Introduction Neisseria meningitidis (Nm) is a Gram-negative commensal of the human upper respiratory tract and asymptomatic carriage of Nm in the nasopharynx is common in healthy adults. In susceptible individuals, Nm can cause septicemia by crossing the mucosal barrier and entering the bloodstream, or can cause meningitis by crossing the blood–brain barrier and multiplying in the cerebrospinal fluid [1]. Invasive meningococcal infections represent a major childhood disease with a mortality rate of 10% and high morbidity in survivors [2]. During the transition from colonization to an invasive bloodstream infection, Nm must adapt to changing environments and host factors. Sequencing of different Neisseria genomes has facilitated the discovery of many previously unknown virulence factors [3]-[7] and the comparison of disease and carrier strains has recently provided new insights into the evolution of virulence traits in this species [6]. In order to better understand how Nm adapts to different interactions with the host, it is necessary to study the gene expression of the bacterium under conditions that approximate the human niches it encounters in vivo. The interactions of Nm with human epithelial and endothelial cells, as well as exposure to human serum, have been analyzed using microarray expression studies, which have provided useful information about the pathogenesis of the bacterium and the function of previously unknown genes, and have also enabled the identification of novel vaccine antigens [8]. However, little is known about how Nm adapts to permit survival and growth in human whole blood, despite the importance of this step in the disease process. An infant rat model of invasive infection has been combined with a signature tagged mutagenesis (STM) approach to identify genes essential for bacteremia [9]. However, Nm is an exclusively human pathogen, and existing animal models may not accurately simulate meningococcal disease. This justifies the use of an experimental system that mimics, as closely as possible, the in vivo situation seen during disease. Human whole blood has been used as an ex vivo model of sepsis for studying the pathogenesis of Nm in terms of complement activation, cytokine production and immunity [10]–[14]. Similar ex vivo models have also been used to understand how pathogens, including Candida albicans, Listeria monocytogenes, group A and group B Streptococcus species, regulate gene expression during exposure to human blood [15]–[18]. In this study we have analyzed the global changes in the transcriptional profile of a virulent Nm serogroup B (NmB) strain in an ex vivo model of bacteremia, using incubation in human whole blood and a time-course oligo-microarray experiment. This approach revealed mechanisms used by Nm to adapt to human blood, and was instrumental in analyzing the role of previously known and newly identified virulence factors whose expression was up-regulated during ex vivo infection. Results and Discussion Transcriptome analysis of Nm gene expression in an ex vivo human whole blood model In order to evaluate the transcriptional response of Nm during growth in blood we used an ex vivo human whole blood model, which enabled meningococcal responses to both host cellular and humoral bactericidal mechanisms to be analyzed. This ex vivo model has shown potential to examine a number of parameters that are likely to be important in the cascade of events associated with acute systemic meningococcal infection [19] and to characterize Nm factors involved in the survival of the bacterium during infection [20], [21]. Freshly isolated whole venous blood collected from four healthy human volunteers (two male and two female) was used. Bacterial loads in patients with fulminant disease can reach up to 109 bacteria/ml [22]–[24]. Therefore, Nm MC58 bacteria (approximately 108, grown in GC medium to early exponential phase) were mixed with blood from each donor in order to mimic disease. Analysis of growth in the blood by colony forming unit (CFU) counting showed that bacterial numbers increased approximately 2-fold over a 90-minute incubation period and that there was no significant difference in the number of CFU between the four donors (Figure 1A). 10.1371/journal.ppat.1002027.g001 Figure 1 Growth of Nm in human whole blood and RNA analysis. (A) Number of bacteria during incubation with human blood. The CFU/ml per single donor is shown during a time course experiment. (B) Analysis of isolated total RNA and enriched Nm RNA using a BioAnalyzer 2100 (Agilent). Upper panel: Total RNA collected from Nm incubated in human whole blood, bacterial RNA (shaded arrowheads) and eukaryotic RNA (open arrowheads) are indicated. Lower panel: Enriched bacterial RNA. In order to evaluate the adaptation of Nm to human blood, samples were collected at six different time points (each time point consisted of triplicate cultures): immediately after mixing bacteria with blood (time 0, reference point), and after 15, 30, 45, 60 and 90 minutes incubation at 37°C. Total RNA extracted at each time point consisted of a mix of eukaryotic and prokaryotic RNA (Figure 1B). Since eukaryotic RNA can compete with bacterial RNA during cDNA synthesis and fluorochrome labeling, we used a procedure that simultaneously removes mammalian rRNA and mRNA [25]–[27]. With this procedure, we were able to significantly enrich the samples for Nm prokaryotic RNA (Figure 1B). We then applied an in vitro transcription amplification/labeling step [28] to produce amplified-labeled cRNA that was then used in competitive hybridization experiments with a 60-mer Nm oligo-microarray. Transcriptional changes throughout the course of Nm incubation in human blood were defined by comparison of expression levels at various time points against time 0 (Figure 2A). Variability between the four blood donors was quantified by measuring the Pearson correlation coefficient ‘r’ between the expression matrices of each pair of donors (r coefficients between pairs of donor samples (i,j = 1-4): r1-2 = 0.77, r1-3 = 0.78, r1-4 = 0.79, r2-3 = 0.74, r2-4 = 0.87, r3-4 = 0.70). We also evaluated the Pearson correlation ‘rijg ’ at the level of each single gene ‘g’ and we represented the distribution of the coefficients both globally and between pairs of donors (Figure S1). This analysis showed an excellent agreement between the gene expression profiles obtained from the four donors. The four data sets were averaged in order to obtain a single data set that was subsequently used to evaluate global gene expression changes (Figure 2B). 10.1371/journal.ppat.1002027.g002 Figure 2 Global changes of Nm gene expression in human whole blood. (A) Experimental design. Human blood isolated from four different donors was incubated with Nm and RNA extracted at the indicated time points (samples from each time point was done in triplicate and then pooled). Time 0 was used as the reference time point. (B) Hierarchical clustering of the differentially expressed genes showing the data of the four different donors (Donors 1–4) and the average dataset (Merge). Clustering showed two well defined partitions of the expression profiles, 360 up-regulated (red) and 277 down-regulated genes (green). Genes were selected based on a fold change of at least two (log2 ratio 1) and a t-test p-value 1 cut-off, is relatively stable in each time point and varies between 24 and 28 genes. The selection criterion was also compared with the results obtained with BETR statistics, which is specifically suitable to discover regulated genes during a time course. The BETR algorithm confirmed 509/637 genes as significantly (p-value 1 or 0.6). (B) Same analysis as performed in A but comparing pairs of data sets between the single blood donors shown in single box-plots. (C) Quartile statistics values are reported in the table for pairs of donors shown in panel B. (TIF) Click here for additional data file. Figure S2 Validation of microarray data by qRT-PCR. (A) Comparison of microarray (white bars) and qRT-PCR (black bars) fold change results for 9 selected genes. Fold change qRT-PCR ratios represent the difference in transcript abundance/signal for these genes after 45 minutes of incubation in human whole blood as compared to time 0. qRT-PCR normalization data was done using 16S rRNA as a reference gene. The qRT-PCR and microarray data are represented as the mean of gene expression of four independent experiments. (B) Correlation analysis of microarray and qRT-PCR transcript measurements for the nine selected genes shown in panel A. The qRT-PCR and microarray log2 values were plotted and the coefficient of correlation was calculated, r = 0.98. (TIF) Click here for additional data file. Figure S3 Distribution of differentially regulated genes within TIGRFAM sub-roles and KEGG pathways. (A) TIGR families sub-roles and (B) KEGG pathways. The groups with five or more differentially regulated genes are represented. Darker bars indicate the most relevant enrichments among the included groups (Fisher s exact test p-value < 0.05). (TIF) Click here for additional data file. Figure S4 Characterization of the Nm deletion mutants. (A) Schematic representation of the allelic replacement with the resistance cassette, used to generate the deletion mutant strains. The gene locus of each deletion mutant was amplified from the genomic DNA using primers specific for each amplicon (indicated as 1 and 2) and the PCR fragments were sequenced using the same primers and primers 3 and 4 for the antibiotic cassette Kan or Ery. (B) The orientation of the resistance cassette for each deletion mutant, determined from sequencing. C. Southern blot analysis was performed using labeled Kan and Ery PCR products as probes. Genomic DNA of wild-type and deletion mutant strains was digested with BglI. The size of the expected fragment for each mutant is reported in panel B. The length of the fragment is approximate since the BglI restriction site is subjected to dam methylation that could occur along the genomic DNA. (EPS) Click here for additional data file. Figure S5 Survival of MC58 wild-type and deletion mutant strains in the ex vivo whole blood model using a second blood donor. Deletion mutants of the selected up-regulated genes were tested for survival using the ex vivo whole human blood model over a time course of 120 minutes. In each panel the phenotype of the specific mutant is compared to MC58 wild-type strain. The MCDfHbp deletion mutant was used as a control. The insets of each panel represent the growth control in GC medium for the same time course of incubation as done with whole blood. (TIF) Click here for additional data file. Table S1 List of genes belonging to the different K-means partitioning clusters and TIGRFAMs main roles. Provided as an Excel file. (XLS) Click here for additional data file. Table S2 Significant correlation (Fisher s exact test p-value < 0.05, underlined if significant with Benjamini-Hochberg correction) between K-means partitioning clusters and TIGRFAM main roles and KEGG pathways. Provided as an Excel file. (XLS) Click here for additional data file. Table S3 Primers used in this study. (DOC) Click here for additional data file. Table S4 Plasmids used in this study. (DOC) Click here for additional data file. Table S5 Nm wild-type, deletion mutants and complementing strains used in this study. (DOC) Click here for additional data file.