and control animals. The dendrogram was generated from the complete gene expression dataset for each animal using the Euclidean distance metric. The division between the BTB and control animal groups was supported by a bootstrap value of 100%. Real time quantitative reverse transcription PCR (qRT-PCR) analysis and validation of Affymetrix ® GeneChip ® results A panel of 23 immune-related genes, including members of the interferon signalling pathway, which was recently shown to be the most significantly over-represented pathway in human patients with active TB [37], were selected for real time qRT-PCR analysis. These were CASP1, CD83, CTLA4, DEFB10, IFNB, IFNAR1, IFNG,  IFNGR1, IFNGR2, IL1A, IL8, IL15, JAK1, KIR3DS1,  MYD88, PTPN2, STAT1, STAT2, TLR3, TLR4, TREM1,  TYK2 and TYROBP. The immune-related function of each gene and the results from these analyses are detailed in Table 1.
Twenty-two of the 23 genes analysed by real time qRT-PCR were represented on the microarray; only TYK2 was not represented. Gene expression profiles for 16/22 (73%) of the genes analysed with real time qRT-PCR were concordant with the results from the microarray analysis. Two genes (IFNGR2 and TLR4) displayed significant expression differences between the two sample groups based on the microarray results, but were not significantly different based on real time qRT-PCR data analysis. Three genes (IFNAR1, IFNGR1 and PTPN2) displayed significant differences between the two groups based on real time qRT-PCR results; however, these genes were not differentially expressed according to the microarray results. Only one gene (STAT1) displayed directionally discordant gene expression profiles between the two methods. In the M. bovisinfected group, STAT1 showed significant reduced relative expression (-1.27-fold) based on real time qRT-PCR results but displayed significant increased relative expression (+1.28-fold) according to the microarray results. The observed discrepancies between the microarray and real time qRT-PCR data may reflect differences in the sensitivity of the two analytical methods used and/or differences in the mRNA transcripts targeted by the probes (microarray) and primer pairs (real time qRT-PCR) used for the two forms of gene expression analysis.
Analysis of differential gene expression using Ingenuity ® Systems Pathway Analysis (IPA) The total number of DE genes that could be mapped to molecules in the Ingenuity ® Knowledge Base was 1,869 from a total of 2,960 DE transcripts. IPA was used to categorise these 1,869 DE genes based on their functional annotation and to assess if a functional gene category contained an over-representation of genes relative to the microarray reference gene list. This analysis showed that the top functional category observed for the DE genes used for IPA was the inflammatory response, which contained 241 genes with P-values ranging from 1.66 × 10 -11 to 1.64 × 10 -2 ( Table 2). The inflammatory response category was further sub-divided with the affects immune response subcategory containing the most molecules (138 DE genes) [ Figure 3]. Further inspection of the individual genes within this subcategory revealed that there was an over-representation of genes displaying reduced relative expression (89 genes) compared to DE genes showing increased relative expression in the BTB animals (49 genes). This observation was in contrast to the other top functional categories where similar numbers of genes showing increased and decreased relative expression were reported. Canonical molecular pathways associated with M. bovis-infection were analysed using IPA. These canonical pathways were ranked according to P-value, which represents the significance of the association between a specific pathway and the genes in the input data set. The majority of the top ranking IPA-identified canonical pathways were involved in cell signalling and communication associated with host innate and adaptive immune responses. Natural killer cell signalling and communication between innate and adaptive immune cells were identified as the top ranking canonical pathways. In addition, TREM1 signalling, dendritic cell maturation, JAK-STAT signalling, T cell signalling, IL6 signalling, chemokine signalling and TLR signalling were among the top twenty IPA-identified canonical pathways (Table 3). Based on the well-documented role of TLR signalling in mycobacterial infection [13,[38][39][40][41][42], this canonical pathway overlaid with gene expression results is shown in Figure 4.

Discussion
The implementation of surveillance and management programmes has done much to reduce the incidence and prevalence of BTB over the past number of decades; however, M. bovis infection remains an important livestock disease worldwide. This is due, in part, to welldocumented limitations of the currently available diagnostics tests (such as the SICTT and IFN-γ tests) leading to a failure to detect all infected animals [43,44].
In recent years, research has shifted from a focus on protein-based diagnostics to functional genomics technologies that interrogate the host transcriptome in response to M. bovis infection. In particular, microarray technologies coupled with the rapid development of more sophisticated bovine genome resources has enabled high-resolution analyses of the genes and cellular pathways governing the host response to infection with M. bovis [23,25,[45][46][47]. In the present study, we have compared the transcriptomes of PBL from noninfected control animals with actively-infected BTB animals using a high-density genome-wide bovine microarray platform.
Modulation of the host PBL transcriptome in response to M. bovis infection was evident from the large number Gene descriptions taken from GeneCards version 3 [82] and the NCBI Entrez Gene database [83]. For the microarray data, genes with fold-change differences in expression were significant after adjustment for multiple testing using the Benjamini-Hochberg method [31] (adjusted P ≤ 0.05). 'Not DE' indicates a gene was not differentially expressed between the two sample groups. of DE genes between the two experimental groups. Statistical analyses of the microarray data identified a total of 2,757 DE genes. Of these, 1,281 (46%) showed increased expression and 1,476 (54%) displayed decreased expression in the BTB group compared to the control animals. It is important to note, however, that the differences in cell subpopulations observed between the M. bovis-infected and control animals ( Figure 1) may have contributed to the gene expression changes detected between the two experimental groups. Also, the haematology analyser results only provided a general description of the PBL cell subsets and do not provide information concerning T lymphocyte subsets in the infected and control animal groups. In addition, the cell  Nervous system development and function 4.59 × 10 -5 -1.18 × 10 -2 8 Reproductive system disease 5.84 × 10 -3 -7.54 × 10 -3 79 Cell cycle 5.94 × 10 -3 -1.18 × 10 -2 20 Endocrine system disorders 5.94 × 10 -3 -5.94 × 10 -3 3 Opthalmic disease 5.94 × 10 -3 -6.60 × 10 -3 7 RNA post-transcriptional modification 6.04 × 10 -3 -6.04 × 10 -3 13 Cardiovascular system development and function 6.68 × 10 -3 -1.30 ×10 -2 19 Cell assembly and modification 6.68 × 10 -3 -1.28 × 10 -2 14 Organ development 6.68 × 10 -3 -1.19 × 10 -2 6 Renal and urological system development and function 1.18 × 10 -2 -1.30 × 10 -2 11 Reproductive system development and function 1.18 × 10 -2 -1.18 × 10 -2 4 The top ranking GO categories identified by IPA are listed according to P-values. P-value ranges are based on subcategory P-values within each parent term.
subset results presented here differ from previous work performed by us [25], most likely due to the different cell sample types examined. Analysis of the DE genes using IPA provided information regarding the immunobiology of active BTB. The highest ranking functional category identified using IPA was inflammatory response and the highest ranking subcategory within this category-affects immune responserevealed a marked bias in the number of genes displaying a decrease in relative expression (64.5%) compared to those showing an increase in relative expression (35.5%) in PBL from the BTB group.
Previous work by our research group demonstrated that suppression of host gene expression was associated with active M. bovis infection in cattle [25]. This earlier work involved the comparison of RNA isolated from PBMC of M. bovis-infected and control animals using an immuno-specific bovine cDNA microarray (BOTL-5). The results presented here, based on the analysis of PBL-derived RNA using a genome-wide microarray, lend further support to our previous study. Indeed, published investigations by other workers suggest that transcriptional suppression is a common feature of mycobacterial infection in mammals [48][49][50].
Further inspection of the inflammatory response functional category in IPA identified several genes that were previously reported to be differentially expressed in cattle and other mammalian species infected with mycobacterial pathogens [23,40,42,51,52]. For example, microarray analysis showed that TLR2 and TLR4 displayed contrasting expression patterns between PBL from the two groups: TLR2 showed decreased expression and TLR4 showed increased expression relative to the control animal group. We have previously observed decreased expression of TLR2 in PBMC from actively infected BTB cattle using the immuno-specific BOTL-5 cDNA microarray; however, contrary to the results of the present study, TLR4 also showed decreased expression with the BOTL-5 cDNA microarray in actively infected animals [25]. The gene expression results obtained by Meade and colleagues using PBMC from M. bovis-infected and control non-infected animals were also used to identify a panel of 15 genes predictive of disease status [25]. Four of these genes were found to be similarly differentially expressed in the current study: UNC84B (now SUN2), GAN, SFPQ and NRP1. Four other of the 15 genes identified previously (TBK1, 28S [now RN28S1], GPR98 and an anonymous BOTL clone [BOTL0100013_F01]) were not present on the Affymetrix ® GeneChip ® Bovine Genome Array. However, the seven remaining genes (NCOR1, PPP2R5B, UCP2, ZDHHC19, NFKB1, NRM and FGFR1) were not differentially expressed in the PBL samples from M. bovis-infected and control non-infected animals used for the present study. This discordance may be due to a number of factors, including: the blood cell sample types used (PBL versus PBMC); differences in sensitivity between the two types of microarray (the single-colour in situ-synthesised Affymetrix ® GeneChip ® versus a dual-colour spotted cDNA array [53,54]); and the requirement for more stringent control of the FDR with larger numbers of genes (24,072 probe sets versus 1,391 duplicate spot features).
These results suggest that genes encoding TLRmediated signalling pathway molecules have a role in governing the host response to BTB and may also serve as targets for immuno-subversion by M. bovis. For example, genes encoding several innate immune receptors and chemokines (such as TLR4, CD83, CCL2, CXCR4, CXCL5, IL1A and IL8)-several of which participate in the initiation of a T cell response during infection [61][62][63][64]-showed increased relative expression in the BTB animal group. Transcriptional profiles suggesting initiation of a T cell response are supported by the comparative analysis of the PBL cell populations in the two animal groups; a significant increase in the mean number of lymphocytes and a significant decrease in the mean number of monocytes were observed in the BTB group relative to the control animals. This difference in the PBL cell composition may represent recruitment of host cytotoxic lymphocytes for the destruction of infected monocytes in the control of M. bovis infection [3,65].
It is important to note, however, that the observed decreased expression of host PRR genes (such as TLR2) and the genes encoding their associated adaptor and signalling pathway molecules (such as MYD88, IRAK4, MAPK13 and MAPK14) may indicate that the adaptive response in BTB animals is inferior due to the repression of these innate immune genes. Indeed, previous work has proposed that mycobacterial antigens, such as the early secreted antigenic target protein 6 (ESAT-6) protein, attenuates the host innate immune response by inhibiting MYD88-IRAK4 binding, thus causing suppression of NF-B-induced transcription of upstream genes required for T cell response initiation [66]. These workers also demonstrated that activation of v-akt murine thymoma viral oncogene homolog kinases (AKTs) is necessary to prevent MYD88-IRAK4 complex formation. Notably, the AKT2 gene displayed increased relative expression (+1.22-fold) in the BTB animal group in the present study.
Repression of host innate immune genes that elicit an adaptive response to M. bovis infection is further supported by the analysis of genes belonging to the interferon signalling pathway, which has been shown to have a role in human tuberculosis [37,[67][68][69][70][71]. IFN-γ is secreted by NK cells and CD4 + T cells upon activation by IL-12 produced by infected macrophages. IFN-γ recruits additional macrophages to the site of infection while also providing the stimulus for activating microbicidal functions in infected macrophages [14,[71][72][73]. IFNγ also induces MHC class II gene expression in infected macrophages by signalling through its receptor (IFN-γreceptor) [74][75][76]. This stimulates the JAK-STAT pathway, resulting in induction of transcriptional activators of MHC class genes, such as the MHC class II transactivator gene (CIITA). Mycobacterial antigen presentation via MHC class II molecules is critical for the recruitment of additional CD4 + T cells and the formation and maintenance of granulomas [38].
The results from the current study support a role for interferon signalling pathways during M. bovis infection. The genes encoding interferon (alpha, beta and omega) receptor 2 (IFNAR2); interferon gamma receptor 2 (IFNGR2); interferon-induced protein with tetratricopeptide repeats 2 (IFIT2); interferon-induced protein with tetratricopeptide repeats 5 (IFIT5); interferon-induced transmembrane protein 3 (IFITM3); protein tyrosine phosphatase, non-receptor type 2 (PTPN2); and signal transducer and activator of transcription 2 (STAT2) displayed differential expression in the BTB animals based on the microarray and/or real time qRT-PCR analyses.
These findings suggest that, in addition to the targeting of TLR-mediated signalling pathways, M. bovis may also target genes involved in the IFN-signalling pathway, resulting in an attenuated T cell response that enables mycobacterial survival and disease progression. It is tempting to speculate that suppression of IFN-signalling in response to M. bovis infection may result in the impairment of the antigen presenting process required for adaptive immunity; however, further work is required to investigate this possibility. Notably, the gene encoding IFN-γ (IFNG) was not differentially expressed in the current study, despite the BTB animals having tested positive for increased IFN-γ based on the Bovi-GAM ® assay. However, it is important to note that unlike the blood samples used for the BoviGAM ® assay in the current work, the PBL fraction from which the RNA was derived in this study was not stimulated with protein purified derivative of tuberculin (PPD), which is required to elicit IFN-γ secretion [4]. In addition, contrary to previous results obtained by Meade and colleagues [25] we did not detect differential expression of the TNF gene between the M. bovis-infected and control animals examined here. The most likely explanation for this apparent discrepancy is the different cell sample types used for gene expression analyses (PBL versus PBMC).
IPA canonical pathway analysis identified a number of DE genes which, to our knowledge, have not previously been reported to be involved in the host response to tuberculosis in cattle or other mammalian species. These included CTLA4 and TLR3. TLR3 encodes an intracellular PRR involved in the recognition of viralderived nucleic acids [77]. In the present study, reduced relative TLR3 expression in BTB animals (-1.55-fold) may suggest some hitherto unknown role for this PRR in intracellular mycobacterial infection. In support, we have observed significant differential expression of TLR3 in bovine monocyte-derived macrophages (MDM) stimulated in vitro with M. bovis when compared to nonstimulated control MDM (unpublished data). CTLA4 encodes an inhibitor of the T cell-mediated response [78,79] and this gene displayed increased relative expression (+3.20-fold, P ≤ 0.05) in the M. bovis infected animals in the present study. The observed increased relative expression of CTLA4 may reflect a mechanism of immuno-modulation used by M. bovis to subvert a host T cell response.
Finally, hierarchical clustering analysis was performed here using a total of 5,388 genes that passed the informative probe filtering criteria. This analysis unambiguously differentiated animals on the basis of their disease status. This result suggests that genome-wide expression profiling of PBL from BTB animals can be used to enable the identification of suitable transcriptional markers for the detection of infected animals within herds and augment current surveillance strategies in countries where control programmes have been implemented [21,22]. However, further work using PBL samples from additional animals infected with M. bovis and other microbial pathogens will be required to identify and validate robust M. bovis-specific transcriptional signatures of infection.

Conclusions
The results presented here support the hypothesis that repression of immune-related genes is an important feature of mycobacterial infections [25,[48][49][50]80,81]. In particular, the gene expression results obtained suggest that M. bovis infection may target the innate immune cellular pathways necessary for the initiation of the appropriate T cell response. Notably, analysis of the cell populations present in the PBL from the BTB animals showed an increase in the number of lymphocytes relative to the control animals, suggesting that the activelyinfected BTB animals do mount a T cell response. However, it is possible that the T cell response elicited by these animals is compromised, resulting in disease progression. Indeed, failure of the adaptive immune response to contain the mycobacterial infection is regarded as the primary cause of the development of active tuberculosis from a latent state of infection [74]. Finally, cluster analysis using all informative mRNA transcripts permitted a clear delineation between healthy and infected animals. These results demonstrate that functional genomics approaches based on transcriptional profiling can be used to supplement current proteinbased diagnostics for BTB.

Additional material
Additional file 1: Perl script used for bootstrapping of cluster analysis results in Figure 2.
Additional file 2: Table S1: Oligonucleotide primers used for real time qRT-PCR validation of microarray results.