In order to conduct a comparative analysis of the transcriptomes of CDR versus drug susceptible C. albicans cells, we performed high-throughput sequencing of cDNA made from poly(A) RNAs obtained from the Gu4 and Gu5 strains. Gu4 is a fluconazole susceptible clinical isolate obtained from an early infection episode. Gu5 is the corresponding fluconazole resistant clinical isolate obtained from later episode in the same patient treated with fluconazole. Gu4 and Gu5 are therefore supposed to be isogenic and mainly differ by their resistance to fluconazole. Gu5 has been characterised as a CDR strain, showing overexpression of CDR1 and CDR2, as compared with Gu4 28. For each strain, two biologically independent samples of RNAs were analysed by RNA seq. We obtained 27 763 991 unique mappable reads for Gu4 (96% of total reads) and 20 570 932 mappable reads for Gu5 (97% of total reads). As noticed in previous RNA seq studies, the background was very low, with no or few reads mapped in intergenic regions. The correlation coefficient between the biological replicates was very high (0.991 for Gu5 and 0.999 for Gu4). We could detect expression for 5727 (92%) of the 6177 ORFs annotated in the CGD. The 450 ORFs which were undetected in our experiments are mostly dubious ORFs or gene without clear function (71%). We arbitrarily grouped the expressed ORFs in five categories: very high expression (more than 1000 reads/nucleotides), high expression (between 100 and 1000 reads/nucleotides), medium expression (between 10 and 100 reads/nucleotides), low expression (below 10 reads/nucleotides) and very low expression (below 3 reads/nucleotides). These categories contained 3%, 20%, 57%, 12% and 8% of the genes, respectively. As described previously, the “very high expression” category was enriched in genes encoding ribosomal proteins, glycolytic enzymes, translation regulators, histones and components of respiratory complexes. The “high expression” category was enriched in genes encoding enzymes involved in various metabolisms. The “medium” and “low” expression categories did not show any significant functional enrichment.

We used our data to define the boundaries of each transcript, as described previously 25. We compared the 5’ and 3’UTR predictions that we obtained with the annotation set up by a previous RNA seq analysis of the Candida albicans reference strain SC5314 25. We found no significant difference in the two annotations for more than 75% of the genes, indicating that our data is reliable and that there is no global change in 5’ and 3’UTR length between those strains. Still, we had some significant discrepancies with the annotation from Bruno et al. for about thirty genes, which showed much longer 5’ or 3’ UTR according to our data ( Additional file 1: Table S1). This is for instance the case of TAC1ADAEC and CZF1 (Figure 1). These genes did not show any particular functional enrichment. Noteworthy, we did not detect significant differences in the size of transcripts between the Gu4 and Gu5 strains, with the notable and sole exception of TAC1 which presented an extension in 5’ which was apparently specific for the azole resistant strain (Figure 1).

Finally, we used our data to detect new transcripts. To do so, the reads from the Gu4 and Gu5 strains were summed and used to identify regions of the genome with a minimum coverage of at least 3 reads/nucleotides (with at least one read in three independent samples) for a stretch of at least 50 nucleotides. With these criteria, we identified 1130 new transcription units, 800 of which partially or totally overlapped new transcribed regions recently identified by RNA seq 2526 or tiling arrays 27 ( Additional file 1: Table S2). We named these new transcripts NGuTs for “New Gu strains Transcripts”. The median size of NGuTs was 425 nucleotides. The majority of these NGuTs did not contain any obvious ORF and are therefore likely to be non-coding RNAs.

The identification of genes differentially expressed in Gu4 and Gu5 was performed combining two different statistical methods (see methods section) and using the gene annotations from the CGD, the new transcribed regions defined by previous studies 252627 and the new transcribed regions defined in this study (see above). We found 130 genes significantly overexpressed in Gu5 and 92 genes which were underexpressed in this strain compared to Gu4 (a full list and description of these genes can be found in Additional file 1: Table S3).

We first compared our results with the one obtained in a previous study, which used microarrays to analyse the transcriptomes of 4 different CDR isolates 18. Among the 130 genes induced in Gu5, 74 (56.5%) were previously shown to be overexpressed in one or several of these CDR clinical isolates ( Additional file 1: Table S3). Among these 74 genes, 53 (72%) were found overexpressed in at least 3 of the 4 CDR strains, suggesting that most of the genes identified in our RNA seq analyses are general features of the CDR isolates. These genes include the CDR1 and CDR2 ABC transporter encoding genes, together with the TAC1 gene encoding their transcriptional regulator. Remarkably, all the previously identified targets of Tac1p are found in our list, including PDR16RTA3 and LCB4.

Of the 92 genes underexpressed in Gu5, 42 (46%) were previously shown to be repressed in CDR isolates. This includes genes involved in iron and metal homeostasis (FTR1, SIT1, FET34 for instance) and amino acid metabolism (MET3, OPT7, GAP4, SAM4 for instance).

We identified 53 ORFs which were overexpressed in Gu5 and were not previously associated with CDR phenotype. About half of them (45%) are uncharacterized and little information is available on their function.

Among the characterized genes (Figure 2 and Additional file 1: Table S3), some of them were known as playing important roles in morphogenesis, hyphal growth and white/opaque switch. This is the case of the transcription factors Czf1 and Sfl2, the kinase Sch9, the membrane sensor Gpr1 and its target Ece1, the signal transducer Srr1, the white specific gene of unknown function ADAEC and the cytoplasmic protein Wh11. Some other induced genes are involved in cell wall maintenance and cell adhesion, like the mannosyl-transferase Mnn4 or the adhesion protein Als1. Finally, two transporters encoding genes belong to this list: Yor1, which is homologous to an oligomycin resistance gene of S. cerevisiae and Hgt10, which encodes a glycerol permease involved in osmotic stress response. Remarkably, four of the CDR induced genes (CZF1, TAC1, ADAEC and orf19.6713) present the particularity to have a long transcribed region upstream of the start codon according to our RNA seq data, which could be interpreted either as long 5’ UTR or as new non coding transcripts (see Figure 1 and chapter below).

Fifty genes were shown for the first time to be downregulated in Gu5. Of these, 24 (49%) are uncharacterized. Among the verified ORFs (Figure 2 and Additional file 1: Table S3), we could find the gene encoding the ABC transporter Cdr4, which is homologous to Cdr1 and Cdr2, and RTA4, which is homologous to the Tac1 target, RTA3. We also found in this group additional genes involved in copper and iron homeostasis (HEM13CRP1 and SMF3) and genes encoding extracellular proteins (PLB1ALS2MAL2). Two genes encoding potential steroid targets, EBP1 and CBP1, were also underexpressed in Gu5, which is interesting considering the previously established link between steroid exposure and CDR response 29.

To support these results, we performed semi-quantitative RT-PCR analyses for 6 repressed and 8 induced genes, using actin as a reference and CDR1 as a positive control (Figure 3). These experiments confirmed the RNA seq results, except for EBP1 and orf19.3769, for which no difference of expression was detected by this method.

Among the new transcripts identified in this study, 5 were overexpressed and 8 underexpressed in Gu5 compared with Gu4. Four of the five induced transcripts were located immediately in 5’ of an induced ORF. Transcript NGuT-chr4-123 (previously identified as NOVEL-Ca21chr4-073 in 25) was in 5’ of CZF1 (Figure 1), NGuT-chr5-037 was in 5’ of TAC1 (Figure 1), NGUT-chr2-038 was in 5’ of ADAEC (Figure 1) and NGuT-chr3-128 (previously identified as TF3W223 in 27), was close to orf19.6713. The fifth transcript, NGuT-chr3-105 (NOVEL-Ca21chr3-056), is located just upstream of the AAF1 gene, which encodes an adhesin and which is not induced in the Gu5 strain.

All the new transcripts which were repressed in Gu5 are located at more than 1 kb from the adjacent ORFs. NGUT-chrR-157 (NOVEL-Ca21chrR-103) and 160 (TFRW269) belong to the same cluster of non coding RNAs. NGUT-chr4-069 (NOVEL-Ca21chr4-040) is located in a zone rich in tRNA genes. NGUT-chrR-041 (NOVEL-Ca21chrR-026) is close to a long terminal repeat.

Semi-quantitative RT-PCR analyses for 4 repressed and 3 induced NGuTs, using actin as a reference confirmed the RNA seq results, except for NGuT-chr3-105 and NGUT-chrR-157 (Figure 3).

RNA seq and tiling arrays experiments have allowed the systematic annotation of transcript boundaries in three yeast species: S. cerevisiaeC. albicans and C. parapsilosis21252735. In all three species, the 5’ and 3’UTR have been shown to be generally short, with an average size ranging between 50 and 80 nucleotides. However, in all these species, many exceptions have been described. In C. albicans, about 100 genes have 5’ UTR longer than 500 bp 27. Interestingly, this group is enriched in genes encoding transcriptional regulators and genes involved in filamentation. These include the CZF1 transcription factor encoding gene, which was induced in our experiments and which has a 5’UTR of about 2 kb (2736; this study). This observation is also true in mammals in which many genes with long 5’UTR encode regulators and proto-oncogenes 37. Long 5’UTR usually have important roles in the post-transcriptional and translational regulation of gene expression. In the yeast Saccharomyces cerevisiae, the expression of several transcription factors, including Gcn4 (amino acid starvation), Yap1 (oxidative stress response) and Yap2 (cadmium response), is regulated by the presence of short upstream ORFs (uORFs) in their long 5’UTR region, which exert negative or positive actions on the translation of the main ORF or on mRNA stability, depending on the cellular context 38. Recent studies have identified about 250 genes containing potential uORFs in this species 21.

As mentioned in the introduction, TAC1 encodes the major transcriptional regulator of CDR genes. Its own mRNA expression is co-induced with those of its main targets, CDR1 and CDR2, possibly due to an autoregulatory loop 1829. In previous transcript annotations, TAC1 mRNA was described as having a 5’UTR of about 300 nt 2527. Our RNA seq data detected a transcribed region up to 1 kb upstream of the TAC1 start codon, which was apparently specific of the Gu5 strain (Figure 1). This could be interpreted as a new transcript expressed only in the CDR strain, as an alternative transcription start site (TSS) specific for Gu5 (alternative TSS had been described in C. albicans, for instance in the case of the EFG1 transcription factor encoding gene 39) or as the normal TSS of TAC1 which could not be detected in drug sensitive strains because expression of TAC1 in those strains and RNA-seq coverage in previous experiments and in this study were too low. 5’RACE and northern blot experiments could resolve if it is an alternative TSS or a new transcript. However, the strong correlation between the expression of the two regions (Figures 1 and 3) suggests that TAC1 mRNA presents, at least in the Gu5 strain, an unusually long 5’UTR, which was previously not detected. Considering the important roles of long 5’UTR in translation regulation, this opens exciting new perspectives to the post-transcriptional regulation of TAC1 expression, which was poorly investigated until now. Especially, we observed in this 5’UTR the presence of one large (about 50 codons) uORF in the same orientation and phase than the TAC1 ORF (Figure 7), which suggests the existence of regulatory mechanisms similar to the ones described for GCN4YAP1 or YAP2 in S. cerevisiae. The analysis of ribosome occupancy at TAC1 mRNA in CDR matched pairs, either measured by global ribosome profiling or by polysomes gradients, will give information on this potential regulation. Moreover, a long 5’UTR would have important implications on the way we consider the transcriptional regulation of TAC1. Indeed, yeast enhancers are usually located close upstream to the TSS. Since 5’UTR regions are also limited in size, people tended to look for DNA consensus motifs within 800 bp upstream of the start codon of the genes. The fact that the actual TSS of TAC1 would be located about 0.8 kb upstream of all previous estimations also change the potential position of its promoter region. For instance, it was shown that Tac1 binds its own promoter, which led to the hypothesis that it is regulated through a positive auto-regulatory loop 18. However, this hypothesis was challenged by the fact that the two potential Tac1 binding motifs (DRE) in this region were located more than 1 kb upstream of TAC1 ATG. In the annotation proposed in Figure 7, these DRE would be located very close to the TSS, which strongly supports their functional role in the activation of TAC1 expression.

Czf1 is a zinc finger transcription factor mostly known for its role in the epigenetic white/opaque switch, which controls cell and colony morphogenesis and properties 40. Czf1 acts synergistically to Wor1 and Wor2 transcription factors, and antagonistically to Efg1, to promote the opaque state. Czf1 also has a role in hyphal growth, (also called filamentation). Czf1 mutants filament normally under standard filamentation induction conditions, but they are defective in hyphal development in anaerobic conditions 41. This phenotype is dependent on the composition of the media, for instance in terms of carbon source. The positive role of Czf1 on hypoxic filamentation is under the negative control of the TOR pathway and more specifically, of the Sch9 kinase 42. As for the white/opaque switch, Efg1 plays an important and complex role in filamentation. In aerobic conditions, it is a positive regulator of hyphal development, whereas in hypoxia, it counteracts the positive action of Czf1 on this process 43. Direct cross-talks exist between Efg1 and Czf1: Efg1 binds to the promoter of CZF1 and repress its expression, while Czf1 has a similar effect on the EFG1 promoter 3644. More recently, Czf1 has been shown to be required for wild type adherence to silicone devices 45. Despite these roles in morphogenesis, the list of Czf1 targets has not been established to date. A partial transcriptome analysis of the czf1 mutant, conducted on 250 genes using the Nanochip technology, detected very few expression changes in this mutant 45.

Our data strongly suggest that Czf1 plays a role in the acquisition of multidrug resistance of the CDR type. First, CZF1 is overexpressed in all CDR mutants that we tested. Noteworthy, some other actors of white/opaque switch and hyphal growth were also overexpressed in the Gu5 strain, including SCH9, ADAEC, WH11 or SFL2. Second, the czf1 null mutant is sensitive to many unrelated drugs, including azoles and molecules targeting the calcineurin pathway. Remarkably, CZF1 does not seem to be involved in drug resistance phenotypes of the MDR kind, since its expression was unchanged in MDR mutant strains. The drug sensitivity associated with the inactivation of Czf1 does not seem to be due to standard drug resistance pathways, since the level of expression of the main actors of the CDR, MDR and calcineurin pathways were unchanged in this mutant. However, the fact that Czf1 does not control the steady-state expression levels of those genes does not mean that it is not involved in their induction following drug exposure. This hypothesis will be tested by inactivating CZF1 in the Gu5 and Gu4 backgrounds and testing the effect of this inactivation on drug resistance and gene expression.

Surprisingly, we observed that the czf1 mutant was resistant to congo red, a cell wall perturbation agent which targets beta-glucans, the most abundant component of Candida albicans cell wall. This phenotype was haploinsufficient, since the heterozygous mutant exhibited an intermediate resistance between the null mutant and the wild type. Congo red is thought to target the glucan architecture by directly interacting with glucan chains, therefore inhibiting synthesis of glucan polymers 46. We observed that the GSL1 gene encoding one of the three beta 1,3 glucan synthase subunits, was strongly overexpressed in the czf1 mutant. This suggests that Czf1 has a negative effect on the expression of some of the actors of the glucan synthesis pathway, which could explain the congo red resistance observed in the null mutant. Many functional connections have been established between cell wall components and hyphal growth. GSL1 itself has been shown to be down-regulated during filamentation 47, a process that is positively regulated by Czf1. Notably, Efg1 is also involved in cell wall architecture and efg1 mutants show a haploinsufficiency phenotype 4748. The deletion of one or two copies of EFG1 leads to a decrease in the expression of many enzymes involved in cell wall biogenesis, including beta1-3 glucan synthases. This causes important changes in cell wall thickness and composition 48. Again, Czf1 and Efg1 seem to antagonize each other in the regulation of cell wall maintenance. The over-expression of some cell wall protein encoding genes has been extensively described as a compensatory mechanism in response to impairments in cell wall biosynthesis or to cell wall damages, both in S. cerevisiae and C. albicans (see for instance 495051). This phenomenon is largely controlled by the cell wall integrity (CWI) pathway 52 but part of it is independent of CWI 53. For instance, treatment with caspofungin (which directly inhibits the activity of the essential beta-glucanase subunit encoded by GSC1) triggers a compensatory process which leads to the overexpression of GSL1, similarly to what was observed in the czf1 null mutant 51. We suggest that Czf1 could play a role in this process.

All the media components were obtained from HiMedia (Mumbai, India). Amphotericin B (AMB), terbinafin, anisomycin, ketoconazole, sodium dodecyl sulphate, FK520, congo red were purchased from Sigma chemicals Co. (St. Louis, MO). Ranbaxy India generously provided FLC. Reverse transcriptase kit was purchased from fermentas, India. Caspofungin was generously provided by Dr. Neeraj Chauhan (Public Health Research Institute and Department of Microbiology and Molecular Genetics, New Jersey, USA). dATP, dGTP, dTTP and dCTP were bought from Sigma chemicals Co. (St. Louis, MO).

Gu4 and Gu5 strains were described elsewhere 28. The strains of Candida albicans mutated in various genes (ALS1MNN4CZF1) that were used for phenotypic tests, were obtained from Dr. Ibrahim, A.S. (David Geffen School of Medicine at UCLA Los Angeles Biomedical Research Institute), Dr. Gow, N. (The university of Aberdeen, Scotland) and Dr. Kumamoto, C.A. (Tufts University, Boston, USA) respectively. The complete list of strains can be found in Additional file 1: Table S4. Cells were grown in YPD rich medium (2% glucose, 1% bactopeptone, 1% yeast extract) at 30°C for 48 hrs.

Twenty ml of exponentially growing (OD = 0.6-0.7) Gu4 or Gu5 cells were flash frozen in 30 ml of cold ethanol, centrifuged for 5 minutes at 3000 g, washed with sterile water and centrifuged again. The cell pellets were stored at −80°C. RNA were extracted using the RNeasy kit (Qiagen), with DNAse treatment. The RNA quality was checked on a bioanalyzer Nanochip (Agilent).

RT-PCR was done using the RevertAidTM H Minus kit (MBI, Fermentas). Briefly, 1 μg isolated RNA was primed with oligo (dT)18 for cDNA synthesis at 42°C for 60 min. Reverse transcription reaction was terminated by heating at 70°C for 5 min. The synthesized cDNA product (2 μl) was directly used for PCR amplification reaction (50 μl) using gene specific forward and reverse primers. The amplified products were gel electrophoresed and quantitated by using Quantity One software by Bio Rad gel documentation system. The RNA samples used for these validation experiments were independent from the samples used for RNA seq. The sequences of the primers used for PCR reactions can be found in Additional file 1: Table S5.

Drug-susceptibility was tested by broth-microdilution assay according to CLSI (Clinical and Laboratory Standards Institute) and serial dilution assay essentially as described previously (CLSI, 2008). Following drugs stocks (in parenthesis) and working concentration were used 2.5 μg/ml of FLC (5 mg/ml), 4 μg/ml of TER (4 mg/ml), 4 μg/ml of AMB (4 mg/ml), 150 μg/ml of ANS (20 mg/ml), 0.5 μg/ml of KTC (1 mg/ml), 50 ng/ml of CAS (1 mg/ml), 0.08% SDS, 200 μg/ml of CR (10 mg/ml), and 200 μg/ml FK520 (20 mg/ml).