Conditional antagonism in co-cultures of Pseudomonas aeruginosa and Candida albicans: An intersection of ethanol and phosphate signaling distilled from dual-seq transcriptomics

Pseudomonas aeruginosa and Candida albicans are opportunistic pathogens whose interactions involve the secreted products ethanol and phenazines. Here, we describe the role of ethanol in mixed-species co-cultures by dual-seq analyses. P. aeruginosa and C. albicans transcriptomes were assessed after growth in mono-culture or co-culture with either ethanol-producing C. albicans or a C. albicans mutant lacking the primary ethanol dehydrogenase, Adh1. Analysis of the RNA-Seq data using KEGG pathway enrichment and eADAGE methods revealed several P. aeruginosa responses to C. albicans-produced ethanol including the induction of a non-canonical low-phosphate response regulated by PhoB. C. albicans wild type, but not C. albicans adh1Δ/Δ, induces P. aeruginosa production of 5-methyl-phenazine-1-carboxylic acid (5-MPCA), which forms a red derivative within fungal cells and exhibits antifungal activity. Here, we show that C. albicans adh1Δ/Δ no longer activates P. aeruginosa PhoB and PhoB-regulated phosphatase activity, that exogenous ethanol complements this defect, and that ethanol is sufficient to activate PhoB in single-species P. aeruginosa cultures at permissive phosphate levels. The intersection of ethanol and phosphate in co-culture is inversely reflected in C. albicans; C. albicans adh1Δ/Δ had increased expression of genes regulated by Pho4, the C. albicans transcription factor that responds to low phosphate, and Pho4-dependent phosphatase activity. Together, these results show that C. albicans-produced ethanol stimulates P. aeruginosa PhoB activity and 5-MPCA-mediated antagonism, and that both responses are dependent on local phosphate concentrations. Further, our data suggest that phosphate scavenging by one species improves phosphate access for the other, thus highlighting the complex dynamics at play in microbial communities.

Like many fermentative organisms, C. albicans produces ethanol. Ethanol is a biologicallyactive metabolite which, at sub-inhibitory concentrations, modulates P. aeruginosa behavior in multiple ways: it induces activity of the sigma factor AlgU through DksA and ppGpp signaling [31]; it promotes Pel matrix production through the Wsp system [29]; it decreases flagellarmediated motility through a surface-sensing pathway [29,32]; it affects programs known to contribute to P. aeruginosa virulence [29,33]; and it fuels fungal antagonism [29]. The broad effects of ethanol apply to many contexts and the response of P. aeruginosa to C. albicans-produced ethanol can serve as a model for how P. aeruginosa may respond to other fermentative fungi and bacteria. We seek to further understand this response and identify common themes which may be implicated in other microbial interactions.
We used P. aeruginosa anti-fungal phenazine production to study the effects of ethanol in co-culture. Previous work has shown that ethanol stimulates the production and secretion of the phenazine 5-methyl-phenazine-carboxylic acid (5-MPCA) by P. aeruginosa and that, in return, P. aeruginosa phenazines promotes C. albicans fermentative metabolism and ethanol production [24][25][26]. P. aeruginosa does not normally secrete 5-MPCA in axenic cultures but in co-culture it does so through the MexGHI-OmpD efflux complex. Consequently, 5-MPCA enters C. albicans cells wherein it reacts with the basic amines such as arginine, disrupting protein function and forming a red pigment whose accumulation causes redox stress and eventually death of C. albicans [24,27,34]. While it is known that ethanol production by the fungus is necessary for P. aeruginosa 5-MPCA release, the mechanisms by which 5-MPCA production is regulated have not yet been described. The mechanisms of stimulation and ensuing consequences of P. aeruginosa 5-MPCA production and accumulation of the red 5-MPCA-derivative within C. albicans cells is a scopic case study for microbial interactions because it is an indicator of general antagonism.
Several studies have described the conditional production of antagonistic factors in response to nutrient availability [35][36][37][38][39][40][41][42][43][44][45][46][47][48][49][50][51][52]. This is often mediated transcriptionally, and such is the case for the low-phosphate response which is mediated by the PhoR-PhoB two-component system. Briefly, inorganic phosphate is sensed through the periplasmic domain of the phosphate transport complex, PstS, that is dependent on the ATPases PstA and PstB. The failure to bind phosphate in low-phosphate environments causes the de-repression of the sensor kinase PhoR which phosphorylates the response regulator PhoB and initiates PhoB DNA-binding to the promoters of many genes, including those in the phenazine biosynthetic pathway, and consequent transcriptional activation. The P. aeruginosa low-phosphate response includes the secretion of an arsenal of phosphatases, phospholipases and DNases that cleave phosphate from diverse macromolecules [40].
In the context of microbial communities, the secretion of these enzymes renders phosphate freely available to any nearby organism. Simultaneous production of antagonistic factors could aid P. aeruginosa in the competition for phosphate amongst other microbes. Indeed, in response to low phosphate and other complex stimuli, P. aeruginosa produces antagonistic factors like phenazines and phospholipases which play important roles in microbial interactions [39,53,54]. How neighboring microbes may alter each other's low-phosphate responses is an open question but it has been reported that P. aeruginosa tailors its low-phosphate response to secondary stimuli: in P. aeruginosa, PhoB interacts with other regulators such as the transcription factor TctD [55] and the sigma factor VreI [37] to orchestrate the expression of its target genes. However, the mechanism by which PhoB exerts condition-specific control over its diverse regulon to manage antagonistic factors in microbial interactions is not yet fully understood.
Analysis of co-culture transcriptomic data from complex environments using curated pathways can be challenging if the complex conditions are not well represented by the data used for pathway definition. Furthermore, pathway definition relies on expert-contributed annotations, yet~38% (2,162 of 5,704) of genes for the PAO1 reference strain (pseudomonas.com) lack descriptions. Recent methods have used unsupervised machine learning to leverage large amounts of transcriptomic data and automatically identify sets of genes with correlated expression patterns across large compendia of samples, agnostic of gene annotations and previously characterized pathways [53,[56][57][58][59][60][61][62][63][64]. In particular, the data-driven tool eADAGE has identified transcriptional signals that contain uncharacterized genes, manifest as small magnitude changes in expression, or are condition-specific.
Here, we demonstrate that P. aeruginosa and C. albicans undergo transcriptional changes in response to one another that are dependent on C. albicans ethanol production. Using eADAGE analysis, we identified a group of PhoB-regulated genes as differentially expressed in response to ethanol, and validated the result using genetic and biochemical assays. Both ethanol and PhoB activity were necessary for the production of the P. aeruginosa antifungal 5-MPCA, and ethanol was sufficient to stimulate PhoB in P. aeruginosa mono-cultures. Further, by examining C. albicans gene expression profiles from the same co-cultures, we found that the C. albicans low-phosphate response is inversely correlated to that of P. aeruginosa. In summary, we show that P. aeruginosa only produces antifungal 5-MPCA when C. albicans is not secreting extracellular phosphatases and the death of neighboring fungi would simultaneously remove a competitor and provide an increase in the availability of the essential nutrient phosphate. We conclude that the enmity of P. aeruginosa-C. albicans interactions is conditional upon phosphate concentrations, and further, that an induction of a phosphate limitation response in one species may improve access to phosphate in the other.

Ethanol is a defining factor in P. aeruginosa-C. albicans interactions, promoting phenazine-mediated antagonism
When grown on a lawn of C. albicans, P. aeruginosa produces the anti-fungal phenazine 5-MPCA which is taken up by C. albicans and modified within the fungal cells to form a red derivative [24,27,29] that can be seen first below and then as a halo surrounding P. aeruginosa colonies (Fig 1A). The 5-MPCA precursor phenazine-1-carboxylic acid (PCA) can be synthesized via enzymes encoded in either of the two highly similar operons, phzA1B1C1D1E1F1G1 (phz1) and phzA2B2C2D2E2F2G2 (phz2) with different regulation; phz1 contributes to phenazine production in liquid while phz2 is responsible for phenazine production in colony biofilms [65]. Analysis of phenazine production in co-culture colony biofilms found that phz1 was dispensable for the formation of red pigment, while phz2 was required (Fig 1A). Consistent with previous results, 5-MPCA-derived red pigment formation required phzM [27], mex-GHI-ompD and soxR [24], and was over-abundant upon inhibition of the conversion of 5-MPCA into another phenazine vis deletion of phzS [66] (S1A Fig). We have previously reported, and reproduced here, that P. aeruginosa 5-MPCA production required C. albicans ethanol production as C. albicans adh1Δ/Δ, which lacks the major ethanol dehydrogenase, did not elicit 5-MPCA-derived red pigment accumulation [29]. Chromosomal complementation of a single copy of ADH1 in C. albicans restored 5-MPCA-derived red pigment production ( Fig 1A). Consistent with previous reports [26,67], HPLC analysis of C. albicans culture supernatants showed ethanol accumulation, and here we found that deletion of ADH1 led to a 10-fold decrease in ethanol concentrations (S2A Fig). There was not a significant difference in concentrations of acetate or glycerol, two other C. albicans fermentation products when WT was compared to adh1Δ/Δ (S2B Fig).
To determine how C. albicans Adh1 activity influenced P. aeruginosa and how 5-MPCAderived red pigment accumulation influenced C. albicans, we took a dual-seq approach: we collected total RNA for simultaneous transcriptome-wide analyses of both organisms from 16 h co-culture colony biofilms of P. aeruginosa with C. albicans WT, in which 5-MPCA-derivatives accumulated, and P. aeruginosa with C. albicans adh1Δ/Δ, in which 5-MPCA products were not observed (see Fig 1B for experimental set up). Single-species P. aeruginosa and C. albicans colony biofilms grown on YPD medium were also analyzed at the same time point. Principle component analysis of gene expression for each organism differentiated mono-culture and co-culture with the first component PC1, which accounted for 40.1% and 44.7% of total variance for P. aeruginosa and C. albicans respectively (Fig 1C and 1D). The presence of ADH1 in C. albicans constituted a defining feature of co-culture captured by PC2 for both organisms, which accounting for 18.8% and 28.2% of total variance for P. aeruginosa and C. albicans respectively (Fig 1C and 1D).
How fungal ethanol shapes co-culture transcriptomes: The C. albicans perspective C. albicans Adh1 is responsible for reducing acetaldehyde to ethanol during fermentation, and we thus expected metabolic shifts between C. albicans WT and adh1Δ/Δ [29]. We identified differentially expressed genes (DEGs) between co-cultures of C. albicans WT and adh1Δ/Δ with P. aeruginosa (S1 Dataset) and conducted KEGG [68][69][70] pathway over-representation analysis. The C. albicans KEGG pathway for fatty acid beta-oxidation was over-represented in the DEGs and the genes it contained (e.g. FAA2-1, FAA2-3) were more highly expressed in C. albicans WT than in C. albicans adh1Δ/Δ (Fig 1E for pathway view, S2 Dataset for over- In co-culture of C. albicans (C.a.) and P. aeruginosa (P.a.), C.a.-produced ethanol stimulates P.a. to produce 5-MPCA and transcriptional responses ensue from both organisms. A) Co-cultures of P.a. wild type (WT) and mutants lacking phenazine biosynthesis operons (Δphz1 or Δphz2) were inoculated onto 72 h-old lawns of C.a. WT, adh1Δ/Δ and adh1Δ/Δ reconstituted with ADH1 (adh1Δ/ADH1). The red pigmentation indicates production of the phenazine 5-MPCA by P.a. B) Dual-seq allowed for parallel analyses of P.a. (green) and C.a. (red or pink) mRNA expression profiles from co-culture lawns to survey the effects of ethanol (green oval) and 5-MPCA (red oval) on gene expression. C) Principle component analysis (PCA) of TPM (transcripts per kilobase per million reads) from transcriptome profiles of P.a. grown alone (No C.a.), P.a. grown with C.a. WT, and P.a. grown with C.a. adh1Δ/Δ. D) PCA of gene expression profiles of C.a. WT and C.a. adh1Δ/Δ grown in mono-culture (No P.a.) or co-cultures with representation analysis). These results are consistent with a previous metabolomics study that reported a phenazine and other mitochondrial inhibitors increasing beta-oxidation in C. albicans WT [71].
The KEGG pathway for glycolysis (e.g. HXK2, PGI1, TDH3, PGK1) was also over-represented in DEGs between C. albicans WT and adh1Δ/Δ from co-cultures with P. aeruginosa with the genes more highly expressed in C. albicans adh1Δ/Δ. This pathway was also more highly expressed in C. albicans adh1Δ/Δ grown alone, likely as metabolic compensation for the inability to ferment to ethanol (Fig 1E, S2 Dataset). There was also over-representation of the KEGG pathways for iron scavenging (e.g. FRP1, FET99) and the DNA damage response genes (e.g. RBT5, CSA1), with these genes more highly expressed in C. albicans adh1Δ/Δ (Fig 1E, S2 Dataset). Since these pathways were over-represented in DEGs between C. albicans WT and adh1Δ/Δ in both co-and mono-culture, they could represent direct or indirect effects of altered C. albicans metabolism or P. aeruginosa secreted products, inclusively: C. albicans intracellular metabolic products at high concentrations in adh1Δ/Δ, such as methylglyoxal [72,73], and phenazines both cause oxidative stress leading to DNA damage.

How fungal ethanol shapes co-culture transcriptomes: The P. aeruginosa perspective
We identified P. aeruginosa DEGs when grown on C. albicans WT compared to on adh1Δ/Δ. On C. albicans WT, P. aeruginosa upregulated genes involved in 5-MPCA biosynthesis including phzM, which is consistent with differences in 5-MPCA-derived red pigment formation between the two co-cultures ( Fig 1E, S3 Dataset). While the last four genes of the phz operons have fewer than three SNPs between each gene pair and are thus not differentiated by alignment, phzA1, phzB1 and phzC1 have substantial enough differences in sequence from phzA2, phzB2 and phzC2 respectively that the two operons can be distinguished, and we found that transcripts from both phz1 and phz2 operons were more highly abundant by at least 2-fold when P. aeruginosa was grown in co-culture on C. albicans WT relative to on adh1Δ/Δ. Although phzS and phzH are not required for 5-MPCA biosynthesis [24, 29], they have been reported to have coordinated expression with other phenazine genes [55,74] (S1D Fig for pathway) and indeed we saw increases in their expression on C. albicans WT compared to on adh1Δ/Δ as well (Fig 1E, S2 Dataset).
We again conducted KEGG pathway over-representation analysis and found over-representation of three KEGG pathways in DEGs from P. aeruginosa grown in co-culture with C. albicans WT compared to with C. albicans adh1Δ/Δ: phenazine biosynthesis, quorum sensing (QS), and pyochelin biosynthesis (Fig 1E, S2 Dataset). Over-representation of the phenazine biosynthesis pathway was expected based on the upregulation of the phz genes as described above. The identification of QS as an over-represented pathway was not surprising in light of the known regulation of phenazine biosynthesis by QS in response to environmental cues [75,76], including in C. albicans co-culture [27]. P. aeruginosa QS involves three major transcriptional regulators: LasR [77], PqsR [78] and RhlR [77]. RhlR and PqsR, along with the PQS biosynthesis enzyme PqsA, were necessary for phenazine production (S1B Fig). Although ΔlasR appeared to produce less 5-MPCA than P. aeruginosa WT on C. albicans WT, consistent with previous data, it produced an abundance of the blue-green phenazines such as pyocyanin and PCN (S1D Fig) [18]. Upon examining the expression of gene targets for these transcription factors (S4 Dataset for gene lists and sources), we did not find evidence for a uniform change in QS that accompanied the activation of phenazine biosynthesis genes (S1C Fig).
The third over-represented KEGG pathway was the biosynthesis of pyochelin, a siderophore [51]. Expressly, genes involved in pyochelin biosynthesis, import and regulation were lower in P. aeruginosa on C. albicans WT compared to on C. albicans adh1Δ/Δ. These data were supported by the observation that P. aeruginosa produces higher levels of fluorescent siderophores on C. albicans adh1Δ/Δ compared to the WT (Fig 1E, inset). The over-representation of low iron responsive genes in both P. aeruginosa and C. albicans demonstrated that the organisms were experiencing simultaneous iron limitation but not in 5-MPCA-permissive conditions. Taken together these data are consistent with C. albicans ethanol production stimulating P. aeruginosa antagonistic 5-MPCA phenazines which affect C. albicans metabolism and ROS stress pathways and the deletion of Adh1 in C. albicans leading to increased competition for iron but not 5-MPCA production.
The lack of compensatory production of other C. albicans fermentation products such as acetate and glycerol, in the adh1Δ/Δ cultures (S1 Fig) led us to wonder if the loss of Adh1 may increase competition for oxygen due to an increased reliance on oxidative phosphorylation for energy generation and redox balancing. However, we did not observe an over-representation of P. aeruginosa genes regulated by Anr in response to low oxygen [79] in DEGs between P. aeruginosa grown on C. albicans WT versus adh1Δ/Δ (p-value = 0.25, hypergeometric test). Further, deletion of Anr did not inhibit red pigment formation in P. aeruginosa grown on C. albicans WT or promote red pigment formation in P. aeruginosa grown on C. albicans adh1Δ/Δ (Table 1). eADAGE analysis of P. aeruginosa transcriptomes revealed additional pathways differentially active in response to ethanol in co-culture with C. albicans While the analysis of DEGs and the KEGG pathways over-represented therein provided insight into key P. aeruginosa-C. albicans interactions, only 19 of the 120 P. aeruginosa strongly DEGs (|logFC| > 2, FDR < 0.01) fell within the three statistically over-represented KEGG pathways: QS (orange bar), phenazine biosynthesis (red bar) and pyochelin biosynthesis (blue bar) (Fig 2A). To look for additional processes that were affected by C. albicans ethanol production, we further identified patterns in the RNA-Seq data using eADAGE, a denoising autoencoder-based tool [53,63,64]. In eADAGE analysis, the activities of previously-defined gene expression signatures are calculated as weighted sums of normalized gene expression values where gene weights are unique to each signature [64]. Using eADAGE, we found 48 differentially active signatures (DASs) in P. aeruginosa grown on C. albicans WT versus P. aeruginosa grown on C. albicans adh1Δ/Δ (S2 Dataset).
As predicted by the DEGs (Fig 2A), there were multiple DASs in which phenazine biosynthesis and pyochelin biosynthesis KEGG pathways were over-represented (Fig 2B, red and blue bars respectively). The detection of multiple signatures enriched in these pathways was expected based on the presence of similar signatures that are redundant in some contexts but discriminating in others [53]. Thirty-two of the DASs ( However, we had previously identified Node108neg (N108n) as significantly more active in low-phosphate media than phosphate replete media across the compendium of gene expression data on which eADAGE was trained [53]. Therefore, upon identifying Node108neg as the most activated eADAGE signature in response to C. albicans ethanol production in co-culture, we further investigated the genes within Node108neg and their connection to the P. aeruginosa low-phosphate response.

eADAGE analysis suggested P. aeruginosa PhoB up-regulated genes in response to C. albicans ethanol production
In the most upregulated eADAGE signature, Node108neg, the set of PhoB-regulated genes (i.e. the PhoB regulon) was significantly over-represented (hypergeometric test: p = 5.9x10 -9 ). The PhoB regulon has been extensively defined through rigorous experimental methods Table 1. P. aeruginosa strain phenotypes for 5-MPCA-accumulation in co-culture with C. albicans WT and ethanol-induced alkaline phosphatase (AP) activity in mono-culture. AP activity was visualized in colonies on agar containing BCIP. P. aeruginosa mutants were defective in Anr, ppGpp-dependent AlgU and DksA signaling, ethanol catabolism, the kinase KinB and sigma factor VreI, which are known to influence PhoB activity, acetyl-phosphate metabolism and ethanol catabolism. including mutant transcriptomics, motif analysis and chromatin immunoprecipitation assays [55]. Of the 32 genes in Node108neg, 11 were also in the PhoB regulon ( Fig 2D) and they all increased in expression in P. aeruginosa grown on C. albicans WT compared to on adh1Δ/Δ, but the Pho regulon defined in Bielecki et al. [55] was heterogeneously expressed overall ( Fig  2E). We more closely examined DEGs in the context of eADAGE signatures in order understand how changes in signature activities embodied the P. aeruginosa response to C. albicans ethanol production and whether P. aeruginosa gene expression changes between growth on C. albicans WT and on adh1Δ/Δ signaled a low-phosphate response. We visualized relationships among DEGs in the eADAGE gene-gene network. The full gene-gene network consists of the 5,549 P. aeruginosa genes used to create the eADAGE model as vertices with similarities in transcriptional patterns as weighted edges (shorter edges represent higher Pearson correlations between gene weights across all signatures in the eADAGE model) [53,63,64]. Here we show strongly DEGs (logFC > 2, FDR < 0.01) between P. aeruginosa grown on C. albicans WT and on adh1Δ/Δ connected by edges whose weights are drawn from the full gene-gene network (edge cutoff ± 0.5) (Fig 2F). DEGs fell into cliques (S1 Table) when visualized as a sub-network within the eADAGE gene-gene network ( Fig 2F). The three largest cliques contained genes relevant to the biological processes of phenazine biosynthesis (clique 1, 17 genes), pyochelin biosynthesis (clique 2, 29 genes), and the low-phosphate response (clique 3, 23 genes) ( Fig 2F). Other cliques contained genes involved in magnesium flux across the membrane (clique 4), isoprenoid catabolism (clique 5), aconitate porins (clique 6), pyrimidine metabolism (clique 7), pyocin biosynthesis (clique 8), spermidine biosynthesis (clique 9), histidine metabolism (clique 10), the heat shock response (clique 11), and the ROS stress response (clique 12). Most notably, many genes within clique 3, which were related to the low-phosphate response, were also in Node108neg.
Genes within clique 3, which were enriched in the DAS Node108neg, clustered into two groups: four genes were more highly expressed in P. aeruginosa grown on C. albicans adh1Δ/Δ (phoBR and pstABC) and nineteen were more highly expressed in P. aeruginosa grown on C. albicans WT (Fig 2G). The group of P. aeruginosa genes that were more highly expressed on C. albicans WT included those that encode the Hxz type II secretion system (PA14_55450, PA14_55460) and its substrate enzyme the phosphatase lapC [80], the TonB-dependent transporter ExbB2/D2, the phosphatase extracellular alkaline phosphatase PhoA, and the secreted phospholipase PlcN. The transcriptomes of P. aeruginosa grown alone were also analyzed: P. aeruginosa did not have an activated low-phosphate response in mono-culture as transcripts associated with the PhoB regulon were very low and not different from levels in P. aeruginosa grown with C. albicans adh1Δ/Δ (S3 Dataset).
The P. aeruginosa low-phosphate response was activated in co-culture with C. albicans WT but not adh1Δ/Δ We confirmed that some PhoB-regulated transcripts were higher in P. aeruginosa co-cultured with C. albicans WT versus adh1Δ/Δ using the NanoString multiplex RNA analysis method. Seventy-five transcripts were monitored, and data were normalized to six housekeeping genes N108n violet dot). D) P.a. gene expression between co-culture with C.a. WT and adh1Δ/Δ, genes in N108n colored violet and genes also in the PhoB regulon in purple with DEGs cutoffs logFC > 1, FDR < 0.05 marked with red dotted lines. E) Same P.a. genes expression as D, genes in the PhoB regulon colored purple. F) Network analysis of DEGs suggest groups of DEGs have correlated patterns across eADAGE: phenazine biosynthesis (1) is inversely expressed with the low iron response (2) and coordinately upregulated with the lowphosphate response (3) upon exposure to ethanol in co-culture. Other cliques of DEGs participate in shared biological pathways. Descriptions of all cliques in S1 Table. G) The Pho Clique (3)

PLOS GENETICS
Conditional antagonism between C. albicans and P. aeruginosa as described in the Methods section. Again, we found that genes encoding phosphate scavenging enzymes were more highly expressed in P. aeruginosa on C. albicans WT than on adh1Δ/Δ (Fig 3A, top section), while phosphate transport-associated genes (phoR, phoB, pstA and phoU) were inversely regulated (Fig 3A, bottom section). Consistent with the phenotypic data, genes involved in phenazine biosynthesis (phzM, phzA, phzC, and phzH) and efflux (mexG) were higher on C. albicans WT than on C. albicans adh1Δ/Δ as well (Fig 3A, second section).
As PhoB activity is controlled by phosphate levels [81], we assessed the effects of phosphate supplementation in co-culture. The addition of 10 mM potassium phosphate to the medium underlying the co-cultures resulted in a decrease in the expression levels of PhoB regulated genes similar to the deletion of P. aeruginosa phoB (Fig 3A top section). The histogram to the right of Fig 3A shows the mean signal of PhoB-dependence (P. aeruginosa WT on C. albicans WT / P. aeruginosa ΔphoB on C. albicans WT). P. aeruginosa ethanol catabolism genes exaA and exaB were also upregulated in P. aeruginosa on C. albicans WT than on adh1Δ/Δ. The expression of these genes was unaffected by phosphate supplementation or deletion of phoB.
PhoB activity has been shown to affect phenazine production previously [39]. We found that the increased expression of P. aeruginosa genes involved in phenazine biosynthesis and efflux (phzM, phzA, phzC, phzH, and mexG) on C. albicans WT versus on adh1Δ/Δ was suppressed by supplementation with 10 mM phosphate and was not observed in P. aeruginosa ΔphoB mutant. As we and others have shown the antifungal effects of phenazines [25, 27, 30], these data provide a strong link between phosphate availability and antagonism in these cocultures.
To accompany the transcript abundance data, we assessed PhoB-dependent promoter activity using a P. aeruginosa pdtA promoter fusion strain [41]. We found that PhoB-dependent pdtA promoter activity decreased in response to 10 mM phosphate in co-culture with C. albicans WT (Fig 3B). In concert, transcript abundance and promoter activity data suggested that in co-culture, the bioavailability of phosphate modulated PhoB activity.

PhoB was necessary for 5-MPCA-derived red pigment accumulation in P. aeruginosa-C. albicans co-culture
Given the dependence on PhoB for the expression of phenazine biosynthesis genes, we determined if PhoB was necessary for the accumulation of 5-MPCA-derived red pigment in co-culture with C. albicans WT. P. aeruginosa ΔphoB did not support the accumulation of the 5-MPCA-derivative as indicated by the lack of red pigment in co-culture with ethanol-producing C. albicans WT, and this was restored by chromosomal complementation with a wild-type copy of phoB in P. aeruginosa, provided that C. albicans had a functional ADH1 gene (Fig 3C). Phenazine production was also dependent on PhoR, the known regulatory kinase of PhoB, and the ΔphoR phenotype was complemented by a wild-type copy of phoR expressed on an extrachromosomal plasmid.
We further demonstrated that PhoB activity was required for 5-MPCA production by phosphate supplementation to the medium prior to co-culture inoculation. Decreased 5-MPCAderived red pigment accumulation in P. aeruginosa WT co-culture with C. albicans WT was seen across a phosphate gradient plate (Fig 3D, top). The ability of phosphate to quench red pigment formation suggested that co-culture biofilms may deplete the phosphate to below a threshold of activation for the P. aeruginosa PhoB response. Increased PhoB activity was sufficient to overcome suppression by phosphate supplementation as a mutant lacking the phosphate transport ATPase PstB with constitutively active PhoB, continued to produce 5-MPCAderived red pigment even at high phosphate concentrations ( Fig 3D). As expected, P. aeruginosa WT did not form any 5-MPCA-derived pigment in co-culture with C. albicans adh1Δ/Δ at any phosphate concentration tested. Surprisingly, P. aeruginosa ΔpstB grown on adh1Δ/Δ only slightly rescued 5-MPCA-derived red pigment formation (Fig 3D). Therefore, a difference in the ability of C. albicans WT and adh1Δ/Δ to deplete phosphate was not fully responsible for differences in P. aeruginosa PhoB-dependent red pigment formation, but rather C. albicans Adh1 activity provided an additional stimulus, such as ethanol, that created conditions conducive to PhoB-regulated P. aeruginosa antifungal phenazine production.

Ethanol was sufficient to activate PhoB at permissive phosphate concentrations
To determine if ethanol could contribute to the stimulation of PhoB in P. aeruginosa-C. albicans co-culture, we tested whether it was sufficient to stimulate PhoB in P. aeruginosa monoculture. To do so, we monitored PhoB-regulated alkaline phosphatase (AP) activity by the conversion of 5-bromo-4-chloro-3-indolylphosphate (BCIP) into a colorimetric substrate according to published methods [82][83][84]. On MOPS (3-morpholinopropane-1-sulfonic acid) buffered minimal medium after growth for 16 h at 37˚C, P. aeruginosa Δphz, wherein all blue coloration can be attributable to BCIP conversion and not phenazines, AP was detected up to approximately 0.55 mM phosphate in the absence of ethanol (Fig 4A). With the addition of 1% ethanol to the medium, P. aeruginosa showed AP activity at higher phosphate concentrations, up to approximately 0.73 mM phosphate (Fig 4A). This induction was also seen in P. aeruginosa WT, but not ΔphoB, at 0.7 mM phosphate on single concentration plates, and was restored upon complementation with a WT copy of phoB (Fig 4B). Quantification of AP activity under these conditions showed a 4-fold increase in the presence of 1% ethanol for P. aeruginosa WT, only trace AP activity for P. aeruginosa ΔphoB and hyperactivity in ΔpstB (Fig 4C).
To determine if PhoB was acting upon the same targets upon ethanol supplementation in mono-culture as shown for co-culture, we monitored the expression of PhoB targets using the same Nanostring codeset described earlier. P. aeruginosa WT and ΔphoB were grown in mono-culture on MOPS minimal medium at 0.7 mM phosphate with and without 1% ethanol. The same subset of PhoB-regulated genes indicated in co-culture expression analysis increased in P. aeruginosa WT upon ethanol supplementation, including: phoA, an alkaline phosphatase; phnD, a phosphonate transporter; glpQ, a glycerophosphoryl diester phosphodiesterase ( Fig  4D, top section). In mono-culture, when ethanol was supplemented into the medium, we saw an increase in phenazine biosynthesis genes however, unlike in co-culture, the magnitudes of their fold-changes between P. aeruginosa WT and ΔphoB was not as high as those of other PhoB-regulated genes suggesting their expression was stimulated by PhoB-independent factors (Fig 4D, second panel). The set of PhoB-regulated genes whose expression was heterogenous and did not trend upward in co-culture were also not activated in mono-culture upon ethanol supplementation (Fig 4D, bottom panel). Stimulation of PhoB by ethanol in mono-culture suggests that ethanol and low phosphate were stimuli that could have both activated PhoB in co-culture. Activation of PhoB in the presence of ethanol may be indicative of physiological changes that increased phosphate utilization by P. aeruginosa. We noted that mono-culture colony biofilms of P. aeruginosa grown on medium with 1% ethanol had no difference in CFUs or cell culture density upon comparison to the ethanol-free condition (S3 Fig). This suggests that ethanol may increase phosphate utilization or decrease phosphate availability without affecting culture growth.
In light of the recent characterization of the P. aeruginosa response to exogenous ethanol [31] that showed ppGpp, synthesized by RelA and SpoT, DksA, and AlgU mounted a transcriptional response to ethanol, we assessed mutants for ethanol-induced PhoB activation in co-culture by monitoring red pigment accumulation and in mono-culture via BCIP assay. We determined that these genes were not necessary to induce PhoB activity in response to ethanol (Table 1). We also ruled out roles for known mechanisms of alternative PhoB activation including contributions of the non-canonical histidine kinase KinB [53] in both co-and  . aeruginosa (P.a.) mono-culture. A) Alkaline phosphatase (AP) activity visualized by blue color derived from cleavage of BCIP in MOPS medium,0.2% glucose, with a gradient of phosphate (P i ) in the absence and presence of ethanol. The P.a. ΔphzA1-G1/A2-G2 strain was used to eliminate color differences due to phenazine production. B) AP activity in the absence and presence of 1% ethanol (E or EtOH) at 0.7 mM P i for P.a. wild type, ΔphoB and the ΔphoB mutant complemented with a wild-type copy of phoB integrated at the native locus. C) AP activity in cells from colony biofilms grown as in B was measured using the colorimetric substrate mono-culture and the extra-cytoplasmic function (ECF) sigma factor VreI [37, 41] in monoculture, but further investigation is required to determine if VreI played a role in regulating 5-MPCA production in co-culture (Table 1).
We also tested the role of ethanol catabolism in PhoB activation in both mono-and co-culture. Mutants in the ExaA-dependent pathway for ethanol catabolism through acetate including exaA and acsA did not show increases in AP activity in response to 1% ethanol (Table 1) indicating that ethanol catabolism was essential for PhoB activation in mono-culture. We hypothesized that ethanol catabolism led to increased levels of acetyl phosphate, a non-canonical phosphate donor for transcription factors including PhoB, [85][86][87][88] but neither acetyl phosphate biosynthesis mediated by AckA nor catabolism by Pta was necessary for ethanolinduced PhoB activity (Table 1). Mutants defective in the ExaA-dependent ethanol catabolic pathway showed only weak 5-MPCA production on C. albicans lawns and the phenotype could be complemented (Table 1). Together these data suggest that ethanol catabolism plays a role in PhoB activation, perhaps by increasing phosphate utilization, and that additional pathways for ethanol catabolism may be present in co-culture conditions. Further investigation is required to determine the mechanism for ethanol induction of PhoB activity, but these results demonstrate ethanol stimulation and PhoB activation participated in non-linear but intersecting pathways.

Low phosphate and ethanol were additive stimulants for PhoB-mediated red pigment formation
We tested whether ethanol stimulation and canonical PhoB de-repression had additive effects on 5-MPCA-derived red pigment formation in P. aeruginosa-C. albicans co-culture. On nonethanol producing C. albicans adh1Δ/Δ, the constitutive PhoB activity in P. aeruginosa ΔpstB was not sufficient to stimulate red pigment to the extent of that on ethanol-producing C. albicans WT (Fig 4E). To determine if repressive stimuli that were present in C. albicans adh1Δ/Δ, but not WT, co-cultures had a negative effect on P. aeruginosa red pigment formation or if the absence of ethanol was the reason for decreased 5MPCA production, we performed an ethanol supplementation experiment. Strikingly, we found that the addition of 1% ethanol to C. albicans adh1Δ/Δ lawns increased red pigment formation in P. aeruginosa ΔpstB but not in P. aeruginosa WT or ΔphoB (Fig 4E). This revealed that in addition to PhoB-independent effects of ethanol on phenazines production, PhoB activation both through the canonical signaling pathway and by means of ethanol stimulation were necessary for 5-MPCA-derived red pigment formation in co-culture (proposed model in Fig 4F).

The C. albicans low-phosphate response was more active in adh1Δ/Δ than in WT
Given the differences between P. aeruginosa PhoB activity in co-culture with C. albicans WT and adh1Δ/Δ, we sought to understand if C. albicans also experienced differences in phosphate pNPP for P.a. WT, ΔphoB, and ΔpstB. � ,p<0.01 by ANOVA with Tukey's multiple comparison test (n � 3). D) Transcripts within the PhoB regulon involved in P i scavenging and transport and genes involved in phenazine production and EtOH catabolism in P.a. WT and ΔphoB cells grown in the absence and presence of EtOH by Nanostring (codeset PAV5). Expression values normalized to loading controls and housekeeping genes, z-score scaled by gene. Middle bar plot shows log 2 fold-change (logFC) between P.a. WT and P.a. ΔphoB on MOPS+1%EtOH. The bar is colored red if expression is PhoB-dependent (logFC P.a. WT / P.a. ΔphoB > 1, FDR < 0.05) else grey. � , FDR < 0.05, #,FDR > 0.05. E) P.a. WT, ΔphoB and ΔpstB grown on C.a. WT, or C.a. adh1Δ/Δ in the absence or presence of 1% exogenous EtOH. F) P i and EtOH are additive stimuli that promote PhoB-dependent expression of AP and phenazine biosynthesis (PHZ). https://doi.org/10.1371/journal.pgen.1008783.g004

PLOS GENETICS
Conditional antagonism between C. albicans and P. aeruginosa availability by examining its low-phosphate response. In C. albicans, low phosphate induces activity of the transcription factor Pho4 which regulates genes involved in phosphate acquisition as well as the homeostasis of cations (e.g. iron), tolerance of stresses including ROS and arsenic, and fitness in murine models [89][90][91][92][93]. Pho4 regulates 133 genes that were previously identified by Ikeh et al. using transcriptomics as differentially expressed both between C. albicans WT and pho4Δ/Δ in low phosphate and between C. albicans WT in high and low phosphate [89] (see S4 Dataset for gene list). We found that Pho4-regulated genes were overrepresented in DEGs between co-cultures of C. albicans WT with P. aeruginosa and C. albicans adh1Δ/Δ with P. aeruginosa (|logFC| > 1, p < 0.05) (hypergeometric test p = 1.9x10 -3 ). Of the top ten most differentially expressed genes between C. albicans WT and pho4Δ/Δ in low phosphate as determined by Ikeh et al. [89], seven were also strongly differentially expressed (logFC > 2) in co-cultures of C. albicans WT with P. aeruginosa compared to C. albicans adh1Δ/Δ with P. aeruginosa including a secreted phospholipase (PHO100) and two secreted phosphatases (PHO112 and PHO113) (Fig 5A, black bars). Given the strong differential expression of these phosphatases, we assessed phosphatase activity via BCIP assay as done for P. aeruginosa. Consistent with the transcriptional data, higher phosphatase activity was observed in C. albicans adh1Δ/Δ compared to C. albicans WT or the complemented strain adh1Δ/ADH1 in the presence of P. aeruginosa (Fig 5B).
To determine if the activation of the C. albicans low-phosphate response in adh1Δ/Δ in coculture was a consequence of competition for phosphate with P. aeruginosa we examined C. albicans gene expression and phosphatase activity in mono-culture. We found that the same Pho4-regulated C. albicans genes that were more highly expressed in C. albicans adh1Δ/Δ than in C. albicans WT in co-culture with P. aeruginosa were also more highly expressed in C. albicans adh1Δ/Δ when each strain was grown alone (Fig 5C). Consistent with the increased expression of phosphatase genes, we also observed higher phosphatase activity in C. albicans adh1Δ/Δ cultures when compared to C. albicans WT or the complemented strain adh1Δ/ ADH1 in mono-culture using the BCIP assay (Fig 5D). For reference, we assayed C. albicans pho4Δ/Δ and found that the low levels of phosphatase activity seen for C. albicans WT and adh1Δ/ADH1 were not evident in pho4Δ/Δ indicating that that phosphatase activity was Pho4-dependent ( Fig 5D). The high Pho4 response in C. albicans adh1Δ/Δ, evident even in mono-culture, suggested that Adh1 activity impacts C. albicans phosphate access, phosphate requirements or Pho4 regulation. By analyzing changes in gene expression between C. albicans adh1Δ/Δ and WT compared to their respective mono-culture conditions, we did not find over-representation of the Pho4 regulon in co-culture conditions compared to C. albicans only cultures (hypergeometric test: p > 0.05) (S1 Dataset). Since phosphatases produced by C. albicans as part of its low-phosphate response are secreted, we hypothesized that their production could affect P. aeruginosa in co-culture, perhaps by providing access to phosphate liberated from macromolecules and this model is discussed further below.

In co-culture with ethanol-producing C. albicans, P. aeruginosa PhoB plays independent roles in phosphate scavenging and phenazine-mediated antagonism
PhoB was important for P. aeruginosa growth in co-culture as ΔphoB formed fewer CFUs than did P. aeruginosa WT on C. albicans WT (Fig 5E). Notably, a comparable number of P. aeruginosa CFUs were recovered from P. aeruginosa WT and Δphz in co-culture suggesting that the lack of phenazines was not a major reason for decreased CFU formation in ΔphoB (Fig 5E). The defect in CFU formation by ΔphoB when compared to WT or Δphz held true on C. albicans with a complemented copy of ADH1. On the C. albicans adh1Δ/Δ mutant, however, P.  C. albicans (C.a.) adh1Δ/Δ has increased expression of the Pho4-mediated low-phosphate response. A) Previously characterized Pho4-regulated genes [89] including a phospholipase and two phosphatases (black bars) were more highly expressed in C.a. adh1Δ/Δ than C.a. WT in P.a. co-cultures (data shown as z-scores of TPM). Pho4-dependence is shown in the right-hand bar plot as log 2 FC C.a. WT/C.a. pho4Δ/Δ using data from [89]. B) Analysis of phosphatase activity in C.a. WT, C.a. adh1Δ/Δ or the complemented strain adh1Δ/ADH1 using the aeruginosa CFUs were similar for P. aeruginosa WT, ΔphoB, and Δphz suggesting that in the absence of C. albicans Adh1 activity, P. aeruginosa PhoB, and its roles in phosphate acquisition or phenazine production, were no longer necessary for fitness. This finding is consistent with the model that elevated phosphatase production by the C. albicans adh1Δ/Δ may eliminate the need for P. aeruginosa to produce phosphatases (Fig 5B and 5D).
C. albicans CFUs were enumerated from the same co-cultures. Consistent with previous reports on the antifungal properties of the phenazine 5-MPCA [25, 27], C. albicans had lower CFUs upon co-culture with P. aeruginosa WT compared to when co-cultured with ΔphoB or Δphz (Fig 5F). In C. albicans adh1Δ/Δ co-cultures which did not support P. aeruginosa PhoBdependent phenazine production (Fig 3), C. albicans CFUs were not different between co-cultures with P. aeruginosa WT, ΔphoB or Δphz.

P. aeruginosa and C. albicans asynchronously activated their lowphosphate responses dependent on C. albicans Adh1
While P. aeruginosa had higher activity of its low-phosphate responsive regulator PhoB and phosphatase production in the presence of ethanol-producing C. albicans WT (Figs 2G, 3A and 3B), C. albicans had higher activity of its low-phosphate responsive transcription factor Pho4 and higher phosphatase levels in C. albicans adh1Δ/Δ (Fig 5). To more completely examine the relationship between the P. aeruginosa and C. albicans low-phosphate responses in coculture, we performed a cross-species correlation analysis on the dual-seq co-culture gene expression data for a subset of P. aeruginosa PhoB-regulated genes (clique 3) and a subset of C. albicans Pho4-regulated genes (those also DEGs between C. albicans WT and adh1Δ/Δ shown in Fig 5A). There was a striking pattern of inverse correlations between P. aeruginosa PhoBregulated genes and C. albicans Pho4-regulated genes (Fig 6A, upper triangle). In addition to being of high magnitudes, many same-species correlations (white background) as well as cross-species correlations (grey background) were statistically significant as indicated by circles (Fig 6A, lower triangle).
The asynchronous and apparently inverse activations of the P. aeruginosa and C. albicans low-phosphate responses in co-culture, in combination with the differences in the requirement for PhoB on C. albicans WT versus adh1Δ/Δ (Fig 5E and 5F), suggest that the two microbes were not responding to common environmental stimuli but rather that they influenced each other. The increased phosphatase activity from C. albicans adh1Δ/Δ even in the absence of P. aeruginosa (Fig 5D) suggested that activation of the C. albicans low-phosphate response may have prevented activation of the P. aeruginosa low-phosphate response. We speculate increased phosphate availability in concert with the lack of ethanol production led to low PhoB activity in P. aeruginosa in co-culture with C. albicans adh1Δ/Δ and a consequent lack of 5-MPCA-derived accumulation.
Taken together, our results led to the model that C. albicans fermentative metabolism promoted P. aeruginosa PhoB-dependent 5-MPCA production through positive stimulation by ethanol and a competition for phosphate that lead to permissively low phosphate concentrations (Fig 6B). PhoB and phenazine production decreased C. albicans fitness (Fig 5F). colorimetric phosphatase substrate BCIP. C) The same Pho4-regulated genes as shown in A in mono-cultures. Righthand bar plot shows Pho4-dependence as in A. D) More phosphatase activity was observed in C.a. adh1Δ/Δ than in strains with ADH1 in in mono-culture. As predicted, phosphatase activity is not evident in the C.a. pho4Δ/Δ strain. Phosphatase activity visualized via BCIP as in B. E) CFUs of P.a. WT, ΔphoB or Δphz after co-culture for 72 h with C.a. WT, C.a. adh1Δ/Δ or adh1Δ/ADH1. � ,p<0.01 by ANOVA with Dunnett's multiple comparisons test (n � 3). F) CFUs of C.a. from the same samples analyzed in E. � ,p<0.01 Student's t-test adjusted for multiple comparisons by Holm-Sidak method (n � 3). https://doi.org/10.1371/journal.pgen.1008783.g005

PLOS GENETICS
Conditional antagonism between C. albicans and P. aeruginosa  6. P. aeruginosa (P.a.) and C. albicans (C.a.) asynchronously active low-phosphate responses. A) Pearson correlation analysis between P.a. (green annotations) and C.a. (orange annotations) low-phosphate responsive genes from co-cultures of P.a. WT with either C.a. WT or adh1Δ/Δ. Inverse cross-species correlations between P.a. PhoB-and C.a. Pho4-regulated genes were apparent. Log 2 FC (p<0.05) between P.a. with either C.a. WT or C.a. adh1Δ/Δ is shown in the right-hand bar plot. Lower half of correlogram shows which correlations are significant (filled circles) and Inversely, loss of Adh1 in C. albicans removed the positive effects of ethanol on P. aeruginosa and increased C. albicans Pho4 activity and phosphatase secretion which we hypothesize liberated sufficient phosphate to suppress P. aeruginosa PhoB-dependent 5-MPCA production (Fig 6B).

Discussion
Here, we show that P. aeruginosa production of the antifungal phenazine 5-MPCA is dependent on PhoB and C. albicans ethanol production. As ethanol is a common metabolite secreted by many fermentative organisms and phosphate is a universally essential nutrient, the processes cross-regulated by ethanol and phosphate signaling in the model organisms P. aeruginosa and C. albicans may be generalizable and aid in understanding the dynamics of complex, diverse microbial communities. These results also highlight the need to consider environmental factors when characterizing microbial interactions.
While we did not elucidate the mechanism by which ethanol influences PhoB activity, we identified ExaA-based ethanol catabolism as a potential regulator of the PhoB-mediated ethanol response including 5-MPCA production [94,95]. Ethanol induction of PhoB did not require either acetyl phosphate or ppGpp (Table 1), two pathways previously linked to ethanol catabolism [96,31], and thus we speculate that the effects could be multifactorial and may involve conditions such as changes in NAD(P)H/NAD(P)+ ratios, the generation of acetaldehyde or acetate as metabolic intermediates, or other roles played by enzymes involved in ethanol oxidation [97]. The influence of ethanol on PhoB and phenazine production may also involve other regulators as it has been linked to PQS and RhlR signaling [36,39,76], and may interface with systems that respond to distinct cues such as nutrient depletion, surface signals or slow-growth states. The Pho regulon has been well characterized in terms of direct PhoB targets [55], but has also been shown to include additional indirect targets across various lowphosphate media [44].
The constant utilization and solubilizing of phosphate in dynamic co-cultures creates an environment which is quite different from conditions in the low-phosphate media in which the canonical low-phosphate response was defined. While the non-canonical PhoB response reported here had an unusual divergence in the expression of the phoBR genes themselves and phosphate transport genes from the expression of other PhoB targets, eADAGE was able to identify this signal. The 600 gene expression patterns observed by eADAGE through its unsupervised training on over 1,000 publicly available P. aeruginosa gene expression profiles found signature Node108neg, which contained this non-canonical subset of the PhoB regulon activated in co-culture and by ethanol. Other signatures in eADAGE, such as Node164pos, contained the canonical PhoB regulon with both phoBR and the pst genes along with genes encoding phosphate scavenging enzymes. Our analysis by eADAGE also facilitated the development of other hypotheses that can be tested in the future. We observed enrichment of genes involved in leucine catabolism which may be indicative of P. aeruginosa response to C. albicans-produced farnesol [18,28,98,99]. Other interesting pathways that emerged in the eADAGE analysis were those for sulfur assimilation, zinc transport, and iron uptake, and coculture conditions may provide an opportunity to understand the connections between these pathways and their roles in stress responses and virulence factor production. This study indicates correlation values by color intensity relative to scale. Same species comparisons have white backgrounds and cross-species correlations have grey backgrounds. B) Model of PhoB activity in P.a.-C.a. co-cultures. PhoB mediates the conditional production of the antagonistic, antifungal phenazine 5-MPCA in response to low phosphate and fungal ethanol production.
https://doi.org/10.1371/journal.pgen.1008783.g006 highlights the power of using unsupervised machine learning methods, in conjunction with a compendium of versatile conditions, to identify higher order gene expression patterns that are not evident in linear correlation-based analyses and have not yet been manually annotated or systematically described.
The intersection of ethanol and phosphate signaling may serve to limit P. aeruginosa production of 5-MPCA and other antifungal products to when ethanol-producing species are nearby and P. aeruginosa is phosphate-limited. This type of two-factor regulation manages a dynamic relationship between these two opportunistically pathogenic model organisms. These findings also suggest that the dynamics of phosphate sensing, transport and enzymatic release by one organism can influence the behavior of other microbes in complex ways, and these relationships must be considered in the prediction of outcomes in relationships with conditional antagonism.

Strains and growth conditions
Bacterial strains and plasmids used in this study are listed in S2 Table. Bacteria were maintained on LB (lysogeny broth) with 1.5% agar [100]. Yeast strains were maintained on YPD (yeast peptone dextrose) with 2% agar. Where stated, ethanol (200-proof) was added to the medium (liquid or molten agar) to a final concentration of 1% and BCIP (Sigma-Aldrich #1158002001, stock solution of 60 μg/mL dimethylformamide) to a final concentration of 6 ng/L. Planktonic cultures were grown on roller drums at 37˚C for P. aeruginosa and at 30˚C for C. albicans.

Construction of in-frame deletion, complementation and expression plasmids
Construction of plasmids, including in-frame deletion and complementation constructs, was completed using yeast cloning techniques in Saccharomyces cerevisiae as previously described [101] or Gibson assembly [102,103]. Primers used for plasmid construction are listed in S3 Table. In-frame deletion and single copy complementation constructs were made using the allelic replacement vector pMQ30 [101]. Promoter fusion constructs were made using a modified pMQ30 vector with lacZ-GFP fusion integrating at the neutral att site on the chromosome. The pdtA promoter region 199 bp upstream of the transcriptional start site (that included a PhoB binding site) was amplified from WT P. aeruginosa PA14 gDNA using the Phusion High-Fidelity DNA polymerase with primer tails homologous to the modified pMQ30 ATT KI vector containing tandem lacZ-gfp reporter genes. All plasmids were purified from yeast using Zymoprep Yeast Plasmid Miniprep II according to manufacturer's protocol and transformed into E. coli strain S17/λpir by electroporation. Plasmids were introduced into P. aeruginosa by conjugation and recombinants were obtained using sucrose counter-selection. Genotypes were screened by PCR and plasmid constructs were confirmed by sequencing.

Co-culture and mono-culture colony biofilms
Co-cultures were inoculated first with 300 μl of a C. albicans culture in YPD grown for 16 then diluted in dH 2 O to OD 600 = 5, which was bead spread onto YPD plates. C. albicans mono-cultures were inoculated with 5 μl of the same cell suspension as spots on YPD plates. C. albicans cultures were incubated for 16 hours at 30˚C then 24 hours at room temperature (~23˚C). Then, 5 μl of P. aeruginosa suspension, prepared from a 5 mL culture in LB grown for 16 hours then diluted in dH 2 O to OD 600 = 2.5, was spotted on top of the C. albicans lawns for co-cultures or onto YPD for mono-cultures. Co-culture and mono-culture colony biofilms were then incubated for 16 hours at 30˚C. For gradient plates (described below) 500 μl of C. albicans suspension was used to spread a lawn that was pre-grown as above and P. aeruginosa was spotted across as 12 evenly spaced 5 μl spots. All images were taken on a Canon EOS Rebel T6i camera. For visualization of siderophores, pictures were taken under UV light.

HPLC analysis
C. albicans cultures were sub-cultured from overnight cultures grown in 5 mL YPD cultures at 30˚C into fresh 5 mL YPD cultures in duplicate and incubated, aerated on a roller drum, at 30˚C for 8 hours before supernatants were collected for HPLC analysis as in [26].

RNA collection
Total RNA was harvested from P. aeruginosa mono-cultures, C. albicans mono-cultures and P. aeruginosa-C. albicans co-culture colony biofilms grown as described above. All samples were collected as cores from agar plates: cores were taken using a single-use straw, cells were suspended by shaking agar plugs in 1 mL dH 2 O on the disrupter genie for three minutes. Cells were spun down and resuspended in 1 mL Trizol and lysed by bead beating with mixed sizes of silicon beads on the Omni Bead Ruptor. Centrifugation induced phase separation and RNA was extracted from the aqueous phase where it was subsequently precipitated out with isopropanol and linear acrylamide. RNA was pelleted, washed with 70% ethanol, resuspended in nuclease-free dH 2 0 and stored at -80˚C. Samples were prepared for sequencing with DNase treatment, ribodepletion and library preparation in accordance with Illumina protocols. Samples were barcoded and multiplexed in a NextSeq run for a total of 4.7x10 8 reads.

RNA-Seq processing
Reads were processed using the CLC Genomics Workbench with which reads were trimmed and filtered for quality using default parameters. For co-cultures, reads were first aligned to the P. aeruginosa UCBPP_PA14 genome from www.pseudomonas.com. All unaligned reads were aligned to C. albicans SC5314 genome Assembly 22 from www.candidagenome.org. For mono-cultures, reads were only aligned to their appropriate reference genome. Results were exported from CLC including total counts, CPM and TPM. R was used for principle component analysis (prcomp, stats library [104]) and consequent plotting (autoplot, ggplot2 [105]) of gene expression TPM data.
R was also used for differential gene expression analysis. EdgeR was used to process both P. aeruginosa and C. albicans gene expression separately [106]. Generalized linear models with mixed effect data design matrices were used to calculate fold-changes, p-values and FDRs for each comparison of interest. Volcano plots using EdgeR output (log 2 fold-change and -log 10 FDR) were made in R as well (ggplot2) [105].

Accession number
Data for our RNA-Seq analysis of P. aeruginosa and C. albicans gene expression has been uploaded to the GEO repository (https://www.ncbi.nlm.nih.gov/geo/) with the accession number GSE148597 and raw reads are available via the SRA database with the accession number. SRP256305 and the associated BioProject accession number PRJNA625101.

eADAGE analysis
We performed an eADAGE analysis in accordance with the workflow published in the ADA-GEpath R package [64]. Briefly, each gene expression profile in counts per million (CPM) from our RNA-Seq experiment was used to calculate lower dimensional representations of the data called signature activity profiles. Then, differentially active signatures were identified by linear model (limma, stats). This resulted in a set of signatures that were significantly different, but which may have been redundant. We applied pareto front optimization of minimal pvalue and maximal absolute fold-change to arrive at a set of candidate signatures that exhibit statistically significant differences and less redundancy. Heatmaps show gene expression (CPM) or eADAGE signature activity scaled by feature (gene or signature) and are hierarchically clustered by sample using the complete method with Euclidean distance (Complex-Heatmap [108]).

NanoString analysis
NanoString analysis was done on RNA isolated as for RNA-Seq (without DNase treatment) and 100 ng were applied to the codeset PaV5 (sequences for probesets used in this study in S4 Dataset) and processed as previously reported [2]. Counts were normalized to the geometric mean of spiked-in technical controls and five housekeeping genes (ppiD, rpoD, soj, dnaN, pepP, dapF). Normalized counts were used for heatmap construction and fold-change calculations.

Measurement of β-galactosidase in reporter fusion strains
For co-culture promoter activity assays, C. albicans lawns were grown as described for RNA-Seq. P. aeruginosa was inoculated onto two polycarbonate filters with 0.22 μM pores (Millipore) placed on the C. albicans lawns to allow for interaction through diffusible compounds but separation of cells for quantification of promoter activity. P. aeruginosa cells were suspended in PBS by disrupter genie and diluted to OD 600 = 0.05. β-Galactosidase (β-Gal) activity was measured as described by Miller [109]. β-Gal activity was measured in P. aeruginosa WT was normalized to that in ΔphoB which acted as a negative control for background-levels of PhoB-independent promoter activity.

Assessment of phosphatase activity
For quantification of AP activity, we used a colorimetric assay using p-Nitrophenyl phosphate (pNPP) (NEB) as a substrate. Briefly, 5 μl of P. aeruginosa overnight culture was inoculated onto MOPS with 0.2% glucose and 0.7 mM phosphate agar plates on filters as for promoter activity assays with and without 1% ethanol and incubated at 37˚C for 16 hours. Colony biofilms were collected from filters as described for the promoter fusion assays. 100 μl of cell suspension was added to 900 μl of 0.01 M Tris-HCl pH 8 buffer. After the addition of 25 μl 0.1% SDS and 50 μl chloroform cells were incubated at 30˚C for 10 minutes. 30 μl of the aqueous phase was transferred to a 96 well plate containing 15 μl reaction buffer (5 μl 0.5 mM MgCl2 and 10 μl 1 M Tris pH 9.5) and 5

Gradient plates
Phosphate gradient plates were made similarly to previously reported methods of creating pH gradient plates [110]. For YPD-based phosphate gradients plates, first 32 mL of molten YPD +10 mM potassium phosphate pH 6 were poured into a 10 cm square petri dish (Corning, BP124-05) that rested in a custom 3D-printed prop that held the plate slanted at a 30˚angle. Once the bottom layer had solidified, the plate was laid flat and 32 mL of molten YPD agar without phosphate supplementation were poured atop. For MOPS-based gradient plates used for P. aeruginosa mono-cultures the procedure was the same except the first layer was 32 mL MOPS minimal media with 0.2% glucose and 1 mM phosphate and the top layer was 32 mL MOPS minimal medium with 0.2%glucose and 0.4 mM phosphate. When needed, BCIP was added as described above.