Transcriptome analysis of HeLa cells response to Brucella melitensis infection: A molecular approach to understand the role of the mucosal epithelium in the onset of the Brucella pathogenesis

Abstract

Brucella spp. infect hosts primarily by adhering and penetrating mucosal surfaces, however the initial molecular phenomena of this host:pathogen interaction remain poorly understood. We hypothesized that characterizing the epithelial-like human HeLa cell line molecular response to wild type Brucella melitensis infection would help to understand the role of the mucosal epithelium at the onset of the Brucella pathogenesis. RNA samples from B. melitensis-infected HeLa cells were taken at 4 and 12 h of infection and hybridized in a cDNA microarray. The analysis using a dynamic Bayesian network modeling approach (DBN) identified several pathways, biological processes, cellular components and molecular functions altered due to infection at 4 h p.i., but almost none at 12 h p.i. The in silico modeling results were experimentally tested by knocking down the expression of MAPK1 by siRNA technology. MAPK1-siRNA transfected cell cultures decreased the internalization and impaired the intracellular replication of the pathogen in HeLa cells after 4 h p.i. DBN analysis provides important insights into the role of the epithelial cells response to Brucella infection and guide research to novel mechanisms identification.

1. Introduction

Brucella is the etiological agent of brucellosis, a worldwide anthropozoonotic infectious disease. Human brucellosis is an occupational-related disease. The highest risk group includes shepherds, animal handlers, farmers and farm workers, butchers, abattoir workers, meat processing plant workers, veterinarians and their assistants, and personnel in microbiologic laboratories. Transmission is associated with accidental contact with infected animals or clinical specimens, inhalation of infected aerosolized particles or foodborne disease associated with the consumption of contaminated animal products [1]. Clinically, human brucellosis is an incapacitating disease that results in intermittent fever, chills, sweats, weakness, myalgia, osteoarthricular complications, endocarditis, depression and anorexia, but low mortality [2]. The severity of the symptoms and signs in humans vary depending on the species of Brucella: B. melitensis causes the most severe and acute symptoms, followed by B. suis, while B. abortus and B. canis tend to produce milder disease and subclinical infections [2]. Among animal species, most mammals are susceptible to brucellosis. Placentitis, abortion and temporary infertility are the principal clinical manifestations of brucellosis in pregnant females. Brucella infection in males causes orchitis and inflammation of the accessory sex organs resulting in permanent or temporary infertility [3].

One of the most attractive topics on Brucella research is to more fully understand the prolonged ability of the pathogen to survive and replicate inside macrophages for long time. It is also important to highlight that Brucella infect susceptible hosts by penetrating mucosal surfaces [4]. Therefore, epithelial cells constitute the first mechanical and immunological barrier against Brucella infection on which few studies have been focused. HeLa cells have been used as a model to understand adhesion, internalization, intracellular trafficking, survival, and replication of brucellae in non-professional phagocytic cells [5–8]. These and other studies have shown that individual Brucella initially attach to non-professional phagocytic cells via receptor molecules containing sialic acid or sulfated residues [5] and within a few minutes are internalized by receptor–mediated phagocytosis [9]. After invasion, Brucella transiently interact with intracellular compartment related to the early endocytic network that is gradually transformed into a multimembranous autophagic vacuole. The expression of virB operon through type IV secretion system (T4SS) allows virulent brucellae to control the maturation of the Brucella-containing vacuole (BCV) to generate a replicative organelle derived from the endoplasmic reticulum in the perinuclear area, where massive intracellular replication occurs [8, 10]. Despite the importance of epithelial cells in the initial Brucella pathogenesis, a detailed molecular response of these cells infected with the intracellular pathogen has not been fully investigated.

Several tools have been developed to study the transcriptional profiles of both pathogen and host [11], the most common of which is cDNA microarray technology. Recently using this approach, we demonstrated that B. melitensis undergo an adaptation period during the first 4 h post HeLa cells infection that is subsequently overcome, facilitating Brucella to replicate intracellularly [12]. With the goal of identifying molecular perturbations in host cells due to B. melitensis infection, we measured the host cells response at 4 and 12 h of infection by a human cDNA microarray, and analyzed the results using a dynamic Bayesian network modeling approach (DBN).

2. Materials and methods

2.1. Cell culture infection and RNA isolation

Eight biological replicas of HeLa cell cultures were infected with a late-log growth phase culture of a virulent B. melitensis 16M, as previously described [12]. Eight other HeLa cell cultures were equally treated with diluent as non-infected controls. Total RNA was extracted from 4 infected and 4 non-infected HeLa cell cultures at 4 and 12 h post-infection (p.i.) using TRI-Reagent® (Ambion, Austin, TX) according to manufacturer’s instructions. Isolated RNA were treated and maintained as previously reported [12].

2.2. Sample preparation and slide hybridization

The labeling and hybridization procedures were adapted from our previous experiments [13]. Briefly, 10 μg of total RNA were reverse transcribed overnight to amino-allyl cDNA using 6 μg of random hexamer primers (Invitrogen), 0.6 μl 50X dNTPs (Invitrogen) / aa-dUTP (Ambion) mix (2:3 aa-dUTP:dTTP) and 400U Superscript III (Invitrogen). cDNA was labeled with Cy5-ester (experimental samples, i.e. infected and non infected samples) or Cy3-ester (human universal human reference RNA, Stratagene, La Jolla, CA) (Amersham Pharmacia Biosciences). After one hour incubation in the dark, uncoupled dye was removed and dye incorporation calculated by NanoDrop^® ND-1000 (NanoDrop). The dried, labeled cDNA samples were re-suspended in 20 μl of nuclease-free water (experimental samples) and human genomic DNA Cot1 (Invitrogen) (reference samples), mixed and heated at 95°C for 10 min, 60°C for 10 min, and then 25°C for 10 min. Samples were kept at 45°C until hybridization. Immediately before hybridization, 40 μl of 2X formamide-based hybridization buffer was added to each sample, well mixed and hybridized to a commercially available 10K human ESTs microarray (Microarray Center, Ontario, Canada). Slides were hybridized at 45°C for ~20 h in a dark, humid chamber (Corning) and washed for 10 min at 45°C with low stringency buffer [1X SSC, 0.2% SDS] followed by two 5-min washes in a higher stringency buffer [0.1X SSC, 0.2% SDS and 0.1X SSC] at room temperature with agitation. Slides were dried by centrifugation at 800X g for 2 min and immediately scanned. Prior to hybridization, microarrays were pre-treated by washing in 0.2% SDS, followed by 3 washes in distilled water and incubated in prehybridization buffer [5X SSC, 0.1% SDS; 1% BSA in 100ml of water] at 45°C for at least 45 min. Immediately before hybridization, the slides were washed 4 times in distilled water, dipped in 100% isopropanol for 10 sec and dried by centrifugation at 1,000X g for 2 min.

2.3. Data acquisition and microarray data analysis

Microarrays were scanned using a commercial laser scanner (GenePix 4100; Axon Instruments Inc., Foster City, CA). The genes represented on the arrays were adjusted for background and normalized to internal controls using image analysis software (GenePixPro 6.0; Axon Instruments Inc.). Genes with fluorescent signal values below background were disregarded in all analyses. Arrays were initially normalized against universal human reference RNA [14] and the resulting data were analyzed and modeled using an integrated platform termed the BioSignature Discovery System (BioSignatureDS^™) (Seralogix, LLC, Austin, TX; www.seralogix.com) explained in detail elsewhere [15–17]. Specifically for the analysis reported herein, the tools were used to: 1) conduct biological system level analysis employing Bayesian network models for scoring and ranking of metabolic and signaling pathways and gene ontology (GO) groups; and 2) conduct Bayesian candidate mechanistic gene analysis to identify genes within the network models that are most responsible for causing pathway and GO group perturbations. Microarray data are deposited in GEO database at NCBI (Accession # GSE14703).

Link for reviewers: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=rdcrpqukokecchw&acc=GSE14703

2.4. Induction of RNAi in HeLa cells

The day before transfection, HeLa cells were cultured in 24 well plates at a concentration of 4 × 10⁴ cells/well in 0.5 ml of cell culture medium and replaced in the incubator. The following day, 50 μl of serum-free cell growth medium (Invitrogen, Carlsbad, CA) was mixed in separate compartments with 30 nM of Silencer® mitogen-activated protein kinase 1 (MAPK1) (ID 1449) Validated siRNA (Ambion, Austin, TX) for each 24 wells of cells to be transfected. Simultaneously, 1 μl of TransFecting lipid reagent (Bio-Rad, Hercules, CA) was diluted into 50 μl of serum-free cell growth medium for each 24-well culture to be transfected. The diluted siRNA was combined and gently mixed with the diluted transfecting reagent. After 20 min incubation at RT, the culture media was removed from the wells and replaced by 100 μl of the siRNA-TransFecting complexes, rocked for 1 min and then filled with 400 μl of F12K cell culture media supplemented with 10% HI-FBS. The next day, the media in the wells was replaced for 0.5 ml of fresh cell culture medium. Forty-eight hours after transfection, HeLa cells were infected with B. melitensis 16M and invasion and survival of intracellular bacteria was determined as described above. Infection of non-transfected cells and HeLa cells transfected with 30 nM Silencer® negative control #1 siRNA (siRNA molecules with no homology on eukaryotic genome) (Ambion) were used as the scrambled control. For validation of RNAi efficiency, RNA from transfected cells was extracted at the same time of infection (i.e. 48 h post-transfection) using RNeasy kit (Qiagen, Valencia, CA) and eluted in 50 μl of DEPC-treated water with 2% DTT and 1% RNase inhibitor (Promega, Madison, WI). RNAs extracted from HeLa cells transfected with 30 nM of Silencer® GAPDH siRNA control and negative control #1 siRNA (Ambion) were used for validation of knocked down gene expression. Contaminant genomic DNA was removed by RNase-free DNase I treatment (Ambion) according to the manufacture’s instructions, and samples were stored at -80°C until used. The concentration of RNA was quantitated by NanoDrop® ND-1000 (NanoDrop, Wilmington, DW), and the quality RNA was assessed using the Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA). Target mRNA levels were measured by qRT-PCR using the following primers (Sigma Genosys, The Woodlands, TX): MAPK1 (Fw 5′-TGGATTCCCTGGTTCTCTCTAAAG-3′, Rv 5′-GGGTCTGTTTTCCGAGGATGA-3′) and GAPDH (Glyceraldehide-3-phosphate dehydrogenase: Fw 5′-AAAAACCTGCCAAATATGATGACA-3′, Rv 5′-AGCTTGACAAAGTGGTCGTTGA-3′).

3. Results

3.1. Dynamic Bayesian modeling analysis of microarray data reveals a scant disturbance of host pathways and GO categories at 4 h that return to a near normal state by 12 h post-B. melitensis infection

To understand the host cells response at the molecular level throughout Brucella infection process, we analyzed the infected host cells transcriptome at the adaptation and the replicative phases of the intracellular pathogen (i.e. 4 and 12 h p.i.) [12]. Four biological replicates of RNA isolated from B. melitensis-infected HeLa cells from every time point (4 and 12 h p.i.) were indirectly labeled and co-hybridized with human universal RNA reference to a commercial 10K human array (n = 8). Gene expression was indirectly compared with RNA isolated from non-infected HeLa cells treated similarly (n = 8). The RNA analyzed from all samples was of good to excellent quality (RNA integrity number (RIN) ≥ 7.0, 28S/18S ratio ≥ 1.6, OD_260/280 ≥ 1.75, OD_260/230 ≥ 1.7).

Mathematical modeling has great potential for discovering and understanding disease mechanisms and biological processes. To better understand the molecular response of host cells after Brucella infection, BioSignatureDS^™ was used to model and score 219 known host metabolic and signaling pathways and more than a 2,000 ontology terms (biological processes, molecular functions and cellular components). The Dynamic Bayesian Gene Group Activation (DBGGA) method which generates Bayesian log likelihood scores that are normalized and transformed to a z-score equivalent (hereafter Bayesian z-score) was employed to identify the perturbations between pathways and GO categories and indicate candidate mechanistic genes.

The analysis identified only 8 pathways significantly altered (Bayesian z-score >|2.24|) at 4 h p.i., but at 12 h p.i., none of the 219 signaling/metabolic pathways associated with gene probes on the human microarrays were perturbed when infected and non-infected host cells were compared (Table 1). Deeper Bayesian analysis identified 37 unique mechanistic genes in the 8 pathways determined to be significantly altered (Table 1). Note that MAPK1 was a mechanistic gene in common with three of the perturbed pathways and was selected for further examination (see paragraph 3.2 below). These data illustrate the potential for cross-talk via inter-pathway interactions and pathways activation in the process of infection and that small alterations in expression of any of these 37 genes can have significant effects on other genes and simultaneously on other pathways, as suggested in the dynamic Bayesian modeling results. As an example, Fig. 1 shows the visualization of the MAPK signaling pathway network model with candidate mechanistic genes highlighted with orange concentric rings. This model was trained with the experiment and control expression data for the two sampling time points. The visualization helps to identify the relationships between genes as well as the expression state of the gene for the selected time point. The MAPK1 gene is identified to have the potential influence on a number of important downstream genes (Fig. 1) such as MYC (v-myc myelocytomatosis viral oncogene homolog) and ELK4 (ETS-domain protein (SRF accessory protein 1).

Table 1. Pathways significantly altered at 4 h p.i. in B. melitensis-infected cells compared with non-infected HeLa cells and key mechanistic genes within each pathway.

This table lists those pathways with a |Bayesian z-score| > 2.24 represents a 98.75% confidence. The key mechanistic genes are those having a |Bayesian z-score| > 1.65 (95% confidence).

Name	Pathway Descriptiona	Category	Pathway Bayesian z-score b 4 h	Key mechanistic genesc
hsa00600	Sphingolipid metabolism	Lipid Metabolism	3.07	GAL3ST1, PHCA PPAP2B
hsa04916	Melanogenesis	Endocrine System	2.92	MAPK1, PRKACB, CALML3, GNAS, DCT, ADCY2
hsa99020	Lectin Induced Complement Pathway	Infectious Diseases	2.82	MBL2, MASP2
hsa00440	Aminophosphonate metabolism	Metabolism of Other Amino Acids	2.68	CHPT1, CARM1
hsa05219	Bladder cancer	Cancers	2.67	MAPK1, THBS1, MDM2, CDKN1A, CDK4
hsa04610	Complement and coagulation cascades	Immune System	2.63	MBL2, F3, F2R C8B, SERPINF2 MASP2, PROS1
hsa04010	MAPK signaling pathway	Signal Transduction	2.34	MAPK1, NFKB2, FLNA, IKBKG, MAP2K5, PRKACB, PRKACB, FGFR3, FGF5, FGF20, CASP8, DUSP14, TGFBR2, TGFB1, DUSP1
hsa05040	Huntington’s disease	Neurodegenerative Diseases	−2.34	CALML3, CASP8, HIP1

Positive Bayesian z-score implies that the pathway is more activated while negative score implies that it is more suppressed

^aPathways are based on Kyoto Encyclopedia of Genes and Genomes (KEGG) descriptions and pathway numbers

^bBayesian z-score: Bayesian log likelihood scores that were normalized and transformed to a z-score equivalent. This score is a one-sided test because it is derived from a log-likelihood modeling score which is a one-sided distribution from the maximum log-likelihood).

^cKey mechanistic genes (determined by Bayesian z-score) intersection points among pathways differentially altered identified by dynamic Bayesian modeling. The mechanistic genes are also determined from a log-likelihood score which were transformed to a one sided z-score.

Figure 1. B. melitensis-infected HeLa cells MAPK signaling pathway with mechanistic genes.

This is an example of the DBN model trained with experimental and control expression data showing the mechanistic genes indicated by concentric rings around the node (each node represents a gene and its expression levels). Increasing up regulation is indicated by the gradient colors from light yellow to dark red and down regulation is from light green to dark green. The thickness of the dark green arcs connecting genes (i.e.,IL1R1->CASP2, FGF7->FGFR2) indicate a strong negative correlation relationship while thicker brown arcs indicate a strong positive correlation (i.e., MAPK1->MYC).

A DBGGA analysis conducted over 2,000 gene ontology (GO) terms, identified 203 GO terms that belonged within the biological processes term (94 up- and 109 down-regulated), 38 belonged to the cellular components term (13 up- and 25 down-regulated) and 28 to molecular functions term (19 up- and 9 down-regulated), highly perturbed (|Bayesian z-score| >2.24) at 4 h p.i. (Tables 2, 3 and 4). However, only 2 biological processes GO categories (vesicle coating –GO:0006901- and regulation of lipid metabolic process –GO:0019216-) but neither cellular component nor molecular function terms were significantly altered (activated) at 12 h p.i. In the ontology of biological processes, the most significant changes occurred in terms related with regulation of biological processes, cell proliferation and developmental processes, metabolic processes, response to stimulus, and immune system processes. The number of terms related with that processes is almost equally activated or repressed in all of them at 4 h post-Brucella infection, except in cell proliferation and developmental processes and immune system processes, where most terms are repressed (Table 2). Examples include central nervous system development (GO:0007417), heart morphogenesis (GO:0003007), embryonic development (GO:0009790) and hemopoiesis (GO:0030097) terms for cell proliferation and developmental processes, while T and B cell differentiation and proliferation (GO:0030217, GO:0030183, GO:0042098, GO:0046651), inflammatory response to antigenic stimulus (GO:0002437), and positive regulation of immune response (GO:0050778) terms are good examples of the repression of the immune system processes. In the GO cellular components, the most significant changes were focused on terms related with nuclear parts (Cajal body –GO:0015030-, chromosome telomeric and centromeric region –GO:0000781 and GO:0000775-, chromosomal part –GO:0044427-), membrane parts (membrane raft –GO:0045121-, plasma membrane part –GO:0044459-, cell projection membrane –GO:0031253-) and cytoskeleton parts (microtubule cytoskeleton –GO:0015630-, actin filament –GO:00005884-) (Table 3), while in the molecular function GO, the most significant changes are equally distributed among binding (RNA binding –GO:0003723-, enzyme binding –GO:0019899-), enzyme regulator activity (small GTPase regulator activity –GO:0005083-, protein kinase regulator activity –GO:0019887-) and catalytic activity (kinase activity –GO:0016301-, pyrophosphatase activity –GO:0016462-) (Table 4).

Table 2. Gene Ontology (GO) biological processes differentially expressed (determined by Bayesian z-score) in infected compare to non-infected HeLa cells at 4 h p.i. identified by dynamic Bayesian modeling.

GO Accession number	Name	Bayesian z-score
GO:0051716	cellular response to stimulus	4.8
GO:0002028	regulation of sodium ion transport	4.55
GO:0007178	transmembrane receptor protein serine/threonine kinase signaling pathway	3.8
GO:0016044	cellular membrane organization	3.55
GO:0010469	regulation of receptor activity	3.53
GO:0009636	response to toxin	3.46
GO:0002682	regulation of immune system process (*)	3.44
GO:0042743	hydrogen peroxide metabolic process	3.42
GO:0019229	regulation of vasoconstriction	3.28
GO:0022607	cellular component assembly	3.21
GO:0006800	oxygen and reactive oxygen species metabolic process	3.14
GO:0042325	regulation of phosphorylation	3
GO:0007219	notch signaling pathway	2.98
GO:0009628	response to abiotic stimulus	2.97
GO:0051341	regulation of oxidoreductase activity	2.95
GO:0007338	single fertilization	2.94
GO:0002688	regulation of leukocyte chemotaxis (*)	2.92
GO:0001824	blastocyst development	2.92
GO:0055067	monovalent inorganic cation homeostasis	2.91
GO:0006875	cellular metal ion homeostasis	2.89
GO:0001990	regulation of systemic arterial blood pressure by hormone	2.88
GO:0051174	regulation of phosphorus metabolic process	2.88
GO:0019220	regulation of phosphate metabolic process	2.88
GO:0051897	positive regulation of protein kinase B signaling cascade	2.88
GO:0030195	negative regulation of blood coagulation (*)	2.88
GO:0006325	chromatin organization	2.88
GO:0010629	negative regulation of gene expression	2.85
GO:0000245	spliceosome assembly	2.76
GO:0001817	regulation of cytokine production (*)	2.76
GO:0002824	positive regulation of adaptive immune response based on somatic recombination of immune receptors built from immunoglobulin superfamily domains (*)	2.76
GO:0008037	cell recognition	2.75
GO:0048709	oligodendrocyte differentiation	2.73
GO:0007566	embryo implantation	2.67
GO:0007193	inhibition of adenylate cyclase activity by G-protein signaling pathway	2.66
GO:0008543	fibroblast growth factor receptor signaling pathway	2.66
GO:0030593	neutrophil chemotaxis (*)	2.65
GO:0043414	macromolecule methylation	2.65
GO:0006958	complement activation, classical pathway (*)	2.64
GO:0006956	complement activation (*)	2.63
GO:0000082	G1/S transition of mitotic cell cycle	2.63
GO:0008217	regulation of blood pressure	2.6
GO:0019935	cyclic-nucleotide-mediated signaling	2.6
GO:0048002	antigen processing and presentation of peptide antigen (*)	2.6
GO:0031128	developmental induction	2.59
GO:0034605	cellular response to heat	2.58
GO:0002504	antigen processing and presentation of peptide or polysaccharide antigen via MHC class II (*)	2.58
GO:0006413	translational initiation	2.56
GO:0042493	response to drug	2.56
GO:0032879	regulation of localization	2.54
GO:0010522	regulation of calcium ion transport into cytosol	2.53
GO:0005482	vesicle targeting	2.5
GO:0043086	negative regulation of catalytic activity	2.5
GO:0030334	regulation of cell migration	2.49
GO:0001775	cell activation	2.49
GO:0051220	cytoplasmic sequestering of protein	2.48
GO:0030005	cellular di-, tri-valent inorganic cation homeostasis	2.48
GO:0051209	release of sequestered calcium ion into cytosol	2.47
GO:0016064	immunoglobulin mediated immune response (*)	2.47
GO:0010038	response to metal ion	2.45
GO:0050865	regulation of cell activation	2.44
GO:0006487	protein amino acid N-linked glycosylation	2.43
GO:0061138	morphogenesis of a branching epithelium	2.43
GO:0051338	regulation of transferase activity	2.42
GO:0016055	Wnt receptor signaling pathway	2.41
GO:0032640	tumor necrosis factor production	2.4
GO:0048610	reproductive cellular process	2.4
GO:0032680	regulation of tumor necrosis factor production	2.4
GO:0001934	positive regulation of protein amino acid phosphorylation	2.4
GO:0031532	actin cytoskeleton reorganization	2.4
GO:0050776	regulation of immune response (*)	2.39
GO:0031325	positive regulation of cellular metabolic process	2.39
GO:0030900	forebrain development	2.38
GO:0050818	regulation of coagulation (*)	2.37
GO:0050817	coagulation (*)	2.37
GO:0000209	protein polyubiquitination	2.36
GO:0006414	translational elongation	2.35
GO:0006796	phosphate metabolic process	2.35
GO:0042308	negative regulation of protein import into nucleus	2.35
GO:0043170	macromolecule metabolic process	2.35
GO:0006508	proteolysis	2.32
GO:0009065	glutamine family amino acid catabolic process	2.31
GO:0008380	RNA splicing	2.31
GO:0016079	synaptic vesicle exocytosis	2.3
GO:0043303	mast cell degranulation	2.3
GO:0044237	cellular metabolic process	2.3
GO:0046824	positive regulation of nucleocytoplasmic transport	2.27
GO:0032270	positive regulation of cellular protein metabolic process	2.26
GO:0048645	organ formation	2.26
GO:0030073	insulin secretion	2.26
GO:0031668	cellular response to extracellular stimulus	2.26
GO:0032526	response to retinoic acid	2.25
GO:0050885	neuromuscular process controlling balance	2.24
GO:0050680	negative regulation of epithelial cell proliferation	2.24
GO:0006396	RNA processing	2.24
GO:0010212	response to ionizing radiation	−2.24
GO:0016192	vesicle-mediated transport	−2.25
GO:0035303	regulation of dephosphorylation	−2.25
GO:0010811	positive regulation of cell-substrate adhesion	−2.25
GO:0030029	actin filament-based process	−2.25
GO:0043010	camera-type eye development	−2.26
GO:0042102	positive regulation of T cell proliferation (*)	−2.27
GO:0060038	cardiac muscle cell proliferation	−2.27
GO:0043627	response to estrogen stimulus	−2.28
GO:0051336	regulation of hydrolase activity	−2.28
GO:0045785	positive regulation of cell adhesion	−2.28
GO:0006898	receptor-mediated endocytosis	−2.29
GO:0070661	leukocyte proliferation (*)	−2.29
GO:0071222	cellular response to lipopolysaccharide	−2.3
GO:0030217	T cell differentiation (*)	−2.31
GO:0050670	regulation of lymphocyte proliferation (*)	−2.32
GO:0001910	regulation of leukocyte mediated cytotoxicity (*)	−2.34
GO:0043408	regulation of MAPKKK cascade	−2.34
GO:0021532	neural tube patterning	−2.34
GO:0006897	endocytosis	−2.34
GO:0006461	protein complex assembly	−2.35
GO:0007602	phototransduction	−2.36
GO:0042113	B cell activation (*)	−2.36
GO:0051817	modification of morphology or physiology of other organism during symbiotic interaction	−2.38
GO:0051851	modification by host of symbiont morphology or physiology	−2.38
GO:0030183	B cell differentiation (*)	−2.39
GO:0007268	synaptic transmission	−2.4
GO:0006687	glycosphingolipid metabolic process	−2.41
GO:0045580	regulation of T cell differentiation (*)	−2.42
GO:0048565	gut development	−2.42
GO:0006664	glycolipid metabolic process	−2.42
GO:0050870	positive regulation of T cell activation (*)	−2.45
GO:0008643	carbohydrate transport	−2.46
GO:0043085	positive regulation of catalytic activity	−2.47
GO:0001666	response to hypoxia	−2.51
GO:0003007	heart morphogenesis	−2.52
GO:0044282	small molecule catabolic process	−2.53
GO:0001912	positive regulation of leukocyte mediated cytotoxicity (*)	−2.53
GO:0045665	negative regulation of neuron differentiation	−2.53
GO:0009100	glycoprotein metabolic process	−2.54
GO:0002831	regulation of response to biotic stimulus	−2.55
GO:0010605	negative regulation of macromolecule metabolic process	−2.55
GO:0033273	response to vitamin	−2.56
GO:0022407	regulation of cell-cell adhesion	−2.57
GO:0007162	negative regulation of cell adhesion	−2.57
GO:0002446	neutrophil mediated immunity (*)	−2.59
GO:0032963	collagen metabolic process	−2.59
GO:0007050	cell cycle arrest	−2.6
GO:0043123	positive regulation of I-kappaB kinase/NF-kappaB cascade	−2.61
GO:0031667	response to nutrient levels	−2.63
GO:0019395	fatty acid oxidation	−2.64
GO:0006665	sphingolipid metabolic process	−2.65
GO:0007283	spermatogenesis	−2.67
GO:0030097	hemopoiesis	−2.67
GO:0060255	regulation of macromolecule metabolic process	−2.7
GO:0002456	T cell mediated immunity (*)	−2.7
GO:0031324	negative regulation of cellular metabolic process	−2.72
GO:0009790	embryonic development	−2.74
GO:0045216	cell-cell junction organization	−2.83
GO:0048584	positive regulation of response to stimulus	−2.84
GO:0000165	MAPKKK cascade	−2.84
GO:0040029	regulation of gene expression, epigenetic	−2.85
GO:0007166	cell surface receptor linked signaling pathway	−2.85
GO:0042490	mechanoreceptor differentiation	−2.87
GO:0002376	immune system process (*)	−2.89
GO:0001816	cytokine production (*)	−2.9
GO:0006919	activation of caspase activity	−2.92
GO:0023052	signaling	−2.93
GO:0050778	positive regulation of immune response (*)	−2.93
GO:0006886	intracellular protein transport	−2.94
GO:0007159	leukocyte cell-cell adhesion	−2.96
GO:0050731	positive regulation of peptidyl-tyrosine phosphorylation	−2.99
GO:0033627	cell adhesion mediated by integrin	−2.99
GO:0042129	regulation of T cell proliferation (*)	−3.01
GO:0046634	regulation of alpha-beta T cell activation (*)	−3.05
GO:0048771	tissue remodeling	−3.1
GO:0051250	negative regulation of lymphocyte activation (*)	−3.1
GO:0043254	regulation of protein complex assembly	−3.12
GO:0001959	regulation of cytokine-mediated signaling pathway	−3.16
GO:0002437	inflammatory response to antigenic stimulus (*)	−3.16
GO:0046631	alpha-beta T cell activation (*)	−3.16
GO:0032355	response to estradiol stimulus	−3.17
GO:0051241	negative regulation of multicellular organismal process	−3.19
GO:0007417	central nervous system development	−3.19
GO:0002709	regulation of T cell mediated immunity (*)	−3.2
GO:0050768	negative regulation of neurogenesis	−3.23
GO:0007405	neuroblast proliferation	−3.24
GO:0051130	positive regulation of cellular component organization	−3.27
GO:0002708	positive regulation of lymphocyte mediated immunity (*)	−3.29
GO:0051251	positive regulation of lymphocyte activation (*)	−3.3
GO:0046651	lymphocyte proliferation (*)	−3.3
GO:0050789	regulation of biological process	−3.4
GO:0042098	T cell proliferation (*)	−3.43
GO:0043406	positive regulation of MAP kinase activity	−3.44
GO:0042130	negative regulation of T cell proliferation (*)	−3.45
GO:0030155	regulation of cell adhesion	−3.47
GO:0030856	regulation of epithelial cell differentiation	−3.51
GO:0002237	response to molecule of bacterial origin	−3.54
GO:0007157	heterophilic cell-cell adhesion	−3.56
GO:0034330	cell junction organization	−3.57
GO:0050850	positive regulation of calcium-mediated signaling	−3.66
GO:0050727	regulation of inflammatory response (*)	−3.68
GO:0044248	cellular catabolic process	−3.71
GO:0045619	regulation of lymphocyte differentiation (*)	−3.74
GO:0001818	negative regulation of cytokine production (*)	−3.79
GO:0050864	regulation of B cell activation (*)	−3.83
GO:0006901	vesicle coating	−4.01
GO:0009607	response to biotic stimulus	−4.03
GO:0034103	regulation of tissue remodeling	−4.4

Positive Bayesian z-score implies that the GO group is more activated while negative score implies a more suppressed state

^(*)Biological processes terms related with immune system processes

Table 3. Gene Ontology (GO) cellular components differentially expressed (determined by Bayesian z-score) in infected compare to non-infected HeLa cells at 4 h p.i. identified by dynamic Bayesian modeling.

GO accession number	Name	Bayesian z-score
GO:0001669	acrosomal vesicle	3.73
GO:0030529	ribonucleoprotein complex	3.35
GO:0015030	Cajal body	2.86
GO:0019861	flagellum	2.73
GO:0030672	synaptic vesicle membrane	2.62
GO:0000180	cytosolic large ribosomal subunit	2.51
GO:0042645	mitochondrial nucleoid	2.5
GO:0000781	chromosome, telomeric region	2.5
GO:0043209	myelin sheath	2.44
GO:0031252	cell leading edge	2.38
GO:0016469	proton-transporting two-sector ATPase complex	2.37
GO:0044455	mitochondrial membrane part	2.36
GO:0045121	membrane raft	2.3
GO:0044427	chromosomal part	−2.27
GO:0005746	mitochondrial respiratory chain	−2.29
GO:0005758	mitochondrial intermembrane space	−2.29
GO:0009986	cell surface	−2.3
GO:0042470	melanosome	−2.33
GO:0043232	intracellular non-membrane-bounded organelle	−2.34
GO:0031901	early endosome membrane	−2.36
GO:0030136	clathrin-coated vesicle	−2.38
GO:0000775	chromosome, centromeric region	−2.39
GO:0015630	microtubule cytoskeleton	−2.4
GO:0044441	cilium part	−2.4
GO:0043292	contractile fiber	−2.5
GO:0005884	actin filament	−2.52
GO:0005624	membrane fraction	−2.56
GO:0046658	anchored to plasma membrane	−2.63
GO:0044459	plasma membrane part	−2.69
GO:0000118	histone deacetylase complex	−2.71
GO:0048471	perinuclear region of cytoplasm	−2.72
GO:0031253	cell projection membrane	−2.88
GO:0005667	transcription factor complex	−2.93
GO:0017053	transcriptional repressor complex	−3.26
GO:0031982	vesicle	−3.38
GO:0005902	microvillus	−3.8
GO:0005901	caveola	−4.15
GO:0031224	intrinsic to membrane	−4.25

Positive Bayesian z-score implies that the GO group is more activated while negative score implies a more suppressed state

Table 4. Gene Ontology (GO) molecular function differentially expressed (determined by Bayesian z-score) in infected compare to non-infected HeLa cells at 4 h p.i. identified by dynamic Bayesian modeling.

GO accession number	Name		Bayesian z-score
GO:0000287	magnesium ion binding		3.88
GO:0003723	RNA binding		3.18
GO:0004402	histone acetyltransferase activity		3.11
GO:0019899	enzyme binding		2.9
GO:0004860	protein kinase inhibitor activity		2.82
GO:0008083	growth factor activity		2.76
GO:0016462	pyrophosphatase activity		2.73
GO:0004713	protein tyrosine kinase activity		2.73
GO:0019901	protein kinase binding		2.73
GO:0015144	carbohydrate transmembrane transporter activity		2.5
GO:0022857	transmembrane transporter activity		2.48
GO:0004659	prenyltransferase activity		2.46
GO:0003735	structural constituent of ribosome		2.45
GO:0005083	small GTPase regulator activity		2.36
GO:0030291	protein serine/threonine kinase inhibitor activity		2.32
GO:0016301	kinase activity		2.32
GO:0043498	cell surface binding		2.31
GO:0008757	S-adenosylmethionine-dependent methyltransferase activity		2.31
GO:0005096	GTPase activator activity		2.29
GO:0046332	SMAD binding		−2.24
GO:0004714	transmembrane receptor protein tyrosine kinase activity		−2.26
GO:0019887	protein kinase regulator activity		−2.34
GO:0019209	kinase activator activity		−2.35
GO:0017016	Ras GTPase binding		−2.36
GO:0016791	phosphatase activity		−2.38
GO:0005160	transforming growth factor beta receptor binding		−2.39
GO:0004871	signal transducer activity		−2.42
GO:0043499	eukaryotic cell surface binding		−3.56

Positive Bayesian z-score implies that the GO group is more activated while negative score implies a more suppressed state

Overall, these results indicate that host metabolic and signaling pathways and GO groups are more pronouncedly perturbed by Brucella infection at very early time post-infection and return to a near normal state by 12 h p.i.

3.2. Role of MAPK1 in Brucella pathogenicity

To evaluate our in silico modeling results, we experimentally tested the role of MAPK1, one of the predicted mechanistic genes during Brucella infection to HeLa cells (Table 1 and Fig. 1). HeLa cells were transfected with MAPK1-validated siRNA molecule and 48 h later were infected with B. melitensis WT. The expression of MAPK1 measured by qRT-PCR in transfected cell cultures were knocked down more than 90% compared with the expression of the gene in non-transfected cells or cells transfected with negative and positive control siRNA molecules (data not shown). The number of viable intracellular Brucella recovered from HeLa cell culture transfected with MAPK1 at T0 was only 40% of the number of bacteria recovered from non-transfected HeLa cells (P < 0.01) (Fig. 2). Additionally, the intracellular replication of Brucella after 4 h p.i. was repressed in HeLa cell cultures transfected with siRNA-MAPK1 compared to controls. These experimental results were in accord with our predictive in silico algorithm and defined the importance of MAPK1 in Brucella replication and survival in non-professional phagocytic cells.

Figure 2. Effect of MAPK-siRNA molecule on the invasion of HeLa cells by Brucella melitensis.

HeLa cells were independently transfected with MAPK1-validated siRNA molecule and 48 h later infected with B. melitensis WT cultures. Non-transfected and cell cultures transfected with siRNA-negative control (siRNA molecules with no homology on eukaryotic genome) were used as controls of infection. The expression of MAPK1 measure by qRT-PCR was knocked-down more than 90% in cells transfected with MAPK1-siRNA molecule. Cell cultures were infected and treated as explained in Materials and Methods. The result suggests the importance of MAPK1 in Brucella invasion to non-professional phagocytic cells. Data are presented as percent of infection and results are the mean + SD (error bars) of 5 independent experiments done in duplicate. Asterisk indicates statistical significant differences (p<0.01) relative to non-transfected cells and cell cultures transfected with siRNA-negative control.

4. Discussion

Using cDNA microarray technology, we have characterized the transcriptional profile of Brucella –infected non phagocytic cells at 4 and 12 h p.i. Recently, we analyzed the transcriptional profile of the intracellular Brucella following Hela cells infection at same time points, which correspond to the adaptation and the replicative phases of the pathogen at the intracellular level [12]. A general overview from the combined analysis of the results from the host and pathogen transcriptional profiles indicates that, while Brucella initially repress and then stimulate its gene expression, the infection initially altered several key pathways and biological processes GO categories in host cells that return to a near normal state by 12 h p.i. All together, our results provide evidence that both non-phagocytic cells and Brucella undergo an adaptation period during the first 4 h p.i. that is overcome by 12 h p.i. permitting Brucella to replicate intracellularly while minimally affecting host physiological processes.

To the best of our knowledge, this is the first study analyzing the transcriptomic profile of non-phagocytic cell response to Brucella infection. Previous studies have described the molecular response of professional phagocytic cells at different time points after Brucella infection [13, 18–21]. Eskra et al. and Covert et al. found quite similar number of genes up- and down-regulated in a murine macrophage cell line infected with different Brucella species at 4 h p.i. [18, 19]. In general, genes associated with immune and inflammatory response were up-regulated, and genes involved in cell cycle/cell division/proliferation and differentiation, intracellular trafficking and metabolism were down-regulated. In another study, He et al. evaluated the transcriptional profile of B. melitensis-infected murine macrophages at different time points. In agreement with our data, they found that the most significant transcriptional changes occurred early after infection (4 h) and returned to normal at later time points (between 24 and 48 h) [20]. Coincidently, in a recent publication Wang et al. also found an active transcriptional activity (> 3,000 host genes) by deep-sequencing analysis of the mouse macrophage response to 2 different B. melitensis strains at 4 h post infection [21]. Altogether, these results suggest that the intracellular presence of Brucella induces an intense early host cell molecular response that normalizes at later time points.

In order to evaluate not only individual genes/proteins, but also to identify which pathways or biological processes were most perturbed in one condition relative to another, we applied the Dynamic Bayesian Gene Group Activation. Three pathways among the 8 significantly altered in B. melitensis-infected HeLa cells (Table 1) such as bladder cancer, Huntington’s disease and MAPK signaling pathway, have been previously found differentially expressed in mouse macrophage response to B. melitensis infection [21]. Mitogen-Activated Protein Kinase (MAPK) signaling pathway has been demonstrated to be implicated in bacterial pathogenesis. Internalization and intracellular survival and replication of different bacterial pathogens are dependent on MAPK pathways activation [22–25]. To test the in silico model and identify the importance of this MAPK signaling pathway in B. melitensis invasion and intracellular survival in HeLa cells, we used siRNA technology to knock-down MAPK1 expression and interrupt the pathway activation. Our results indicated that the internalization of Brucella decreased when the gene was knocked-down and the intracellular replication of the pathogen was impaired after 4 h p.i., highlighting the importance of the pathway integrity to Brucella invasion and survival process in these cells. These results are in agreement with Guzman-Verri et al., who reported that pretreatment of HeLa cells with PD098059, an ERK1/2 pathway inhibitor, resulted in a 50% decrease in Brucella internalization [6]. MAPKs family of proteins regulates cellular activities by phosphorylating target protein substrates such as cytoskeletal proteins. Those authors have demonstrated the participation of GTPases of the Rho/Rac/Cdc42 subfamily in B. abortus internalization in non-phagocytic cells [6]. MAPK1 is in a downstream pathway activated by these small GTPase subfamily proteins [26], thus it is plausible that any interruption of this signaling pathway may adversely impact the process of Brucella invasion in epithelial cells. Interestingly, activation of MAP kinases MEK-1 and ERK-2 genes (also known as MAPK1) is also part of the signaling required for uptake of Listeria monocytogenes by epithelial cells, but not for Salmonella invasion [25]. These findings could probably be related with the very different invasion mechanisms evolved from each pathogen: while Listeria induce a “zipper” mechanism, Salmonella is internalized by a “trigger” mechanism [27]. According to these and previous results [6], Brucella would exploit similar cell signal transduction pathways similar to Listeria for invasion of epithelial cells. Contrary, virulent smooth Brucella do not activate MAPK pathways when invade mouse macrophages [21, 28], probably as a self-defensive mechanism, because MAPK pathway activation contributes to eliminate intracellular Brucella by inducing immune responses [28, 29]. Based on this and previous results, we hypothesize that some Brucella virulent factors other than LPS [28] are activated upon contact to epithelial cells, but not to macrophages, to induce MAPK signaling pathway activation and invasion stimulation. Collectively, these results provide other examples of how pathogens manipulate host metabolic and signaling pathways for it own benefit. Additional studies focused on MAPK pathway regulation during Brucella infection will further clarify the role(s) of this signaling pathway in Brucella pathogenesis.

Particularly important in infectious disease is the host immune response to pathogen invasion. The influence of the epithelium in the initiation of the immune response in Brucella infection has been inadequately studied. Salmonella typhimurium is known to stimulate the Toll-like receptor signaling pathway in intestinal epithelial cells, resulting in IL8 secretion and a massive neutrophil influx into the intestinal lumen [22]. Also Legionella pneumophila, another intracellular pathogen, is reported to induce secretion of several cytokines from the lung epithelium after infection, which contributes to the immune response in legionellosis [30]. Brucella has developed a stealth strategy that allows it to reach its replication niche before activation of antimicrobial mechanisms through the immune response [31]. In a recent publication, Ferrero et al. showed that Brucella-infected human bronchial epithelial cell lines collaborate in mounting a host innate immune response [32]. An overview of our DBGGA analysis indicates that GO biological processes terms related with the immune system process had reduced expression. Deeper analysis found that “complement and coagulation cascades” pathway as well as biological processes GO terms related to complement activation, coagulation cascade and antigen processing and presentation were significantly activated, while those related to lymphocyte differentiation and proliferation were repressed (Table 2, terms marked with *). MHC-I is present in all nucleated cells of the body, while MHC-II expression is restricted to immune cells. Lapaque et al. have shown that Brucella LPS has no effect on MHC class I antigen presentation in infected macrophages [33]. This is in agreement with our in silico modeling results and may indicate a pathogen manipulation of host defensive mechanisms, as it was observed that cells with reduced levels of MHC-I molecules are target of NK cells. Also in support of our analysis, Brucella infection consumes complement, although less than when compared to Salmonella [31]. Moreover, Brucella LPS is directly involved in deficient CD4+ T cells activation [34] which is consistent with our data that indicating repressed lymphocyte differentiation, proliferation and activation.

In conclusion, we have documented the molecular response of the epithelium-like HeLa cell line during the first 12h post-Brucella infection. Our results indicate that Brucella infection perturbs epithelial cell biology at very early time post-infection and then returns to near normal state by 12 h p.i. without significantly interfering with the fate of the pathogen. In vivo, Brucella invade and traverse the epithelium layer of susceptible hosts and are endocytosed by mucosal macrophages. Based on our results and published literature, we propose that quickly after invasion, Brucella drive the infected cells through an active transcriptional activity toward an adaptive period that appears to be necessary and crucial for successful persistence of intracellular pathogen. During this adaptive period Brucella not only regulate their own survival but also modulate host epithelial cells response for their benefit. For example, here we showed that MAPK1 expression is necessary for Brucella invasion and intracellular replication in non-phagocytic cells. Simultaneously, at very early times post-infection Brucella modulate host adaptive immune response through epithelial cells, down-regulating the lymphocytes differentiation, proliferation and activation (i.e. adaptive immunity). All these mechanisms contribute to the ability of Brucella to quickly adapt to an intracellular life and reach their replication niche, where they initiate replication without significant disruption of host processes. Further analysis of the computational integrated results of host and pathogen molecular response, along with experimental confirmation are expected to further unravel the initial molecular pathogenesis of Brucella.