Effect of hypoxia on transcript and protein levels in batch-fermentation culture
To study the effect of hypoxia on the transcriptome and proteome of Aspergillus fumigatus two wild type strains, CBS144.89 (transcriptomics) and ATCC 46645 (proteomics) were cultured in an oxygen-controlled fermenter with glucose as the sole carbon source. After pre-cultivation under normoxic conditions (21.0% O2) for 14 hours the oxygen partial pressure was shifted to 0.21% O2 and samples were taken at 0, 2, 6, 12, and 24 hours after exposure to hypoxia. Of note, all transcript and protein level data for each time point is relative to time 0. For the first proteomics time point in hypoxia, we elected to wait one hour compared to the transcriptome time point. During cultivation, the pH of the growth medium remained constant and fungal growth occurred under both normoxic and hypoxic conditions (Figure 1A). While growth continued under hypoxia, it is clear that a reduction in growth rate occurred in response to hypoxia (Figure 1A). We elected to allow glucose to be consumed during the fermentation to more closely mimic what likely occurs during in vivo growth in the mammalian lung. However, glucose levels remained high even at the final time point (Figure 1A, consumption of half of the amount of glucose used in the medium), and thus are unlikely to have affected the growth rate in hypoxia. To determine the production of typical fermentation products, the concentration of lactate, ethanol and acetate in the culture supernatant was determined by enzymatic assays. The average D/L-lactate concentration in the supernatant from the culture at the 24-hour time point was determined to be 0.066 mM, and ethanol concentration was 0.049 mM. Acetate concentration did increase over time, and was approximately 0.4 mM at the end of the experiment (Figure 1B). Based on these data, we expect that the majority of transcript and protein level changes are due to the hypoxia response. However, we cannot rule out that some observed changes might be carbon-source mediated.
Assessing the transcriptome across all time points, significance analysis of microarray (SAM) revealed that approximately 800 transcripts, or 10% of the genes on the microarray, were significantly altered in response to hypoxia (FDR < 0.05, Additional file 1). Considering all transcripts represented in the microarray, gene set enrichment analysis (GSEA) was performed to gain a general understanding of functional categories affected by hypoxia (Additional File 2). GSEA results showed that transcripts from steroid (ergosterol) biosynthesis and other metabolic processes were significantly increased during the course of the experiment. Specific categories of transcripts that were significantly increased at two hours were associated with histone deacetylases, fungal transcription factors and major facilitator superfamily genes (p < 0.01, Additional file 2). At six hours, similar transcripts were represented, but the histone deacetylases were no longer significant. Thus, it is intriguing to speculate that the early adaptation to hypoxia requires changes in chromatin structure and/or post-translational modification of key regulatory proteins. At 12 hours in hypoxia, transcriptional regulation and major facilitator superfamily transcripts continued to be increased. After 24 hours in hypoxia, few categories were significantly increased, however transcripts related to sterol related processes were highly significant. Using a self-organizing map algorithm (SOMA) to cluster genes, 64 clusters grouped the majority of steroid genes into a single cluster, highlighted in Figure 2. Thus, a major part of the A. fumigatus early hypoxia response is characterized by increases in ergosterol biosynthesis associated transcripts. This result is consistent with previous observations regarding the A. fumigatus transcriptional regulator, SrbA, which is required for growth in hypoxia and directly regulates genes in the ergosterol biosynthesis pathway [10, 27].
GSEA also identified categories that are significantly reduced in response to hypoxia, and include ribosome biosynthesis, proteasome activity, pyrimidine and purine metabolism, and oxidative phosphorylation across all time points. Additionally, gene groupings associated with heat shock proteins, RNA recognition motifs, and intracellular trafficking and secretion factors were also reduced. There were fewer differences among the reduced transcripts than the increased transcripts across all time points, and more categories and domains were significant (Additional file 2). Thus, a large part of the mRNA response to hypoxia in A. fumigatus involves a reduction in transcript levels of genes associated with core cellular processes.
With regard to the proteomic response of A. fumigatus to hypoxia, we analyzed the mycelial proteome of A. fumigatus wild type strain ATCC 46645 after 0 (21% pO2), 3, 6, 12 and 24 h growth under hypoxic conditions (0.21% pO2). The use of ATCC46645 allowed us to compare our results with a previous study on the proteomic changes during long-term adaptation to low oxygen levels using chemostat culture [25]. To obtain a high spatial resolution of protein spots we analyzed the intracellular proteins by two dimensional gel electrophoresis using two different immobilized pH gradient strips covering an acidic (pH 3-7) and basic pH range (pH 7-11), respectively. In total 86 differentially regulated proteins could be identified by comparing gels from different time points of hypoxic cultivation (Figure 3, Additional file 3). Under hypoxia, 52 different proteins showed an increase in abundance, whereas 34 proteins showed a decrease. Changes in protein levels occurred as early as 3 hours of hypoxic cultivation, indicating that the cellular response to hypoxia is rapid.
Looking at different primary metabolic pathways and performing protein set enrichment analysis with KEGG categories, growth under hypoxia resulted in an increased level of proteins involved in glycolysis, ethanol fermentation, electron transport, alanine, aspartate, glutamate metabolism and the oxidative stress response. In contrast, the abundance of several enzymes in the TCA cycle, the pentose phosphate shunt, and cysteine/methionine metabolism decreased during hypoxic growth conditions. Furthermore, a decreased level of proteins involved in ribosome biogenesis, sulfate assimilation and purine metabolism was also observed.
Correlation of Proteomics and Gene Expression
Proteomic analysis was performed under similar conditions as the transcriptomic analysis, however with a different wild type strain and one different early time point was analyzed. Despite the difference in wild-type strains, a correlation was observed between the proteomic and transcriptomic data. 86 protein spots showed significant changes in abundance among all experiments. Of the 86 proteins, 85 were identified in the transcriptomics data. Statistical analysis of the correlation between protein and microarray experiments shows a consistent trend of correlation among the datasets (Figure 4). Linear regressions were forced with X and Y-intercepts at zero, and varied between 0.18 and 0.25, with p < 0.05 for all correlations. A heat map summarizes the differences and similarities observed between protein and transcript data for selected pathways (Figure 5). No protein data was obtained for the terpenoid or steroid biosynthesis pathways. Additionally, glycolysis was generally not significantly over represented among either increased or decreased transcripts, however it was significant for the protein dataset. The TCA cycle and glyoxylate and dicarboxylate metabolism were decreased significantly in the transcript data, whereas both categories were significantly positively and negatively changed for the protein data. Only ribosome biogenesis was significantly down for both the transcript and protein data.
Of note, in the correlation of proteomics and transcriptomics data (Figure 4), we observed very few points located in the top-left quadrant, and therefore very few instances occurred where mRNA levels increased and protein levels decreased. Two examples of this are represented by transcripts that were increased for the entire time course, yet the proteins showed decreased levels: a nuclear segregation protein Brf1 (Afu1g14120) and a hypothetical protein (Afu7g00350) (marked with # in Figure 6). In general, if the mRNA abundance was enhanced, the protein levels were also high. However, there were exceptions with genes that had reduced mRNA levels and enhanced protein abundance. Comparing the fold change values for proteins with the microarray data, 25 of the genes show different patterns of transcript and protein levels (Additional file 4). Two functional categories, glycolysis and amino acid metabolism, are noted on the heat map of comparison (Figure 6). Several of the transcripts in these groups are decreased in abundance, while the protein levels are increased. Three additional pairs are noted with an asterisk, a 40S ribosomal protein (Afu6g12660), the Asp-hemolysin (Afu3g00590) and the proteasome component Pre2 (Afu6g08310). It is possible that enhanced protein levels could be achieved through mRNA stabilization and differences in turnover rates, and it seems clear that in addition to the apparent transcript regulation of the hypoxia response, other post-transcriptional and post-translational regulatory mechanisms are active. Further identification of these regulatory mechanisms could identify new targets for gene replacement studies to determine whether these mechanisms are critical for in vivo growth of the fungus during hypoxia adaptation and pathogenesis.
Comparison of the proteomic response of A. fumigatus to short- and long-term incubation under hypoxic conditions
In a previous study we analyzed the proteome of A. fumigatus cultivated in an oxygen-controlled, glucose-limited chemostat under normoxic (21% pO2) and hypoxic (0.2% pO2) growth conditions. In contrast to the experimental set-up chosen for the work described here, the fungus was cultivated under conditions of glucose depletion for a long time period (10 days), and the medium was exchanged continuously. Only one third of the differentially regulated proteins found in our study were also found to be differentially expressed in chemostat cultures [25]. Under both conditions, exposure of A. fumigatus to hypoxia caused an increase in the abundance of glycolytic enzymes and the NO-detoxifying flavohemoprotein Afu4g03410, whereas fatty acid metabolism associated protein levels were reduced. In contrast, the level of proteins associated with the pentose phosphate pathway and the citric acid cycle decreased during short-term incubation at low oxygen-levels, but increased after long-term exposure. The short-term response to hypoxia was also characterized by the activation of ethanol fermentation (see below), which could not be observed after cultivation of A. fumigatus in a chemostat. Furthermore, in comparison to the previous report the pseurotin A cluster was not significantly induced during the short-term exposure to hypoxia. It is interesting to speculate that either the formation of pseurotin A is only activated after longer periods of hypoxia, or glucose depletion may be a factor for derepression of the pseurotin A biosynthesis gene cluster [25].
Glycolysis and Fermentation
Previous work with A. nidulans and C. albicans suggested that transcripts encoding enzymes of glycolysis are strongly induced in response to hypoxia [11, 16, 17, 23]. With regard to A. fumigatus, the majority of glycolytic transcripts in the microarray were unaffected by hypoxia which is similar to observations in Cryptococcus neoformans (Figure 7A) [14]. However, increases in glycolytic protein levels were observed from the proteomics data. Taken together, these data suggest that the primary regulation of glycolysis in response to hypoxia in A. fumigatus may occur post-transcriptionally. Intriguingly, the glycolytic transcript most induced by hypoxia in A. fumigatus was the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) Afu5g01030, and this was verified with RT-PCR (Figure 7B). This enzyme is one of three GAPDHs in the genome of A. fumigatus that likely catalyzes the 6th step in glycolysis. The other two GAPDHs, Afu5g01970 and Afu8g02560, were slightly negatively altered in transcript abundance (Additional file 5). The increase in mRNA expression of Afu5g01030 was further validated by proteomics, where this was one of the most abundant proteins in hypoxia (Additional file 3). It is likely important that this glycolytic gene appears to be regulated at the level of transcription unlike other glycolytic genes in A. fumigatus [28]. In the genome of S. cerevisiae, three GAPDH genes also are found, and these genes show different growth-phase dependent regulation. Most interestingly, the corresponding proteins are not only located in the cytosol but also present in the cell wall. Thus, it has been speculated that they mediate cell adhesion [29]. Whether additional roles for GAPDHs in fungal biology and pathogenesis exist remain to be determined.
Also related to glycolysis, both transcripts that are required for the conversion of pyruvate to ethanol are induced in response to hypoxia: pdcA (Afu3g11070) encoding a pyruvate decarboxylase, is one of the most highly induced transcripts, along with two alcohol dehydrogenases, alcC (Afu5g06240) and adh2 (Afu2g10960). PdcA and AlcC were both recently confirmed to be required for ethanol fermentation in A. fumigatus, and AlcC was demonstrated to have an important but undefined role in fungal pathogenesis [7]. Consistent with the transcript level results, the protein level of pyruvate decarboxylase PdcA was increased in hypoxia (Additional File 5). In addition, the observed increase in ethanol fermentation is in contrast to the previous proteomic analysis of the A. fumigatus hypoxic response, which did not find evidence of an NAD+ regenerating system [25]. This is likely explained by the different growth conditions used between the two studies. However, our results here with the short term hypoxia response are consistent with recent findings in a murine model of IPA that suggest ethanol fermentation is part of in vivo growth mechanisms of A. fumigatus [7]. In addition, in A. nidulans, an ethanol fermentation response to hypoxia was also observed [23, 24]. However, loss of ethanol fermentation genes in A. fumigatus did not dramatically affect the in vitro growth rate of the fungus in hypoxic conditions. Thus, other mechanisms of NAD+ regeneration are likely present in A. fumigatus and it is unclear what role ethanol fermentation plays in the hypoxia response. One possible alternative fermentation mechanism is lactate fermentation as has been observed in A. nidulans. In support of this hypothesis, a mitochondrial lactate dehydrogenase transcript (Afu1g00510) was transcriptionally increased in response to hypoxia. However, the determined lactate concentration in the culture supernatant was only in the micromolar range (see Figure 1B). Thus, the role of fermentation and NAD+ regeneration in the A. fumigatus hypoxia response awaits further mechanistic characterization.
In contrast to transcript levels, high protein levels of AlcA (Afu7g01010) suggest a critical role for this protein in the hypoxia response, as it was one of the highest induced proteins (Additional files 3 and 4). In A. nidulans, AlcA is involved in utilization of ethanol as a carbon source [30] and in accordance with this, the level of the AlcA protein highly increased during growth of A. fumigatus on ethanol as shown by a 2D gel electrophoresis study [31]. Thus, it is possible that as ethanol production increases in response to hypoxia, A. fumigatus is then able to utilize the produced ethanol as a carbon source, which may have potential implications for fungal pathogenesis. Alternatively, as these experiments were conducted in a glucose rich environment, the role of ethanol fermentation and/or utilization in the hypoxia response may be overstated and depend on the availability of fermentable substrates.
To further support the observed changes in transcript levels from the microarray data, we chose six transcripts involved in glycolysis for additional analyses using real-time RT-PCR (Figure 7C). Glycolysis starts with the conversion of D-glucose to D-glucose 6-phosphate by hexokinase or glucokinase. There are three hexokinases annotated in the KEGG database (Afu2g00450, Afu2g05910, and Afu6g03980) in A. fumigatus, and these tended to be transcriptionally reduced or unchanged in response to hypoxia, whereas the two glucokinases (Afu6g02230 and Afu2g16330) were induced at later time points when glucose levels began to drop. Thus, it is unclear whether these glucose responsive transcripts were altered due to hypoxia or changes in the glucose level in the culture medium at later time-points. We chose to look at Afu6g02230 (glkA) and Afu2g05910 (hxkA) as these are thought to be the main active enzymes in A. fumigatus for this part of glycolysis [32]. As predicted by the microarray data, the hxkA transcript was repressed and the glkA transcript was enhanced. This regulation may be explained either by a stress-induced expression of hxkA [33] or by an increase in transcript levels of the high-affinity sugar kinase glkA when the glucose concentration decreases.
Afu4g00960 transcript, 6-phosphofructokinase, one of the early steps in converting glucose 6-phosphate to glyceraldehyde 3-phosphate was slightly reduced in the microarray, and relatively unchanged in PCR. Afu1g10350, phosphoglycerate kinase (pgkA) at the midpoint of the glycolytic pathway was not significantly changed in the microarray. However, RT-PCR data suggest that this transcript is increased in response to hypoxia, which is consistent with findings in mammals for this transcript [34]. Proteomic data confirms the PCR values and shows that PgkA protein levels were also enhanced in response to hypoxia (Additional files 3 and 4). Afu6g07430 transcript, a putative pyruvate kinase, is the final step in the glycolytic pathway and transcript levels were reduced in the microarray and verified with RT-PCR (Figure 7C). However, changes in protein levels of this enzyme were not detected.
Another gene associated with glycolysis investigated with qRT-PCR was Afu6g11430 (aldA), aldehyde dehydrogenase A, which is necessary for conversion of acetaldehyde to acetate for central carbohydrate and lipid metabolism. Microarray results suggest that this transcript was reduced in response to hypoxia and RT-PCR results confirm this observation (Figure 7C). However, there are four additional aldehyde dehydrogenase encoding genes in A. fumigatus, Afu2g00720 and Afu7g01000 (NAD+) and Afu4g08600 and Afu4g13500 (NAD(P)+), so it is possible that these enzymes, and not AldA, are utilized under hypoxic conditions to metabolize ethanol. Afu4g08600 does show a gradual increase in transcript by 24 hours in hypoxia when ethanol levels are assumed to be higher, whereas Afu7g01000 and Afu4g13500 transcripts are reduced, although not to the same degree as aldA (Additional file 5). A recent proteomics study indicated that Afu7g01000 is most probably the major aldehyde dehydrogenase involved in the metabolism of ethanol, as it was highly expressed in A. fumigatus during growth on ethanol [31]. Afu2g00720 was not on the microarray. Taken together, these results suggest that hypoxia plays a small role in altering the transcription of glycolysis encoding enzymes, however, the increase in protein levels of several enzymes coupled with the apparent increase in fermentation suggest that glycolytic activity is likely increased in response to hypoxia in A. fumigatus and regulation may occur at posttranscriptional levels. This observation confirms a previous study, which showed post-transcriptional regulation of glycolysis in anaerobic S. cerevisiae cultures [35].
Of critical relevance to fungal pathogenesis, cell wall biosynthesis transcripts are affected by hypoxia (Figure 7D) [36]. Specifically, the 1,3-β-glucan synthase fksA (Afu6g12400) and the 1,3-α-glucan synthase agsA (Afu3g00910) are increased along with transcripts for glucanosyltransferases gelC (Afu2g12850) and gelA (Afu2g01170). It is also noted that chitin synthases are altered in expression. Specifically, chsE (Afu2g13440) and chsF (Afu8g056300) transcripts are increased, while the remainder of chitin synthase transcripts are unchanged or reduced. Many cell wall polymers require precursors from the glycolysis pathway and thus alterations in glycolytic flux in response to hypoxia may alter available cell wall precursor levels. Taken together, these results suggest that cell wall composition changes occur in response to hypoxia, which has important ramifications for the effect of cell wall-targeting antifungals, host immune system recognition of A. fumigatus, and the development of the inflammatory response to the invading fungus.
Ergosterol Biosynthesis
Ergosterol is the fungal equivalent of mammalian cholesterol and in addition to being a critical component of plasma membranes, is also a target of the most commonly used antifungal drug against A. fumigatus, the triazole voriconazole. Importantly, previous research has identified sterols as an oxygen sensing system in fungi including Saccharomyces cerevisiae and Schizosaccharomyces pombe [37, 38]. This is likely due to the large oxygen requirement for sterol biosynthesis. Previous transcript profiling experiments of hypoxia responses in the human fungal pathogens C. neoformans and C. albicans identified the ergosterol biosynthesis pathway as one of the most affected by hypoxia [14, 16]. In addition, we previously observed a significant reduction in several ergosterol biosynthesis enzyme-encoding transcripts in response to hypoxia in the A. fumigatus SREBP null mutant (SrbA, Afu2g01260) [10]. Thus, it is not surprising that one of the most affected metabolic pathways in our transcript profiling experiments in this study was steroid biosynthesis. To fully explore the observed significance of steroid biosynthesis genes, we compiled a complete list of ergosterol genes from previously published work that includes the entire pathway from acetyl-CoA to ergosterol, including terpenoid backbone biosynthesis [39]. Transcripts of ergosterol biosynthesis enzymes previously shown to require oxygen were highly induced in response to hypoxia and largely clustered together, particularly the C-14 sterol reductases erg24A (Afu1g03150) and erg24B (Afu1g05720), and the C-4 methyl sterol oxidases erg25A (Afu8g02440) and erg25B (Afu4g04820), which are all enzymes needing three molecules of oxygen for full function (Figure 8A) [40]. Unfortunately, another oxygen dependent enzyme, the target of the triazole drugs, Erg11A (Cyp51A, Afu4g06890) was not on the microarray. However, previous work has established that erg11A transcript is generally increased in response to hypoxia [41].
In addition to these, several other ergosterol biosynthesis genes are duplicated in the A. fumigatus genome. The three copies of sterol desaturases, erg3A (Afu6g05140), erg3B (Afu2g00320) and erg3C (Afu8g01070), convert episterol to 5,7,24(28)-ergostatrienol, and require oxygen and NADPH [42]. Erg3A and erg3B are significantly increased at all time points, whereas the erg3C transcript is slightly decreased. The two copies of C-24(28) sterol reductase genes, erg4A (Afu5g14350) and erg4B (Afu1g07140), are the last step of ergosterol biosynthesis, and both show decreased transcript levels in response to hypoxia. Three copies of the lanosterol cyclase, erg7A (Afu5g04080), erg7B (Afu4g012040) and erg7C (Afu4g14770), are involved in the transition from epoxysqualene to lanosterol. All transcripts are increased, and most significantly at 6 and 12 hours. Erg7A and erg7B are slightly decreased in expression at 24 hours. The two copies of acetyl-CoA-acetyltransferase, erg10A (Afu6g14200) and erg10B (Afu8g04000) are relatively unchanged in expression over the 24-hour period of hypoxia exposure. These enzymes are at the earliest stage of the terpenoid biosynthesis pathway, converting 2 acetyl-CoA to acetoacetyl-CoA. This important pathway is upstream of the ergosterol pathway and condensation of terpenoid building blocks (five-carbon isoprene units) leads to precursors for sterol biosynthesis. Next, erg13A (Afu8g07210) and erg13B (Afu3g10660) are significantly decreased, with erg13B more so than erg13A. This is the next step in the terpenoid biosynthesis pathway, converting acetoacetyl-CoA to 3-hydroxy-3-methyl-glutaryl-CoA. Finally, the next step in the terpenoid pathway converts 3-hydroxy-3-methyl-glutaryl-CoA to mevalonate. HMG-CoA reductase (HMGR) is the rate-limiting step in ergosterol biosynthesis and both hmgA (Afu2g03700) and hmgB (Afu1g11230) have reductions in transcript levels in response to hypoxia. Generally, we observe that the duplicated genes in the ergosterol pathway share similar, although not identical, patterns of expression in response to hypoxia.
An intriguing question is the mechanism behind the increase of specific ergosterol biosynthesis transcripts in response to hypoxia. One mechanism is thought to be through direct transcriptional regulation by the fungal SREBP transcription factor, which has been shown to be a key transcriptional regulator of ergosterol biosynthesis [10, 13, 14, 27, 43]. For S. pombe, activation of the SREBP Sre1 by proteolytic cleavage has been linked to sensing of total ergosterol levels in the cell [44]. As sterol biosynthesis decreases upon reduction of available oxygen, the SREBP pathway senses the drop in total ergosterol levels and is capable of restoring sterol homeostasis through induction of sterol biosynthesis genes [44]. However, another mechanism may be suggested in our transcript profiling data, which illustrates a reduction in transcripts associated with terpenoid biosynthesis, which lies upstream of ergosterol production.
As previously stated, several transcripts in the terpenoid (isoprenoid) backbone biosynthesis pathway are reduced in response to hypoxia (Figure 8B). All three initial steps in the terpenoid/mevalonate biosynthesis pathway (erg10, erg13 and hmg) have reductions in transcript levels in response to hypoxia (Figure 8B). Thus, a decrease in needed precursors, such as mevalonate, for sterol biosynthesis may also stimulate an increase in transcript levels of enzymes further down the pathway in response to the growing reduction in important precursor levels. The terpenoid (isoprenoid) biosynthesis pathway also leads to N-glycan and ubiquinone and other terpenoid-quinone biosynthesis pathways in A. fumigatus. Moreover, mevalonate was shown to be a key intermediate in the biosynthesis of siderophores in A. fumigatus under iron-limited conditions [45]. Transcripts in the N-glycan biosynthesis pathway remain mostly reduced in hypoxia, with the exception of three genes: Afu2g14630, cell wall glycosyl hydrolase; Afu6g04210, mannosyl-oligosaccharide glucosidase; and Afu6g09770, geranylgeranyl diphosphate synthase (Additional files 2 and 5). Transcripts associated with ubiquinone biosynthesis are also slightly reduced, but mostly remain unchanged (Additional file 2). Thus, only transcripts in the ergosterol biosynthesis pathway are significantly increased downstream of the terpenoid biosynthesis pathway. A direct regulatory link between terpenoid and ergosterol biosynthesis remains unclear. Most likely, HmgR plays a significant role in coordinately altering the biosynthetic activity of these pathways [46].
Iron acquisition and siderophore biosynthesis
Taking only transcripts that were most significantly changed in abundance using SAM (Additional file 1), analysis with FungiFun https://sbi.hki-jena.de/FungiFun/FungiFun.cgi showed additional significant categories. In addition to the previous pathways identified by GSEA, oxidation-reduction and iron related transcripts are significantly increased in response to hypoxia (Additional file 6). Importantly, iron is a required cofactor for many of the transcripts coding for enzymes associated with ergosterol biosynthesis steps that were increased in response to hypoxia (erg 25 and erg 11). A literature search provides a list of iron-related transcripts to investigate the microarray results [47–49]. The majority of transcripts associated with iron acquisition and biosynthesis are somewhat increased in response to hypoxia (Figure 9). However, transcripts related to siderophore biosynthesis (i.e. sidA, Afu2g07680) and iron uptake (i.e. sit1, Afu7g06060) were slightly transcriptionally reduced. SreA (Afu5g11260), a previously identified negative regulator of iron homeostasis, showed increased transcript levels at 2, 6, and 12 hours, so that the iron-limitation response was apparently not induced under the tested conditions. Importantly, we must distinguish A. fumigatus SreA from SrbA the sterol regulatory element binding protein in this mold (SREBPs in yeast have been named Sre1). There is also an apparent decrease in terpenoid intermediates in response to hypoxia which may negatively affect siderophore biosynthesis [45]. A further discussion on the implications in changes in iron homeostasis mechanisms is below in the transcription factor analysis. In general, these results are consistent with a requirement for iron in the hypoxia response that has been observed in mammalian systems and was shown for A. fumigatus when it was cultivated in a glucose-limited chemostat at low oxygen levels [21, 25]. Moreover, in C. albicans an increase in iron uptake transcripts was also observed upon exposure to hypoxia [11, 17]. Recently, we observed that SrbA also regulates iron uptake and homeostasis in iron replete conditions under hypoxia and in low iron conditions, further suggesting an important link between iron uptake and the A. fumigatus hypoxia response that remains to be further explored [27].
Tricarboxylic acid cycle, GABA Shunt, and Respiration
For many organisms, exposure to hypoxia results in decreased flux through the TCA cycle concomitant with a decrease in aerobic respiration [16, 43]. Correspondingly, most transcripts associated with oxidative phosphorylation complexes were decreased; although cytochrome C oxidase associated transcripts were overall increased (Figure 10A). Additionally, many transcripts associated with the TCA cycle were reduced in response to hypoxia at the time points we examined in our experiments (Figure 10B). In contrast, several transcripts in the TCA cycle are transcriptionally increased in response hypoxia in A. nidulans [23]. In addition, the industrial mold Aspergillus niger has been observed to increase production of TCA cycle intermediates during hypoxic cultivation [50]. However, A. fumigatus TCA cycle gene transcripts were largely reduced in response to hypoxia, although the conditions of the experiments were not identical to the previous Aspergillus experiments (Figure 10B, Additional file 5). For example, transcripts for fumarate reductase OsmA (Afu8g05530), phosphoenolpyruvate carboxykinase AcuF (Afu6g07720), and putative malate dehydrogenase homologs (Afu2g13800 and Afu7g05740) were reduced in response to hypoxia. These transcripts are key factors in the reductive branch of the TCA cycle, which in some organisms such as Mycobacterium tuberculosis are important for the reoxidation of intracellular NADH during hypoxic growth [51]. In further support of these data, putative succinate dehydrogenase transcripts sdh1 (Afu3G07810) and sdh2 (Afu5g10370) were also transcriptionally reduced in response to hypoxia (Additional File 4). Thus, hypoxic responses even amongst the genus Aspergillus are divergent, and it appears that A. fumigatus does not use the reductive branch of the TCA cycle to adapt to the hypoxia culture conditions utilized in our study.
As observed with A. nidulans, there is some evidence of the GABA shunt being utilized by A. fumigatus in response to hypoxia (Figure 10B, C). The GABA shunt is hypothesized to help organisms avoid the accumulation of high NADH levels in the absence of a terminal electron acceptor such as oxygen and also contributes to glutamate formation. Of the six A. nidulans genes identified involved in the GABA shunt [23], there are five potential homologues in A. fumigatus: Afu4g06620 and Afu2g06000 (glutamate dehydrogenase), Afu6g13490 (glutamate decarboxylase), Afu5g6680 (aminobutyrate transaminase, gatA) and Afu3g07150 (succinate-semialdehyde dehydrogenase). Of these five, four were on the A. fumigatus microarray, missing Afu2g06000. Transcripts for two of these genes were increased in response to hypoxia: glutamate decarboxylase Afu6g13490 and glutamate dehydrogenase Afu4g06620 (Figure 10B, C). In the proteomic data, Afu4g06620 protein levels are increased after 24 hours in hypoxia. However, this pathway has not been specifically studied in A. fumigatus, so it is possible that more distant homologues may function in this pathway. Moreover, with the reduction in TCA cycle transcripts, it is unclear what the available levels of 2-oxoglutarate would be to allow flux through the GABA shunt in A. fumigatus. Investigating the KEGG defined glutamate pathway shows that glutamate decarboxylase (Afu6g13490) is also associated with glutamate biosynthesis, and a great number of transcripts do show increased levels in this pathway in response to hypoxia (Figure 10C). Glutamate decarboxylases have a wide range of functions depending on the organism. For example, in S. cerevisiae, GAD1 is critical for normal tolerance to oxidative stress [52]. Thus, the increased levels of transcripts associated with the GABA shunt and glutamate biosynthesis in A. fumigatus may be a response to the changing redox status of the cell in response to hypoxia.
Novel genes showing positive regulation
Novel and interesting transcripts were increased significantly in response to hypoxia in the microarray analysis, and not associated with a KEGG pathway. Some of these transcripts were verified with RT-PCR. For example, Afu3g14170 (hxtA), a major facilitator superfamily glucose transporter and putative high-affinity hexose transporter, was up at all time points (Figure 11A). The homolog in A. nidulans is associated with nutrient starvation and sexual reproduction [53]. The homolog in S. cerevisiae senses low glucose, mannose and fructose, and induces hexose transporters [54]. The trend of increasing expression continues over time, consistent with reduction in glucose in the growth media and the increased mRNA levels of the glucokinase glkA, which is characterized by a higher affinity to glucose in comparison to the hexokinase HxtA.
Afu2g09590 (udpA), an UDP-N-acetylglucosamine 1-carboxyvinyltransferase family member, was strongly increased in hypoxia, and is a protein generally associated with glycan biosynthesis in bacteria, where this gene is commonly named MurA. This gene is associated with the first step in peptidoglycan biosynthesis in bacteria, and is a target for the antibacterial drug fosfomycin [55]. However, a database query http://aspgd.broadinstitute.org/cgi-bin/asp2_v3/shared/show_protein_cluster.cgi?site=asp2_v3&id=960184 shows a similar gene in several Aspergillus species syntenic within the inspected region. The function of this protein in fungal species is unknown.
Afu3g11590 (atg11) transcript was increased at all time points. The gene is homologous to autophagy related protein 11 (ATG11) in S. cerevisiae, and has been shown to be involved in trafficking of autophagosomes and cytoplasm to vacuole (Cvt) vesicles [56]. ATG genes are conserved among eukaryotes, however the function of the Atg11 protein appears to be fungal specific and depends on the interaction with other proteins in the autophagy pathway [57]. Autophagy is stress induced, and it is thought that it can function to recycle cellular components that are no longer necessary into pathways that are critical for survival. In this way, autophagy helps to reduce the energy costs for significant changes in cell physiology [58]. Thus, it is tempting to speculate that autophagy is one mechanism by which A. fumigatus adapts to hypoxic stress.
Afu5g00900 (rgsA), a G-protein signal regulator with an RGS domain, is predicted to be involved in attenuation of G-protein signaling through activation of the intrinsic GTPases of G-proteins [59]. This gene was one of the most highly induced genes in the microarray analysis. Although this gene has not been investigated in A. fumigatus, it is likely that this gene allows for the rapid response of the cell to environmental changes. In A. nidulans, the homologue of this gene (rgsA) negatively regulates the G protein α subunit GanB, which is involved in the activation of various stress responses and the inhibition of asexual conidiation. During hypoxia RgsA is putatively required to down-regulate these energy-consuming stress responses [60]. The GanB ortholog GpaB in A. fumigatus was shown to regulate the virulence gene pksP and loss of GpaB function caused an increased susceptibility to killing by macrophages [61]. In mammalian solid tumors, Rgs2 is associated with hypoxia and pro-tumor functions [18]. Rgs2 appears to be a critical regulator of the pro-angiogenic function of myeloid derived suppressor cells, and deletion of Rgs2 in a murine model significantly reduced tumor growth. Thus, Rgs2 is an intriguing candidate for future study of its role in fungal hypoxia adaptation and pathogenesis.
Novel genes whose transcripts decrease in response to hypoxia
Analysing only genes in the SAM dataset (Additional file 1) with FungiFun, functional categories with transcripts largely reduced in abundance are similar to the GSEA analysis and include ribosome and RNA processing and the TCA cycle (Additional file 7). An additional selection of transcripts whose levels were significantly reduced in hypoxia was also validated with RT-PCR. First, Afu5g12510 (afeA), an adenylate-forming enzyme, was down in the 2 and 6 hour time points and up at 12 and 24 hours in the microarray experiment, and this was confirmed by PCR (Figure 11B). Adenylating enzymes generally activate otherwise unreactive carboxylic acids and can be used to form a wide array of natural products [62]. This gene is a member of the acyl-CoA synthetases (AMP-forming)/AMP-acid ligases family, and potentially involved in lipid metabolism (COG ontology). It is also identified in PANTHER as an ATP-dependent AMP-binding enzyme family member, with a 4-coumerate CoA ligase sub-family designation. Among the most highly reduced proteins, the level of the acetate activating acetyl-coenzyme-A synthetase FacA decreased most significantly in response to hypoxia. This finding suggests an oxygen-dependent regulation of the acetate-activating enzyme FacA that has also been reported for the yeast orthologue Acs1p [63]. Taken together, these results suggest that fatty acid metabolism may be reduced in A. fumigatus in response to hypoxia, contrary to results in other organisms such as the pathogenic yeast C. neoformans [14].
Next, Afu6g05160 (azf1) is a zinc-finger C2H2-type transcription factor, which may be involved in the regulation of cell cycle, the G2/M transition, and is activated in non-fermenting conditions in S. cerevisiae [64]. Consistent with this, the transcript level was reduced in microarrays after 2 hours in hypoxia, and PCR results confirmed this observation (Figure 11B). Additionally, this transcription factor is a positive regulator of the G1 cyclin CLN3 in S. cerevisiae in response to glucose [65]. CLN3 is involved in the progression of the cell cycle, and activates Cdc28p kinase which promotes the G1 to S phase transition [66]. In general, the transcriptomics and proteomics data along with the actual growth curve of the fungus in hypoxia all suggest a reduction in cellular metabolism, which likely would correspond to changes in the regulation of cell cycle length.
Finally, transcript levels of Afu1g03210 (flbD), homologous to a Myb family conidiophore development gene in A. nidulans [67–69], were reduced at all time points, and this was verified by PCR (Figure 11B). Therefore, it is likely that in response to hypoxia, asexual reproduction is inhibited. Consistent with this hypothesis, the target of FlbD, BrlA (Afu1g16590), is also repressed in expression in the microarrays at all time points. A bZIP protein, FlbB (Afu2g14680), which acts upstream of FlbD, is also repressed in expression, almost 2-fold by 24-hour time point (Additional file 5) [70]. Interestingly, the FlbB locus is also associated with gliotoxin production. Moreover, it is well known that in vivo during invasive aspergillosis that asexual reproduction generally does not occur with A. fumigatus. Thus, the in vivo hypoxic microenvironment might repress the transcriptional program needed for asexual development in response to the host environment.
Transcriptome analysis of transcription factors
Several significant functional categories of positively affected transcripts were associated with transcription factors (Additional file 2). To investigate this more completely, a list of 430 genes annotated as transcription factors was compiled from the genome of A. fumigatus. Of those, 414 were present on the array (TF tab, Additional file 5). The majority of these genes have not been directly investigated in A. fumigatus. Self-organizing tree algorithm (SOTA) analysis [71, 72] reveals distinct patterns of changes in transcript abundance over the course of the experiment (Figure 12). Compared to other classes of genes in our analyses, transcription factors are largely positively regulated in response to hypoxia. As expected, transcript levels of a transcriptional regulator of hypoxia adaptation in A. fumigatus, SrbA (Afu2g01260), increased over the course of the experiment. This correlates with the observed increase in ergosterol biosynthesis gene transcripts that are likely directly regulated by SrbA [10, 27, 41]. The cluster associated with the SrbA transcript contains 63 transcription factors that mirror the expression pattern of a gene known to be critical for adaptation to hypoxia and virulence, thus these transcription factors may be exciting targets for gene replacement studies to determine their role in hypoxia adaptation and virulence (Figure 13).
Transcription factors involved in regulation of iron homeostasis were also responsive to hypoxic cultivation. The GATA type transcription factor SreA (Afu5g11260) has increased transcript levels at 2, 6, and 12 hours hypoxia, while HapX (Afu5g03920) transcripts are correspondingly slightly reduced. This is consistent with SreA's role as a repressor of HapX [73, 74]. The transcript profile of these two key iron regulators suggests that HapX may be slowly released from SreA repression as the exposure to hypoxia increases. This would be consistent with hypoxic cells initially being iron replete, but as the iron supplies become utilized by key pathways required for hypoxia adaptation (ergosterol biosynthesis, respiration), SreA repression is alleviated and transcripts critical for iron acquisition, including HapX, become more abundant. Also consistent with the observed SreA and HapX transcript profiles, AcuM (Afu2g12330) transcript levels were reduced at 2 and 6 hours growth in hypoxia and slightly increased at 12 and 24 hours. AcuM, a transcriptional regulator of gluconeogenic genes, has recently been suggested to repress SreA and thereby induce HapX transcripts via transcriptional profiling experiments [75]. It is unclear, however, whether acuM transcript levels were affected by changes in glucose levels between early and late time points or hypoxia itself. Importantly, the transcript profile of all the iron acquisition genes does not exactly mimic the expected profile based on what is known about the SreA and HapX regulons. Thus, it seems apparent that additional regulators of iron acquisition and homeostasis are present in A. fumigatus and operative under hypoxia.