Cell infection for microarray analysis
To understand the extent of host gene modulation by the murine model malaria parasite P. berghei ANKA, a mouse hepatoma cell line, Hepa1-6, was infected with sporozoites obtained from Anopheles stephensi salivary glands, freshly dissected as previously described [12]. Plasmodium sporozoite infection in vitro shows very low efficiency ([11] and Figure 1A, top panel, parasites are green), with an average of 2.7 ± 0.7% infected cells (n = 8) in our chosen model. To overcome the "dilution effect" of infected cells by non-infected ones, Hepa1-6 cells were infected with GFP-expressing P. berghei ANKA parasites [13] and infected cells were purified by Fluorescence Activated Cell Sorting (FACS) at 6, 12, 18 and 24 h post-infection (p.i., Figure 1B) to capture as many parasite-induced cellular events as possible: at 6 h p.i. the parasite starts to change its elongated sporozoite shape into a rounder form, which grows in size (12 h p.i) and starts multiplying, with several nuclei present at 18 h p.i.; by 24 h p.i. sporozoite specific organelles are no longer visible and the parasite is bigger and round-shaped (Figure 1A, bottom panel, parasites are green). The viability of sorted infected and non-infected cells was confirmed by microscopic observation of cells seeded on coverslips immediately after sorting and allowed to adhere for at least 2 hours (Figure 1C).
Global gene expression analysis of hepatoma cells during Plasmodium infection
Expression profiling of P. berghei ANKA-infected and non-infected Hepa1-6 cells was performed at the four developmental time-points selected, using the Mouse Genome 430 2.0 Array (Affymetrix, Santa Clara, CA). To exclude potential false positives due to cross-hybridization of parasite RNA with mouse probes, cRNA obtained from axenic cultures of P. berghei sporozoites [14] at the same selected time points was pooled and hybridized as an extra sample on the Mouse 430 2.0 array. No cross-hybridization was found in the list of genes differentially expressed (DE) in P. berghei-infected cells with the parasite cRNA control array (data not shown).
The experimental design aimed at two complementary goals: (i) identifying host genes DE in Plasmodium infected compared to non-infected hepatoma cells, and (ii) following the patterns of differential expression across infection and parasite development within the cell. After robust normalization ~55% of the 45000 probesets represented on the array were classified as present by the MAS5.0 software (Affymetrix). Empirical Bayes moderated t-statistics ("limma" software package of R/Bioconductor [15]) were used to detect up- and down-regulated transcripts in response to Plasmodium infection at any of the four selected time-points. One thousand one hundred and seven probesets (or transcripts) representing 1064 unique genes proved to be significantly DE in at least one time point during the course of Hepa1-6 cells infection by GFP+ P. berghei ANKA (see Additional File 1). This set of data was used in all subsequent analyses.
We used the expression values of each of the 1107 transcripts for hierarchical clustering where each time point was plotted for non-infected and infected samples (Figure 2A). Different profiles can clearly be identified between cells that either have or have not been infected with sporozoites (Figure 2A, green and yellow bars, respectively). Two main clusters separate genes that are highly expressed as a result of malaria infection (Figure 2A, red bars), while their expression in non-infected cells is relatively low (Figure 2A, blue bars), and vice versa. In addition, within the two large clusters, it is possible to identify smaller groups of probesets displaying an increasing (or decreasing) signal throughout time within both the infected and the non-infected cell categories (Figure 2A, grey shaded side bars). These changes indicate not only that malaria parasites modify the host cell transcriptome during the course of infection but also that they manipulate it in a timely fashion, presumably to fit their developmental needs.
In order to evaluate the overall structure of the data, we have plotted the first three principal components of a principal component analysis (PCA), capturing the overall variance of the samples in 3 dimensions (Figure 2B). This analysis clearly separates the data into 4 subgroups, clustering together the biological replicates, segregating infected from non-infected cells, and separating the samples by time after infection. Intriguingly, PCA separates genes expressed at 12 h p.i. in infected cells (Figure 2B, red dots) from those expressed at 6, 18 and 24 h p.i. (Figure 2B, black, green and blue dots, respectively). In non-infected cells, early to mid-term cellular events (6–12 h p.i, turquoise and pink dots, respectively) can be differentiated from later ones (18–24 h p.i., yellow and grey dots, respectively).
To validate the microarray results, we analyzed 6 transcripts by quantitative RT-PCR (qRT-PCR), selected for their top ranking positions on the DE gene list at 6 h p.i. (see Additional File 1) or for their known function in the cell. Correlation analysis was performed by comparing expression ratios from microarray results with the ratios determined by the qRT-PCR analysis (Figure 3). Although the fold changes were under-estimated in the microarray results, a significant correlation was observed between the two assays (Pearson correlation coefficient r = 0.88, R2 = 0.78).
Overall, these results show more under-expressed than up-regulated genes in infected hepatoma cells (Figure 4A, blue stacked bars); seven hundred and eighty five transcripts are uniquely DE at a single time point while 323 probesets are so in at least two time points (Figure 4B). We also find that the number of DE host genes steadily decreases, from 417 at 6 h p.i. to 163 at 24 h p.i. (Figure 4A). These data suggest that host responses to parasite infection are more prominent at early time-points, causing more profound changes in gene expression during the initial stages of infection than during subsequent developmental processes occurring within the study's timeframe. In summary, our global analysis indicates that infection by Plasmodium leads to a sequence of cellular events that involves both discrete and sustained expression modulation responses in the host cell, as well as responses that are controlled in a temporal fashion during parasite development (Figure 4C).
Time-course independent host cell responses to Plasmodium infection
Interestingly, 24 genes (32 transcripts) playing a role in transcription/translation, transport, signaling, apoptosis, cell growth and differentiation, proteolysis, cytoskeleton, and steroid biosynthesis are DE at all times in infected cells compared to non-infected ones (Figure 4B and see Additional File 2). Although no common pathway or gene ontology can be clearly identified for the whole set, 14 genes are up-regulated in all the time points assessed. They include three endoplasmic reticulum (ER)-stress related genes: trib3, nupr1 and ddit3 that can act in concert to induce apoptosis under the right stimuli [16] as shown during Hepatitis C virus infection [17]. As a matter of fact two of these genes, trib3 and nupr1, systematically ranked in the top 5 of genes DE throughout parasite development (see Additional File 3). However, the same ER-stress induced complex, involving Atf4, which binds to Ddit3 through a DNAzip binding domain, has been shown to regulate the heme oxygenase 1 (hmox-1) gene expression as a defense mechanism against oxidative stress [18]. Interestingly, atf4 is also up-regulated in our analysis from 6 to 18 h p.i, as is hmox-1 (top ranking at 18 h p.i; see Additional Files 4 and 1), suggesting a clear definitive and sustained response to oxidative stress caused by the parasite (see Additional File 1). Those ER-stress induced factors are attractive candidates to stimulate Hmox-1 expression, which we have shown to promote P. berghei liver infection [19]. Trib3 has also a clear negative modulatory effect on the insulin signaling pathway in hepatocytes that is responsible for glucose homeostasis [20]. Whether these metabolic pathways are important for Plasmodium development inside the hepatocyte remains to be elucidated.
It is also compelling to find that some host molecules are down-regulated in infected cells throughout the entire period of the experiment. Among those 10 transcripts, there are: 2 G-protein coupled receptors, which have a role in signal transduction; protein C, which is a regulator of the blood coagulation cascade; maf, which encodes an oncoprotein promoting cell differentiation; and 2 solute carriers, MCT4 and ZIP8, which, if down-modulated, have been shown respectively to help restore calcium homeostasis [21] and impair mitochondrion function in response to TNFα [22]. Mct4 and zip8 rank in the top 5 DE genes in our analysis, along with an ABC transporter (abcd2) that is also down-regulated during the whole course of parasite development, and trib3 and nupr1 (see above) that are up-regulated at all the time points assessed (see Additional file 3 and 1).
Time-course specific host cell responses to Plasmodium infection
To characterize all the DE transcripts biologically, an enrichment analysis of modulated genes by their functional annotation with Gene Ontology (GO) and KEGG metabolic pathways was performed at each time point (Figure 5A and 5B, respectively).
According to the GOslim analysis, most receptor binding activity, as well as stress and stimulus response-related genes, are modulated by the parasite in the early stages of infection. (Figure 5A, pink stacked bars). On the other hand, this analysis further shows that metabolic processes are then engaged at a later stage of the infection process. (Figure 5A, purple and blue stacked bars). The early responses are possibly mediated by cytokine and subsequent pro-inflammatory signaling events (Figure 6A, 6 and 12 h p.i in infected cells) that, in turn, induce fast cell cycling, as indicated by the KEGG metabolic pathways specifically enriched at 6 h p.i. (Figure 5B, pink stacked bars). This type of response is somewhat similar to that of human foreskin fibroblasts (HFF) infected by Toxoplasma gondii where both stress and cytokine/chemokine host genes are DE in the early stages of infection [23]. In T. gondii infection this response seems to be triggered by induction of il-1β and tnf-α and followed by the JAK/STAT1 signaling pathway [24]. However, in P. berghei-infected cells only TNF signaling is significantly modulated. Both tnfaip2 and tnfaip3 are up-regulated at 6 h p.i. and tnf receptors are down-regulated from 6 to 18 h p.i. (except for the anti-apoptotic receptor molecule traf1, which is up-regulated at 6 h p.i.). Neither jak nor stat1 are in our final list of DE genes (see Additional File 1) although stat5a is up-regulated at 6 h p.i. and jak2 is down-regulated at 12 and 18 h p.i., albeit not significantly. Thus,, it is not clear from the present experiment whether malaria parasites modulate the JAK/STAT1 pathway as happens during Toxoplasma infection [24].
Similarly to T. gondii [25, 26], P. berghei parasites manipulate host cell NF-κB signaling. NF-κB transcription factors have been proposed to act as a "first responder" to the cellular stimuli caused by traversal of host cells by Plasmodium prior to invasion [27]. To identify targets or interacting molecules, we cross-checked the repertoire of cytokines and chemokines produced by infected liver cells with known target genes of NF-κB http://people.bu.edu/gilmore/nf-kb/target/index.html. Ccl2, a member of the CC subfamily of chemokines, is up-regulated throughout parasite development and ccl5 (rantes) is up-regulated between 6 and 12 h p.i (see Additional Files 1 and 5). Cxcl1 (gro-α), and cxcl2 are members of the CXC subfamily of chemokines that are also up-regulated, from 6 to 18 h p.i. or at all time points, respectively (see Additional Files 3, 1 and 5). Additional NF-κB-induced factors DE at early stages of infection include cell adhesion molecules icam1, and vcam1, as well as the transcription factors nfkb1b and nfkb1z (see Additional File 1).
Interestingly, it has been previously shown that T. gondii also manipulates the MAPK signaling pathway through LPS/TLR4 engagement [28]. Although in the present experiment none of the Toll-Like Receptors were DE, the MAPK pathway seems to be up-regulated through the production of gadd45α/β and γ from 6 to 24 h p.i. (see Additional File 4), mkp from 6 to 12 h p.i. and ikk from 18 to 24 h p.i. (see Additional Files 1 and 5). The MAPK pathway was also shown to be up-regulated in B-type hepatocellular carcinoma and during Hepatitis B Virus (HBV) infection [29, 30], indicating that up-regulation of signal transduction-related genes is not pathogen-specific. This suggests that cell traversal or invasion by sporozoites may engage several host receptors and that these events constitute, perhaps not surprisingly, a stress-inducing stimulus for the cells, leading to a cascade of signaling events through chemokines and cytokines that might have positive or negative effects on the establishment of the liver infection. In fact, one such gene is hmox-1, which is up-regulated in infected cells at all the time points assessed, significantly so at 18 h p.i. (see Additional Files 4, 1 and 5), as well as in macrophages and Kupffer cells in the liver [19]. As mentioned above, Hmox-1 is a major stress-response molecule that we have shown to promote liver infection by controlling the extent of the host inflammatory response [19]. Interestingly, Hmox-1 has pronounced antiviral activity in HBV infection [31] while it can promote Mycobacterium tuberculosis latent infection in host [32].
At 12 h p.i., with 287 DE genes, the parasite seems to primarily modulate the host cell's transcriptome towards biosynthetic pathways including carbohydrate and fatty acid metabolism. This correlates well with host responses observed during early T. gondii infection that has been reported to be a general reaction to stress rather than a specific parasite induced response [23]. Genes involved in cellular homeostasis, transport and cell communication are also modulated following productive invasion.
At 18 h of parasite development 451 DE genes playing a role in lipid metabolism (35 genes), amino acid and nucleic acid metabolism (39 genes), as well as cell cycle (58 genes) and catalytic activity (33 genes) are significantly modulated (Figure 5; see Additional Files 1 and 5). This picture may suggest that once a host cell is successfully infected by Plasmodium, its resources remain engaged in energy production and homeostasis, possibly in an attempt to cope with the imbalance caused by the intracellular presence of the parasite. Indeed, the liver stage parasite goes through an extensive proliferative phase during this period, likely requiring large amounts of proteins and lipids to produce tens of thousands of merozoites that will in turn invade the red blood cells (reviewed in [5]). Consistent with this notion, our analysis shows that the parasite's presence down-modulates signaling trough the PPAR and adipocytokine pathways around 12 to 18 h p.i. (Figure 5B, purple and blue stacked bars) leading to increased lipid metabolic processes (Figure 5A, purple and blue stacked bars; Figure 6B; see Additional File 5). It has been described that a parasitophorous vacuole membrane protein encoded by the parasite, UIS3, specifically binds to the liver-fatty-acid binding protein (L-FABP) [33]. By engaging L-FABP, the parasite would reduce PPAR activity, resulting in an increase in fatty acids available to it. As such, it is tempting to speculate that the parasite influences host cell metabolism in its own interest. Also, transport molecules are engaged from 6 to 18 h p.i Whether this is important to provide the developing parasite with the components it requires for its rapid proliferation and/or host cell survival remains to be established (Figure 5A; see Additional File 5).
At 24 h p.i., with 163 DE genes, glutathione metabolism is the major host pathway modulated by the parasite (Figure 5B). Glutathione metabolism and transport participate in many cellular reactions including antioxidant defence of the cell, drug detoxification and cell signalling (involved in the regulation of gene expression, apoptosis and cell proliferation). Alterations in glutathione concentration have also been shown to be a common feature of many pathological conditions including diabetes, cancer, AIDS, neurodegenerative and liver diseases [34]. Therefore, one can speculate that the parasite stimulates the glutathione pathway to help sustain the host cell in the final stages of development.
Finally, it is interesting to note that genes involved in cell differentiation and death, through the p53 signalling pathway, are DE throughout the whole time span covered in this study (Figure 5A and 5B, pink, purple, blue and green stacked bars; see Additional Files 5). From the earliest stages of infection, cell survival and apoptosis mechanisms are actively at play (Figure 6C, mainly down-regulated transcripts in infected cells in early time points). Malaria hepatic infection involves the activation of the tyrosine kinase HGF receptor (MET) by secreted HGF [35] and down modulation of MET leads to a decrease in infection [8, 35]. In our analysis, met is over-expressed at 6 h p.i (see Additional File 4). It has been shown that the malaria parasite exploits MET not as a primary binding site, but as a mediator of signals that make the host susceptible to infection. This signalling pathway might influence host actin cytoskeleton [35] and apoptosis [36] during infection (Figure 6D and 6C, respectively) and play an important role in the survival of the infected cell. Indeed, P. berghei-infected hepatocytes are more resistant to apoptosis induced by different stimuli throughout infection [36, 37], and require an actively developing parasite [38]. Hence, one might therefore suggest that pro- and anti-apoptotic genes are actively modulated throughout infection, in a constant struggle between the infected cell's tendency to undergo apoptosis and the parasite's need for the host cell to survive (Figure 6A and 6C).
Similarities between the rodent malaria species P. yoelii and P. berghei
A complementary in vivo assay was carried out on liver cells from mice infected with P. yoelii at 24 h, when most liver stage parasites are at a small early schizont stage and have undergone two to four rounds of nuclear division, and 40 h p.i., when most liver stage parasites are large late schizonts and have undergone up to 13 rounds of nuclear division [39], and compared to the findings described above. We found that 67% of genes DE in P. berghei-infected Hepa1-6 cells are also expressed in P. yoelii-infected livers. Although parasite development in cell culture and in vivo is known to proceed at different rates [40], and two different microarray platforms were used, for two different parasites species, a sizeable 42% of the genes DE in P. berghei infected cells are also found DE in P. yoelii-infected livers in at least one time point (see Additional File 6). Twenty-six transcripts are found DE in both experiments with the highest agreement in terms of up or down-modulation (16 transcripts) and common time point (10 transcripts at 24 h p.i.) (Grey shaded boxes in Additional Files 6 and 7). It is reassuring to find that some of those transcripts have been discussed above as identified in top ranking positions during P. berghei infection and include ccl2, cxcl2, G protein-coupled receptor gpr137B, and gpr155, nr4a2, abcd2, hmox-1 and klf4. These findings corroborate the validity and strength of our in vitro model, and highlight those genes or the pathways they are involved in as common and important factors in host response to rodent parasites.