Carbon Dioxide Mediates the Response to Temperature and Water Activity Levels in Aspergillus flavus during Infection of Maize Kernels

Aspergillus flavus is a saprophytic fungus that may colonize several important crops, including cotton, maize, peanuts and tree nuts. Concomitant with A. flavus colonization is its potential to secrete mycotoxins, of which the most prominent is aflatoxin. Temperature, water activity (aw) and carbon dioxide (CO2) are three environmental factors shown to influence the fungus-plant interaction, which are predicted to undergo significant changes in the next century. In this study, we used RNA sequencing to better understand the transcriptomic response of the fungus to aw, temperature, and elevated CO2 levels. We demonstrate that aflatoxin (AFB1) production on maize grain was altered by water availability, temperature and CO2. RNA-Sequencing data indicated that several genes, and in particular those involved in the biosynthesis of secondary metabolites, exhibit different responses to water availability or temperature stress depending on the atmospheric CO2 content. Other gene categories affected by CO2 levels alone (350 ppm vs. 1000 ppm at 30 °C/0.99 aw), included amino acid metabolism and folate biosynthesis. Finally, we identified two gene networks significantly influenced by changes in CO2 levels that contain several genes related to cellular replication and transcription. These results demonstrate that changes in atmospheric CO2 under climate change scenarios greatly influences the response of A. flavus to water and temperature when colonizing maize grain.


Introduction
Aspergillus flavus is a saprophytic fungus that infects several crops of agronomic importance, including corn, cotton, peanuts and tree nuts. Prior to harvest, A. flavus may infect the fruiting bodies or seeds in crops and produce several toxic secondary metabolites, including the polyketide derived aflatoxins (AFs), cyclopiazonic acid and aflatrem. In the USA and other industrialized countries, the establishment of contamination thresholds by regulatory agencies and close monitoring of crops have minimized the direct impact on human health. However, economic losses remain significant. Regarding AF contamination in the USA alone, estimates of losses are between $163 million for maize crops to $500 million annually for maize, peanuts and other crops [1,2]. In lower middle income countries (LMCs) where the regulatory controls either do not exist or are not enforced, especially in

The Interaction of Water, Temperature, and CO 2 Impact AFB 1 Production in Maize Grain
Overall, AFB 1 production showed a positive correlation with water availability and CO 2 levels. AFB 1 was extracted and quantified after 10 days incubation at 30 • C or 37 • C, 0.91 or 0.99 a w , and 350, 650, or 1000 ppm CO 2 . At low a w ( Figure 1A) the quantity of AF produced was 5-to 10-fold lower than at high a w levels ( Figure 1B). However, the effect of CO 2 was positively correlated with AFB 1 production at both high and low a w levels. With an elevated CO 2 level of 650 ppm, there was an interaction effect with temperature and a w . At 0.91 a w , AFB 1 production was highest at 37 • C. Conversely, at 0.99 a w the temperature condition of 30 • C exhibited higher toxin levels. The trends observed indicate that CO 2 affects toxin production. The maximum quantity of toxin was observed at elevated levels of CO 2 (650 ppm) with low water activity and at 30 • C, however, no additional accumulation of AFB 1 was observed at 1000 ppm.
Toxins 2018, 10, 5 3 of 15 on AFB1 production. We used functional genomics to better understand the transcriptomic response of the fungus to CC parameters of aw, temperature and CO2 levels, with a view towards predicting changes in fungal infection and toxin production associated with resilience to such climatic stress factors.

The Interaction of Water, Temperature, and CO2 Impact AFB1 Production in Maize Grain
Overall, AFB1 production showed a positive correlation with water availability and CO2 levels. AFB1 was extracted and quantified after 10 days incubation at 30 °C or 37 °C, 0.91 or 0.99 aw, and 350, 650, or 1000 ppm CO2. At low aw ( Figure 1A) the quantity of AF produced was 5-to 10-fold lower than at high aw levels ( Figure 1B). However, the effect of CO2 was positively correlated with AFB1 production at both high and low aw levels. With an elevated CO2 level of 650 ppm, there was an interaction effect with temperature and aw. At 0.91 aw, AFB1 production was highest at 37 °C. Conversely, at 0.99 aw the temperature condition of 30 °C exhibited higher toxin levels. The trends observed indicate that CO2 affects toxin production. The maximum quantity of toxin was observed at elevated levels of CO2 (650 ppm) with low water activity and at 30 °C, however, no additional accumulation of AFB1 was observed at 1000 ppm.

Effect of Three-Way Interacting CC Conditions on Gene Expression
RNA-seq of A. flavus genes showed a large global effect with water and temperature stress, but a limited effect with elevated CO2 levels. Between 6.22 × 10 5 and 3.37 × 10 7 reads mapped to exogenic regions of A. flavus strain 3357 (Table 1). Principle component analysis (PCA) indicated that aw caused the largest variance observed (Figure 2A), accounting for 67% of the variance, however samples with similar water activity cluster together on PC1, indicating there little sample variation for this metric. The second principle component indicated that a change in temperature was the second most important factor, but only for samples stored with freely available water (0.99 aw). All samples stored under water stress (0.91 aw) clustered together and showed relatively little inter-sample variance. At high aw, a change in temperature resulted in a high variance in gene expression. The total number of genes affected under the various conditions are illustrated in Figure 2B. For each data point in Figure 2B, the unidentified variable (aw level, temperature or CO2 level) is the baseline condition (30 °C, 0.99 aw, or 350 ppm CO2) for those samples being and under low water activity levels of 0.91 a w ("30/91"), the effects of CO 2 are minimal; however, at 37 • C ("37/91") there are increases in AFB 1 production correlating with higher CO 2 levels. (B) At high water activity levels the biosynthesis of AFB 1 was significantly elevated at both 30 • C ("30/99") and 37 • C ("37/99"). ANOVA: analysis of variance statistical analysis. * p ≤ 0.05.

Effect of Three-Way Interacting CC Conditions on Gene Expression
RNA-seq of A. flavus genes showed a large global effect with water and temperature stress, but a limited effect with elevated CO 2 levels. Between 6.22 × 10 5 and 3.37 × 10 7 reads mapped to exogenic regions of A. flavus strain 3357 (Table 1). Principle component analysis (PCA) indicated that a w caused the largest variance observed (Figure 2A), accounting for 67% of the variance, however samples with similar water activity cluster together on PC1, indicating there little sample variation for this metric. The second principle component indicated that a change in temperature was the second most important factor, but only for samples stored with freely available water (0.99 a w ). All samples stored under water stress (0.91 a w ) clustered together and showed relatively little inter-sample variance. At high a w , a change in temperature resulted in a high variance in gene expression. The total number of genes affected under the various conditions are illustrated in Figure 2B. For each data point in Figure 2B, the unidentified variable (a w level, temperature or CO 2 level) is the baseline condition (30 • C, 0.99 a w , or 350 ppm CO 2 ) for those samples being compared. Of note,~10-fold more genes responded to temperature and CO 2 stress at high a w levels ( Figure 2B, upper graph). Likewise,~3x more genes were affected by temperature and~25x more genes were affected by CO 2 in the non-stressed temperature conditions (30 • C; Figure 2B, middle graph). High CO 2 levels (1000 ppm) had a significant impact on the number of genes expressed, decreasing the number of genes affected by water and temperature bỹ 3 fold ( Figure 2B bottom graph). In summary, elevated temperature, water stress, and increased CO 2 decreased the number of differentially expressed genes when compared to the non-stressed conditions. 2B, upper graph). Likewise, ~3x more genes were affected by temperature and ~25x more genes were affected by CO2 in the non-stressed temperature conditions (30 °C; Figure 2B, middle graph). High CO2 levels (1000 ppm) had a significant impact on the number of genes expressed, decreasing the number of genes affected by water and temperature by ~3 fold (Figure 2B bottom graph). In summary, elevated temperature, water stress, and increased CO2 decreased the number of differentially expressed genes when compared to the non-stressed conditions.  A significantly larger number of genes were affected by water stress, whereas approximately equal numbers were affected by temperature and CO2 (Figure 3, left and center). Most of the genes that were affected by changing temperature and CO2 levels were also affected by low aw, with 472 genes being upregulated by all three environmental conditions and 564 genes being downregulated. Figure 3 illustrates A significantly larger number of genes were affected by water stress, whereas approximately equal numbers were affected by temperature and CO 2 ( Figure 3, left and center). Most of the genes that were affected by changing temperature and CO 2 levels were also affected by low a w , with 472 genes being upregulated by all three environmental conditions and 564 genes being downregulated. Figure 3 illustrates how changing all three environmental variables simultaneously resulted in a significantly larger number of genes (4853) being affected than changing variables independently. In other words, there is a cumulative effect of changing environmental conditions on global gene expression. The group labeled "Genes affected individually" refers to genes that are affected by just a w , temperature, or CO 2 levels when examined separately. Sixty-nine genes were affected by each of the three conditions individually, but not affected when all three conditions are applied simultaneously.
how changing all three environmental variables simultaneously resulted in a significantly larger number of genes (4853) being affected than changing variables independently. In other words, there is a cumulative effect of changing environmental conditions on global gene expression. The group labeled "Genes affected individually" refers to genes that are affected by just aw, temperature, or CO2 levels when examined separately. Sixty-nine genes were affected by each of the three conditions individually, but not affected when all three conditions are applied simultaneously. . When all three conditions are changed simultaneously, 5820 genes are affected, however when only one environmental condition is changed, only 967 of these genes are affected, indicating a significant difference between the cumulative and individual effects of environmental changes (right).

Effect of Three-Way Interacting CC Conditions on Biological Processes
After 10 days A. flavus colonization of the maize kernels, quantitative PCR was conducted on two AF cluster genes: aflR, the regulating transcription factor, and aflD, a reductase ( Figure 4A). Gene expression levels are shown relative to a control condition of 30 °C, 0.99 aw, and 350 ppm CO2. Four important observations were made: (1) at 30 °C/0.99 aw there were decreases in aflR and aflD expression, even at elevated CO2 levels (650 ppm; 1000 ppm) relative to the control (350 ppm). This is in contrast to what was observed at 37 °C, where aflD is activated at higher levels in the 650 and 1000 ppm CO2 treatments at 0.99 aw; (2) in two separate stressed conditions 30 °C/0.91 aw (water stressed conditions) and 37 °C/0.99 aw (high temperature), there are sharp differences in aflR expression levels between 650 and 1000 ppm, indicating that some CO2 level between the two may represent a threshold; (3) at low aw and high temperature, there were higher levels of activity than at 0.99 aw/350 ppm; (4) at 37 °C/0.99 aw, a dose response in relation to CO2 was observed. . When all three conditions are changed simultaneously, 5820 genes are affected, however when only one environmental condition is changed, only 967 of these genes are affected, indicating a significant difference between the cumulative and individual effects of environmental changes (right).

Effect of Three-Way Interacting CC Conditions on Biological Processes
After 10 days A. flavus colonization of the maize kernels, quantitative PCR was conducted on two AF cluster genes: aflR, the regulating transcription factor, and aflD, a reductase ( Figure 4A). Gene expression levels are shown relative to a control condition of 30 • C, 0.99 a w , and 350 ppm CO 2 . Four important observations were made: (1) at 30 • C/0.99 a w there were decreases in aflR and aflD expression, even at elevated CO 2 levels (650 ppm; 1000 ppm) relative to the control (350 ppm). This is in contrast to what was observed at 37 • C, where aflD is activated at higher levels in the 650 and 1000 ppm CO 2 treatments at 0.99 a w ; (2) in two separate stressed conditions 30 • C/0.91 a w (water stressed conditions) and 37 • C/0.99 a w (high temperature), there are sharp differences in aflR expression levels between 650 and 1000 ppm, indicating that some CO 2 level between the two may represent a threshold; (3) at low a w and high temperature, there were higher levels of activity than at 0.99 a w /350 ppm; (4) at 37 • C/0.99 a w , a dose response in relation to CO 2 was observed. The heat map ( Figure 4B) indicates relative expression levels for all genes in the AF gene cluster obtained from RNA-seq. Hierarchical clustering of sample conditions (top brackets) and genes (left side) are indicated. The AF cluster genes aflD and aflR are clustered relatively close together (genes 3 and 6, respectively), indicating similar expression profiles, which is in agreement with qPCR results where 9 of the 12 samples are similarly expressed. Hierarchical clustering of the samples indicated water availability as the primary determinant of expression patterns. After 10 days, the 30 °C/0.99 aw/350 ppm CO2 condition exhibited higher overall expression levels, distinct from any of the stressed conditions. Notably, expression at 30 °C/0.99 aw/350 ppm is higher than at 650 ppm and 1000 ppm CO2. Three of the genes in the AF biosynthesis pathway, fatty acid synthases aflA and aflB, and the polyketide synthase aflC, all of which are responsible for the synthesis of norsolorinic acid, were enriched in all samples. The reductases coded by aflD and aflF, responsible for the reduction of norsolorinic acid, were also enriched at relatively high levels; however, a third reductase, aflE, exhibited the highest enrichment level in the most stressed condition examined (37 °C/0.91 aw and 1000 ppm CO2).
Enrichment analysis was conducted to identify KEGG biological processes with an overrepresentation of genes being up-or downregulated. Table 2 lists the KEGG categories affected by changes in conditions, both individually and combined, from which several important observations can be made. Primarily, energy-related metabolic processes such as glycolysis/gluconeogenesis, purine/pyrimidine metabolism, and starch/sucrose metabolism are prominent among conditions analyzed. Most genes in this category that are essential for glycolysis, including triosephosphate isomerase (AFLA_094630), fructose bis phosphate aldolase (AFLA_030930), and pyruvate kinase (AFLA_087900) are all downregulated in all the After 10 days of incubation on maize kernels gene levels at 30 • C generally decrease, however the effects of high CO 2 (1000 ppm) levels at 0.91 a w indicate decrease values (left), possibly in response to elevated AFB 1 levels. At 37 • C gene levels remain high, however, again at 1000 ppm CO 2 the transcription factor aflR is decreased. (B) The heat map of regularized log transformed counts indicate hierarchal clustering associated with water activity levels. The clustering also indicates expression patterns that suggest early genes in the pathway may be responsive to high CO 2 levels (See Results).
The heat map ( Figure 4B) indicates relative expression levels for all genes in the AF gene cluster obtained from RNA-seq. Hierarchical clustering of sample conditions (top brackets) and genes (left side) are indicated. The AF cluster genes aflD and aflR are clustered relatively close together (genes 3 and 6, respectively), indicating similar expression profiles, which is in agreement with qPCR results where 9 of the 12 samples are similarly expressed. Hierarchical clustering of the samples indicated water availability as the primary determinant of expression patterns. After 10 days, the 30 • C/0.99 a w / 350 ppm CO 2 condition exhibited higher overall expression levels, distinct from any of the stressed conditions. Notably, expression at 30 • C/0.99 a w /350 ppm is higher than at 650 ppm and 1000 ppm CO 2 . Three of the genes in the AF biosynthesis pathway, fatty acid synthases aflA and aflB, and the polyketide synthase aflC, all of which are responsible for the synthesis of norsolorinic acid, were enriched in all samples. The reductases coded by aflD and aflF, responsible for the reduction of norsolorinic acid, were also enriched at relatively high levels; however, a third reductase, aflE, exhibited the highest enrichment level in the most stressed condition examined (37 • C/0.91 a w and 1000 ppm CO 2 ).
Enrichment analysis was conducted to identify KEGG biological processes with an over-representation of genes being up-or downregulated. Table 2 lists the KEGG categories affected by changes in conditions, both individually and combined, from which several important observations can be made. Primarily, energy-related metabolic processes such as glycolysis/gluconeogenesis, purine/pyrimidine metabolism, and starch/sucrose metabolism are prominent among conditions analyzed. Most genes in this category that are essential for glycolysis, including triosephosphate isomerase (AFLA_094630), fructose bis phosphate aldolase (AFLA_030930), and pyruvate kinase (AFLA_087900) are all downregulated in all the conditions examined. The KEGG categories affected, that are unique to changes in CO 2 concentration (30 • C/0.99 a w /350 ppm CO 2 vs. 30 • C/0.99 a w /1000 ppm CO 2 ), are cysteine and methionine metabolism (p = 0.009) and folate biosynthesis (p = 0.030). Three categories of KEGG were affected only when all three environmental conditions were changed simultaneously

Effect of CO 2 and Interactions with Other Abiotic Factors on Secondary Metabolite Gene Clusters
To identify secondary metabolic gene clusters that were affected by the interacting environmental conditions tested according to the RNA-seq results, a Secondary Metabolite Unique Regions Finder (SMURF) analysis of the A. flavus genome was conducted. Table 3 lists several SMURF-identified secondary metabolic gene clusters labeled according to Georgianna, et al. [18] and their corresponding relative expression values (log 2 fold change). The full list of secondary metabolite-associated genes identified by SMURF is provided in Supplementary Table S1.
Many of the genes listed in Table 3 have high sequence identity to previously characterized genes (in black). Of note, four of the genes (shown in red) are affected by both of the elevated CO 2 levels (650 ppm and 1000 ppm). These consist of two dimethylallyl tryptophan synthases (DMTS), a terpene cyclase, and the hybrid NRPS/PKS shown to be responsible for leporin biosynthesis (see Discussion).

Identification of Gene Networks Affected by CO 2 Levels
Gene co-expression networks were determined by analyzing all RNA sequencing results using WGNCA. The results were then visualized using Cytoscape. This analysis revealed two prominent gene networks with significant numbers of genes affected by CO 2 levels ( Figure 5).

Identification of Gene Networks Affected by CO2 Levels
Gene co-expression networks were determined by analyzing all RNA sequencing results using WGNCA. The results were then visualized using Cytoscape. This analysis revealed two prominent gene networks with significant numbers of genes affected by CO2 levels ( Figure 5). Genes with altered differential expression at 30 °C, 0.99 aw and 1000 ppm CO2, compared with 350 ppm, are colored and genes not differentially expressed are shown in grey. The network illustrated in Figure 5A shows 905 genes interacting in total, with 268 genes differentially expressed. The network has three distinct clusters on the left, right, and center, and the center cluster is coexpressing with genes in the outer clusters (indicated by black lines (edges)). The left and right clusters (95 and 67 genes, respectively) have edges connecting to the 45 genes in the center cluster. Gene ontology (GO) enrichment analysis of the individual clusters indicated the molecular function of genes on the left involve primarily ATP-binding protein kinases (e.g., GO:0005524, ATP binding; GO:0004672, protein kinase activity) (See Supplemental Table S2). The right cluster is enriched in structural and secretory-related genes (e.g., GO:0000166, nucleotide binding, Rheb, and Myo5; GO:0051056 regulation of small GTPase mediated signal transduction, Sar1). The center of the network contained no enriched GO categories, however it does contain several transcription factors and transcriptional regulatory elements (AFLA_029620, AFLA_114920, AFLA_003630). Figure 5B is a second network (network 2) identified by WGNCA analysis and Cytoscape rendering. It shows 415 genes in the network, 114 of which are significantly affected by changing CO2 levels. Network 2 consists of a significantly enriched number of ribosomal-related genes (GO:0003735, structural constituent of ribosome). Other non-ribosomal genes include and Hsp90-binding chaperone sba1 (AFLA_095590), a mycelial catalase cat1 (AFLA_090690), and the AF cluster gene aflT (AFLA_139420) (Supplementary Table S2).

Discussion
To date, the majority of research pertaining to environmental effects on AF production, fungal growth, and plant pathogenicity have focused on the effects of water availability and temperature. Genes with altered differential expression at 30 • C, 0.99 a w and 1000 ppm CO 2 , compared with 350 ppm, are colored and genes not differentially expressed are shown in grey. The network illustrated in Figure 5A shows 905 genes interacting in total, with 268 genes differentially expressed. The network has three distinct clusters on the left, right, and center, and the center cluster is co-expressing with genes in the outer clusters (indicated by black lines (edges)). The left and right clusters (95 and 67 genes, respectively) have edges connecting to the 45 genes in the center cluster. Gene ontology (GO) enrichment analysis of the individual clusters indicated the molecular function of genes on the left involve primarily ATP-binding protein kinases (e.g., GO:0005524, ATP binding; GO:0004672, protein kinase activity) (See Supplemental Table S2). The right cluster is enriched in structural and secretory-related genes (e.g., GO:0000166, nucleotide binding, Rheb, and Myo5; GO:0051056 regulation of small GTPase mediated signal transduction, Sar1). The center of the network contained no enriched GO categories, however it does contain several transcription factors and transcriptional regulatory elements (AFLA_029620, AFLA_114920, AFLA_003630). Figure 5B is a second network (network 2) identified by WGNCA analysis and Cytoscape rendering. It shows 415 genes in the network, 114 of which are significantly affected by changing CO 2 levels. Network 2 consists of a significantly enriched number of ribosomal-related genes (GO:0003735, structural constituent of ribosome). Other non-ribosomal genes include and Hsp90-binding chaperone sba1 (AFLA_095590), a mycelial catalase cat1 (AFLA_090690), and the AF cluster gene aflT (AFLA_139420) (Supplementary Table S2).

Discussion
To date, the majority of research pertaining to environmental effects on AF production, fungal growth, and plant pathogenicity have focused on the effects of water availability and temperature. Work by Medina et al. [19,20] was among the first attempts to examine the effects of a w and temperature in the context of higher CO 2 levels. They found that the interactive effect of water stress, high temperature and high CO 2 increased both gene expression of the AF biosynthesis genes aflR and aflD, and AFB 1 production in maize kernels. While the experimental procedure here permitted us to observe changes in gene expression and AF production after 10 days of growth, it is limited in that we cannot observe a temporal response of gene activation followed immediately by AF biosynthesis. However, we further expand on previous findings, allowing us to characterize putative secondary metabolic gene clusters, important developmental genes, and identify networks of co-expressed genes affected by CO 2 levels.
Most evidence involving fungal carbon metabolism involves the release of CO 2 via cellular respiration, however CO 2 serves other functions. Hall et al. [21] reported that CO 2 serves as an intra-colony signaling molecule important for colonization and pathogenesis in the fungus Candida albicans. It has also been shown that insects, in symbiosis with fungi, demonstrate preferences for certain ranges of elevated CO 2 in which to conduct their fungal-rearing [22]. Furthermore, elevated CO 2 present in the ambient environment of soil samples containing a mixed fungal population decreases respiration activity [23] and decreases AF accumulation in A. parasiticus [24]. The data here demonstrates that while marked effects in transcription (and by consequence growth and toxin production) are primarily a result of water availability, and secondarily temperature, changing CO 2 levels altered fungal response to both water and temperature changes. This is made apparent by the total numbers of genes affected and by our observing that changes brought on by low water or high temperature stresses can vary depending on CO 2 availability. Furthermore, evidence indicates that the combination of environmental stressors, including high CO 2 , have a compounding effect compared to that of individual stressors.
The production of secondary metabolites under different environmental conditions is of concern due to potential shifts in toxigenic potential. The biosynthetic gene cluster 19 in A. flavus contains a dimethylallyl tryptophan synthase (DMAT), which prenylates tryptophan or other aromatic substrates, however the secondary metabolite biosynthesized has yet to be identified. While the expression profile of individual cluster genes does not necessarily correlate with metabolite biosynthesis, DMATS genes are involved in the biosynthesis of several toxic metabolites produced by fungi that infect and contaminate crops such as cyclopiazonic acid [25] in A. flavus and ergotamine in Claviceps purpurea [26]. Toxic metabolites can serve offensive or defensive functions, increasing virulence of the fungus toward its host, or protecting the fungus from threats by other microorganisms or predation. The backbone gene in cluster 23, a hybrid Non-ribosomal Peptide Synthase/Polyketide Synthase (NRPS/PKS), is upregulated with increasing CO 2 levels. This cluster is responsible for the biosynthesis of leporins [27]. Leporin A has anti-insectan properties [28], while leporin B [29] can chelate iron, forming a trimer with Fe 3+ [27]. Siderophore biosynthesis and siderophore-mediated iron uptake have been found to increase under hypoxic conditions in A. fumigatus [30]. AFLA_125760, which was also upregulated at elevated CO 2 levels, encodes for a terpene cyclase flanked by a putative P450 alkane hydroxylase (AFLA_125750) and a steroid alpha reductase (AFLA_125740). These types of cyclases convert the linear triterpene squalene into cyclized products under hypoxic conditions. Cyclized triterpenes, such as fungal ergosterol and bacterial hopenoids, provide cell membrane structural integrity and fluidity. Production of ergosterol, the most prevalent cyclized triterpene found in fungal cell membranes, is affected by mechanical and oxidative stress [30,31]. Other cyclic triterpenes may also be expressed under hypoxic and stressful conditions to improve cell membrane integrity.
Here we have demonstrated for the first time transcriptome-wide changes that occur in a pathogenic fungus while colonizing maize kernels under these interacting CC-related environmental conditions. AF biosynthesis and genes from the AF biosynthetic cluster responded to elevated CO 2 levels, as did several other identified secondary-metabolic gene clusters. Further, there is a global change in the transcriptome in response to water and temperature stress under high CO 2 conditions. This works lays a solid foundation for further research to establish which genes and gene networks should be targeted for fungal inhibition in future CC scenarios.

Sample Preparation and Treatment
Undamaged French feed maize kernels were used in this study. Initially a standard curve was conducted whereby known amounts of water were added to multiple 10 g samples of kernels, incubated at 4 • C for 48 h, and the resulting a w of the kernels was determined using an Aqualab 4 TE water activity meter (Decagon Devices, Pullman, WA, USA).
Based on a w standard curve results, a volume of water was added to each 10 g sample of maize kernels to obtain a w levels of 0.99 and 0.91 (minus 200 µL for the later additions of A. flavus spore suspension). The maize kernels were then placed in glass culture vessels containing a microporous lid, which allows for moisture and air exchange (Magenta, Sigma Ltd., Castleford, UK). Subsequently, 200 µL of spore suspension (approx. 10 6 spores/mL) were added to make up the predetermined amounts of water required and thoroughly mixed. The inoculated vessels for each treatment were placed in enclosed environmental chambers. To control humidity levels, 2 glass jars (500 mL) containing glycerol-water solutions, appropriate to maintaining the equilibrium relative humidity at the target a w level, were placed in each chamber. The chambers were incubated at 30 and 37 • C for 10 days. To control CO 2 and maintain humidity, specialty certified CO 2 gas cylinders (British Oxygen Company, Jierfude, UK) containing either air, 650 ppm CO 2 or 1000 ppm CO 2 were used for flushing the environmental chambers. The gases were bubbled through a glycerol:water solution of the required a w level before flushing through the chambers and the valves closed. CO 2 flushing was performed every day as described previously [13]. The glycerol-water solutions in the chambers were replaced with fresh solutions every 2 days during the incubation period. Three replicates per treatment were used in all cases. At the end of the incubation period, samples were snap frozen using liquid N 2 and kept at −80 • C until a portion of each sample could be used for RNA extraction and purification, or dried for AFB 1 extraction and clean-up prior to quantification using HPLC analysis.

Aflatoxin Analysis
AFB 1 extraction was performed using AflaStar™ ® -Immunoaffinity Columns (IAC, Romer Labs Inc., St. Louis, MO, USA), following the manufacturer's instructions. Briefly, 5 g of the sample were dried overnight at 80 • C and stored at room temperature. The samples were ground and 4 g was placed into a 50 mL Falcon tube, to which 16 mL of a methanol:water (60:40 v:v) solution was added. The samples were shaken for 1 h at room temperature, and then filtered through qualitative filter paper (QL 110, Fisher Scientific UK Ltd., Loughborough, UK). The extract (1 mL) was diluted in a 15 mL Falcon tube with 9 mL of PBS buffer (0.05 M/0.15 M NaCl, pH 7.4, Fisher Bioreagents ® , Fisher Scientific UK Ltd., Loughborough, UK), and pH was checked with pH strips. The diluted extract was applied to the IAC, and allowed to drip through. After further cleaning, 3 mL of Methanol (HPLC grade) was used to elute the AFs. The eluent was dried and standards were prepared using 200 µL AF (R-Biopharm Rhône Ltd., Darmstadt, Germany) stock solution comprised of 1 ng/uL AFB 1 . The stock solution was pipetted into 2 mL Eppendorf tubes and left to evaporate to dryness overnight inside a fume hood. For quantification of AFs, 200 µL hexane was added to the residue followed by the addition of 50 µL triflouroacetic acid (TFA). The mixture was then vortexed for 30 s and then left for 5 min. Thereafter, a mixture of water:acetonitrile (9:1, v:v) was added and the entire contents of the tube were vortexed for 30 s, after which the mixture was left for 10 min to allow for thorough separation of layers. The hexane layer was discarded and the aqueous layer filtered through nylon syringe filters (13 mm × 0.22 µm; Jaytee Biosciences Ltd., Herne Bay, UK) directly into amber salinized 2 mL HPLC.
A reversed-phase Agilent 1200 series HPLC system with fluorescence detection was used to confirm the identity and quantify AFB 1 . This consisted of an in-line degasser, auto sampler, binary pump and a fluorescence detector (excitation and emission wavelength of 360 and 440 nm, respectively). Separation was achieved through the use of a C 18 column (Agilent Zorbax Eclipse plus C 18 4.6 mm × 150 mm, 3.5 µm particle size; Agilent, Berks, UK) preceded by a guard cartridge with the same packing material. Isocratic elution, with a mobile phase that included methanol:water:acetonitrile (30:60:10, v:v:v), was performed at a flow rate of 1.0 mL/min. The injection volume was 20 µL. A set of standards was injected (1 to 5 ng AFB 1 , B 2 , G 1 and G 2 per injection) and standard curves were generated by plotting the area underneath the peaks against the amounts of AFB 1 standard injected. Linear regression was performed in order to establish a correlation relationship (correlation coefficient, R 2 = 0.99).

Total RNA Extraction
For RNA-seq, tissue harvest was performed after 10 days using three replicates. This time frame was chosen because previous studies with both A. flavus and A. parasiticus suggested that gene expression of many of the biosynthetic genes was optimal after 8-10 days growth, although there does appear to be a sequential expression of groups of AF biosynthetic genes [39][40][41]. Studies on stored maize grain have also shown optimum AFB 1 production at between 5-10 days at 0.98 a w and 10 days at 0.95 a w [42]. We have compromised and used 10 days in this study so that we can obtain molecular information and relevant toxin data. After 10 days of kernel infection (see above), 1 g of frozen milled maize was ground to powder using a mortar and pestle in the presence of liquid nitrogen, and then placed into a 2 mL extraction tube for isolation of total RNA. Total RNA was extracted using the RNAeasy Plant Mini kit (Qiagen, Hilden, Germany). One hundred milligram of the resulting powder was used for isolation of total RNA. The powder was resuspended in 1 mL lysis buffer supplemented with 10 µL β-mercaptoethanol in a 2 mL RNase free micro reaction tube. After vortexing, the tube was quickly frozen in liquid nitrogen. The sample was then thawed on ice. All further procedures were essentially the same as recommended by the manufacturer's protocol.

RNA Sequencing
Library preparation for RNA sequencing was conducted using the NEB Ultra Directional RNA Library Prep Kit. Sequencing was performed on an Illumina HiSeq 2000 instrument. All samples had three biological replicates except the following samples: 30 • C/0.99 a w /350 ppm, 30 • C/0.99 a w /650 ppm and 37 • C/0.99 a w /650 ppm, which had two replicates, and 30 • C/ 0.99 a w /1000 ppm, 37 • C/0.99 a w /1000 ppm and 37 • C/0.99 a w /350 ppm which consisted of one replicate. Untrimmed sequencing reads were mapped to the A. flavus NRRL3357 (assembly JCVI-afl1-v2.0, http://www.ncbi.nlm.nih.gov/genome/360?genome_assembly_id=28730) reference sequence using GSNAP [43,44]. Reads (www.ncbi.nih.gov, accession ID: PRJNA380582) mapping to exons were counted using featureCounts [45] followed by differential expression testing with DESeq2 [46]. Genes were considered differentially expressed if they had an adjusted p-value < 0.05. Gene ontology and KEGG term enrichment was done using the GOSeq R Bioconductor package. The gene co-expression network was made using WGCNA (Weighted Gene Network Co-expression