By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
The human pathogenic fungus Candida albicans can cause a wide range of infections and invade multiple organs. To identify C. albicans genes that are expressed during invasion of the liver, we used genome-wide transcriptional profiling in vivo and ex vivo. By analysing the different phases of intraperitoneal infection from attachment to tissue penetration in a time-course experiment and by comparing the profiles of an invasive with those of a non-invasive strain, we identified genes and transcriptional pattern which are associated with the invasion process. This includes genes involved in metabolism, stress, and nutrient uptake, as well as transcriptional programmes regulating morphology and environmental sensing. One of the genes identified as associated with liver invasion was DFG16, a gene crucial for pH-dependent hyphal formation, correct pH sensing, invasion at physiological pH and systemic infection.
The human fungal pathogen Candida albicans can cause a variety of infections ranging from superficial mucosal infections to haematogenously disseminated infections. Intra-abdominal or invasive C. albicans infections are life-threatening diseases after liver transplantation with mortality rates up to 60% (Boos et al., 2005; Fischer and Sterneck, 2005). Furthermore, C. albicans may cause peritonitis during continuous ambulatory peritoneal dialysis (Michel et al., 1994). During the course of this type of infection, C. albicans can penetrate almost all organs within the intraperitoneal cavity except the kidneys (Kretschmar et al., 1999a). Once inside the organ tissues, the fungus can enter the blood stream and may disseminate throughout the entire body, often causing systemic multi-organ infections (Filler and Kullberg, 2002).
Depending on the site of infection and the events associated with the infection process, C. albicans has to adapt rapidly to changing environments (Hube, 2004). This rapid adaptation can be observed at the transcriptional level. For example, it has been shown that C. albicans can modify its global transcriptional profile within 10 min of exposure to human blood (Fradin et al., 2003). In this experimental setting, the granulocytes of the human blood are the dominant host cells that alter fungal gene expression (Fradin et al., 2005). Other experimental conditions that mimic characteristics of certain host environments, such as changes in pH (De Bernardis et al., 1998; Bensen et al., 2004), iron starvation (Lan et al., 2004), exposure to nitric oxide (Hromatka et al., 2005), stress (Enjalbert et al., 2006) and interaction with macrophages (Lorenz et al., 2004) or neutrophils (Rubin-Bejerano et al., 2003) also affect the expression of sets of genes that are particularly important for the adaptation to each of these conditions. Additionally culture conditions induce transcriptional programmes such as the one that regulates the yeast-to-hyphal transition (Nantel et al., 2002; Sohn et al., 2003). This morphological programme has been associated with multiple functions, in particular virulence attributes (Kumamoto, 2005; Kumamoto and Vinces, 2005). It is likely that numerous genes are crucial for growth and survival in each of the microenvironments within the host during commensal growth and/or induction of disease. However, we postulate that subsets of genes are important and optimized for infection in a stage- and tissue-specific manner.
For a number of different environmental and culture conditions, marker genes have been identified whose expression indicates distinct biological functions. For example, the yeast-to-hyphae transition of C. albicans, known to occur within tissues, correlates with the expression of genes associated with this morphological transition. Hyphae-associated genes include SAP4, SAP5 and SAP6, which encode secreted aspartic proteinases (Hube et al., 1994; Felk et al., 2002), ECE1, the cell elongation gene (Birse et al., 1993), HWP1, a gene encoding a cell wall-linked adhesion protein acting as a substrate for host transglutaminase (Staab et al., 1996; Staab and Sundstrom, 1998), members of the ALS adhesin gene family, in particular ALS3 (Zhao et al., 2004; Kadosh and Johnson, 2005), and Tup1-regulated genes such as RBT5 (Braun et al., 2000). RBT5 codes for a haem binding protein associated with iron uptake and it is upregulated under low-iron and alkaline conditions (Bensen et al., 2004; Lan et al., 2004). PHR1 and PHR2 encode glycosidases and are marker genes for the environmental pH. PHR1 is expressed at neutral to alkaline conditions, whereas PHR2 is expressed at acidic pH (Saporito-Irwin et al., 1995; Muhlschlegel and Fonzi, 1997). Detecting the expression of each of these marker genes during infection may provide valuable information about the microenvironments that fungal cells are exposed to in the host and about the pathogenesis of fungal infections in general. However, the genes that are expressed during the different types of infection and during tissue invasion are largely unknown.
Even slight alterations in the host's physiological state can turn the normally harmless commensal yeast C. albicans into an aggressive pathogen (Hube, 2004). However, it is not only the host that governs the transition from commensalisms to infection or the outcome of an infection. For example, it has been shown that different C. albicans isolates have a different capacity to invade epithelial tissue (Bartie et al., 2004) or to respond to macrophages (Tavanti et al., 2006). In particular, the C. albicans strains SC5314 and ATCC10231 differ dramatically in their ability to invade into host tissue. Strain SC5314 (Gillum et al., 1984) is a widely used laboratory strain with high virulence potential that was chosen for the C. albicans genome sequencing project (Jones et al., 2004). In contrast, strain ATCC10231, which was isolated from a patient with bronchomycosis (http://www.lgcpromochem-atcc.com/) and is commonly used for drug susceptibility testing (Feng et al., 2005; Blanco et al., 2006), is non-invasive (Kretschmar et al., 1999a,b) and strongly reduced in its virulence potential in all animal models tested (Balish and Phillips, 1966; Phillips and Balish, 1966; Schmidt and Geschke, 1996). However, the mechanisms underlying the different invasive properties of these two strains are unknown.
The aim of the current study was the identification of genes that are expressed during invasion of parenchymal organs such as the liver by (i) using genome-wide in vivo transcriptional profiling of the different phases of infection from attachment to tissue invasion, (ii) comparing the in vivo transcriptional profiles of an invasive strain (SC5314) with those of a non-invasive strain (ATCC10231) during intraperitoneal infection, and (iii) comparing data obtained from the intraperitoneal infection model with those obtained from an ex vivo model of infection. By analysing these transcriptional profiles, we identified genes and transcriptional pattern that are associated with the invasion process. One of the genes identified as associated with liver invasion was DFG16, a gene essential for invasion at physiological or alkaline pH.
Strain ATCC10231 is non-invasive in a transmigration assay
To enable us to compare the in vivo transcriptional profiles of an invasive with those of a non-invasive strain, we searched for a strain with reduced ability to invade tissue and extracellular matrix. The invasive properties of different C. albicans strains have been investigated in vivo (Kretschmar et al., 1999a), in vitro using cell monolayers (Zink et al., 1996), and tissue models (Bartie et al., 2004). Strain ATCC10231 has been shown to have dramatically reduced capacity to invade the liver in a mouse model of Candida peritonitis (Kretschmar et al., 1999a). We therefore compared the transmigration through extracellular matrix (Matrigel) of this strain with that of strain SC5314 using a modification of the invasion assay developed by Crowe et al. (2003). As shown in Fig. 1, strain ATCC10231 was almost non-invasive in this transmigration assay while large numbers of SC5314 cells penetrated through the matrix. During transmigration, SC5314 cells produced normal hyphae, whereas strain ATCC10231 produced only very short filaments (pseudo- and true hyphae) on the surface and within the matrix (not shown). However, cells of strain ATCC10231 were able to produce slightly shorter, but otherwise normal hyphae under most in vitro hyphae-inducing conditions tested (except when incubated in medium with N-acetylglucosamine; see below). These data confirmed the different invasion properties previously observed in vivo (Kretschmar et al., 1999a). Therefore, we used strain ATCC10231 for our subsequent investigations and used the transmigration assay to investigate strains with potential defects in invasion properties (see below).
Intraperitoneal infections in mice
In order to obtain infected tissue, mice were infected intraperitoneally with the C. albicans strains SC5314 or ATCC10231 as described in the Experimental procedures. Liver sections were stained for histology to follow the infection process of both strains microscopically (Kretschmar et al., 1999a; Felk et al., 2002). We observed that cells of strain SC5314 begun to attach to the liver 30 min after injection. Between 3 h and 5 h, hyphae of strain SC5314 penetrated through the liver capsule and into the liver tissue. In contrast, cells of strain ATCC10231 attached to the liver capsule after 30 min, but remained on the surface between 3 h and 5 h, unable to penetrate into the tissue. Both strains of C. albicans formed hyphae after attachment to the liver for 3 h. Although it was incapable of invading into the liver (or the spleen and pancreas) strain ATCC10231 persisted in the intraperitoneal cavity and proliferated during the infection. At later time points, it spread via haematogenous dissemination to the lungs, heart and kidney (Kretschmar et al., 1999a).
An ex vivo infection of liver mimics an in vivo infection
Though the intraperitoneal mouse model is an excellent model to study invasion of C. albicans into parenchymal organs (Kretschmar et al., 1999a), the total amount of infected liver tissue that can be dissected for microarray analysis is low. To obtain larger amounts of infected tissue in a highly reproducible model, we used an ex vivo haemoperfused porcine liver infection model (Fig. 2A). We have recently shown that this model is very useful for studying invasion of C. albicans into a parenchymal organ (Thewes et al., 2007). Importantly for our study, this model allows the dissection of large amounts of infected tissue. Furthermore, in this model, we were able to investigate the infection process of the two C. albicans strains on different regions of the same liver, which minimized animal-to-animal variations in the infection process, and thus the transcriptional response of the fungal cells. Histological analysis of sections at the sites of infection revealed that strain SC5314 invaded the liver similarly both ex vivo and in vivo, although adherence to and invasion of the liver occurred more slowly ex vivo (Fig. 2B and Felk et al., 2002). According to the different stages of the infection process in the ex vivo model, we isolated infected tissue at time points 5 h (attachment), 7 h (superficial invasion) and 12 h (deep invasion) post infection.
In vivo and ex vivo transcriptional profiling
In vivo transcriptional profiling of microbes infecting a host is restricted by a number of technical problems. These include the fact that almost all samples contain low numbers of the infecting cells and high amounts of host cells. In our hands, even low amounts of C. albicans cells can provide robust in vivo and ex vivo gene expression data when the isolated mRNA was amplified and labelled with a linear amplification system. Linear amplification systems have been evaluated in our and several other laboratories and was shown to be reliable and highly reproducible (Wang et al., 2000; Pabon et al., 2001; Feldman et al., 2002; Wang, 2005). In addition, the presence of host cell mRNA in the infection samples was not a central problem per se because hybridization with labelled cRNA from isolated host cell tissue did not show any substantial cross-hybridization with the PCR products on the C. albicans microarrays used (not shown). However, we found that the relation of host cells to C. albicans cells in the samples was of crucial importance. High ratios (> 5:1) resulted in weak signals on the microarrays (not shown). Differential lysis approaches to remove host cells were rejected as our own data showed that C. albicans cells change their cellular transcripts within minutes in adaptation to any environmental changes including those required to lyse host cells. Furthermore, yeast cells were found to be more resistant to the lysis procedures than were hyphae (data not shown). Consequently, any differential lysis protocol would enrich yeast cells and thus transcripts from yeast cells. Therefore, only tissue biopsies containing at least 20% C. albicans cells were shock frozen and directly used for RNA isolation and microarray analysis.
Experimental design and global analysis of the in vivo and ex vivo transcriptional data
Genome-wide transcriptional profiling was used to identify genes that are expressed during attachment and invasion of parenchymal organs after intraperitoneal infection and ex vivo infection. The experimental design was based on two different strategies: (i) identification of genes expressed during a time-course after in vivo intraperitoneal or ex vivo hepatic infections (Fig. 2A) and (ii) comparison of the expression pattern obtained from the invasive and the non-invasive C. albicans strains. Based on our histological observations, we isolated RNA from cells of strain SC5314 or ATCC10231 by lavage of the intraperitoneal cavity 30 min after infection and from infected liver tissue from mice after 3 h and 5 h. We also isolated RNA from these strains from porcine liver after 5 h, 7 h and 12 h of infection in the ex vivo model. Furthermore, RNA from cells of the broth cultures that were used for both infection models was isolated and used for comparison (time point 0 h).
Our analysis of the transcriptional profiling results was based on the following hypotheses. (i) we expected that genes necessary for the invasion process would be expressed as soon as cells began to invade the tissue, but not in the early stages of infections, when cells were in the peritoneal cavity without contact to the liver or during the early attachment phases, (ii) we reasoned that genes necessary for invasion of tissue would be expressed in strain SC5314 between 3 h and 5 h in the in vivo model, and between 7 h and 12 h in the ex vivo model. However, these genes would not be expressed by the non-invasive strain (ATCC10231) at the same time points. To make each of the different data sets comparable, we used a common reference (labelled cRNA from mid-log grown SC5314 cells) for each microarray hybridization.
From the in vivo time-course experiment, we identified 1258 genes whose transcripts were detectable in at least three out of the four time points (0 h, 0.5 h, 3 h, 5 h) for both strains (combined analysis) using stringent conditions (see Experimental procedures). Only these technically reliable data were included in the subsequent analyses. The data sets were then used for a hierarchical clustering (Fig. 3A) and a vertical analysis (Murillo et al., 2005) where we compared gene expression data at each time point between the two strains (i.e. 0.5 h from SC5314 with 0.5 h from ATCC10231). Furthermore, in a horizontal analysis, where each time point was compared with the broth cultures of each strain separately (Murillo et al., 2005), we identified 787 genes for strain SC5314 and 1195 genes for strain ATCC10231 where a transcript was detectable in at least three out of four time points. These data sets were used for a detailed expression analysis to characterise the different stages of infection. Additionally, the expression of selected genes, for example genes known to be expressed under distinct environmental or physiological conditions or associated with morphology (marker genes) was analysed. Finally, we compared these data with those obtained from the ex vivo infection model.
Transcriptional profiles of SC5314 and ATCC10231 are similar prior invasion but differ during the invasion stages
Based on our histological investigation, we expected different global transcriptional patterns according to the different experimental stages: (i) growth in the broth culture, (ii) exposure to the peritoneal cavity, (iii) attachment to the liver and (iv) invasion into the liver.
Hierarchical clustering of the expression profiles of the 1258 gene whose transcripts were detectable at most time points from the two different strains in fact reflected these stages. As shown in Fig. 3A, the expression profiles of both strains were very similar in the broth culture and 0.5 h after injection into the peritoneal cavity. In contrast, the profiles from the broth culture were clearly different from all other profiles. Furthermore, cluster analysis showed that the profiles for the later infection stages (3 h and 5 h post infection) differed markedly between the invasive strain SC5314 and the non-invasive strain ATCC10231. Accordingly, the 3 h and 5 h profiles from strain ATCC10231 were more similar to each other as compared with the corresponding profiles from strain SC5314.
Genes expressed during the different stages of experimental infection in strain SC5314
When we analysed the expression pattern of the 787 genes of strain SC5314 during the course of infection in mice (horizontal analysis), we identified 255 genes that were upregulated at least twofold at the later time points as compared with the broth culture (time point 0 h; supplementary data S1). These genes could be divided into three groups: (i) upregulated at all time points in vivo (group A); (ii) upregulated at 0.5 h and 3 h (group B); and (iii) upregulated at 5 h only (group C) (Fig. 3B). This distribution reflected the different stages of infection: (i) general adaptation to the environment of the peritoneal cavity and liver tissue, (ii) attachment of fungal cells to the liver surface, and (iii) invasion into the liver. We also identified 221 genes that were downregulated at one or more of these time points compared with the broth culture (supplementary data S1). However, in contrast to the upregulated genes, we did not observe a clear division into specific subgroups (Fig. 3B).
A notable number (29%) of the upregulated genes of group A belongs to the functional categories metabolism and energy (supplementary Table T1). These genes are likely important for a general adaptation to the peritoneal cavity and survival within tissue. Expression of some of these genes indicates that the fungal cells utilize sugars as carbon source and use respiration rather than fermentation to produce energy. For example, PFK2, PDA1, PDX1, KGD1 and KGD2 were upregulated in C. albicans cells at all time points. PFK2 encodes phosphofructokinase, a key enzyme of glycolysis. PDA1 and PDX1 encode subunits of pyruvate dehydrogenase involved in acetyl-CoA biosynthesis, and KGD1 and KGD2 encode key enzymes of the tricarboxylic acid (TCA) cycle. However, at least a subpopulation of cells also seemed to be exposed to a low glucose environment, because we also observed the upregulation of PCK1, the gene encoding phosphoenolpyruvate carboxykinase, which is the key enzyme of the gluconeogenesis.
A second observation was that some genes associated with stress response were upregulated at all time points. These genes include those encoding the heat shock proteins Hsp78 and Hsp90, as well as the stress-associated protein Ddr48. Additional genes known to be expressed during thermal stress, such as HSP104, HSP12 and SSA4 were upregulated during the infection process. In contrast, we found no evidence that the organisms were exposed to oxidative, nitrosative or osmotic stress, as genes coding for catalase (CAT1), flavohaemoglobin (YHB1), or proteins involved in glycerol metabolism were not upregulated at any time point.
Other genes which were upregulated at all time points encode well-known factors associated with virulence or adaptation to the host environment such as SAP5, ALS3 or PHR1 (see below). Finally, expression levels of ACT1 coding for actin did not show any significant fluctuations during the infection process and was therefore used as a control (see also Fig. 3C and D).
The genes that were upregulated 0.5 h and 3 h post infection in mice (group B) included CDC61, FRS1, GRS1 and ILS1, which all encode tRNA synthetases. These results indicate that the protein synthesis machinery was active and that proteins possibly necessary for the new environmental conditions were produced (supplementary Table T1: protein synthesis). We also observed that the expression of genes encoding ribosomal proteins went down in the later stages of infection. Both of these results are similar to our previous findings, where we observed that C. albicans cells transferred from a broth culture to blood showed a high level of transcripts involved in protein synthesis as early as 10 min after inoculation and that genes encoding ribosomal proteins had reduced expression at the later time points (Fradin et al., 2003).
Although multiple genes were upregulated at 5 h post infection when cells had begun to penetrate into the tissue (group C), it was difficult to subdivide these genes into functional categories (supplementary Table T1). This finding was likely due to heterogeneity among the conditions to which the organisms were exposed. For example, at this time point, some organisms were still attached to the liver surface, whereas other organisms had already invaded into this organ. Nevertheless, we reasoned that some genes associated with invasion or survival within tissue should be upregulated at this time point. The liver appears to be an iron-poor environment for C. albicans, as indicated by the finding that genes associated with iron acquisition were upregulated at this time point. These genes include FET5, FTR1 and ZRT1, which encode proteins involved in iron and zinc uptake. Other genes known to be associated with iron deprivation (Lan et al., 2004) had higher expression levels at 5 h post infection. They included the iron uptake genes CFL1, RBT5 and FRE5, and the copper transporter gene CTR1. The upregulation of all these genes suggests that the availability of iron within the liver tissue is limited. Availability of iron for fungal cells within a human host is even more complicated in neutral or slight alkaline pH conditions such as those found in the liver tissue (pH 7.4) because the balance between the soluble Fe2+ ion and the more insoluble ferric form Fe3+ shifts towards the insoluble form. One of the proteins known to be responsible for intracellular pH regulation and, indirectly, for nutrient transport in C. albicans (Rao et al., 1993), is the major plasma membrane H+-ATPase Pma1. We found that PMA1 was upregulated at 5 h post infection, suggesting that the organisms might counteract the alkaline pH and the poor iron situation while in the liver. The transcript levels of other key genes known to be regulated in response to the external pH also indicate that cells were in fact exposed to alkaline pH during the course of infection. This includes the expression of the two key pH marker genes PHR1 and PHR2. The expression of PHR1, shown to be induced under alkaline conditions (Saporito-Irwin et al., 1995; Muhlschlegel and Fonzi, 1997), increased when cells from the broth culture (pH ∼5.5) were injected into the peritoneal cavity, attached to the liver surface and subsequently penetrated into the tissue (Fig. 3C). In contrast, the expression of PHR2, the acid-induced counterpart of PHR1 (Saporito-Irwin et al., 1995; Muhlschlegel and Fonzi, 1997), decreased dramatically within the first 30 min of infection. Interestingly, expression of PHR2 increased again slightly during the course of infection indicating that at least some cells are again exposed to an acidic environment. This is likely due to heterogeneity of the microenvironments of the organisms, for example, some organisms could be inside acidic phagosomes, whereas other cells were exposed to a more alkaline pH. Alternatively, this may be explained by the above mentioned upregulation of PMA1 which may be responsible for an active acidification of the direct surrounding tissue in the host. Other genes which were found to be upregulated under alkaline conditions (Bensen et al., 2004) and which were also upregulated in our in vivo data were hyphal-associated genes like ECE1 and HWP1, the phosphate transporter gene PHO84, and the Na+ efflux transporter gene ENA22. The last two genes were in fact most highly upregulated at time points 0.5 and 3 h (group B).
While iron seemed to be limited during infection, nitrogen seemed to be available in sufficient amounts because the expression level of genes associated with amino acid deprivation (Rubin-Bejerano et al., 2003) remained unchanged or were even downregulated over time as compared with the broth culture (supplementary data S2). These genes included GCN4, which encodes the key transcriptional activator of amino acid metabolism and genes associated with methionine or arginine biosynthesis or other genes involved in amino acid metabolism. Therefore, one can conclude that C. albicans is able to acquire peptides and amino acids during the invasion process from the host. Acquisition of the peptides and amino acids from the host may be facilitated by the activity of secreted aspartic proteases. These proteases, which are encoded by 10 SAP genes, are one of the best investigated virulence attributes of C. albicans and are known to play important roles in host/pathogen interaction (Naglik et al., 2004). The expression of SAP1 and SAP3 was not increased over time (supplementary data S2). In contrast, the expression of SAP2 was low at the early time points but increased after 5 h of infection. The hyphal associated genes SAP4, SAP5 and SAP6 were all upregulated as soon as cells where exposed to the host environment, with SAP5 being the most strongly expressed SAP gene upregulated at all time points in vivo.
Most, if not all cells that penetrated into the liver at the 5 h time point grew in the hyphal form. Therefore, it was not surprisingly that we detected the upregulation of more genes (in addition to SAP4–6) that either regulate the yeast to hyphae transition or are known to be expressed by hyphae. The genes include those encoding proteins such as the 14-3-3 protein Bmh2, the forkhead transcription factor Fkh2, and the transcriptional regulator Ssn6, all of which have been shown to govern the yeast to hyphae transition (Bensen et al., 2002; Cognetti et al., 2002; Hwang et al., 2003) (supplementary data S1). Similarly, the hyphae-associated genes, DDR48, SOD5, PHR1, ECE1, ALS3, HWP1 and RBT5 (Nantel et al., 2002; Kadosh and Johnson, 2005) were also upregulated after 5 h of infection (supplementary data S2).
Some of these genes, including the adhesin genes ALS3 and HWP1, were expressed at all time points in vivo, suggesting that hyphal proteins involved in the attachment to host surfaces are already produced at early time points during the course of infection. Finally, orthologues of genes known to be upregulated in Saccharomyces cerevisiae under anaerobic conditions (ter Linde et al., 1999) such as GAL10 and AGP1 where found to be higher expressed at later time points (supplementary data S1), which may reflect that some fungal cells invading into the liver tissue were exposed to a more anaerobic environment as compared with the intraperitoneal cavity.
Interestingly, the proportion of genes with unknown function within the three distinct groups rose from 13% at all time points (group A), to 19% at 0.5 h and 3 h (group B), to 26% at 5 h (group C) (supplementary Table T1). Among these genes of unknown function were ones that have no homologues in S. cerevisiae. We speculate that some of these genes might be associated with the pathogenic lifestyle of C. albicans. Approximately 25% of the genes in group B with unknown function and 50% of those genes in groups A and C have no homologues in S. cerevisiae (supplementary data S2). As these genes are induced upon invasion, one possible explanation may be that these genes have distinct functions during the invasion process.
Genes expressed in the non-invasive strain ATCC10231 and vertical analysis of the in vivo expression profiles
During intraperitoneal infection, C. albicans strain ATCC10231 was able to adhere to the liver surface and produce hyphae. In contrast to strain SC5314, it was completely unable to invade the liver (Kretschmar et al., 1999b; Thewes et al., 2007). By analysing the microarray data from the time-course experiment with strain ATCC10231, we identified 1195 genes whose transcripts were detectable in at least three out of four time points (horizontal analysis). However, in contrast to data obtained with strain SC5314, the 468 upregulated genes identified in the horizontal analysis did not show a clear distribution within the time points (Fig. 3B and supplementary data S3). Furthermore, we identified 92 genes whose expression levels were increased as compared with cells of the broth culture at one or more time points in both strains (supplementary data S4). However, multiple other genes had different transcriptional levels in strain ATCC10231 as compared with the invasive strain (supplementary Table T2 and supplementary data S3). For example, genes associated with iron and copper transport (such as FTR1 and CTR1) were not upregulated during infection in strain ATCC10231, suggesting that this strain has defects in environmental sensing, that the upregulation of iron-uptake genes in SC5314 is associated with invasion, or that the organism is only exposed to iron-limiting conditions when it is inside the liver. Similarly, although PHR1 was upregulated during the time-course in ATCC10231, the expression of PHR2 did not decrease within the first 30 min after infection. Therefore, cells of ATCC10231 either were exposed to both alkaline and acidic environment, or had a defect in sensing the external pH.
Using a vertical analysis approach (Murillo et al., 2005), we identified 248 genes that were expressed at least two times more strongly in strain SC5314 at one or more time points as compared with strain ATCC10231 (supplementary data S5). We also found 302 genes that were at least twofold more strongly expressed in strain ATCC10231 at one or more time points compared with strain SC5314 (supplementary data S5). One of the most striking groups of genes that were expressed more strongly in strain ATCC10231 were genes involved in ergosterol metabolism, such as ERG2, ERG7, ERG11 (ERG16), ERG26, ERG27 and ERG251 (supplementary data S5). Other metabolic genes indicated activation of the glyoxylate cycle in strain ATCC10231 (supplementary data S5). For example, the isocitrate synthase gene ICL1, the citrate synthase gene CIT1, the malate dehydrogenase gene MDH11 and the acetyl-CoA-synthetase gene ACS1 were all expressed more highly in strain ATCC10231 at one or more time points compared with strain SC5314, which only showed a moderate increase of ICL1 transcript levels at the last time point investigated (5 h).
Additionally, some of the genes described above that were expressed by strain SC5314 during invasion were expressed at a much lower level in strain ATCC10231. These genes included FKH2, FTR1 and PMA1. All three genes were only upregulated in strain SC5314 at the later time points of infection (3–5 h; supplementary data S5) and were not upregulated in strain ATCC10231, supporting the view that these genes may be associated with invasion.
Identification of genes associated with invasion of liver tissue
To enhance the power of our analysis, we compared the transcriptional profiles of the two C. albicans strains SC5314 and ATCC10231 in an ex vivo haemoperfused liver invasion model (Thewes et al., 2007). This approach enabled us to identify common genes that were associated with the invasion of C. albicans into a parenchymal organ (either mouse or pig liver). As histological examinations revealed that the invasion process was delayed in the ex vivo model as compared with in vivo invasion of liver in mice (see above), we analysed the transcriptional profiles of both strains at time points 5 h (attachment), 7 h (superficial invasion) and 12 h (deep invasion) post infection (see Fig. 2B).
Hierarchical clustering of the genes expressed in vivo at time points 3 h and 5 h and ex vivo at time points 5 h, 7 h and 12 h, showed that the 12 h time point from the ex vivo model was most similar to the 3 h and 5 h time points from our in vivo infection (data not shown). Therefore we concentrated our analysis on the comparison between the in vivo time points 3 h and 5 h after infection and the ex vivo time point 12 h after infection of strain SC5314.
We identified 63 genes that had at least twofold higher expression levels in SC5314 at these three time points as compared with the broth culture (Fig. 4A and supplementary data S6). Among these genes, were the heat shock genes HSP90, HSP78 and DDR48, the glycolytic marker gene PFK2, and the anaerobic marker genes AGP1 and GAL10. Furthermore, we identified ALS3 and PHR1 as upregulated during the invasion phases in both models. These data further support the observation made for the in vivo model in terms of metabolism, stress proteins, glycolysis and anaerobic growth.
Next, we eliminated from our analysis those genes that were expressed by both strains in both models because these genes are unlikely to be involved in invasion. By this process, we were able to identify genes that were more highly expressed by the invasive strain SC5314 during tissue invasion, but not by the non-invasive strain at the same time points. Again, we concentrated our analysis on the 3 h and 5 h in vivo time points and the 12 h ex vivo time point. We identified 142 (3 h) and 75 (5 h) genes for the in vivo time points and 358 genes for the ex vivo time point (12 h) that were (i) significantly more highly expressed compared with the common reference, and (ii) more highly expressed in strain SC5314 compared with strain ATCC10231 at all time points in both infection models. As shown in Fig. 4B, the majority of the genes that were more highly expressed in the ex vivo invasion phase were specific for this model (341 of 358). Similar results were obtained for genes that were more highly expressed in the in vivo model (77 of 142 for 3 h, 16 of 75 for 5 h and 51 in the intersection of these two time points). Nevertheless we identified five genes that were upregulated in the invasion phase in both infection models (Fig. 4B). These genes were orf19.6777 (unknown function), orf19.5446, a gene encoding a putative membrane protein with one membrane-spanning domain and regulated by the hyphal growth regulator Ssn6 (Garcia-Sanchez et al., 2005), FUM12, a putative fumarate hydratase, orf19.3428, a gene encoding a protein with a predicted PWWP domain for protein–protein interactions, and DFG16, a gene coding for a putative seven transmembrane receptor. As we expected that these genes were very likely to be associated with invasion of strain SC5314 into liver tissue, we decided to further analyse one of them in more detail.
DFG16, a gene predominantly expressed during tissue invasion of liver
One of the genes that was upregulated during the invasion stage in strain SC5314 in both the in vivo and the ex vivo model, but not upregulated in strain ATCC10231 at the same time points, was DFG16 (orf19.881). DFG16 shares homology with S. cerevisiae DFG16 which is involved in invasive growth upon nitrogen starvation (Mosch and Fink, 1997) and PalH of Aspergillus nidulans, which is putatively involved in pH signalling (Negrete-Urtasun et al., 1999; Herranz et al., 2005). Barwell et al. recently discovered that DFG16 is a member of the Rim101 pathway, which is responsible for the pH response of S. cerevisiae and C. albicans (Barwell et al., 2005). They showed that the corresponding gene, DFG16, in C. albicans is involved in alkaline pH-induced filamentation (Barwell et al., 2005). Based on our data and these results, we decided to further analyse DFG16 and its role in the invasion process of C. albicans.
A mutant lacking DFG16 does not produce hyphae and has growth defects at alkaline but not at acidic conditions
To analyse the function of DFG16 of C. albicans, we deleted 1172 bp of both alleles of the open reading frame with the URA-blaster protocol (see Experimental procedures). Ura-negative Δdfg16 mutants were transformed either with the empty URA3 containing plasmid CIp10 or with CIp10 containing the entire orf of DFG16, including 1025 bp of the 5′-untranslated region and 293 bp of the 3′-untranslated region. The mutant was screened under a broad range of conditions and on several different media. Only few, but very distinct phenotypes were discovered.
No yeast growth defects were observed under most conditions and filamentous growth was normal under hypha-inducing conditions, such as serum- or N-acetylglucosamine (GlcNac) containing media (Fig. 5A). In contrast, one of the most prominent phenotypes was the inability of Δdfg16 to form filaments on solid media at pH 8 (Fig. 5A). In addition, we found that Δdfg16 grew poorly under iron-limiting conditions at pH 8 (Fig. 5B). Similarly, Δdfg16 was more sensitive to elevated cation concentrations at pH 8 as compared with the wild type. These phenotypes were also observed at physiological pH (7.4), but were not observed at pH 5 and were restored in the DFG16-complemented strain (Fig. 5B). Interestingly, C. albicans strain ATCC10231 also showed a slight defect in hypha formation on M199 plates at pH 8. However, in contrast to the Δdfg16 mutant, strain ATCC10231 was unable to form hyphae on GlcNac plates (Fig. 5A). A further major phenotype was observed when the Δdfg16 mutant was grown under phosphate-limited conditions. Here, similar to the defects described above, the Δdfg16 mutant did not grow under low phosphate conditions at pH 8 or pH 7.4. In contrast, no differences to the wild type were detected at pH 5 (Fig. 5B) or in media with elevated phosphate concentrations (data not shown).
Dfg16 is located in the plasma membrane
As DFG16 codes for a protein with seven putative membrane-spanning regions similar to PalH of A. nidulans, it was expected that Dfg16 is located in the plasma membrane. For cellular localization of Dfg16 of C. albicans we constructed Dfg16–Gfp fusion proteins by adding the GFP gene at the 3′ end of DFG16. As a control we used transformants carrying pGFP or pACT1-GFP (Fradin et al., 2005). Immunoelectron microscopy using an anti-Gfp antibody revealed that the fusion protein is located predominantly in the plasma membrane of C. albicans (Fig. 6).
The transcriptional profile of Δdfg16 reflects a defect in pH sensing
Based on the observation that Dfg16 has seven putative membrane-spanning domains and a long hydrophilic C-terminal region similar to pH sensors in other fungi, Barwell et al. predicted that Dfg16 may be a surface receptor (Barwell et al., 2005). Our data obtained with the Dfg16–Gfp fusion protein and the Δdfg16 mutant support this view and further suggest that DFG16 is not only involved in alkaline-induced hyphal formation, but also iron and phosphate metabolism. If DFG16 is in fact a receptor responding to external pH, one would expect that the mutant has defects in sensing extracellular pH which should be reflected on the transcriptional level.
Therefore, we performed transcriptional profiling of the Δdfg16 mutant under mild alkaline conditions (physiological pH 7.4) in comparison with the wild-type strain. Sixty-one genes were significantly differentially regulated between the wild type and the mutant (30 up- and 31 downregulated; Table 1). Of these 61 genes, 18 (29.5%) had a putative Rim101 binding site (Ramon and Fonzi, 2003) in their promoter regions. For comparison, only 11.7% of all genes in the entire genome of C. albicans have a putative Rim101 binding site in their promoter regions.
Table 1. Genes differentially expressed in the Δdfg16 mutant compared with the wild type at pH 7.4.
Member of the phosphate permease family (by homology)
Sulphite sensitivity protein (by homology)
Putative transcription factor (by homology)
Amino acid permease (by homology)
pH-regulated protein 2
F1F0-ATPase complex, F1 beta subunit (by homology)
Repressed by TUP1 protein 5
GPI-anchored pH responsive glycosyl transferase
Chitinase 2 precursor
DNA-directed RNA polymerase I (by homology)
Peroxysomal 3-ketoacyl-CoA thiolase A (by homology)
Na+-coupled phosphate transport (by homology)
F1F0-ATPase complex, F1 alpha subunit
Amino-acid permease (by homology)
Inorganic phosphate transport protein, 3-prime end
High-affinity iron permease
Ferric reductase (by homology)
Nucleolar protein required for pre-18S rRNA processing
Component of the U3 small nucleolar ribonucleoprotein
Aconitate hydratase (by homology)
Glucan synthase subunit, 3-prime end
P-type ATPase involved in Na+ efflux (by homology)
fatty-acyl-CoA synthase, beta chain
Phospholipase B (by homology)
Putative cell wall protein (by homology)
Translation initiation factor eIF4B (by homology)
Zuotin, a putative Z-DNA binding (by homology)
Translation elongation factor 3
Heat shock protein 70
Similar to S. cerevisiae Pab1p
Mycelial surface antigen precursor
Among the genes that were upregulated in the Δdfg16 mutant strain were three genes, DDR48, RBT2 and PHO87 that are normally repressed under alkaline conditions by Rim101 (Bensen et al., 2004). These results suggest that the expression of these genes is derepressed in the Δdfg16 mutant and confirm that Dfg16 is involved in the Rim101 pathway. In contrast, SOD1, a gene coding for a superoxide dismutase and found to be upregulated in wild-type cells under alkaline conditions (Bensen et al., 2004), was also upregulated in the Δdfg16 mutant. Within the group of genes that were downregulated in the Δdfg16 mutant were a number of genes whose repression may explain the phenotypes of this strain. For example, FTR1, which encodes a high-affinity iron permease that is normally induced under low-iron conditions (Ramanan and Wang, 2000), was downregulated in the Δdfg16 mutant. Mutants lacking FTR1 exhibited a severe growth defect in iron-deficient medium and have significantly reduced virulence in mice (Ramanan and Wang, 2000). As alkaline conditions cause iron starvation in C. albicans (Bensen et al., 2004), repression of this gene under alkaline conditions in the Δdfg16 mutant should lead to a growth defect (Fig. 5B). In this context, it is worth noting the upregulation of HSP12 in the Δdfg16 mutant. Hsp12 has been shown to be induced under several stress conditions including low iron (Lan et al., 2004). Also, it was recently shown to be involved in copper metabolism, which is closely linked to iron metabolism (van Bakel et al., 2005).
Similarly, one reason for the higher sensitivity of the Δdfg16 mutant to cations may be a downregulation of genes involved in ion transport. In fact, we found that ENA22, a gene coding for an ATPase involved in Na+ efflux and normally upregulated in alkaline conditions (Bensen et al., 2004), is repressed in the Δdfg16 mutant. Downregulation of this gene in the Δdfg16 mutant at pH 7.4 may therefore lead to an increased sensitivity against high Na+ concentrations at alkaline pH (Fig. 5B). Additional genes repressed at alkaline pH in the mutant but upregulated in the wild type include genes involved in phosphate uptake such as PHO89 and PHO84. Both genes were previously shown to be alkaline upregulated with PHO89 being directly regulated by Rim101 (Bensen et al., 2004). This lack of upregulation in the Δdfg16 mutant may explain its growth defects under phosphate-limited conditions.
Finally, the most striking evidence for a dysregulation of the pH response in the Δdfg16 mutant strain under alkaline conditions was the expression pattern of the two key pH-regulated genes PHR1 and PHR2 (Saporito-Irwin et al., 1995; Muhlschlegel and Fonzi, 1997): the normally alkaline-induced gene PHR1 was repressed and the normally alkaline-repressed gene PHR2 was induced in the Δdfg16 mutant under alkaline conditions. Both genes are known to be regulated by Rim101 and the same dysregulation was discovered in a Δrim101 mutant (Ramon et al., 1999; Davis et al., 2000a; Bensen et al., 2004).
Δdfg16 is non-invasive in a transmigration assay
To determine if Dfg16 governs C. albicans invasion, we analysed the capacity of the Δdfg16 mutant to invade extracellular matrix (Matrigel) in vitro. As shown in (Fig. 7A), the Δdfg16 mutant was completely unable to invade into and transmigrate through the Matrigel while the wild-type and DFG16-complemented strains invaded the Matrigel in large numbers. Similar results were observed with a Δrim101 mutant (not shown). To investigate whether the reduced invasion was due to reduced adhesion to host cells, we tested the adherence of the Δdfg16 mutant to a fibroblast monolayer. No significant differences were seen in the ability of the Δdfg16 mutant to adhere to these host cells as compared with the wild type (data not shown).
Next, we investigated whether a reduced ability to form hyphae might be the cause of the non-invasive phenotype of the Δdfg16 mutant. Both the Δdfg16 and the Δrim101 mutants were unable to produce hyphae in contact with the Matrigel. As the pH of the surrounding cell culture medium was alkaline, we analysed whether the lack of hyphal formation was due to a defect in contact induction (Kumamoto, 2005) or a defect in pH response. Using the same culture medium, but a plastic surface, both mutants were still able to produce short hyphae at alkaline pH and almost normal hyphal length at acidic pH (Fig. 7B). Therefore, the non-invasive phenotype of the Δdfg16 mutant can be explained by its inability to form hyphae, which is due to both a defect in contact induction when in contact with Matrigel and a defect in alkaline pH sensing. Interestingly, the hyphal length of the Δrim101 mutant was significantly shorter than that of the Δdfg16 mutant (Fig. 7B). Similar differences in hyphal length between the wild-type strain and both mutants were also observed during contact to fibroblasts (data not shown). As a reduced invasive property into host tissue may be explained by a reduced induced endocytosis of the host cells (Filler et al., 1995) we analysed the ability of the Δrim101 mutant and the Δdfg16 mutant to induce their own endocytosis by fibroblasts. However, the observed difference in hyphal length had no effect in the endocytosis assay, as the fibroblasts endocytosed the Δdfg16 mutant similarly to the wild-type and DFG16-complemented strains (Fig. 8A).
DFG16 is essential for systemic infection, but not for epithelial tissue damage
Both the in vivo and ex vivo transcriptional data and the results from the in vitro invasion assay with the Δdfg16 mutant suggested that DFG16 expression is associated with invasion and that this gene is essential for invasion of extracellular matrix (Matrigel) or certain tissues. Therefore, we investigated the virulence properties of the Δdfg16 mutant. This mutant and the wild-type strain had similar capacity to damage epithelial cells in an in vitro model of oral infections [RHE model (Schaller et al., 1998); data not shown]. However, the Δdfg16 mutant had markedly attenuated virulence in a mouse model of haematogenously disseminated infection. Mice infected with this mutant had significantly improved survival, as well as significantly reduced renal fungal burden (Fig. 8B and C). The virulence of the Δdfg16 mutant was restored by reintegration of a wild-type allele of DFG16 (Figs 7 and 8).
In vivo transcriptional profiling
Genome-wide transcriptional expression profiles of microbes during infection provides information about the host microenvironment to which the microbe is exposed, as well as the gene expression changes that enable the microbe to adapt to its host niche (Hinton et al., 2004). Due to the technical problems associated with in vivo transcriptional profiling, very few studies dealing with genome-wide expression profiles of bacteria within host tissues have been reported (Hinton et al., 2004). Andes et al. have recently described a protocol to analyse the transcript profile from C. albicans populations in pooled kidneys from several mice (Andes et al., 2005). This study and other bacterial studies are based on a differential lysis approach that involves the physical separation of microbes from host tissue. However, as microbes respond extremely rapidly to changes in the environment, separation procedures have the potential to change the entire transcriptional profile of the microbe and make it difficult to discern their response to the host (Hinton et al., 2004). Here we describe the first detailed in vivo transcriptional profile of a human pathogenic fungus during different stages of tissue invasion into liver. Samples were obtained directly from infected tissue in vivo and ex vivo without separation and enrichment, and they were immediately frozen, avoiding environmental influences on the expression profile during any technical procedures. In addition, we took biopsies at distinct time points that represent characteristic stages of tissue interaction. Furthermore, we analysed the transcriptional profile of an invasive and a non-invasive strain to identify genes that are associated with the invasion process. The use of a common control allowed the comparison of all samples of both strains in vivo and ex vivo.
Stage- and strain-specific expression profiles
This experimental design allowed the isolation of multiple cells that were at the same stage of infection, exposed to a similar environment, and interacting with host cells in a similar manner (e.g. attaching and invading cells). However, we were aware of the fact that we still analysed a heterogeneous population of cells with overlapping transcriptional profiles. However, hierarchical clustering and careful analyses of distinct groups of genes clearly reflected the different stages of infection. For example, the pattern of the invasive (SC5314) and the non-invasive (ATCC10231) strain were similar at the early stages of infection, but differed at later stages when cells of the strain SC5314, but not of strain ATCC10231, invaded into the tissue. Therefore, the expression pattern changed when a new population of fungal cells appeared: those that had invaded the tissue. Similarly, the expression pattern of strain SC5314 at the 12 h time point from the ex vivo model was most similar to the 3 h and 5 h time points from our in vivo infection when most cells invaded the liver. Consequently, gene expression at these stages should reflect general adaptation to the environment of the peritoneal cavity and liver tissue, attachment of fungal cells to the liver surface and invasion into the liver.
Metabolism, stress and nutrients
As previously observed (Fradin et al., 2003), adaptation of C. albicans to changing environments appears to be rapid and involves the upregulation of the protein machinery (e.g. tRNA-synthase genes) to produce proteins required for the demands of a new milieu. Genes generally expressed in response to the peritoneal cavity include those that are expressed at both early and late time points in strain SC5314. For example, metabolic genes of the glycolysis, acetyl-CoA biosynthesis and the TCA cycle (PFK2, PDA1, PDX1, KGD1 and KGD2) were all upregulated in C. albicans cells at all time points. This may reflect that cells used six-carbon compounds via the glycolytic pathway and used respiration via the production of NADH2 in the TCA cycle to produce energy. However, some cells seemed to lack glucose, because PCK1, the gene encoding the key enzyme of the gluconeogenesis, was upregulated suggesting that two-carbon compounds were used. Similar observations were made by Barelle et al. who concluded that C. albicans displays a metabolic programme with activated gluconeogenesis and glycolysis at different time points of systemic infection (Barelle et al., 2006). Such different populations of cells are likely to overlap on the transcriptional level in our time-course experiments, which would also explain why we observed the expression of genes possibly reflecting anaerobic growth (GAL10, AGP1) at the same time points when genes of the TCA cycle were upregulated.
Interestingly, strain ATCC10231 did not show any upregulation of glycolytic genes, but had increased expression levels of genes involved in the glyoxylate cycle (ICL1, CIT1, MDH11 or ACS1), which are known to be upregulated in cells phagocytosed by neutrophils or macrophages or in other glucose-deficient environments (Lorenz and Fink, 2001; Lorenz et al., 2004; Fradin et al., 2005). Although these data suggest that a large portion of cells may be exposed to phagocytic cells, we found no further indications for interactions with macrophages or neutrophils. For examples, expression of marker genes for oxidative (CAT1) or nitrosative (YHB1) stress was low during the infection. In contrast, stress-associated genes such as HSP78, HSP90, DDR48, HSP104, HSP12 or SSA4 were upregulated during the infection process. However, as HSP78, HSP90 and SSA4 were shown to be downregulated in cells exposed to blood (Fradin et al., 2003) it is unlikely that phagocytic cells caused the higher transcript levels of these genes. Nevertheless, as Hsp90 is known to play a crucial role during infection as this essential protein has an immunogenic role and is the major target of protective antibodies against systemic Candida infections (Matthews et al., 1991; Matthews and Burnie, 1992), the expression of HSP90 during intraperitoneal infection is noteworthy. The exact reason for the upregulation of the heat shock protein genes remains unclear; however, at least one of these heat shock genes (HSP12) is known to be upregulated under iron/copper starvation (Lan et al., 2004; van Bakel et al., 2005).
The view that phagocytosis by neutrophils is a rather rare event during intraperitoneal infections is furthersupported by our observation that the expression profile did not indicate amino acid deprivation, which is induced when C. albicans is phagocytosed by neutrophils in vitro (Rubin-Bejerano et al., 2003; Fradin et al., 2005). Key genes known to be involved in nitrogen metabolism and expressed under low-nitrogen conditions such as GCN4 or genes associated with amino acid biosynthesis remained unchanged or were even downregulated during the course of infection. Aspartic proteases secreted by C. albicans may assist the fungus to gain peptides as a source of nitrogen within the host. For example, the hyphal associated protease genes SAP4–6, in particular SAP5, were expressed at high levels at all times points. Also, SAP2, which encodes the major protease for the utilization of proteins as a nitrogen source was upregulated at the later time points. Similar results have been found by Staib et al. (1999) using in vivo expression technology during intraperitoneal infection. Other nutrients essential for growth must be transported from the host environment into the fungal cells during infection and the gene expression profile observed in our study suggests that genes involved in transport of iron, copper, zinc and phosphate are upregulated (FTR1, CTR1, ZRT1, PHO84, PHO89). The importance of such genes for survival, virulence and invasion has been reported for C. albicans and other microbes (Lucas and Lee, 2000; Ramanan and Wang, 2000; Davis et al., 2002; Collins et al., 2003; Marvin et al., 2004; Peirs et al., 2005).
Attachment and hyphal formation
To attach to host surfaces, C. albicans needs to express adhesion factors. The most obvious genes known to be associated with adhesion are hyphal associated genes such as ALS3 and HWP1. These genes were in fact expressed at all in vivo time points suggesting that hyphal proteins involved in the attachment to host surfaces are already produced before cells were attached to liver, during attachment and during the invasion phase. However, the finding that these genes were expressed by both the invasive (SC5314) and non-invasive (ATCC10231) strain suggests that the products of these genes are not sufficient to induce tissue invasion.
Clearly, hyphal formation has a central function for C. albicans during invasion. Histology data show that almost all cells that penetrate through the liver capsule and into the hepatocytic tissue are hyphae (Fig. 2B). A distinct transcriptional programme is associated with hyphal formation, and includes, for example, the expression of SAP4–6, which has been shown to be crucial for invasion of organs during both intraperitoneal and disseminated infections (Sanglard et al., 1997; Felk et al., 2002). However, hyphal formation and the expression of hyphal associated genes is not sufficient to guarantee invasion as the non-invasive strain ATCC10231 did in fact produce hyphae and express these genes at a high level. Although it may be possible that some of these genes do not encode fully functional proteins in strain ATCC10231, it is more likely that this strain failed to express other genes that are important for tissue invasion.
The role of pH regulation during invasion
From the numerous in vitro data available, one can conclude that serum factors, starvation, anaerobic conditions, or contact to surface are all likely factors or conditions which may contribute to the morphological switch in vivo (Sanchez-Martinez and Perez-Martin, 2001). One further well-known condition that favours the yeast to hyphae transition is an elevated pH (Davis, 2003), and our data suggest that pH sensing plays a critical role in hyphal formation, tissue invasion and survival of C. albicans in vivo. The pH of the intraperitoneal cavity and the liver is expected to be around pH 7.4. Our transcriptional data suggest that such a pH was sensed by C. albicans (e.g. upregulation of PHR1) during intraperitoneal infections and that C. albicans responded to this pH not only by producing hyphal cells, but also by cellular adaptation.
One of the physiological consequences of a neutral pH is the reduced availability of soluble Fe2+ ions. Furthermore, a number of host proteins play key defensive roles against microorganisms by sequestering free iron, a defence system that has been called ‘nutritional immunity’ (Weissman and Kornitzer, 2004). By necessity, successful pathogens have therefore had to evolve mechanisms to acquire iron from host storage proteins. Not surprisingly, C. albicans has multiple ways of obtaining iron. This includes a low affinity and a high-affinity iron-transport system, the latter being essential under low-iron conditions such as those found in most host environments. Transcriptional profiling showed that genes involved in iron acquisition (FET5, FTR1) are upregulated in the 5 h cell population or had higher expression levels during the course of the infection process (CFL1, RBT5, FRE5, CTR1). As the expression of these genes was highest in the invasion phase, but not detected in the non-invasive strain ATCC10231, we conclude that these iron acquisition genes are particularly important during penetration of tissue, for example, because organisms that have invaded the liver are exposed to more iron-limited conditions compared with organisms that remain either on the surface of the liver or in the peritoneal cavity. These data are consistent with in vivo transcriptional data from invasive bacteria such as Yersinia pestis (Lathem et al., 2005; Sebbane et al., 2006), which showed that tissue invasion is associated with the upregulation of genes involved in iron and haem uptake.
Phosphate is another nutrient for which availability to C. albicans during infection seems to be linked to the environmental pH. Phosphate is an essential macronutrient important for the synthesis of nucleic acids and phospholipids. Adapting to phosphate-limited conditions is also crucial for persistence of intracellular and invasive bacteria such as Mycobacterium bovis, Mycobacterium tuberculosis and Salmonella typhimurium (Lucas and Lee, 2000; Collins et al., 2003; Peirs et al., 2005).
PHO84, encoding a putative high-affinity inorganic phosphate transporter, was shown to be upregulated during intraperitoneal infection with C. albicans. PHO84 and another phosphate transporter gene, PHO89, were previously shown to be upregulated under alkaline conditions (Bensen et al., 2004) and were regulated upon white-opaque switching (Lan et al., 2002). In addition, two homologous transporter genes in Yarrowia lipolytica were demonstrated to be under the control of extracellular phosphate and pH (Zvyagilskaya et al., 2001) and PHO84 of S. cerevisiae was transiently expressed upon phosphate starvation (Thomas and O'Shea, 2005). As PHO84 of C. albicans was most highly expressed 3 h post infection, but had lower expression levels during the invasion phase, it can be speculated that access to phosphate was limited in the peritoneal cavity, but was available during the invasion process.
During infection, the fungus may respond to changes in the pH of the host environment or may even actively modify the pH. For example, our data show that PMA1, encoding the major plasma membrane H+-ATPase of C. albicans, is upregulated during infection. Pma1 can act as a proton pump to regulate the intracellular pH of C. albicans, but also helps to create a sufficient membrane potential for secondary nutrient uptake, for example transport of phosphate.
In summary, these data show that C. albicans can respond and adapt to the host environment during intraperitoneal infections. One of the key events for this adaptation is the correct sensing of the environmental changes, such as changes in external pH.
DFG16 encodes a putative membrane sensor of the Rim101 pathway
One of the five genes that were shown to be closely associated with the invasion stage was DFG16. As previously shown, DFG16 is a member of the Rim101 pathway, which is responsible for the pH response of S. cerevisiae and C. albicans and is involved in alkaline pH-induced filamentation (Barwell et al., 2005). Similar to the gene encoding the pH sensor PalH of A. nidulans, DFG16 of C. albicans encodes a protein with seven putative transmembrane regions and a cytoplasmatic tail. Therefore, it is likely that Dfg16 is also a sensor located within the cell membrane. Our data obtained with a Dfg16–Gfp fusion protein show that Dfg16 is in fact located in the cell membrane, consistent with this hypothesis.
One of the most prominent phenotypes of the Δdfg16 mutant was its inability to form filaments on agar plates at pH 8 (see Fig. 5A and Barwell et al., 2005). Furthermore, the Δdfg16 mutant failed to produce hyphae and invade Matrigel at alkaline pH. Therefore, pH is a powerful inducing condition whose effects cannot be compensated by contact to surfaces (Kumamoto, 2005). However, our data show that induction of hyphal formation via contact with a surface depends on the nature of the surface. Although agar or Matrigel surfaces did not induce hyphal formation of Δdfg16, contact to hydrophobic plastic partially compensated for the pH-sensing defect. Therefore, it is very likely that hyphal formation of C. albicans cells in vivo is mediated by a mixture of environmental signals.
Aberrant sensing of the environmental pH due to the lack of Dfg16 has more effects: the Δdfg16 mutant grew poorly under iron-limiting conditions at pH 8, a phenotype also observed for Δrim101 mutants of C. albicans and S. cerevisiae (Lamb et al., 2001; Bensen et al., 2004). This mutant was also more sensitive to elevated cation concentrations at pH 8 (Lamb et al., 2001). The view that these defects are caused by an inability of the Δdfg16 cells to sense the environmental pH via the Rim101 pathway is supported by the observed changes on the transcriptional level. Several genes which were upregulated in wild-type cells at neutral or alkaline pH and during intraperitoneal infections, including genes associated with hyphal formation, iron or phosphate acquisition were downregulated in the mutant. Some of these genes were also dysregulated in the Δrim101 mutant (Bensen et al., 2004). Therefore, the Δdfg16 mutant was unable to respond properly to a changing environment, a vital attribute for survival and proliferation in the host. Not surprisingly, mutants lacking Dfg16 were almost avirulent in a mouse model of systemic infection, similar to mutants lacking the key transcriptional factor of the pH-sensing pathway of C. albicans, Rim101 (Davis et al., 2000b).
It should be noted that the Δdfg16 and Δrim101 mutants did not show entirely identical phenotypes and transcriptional profiles. For example, the Δrim101 mutant formed shorter hyphae than did Δdfg16, and genes such as CSA1, ECE1, HWP1, HYR1, IHD1 and RBT1 were downregulated in the Δrim101 mutant (Bensen et al., 2004), but not in the Δdfg16 mutant. These divergent results may be explained by differences in experimental conditions under which the two strains were analysed in the different laboratories or by the presence of an alternative pH sensor. Sequence analysis shows that Dfg16 possesses a LGR motif, which is highly conserved within PalH-like proteins. This motif has been shown to be necessary for interaction with PalF, the protein of the pH sensing pathway that links the sensor PalH with the downstream proteins of the PacC signalling pathway in A. nidulans (Herranz et al., 2005). The protein Rim21 of C. albicans and S. cerevisiae also belongs to this class of PalH-like proteins. blast analysis of the amino acid sequences of Dfg16 and Rim21 within the sequenced fungal genomes showed that these PalH homologues seem to be specific for eu- and hemiascomycotic fungi. Phylogenetic analysis showed that C. albicans, C. dubliniensis, S. cerevisiae (and other sensu stricto species), Debaryomyces hansenii and Yarrowia lipolytica were the only species that possess two different homologues of PalH, which are similar to either Dfg16 or Rim21. Therefore, the postulated second PalH-like protein may be Rim21.
Our data suggest that invasion of C. albicans is associated with the expression of a distinct set of genes and that the regulation of these genes depends on the sensing of the surrounding environment. At least two groups of genes are expected to be involved with invasion: (i) ones that are necessary to initiate and perform the invasion through barriers and tissue, and (ii) ones that respond to the changing environment during the invasion. It can be expected that some of these genes are expressed prior to invasion (to initiate invasion) and others during the invasion (to support deeper penetration and to respond to the changing host microenvironment).
Each of these genes may also function during a non-invasive lifestyle (e.g. during surface infection or even commensal growth), and invasion may depend on the co-ordinate regulation of these genes as well as on the host response. It is interesting to note that this view is similar to observations made for mammalian tumour invasion (Ozanne et al., 2006). Here, the authors report that certain receptors normally involved in cellular functions such as cytoskeleton organization or cell development are upregulated upon invasion, for example, the tumour invasion receptor CD44 (Marhaba and Zoller, 2004). These sensors initiate a genetic programme regulated by transcription factors such as AP-1 (Eferl and Wagner, 2003; Spence et al., 2006). This genetic programme includes genes encoding members of the matrix metalloprotease family of extracellular proteases that have been associated with invasion (Westermarck and Kahari, 1999). Inhibition or blocking of one of the sensors involved can prevent invasiveness of transformed cells (Ozanne et al., 2006). Identifying cellular receptors whose inhibition prevents tumour invasion and discovering inhibitors that can block these receptors are central aims of cancer research. Similarly, fungal specific sensors necessary for fungal invasion would be attractive drug targets.
In this study we have shown that DFG16 is one of the genes of C. albicans associated with invasion of liver. Mutants lacking this gene have multiple pH-dependent phenotypes and are avirulent. Other genes identified as invasion associated in our study may have similar important functions during invasive diseases of C. albicans.
YPD media buffered at pH 5 or pH 8 and depleted for iron were prepared as described containing 150 μM bathophenanthroline disulphonate (BPS) (Bensen et al., 2004). Cation sensitivity was tested using YPD medium buffered with 100 mM HEPES at pH 5 or pH 8 and supplemented with 1 M NaCl. Control plates were not supplemented with NaCl. Hyphal formation was investigated on M199 (Sigma) plates buffered at pH 8 and N-acetylglucosamine (GlcNac) plates (Mattia et al., 1982). SD medium depleted for inorganic phosphate (Pi-depleted) was prepared as described (O'Connell and Baker, 1992) and the pH value adjusted to pH 5 or pH 8.
In vitro invasion assay
The in vitro invasion assay was performed as described (Crowe et al., 2003) with modifications. Briefly, wells (12 mm diameter with 3 μm pore polycarbonate filters; Corning) were coated with 55 μl Matrigel (Becton Dickinson) diluted 1:1 with Dulbecco's modified Eagel's medium (DMEM; Biochrom) and incubated for 30 min at 37°C. After gel polymerization, 1.5 ml of DMEM was added into the lower compartment. Stationary phase wild-type or mutant strains were added to the upper compartment at 2.4 × 105 cells in 0.5 ml DMEM. Wells were incubated at 37°C for 24 h and the number of C. albicans cells in the lower compartment was determined by counting the colony-forming units on YPD agar after incubating overnight at 37°C.
In vivo and ex vivo infection models
For the in vivo transcriptional profiles 8- to 12-week-old female Balb/C mice were infected intraperitoneally with the indicated strains (5 × 107 cells in 0.5 ml PBS) and killed at the indicated time points as described (Kretschmar et al., 1999a,b). For the 30 min time point the intraperitoneal cavity was lavaged with 10 ml of PBS, centrifuged, and the fungal pellet was immediately frozen in liquid nitrogen. At later time points (3 h and 5 h post infection) biopsies were taken from infected liver tissue and immediately frozen in liquid nitrogen. Samples were stored at −70°C until further use.
The ex vivo infection of pig liver was performed as described (Thewes et al., 2007) and samples for RNA extraction were taken as described above. Pig liver samples for histological examination were fixed immediately in 2.5% glutaraldehyde with 2% paraformaldehyde in PBS (pH 7.4) and stored at room temperature until further use. Embedding, cutting and staining were performed as described (Schaller et al., 1999).
To induce haematogenously disseminated candidiasis, male Balb/C mice were infected via the lateral tail vein with 5 × 105 blastospores of C. albicans as previously described (Sanchez et al., 2004). For the survival study, groups of 11–12 mice were infected with each strain and monitored daily for 21 days. To determine tissue fungal burden, groups of 12 mice were infected with each strain. Six mice per strain were sacrificed on days 1 and 2 post infection. Their kidneys were harvested, weighed, homogenised and then quantitatively cultured.
RNA isolation and labelling
Frozen infected organ samples were lysed and homogenised by vortexing in PeqGOLD RNApure reagent (Peqlab) with acid washed glass beads (0.4–0.6 mm; B. Braun Biotech International) using a FastPrep machine (MP Biomedicals) for 2 × 30 s at 5.5 m s−1. For microarray analysis total RNA was extracted (Fradin et al., 2005) and labelled cRNA was prepared from 5 μg total RNA using the low input fluorescent linear amplification kit (Agilent Technologies).
For in vitro transcriptional profiling the C. albicans strains CAI-4 + CIp10 and Δdfg16 + CIp10 were grown overnight at 37°C at 180 r.p.m. in YPD buffered with 100 mM HEPES at pH 5 and used to inoculate YPD buffered at either pH 5 or pH 7.4. After 4 h cells were harvested and immediately frozen in liquid nitrogen. Total RNA was isolated and labelled as described above.
For C. albicans cells isolated from infection models, 2 μg of Cy3 labelled cRNA from the common reference (RNA from SC5314 grown in YPD to mid-log phase) and 2 μg of Cy5 labelled cRNA from tissue were co-hybridised to the microarrays overnight at 42°C in DIG Easy Hyb solution (Roche). Biological independent duplicates were performed for each time point. For transcriptional profiling of the Δdfg16 mutant at pH 7.4, 2 μg of labelled cRNA was co-hybridised with 2 μg of labelled cRNA from the wild type in biological triplicate including one dye swap as described above. The slides were washed at room temperature for 5 min in 2× SSC, 0.06% SDS, for 5 min in 0.4× SSC and for 5 min in 0.1× SSC and dried by centrifugation at low speed for 5 min. Hybridised slides were scanned immediately with an Axon 4000B scanner at a resolution of 10 μm and the data extracted using GenePix 4.1 software (Axon). Spots were flagged as ‘present’ if the median intensity was at least one standard deviation above the median background of the spot in both channels. ‘Marginal’ genes fulfilled this criterion only in one channel and spots were flagged as ‘absent’ if in both channels the median intensity of the spot was below 1× standard deviation of the background. The complete transcriptional data are available at http://www.galarfungail.org/data.htm.
All data were analysed using GeneSpring 7.2 (Agilent Technologies) after an intensity-dependent data normalisation (LOWESS). Robust gene expression of present and marginal genes was defined as normalised expression, which did not vary more than a standard deviation of 1.5 within the replicate arrays. Genes were defined as differentially expressed if their expression was at least 2× stronger at a given time point compared with the pre-culture (time point 0 h; horizontal analysis in vivo and ex vivo) or if their expression was at least 2× stronger in one strain compared with another at a given time point (vertical analysis; in vivo only). For the ex vivo profiles, genes were defined as highly expressed if the expression was at least 2× above the expression of the common reference. For in vitro transcriptional profiling, a gene was defined as differentially expressed if it was at least 2× stronger expressed in one strain compared with the other using a P-value cut-off of 0.05 (Student's t-test). Clustering of the in vivo profiles was performed in GeneSpring using a hierarchical clustering algorithm based on the standard correlation (Eisen et al., 1998).
Confirmation of microarray data by real-time RT-PCR
For the confirmation of the in vivo and ex vivo microarray data, expression of selected genes was analysed by real-time RT-PCR. For this, RNA abundance of selected genes was measured using the QuantiTect Probe RT-PCR kit and protocol (Qiagen) on a 7500 Real-Time PCR System (Applied Biosystems). For primer and probe sequences see supplementary table T4. Relative transcript abundance was determined with the 2−ΔΔCT method (Livak and Schmittgen, 2001) with RNA from the common reference (see above) as a calibrator and ACT1-abundance as an endogenous control. For each in vivo infection time point two independent RNA samples from two independent mice were analysed. For the ex vivo time point 12 h only one sample was analysed.
Expression of selected genes from the in vitro microarray data was confirmed by Northern blot (data not shown).
We used a modified Ura-blaster protocol (Fonzi and Irwin, 1993) for the disruption of DFG16. The DFG16 region from −199 to +193 was PCR amplified using the primers IPF9013-fwd (5′-TAACGAGCTCGGGTTTTGTTAGGACAGC-3′) and IPF9013-rev (5′-ATGTAGCATGCTTCAGGACCTATAATG-3′). The resulting fragment was cloned into the pCR2.1TOPO vector (Invitrogen). The plasmid was amplified with an inverse PCR from the DFG16-position 269–1441 using the primers 9013-KO-2 (5′-TAATAGATCTAGCATCCACATGGAAACAC-3′) and 9013-KO-5 (5′-AAACTGCAGAAGATGACGATGTCGAG-3′) containing a BglII and a PstI restriction site (underlined). The restricted fragment was ligated with the BglII/PstI digested 3 kb hisG-URA3-hisG fragment from pMB7 (Fonzi and Irwin, 1993) to give pKO-DFG16. This plasmid was linearised and transformed into strain CAI-4 as described (Walther and Wendland, 2003). Two rounds of Ura-blasting were performed to disrupt both alleles of DFG16 to give Δdfg16. Correct integration of the cassette was confirmed by PCR and Southern blot.
For the reintegration, a PCR fragment containing the native DFG16 gene was cloned into CIp10 (Murad et al., 2000), then amplified from position −1025 to +293 with the primers 9013-retrafo-fwd (5′-TGATGACGGAAAAGCAGGAG-3′) and 9013-retrafo-rev (5′-TGAGTTTGAAGGGAGAAGGG-3′) and cloned into pCR2.1TOPO. After digestion of the resulting plasmid with KpnI/XhoI the fragment was subcloned into CIp10 to give CIp10-DFG16. The plasmid was transformed after linearisation into Δdfg16 (Ura–) and single integration of the plasmid at the RPS1 (previously named RPS10) (Murad et al., 2000) locus was confirmed by Southern blot. As a control, the empty CIp10 plasmid was transformed into Δdfg16 (Ura–).
Dfg16/Gfp protein fusion
To construct pACT1-DFG16/GFP, GFP was fused to the 3′-end of DFG16. To achieve this, the actin promoter was PCR amplified from pACT1-GFP (Barelle et al., 2004) using the primers P-ACT1-fwd (5′-ATCGCTCGAGCTATTAAGATCACCAGCCTC-3′) and P-ACT1-rev (5′-ACCACCTCTAGATTTGAATGATATATTTT-3′) containing an XhoI and XbaI restriction site (underlined). DFG16 was PCR amplified from genomic DNA from position 1–1713 using the primer pair IPF9013-Fus1 (5′-AACCTCTAGAATGGGCTGTTCTGTTCTATATAC-3′) and IPF9013-Fus2 (5′-ACAAGCATGCGCCTTCCTTTCGGTCGTTC-3′) containing an XbaI and SphI restriction site (underlined). GFP was PCR amplified from pACT1-GFP using the primers GFP-1 (5′-GCATGCGGTGGTGGTATGTCTAAAGGTGAA-3′) and GFP-2 (5′-ACCAGCTAGCTTATTTGTACAATTC-3′) containing a SphI and NheI restriction site (underlined), and a glycine linker (italics). All PCR products were cloned into pCR2.1TOPO, digested and ligated. The ligation product was PCR amplified using the primers P-ACT1-fwd and GFP-2. The PCR product was gel purified and cloned into pCR2.1TOPO. After digestion the insert was used to replace GFP in pGFP (Barelle et al., 2004) to obtain pACT1-DFG16/GFP. The plasmid was integrated into the RPS1 locus as described above. Transformants carrying pGFP or pACT1-GFP were used as control. Immunoelectron microscopy for ultrastructural localisation of Dfg16 was performed as described (Fradin et al., 2005).
Hyphal length measurement and endocytosis
Hyphal length of the indicated C. albicans strains was measured after contact with plastic surface. Briefly, YPD-grown pre-culture cells were diluted (2 × 104 cells ml−1) in M199 buffered at pH 5 or pH 8 and incubated for 4 h in 24 well plastic dishes at 37°C. Hyphal length was measured microscopically using the Axiovert 200 microscope and the AxioVision 3.1 software (Zeiss).
For the endocytosis assay, fibroblasts (NIH 3T3) were routinely cultured in DMEM containing 15% FCS. The assay was performed as described (Jacquinot et al., 1998) with modifications: 105 fibroblasts were seeded onto 12 mm diameter glass coverslips coated with fibronectin and placed in a 24 well cell culture plate overnight. Near confluent fibroblasts were infected with 105Candida cells in DMEM without FCS. After 4 h of incubation at 37°C and 5% CO2 the medium was aspirated and non-adherent cells were removed by washing with PBS. Infected fibroblasts were fixed with 3% paraformaldehyde (Histofix; Sigma) for 30 min, washed with PBS and Candida cells were stained for 1 h with rat anti-C. albicans monoclonal antibody CA-1. After washing with PBS cells were counterstained with anti-rat IgM conjugated with Alexa Fluor 568 (Invitrogen) for 30 min. Fibroblasts were then permeabilised with 0.2% Triton X-100 and Candida cells were stained with calcofluor white for 20 min. Actin was stained with phalloidine for 20 min. The coverslips were dried and mounted inverted on a microscope slide. Morphology and endocytosis of the wild-type and mutant cells was observed under epifluorescence (Eclipse E600, Nikon) using filter sets to detect Alexa Fluor 568, calcofluor white and phalloidine. Images were taken with the Digital Camera DXM1200 and the ACT-1 2.63 software (Nikon).
The survival of mice infected with the different strains of C. albicans was compared using the Log-Rank test. Differences in organ fungal burden were analysed by the Wilcoxon Rank Sum test. P-values ≤ 0.05 were considered to be significant. All other data are presented as mean ± standard deviation. Differences were analysed by the Student's t-test. Again, P-values ≤ 0.05 were considered to be significant.
This work was supported by the Robert Koch-Institut, the Deutsche Forschungsgemeinschaft and the European Commission. S.G.F. was supported by Grant R01AI054928 from the National Institutes of Health, USA. We thank Aaron Mitchell and Alistair Brown for providing strains and plasmids and Julian Naglik for help in preparing the manuscript. Sequence data from Candida albicans were obtained from the Stanford DNA Sequencing and Technology Center website at http://www.sequence.stanford.edu/group/candida/index.html. Sequencing of C. albicans at the Stanford DNA Sequencing and Technology Center was accomplished with the support of the NIDR and the Burroughs Wellcome Fund.