Transcriptome Analyses of Candida albicans Biofilms, Exposed to Arachidonic Acid and Fluconazole, Indicates Potential Drug Targets

Candida albicans is an opportunistic yeast pathogen within the human microbiota with significant medical importance because of its pathogenic potential. The yeast produces highly resistant biofilms, which are crucial for maintaining infections. Though antifungals are available, their effectiveness is dwindling due to resistance. Alternate options that comprise the combination of existing azoles and polyunsaturated fatty acids, such as arachidonic acid (AA), have been shown to increase azoles susceptibility of C. albicans biofilms; however, the mechanisms are still unknown. Therefore, transcriptome analysis was conducted on biofilms exposed to sub-inhibitory concentrations of AA alone, fluconazole alone, and AA combined with fluconazole to understand the possible mechanism involved with the phenomenon. Protein ANalysis THrough Evolutionary Relationships (PANTHER) analysis from the differentially expressed genes revealed that the combination of AA and fluconazole influences biological processes associated with essential processes including methionine synthesis and those involved in ATP generation, such as AMP biosynthesis, fumarate metabolism and fatty acid oxidation. These observations suggests that the interference of AA with these processes may be a possible mechanisms to induce increased antifungal susceptibility.

Therefore, this investigation aims to unravel the mechanism involved by conducting transcriptome analysis on C. albicans biofilms, grown in the presence of AA, fluconazole, or a combination of the two to identify differentially expressed genes under the different conditions.

Stock solutions
Stock solutions of fluconazole (Sigma-Aldrich, USA) and the fatty acid, AA (Sigma-Aldrich, USA) were dissolved in dimethyl sulfoxide (DMSO) (Sigma-Aldrich, USA) and ethanol (EtOH) (Merck, RSA), respectively. Working concentrations were added to the media, ensuring that the solvent concentration was below 1%. Fluconazole was added to a final concentration of 1 mg.L -1 , while a concentration of 1 mM was used for AA.

Biofilm formation and storage
The revived Candida albicans CBS 8758 (SC5314) was used to inoculate 10 mL Yeast Nitrogen Base (YNB) broth (6.7 g.L -1 YNB, 10 g.L -1 glucose) and incubated on a shaker at 30°for 24 h. The cells were harvested by centrifugation at 3080 g for 5 min and washed thrice with sterile phosphate buffer saline (PBS) (Sigma-Aldrich, USA). Cell count was determined with a hemocytometer, and the cells diluted to 1 · 10 6 cells.mL -1 in 20 mL sterile RPMI 1640 media (2 g.L -1 glucose) (Sigma-Aldrich, USA). Biofilm formation was initiated in a polystyrene Petri dish and incubated without agitation at 37°for 90 min, then washed with PBS to remove non-adherent cells (Mishra et al., 2014). The medium was replenished with 20 mL RPMI medium containing the desired concentration of fluconazole with AA, or an equivalent volume of the drug vehicles, which was below 1%. The investigated conditions include AA (dissolved in EtOH) and DMSO (DA), fluconazole (dissolved in DMSO), and EtOH (FE) as well as fluconazole (dissolved in DMSO) and AA (dissolved in EtOH) (FA). The control contained both solvents, DMSO, and EtOH (DE). A total of 5 biofilms were prepared in triplicate for each condition. The plates were covered with parafilm before incubation for 6 h at 37°t o allow for biofilm formation. The biofilms and the growth medium used were scraped off the surface of the plate using a cell scraper and centrifuged at 1971 g for 3 min at 4°. The supernatant was aspirated and replaced with 2 mL of Qiagen RNAlater (Qiagen) to prevent RNA degradation before storage at -80°.
Total RNA extraction and sequencing Stored biofilms samples were thawed on ice and centrifuged at 4000 g for 5 min. RNAlater was aspirated, and total RNA extraction was carried out using the RNeasy protect mini kit (Qiagen). At the same time, the DNA present was removed with RNase-Free DNase Set (Qiagen), according to the manufacturer's instructions. The quality of RNA in each sample was determined at the Centre for Proteomic and Genomic Research (CPGR) before sequencing. Quality tests carried out include checking for contaminants with the NanoDrop ND1000, absolute concentration using the Qubit RNA HS Assay Kit, and an Agilent Bioanalyzer Nano Assay to evaluate the sample integrity. Samples having contaminations were cleaned up using an Agencourt RNAClean XP kit (Beckman Coulter), and the quality confirmed. Ribosomal RNA was removed from 1 mg of RNA per sample, using the Illumina Ribo-zero rRNA removal kit, and purified with an Agencourt RNAClean XP kit. Indexed libraries were prepared using the ScriptSeq v2 RNA-Seq Library Preparation Kit and ScriptSeq Index PCR Primers-Set 1 (Illumina). The library size was profiled with the Bioanalyzer High Sensitivity Assay Kit (Agilent) and quantified with the Qubit HS DNA Assay Kit. The samples were diluted and spiked with a Phix control library (Illumina). Sequencing was completed on a Nextseq 500 (Illumina) using the Nextseq 500 High Output (150 cycles) kit, which yielded paired-end reads with 96% and 97% of total data having bases with a Phred score % $ Q30.
Sequence analysis for differentially regulated genes Fastq files, obtained from the sequencing, were analyzed for quality using FastQC v0.11.5; (Andrews 2010) and the reads with low quality (,Q30) were discarded, using PRINSEQ-lite v0.20.4 (Schmieder and Edwards 2011). The sequences were aligned to the C. albicans SC5314 genome assembly 21 (Skrzypek et al., 2017) using TopHat2 (Trapnell et al., 2012;Kim et al., 2013) with the fr-secondstrand, -r 250 matestd-I 10000 -G option (Dutton et al., 2016). Subsequently, the aligned files were merged with SAMtools (Li et al., 2009). Gene expression count tables were built from BAM files, created with TopHat2 using the BEDTools multicov command (Quinlan and Hall 2010). These counts were used to evaluate the differential expression of genes with DESeq2 (Love et al., 2014). Cuffdiff (Trapnell et al., 2012) was also used to generate differential expression data using aligned reads obtained from TopHat2 (Trapnell et al., 2012;Kim et al., 2013) which were assembled with Cufflinks and merged with Cuffmerge (Trapnell et al., 2012). Differentially expressed genes identified from DESeq2 and Cuffdiff with a P # 0.05 were considered significant. Heat maps, volcano plots, and principal component analysis were done as described by Love and co-workers (2014) with a modification of the script as described on github.com/stephenturner. The Protein ANalysis THrough Evolutionary Relationships (PANTHER) classification system was used for enrichment analysis of biological processes of the significant differentially regulated genes in the different data sets (Ashburner et al., 2000;Thomas 2003;Mi et al., 2017; The Gene Ontology Consortium 2019).
NanoString analysis RNA extracted from the different treatment conditions [DA= AA (dissolved in EtOH) and DMSO, FE = fluconazole (dissolved in DMSO) and EtOH, FA = fluconazole (dissolved in DMSO), AA (dissolved in EtOH), and DE = the solvents, DMSO and EtOH] were analyzed with the NanoString nCounter analysis system (Geiss et al., 2008) at the University of the Witwatersrand, Department of Internal Medicine, using a gene expression TagSet that targets 36 genes including three housekeeping genes (orf19.1191, SLF1, orf19.2184). Custom-designed TagSets, which comprise a fluorescent-labeled reporter tag and a biotinylated universal capture tag, were designed for each gene. Both the reporter tag and capture tag (referred together to as CodeSet) was hybridized with 100 ng of total RNA at 67°in a preheated thermocycler for 16 h. Subsequently, the hybridized samples were transferred into an nCounter SPRINT Cartridge and loaded into a nCounter SPRINT Profiler to quantify the transcripts. The nCounter raw expression data file (.RCC) obtained was uploaded into the nSolver Analysis Software 4.0 for review of quality control metrics. The data were normalized by subtracting the geometric mean of the negative control of each sample from the raw counts. The data were grouped between the experiments and control, and their expression ratio was determined. The Log 2 fold change obtained with nSolver was compared to the values obtained with RNA-seq analysis. Figure S1 are graphs showing the inhibitory effect of fluconazole, arachidonic acid and the combination of fluconazole and arachidonic acid. Figure S2 is a principle component analyses of the data and Figure S3 is a volcano plot showing the statistical significance vs. the magnitude of change between the control sample and the different test conditions. Table S1-S9 comprises the PANTHER classification gene ontology terms for the different test conditions. Table S10 shows a comparison between the transcripts obtained using RNA-Seq and NanoString. All raw sequences and analyzed data for this investigation have been deposited in the NCBI's GEO database under the accession number GSE137423. Supplemental material available at figshare: https://doi.org/10.25387/g3.12473012.

RESULTS AND DISCUSSION
Different conditions produced unique transcriptome profiles RNA sequencing has become an important technique to understand genomic functions (Hrdlickova et al., 2017). The most common application of RNA-seq is for the identification of differentially regulated genes under two or more conditions (McDermaid et al., 2018). In this study, RNA-seq was applied to identify differentially regulated genes in C. albicans biofilms grown in fluconazole, AA, or the combination of the two, compared to the drug vehicles as controls after early biofilm formation (6 h). The use of 1 mM AA was based on our observation that AA concentrations below 1 mM were noninhibitory to C. albicans biofilms ( Figure S1a) as well as the report by Ells and co-workers (2009), which reported an increase in susceptibility of C. albicans biofilms to 1.25 mg.L -1 clotrimazole increased by approximately 160% in the presence of 1 mM AA, in comparison with biofilms without AA. We made a similar observation with fluconazole in combination with 1 mM AA after measurement of biofilm metabolic activity using XTT assay ( Figure S1b). The most significant (P # 0.05) increase in susceptibility was observed when 1 mM AA was combined with 1 mg.L -1 of fluconazole assay ( Figure S1c). Mishra and co-workers (2014) demonstrated that a combination of 500 mM AA and half the fluconazole MIC did not cause a significant increase in fluconazole susceptibility, indicating that the presence of high AA concentration is vital to the observed increase in fluconazole susceptibility. Although the AA concentration used is much higher than the physiological concentration, which is between 0.1-50 mM in human plasma (Pompeia et al., 2003;Tallima and Ridi 2018), the aim of this study was to investigate the antifungal mechanism of AA, thus the higher, inhibitory concentration was used.
The independent Illumina sequencing runs of C. albicans biofilms biological replicates produced a total of 325 million raw reads, with 92.4% having a Phred score % $ Q30. Principal component analysis between the different replicates obtained showed a close correlation between the number of reads obtained for each run, as shown by the close clustering of the spots, which represents individual runs ( Figure S2). The variation between the different test conditions (FA, DA, and FE) and the control (DE) was minimal, as represented by the distance between the spots. The high reproducibility among the data sets was demonstrated with a heat map that showed close clustering between identical replicates, which can be seen with the corresponding color intensity. The relatively higher variability among the replicates of DE, may be due to higher morphological diversity in these biofilms (i.e., yeast, hyphae) compared to biofilms grown in the presence of AA (Clément et al., 2007) or fluconazole (Ha & White 1999). Volcano plots to show the number of statistically significant differentially regulated genes are shown in Figure S3. As shown, there were more differentially regulated genes in FA and FE than in DA.
The data obtained after RNA-Seq were initially analyzed with Cuffdiff to identify differentially expressed genes from different conditions. The biofilms which were grown in the presence of DE served as the control and was compared with the other three experimental conditions of DA, FA, and FE to identify differentially expressed genes. The open reading frames were identified with pipelines by aligning the reads to the C. albicans genome haploid assembly 21 from the Candida genome database (CGD) (Skrzypek et al., 2017). Generally, Cuffdiff converts aligned read counts to FPKM (Fragments Per Kilobase of exon per Million fragments mapped), then calculates the observed change in expression and the statistical significance. However, the program assumes that the number of reads produced by each transcript is proportional to its abundance (Trapnell et al., 2012). An observation made using Cuffdiff was genes that are close to a particular locus were assigned the same fold change value. Therefore, to resolve this challenge, the data were analyzed with DESeq2, which uses the count matrix (Love et al., 2014).
According to the analysis, the number of significantly regulated (P # 0.05) genes in each of the tested conditions differed between Cuffdiff and DESeq2, with DESeq2 identifying more differentially regulated genes than Cuffdiff. Studies to compare the different software packages for differential expression analysis have also reported the presence of more genes using DESeq2 (Nookaew et al., 2012;Seyednasrollah et al., 2015). An evaluation of the performance of different software packages for differential expression analysis conducted by Zhang and co-workers (2014) showed that the performance of DESeq and Cuffdiff2 was similar in the analysis of sequencing dept $ 20 million reads. The investigation suggested that if the number of false positives is a concern, differentially regulated genes shared between the two programs should preferably be used for further analysis. Therefore, in this study, only the significantly differentially regulated genes (P # 0.05), which intersect between Cuffdiff and DESeq2, were considered for further investigation. In the different conditions tested, the number of genes intersecting between the two programs are indicated in Table 1. Genes with a Log 2 fold change of more than one (Log 2 $ 1) were identified from the gene list in each of the experimental conditions and for further analysis.
The distribution of the differentially regulated genes between the different test conditions was investigated (Figure 1). Most genes present are unique to a treatment condition, suggesting that the different treatment conditions may have induced unique processes or proteins.

Enrichment analysis of differentially regulated genes
Biological process enrichment analysis was conducted to assign gene ontology (GO) terms to the genes obtained from different conditions, using Protein ANalysis THrough Evolutionary Relationships (PANTHER) Gene Ontology (GO) (Ashburner et al., 2000;Mi et al., 2017; The Gene Ontology Consortium 2019). Different clusters shown on the Venn diagram for both upregulated and downregulated genes were individually analyzed. The downregulated and upregulated genes shared by all the tested conditions were not enriched for specific GO terms, possibly because most of the genes present in this section are yet to be characterized. However, a search of the possible role of the downregulated genes indicates that they are associated with filamentous growth, cell wall, and membrane transport processes, with the upregulated genes linked to roles in transcriptional regulation, secretory functions, and transport. Figure 2 is a summary of processes with enriched GO terms for the different test conditions.

Processes enriched by growth with arachidonic acid
The unique downregulated genes associated with biofilms grown in the presence of AA (DA) were enriched for four GO terms, namely, valine catabolic process, aerobic respiration process, drug metabolic processes, and oxidation-reduction (Table S1). In the valine catabolic process ontology, HPD1 (which encodes hydroxypropionate dehydrogenase) and ALD6 (which encodes malonate aldehyde dehydrogenase) were identified. In C. albicans, these genes are involved in the degradation of propionyl-CoA via a modified b-oxidation pathway to produce acetyl-CoA (Otzen et al., 2014). Propionyl-CoA is a toxic intermediate produced by yeast cells from the breakdown of cholesterol, valine, odd-chain fatty acids, methionine, isoleucine, and threonine (Brock and Buckel 2004;Zhang et al., 2004). Although C. albicans can use propionyl-CoA as a sole carbon source, glucose is still the preferred carbon source for C. albicans (Otzen et al., 2014). Therefore, the downregulation of propionyl-CoA degrading genes in biofilms grown in the presence of AA could imply either the sufficient availability of glucose or the absence of propionyl-CoA.
Aerobic respiration GO term was associated with NDH51 (NADH dehydrogenase complex I (CI)) and ACO1 (aconitase) genes. NADH dehydrogenase complex I is a primary component of the C. albicans mitochondrial electron transport chain (Nosek and Fukuhara 1994). Aconitase, an iron-sulfur containing protein, catalyzes the conversion of citrate to isocitrate in the tricarboxylic acid cycle (Gangloff et al., 1990). The downregulation of these genes could be due to glucose availability because ACO1 is regulated by catabolite repression in the presence of glucose and glutamate (Gangloff et al., 1990;Flores et al., 2000). Glucose availability also influences NDH51 expression (McDonough et al., 2002;Vellucci et al., 2007).
The CAT1 gene, which encodes the catalase enzyme (Wysong et al., 1998), was assigned to the drug metabolic process GO term. Glucose availability has been reported to repress the expression of catalase in C. albicans, even when treated with hydrogen peroxide (Nakagawa et al., 1999). The downregulation of the genes associated with these processes suggests that the presence of AA may be interfering with glucose uptake and metabolism.
The enriched upregulated genes, which were unique to biofilms grown in the presence of AA, were associated with four GO terms, namely, ribosomal large subunit biogenesis, rRNA processing, singlespecies submerged biofilm formation, translation and cellular response to drugs (Table S2). The process of ribosomal biogenesis entails highly coordinated procedures with intense energy demands for the synthesis, processing, transport and ribosome assembly (Thomson et al., 2013). DEAD-box proteins DRS1, DBP3, and DBP7 are involved in RNA metabolism processes, which include the unwinding of double-stranded RNA molecules (Linder 2006). However, DRS1, DBP3, and DBP7 are involved in the maturation of ribosomal RNA (Issi et al., 2017). Candida albicans ribosome biogenesis processes are downregulated due to the high energy required when the cell undergoes stress due to reactive oxygen species (ROS) and nutrient depletion (Loar et al., 2004;Kos-Braun and Ko s 2017). However, the upregulation of genes associated with ribosomal biogenesis indicates the presence of acceptable energy levels in the cells n■ Table 1 The number of significant differentially expressed genes identified from different conditions. DA (DMSO and Arachidonic acid), FA (Fluconazole and Arachidonic acid), and FE (Fluconazole and EtOH)

Conditions
Significant regulation Genes common to Cuffdiff and DESeq2 Genes with fold change (Log 2 $ 1) Cuffdiff DESeq2  Downregulated  Upregulated  DA  1200  2124  783  183  244  FA  1790  2853  1253  343  407  FE  1089  1954  803  229  312  Total  when grown in the presence of DA. Also, hyphae inducers such as serum and N-acetyl-D-glucosamine causes a reduction in ribosomal RNA (Fleischmann and Rocha 2015) and the fact that AA inhibits yeast to hyphae transition in C. albicans (Clément et al., 2007;Shareck and Belhumeur 2011), may explain the opposite effect observed here. Genes that play essential roles in biofilm formation in C. albicans such as CSH1, GCA1, NDT80, ADH1, TEC1, and TYE7 were assigned to the single-species submerged biofilm formation GO term, which was enriched in the presence of AA. The CSH1 gene product (Csh1), is a vital virulence factor in C. albicans, which affects cell surface hydrophobicity (Antley and Hazen 1988). It is also involved in the irreversible adhesion of cells during the early biofilm formation (Chandra et al., 2001;Bujdáková et al., 2013). Gca1 is a plasma membrane-associated glucoamylase enzyme with a role in the formation of C. albicans biofilm matrix (Nobile et al., 2009) and has been reported to assist with starch degradation and energy generation (Maicas et al., 2016). Alcohol dehydrogenase encoding ADH1 catalyzes the reversible conversion of aldehyde to ethanol (Bertram et al., 1996), and is downregulated in C. albicans biofilms (Mukherjee et al., 2006). Transcriptional regulators NDT80, TEC1 and TYE7 were also associated with this ontology, while NDT80 and TEC1 are among the six C. albicans biofilm master circuit transcriptional regulators (Nobile et al., 2012), TYE7 is involved in the activation of the glycolytic pathway (Askew et al., 2009). Transcriptional factors NDT80 and TEC1 are required for hyphal growth, although NDT80 is not always required in all growth conditions, such as during the use of glycerol as a carbon source (Daniels et al., 2015;Min et al., 2018). The upregulation of genes associated with the biofilm formation ontology may be a possible attempt by the yeast to compensate for decreased biofilm formation after treatment with DA ( Figure S1a).
Processes enriched by growth with fluconazole Gene ontologies enriched in downregulated genes that are unique to biofilms grown in the presence of fluconazole (FE) include DNA methylation or demethylation, negative regulation of helicase activity, DNA replication checkpoint, DNA strand elongation involved in DNA replication and DNA-dependent DNA replication maintenance of fidelity (Table S3). DNA methylation and demethylation processes are epigenetic modification, which represses transcription and transposon control (Finnegan et al., 2000;Martienssen and Colot 2001). In C. albicans, this process is associated with the repression of genes involved with phenotypic switching, structural genes and gene regulation. Additionally, the presence of environmental stressors can also induce C. albicans DNA methylation (Mishra et al., 2011). The presence of fluconazole has been associated with stresses that have been explicitly identified as the primary cause for the formation of trimera, leading to aneuploidy in C. albicans (Harrison et al., 2014). The histone H3 isoform encoding gene, HHT21, was associated with the DNA methylation or demethylation ontology. According to an investigation by Zacchi and co-workers (2010), transcription and chromosome segregations are affected by a change in the number of C. albicans histones, and this results in growth defects and aneuploid formation.
Upregulated genes unique to biofilms grown in the presence of FE were enriched for GO terms, ergosterol biosynthetic process, farnesyl diphosphate metabolic process, and cell wall chitin biosynthetic process (Table S4). Ergosterol is essential in yeast because it assists with the maintenance of membrane fluidity, permeability, and integrity (Lv et al., 2016). Fluconazole inhibits ergosterol synthesis by binding to lanosterol demethylase enzyme, encoded by ERG11 (Bondaryk et al., 2013). The upregulation of ergosterol synthesis genes is one of the known resistance mechanism of C. albicans biofilms against fluconazole treatment (Sanglard et al., 2003;Tsui et al., 2016). Several genes associated with ergosterol synthesis, including ERG3, ERG1, ERG11, ERG9, and CYB5, were linked with this ontology term. The upregulation of processes associated with ergosterol synthesis in the presence of fluconazole indicates a resistance response to compensate for the presence of the antifungal (Bondaryk et al., 2013).
Cell wall chitin biosynthetic process was also associated with genes upregulated in biofilms grown in the presence of fluconazole. Fluconazole induces membrane and cell wall stress, by sterol disruption and increasing the fluidity of the plasma membrane (Abe et al., 2009;Klis and Brul 2015). In response to cell wall stress, C. albicans increases the amount of chitin and expression of cell wall repair proteins (Munro et al., 2007;Sorgo et al., 2011).

Processes enriched by growth with both fluconazole and arachidonic acid
The downregulated genes, which were unique to biofilms grown in the presence of both fluconazole and AA (FA), were not enriched for any GO term. Most of the genes in this cluster are uncharacterized. However, those with known functions have roles in zinc homeostasis, oxidative stress, and copper and zinc transport.
Upregulated genes, unique to cells grown in the presence FA, are enriched for AMP biosynthetic processes, fumarate metabolic process, ATP synthesis coupled proton transport, fatty acid oxidation, tricarboxylic acid process and acyl-CoA metabolic process (Table S5).
Purine synthesis in C. albicans occurs via the de novo biosynthesis pathway and the salvage pathways (Rolfes 2006). Adenylosuccinate synthase (encoded by ADE12), which catalyzes the conversion of inosine monophosphate (IMP) to adenylosuccinate, and adenosine kinase (encoded by ADO1), which recycles adenosine to AMP, were assigned to the AMP biosynthetic process ontology. The energy demand for purine biosynthesis is high; therefore, the regulation of this process is via feedback inhibition by the ATP and ADP end products and the presence of excess purine in the cell-extracellular environment (Rébora et al., 2001;Rolfes 2006). The upregulation of genes associated with purine synthesis in biofilms grown in the presence of fluconazole and AA indicates an attempt to increase ATP generation under these conditions. ATP synthesis coupled proton transport GO term was represented by ATP synthase alpha subunit (encoded by ATP1), F1F0-ATP synthase beta subunit (encoded by ATP2), oligomycin-sensitive conferring protein (OSCP) (encoded by ATP5) and F1F0 ATP synthase subunit h (encoded by ATP14), these proteins constitute different components of the mitochondrial ATP synthase, required for the phosphorylation of ADP to ATP using the energy provided by the electrochemical gradient (Arnold et al., 1998;Xu et al., 2015). The absence of ATP1 or ATP2 in C. albicans result in the cells being avirulent, unable to grow on non-fermentable or non-glucose carbon sources, defective in filamentation, biofilm formation, and ATP production (Li et al., 2017;Li et al., 2018). Lack of growth in media containing glycerol and lactate as a carbon source was observed in Saccharomyces cerevisiae ATP14 homozygous mutant (Arselin et al., 1996). This also supports the possible need for ATP generation in biofilms grown in the presence of both fluconazole and AA.
Fumarate metabolic process and tricarboxylic acid process are represented by genes encoding enzymes within the tricarboxylic acid (TCA) cycle. Fumarate hydratase (encoded by FUM12) was assigned to both ontologies. In contrast, adenylosuccinate synthase (encoded by ADE12) was assigned to the fumarate metabolic process and malate dehydrogenase (encoded by MDH1-1) and succinate-CoA ligase (encoded by LSC1) to the tricarboxylic acid process. Fumarase is present in the cytoplasm and mitochondria of eukaryotic cells. The mitochondrial fumarase catalyzes the reversible hydration of fumarate to L-malate in the TCA cycle, while the cytosolic fumarase is synthesized in response to DNA damage (Yogev et al., 2010). The upregulation of genes associated with the TCA cycle in the presence of both fluconazole and AA may also be as a result of the need for an increase in ATP generation.
Putative pyruvate dehydrogenase (encoded by PDB1), carnitine acetyltransferase (encoded by CAT2), and succinate-CoA ligase (encoded by LSC1) were associated with the GO term acyl-CoA metabolic process. The functions of CAT2 and LSC1 in the cell have been discussed earlier, but PDB1, which encodes a b-subunit of the pyruvate dehydrogenase complex, catalyzes the conversion of pyruvate to acetyl-CoA and CO 2 (Vellucci et al., 2007). The GO terms enriched in the upregulated genes of C. albicans cells grown in the presence of fluconazole and AA are mostly associated with energy generation. This suggests that under this treatment condition, the cell attempts to increase its energy output, and this indicates that there is a demand for energy.
Overlapping processes enriched by growth with either arachidonic acid or fluconazole The upregulated and downregulated genes shared between biofilms grown in the presence of either AA (DA) or fluconazole (FE) were not enriched for any specific GO terms. The reported descriptions for the downregulated genes include phosphatase complex subunit, SIN3binding protein, a heme oxygenase, and a high-affinity iron permease, while the upregulated genes are associated with the cell wall and DNA synthesis.
Overlapping processes enriched by growth with either arachidonic acid or a combination of fluconazole and arachidonic acid The downregulated genes which were shared between biofilms grown in the presence of either AA (DA) or a combination of fluconazole and AA (FA) were enriched for GO terms, which include hydrogen sulfide biosynthetic process, sulfate assimilation and sulfur amino acid biosynthetic process (Table S6). The different processes mentioned above are represented by ATP sulfurylase (encoded by MET3), putative adenylylsulfate kinase (encoded by MET14), putative sulfite reductase beta subunit (encoded by ECM17) and O-acetylhomoserine O-acetylserine sulfhydrylase (encoded by MET15) which are genes in the C. albicans methionine/cysteine biosynthesis pathway. Methionine/cysteine synthesis is essential for biofilm formation in C. albicans (García-Sánchez et al., 2004;Murillo et al., 2005). A failure in the formation of proper biofilm, diminished adhesion and a decrease in filamentation was observed in homozygous ecm17Δ/Δ C. albicans strains and with the absence of methionine or cysteine (Li et al., 2013). MET3, which is an activator of the methionine/cysteine pathway for sulfur assimilation, is one of the most highly upregulated genes during biofilm formation in C. albicans (Murillo et al., 2005). However, the presence of methionine or cysteine at a concentration higher than 1 mM represses the expression of MET3 (Care et al., 1999). The concentrations of methionine and cysteine in RPMI media are 0.43 mM and 0.062 mM, respectively. Hence, the downregulation in methionine/cystine processes may not be due to catabolite repression but rather the presence of AA in both these conditions, which may repress the methionine/cysteine pathway, resulting in low biofilm formation.
The GO terms enriched in upregulated genes shared between biofilms grown in the presence of either DA or FA include xenobiotic transport, spermine transport, and spermidine transport (Table S7). These GO terms were represented by an ATP-binding cassette (ABC) transporter (encoded by YOR1) as well as FLU1, which encodes for fluconazole resistance 1 protein, an efflux pump belonging to the major facilitator superfamily (MFS) class of transporters involved with drug efflux (Cannon et al., 2009). However, contrary to expectations of its role in fluconazole efflux, the disruption of FLU1 in C. albicans has a negligible effect on its fluconazole susceptibility, although its deletion causes susceptibility to mycophenolic acid (Calabrese et al., 2000).
The C. albicans Yor1 is similar to the S. cerevisiae Yor1, which is an ATP-binding cassette (ABC) transporter, which acts as an oligomycin resistance permease (Gaur et al., 2005). However, Yor1 in C. albicans effluxes beauvericin, a compound that enhances azole susceptibility, from the cell (Shekhar-Guturja et al. 2016). Although the upregulated genes associated with this GO term in DA and FA are involved in transport, their role in the transport of either fluconazole or AA has not been identified.
Overlapping processes enriched by growth with either fluconazole or a combination of fluconazole and arachidonic acid The GO terms enriched for downregulated genes shared between biofilms grown in the presence of either fluconazole (FE) and a combination of fluconazole and AA (FA) include filamentous growth regulation and cytoskeleton-dependent cytokinesis ( Table S8). The genes linked with the filamentous growth ontology include Zn(II)2-Cys6 transcription factors (encoded by ROB1, AHR1, and UME6) as well as a G-protein alpha subunit (encoded by GPA2). Biofilm formation in C. albicans is under the control of six master circuit transcriptional factors, including Rob1 (Nobile et al., 2012). The importance of ROB1 to C. albicans biofilm formation is demonstrated with the failure of rob1Δ/Δ mutants to form biofilms and the haploinsufficiency of the heterozygous mutant (Nobile et al., 2012;Glazier et al., 2017). The zinc cluster transcription factor, Ahr1, activates adhesion, and hyphal regulatory genes (Askew et al., 2011). Cell adhesion to host surface and hyphae formation are essential processes of C. albicans biofilms formation and also contributes to the biofilm structure (Chandra et al., 2001;Finkel and Mitchell 2011). UME6 is an essential regulator for hyphal extension and virulence in C. albicans (Banerjee et al., 2008). This is also seen in the enrichment of the biological adhesion GO term, which was represented by agglutinin-like protein 3 (encoded by ALS3), Zn(II)2Cys6 transcription factor (encoded by UME6), and the zinc cluster transcription factor (encoded by AHR1). Candida albicans agglutinin-like protein 3 (Als3) is known to function in adhesion, biofilm formation, and iron acquisition processes (Almeida et al., 2008;Liu and Filler 2011). However, its expression is only observed in the hyphae and pseudohyphae (Hoyer et al., 1998). The deletion of ALS3 in C. albicans resulted in the formation of thin and disorganized biofilms in vitro (Nobile et al., 2006). The downregulation of critical transcriptional regulatory genes in biofilms grown in FE and FA may contribute to the presence of poorly formed biofilms under these conditions ( Figure S1b).
The cytoskeleton-dependent cytokinesis GO term was represented by septin (encoded by CDC10), formin-binding protein (encoded by HOF1), and Ras GTPase-activating protein (encoded by IQG1). Septins are filament-forming proteins which comprise Cdc3, Cdc11, Cdc12 and Cdc10. (Longtine et al., 1996;Frazier et al., 1998). Although C. albicans viability or growth rate is not affected by a mutation of CDC10, the observed defects in C. albicans CDC10 mutants include a failed cytokinesis due to abnormal cell morphology of elongation and enlargement of cells without separation as well as defects in chitin localization (Warenda and Konopka 2002). HOF1 encodes an F-box protein involved in cytokinesis and the formation of a ring around the bud neck of the mother and daughter cells. In the absence of HOF1, C. albicans cells suffer severe cytokinesis and separation defects, which is evident in the formation of chains of cells (Li et al., 2008;Jonkers and Rep 2009). IQG1 also localizes to the bud neck, contributing to neck formation with the assembly and contraction of the actomyosin ring. In the absence of IQG1, cell necks are swollen and broaden (Li et al., 2008). The downregulation of processes associated with cytokinesis process in biofilms grown in the presence of fluconazole as well as a combination of fluconazole and AA may be due to the presence of fluconazole as it has been shown to induce trimeras as discussed previously (Harrison et al., 2014).
The GO term enriched in the upregulated genes shared by biofilms grown in FE) and FA includes the glyoxylate cycle, acetate catabolic process, carbon utilization and fatty acid catabolic process (Table S9). All these processes were represented by the isocitrate lyase (encoded by ICL1), malate synthase (encoded by MLS1), and a multifunctional enzyme of the b-oxidation pathway (encoded by FOX2). ICL1 and MLS1 are the first two genes of the glyoxylate cycle, an alternative pathway that allows the conversion of two-carbon acetate into a fourcarbon dicarboxylic acid, which replenishes the TCA cycle or serves as a precursor for either amino acid or carbohydrate synthesis (Kunze et al., 2006).
Conversely, Fox2 catalyzes both enoyl-CoA hydratase and 3-hydroxyacyl-CoA dehydrogenase activities in the b-oxidation pathway (Hiltunen et al., 2003). Alternative energy-generating pathways, such as the b-oxidation pathway and the glyoxylate pathway, allow the yeast to grow on alternative carbon sources such as fatty acids, ethanol and acetate (Kunze and Hartig 2013). Treatment of C. albicans cells with fluconazole has been shown to increase metabolites connected with the central carbon metabolic pathway (Katragkou et al., 2016). Therefore, the upregulation of glyoxylate pathway genes and FOX2 from the b-oxidation pathway may indicate an increase in the energy requirements of the biofilms under these conditions.

Confirmation of RNA-Seq data with NanoString
The fold change obtained using RNA-Seq compared with that from NanoString, using selected genes of interest is shown in Table S10. As observed, the similarity between the transcript levels obtained from the two platforms indicates consistency in the data generated for the analysis.

CONCLUSIONS
Differentially regulated genes were obtained in all the tested conditions with the most differentially regulated genes obtained with the combination of fluconazole and AA (FA). Analysis of the biological processes indicated that in the presence of AA (DA), energy generation processes were downregulated, which may be due to an inhibition of glucose uptake. The low energy generation in the presence of AA may also be associated with the possible upregulation of biofilm formation processes to compensate for low biomass. As expected, the presence of fluconazole enriched for processes involved in ergosterol synthesis and an increase in chitin content. Also, processes associated with abnormal cell cycle, such as aneuploidy, was enriched in the fluconazole treated cells. In cells grown with the combination of fluconazole and AA, several processes associated with ATP synthesis was enriched, indicating the need to produce ATP via alternative processes. Methionine/cysteine synthesis processes, which are required for biofilm formation, were downregulated in the presence of AA. It can be suggested that the presence of AA interferes with the energy generation and methionine synthesis processes as a mechanism for increased susceptibility. This lays the ground for more detailed studies regarding these processes as possible drug targets.

ACKNOWLEDGMENTS
This work was supported by the National Research Foundation of South Africa [grant number 115566 to CHP]. The funders played no role in the study or in the preparation of the article or decision to publish.