A Core Filamentation Response Network in Candida albicans Is Restricted to Eight Genes

1The formation of filaments by Candida albicans is an essential attribute of this species with direct implications for tissue invasion and virulence. An extensive list of in vitro conditions including pH, temperature, nutrient sources, CO2 concentration and serum has been described to favor growth of C. albicans in either yeast or filamentous morphotype,. Whereas in early reports filamentous forms were described as “virulent” or “invasive” in contrast to the “commensal” yeast morphotype, numerous studies now suggest that morphological plasticity rather than a single morphotype are required during infection. This hypothesis is backed by the observation that mutants locked in either yeast (efg1Δ/cph1Δ) or filamentous (tup1Δ/nrg1Δ) morphotype show reduced virulence in infection models,. The central role of the shift towards filamentous growth has been emphasized in studies using mutant strains that retained their switching ability but showed either extended filamentation after derepression of UME6 from a tet-controlled promotor or induced reversion to yeast morphology due to tet-induced expression of NRG1,. Filamentous growth may directly contribute to host damage by invasion of epithelial or dendritic cells, and modulate the antifungal immune response,. In addition, among the several genes that have been identified to be expressed in a morphotype dependent manner, “filament-specific” genes and their products have frequently been characterized as virulence factors. The most prominent examples for this include the adhesin and iron-recruitment protein Als3, and the adhesin Hwp1. However, many other genes described as “filament-specific” have only been studied under selected conditions and the transcriptomic changes associated with filamentation of C. albicans independent of the external stimulus have not been fully elucidated.

2Two major signal cascades, the cAMP pathway and the MAP kinase cascade and their terminal transcription factors Efg1 and Cph1 control the formation of filaments,,. Depending on the external stimulus, other pathways like the pH response cascade may contribute to filamentation, but Efg1 often remains the terminal transcription factor of these pathways,. Besides activation, derepression is also crucial for the induction of filamentous growth. The major repressor complex consists of the regulator Tup1 and its DNA-binding partner Nrg1 and prevents expression of hyphae-associated genes,,. Detachment of this complex from target gene promoters is controlled by remodeling of chromatin structures and contributes to filamentous growth,. Overexpression or constitutive expression of NRG1 prevents filamentation, and deletion of both or either TUP1 or NRG1 results in a hyperfilamentous phenotype (,,. Within this study we used three different, well- defined stimuli to induce hyphal growth of C. albicans. A combination of transcriptome analyses and network modeling helped to define a core filamentation response for this fungal pathogen. Only eight genes were part of a group of genes with highly correlated expression pattern which was up- regulated in hyphae independent from the environmental condition. Analysis of expression patterns in C. albicans regulatory mutants confirmed that regulation of this core filamentation response is complex and depends on the environmental stimulus.

3Three well- defined stimuli were used to induce filamentation in stationary phase C. albicans yeast cells: (i) a shift from pH4 to pH8, (ii) the addition of 10% human AB serum to the medium and (iii) the change of the carbon source from 2% glucose to N- acetylglucosamine (for details see Material and Methods section). Human serum was used to be as close as possible to the natural environment in the human host. In time course experiments, we observed two stages of filamentation for all three shifts, germ tube formation during the first two hours after stimulation followed by hyphal elongation and branching (Figure 1). For transcriptional profiling, filament- inducing conditions were optimized to ensure a maximum comparability of germ tube formation kinetics.

4For each condition, 1×106 cells/ml from the overnight preculture were transferred into prewarmed medium (at 37°C) which either promoted yeast or hyphal growth. By this, we avoided a temporary shift from 37°C to lower temperatures and back which might delay hyphal development. Using a serum concentration of 10% resulted in kinetics which were most closely related to the pH and GlcNAc shift. For each time point, cells were checked for germ tube formation (Figure 1, 1 h and 2 h) or hyphae with visible septa and/or branches (Figure 1, 3–12 h). Control cultures were checked before RNA isolation for stable yeast morphology.

5Serum-induced hyphae reverted to budding filaments and even yeast cells again, whereas reversions were barely detected throughout the observation periods when using the other two protocols (Figure 1, 12 h).

6Total RNA was isolated for each of the standardized conditions at five distinct time points (1 h, 2 h, 3 h, 6 h and 12 h) in three independent biological replicates and used for transcriptome analysis. Raw data are available at ArrayExpress (http://www.ebi.ac.uk/arrayexpress/, accession number E-MEXP-3675).

7To confirm that the three experimental shifts induced specific transcriptional adaptation, overall transcription patterns were analyzed. A total of 704 genes changed their expression in response to pH shift from pH4 to pH8 at the 1 h time point. More than 50% (401 out of 704, Figure 2 A and Table S1) were specific for the pH shift, while 303 genes were also differentially expressed in either the serum or the GlcNAc or both (Figure 2 A). As expected we detected the down- regulation of PHR2, which is known to be expressed at acidic pH and the simultanous up- regulation of PHR1 (Figure 2 B). In addition, the alkaline induced gene PRA1 showed an increased expression at 6 h and 12 h (Figure 2 B). In response to the serum shift, 573 genes showed either an up- or a down- regulation at the 1 h time point. Less than half of them (238) were specific for serum- induced germ tubes (Figure 2 A and Table S1). CSA1, DDR48, SAP6 and HYR1 were highly up- regulated in serum- stimulated hyphae as reported previously,. Consistent with reversion to yeast morphology they were no longer expressed or even down- regulated after the hyphae- to yeast reversion at 12 h (Figure 2 C). In total, only 176 genes were differentially expressed 1 h after the carbon source change (Table S1) with 61 genes specific for the GlcNAc shift. Among those were genes associated with uptake and processing of GlcNAc, including NGT1 (encoding a GlcNac specific transporter), HXK1 (encoding a GlcNAc- kinase), NAG1 (encoding a Glucosamine-6-phosphate deaminase) and DAC1 (encoding a N-acetylglucosamine-6-phosphate deacetylase), which have previously shown to reflect adaptation to GlcNAc,, (Figure 2 D). In addition, genes GIG1, GAL1, GAL7 and GAL10 were also up- regulated during the GlcNAc shift as shown previously,, (Figure 2 D).

8Taken together, these analyses clearly show, that each of the three shift induces a highly specific transcriptional adaptation which is predicted to involve several different signaling cascades based on prior knowledge.

9For further analyses, we focused on genes which were differentially expressed for all three protocols. For this term, we only took the first 6 hours into account as at the 12 h time point reversion of serum- induced hyphae to yeast cells was observed (Figure 1). For all differentially regulated transcripts we calculated pairwise Pearson correlation coefficients and their significance over all 12 test conditions of the experiment using R software 2.14.1. For visualization, we selected all pairs of transcripts with correlation coefficients larger than 0.75 (all showing a p-value smaller than 0.0013). Visualization was done with Cytoscape 2.8.1 using the edge-weighted spring embedded layout. Based on gene expression correlation, we identified two different networks during the event of germ tube formation. These networks comprise all transcripts that are differentially regulated at 1 h and/or 2 h by all three stimuli. Only genes which were directly linked with each other are shown in the networks. The first network, consisting of down- regulated genes, contains genes that are related to either ribosome function, RNA metabolism or chromatin remodeling and histones (germ-tube formation network [GFN]; Figure 3 B). The central hubs of this network were the C. albicans BRN1 homolog, which encodes a putative condensin I and POL2, encoding DNA polymerase epsilon (Figure 3 B) Interestingly, this network was linked to the first hour of stimulation only, as most of these genes were no longer differentially expressed at later time points. However, some of them were up- regulated at the 12 h time point (Table S1). The second network was characterized by up- regulated genes which mainly have already been linked to filamentation such as ALS3, ECE1, RBT1 and IHD1 (early filamentation network [EFN], Figure 3 B). Germ tube formation was followed by hyphal elongation, characterized by longer and branching hyphae (Figure 1). In contrast to the earlier stage of hyphal growth, only a single network of up- regulated genes was identified at this stage, which was highly related to the EFN (late filamentation network [LFN], Figure 3 B). The LFN consisted of known hyphae- associated genes including ALS3, ECE1, HWP1 and IHD1 as well as of the regulatory gene UME6, which was shown to be crucial for hyphal elongation (Figure 3 B). PHR1, whose up- regulation is a typical transcriptional response to alkaline pH (Figure 2 B), was up- regulated in elongated hyphae in all three shifts and was therefore included within the network, sugesting a role which is independent from pH sensing (Figure 3 B). It should be noted that the networks are based on correlation in expression profiles. Therefore, some genes which were differentially expressed in all three stimuli were not showing up as they were not linked to the nodes of the aforementioned networks. The two most prominent examples are EED1, a regulator of hyphal elongation which was only up- regulated at 6 h (Table S1) and the repressor gene NRG1, the only gene which was down- regulated during hyphal elongation phase in all three shifts (Table S1, 3 h and 6 h). Interestingly, this down- regulation disappeared in serum- stimulated cells during the process of hyphae- to yeast- reversion while it was still down- regulated in pH- and GlcNAc- stimulated hyphae which did not undergo reversion (Table S1, 12 h). A combination of the EFN and the LFN resulted in a minimal set of eight genes which were part of the early filamentation network as well as of the late filamentation network: ALS3, DCK1, ECE1, HGT2, IHD1, HWP1, RBT1, orf19.2457. All of them could be linked to each other into a network, which was defined as the core filamentation response network (CFR) of C. albicans (Figure 3 C). This group of genes was up- regulated in germ tubes and hyphae, independently from the growth phase as well as the stimulating environmental condition.

10The genes IHD1 and orf19.2457 were central hubs of the core filamentation response network (Figure 3 C). Both not yet characterized ORFs were deleted in a PCR- based gene targeting approach and the resulting homozygous mutants were tested for their ability to develop hyphae or not. Neither the deletion of IHD1 nor orf19.2457 did affect the ability to form hyphae under the tested conditions (Figure 4), indicating none of them is actually required for the yeast to hyphae transition. These findings fit to the fact that previously described mutants lacking CFR genes ALS3, ECE1 and RBT1 were still able to form filaments. In contrast, mutants lacking HWP1 and DCK1, displayed defects in hyphal development. No information is available for the effects of HGT2 deletion on filamentation.

11To validate our findings, we analyzed the expression of the identified CFR genes in published transcriptome analyzes dealing with filamentation of C. albicans in similar or more complex experimental settings such as the interaction of fungal and human host cells. For that purpose studies using whole genome C. albicans expression arrays were identified from PubMed focusing on two groups of datasets: (i) whole- genome approaches dealing with hyphal induction with single, well- defined stimuli such as serum or cAMP, (ii) transcriptome analyses dealing with more complex conditions during host- pathogen interaction with a clearly defined yeast morphology as starting point and filamentation occurring within the analyzed time-frame. In addition, two recent studies using RNAseq or tiling arrays were included into the analysis.

12Wächtler et al. analyzed the transcriptional dynamics of C. albicans during invasion of human oral epithelial cells using the same array as in our analyses. In this study, very early time points of the invasion process (20 and 60 minutes) as well as later time point (180 minutes) were examined. The CFR as defined by our analyses was up- regulated for all time points (Figure 5 A). With only one exception (ADE4), genes of the GFN showed a decreased expression during the early time points 20 and 60 minutes, but were no longer differentially expressed at the later 180 time point (Figure 5 A). In addition, all genes of the EFN and LFN except ZDS1 showed strong up-regulation at 60 min post infection (Figure 5 A).

13Within the core filamentation response genes, ALS3, ECE1, HWP1, IHD1 and RBT1 were mostly found to be up- regulated during the switch from yeast to hyphae in experiments including serum shift, pH shift, cAMP shift, induction via plastic surfaces, biofilms or fungal interaction with oral epithelial cells, blood or caecum cells (Figure 5 B). Fitting best to our findings were the results from a RNA- seq based transcriptional profile of serum- induced hyphae were all aforementioned CFR genes were up- regulated (Figure 4 B). As ALS3, ECE1, HWP1 and RBT1 were long time predicted hyphae- specific genes, it was of interest to find out the dynamics of the four additional CFR genes. IHD1 was up- regulated in all analyzed transcriptome studies including an in vivo study with clinical samples from patients with oral candidosis (Figure 5 B). DCK1 was up- regulated in studies analyzing filaments induced by serum,, cAMP, contact with plastic surfaces or human oral epithelial cells caecum cells and human blood,, although other studies failed to detect a differential expression of this gene (Figure 5 B). DCK1 was clearly up- regulated in our study of alkaline- induced hyphae, but not in a previous one (Figure 5 B).In contrast to the other CFR genes, HGT2 was only found to up- regulated in one serum study and caecum cells, but interestingly also in patients suffering from oral candidosis. In another study, however, HGT2 was down- regulated in serum- induced hyphae (Figure 5 B).

14Except for DCK1, all of the CFR genes are characterized by large 5′ intergenic regions before the open reading frames with sizes between 2 kb and 6 kb (Table 1). On average, 5′ intergenic regions including promoters in C. albicans are 500 to 1000 bp. Argimon et al. have previously suggested that 5′ intergenic regions of hyphae- specific genes are unusually long compared to C. albicans genes in general. All of the CFR promoters contained putative binding motifs for transcription factor Efg1 and with the only exception orf19.2457 they also contained Nrg1 response elements (Table 1). Based on these in silico observations, the expression of the CFR genes was analyzed in non- filamentous mutants lacking EFG1 (efg1Δ and cph1Δ/efg1Δ) and the hyperfilamentous nrg1Δ mutant. Additionally, we tested the rim101Δ mutant, which formed wild type- like filaments during the serum and GlcNAc shift, but not after the change from pH4 to pH8. An increased expression of the CFR genes was observed in the wild type after 3 h hyphal growth in all three shifts (Figure 6). Interestingly, ECE1, HWP1, HGT2 and IHD1 were strongly expressed in efg1Δ and partially in cph1Δ/efg1Δ but not in rim101Δ during the pH shift, suggesting that Rim101 is more important for the regulation of these genes under pH inducing conditions than Efg1 (Figure 6). For the two other protocols, there was no induction of CFR genes in the cph1Δ/efg1Δ double mutant. The results suggest an impact of Cph1 on the regulation of both ECE1 and HWP1, as there was still an increase of expression of these genes in efg1Δ, but not in cph1Δ/efg1Δ (Figure 6). In the filamentous nrg1Δ mutant, the basic level of CFR gene expression was higher than in wild type and therefore the increase of expression was often not significant. However, some genes were still strongly induced in this mutant in response to filament inducing conditions, such as ECE1 during serum shift, indicating that a combination of derepression and activation contributes to the high levels of expression (Figure 6).

15Many different stimuli can induce filamentous growth of C. albicans and a lot of protocolls were used to trigger the morphological change from unicellular yeasts into germ tubes and later hyphae. For this study, we have chosen three well- defined stimuli which require change of only one condition compared to yeast-promoting media. In addition, all three shifts could be performed at a temperature of 37°C which reflects temperature in the natural human habitat of C. albicans. Dilution of stationary phase culture has previously been described to result in the formation of germ-tubes due to the release of quorum-sensing induced inhibition of filamentation. However, in our experiments, control conditions favoring yeast growth were diluted in an identical way as experimental conditions and no induction of filamentation was observed. In contrast to the aforementioned study we have not used YPD with a neutral pH as preculture and medium for yeast- like growth which might explain the absence of a quorum- sensing effect. In concordance with a trigger-specific induction of filamentation. C. albicans responded differently to each of the three shifts leading to gene expression pattern which were characteristic for each shift (Figure 2). Of the three shifts we used, human serum was the one where hyphae to yeast reversion occurred at the earliest time point (between 6 and 12 h), whereas this process was barely detected for the other two shifts in the 12 h time course (Figure 1).

16A combination of classical transcriptome analysis tools with network modeling led to the identification of a surprisingly small core filamentation response (CFR) in C. albicans which is differentially regulated during the formation of hyphae independent of the external stimulus. The algorithms employed for network modeling have originally been developed to analyze and visualize large datasets for protein-protein or protein-DNA interaction together with mRNA quantification In the correlation- based networks used for analysis of large-scale transcriptome data in this study, the algorithm allows a clustering of co-regulated genes and – in the setting of different input conditions leading to an identical response (in this case filamentation) – the definition of core gene sets that are directly related to the response and independent of the input condition. Using this approach, only eight genes fulfilled the criteria for a core filamentation response: ALS3, DCK1, ECE1, HGT2, HWP1, IHD1, RBT1 and orf19.2457. These genes constitute a core filamentation network of co-regulated genes. ALS3, ECE1 and HWP1 have long been considered prototypic hyphal associated genes and their identification confirms the validity of our approach. All of those encode for cell-wall/cell-membrane proteins, which is also true for DCK1, HGT2, IHD1 and RBT1 according to published data or sequence based predictions. Only the protein encoded by orf19.2457 may have no surface localization based on in silico analyses and nothing is known about its function.

17Hardly any information is available for HGT2, a gene homologous to the high-affinity glucose transporter HGT1. However, the gene is located next to a cluster containing also the galactose metabolism genes GAL10, GAL7 and GAL1, although it is separated from the other three genes by a very large intergenic region. These genes share similar or equal motifs in their promotor regions suggesting a regulatory function of the transcription factor Cph1. However, the expression dynamics were quite different. While the GAL genes were very specific for the GlcNAc shift (Figure 2 D, Table S1), HGT2 was up- regulated in all three shifts for the majority of time points (Table S1). Therefore, its expression seems to be separately regulated from the GAL genes and is at least under pH8 inducing conditions independent from Cph1 (Figure 6).

18The central hub of the core filamentation response network is the gene IHD1. The function of the encoded protein is unknown so far, although it might be GPI- anchored and therefore should localize on the fungal surface. Previously, IHD1 has been described as a hyphal induced for serum and temperature triggered filamentation. Rbt1 is an adhesin related to Hwp1 and has been involved in mating and biofilm formation. Dck1, a guanine nucleotide exchange factor for the Rac1 GTPase, is dispensable for serum-induced filamentation but required for matrix-embedded induced filamentous growth, a condition that was not tested in our study. Our data indicate that Dck1-Rac1 may have a stimulus independent function in filamentation. As RAC1 itself was not differentially expressed, this might suggest a Rac1- independent function of Dck1. Based on our data, DCK1 and HWP1 are the only genes in the core filamention response that are necessary for filamentation at least in certain conditions, while all other CFR network genes are effectors of a filamentation rather than being required for this morphological process. This is also reflected by the fact that well known regulators of filamentation including EFG1, HGC1, UME6, CPH1 and many others are not part of the CFR.

19According to several transcriptome studies from the past decade and to the list of hyphae- induced genes which was published by Sudbery, genes like DDR48, HYR1, SAP4-6 and SOD5 might have been expected to be part of the core filamentation response. However, none of these genes was included due to the fact that they were up- regulated in one or two filament- inducing shifts but not simultaneously in all three shifts. All of them were up- regulated during the serum shift and partially the pH (DDR48, SOD5) or the GlcNAc shift (HYR1 and SAP6). DDR48, HYR1 and SOD5 were however not differentially expressed or even down- regulated after serum- induced hyphae reverted to yeasts at the 12 h time point (Table S1). Taken together, these four genes are associated to the filamentation response in a broader meaning, but they are not as stimulus-independent as the aforementioned core filamentation response genes.

20Validation of the CFR network with data from other experimental settings confirms that it defines a minimum number of genes which can be used to assess morphogenesis in most if not all experimental settings. This is of potential interest for several applications: Using transcriptional quantification of these genes, newly generated mutants defective in morphogenesis can easily be tested for an expression pattern that resembles yeast or hyphal morphology, aiding in the identification of mutants displaying a dissociation of phenotype from transcriptional pattern as described for the C. albicans Δhgc1 mutant. Similarily, quantification of genes regulated during early morphogenesis could assist in further elucidating the role of filamentation in interaction of C. albicans with immune cells by providing a tool for early detection of a switch towards filamentation in the transcriptional program.

21Despite the low number of CFR genes, the regulation of these genes is apparently complex. Quantification of CFR expression in several regulator mutants suggested that CFR expression is governed by overlapping and at least partially redundant molecular mechanisms. In addition, whereas regulation of the CFR genes occurs independent of the external stimulus for filamentation, the regulatory mechanisms leading to induction differ between the stimuli, suggesting that stimulus-dependent contributions of several regulators rather than a universal master regulator govern CFR expression. A prominent example for this is the fact, that Rim101 is absolutely required for induction of all CFR genes in pH-shift induced filamentation, whereas it is dispensable in the other conditions.

22Furthermore, the identification of the GFN which is specifically downregulated during the very early phase of hyphal development is consistent with the involvement of chromatin modification in the regulatory processes. Modification of histone proteins has shown to be involved in regulation of morphogenesis and it was recently shown that chromatin remodeling plays an important role in allowing Nrg1 to get access to its target DNA sequences,.

23In summary, we could show that correlation-based modeling of transcriptional analyses has proven useful to identify a core set of marker-genes characteristic for the important biological process of filamentation in C. albicans. Therefore, correlation- based network modeling might be suitable tool for further analyses of other processes in C. albicans.

24All C. albicans strains used in this study are listed in Table 2. Strains were routinely grown at 37°C in either M199 medium (9.8 g/l M199 powder, PAA; 35.7 g/l HEPES, 2.2 g/l sodium carbonate; adjusted to different pH values with either sodium hydroxide or hydrochloride acid) with pH4 or SDG medium (6.7 g/l YNB without amino acids, DIFCO, 20 g/l glucose). For hyphal induction, 1×106 cells/ml were transferred from a stationary phase culture in M199 pH4 to M199pH8 (“pH shift”) or from a stationary phase culture in SDG to SDG with 10% human serum (PAA, “serum shift”) or SDN (6.7 g/l YNB without amino acids, DIFCO, 20 g/l N- acetlyglucosamine, “GlcNAc shift”). Cells were then incubated at 37°C for a total of 12 h.

25The genes GIT2, IHD1, orf19.1344 and orf19.2457 were deleted in the background strain SN87 which is auxotrophic for histidine and leucine. Gene deletions were performed with a PCR- based strategy using the plasmids pFA-CdHIS1 and pFA-CmLEU2. Transformations of C. albicans strains were performed with the established lithium- acetate method.

26Routinely, cells were analyzed by DIC microscopy, which was performed on a Zeiss AxioObserver Z.1 (Carl Zeiss, Göttingen and Jena, Germany).

27Total fungal RNA was isolated by a hot phenol- chloroform method previously described in more detail. Quality and quantity of the RNA were checked by Agilent Bioanalyzer 2100 (Agilent Technologies). For whole genome expression studies, sample RNA was labeled with Cy5-CTP and hybridized with a Cy3- labeled common reference RNA on C. albicans DNA microarrays (ClinEuroDiag, Brussels, Belgium). Slides were hybridized, washed and scanned with a Genepix 4000B (Molecular Devices) as described previously. After quality control, arrays were preprocessed using R software version 2.14.1 (http://www.r-project.org). Printtiploess and Gquantile methods were used for normalization. A linear model was fit to the normalized data. Transcripts were regarded as being significantly differentially expressed when they showed an absolute fold change of larger than 1.5 and an FDR adjusted t-test p-value of less than 0.05. Raw data of the microarrays are available at ArrayExpress (http://www.ebi.ac.uk/arrayexpress/) with the accession number E-MEXP-3675 for wild type arrays.

28For all significant differentially regulated transcripts we calculated pairwise Pearson correlation coefficients and their significance over all 15 test conditions of the experiment using R software 2.14.1. For visualization, we selected all pairs of transcripts with correlation coefficients larger than 0.75 (all showing a p-value smaller than 0.0013). Visualization was done with Cytoscape 2.8.1 using the edge-weighted spring embedded layout.

29Quantitative RT PCR was performed using the Brilliant III Ultra Fast SYBR Green qRT PCR Kit (Agilent Technologies) on a Stratagene Mx3005P (Agilent Technologies) with 100 ng/µl RNA as template. To determine the fold changes of gene expression we used the ΔΔCt method. Gene expression was normalized against ACT1 as housekeeping gene and the aforementioned common reference RNA as control.