Research article | Open | Published:
Simultaneous host and parasite expression profiling identifies tissue-specific transcriptional programs associated with susceptibility or resistance to experimental cerebral malaria
BMC Genomicsvolume 7, Article number: 295 (2006)
The development and outcome of cerebral malaria (CM) reflects a complex interplay between parasite-expressed virulence factors and host response to infection. The murine CM model, Plasmodium berghei ANKA (PbA), which simulates many of the features of human CM, provides an excellent system to study this host/parasite interface. We designed "combination" microarrays that concurrently detect genome-wide transcripts of both PbA and mouse, and examined parasite and host transcriptional programs during infection of CM-susceptible (C57BL/6) and CM-resistant (BALB/c) mice.
Analysis of expression data from brain, lung, liver, and spleen of PbA infected mice showed that both host and parasite gene expression can be examined using a single microarray, and parasite transcripts can be detected within whole organs at a time when peripheral blood parasitemia is low. Parasites display a unique transcriptional signature in each tissue, and lung appears to be a large reservoir for metabolically active parasites. In comparisons of susceptible versus resistant animals, both host and parasite display distinct, organ-specific transcriptional profiles. Differentially expressed mouse genes were related to humoral immune response, complement activation, or cell-cell interactions. PbA displayed differential expression of genes related to biosynthetic activities.
These data show that host and parasite gene expression profiles can be simultaneously analysed using a single "combination" microarray, and that both the mouse and malaria parasite display distinct tissue- and strain-specific responses during infection. This technology facilitates the dissection of host-pathogen interactions in experimental cerebral malaria and could be extended to other disease models.
Cerebral malaria (CM) is a major cause of global morbidity and mortality. The molecular basis of this syndrome remains incompletely defined ; however, the development, severity and ultimate outcome of malaria infections are known to be influenced by genetic factors in both the host and the parasite [1–3]. Murine models have been informative in clarifying the molecular mechanisms underlying CM. Plasmodium berghei ANKA (PbA) infection of mice simulates many of the features of human CM  and infection of genetically defined inbred mice provides an opportunity to dissect the host response to infection. C57BL/6 (B6) and other susceptible mice [4–6] infected with PbA develop malaria-associated encephalopathy similar to CM in humans. In contrast, BALB/c mice do not develop encephalopathy, although they become infected and achieve similar levels of parasite density . Host factors, including differential intensity and timing of pro- and anti-inflammatory cytokine responses to infection, have been implicated in susceptibility to CM in this model .
Parasite-dependent factors also influence disease outcome. PbA infection of B6 mice results in CM, while infection of the same host with the closely related parasite line P. berghei K173 does not . By comparing these models, it has been postulated that parasite-dependent modulation of host immune responses may contribute to the pathogenesis of CM .
Genome-wide expression profiling is being increasingly applied to dissect the complex details of the host response to malaria infection [10–16]. Delahaye et al. identified candidate mouse resistance genes to PbA by profiling brain gene expression in resistant and susceptible mice . Sexton et al. examined transcriptional responses in the spleens of B6 mice infected with PbA and reported gene expression patterns suggestive of suppressed erythropoiesis, up-regulated host glycolysis and an interferon-inducible target response . Several studies have also analyzed parasite expression patterns in vitro and in vivo [17, 18]; however, concurrent examination of the host-parasite interaction has only been used to analyze the vector-parasite relationship between Anopheles stephensi and P. berghei .
Malaria infections result in diverse clinical outcomes presumably because of the dynamic relationship between parasite-expressed virulence factors and individual host response to these determinants. The objective of this study was to simultaneously examine both sides of the parasite-host interface to identify corresponding PbA and murine organ-specific expression profiles associated with resistance or susceptibility to CM. We used custom-designed "combination" microarrays containing both murine and plasmodial genes to define host and parasite transcriptional programs in target organs during PbA infection of CM-susceptible (B6) and resistant (BALB/c) mice. Malaria gene expression was readily detected within host tissue, and differential expression of parasite and murine genes discriminated infection in resistant from susceptible mice, identifying both host and parasite transcriptional programs which may contribute to CM.
Results and discussion
A combination murine/P. berghei ANKA microarray
We designed and tested a microarray (the "combination array") composed of 42,034 sixty-mer probes designed to detect over 20,000 known and predicted mouse transcripts  as well as 17,313 known or predicted P. berghei genes and ESTs. Because both the mouse and malaria genomes are still being annotated, and the statistical analyses employed are sensitive to artefacts that can arise from duplicated genes, and inclusion of pseudogenes and unverified genes, we only considered probes in our computational analyses that uniquely target 9,035 single-copy mouse genes and 8,577 independent P. berghei genes or ESTs (henceforth referred to collectively as "genes") annotated in current genome databases.
To test the specificity of our combination array, we hybridized one microarray with Cy3-labelled cDNA from uninfected B6 brain and Cy5-labelled cDNA from PbA infected B6 blood, containing approximately 20% parasitized red blood cells. In uninfected brain, the spot intensity distribution from malaria probes was indistinguishable from that of random-sequence negative control probes on the array, while 20% of the mouse probes exceeded the 99th percentile of negative control probe intensities (Figure 1), presumably representing the fraction of genes expressed in brain. In contrast, 25% of the malaria probes exceeded the 99th percentile of negative control probe intensities in infected blood. This indicates that the malaria probes were functional under our standard hybridization conditions and had minimal cross-hybridization to mouse cDNAs. Therefore, the combination array was able to specifically detect parasite and host gene expression.
Host and parasite tissue-specific expression patterns
We next hybridized combination arrays with cDNA from four malaria target organs (lung, brain, liver, and spleen) from B6 (susceptible) and BALB/c (resistant) mice over a time-course of infection (pre-infection [day 0], at an early time point with low parasitemia [day 3] and just prior to the development of CM in susceptible mice with high parasitemia [day 6]). In total, we detected 2,829 mouse genes with expression above the 99th percentile of negative-control probes in at least one tissue, among which 2,268 were detected at least 6-fold higher in at least one tissue relative to the gene median expression across all tissues (Figure 2A). We confirmed that these transcripts displayed tissue-specific gene expression patterns similar to those previously described , including enrichment of expected Gene Ontology Biological Process (GO-BP) categories  among the genes expressed in different tissues (Figure 2A). For example, genes highly expressed in the spleen were significantly associated with the GO-BP categories of 'Immune Response' and 'Response to Biotic Stimulus'.
We detected 5,632 malaria genes with expression above the 99th percentile of negative-control probes in at least one tissue, among which 2,159 were detected at least 15-fold higher in at least one infected tissue relative to uninfected tissue. This indicated that they were unlikely to represent cross-hybridization to mouse transcripts in the same samples and were roughly consistent with the number of malaria genes previously found to be expressed in infected blood [17, 22]. PbA transcripts also displayed an organ-specific "signature" of gene expression that was modulated over the course of infection (Figure 2B).
Strikingly, the majority of PbA transcripts with detected expression were present in the lung (Figure 2B). Although information about the PbA genes is sparse, the lung-expressed genes encoded apparent heat shock proteins, ribosomal proteins, and proteasome components, and the cluster is significantly enriched with GO-BP annotations of nucleic acid metabolism, entry into host, and metabolism (P < 0.05). Further analysis of the lung-expressed genes, using P. falciparum Protein-Protein Interaction (PPI) sub-networks identified by LaCount et al. , showed that a number of gene pairs in the PPI network were significantly over-represented in the group of PbA genes expressed in the lung (P < 0.001), including some associated with cell invasion (P < 0.05) [See Additional file 1]. These findings suggest that the lung may be a preferential site of PbA biosynthesis, metabolism and proliferation. Although this intriguing observation may reflect, at least in part, pulmonary blood volume, models using PbA in mice  and P. falciparum in rats  indicate that parasites do sequester in the lung. Examination of parasite transcription profiles shows that parasites are undergoing replication and metabolic functions in the lung and are displaying activity other than simple adhesion. Parasite sequestration and replication in the lung would be expected to induce the release of parasite products, cytokine production, and recruitment of immune cells, thereby causing pulmonary inflammation and/or injury. Indeed pulmonary pathology has been reported in previous studies of PbA infection, including increased pulmonary vascular permeability and edema [8, 26–28]. Additionally, respiratory distress is associated with severe malaria due to P. falciparum [29, 30], and pulmonary edema often occurs in individuals who develop CM [31, 32].
Baseline differences and unique temporal responses to PbA infection in CM-resistant (BALB/c) versus CM-susceptible (B6) mice are predominantly associated with immune function
Different clinical outcomes to malaria may be attributable to the differential immunological resting state of resistant (BALB/c) and susceptible (B6) hosts – that is, differences in transcriptional status at baseline may influence the subsequent course of infection. Additionally, host response during infection is likely to contribute to outcome. To test these hypotheses, a linear model was employed to compare expression profiles in specific tissues of both mouse strains at baseline (day 0) and between baseline and post-infection (days 3 and 6 versus day 0). Six hundred and eight (608) mouse genes were expressed differentially at baseline (P < 0.05), and 607 genes showed a significant differential response to infection (P < 0.05) between resistant and susceptible mice. Two hundred and sixty-six (266) genes were differentially expressed both at baseline and over infection, and 949 genes were differentially expressed at baseline and/or during the course of infection. To further characterize these 949 genes, we performed clustering analysis of gene expression at each time point in all tissues (Figure 3). This revealed several groups of genes with distinctive profiles over the time course, many of which are enriched for immunological GO-BP categories.
The expression of one group of 191 genes was consistently greater in resistant mice at baseline and over infection (Figure 3 lower panel, metallic blue bar). Functional categories significantly enriched in this group included 'defence response', 'immune response', 'complement activation' and 'humoral immune response'. Immune-related genes within these groups included interferon-inducible guanylate nucleotide binding protein 1 (gbp1) and chemokine (C-C motif) receptor 4 (ccr4).
Of interest was a cluster of 265 genes with significant enrichment in immune related GO categories. This cluster was highly expressed at baseline in livers of resistant animals, up-regulated in livers of susceptible mice over the course of infection, as well as being upregulated in lungs and spleens of resistant mice during infection (Figure 3 lower panel, maroon bar). Functional annotations associated with this cluster included 'cell adhesion', 'cell communication', 'response to stimulus', 'external stimulus and biotic stimulus', and 'defence, immune and humoral immune response'. Several immunologic cell surface molecules were clustered within this group (e.g., vcam1), as were numerous genes containing immunoglobulin (Ig) domains. Additionally, 38 transcripts encoded or putatively encoded Ig family member genes. Other potentially important transcripts figure included cytokine receptors, 2 subcomponents of complement component 1q, at least 9 interferon-inducible genes and several signalling molecule transcripts.
A third gene cluster of interest (Figure 3 lower panel, turquoise bar), significantly associated with the GO annotations of 'signal transduction', 'cell communication', 'cell surface receptor linked signal transduction' and 'G-protein coupled receptor signalling pathway', was up-regulated in susceptible mice at baseline and in response to infection. Genes encoding proteins likely involved in cell communication included platelet-derived growth factor receptor-like (pdgfr1) and members of the uPAR/Ly6/CD59 family. C-reative protein (crp) was also highly up-regulated in the lungs of resistant mice at baseline, decreasing over infection.
Additional analysis of the genes selected by the linear model, using the Ingenuity pathways analysis program, identified pathways and gene networks enriched in the expression data and therefore likely to be important in CM pathogenesis. Pathways significantly associated with all genes selected by the linear model (right-tailed Fisher's Exact Test, p < 0.05) include pyruvate metabolism, p38 MAPK signalling and platelet-derived growth factor (PDGF) signalling [see Additional file 2]. Pyruvate kinase deficiency has been shown to protect animals from severe disease in another murine malaria model, P. chabaudi chabaudi . In vitro studies have demonstrated that P. falciparum glycosylphosphatidylinositol (GPI) moities stimulate macrophages to produce the proinflammatory cytokines IL-6 and IL-12 via a p38 MAPK-dependent pathway . Although no studies have directly linked PDGF signaling with malaria infection, PDGF is known to stimulate proliferation and chemotaxis in many cell types, including leukocytes; Moreover, platelets are a source of PDGF and may be important in PbA pathogenesis [5, 35].
Analysis of genes differentially expressed in the liver at day 0 identified a network involving several immune-related transcripts, including signal transducer and activator of transcription 1 (stat1), a transcription factor involved in several inflammatory signaling processes (Figure 4A). Many genes in this network were highly expressed in resistant mice at baseline and became up-regulated in susceptible mice over the course of infection (Figure 4B). Therefore, Stat1-mediated signaling, which includes the interferon-α/β and interferon-γ signaling pathways, may moderate survival early in PbA infection. Our analysis shows that resistant BALB/c mice are primed for this response at baseline, while B6 mice show a delayed response, which may contribute to poor outcome. Both interferon-α/β and interferon-γ signaling have previously been associated with malaria infection. Development of CM in PbA infection requires both IFNγ and IFNγ receptor [28, 36]. However, early production of IFNγ, as observed in the K173 murine model of non-cerebral malaria may protect against subsequent progression to cerebral malaria . Moreover, IFNα and IFNγ receptor polymorphisms have been linked to protection from severe and cerebral malaria in humans [37, 38]. Recent work comparing transcriptional responses in presymptomatic versus clinically apparent P. falciparum infection identified IFNγ signalling pathway members, including STAT1, to be induced early in infection , lending further support to the role of this signalling pathway. Since the differential response appears to involve a temporal sequence (i.e., early expression occurs in mice that survive acute infection and late expression involves CM development), enhancement of this pathway at appropriate times during the course of infection could improve the outcome of susceptible animals. Further experiments in a broader range of mouse strains will be required to examine how Stat1-mediated signalling, including early interferon-α/β and/or-γ mediated responses, modulates the progression of plasmodial infections and whether it could be used as a biomarker to predict resistance to CM.
Malaria gene expression is modified by host genetic background and tissue microenvironment
We hypothesized that malaria gene expression would differ depending on the susceptibility status of the mammalian host. Therefore, we examined differential PbA gene expression between mouse strains. Using a linear model, 469 malaria genes were identified as differentially expressed (P < 0.05, Figure 5). Of note, when clustered, these genes showed striking differences, not only between resistant and susceptible hosts, but also between the target tissues examined (Figure 5).
An initial group of 60 genes, expressed in the lung and liver of susceptible (B6) mice (Figure 5 lower panel, turquoise bar) was over-represented in the GO category of 'cell communication'. One major cluster defined a set of 53 co-expressed genes highly up-regulated in the lungs and livers of susceptible (B6) mice and spleens of resistant (BALB/c) mice (Figure 5 lower panel, navy bar). GO categories significantly enriched in this cluster included 'response to stimulus and temperature', 'glucose, amino acid, hypusine and pigment metabolism' and the four predominant GO categories of 'biosynthesis', 'cellular process', 'metabolism' and 'physiological process'. The cluster representing 54 highly expressed genes in spleens of susceptible mice (Figure 5 lower panel, maroon bar) had limited functional annotations, but was significantly represented in the GO categories of 'response to stimulus' and 'drug transport'. A smaller cluster of 33 genes, which was up-regulated in the lungs of resistant mice and to a lesser extent in the livers of susceptible animals, was associated with regulation of enzyme activity and mitotic interphase (Figure 5 lower panel, light blue bar). Lastly, a prominent cluster containing predominantly putative genes lacking GO annotations appeared in the brains of resistant mice (Figure 5 lower panel, purple bar).
Although it was anticipated that parasites might respond differently in resistant versus susceptible hosts, the unique transcriptional signature that the parasite displayed in each of the four organs examined provides in vivo evidence that malaria parasites are profoundly influenced by the host genetic background and tissue microenvironment in which they replicate or sequester. The systematic transcriptional profiling of the same parasite isolate in multiple organs identified differential expression of a number of enzymes involved with energy pathways, cell signalling molecules, and genes encoding heat-shock proteins, some of which may constitute the parasites' own response to host defence. However, many of the transcripts remain annotated as hypothetical proteins and several evident gene clusters, especially those transcribed in the brains, have no gene ontology or protein information associated with them. Regardless, further investigation of these differentially expressed clusters may yield new potential targets for drug development and provide further insights into the regulation of plasmodial virulence determinants in vivo.
Verification of expression analysis using quantitative real-time RT-PCR
Quantitative real-time RT-PCR (qRT-PCR) was performed to analyze the expression of a number of representative mouse genes (including cell adhesion molecules, cytokines, and interferon-inducible transcripts) and PbA genes, which were either identified by our statistical analysis or previously shown to be important in other studies but not identified by our analysis (e.g. TNF-α). qRT-PCR results correlated well with normalized intensity data in each organ at each time point, yielding a median overall correlation of 0.724 with the microarray data, and thereby confirming the validity of observed gene expression patterns [see Additional file 3].
Analysis of mouse and malaria gene groups previously associated with malaria disease pathogenesis
Previous studies in various experimental models of malaria have demonstrated that specific host responses may influence clinical outcome [14, 15] and that certain malaria genes may contribute to parasite virulence or disease pathogenesis. Based on the results from these previous studies, we chose to examine expression patterns of genes represented on the array in specific categories associated with malaria infection. For mouse genes, functional categories included 'cell adhesion', 'toll-like receptors', 'immune response' (cytokines, chemokines and their receptors), 'interferon responsive genes', 'acute phase proteins', 'erythroid-associated transcripts', 'complement activation', 'glycolysis' and 'hemostasis' [see Additional file 4]. PbA transcripts examined included the bir family of genes, orthologues of the P. falciparum rif/stevor genes  which encode for variant surface antigens ; an orthologue of sequestrin and the GO-BP category of 'invasion' [see Additional file 5]. While many of the PbA transcripts were in the top quartile of expressed genes, especially in the lungs, none of the chosen groups showed a significant enrichment of genes selected by the linear model (hypergeometric distribution, P < 0.05) and only two were differentially expressed between resistant and susceptible animals. This finding suggests that while parasite surface molecules and invasion-associated genes may be important in infection, they are not associated with a parasite phenotype that promotes the development of CM.
However, in several host gene groups, there was a significant enrichment of differentially expressed genes selected by the linear model (Figure 6). This emphasizes the role of toll-like receptors; cytokines, chemokines and their receptors; interferon responsive genes; acute phase proteins and complement activation groups in the PbA CM model. In our analysis, many IFN-inducible genes were identified as differentially expressed (Figure 6C), lending further support for the importance of the timing and magnitude of interferon responses in mediating outcomes in PbA infection. With this and other microarray studies [13, 14, 16] identifying the expression of interferon-inducible genes in response to P. berghei, it could be argued that differences in constitutive and early innate immune responses by IFN-regulated genes, could contribute to innate resistance in BALB/c mice, and may be prognostic of outcome.
This analysis also identified complement-related genes as a group likely to be important to PbA susceptibility (Figure 6E). Epidemiological evidence links decreased complement receptor 1 (CR1) expression with reduced malaria incidence and disease severity . Additionally, increased expression of C1qbeta is found in the brains of susceptible mice, compared to resistant ones, in PbA infection . As with unregulated cytokines, a defective or deregulated complement response may also contribute to the pathophysiology of malaria.
In this study, we designed and applied a combination oligonucleotide microarray to simultaneously interrogate organ-specific transcriptional responses on both sides of the pathogen-host interface in experimental murine malaria. The use of this and other host and/or plasmodium microarrays to examine parasite transcriptional profiles in vivo [10–15, 17], demonstrates important features of how both host and pathogen respond during infection. A host/parasite combination array not only offers the advantage of providing concurrent information about both organisms, but is also more time- and cost-effective, and does not provide additional challenges for data analysis than using separate platforms.
The combination microarray data generated by this study provide an initial understanding of the dynamic tissue-specific bi-directional interaction between pathogen and host in malaria. Several unique insights emerged from this study. From the pathogen perspective, the lung is an important site of parasite metabolism and proliferation and that parasite transcriptional differences occurred not only between different hosts, but also within different organ environments. The differential response between resistant and susceptible mice was primarily associated with immune function and our analysis provides evidence for the importance of interferon signalling, interferon-responsive genes and complement-related genes in CM pathogenesis. Ultimately, when combined with complementary genetic strategies including targeted gene disruption, or extended to other host-pathogen pairs including humans and P. falciparum, such work may provide further insights into the pathogenesis of CM and facilitate the design of novel interventions targeting both host and parasite gene networks to improve clinical outcome in malaria.
Mouse/PbA combination microarray design
The 21,626 mouse probes were identical to those in a previous study for which hybridization was detected in at least one of 55 mouse tissues analyzed . 12,328 known and predicted P. berghei gene sequences were downloaded from the Plasmodium falciparum Genome Projects website at the Wellcome Trust Sanger Institute. P. berghei EST (5,582) and GSS (5,482) sequences were gathered from the University of Florida P. berghei Genome Tag Sequences website in August 2004 . Redundancy checks were performed using sequence alignments (BLAST) with an e-value threshold of e-8. From the University of Florida sequences, 4,520 GSS sequences were found to be unique from ESTs. From the Sanger set, 7,211 sequences were found to be unique from ESTs. All University of Florida EST sequences were included in the final sequence set which totaled 17,313 malaria sequences. Sixty-mer probes were designed using a protocol in which Tm-balanced sequences beginning every 10 bases in each target gene are evaluated for secondary structure, repeat content, and uniqueness, and a single probe that best balances these parameters was selected. One or two probes were designed for each of the malaria sequences. Probe sequences were submitted to Agilent Technologies (Palo Alto CA) for array manufacture.
Mice and parasites
Experiments were approved by and conducted according to the University of Toronto animal ethics guidelines. Male B6 and BALB/c mice, 6–8 weeks of age, were obtained from Charles River Laboratories (Senneville QC). Cryopreserved P. berghei ANKA (MR4, Manassas VA) was thawed and passaged through naïve B6 donor mice.
On day 0, prior to infection, 5 mice from each strain were euthanized and served as a baseline control for all experimental mice. Immediately following euthanasia, whole blood was collected by direct cardiac puncture, and intact brains, livers, spleens and lungs were excised, snap-frozen in liquid nitrogen and stored at -80°C until use. 10 mice from each strain were infected with 1 × 106 freshly isolated P. berghei ANKA parasitized erythrocytes by intraperitoneal injection. Parasitemia was monitored daily using thin blood smears stained with modified Giemsa (Protocol Hema 3 Stain Set, Sigma, Oakville ON). Five mice from each strain were sacrificed at both day 3 and day 6 following infection. Blood and organs were harvested as above.
Total RNA was extracted by homogenizing organs in Trizol reagent (Invitrogen, Burlington ON) according to the manufacturer's instructions, and mRNA was purified as described previously [20, 44]. Briefly, RNA samples were denatured and loaded onto a 0.25 ml Oligo-dT cellulose (NEB, Mississauga ON) column in 40 mM Tris pH7.5, 1 M NaCl, 2 mM EDTA and 0.2% SDS. The columns were washed, and RNA was eluted in TE. This process was repeated once. Eluted mRNA was ethanol precipitated, resuspended in dH20 and stored at -80°C until use. Integrity of RNA was assessed by formaldehyde agarose gel electrophoresis.
cDNA labeling and microarray hybridization
cDNA was reverse-transcribed from 1–-2 g mRNA using Superscript II reverse μ transcriptase (Invitrogen) with 1 μg random nonamer primers and 0.25 g T18VN per μ reaction. The reaction mix contained a final concentration of 1X RT Buffer, 10 mM DTT, 0.5 mM each dNTP and 0.5 mM 5-(3-aminoallyl)-2'deoxyuridine-5'-triphosphate (AAdUTP, Sigma). Following the RT reaction, RNA was hydrolyzed using NaOH/EDTA and cDNA was purified using QIAquick PCR Purification columns (Qiagen, Mississauga ON), washed with 80% ethanol and eluted in water. Purified cDNA was coupled with N-hydroxysuccinimide esters of Cy3 or Cy5 (GE Lifesciences, Baie d'Urfe QC) in bicarbonate buffer, following the manufacturer's instructions. Dye reactions were quenched by adding hydroxylamine and labeled cDNAs were separated from free dye molecules using QIAquick columns. Cy3 and Cy5-labeled cDNA pairs and Agilent control spots were added to a final volume of 0.5 ml hybridization buffer (1 M NaCl, 0.5% sodium sarcosine, 50 mM methyl ethane sulfonate (MES), pH 6.5, 33% formamide and 40 g salmon sperm DNA (Invitrogen)). Hybridizations were performed in Agilent hybridization chambers at 42°C with rotation for 18–24 hours. Slides were washed in 6X SSPE, 0.005% sarcosine, followed by 0.06X SSPE, allowed to dry and scanned with a 4000A microarray scanner. TIFF images were quantified with GenePix (Axon Instruments, Union City CA). Each array was hybridized with two samples representing equivalent tissues at equivalent time points in the different mouse strains.
The median intensity measurements extracted from the GenePix files were spatially detrended . We applied variance stabilizing normalization (VSN)  in Bioconductor  and transformed to log2 scale. To compensate for potential cross-hybridization of mouse mRNA to malaria probes, the day 0 intensity value of each probe was subtracted from the day 3 and 6 intensity values in the same tissue. The intensity scale data (shown in Figures 2, 3 and 4) was median-subtracted across tissues. Because most genes are not expressed in most tissues, this prevents the clustering algorithm from seizing on noise amplified in the log scale. After the cross-hybridization removal step and median subtraction, all negative values were set to zero. Finally, the ratios of BALB/c to C57BL/6 intensity measurements in log2 scale (henceforth referred to as log ratios) were obtained after loess smoothing using Bioconductor  of VSN-normalized arrays.
Microarray probe annotation
Gene names were obtained by pairwise alignment to Refseq (m33 May 2005) and Ensembl (v31 May 2005) using megablast . The corresponding official gene names to the RefSeq and Ensembl IDs were obtained from the Mouse Genome Informatics (MGI). GO-Biological Process (GO-BP) annotations for mouse probes were from Zhang et al. . PbA probes were aligned using blat  to the gene sequences provided by the P. berghei genome project at the Sanger Institute downloaded in July 2005 . Since, to the best of our knowledge, there is no P. berghei functional annotation data, we annotated PbA genes by their P. falciparum orthologues using homology tables were obtained from Hall et al. . Among the 8,577 PbA genes in the arrays, 2,616 have a P. falciparum orthologous gene whose identifier starts with "PF" or "MAL". P. falciparum GO-BP annotations were obtained from the Gene Ontology database downloaded in May 2005 and PlasmoDB annotations released in October 2002 . Only 813 out of the 2,616 P. falciparum genes with a homologous PbA gene in the arrays have a functional annotation. Pfam assignments available for 8,502 malaria genes in the arrays were downloaded from GeneDB on November 24th, 2005 . Of these, only 1,106 or 13% have a product name other than "hypothetical protein".
For statistical analysis, linear models for microarray data  were applied to the normalized log ratios using the Limma software package version 1.9.6  in Bioconductor. Linear models are a statistical approach similar to ANOVA, which generalize both ANOVAs and linear regressions. B6 samples were taken as reference in the design matrix and two contrast matrices were used. The first contrast matrix was used to estimate the difference between susceptible and resistant mice at day 0. The second contrast matrix was used to estimate the difference in the response to the infection between susceptible and resistant mice (i.e., to compare the difference between the expression level at day 0 and the average expression level of day 3 and 6 between the two mouse strains). Design and contrast matrices are available upon request. P-values for the four tissues were combined in a p-value per probe using the F-distribution. F-pvalues of probes mapped to the same gene were combined using Fischer method. Genes with a combined F-p-value < 0.05 were considered differentially expressed. P-values were not corrected for multiple testing.
We obtained the log ratios per gene by averaging the log ratios of probes mapped to the same gene. Genes were clustered according to their log ratio using either Pearson (Figure 2) or cosine-angle correlation (Figures 3, 4, 5, 6), and hierarchical clustering functions available in Bioconductor and R. The intensity diagrams show the median-subtracted intensity per gene obtained by averaging the relative measurements of probes mapped to the same gene. Clusters were selected for functional/annotation analysis by partitioning the corresponding dendrogram at a fixed height (between 1.2 and 1.7). The exact height was manually determined for each dendrogram based on the visual homogeneity of the sub-clusters and their size.
GO annotations were up-propagated using the GO-graph available in the GO package version 1.8.2 in Bioconductor. Enrichment of functional (GO-BP) or protein (Pfam) annotations on each sub-cluster was scored using the hypergeometric distribution. Annotations with a hypergeometric p-value smaller than 0.05 were considered to be significantly over-represented in a sub-cluster. For visualization, similar significant GO categories were joined together under their most specific common ancestor.
Network and pathway analysis
Networks and pathways involving genes selected by the linear model were identified using Ingenuity Pathway Analysis software . The analysis was performed on all 949 genes differentially expressed at baseline and across the course of infection (Figure 3), and also on genes differentially expressed per tissue. Significance values were calculated using a right-tailed Fisher's Exact Test. The one-tailed version of Fisher's Exact Test is identical to the corresponding test based on the hypergeometric distribution.
Comparison between the malaria transcriptome and interactome
To explore whether there is a relation between the malaria expression profiles obtained in this work and protein-protein interaction (PPI) data, we compared the PPI network of P. falciparum described by Lacount, DJ et al.  with the cluster graph shown in Figure 2. Enrichment of protein-protein interactions in clusters was determined by random permutation of gene labels in the interaction network. There are 216 interacting protein pairs belonging to the same cluster among the 602 P. falciparum genes in the PPI network with a PbA homologue in the array. This intersection is significantly higher than expected by chance (P-value < 0.0008) when compared with the intersection obtained with 10,000 random networks. To determine whether any of the clusters were enriched for interacting proteins from any of the sub-networks , we used the hypergeometric distribution and selected those sub-networks with a p-value < 0.05.
Quantitative real-time RT-PCR (qPCR)
cDNA was synthesized from 0.5 μg of mRNA using Superscript II reverse transcriptase with Oligo (dT)12–18 primers (Invitrogen). Serial dilutions of mouse or PbA genomic DNA purified from blood (Qiagen) were used as standards. gDNA standards or cDNA were added to the qPCR reaction containing 1X Power Sybr Green Master Mix (Applied Biosystems, Streetsville ON) and 0.5 μM primers in a final volume of 10 μl. qPCR was performed using the ABI Prism® 7900HT Sequence Detection System (Applied Biosystems). PCR primer sequences are posted on the website (below). Copy numbers were normalized to 5 mouse housekeeping genes (mouse)  or to cDNA concentration (PbA). qRT-PCR data and microarray data (normalized intensities) were compared by calculating the overall correlation of all organs at all time points in both mouse strains for each gene.
Detailed protocols, probe sequences and mappings, gene annotations, all microarray data, and information about the microarray hybridizations are posted on the project website . The microarray platform (GPL4220 [NCBI #15193528]) and data series (GSE5672 [NCBI tracking system #15195573]) have been submitted to the Gene Expression Omnibus (GEO).
Plasmodium berghei ANKA
C57BL/6 mouse strain
Gene Ontology Biological ProcessPPI: Protein-Protein Interaction
- stat1 :
signal transducer and activator of transcription 1
quantitative real-time RT-PCR
Mackintosh CL, Beeson JG, Marsh K: Clinical features and pathogenesis of severe malaria. Trends Parasitol. 2004, 20 (12): 597-603. 10.1016/j.pt.2004.09.006.
Fortin A, Stevenson MM, Gros P: Susceptibility to malaria as a complex trait: big pressure from a tiny creature. Hum Mol Genet. 2002, 11 (20): 2469-2478. 10.1093/hmg/11.20.2469.
Hunt NH, Grau GE: Cytokines: accelerators and brakes in the pathogenesis of cerebral malaria. Trends Immunol. 2003, 24 (9): 491-499. 10.1016/S1471-4906(03)00229-1.
Rest JR: Cerebral malaria in inbred mice. I. A new model and its pathology. Trans R Soc Trop Med Hyg. 1982, 76 (3): 410-415. 10.1016/0035-9203(82)90203-6.
Lou J, Lucas R, Grau GE: Pathogenesis of cerebral malaria: recent experimental data and possible applications for humans. Clin Microbiol Rev. 2001, 14 (4): 810-20, table of contents. 10.1128/CMR.14.4.810-820.2001.
de Souza JB, Riley EM: Cerebral malaria: the contribution of studies in animal models to our understanding of immunopathogenesis. Microbes Infect. 2002, 4 (3): 291-300. 10.1016/S1286-4579(02)01541-1.
de Kossodo S, Grau GE: Profiles of cytokine production in relation with susceptibility to cerebral malaria. J Immunol. 1993, 151 (9): 4811-4820.
Neill AL, Hunt NH: Pathology of fatal and resolving Plasmodium berghei cerebral malaria in mice. Parasitology. 1992, 105 ( Pt 2): 165-175.
Mitchell AJ, Hansen AM, Hee L, Ball HJ, Potter SM, Walker JC, Hunt NH: Early cytokine production is associated with protection from murine cerebral malaria. Infect Immun. 2005, 73 (9): 5645-5653. 10.1128/IAI.73.9.5645-5653.2005.
Schaecher K, Kumar S, Yadava A, Vahey M, Ockenhouse CF: Genome-wide expression profiling in malaria infection reveals transcriptional changes associated with lethal and nonlethal outcomes. Infect Immun. 2005, 73 (9): 6091-6100. 10.1128/IAI.73.9.6091-6100.2005.
Griffiths MJ, Shafi MJ, Popper SJ, Hemingway CA, Kortok MM, Wathen A, Rockett KA, Mott R, Levin M, Newton CR, Marsh K, Relman DA, Kwiatkowski DP: Genomewide analysis of the host response to malaria in Kenyan children. J Infect Dis. 2005, 191 (10): 1599-1611. 10.1086/429297.
Ylostalo J, Randall AC, Myers TA, Metzger M, Krogstad DJ, Cogswell FB: Transcriptome profiles of host gene expression in a monkey model of human malaria. J Infect Dis. 2005, 191 (3): 400-409. 10.1086/426868.
Hansen DS, Evans KJ, D'Ombrain MC, Bernard NJ, Sexton AC, Buckingham L, Scalzo AA, Schofield L: The natural killer complex regulates severe malarial pathogenesis and influences acquired immune responses to Plasmodium berghei ANKA. Infect Immun. 2005, 73 (4): 2288-2297. 10.1128/IAI.73.4.2288-2297.2005.
Sexton AC, Good RT, Hansen DS, D'Ombrain MC, Buckingham L, Simpson K, Schofield L: Transcriptional profiling reveals suppressed erythropoiesis, up-regulated glycolysis, and interferon-associated responses in murine malaria. J Infect Dis. 2004, 189 (7): 1245-1256. 10.1086/382596.
Delahaye NF, Coltel N, Puthier D, Flori L, Houlgatte R, Iraqi FA, Nguyen C, Grau GE, Rihet P: Gene-Expression Profiling Discriminates between Cerebral Malaria (CM)-Susceptible Mice and CM-Resistant Mice. J Infect Dis. 2006, 193 (2): 312-321. 10.1086/498579.
Ockenhouse CF, Hu WC, Kester KE, Cummings JF, Stewart A, Heppner DG, Jedlicka AE, Scott AL, Wolfe ND, Vahey M, Burke DS: Common and Divergent Immune Response Signaling Pathways Discovered in Peripheral Blood Mononuclear Cell Gene Expression Patterns in Presymptomatic and Clinically Apparent Malaria. Infect Immun. 2006, 74 (10): 5561-5573. 10.1128/IAI.00408-06.
Daily JP, Le Roch KG, Sarr O, Ndiaye D, Lukens A, Zhou Y, Ndir O, Mboup S, Sultan A, Winzeler EA, Wirth DF: In vivo transcriptome of Plasmodium falciparum reveals overexpression of transcripts that encode surface proteins. J Infect Dis. 2005, 191 (7): 1196-1203. 10.1086/428289.
Hall N, Karras M, Raine JD, Carlton JM, Kooij TW, Berriman M, Florens L, Janssen CS, Pain A, Christophides GK, James K, Rutherford K, Harris B, Harris D, Churcher C, Quail MA, Ormond D, Doggett J, Trueman HE, Mendoza J, Bidwell SL, Rajandream MA, Carucci DJ, Yates JR, Kafatos FC, Janse CJ, Barrell B, Turner CM, Waters AP, Sinden RE: A comprehensive survey of the Plasmodium life cycle by genomic, transcriptomic, and proteomic analyses. Science. 2005, 307 (5706): 82-86. 10.1126/science.1103717.
Xu X, Dong Y, Abraham EG, Kocan A, Srinivasan P, Ghosh AK, Sinden RE, Ribeiro JM, Jacobs-Lorena M, Kafatos FC, Dimopoulos G: Transcriptome analysis of Anopheles stephensi-Plasmodium berghei interactions. Mol Biochem Parasitol. 2005, 142 (1): 76-87. 10.1016/j.molbiopara.2005.02.013.
Zhang W, Morris QD, Chang R, Shai O, Bakowski MA, Mitsakakis N, Mohammad N, Robinson MD, Zirngibl R, Somogyi E, Laurin N, Eftekharpour E, Sat E, Grigull J, Pan Q, Peng WT, Krogan N, Greenblatt J, Fehlings M, van der Kooy D, Aubin J, Bruneau BG, Rossant J, Blencowe BJ, Frey BJ, Hughes TR: The functional landscape of mouse gene expression. J Biol. 2004, 3 (5): 21-10.1186/jbiol16.
Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP, Dolinski K, Dwight SS, Eppig JT, Harris MA, Hill DP, Issel-Tarver L, Kasarskis A, Lewis S, Matese JC, Richardson JE, Ringwald M, Rubin GM, Sherlock G: Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet. 2000, 25 (1): 25-29. 10.1038/75556.
Bozdech Z, Llinas M, Pulliam BL, Wong ED, Zhu J, DeRisi JL: The transcriptome of the intraerythrocytic developmental cycle of Plasmodium falciparum. PLoS Biol. 2003, 1 (1): E5-10.1371/journal.pbio.0000005.
LaCount DJ, Vignali M, Chettier R, Phansalkar A, Bell R, Hesselberth JR, Schoenfeld LW, Ota I, Sahasrabudhe S, Kurschner C, Fields S, Hughes RE: A protein interaction network of the malaria parasite Plasmodium falciparum. Nature. 2005, 438 (7064): 103-107. 10.1038/nature04104.
Franke-Fayard B, Janse CJ, Cunha-Rodrigues M, Ramesar J, Buscher P, Que I, Lowik C, Voshol PJ, den Boer MA, van Duinen SG, Febbraio M, Mota MM, Waters AP: Murine malaria parasite sequestration: CD36 is the major receptor, but cerebral pathology is unlinked to sequestration. Proc Natl Acad Sci U S A. 2005, 102 (32): 11468-11473. 10.1073/pnas.0503386102.
Pettersson F, Vogt AM, Jonsson C, Mok BW, Shamaei-Tousi A, Bergstrom S, Chen Q, Wahlgren M: Whole-body imaging of sequestration of Plasmodium falciparum in the rat. Infect Immun. 2005, 73 (11): 7736-7746. 10.1128/IAI.73.11.7736-7746.2005.
Senaldi G, Vesin C, Chang R, Grau GE, Piguet PF: Role of polymorphonuclear neutrophil leukocytes and their integrin CD11a (LFA-1) in the pathogenesis of severe murine malaria. Infect Immun. 1994, 62 (4): 1144-1149.
Weiss ML, Kubat K: Plasmodium berghei: a mouse model for the "sudden death" and "malarial lung" syndromes. Exp Parasitol. 1983, 56 (1): 143-151. 10.1016/0014-4894(83)90105-4.
Yanez DM, Manning DD, Cooley AJ, Weidanz WP, van der Heyde HC: Participation of lymphocyte subpopulations in the pathogenesis of experimental murine cerebral malaria. J Immunol. 1996, 157 (4): 1620-1624.
Marsh K, Forster D, Waruiru C, Mwangi I, Winstanley M, Marsh V, Newton C, Winstanley P, Warn P, Peshu N, et al: Indicators of life-threatening malaria in African children. N Engl J Med. 1995, 332 (21): 1399-1404. 10.1056/NEJM199505253322102.
Maguire GP, Handojo T, Pain MC, Kenangalem E, Price RN, Tjitra E, Anstey NM: Lung injury in uncomplicated and severe falciparum malaria: a longitudinal study in papua, Indonesia. J Infect Dis. 2005, 192 (11): 1966-1974. 10.1086/497697.
Aursudkij B, Wilairatana P, Vannaphan S, Walsh DS, Gordeux VR, Looareesuwan S: Pulmonary edema in cerebral malaria patients in Thailand. Southeast Asian J Trop Med Public Health. 1998, 29 (3): 541-545.
Taylor WR, White NJ: Malaria and the lung. Clin Chest Med. 2002, 23 (2): 457-468. 10.1016/S0272-5231(02)00004-7.
Min-Oo G, Fortin A, Tam MF, Nantel A, Stevenson MM, Gros P: Pyruvate kinase deficiency in mice protects against malaria. Nat Genet. 2003, 35 (4): 357-362. 10.1038/ng1260.
Zhu J, Krishnegowda G, Gowda DC: Induction of proinflammatory responses in macrophages by the glycosylphosphatidylinositols of Plasmodium falciparum: the requirement of extracellular signal-regulated kinase, p38, c-Jun N-terminal kinase and NF-kappaB pathways for the expression of proinflammatory cytokines and nitric oxide. J Biol Chem. 2005, 280 (9): 8617-8627. 10.1074/jbc.M413539200.
Polack B, Delolme F, Peyron F: Protective role of platelets in chronic (Balb/C) and acute (CBA/J) Plasmodium berghei murine malaria. Haemostasis. 1997, 27 (6): 278-285.
Amani V, Vigario AM, Belnoue E, Marussig M, Fonseca L, Mazier D, Renia L: Involvement of IFN-gamma receptor-medicated signaling in pathology and anti-malarial immunity induced by Plasmodium berghei infection. Eur J Immunol. 2000, 30 (6): 1646-1655. 10.1002/1521-4141(200006)30:6<1646::AID-IMMU1646>3.0.CO;2-0.
Koch O, Kwiatkowski DP, Udalova IA: Context-specific functional effects of IFNGR1 promoter polymorphism. Hum Mol Genet. 2006, 15 (9): 1475-1481. 10.1093/hmg/ddl071.
Aucan C, Walley AJ, Hennig BJ, Fitness J, Frodsham A, Zhang L, Kwiatkowski D, Hill AV: Interferon-alpha receptor-1 (IFNAR1) variants are associated with protection against cerebral malaria in the Gambia. Genes Immun. 2003, 4 (4): 275-282. 10.1038/sj.gene.6363962.
Janssen CS, Phillips RS, Turner CM, Barrett MP: Plasmodium interspersed repeats: the major multigene superfamily of malaria parasites. Nucleic Acids Res. 2004, 32 (19): 5712-5720. 10.1093/nar/gkh907.
Cheng Q, Cloonan N, Fischer K, Thompson J, Waine G, Lanzer M, Saul A: stevor and rif are Plasmodium falciparum multicopy gene families which potentially encode variant antigens. Mol Biochem Parasitol. 1998, 97 (1-2): 161-176. 10.1016/S0166-6851(98)00144-3.
Cockburn IA, Mackinnon MJ, O'Donnell A, Allen SJ, Moulds JM, Baisor M, Bockarie M, Reeder JC, Rowe JA: A human complement receptor 1 polymorphism that reduces Plasmodium falciparum rosetting confers protection against severe malaria. Proc Natl Acad Sci U S A. 2004, 101 (1): 272-277. 10.1073/pnas.0305306101.
Wellcome Trust Sanger Institute Plasmodium berghei genome project. [http://www.sanger.ac.uk/Projects/P_berghei/]
University of Florida P. berghei Genome Tag Project. [http://parasite.vetmed.ufl.edu]
Hughes TR, Marton MJ, Jones AR, Roberts CJ, Stoughton R, Armour CD, Bennett HA, Coffey E, Dai H, He YD, Kidd MJ, King AM, Meyer MR, Slade D, Lum PY, Stepaniants SB, Shoemaker DD, Gachotte D, Chakraburtty K, Simon J, Bard M, Friend SH: Functional discovery via a compendium of expression profiles. Cell. 2000, 102 (1): 109-126. 10.1016/S0092-8674(00)00015-5.
Huber W, von Heydebreck A, Sultmann H, Poustka A, Vingron M: Variance stabilization applied to microarray data calibration and to the quantification of differential expression. Bioinformatics. 2002, 18 Suppl 1: S96-104.
Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, Ellis B, Gautier L, Ge Y, Gentry J, Hornik K, Hothorn T, Huber W, Iacus S, Irizarry R, Leisch F, Li C, Maechler M, Rossini AJ, Sawitzki G, Smith C, Smyth G, Tierney L, Yang JY, Zhang J: Bioconductor: open software development for computational biology and bioinformatics. Genome Biol. 2004, 5 (10): R80-10.1186/gb-2004-5-10-r80.
Zhang Z, Schwartz S, Wagner L, Miller W: A greedy algorithm for aligning DNA sequences. J Comput Biol. 2000, 7 (1-2): 203-214. 10.1089/10665270050081478.
Kent WJ: BLAT--the BLAST-like alignment tool. Genome Res. 2002, 12 (4): 656-664. 10.1101/gr.229202. Article published online before March 2002.
Plasmodium berghei GeneDB. [http://www.genedb.org/genedb/idListForm.jsp]
Smyth GK: Linear Models and Empirical Bayes Methods for Assessing Differential Expression in Microarray Experiments. Statistical Applications in Genetics and Molecular Biology. 2004, 3 (1): Article 3-10.2202/1544-6115.1027.
Smyth GK: limma: Linear Models for Microarray Data. Bioinformatics and computational biology solutions using R and Bioconductor. Edited by: Gentleman RCVHWIRDS. 2005, New York , Springer, 397-420:
Ingenuity® Systems. [http://www.ingenuity.com]
Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, Speleman F: Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol. 2002, 3 (7): RESEARCH0034-10.1186/gb-2002-3-7-research0034.
Hughes' Lab Malaria Project Website. [http://hugheslab.med.utoronto.ca/malaria/]
We would like to thank Jay Hesselberth (University of Washington, Seattle WA) for providing PPI information and members of Sandy Der's laboratory (University of Toronto) for assistance with qRT-PCR. This work was financially supported by the Canadian Institutes of Health Research (Team Grant in Malaria [TRH and KCK] and MT-13721 [KCK]) and by Genome Canada through the Ontario Genomics Institute. FEL is supported by a CIHR MD/PhD Studentship, WCL by a CIHR Canada Research Chair in Infectious Diseases and Inflamation, and KCK by a CIHR Canada Research Chair in Molecular Parasitology.
FEL carried out the animal studies, organ and RNA isolation, performed microarray hybridizations, qRT-PCR and prepared the manuscript. LP-C performed all data analysis including probe annotation, normalization and statistical/functional analysis; and prepared the figures. NM wrote the software used for probe design, designed microarray probes and performed microarray hybridizations. WCL assisted with data interpretation. TRH conceived the "combination array" concept, designed microarrays and contributed to study design and coordination. KCK conceived the study, and contributed to its design and coordination. All authors helped to draft and approved the final manuscript.
Fiona E Lovegrove, Lourdes Peña-Castillo contributed equally to this work.