Reprogramming of carbon metabolism by the transcriptional activators AcuK and AcuM in Aspergillus nidulans



The ability of fungi to use carbon sources metabolized via the TCA cycle requires gluconeogenesis. In Aspergillus nidulans the AcuK and AcuM transcription factors regulate the expression of the gluconeogenic genes acuF, encoding phosphoenolpyruvate carboxykinase, and acuG, encoding fructose-1,6-bisphosphatase. Expressed proteins containing the AcuK/AcuM N-terminal DNA-binding domains bind together in vitro to motifs containing repeats of CGG separated by seven bases (CCGN7CCG) and the functionality of these sequences was verified in vivo by acuF–lacZ reporter studies. Chromatin immunoprecipitation analysis showed inter-dependent DNA binding of the proteins to the promoters of gluconeogenic genes in vivo independent of the carbon source. Deletion of the mdhC gene encoding a cytoplasmic/peroxisomal malate dehydrogenase showed that this activity is not essential for gluconeogenesis and indicated that induction of AcuK/AcuM regulated genes might result from malate accumulation. Deletion of the gene for the alternative oxidase did not affect growth on gluconeogenic carbon sources; however, expression was absolutely dependent on AcuK and AcuM. Orthologues of AcuK and AcuM, are present in a wide range of fungal taxa and the CCGN7CCG motif is present in the 5′ of many genes involved in gluconeogenesis indicating a fundamental role for these transcription factors in reprogramming fungal carbon metabolism.


Annotation of fungal genomes has revealed the presence of hundreds of genes encoding transcription factors. A major challenge therefore is to determine their function. In Aspergillus nidulans, proteins containing the fungal-specific Cys6 Zn binuclear cluster (Zn cluster) family are the major class (more than 250) and the function of only a small proportion is known – usually by cloning genes identified by mutant phenotypes (Caddick and Dobson, 2007). Reflecting the metabolic versatility of fungi, a common function of this class of transcription factor is to mediate the induction of carbon and/or nitrogen catabolic or secondary metabolite pathways by acting as pathway-specific activators. Here instead we have characterized the function of two widely conserved Zn cluster proteins that act together to regulate the expression of genes involved in central carbon metabolism.

Fungi are capable of using a diverse range of carbon sources and need to rearrange metabolism to adapt to changes in substrate availability. This is of fundamental significance not only for saprophytic fungi growing in the environment but also for pathogenic fungi adapting to the animal or plant host environment upon infection. In industry, an understanding of carbon metabolism is an important factor in efforts to manipulate the energy-efficient production of products such as commercial enzymes, citric acid, antibiotics and of substrates for ethanol production. While fungi can grow on a range of carbohydrates and sugars using glycolysis, they can also use many alternative carbon sources such as ethanol, amino acids, lipids and organic acids which are metabolized solely via the TCA cycle and hence require the formation of sugars in the process of gluconeogenesis; a reversal of glycolysis in which the TCA-cycle intermediate oxaloacetate is converted to sugars. Regulation of enzyme activities is absolutely vital to avoid futile cycling by glycolysis opposed by gluconeogenesis. Alterations in the available carbon source can result in the induction of enzymes specific to the breakdown of the particular compound and also lead to rearrangement in the flow of metabolites through the central pathways of carbon metabolism. Therefore the expression of genes specific to these pathways must be regulated in response to changes in the carbon source.

The key enzymes for gluconeogenesis are: phosphoenolpyruvate carboxykinase (PCK, E.C. which converts oxaloacetate to phosphoenolpyruvate and fructose-1,6-bisphosphatase (FBP, E.C. the final step in hexose monophosphate formation. Transcriptional regulation of these enzymes, and other aspects of gluconeogenesis, have been studied in Saccharomyces cerevisiae. S. cerevisiae is specialized with a strong preference for growth on fermentable monosaccharides generating ATP via glycolysis and not by respiration. In the absence of a fermentable carbon source, metabolism switches to a respiratory mode for the use of alternative carbon sources (Haurie et al., 2001; Schüller, 2003). S. cerevisiae can only use a limited number of substrates that result in the generation of TCA-cycle intermediates as sole carbon sources – ethanol, acetate and fatty acids all of which result in the production of acetyl-CoA. The glyoxylate cycle is necessary for the net conversion of acetyl-CoA to oxaloacetate which is then used for gluconeogenesis. The extensive changes in gene expression accompanying growth on non-fermentable carbon sources is determined by the Snf1 kinase which, in response to glucose limitation, relieves repression by the Cys2His2 zinc finger Mig1 repressor and stimulates the transcriptional activators, Adr1, Cat8 and Sip4 (which is activated by Cat8) to turn on genes involved in ethanol breakdown, acetyl-CoA synthesis, the glyoxylate by-pass and gluconeogenesis, including FBP and PCK (Haurie et al., 2001; Schüller, 2003). Different genes are dependent to varying extents on Cat8 and Sip4 binding to carbon source response elements (CSRE) and Adr1 to UAS1 sequences in their 5′ regions (Young et al., 2003; Tachibana et al., 2005). More recently additional Zn cluster transcription factors, Rds2, Ert1 and Gsm1 regulating non-fermentable carbon source metabolism have been identified (Soontorngun et al., 2007; Turcotte et al., 2010).

In contrast most other fungi are obligate aerobes with the TCA cycle and respiration used during growth on glycolytic carbon sources. In addition they can also utilize not only carbon sources that are metabolized via acetyl-CoA but also alternative carbon sources that directly result in the formation of TCA cycle intermediates, e.g. amino acids. This results in significant differences between the strategies employed by S. cerevisiae and other fungi.

Aspergillus nidulans is capable of using a diverse range of gluconeogenic carbon sources – ethanol, fatty acids and acetate which are metabolized via acetyl-CoA; amino acids, such as glutamate and proline, as well as aromatic compounds (Hondmann and Visser, 1994). Specific catabolic pathways are controlled by pathway-specific induction by transcriptional activators with glucose repression of these pathways mediated by the CreA repressor (Dowzer and Kelly, 1991). Pathway-specific Zn cluster transcription factors include PrnA, which controls proline induction of activities required for the conversion of proline to glutamate (Gómez et al., 2002), AlcR, controlling conversion of ethanol to acetate (Felenbok et al., 2001), and FarA and FarB controlling the expression of proteins required for fatty acid utilization, including glyoxylate cycle enzymes (Hynes et al., 2006). In response to acetate, FacB activates genes specifically required for growth on acetate, such as the gene for acetyl-CoA synthetase, as well as the glyoxylate cycle genes (Todd et al., 1997; 1998). FacB has some similarities to Cat8 and Sip4 of S. cerevisiae: critically however facB mutations do not affect growth on gluconeogenic carbon sources which are not metabolized via acetyl-CoA showing that FacB is a specific regulator of acetate metabolism and does not directly control gluconeogenesis (Kelly and Hynes, 1981; Hynes et al., 2002; 2007).

Therefore genes required for gluconeogenesis must be regulated independently of the particular carbon source by a central control that responds to the build up of TCA cycle intermediates and the requirement for hexose synthesis. The acuF gene, encoding PCK, is induced not only by acetate but also by glutamate, proline and other sources of TCA cycle intermediates (Hynes et al., 2002). The acuG gene, encoding FBP, is required for growth on glycerol as well as on gluconeogenic carbon sources and is subject to control by CreA mediated carbon catabolite repression as well as elevated expression under conditions where TCA-cycle intermediates accumulate (Hynes et al., 2007). Other enzymes of gluconeogenesis are reversible and required also for glycolysis. Enolase (encoded by acuN) is expressed from two transcription starts – one controlled by glycolytic carbon sources and one by gluconeogenic carbon sources (Hynes et al., 2007). Presumably other reversible enzymes of glycolysis/gluconeogenesis have similar dual control.

The acuK and acuM genes were originally identified among mutants unable to utilize acetate (Armitt et al., 1976). Mutations in either gene result in loss of growth on all carbon sources requiring gluconeogenesis. Consistent with this, acuK and acuM mutations abolish induction of acuF, acuG and acuN by sources of TCA cycle intermediates such as acetate and proline. The acuK and acuM genes encode related transcription factors with Zn cluster DNA-binding domains (DBDs). Therefore AcuK and AcuM are proposed to be the key regulators controlling the transcriptional reprogramming of metabolism in response to growth on gluconeogenic carbon sources (Hynes et al., 2007).

Here we describe an analysis of the function of these transcription factors in A. nidulans. It is proposed that, independently of the carbon source, AcuK and AcuM bind only as heterodimers to a sequence motif containing direct repeats of CCG separated by seven bases (CCGN7CCG) and activate gene expression in response to one or more metabolites generated by metabolism of gluconeogenic carbon sources. Predicted orthologues of AcuK and AcuM, with highly conserved DBDs and C-terminal sequences, are present in fungal taxa ranging from the basal chytrids to the basidiomycetes. Analysis of the 5′ sequences of genes relevant to gluconeogenesis in a range of species indicates that the CCGN7CCG motif is widely conserved. The extraordinary extent of phylogenetic and sequence conservation indicates a fundamental role for these transcription factors in controlling fungal metabolism.


Sequence features and conservation of AcuK and AcuM

The acuK and acuM genes were originally cloned by complementation of the acuK248 and acuM301 mutations (Hynes et al., 2007). The genes were found to correspond to AN7468 and AN6293 respectively in the A. nidulans genome databases. acuK was predicted to have one intron close to the 5′ end. Subsequently, however, an additional intron has been found at the 3′ end of the coding region and confirmed by sequencing of cDNA. Orthologues in related filamentous ascomycetes possess this intron, although the intron in Neurospora crassa is longer than that of other species (Chae et al., 2007). AcuK and AcuM are conserved in euascomycetes and are related to each other with the Zn2Cys6 binuclear cluster DBDs showing significant conservation but also conserved sequence features unique to each protein. blastp analysis in fungal databases showed that AcuK and AcuM are widely conserved in fungal taxa with clear orthologues found in ascomycete, basidiomycete and zygomycete species. In addition the chytrids, Allomyces macrogynus and Spizellomyces punctatus have orthologues of both proteins while the sequenced isolate of Batrachochytrium dendrobatidis has an AcuM orthologue but is missing an annotated AcuK orthologue. In striking contrast the A. nidulans transcription factors specifically controlling acetate and proline utilization, FacB and PrnA respectively (Todd et al., 1997; Cazelle et al., 1998), are not highly conserved outside the ascomycetes.

Alignments of selected species are shown in Fig. S1. Conservation extends to the C-terminal sequences with the AcuK orthologues being somewhat less conserved than those of AcuM. The proteins lack the middle homology region found in many proteins containing the Zn cluster DBD and also lack obvious dimerization domains adjacent to the DBD (Schjerling and Holmberg, 1996; MacPherson et al., 2006). Conserved predicted Per–Arnt–Sim (PAS) domains (Henry and Crosson, 2011) are found in the C-terminal regions. Interestingly the S. cerevisiae DBDs are less conserved than those of all other species including the hemi-ascomycete, Candida albicans. It should be noted that S. cerevisiae has an additional transcription factor, Gsm1, that is clearly related to AcuK and has a C. albicans orthologue. However these proteins have more divergent DBDs (Fig. S1) and also lack a sequence in the conserved C-terminal region preceding the PAS domain (not shown).

Deletion of acuK and acuM

The mutations, acuK248 and acuM301, result from single base pair changes resulting in premature stop codons causing loss of the conserved C-terminal sequences (Hynes et al., 2007). To confirm that these mutations represented a complete loss of function, deletions, made by gene replacement, were created. The acuKΔ was made by replacing a 1064 bp region (−22 to +1041), which includes the translation start and the DBD with the A. nidulans riboB gene. For the acuMΔ a 1458 bp region (−436 to +1021), which includes the acuM translation start and the DBD, was replaced with the A. nidulans pyrG gene. The phenotypes of these mutants and the double acuKΔ acuMΔ mutant on gluconeogenic carbon sources were identical to those of the original acuK248 and acuM301 mutants (see below). It is important to note that deletion of these genes has no detectable effect on growth on glycolytic carbon sources or on asexual or sexual development.

Expression and nuclear localization of AcuK and AcuM

The possibility that AcuK and AcuM expression is regulated by the carbon source present has been investigated. RT-PCR analysis showed that both acuK and acuM are expressed in the presence of glucose (results not shown). In agreement with this, microarray analysis has shown expression of both genes in glucose-grown cells with the level of expression elevated about threefold in ethanol-grown cells (David et al., 2006) and the N. crassa orthologues are transcribed in sucrose grown (Vogel's medium) mycelia (Chae et al., 2007). Western blot analysis of Myc tagged proteins was performed. Using the gene replacement procedure described in later experiments, strains expressing N-terminal Myc13-tagged proteins under the control of their cognate native promoters were isolated. These strains had a wild-type phenotype indicating functionality of the modified proteins. Both Myc-AcuM and Myc-AcuK were present in glucose grown mycelia with levels increased in mycelia transferred to acetate and proline containing media (Fig. 1A).

Figure 1.

Regulation and nuclear localization of AcuK and AcuM. A. Regulation of expression of N-terminal tagged Myc proteins. Western blot of extracts (50 µg of total protein) of mycelium of strains grown for 16 h in 1% glucose minimal liquid medium with 10 mM ammonium tartrate as the nitrogen source and then transferred for 6 h to either the same medium (glucose) or carbon-free minimal medium with 50 mM proline (proline), 50 mM acetate (acetate) with 10 mM ammonium chloride as the nitrogen source. α-tubulin was used as a loading control. B. Nuclear localization of GFP–AcuK and GFP–AcuM. Strains were grown on coverslips for 12 h at 37°C in 1% glucose minimal liquid media and then transferred to the same media, 50 mM proline or 50 mM acetate with ammonium chloride as the nitrogen source for a further 4 h before fixing for microscopy. DIC, differential interference contrast. Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI). Equivalent results were observed for an additional two independent complementing transformants.

The activity of AcuK and AcuM could be regulated by their subcellular localization in response to different carbon sources. GFP was fused to the N-terminal region of both AcuK and AcuM and the fusion proteins were placed under the control of the constitutively expressed A. nidulans gpdA promoter. pYS6604, containing the gpdA(p)gfp–acuK fusion construct with Aspergillus fumigatus riboB+ as a selectable marker, was transformed into the loss of function mutant acuK248 (MH9416). Twelve of 26 independent RiboB+ transformants complemented the acuK248 phenotype for growth on proline as a carbon source. pYS6605, containing the gpdA(p)gfp–acuM fusion construct with A. fumigatus pyroA+ as a selectable marker was transformed into the loss of function acuM301 mutant (MH10131). Thirteen of 26 independent PyroA+ transformants complemented the acuM301 phenotype for growth on proline. Therefore both GFP fusion proteins expressed from the gpdA promoter were functional. Three individual complementing transformants for each construct were selected for microscopy. Both GFP–AcuK and GFP–AcuM were found to be localized to the nucleus and this was independent of the presence of glucose, proline or acetate as carbon sources (Fig. 1B). Therefore regulation by the AcuK and AcuM proteins is not controlled by differential nuclear localization in response to the carbon source.

Analysis of binding sites of AcuK and AcuM

Regulatory Sequence Analysis Tools (van Helden, 2003) were used to identify potential binding sites. The promoter regions (1 kb) 5′ to the predicted start codon of four AcuK/M regulated genes (encoding FBP, PCK, enolase and malic enzyme) from A. nidulans, A. fumigatus, A. oryzae and A. terreus were searched for dyad repeats in direct or indirect orientation separated by variable spacing. The motif CCGN7CCG was found to be significantly over-represented in these sequences compared with the occurrence expected in random sequences. The occurrence of this motif was investigated in a broader range of Aspergillus spp. and was frequently found in the promoter region of these genes, conserved in position relative to the ATG and there was also conservation in sequences flanking the core motif (Table S1). Conservation of motifs in orthologues of the acuF gene is shown in Fig. 2A and B. For all species analysed, there are two motifs (sites A and B) on opposite DNA strands arranged in the same orientation with similar spacing (43–88 bp) within 200 bp upstream of the start codons, and with the proximal site (site A) being more highly conserved than the distal site (site B). Additional motifs further upstream are found for some species including A. nidulans (Fig. 2A).

Figure 2.

Identification of binding sites for AcuK and AcuM. A. Presence of the CCGN7CCG motif in 1 kb upstream of the ATG codon in PCK genes of Aspergillus species detected by RSAT analysis ( The motif is shown as bars – above the line indicates a direct orientation while below the line indicates a reverse orientation. Sequences were obtained from the Broad Institute Aspergillus Comparative Site ( A. fumigatus (Afu6g07720), A. nidulans (AN1918), A. oryzae (AO090003000174), A. terreus (ATEG_05895), A. niger (est_GWPlus_C_40489), A. clavatus (ACLA_081870). B. Consensus sequences of sites A and B determined by the sequence logo generator Weblogo ( C. Schematic of probes derived from the 5′ region of acuF used for EMSA analysis. Numbers represent base-pair co-ordinates relative to the ATG. Foligo1mut is derived from Foligo1 with the base substitutions shown. D and E. EMSA assays using MBP–AcuK(1–123) and MBP–AcuM(1–124) fusion protein extracts (designated as AcuK and AcuM respectively). Lanes designated Probe contained no extract control reactions and lanes designated MBP contained control reactions with MBP protein extract. The probes used are indicated at the bottom and FP represents free probe.

Electrophoretic mobility shift assays (EMSA) were performed using bacterially expressed AcuK and AcuM fusion proteins with probes derived from the acuF 5′ sequence. Proteins were expressed in Escherichia coli as N-terminal maltose-binding protein (MBP) fusions to aa 1–123 of AcuK and to aa 1–124 of AcuM with each containing the Zn cluster domain. The probes used are shown in Fig. 2C. Binding to fragments containing both sites (F2), site B (F3) and site A (F4) was observed only when binding reactions contained both AcuK and AcuM fusion proteins and not when either fusion protein was present alone (Fig 2D). Similarly, a mobility shift was only observed with oligonucleotide probes encompassing site A (Foligo1) and site B (Foligo2) when both fusion proteins were present and, for site A, this was abolished by single base-pair substitutions in both of the CCG repeats of the CCGN7CCG motif (Fig. 2E). Similar results were observed when MBP fusion proteins containing full-length sequences of AcuK and AcuM were used (results not shown). These results strongly indicated that binding by AcuK and AcuM to DNA containing the CCGN7CCG motif is interdependent, at least in vitro. Because Zn cluster proteins almost invariably bind as dimers, binding is highly likely to be as a heterodimer. Furthermore, this binding specificity only requires the N-terminal regions of the proteins.

Functional analysis of the binding motifs in the acuF promoter

AcuK- and AcuM-dependent induction of an acuF–lacZ reporter gene integrated at the argB locus was previously shown (Hynes et al., 2002; 2007). A promoter length of 486 bp was sufficient for induction by gluconeogenic carbon sources (Hynes et al., 2002). The function of sites A and B containing the CCGN7CCG motif was investigated by the construction of strains containing various acuF promoter sequences driving expression of the acuF–lacZ reporter gene. A novel approach was used in which fusion constructs with a truncated lacZ reporter gene were targeted by homologous crossing over with an amdS–lacZ fusion gene replacement at the amdS locus (Davis et al., 1988) resulting in a truncated amdS–lacZ fusion and an intact acuF–lacZ reporter gene (see Fig. 3A). The promoter variants used are diagrammed in Fig. 3B.

Figure 3.

In vivo significance of AcuK and AcuM binding sites for regulation of an AcuF–LacZ reporter. A. Strategy for the creation of acuF–lacZ strains with acuF promoter modifications. Transformation of an nkuAΔ pyroA4 strain containing an amdS–lacZ fusion and selection for PyroA+ with plasmids carrying the pyroA gene from A. fumigatus and acuF promoter (acuF(p)) variants upstream of a truncated lacZ gene (lacZ′) fused to acuF enabled the generation of a full-length acuF–lacZ reporter by a single cross-over accompanied by the formation of a truncated amdS–lacZ′. Diagram is not to scale. B. acuF promoter variants studied. Co-ordinates are shown relative to the ATG. −198 mut contains a deletion of three bases converting the sequence of site A from GCCCCGACATCTCCCG to GCCACATCTCCCG. C. Expression of acuF–lacZ reporters. Strains are designated according to the promoter variants in (B). Mycelium was grown for 16 h in 1% glucose medium and then transferred to fresh medium containing 1% glucose, 50 mM proline or 50 mM acetate medium for 6 h before harvesting. Standard errors based on three independent experiments are shown. Note the difference in scale between the two parts of (C).

Strong AcuK- and AcuM-dependent induction by proline and acetate was observed for −486 and −198 promoters containing both sites A and B (Fig. 3C). Induction was greatly reduced (approximately sixfold) by mutation of site A (−198 mut) and more than 10-fold when only site A was present (−122). It should be noted that levels on glucose (reflecting endogenous induction) were also reduced in these strains and also by acuKΔ and acuMΔ (see below). These results showed that sites A and B, identified by conservation between Aspergillus spp. and by EMSA analysis, were sufficient to result in induction and therefore functional in vivo and furthermore the sites were strongly synergistic in their effects on expression.

AcuK and AcuM binding in vivo is independent of the carbon source

Chromatin immunoprecipitation (ChIP) was used to determine whether AcuK and AcuM can bind to the two sites in the acuF promoter, whether binding to DNA is interdependent, as observed in vitro, and also whether binding is carbon source dependent. Strains expressing epitope tagged proteins were generated by the gene replacement method described in later experiments such that each protein was expressed from its native promoter. AcuK and AcuM were N-terminally tagged with 13 copies of Myc and three copies of HA respectively. Crossing was used to obtain strains expressing both the fusion proteins and to obtain the relevant acuKΔ or acuMΔ double mutants. Strains expressing the fusion proteins were phenotypically wild type. ChIP analysis (Fig. 4) showed binding to sequences in the 5′ regions of acuF, as well to the promoters of acuG (encoding FBP) and aodA (encoding the alternative oxidase – see below). For all three genes no significant binding of single fusion proteins was observed in the absence of the corresponding binding partner and the levels of occupancy of AcuK and AcuM were extremely similar, strongly indicating that AcuK and AcuM only bind together – presumably as heterodimers. Importantly, strong binding was observed in cultures grown on glucose as well as the gluconeogenic carbon sources, acetate and proline. Binding in acetate grown cultures was somewhat lower than for the other carbon sources for acuF and aodA. Induction of acuF by acetate and proline is generally similar (Hynes et al., 2002; 2007) and the significance (if any) of this observation is unknown. These data were consistent with nuclear localization being independent of the carbon source (Fig. 1B) and also indicated that the expression level during growth on glucose is not limiting for binding.

Figure 4.

AcuK and AcuM binding is interdependent in vivo. ChIP analysis was performed on chromatin prepared from strains grown for 16 h in 1% glucose medium and then transferred to fresh medium containing 1% glucose, 50 mM proline or 50 mM acetate medium for 6 h. Myc13-tagged AcuK and HA3-tagged AcuM were precipitated with anti-Myc and anti-HA respectively. Quantitative real-time PCR was used to detect binding to DNA from the 5′ regions of acuF, acuG and aodA using primers corresponding to the indicated co-ordinates. Results are expressed as fold enrichment relative to a region within the coding region of the benA gene (+968 to +1048). Error bars show standard errors from three replicates for the myc::acuK; HA::acuM strain and two replicates for the other strains. The schematic shows the position of CCGN7CCG motifs relative to primer positions (arrowheads).

Presence of the CCGN7CCG motif in the upstream regions of genes

In order to assess the conservation of gene regulation by AcuK and AcuM orthologues binding to the CCGN7CCG motif, analysis of the presence of the motif in the 5′ region of some relevant genes in fungi was performed. In Aspergillus spp. the motif is commonly conserved in relative position and there was striking conservation of sequences including flanking sites suggesting similar binding affinities for AcuK and AcuM (Table S1). Notably most genes coding for TCA cycle enzymes contained one or more motifs consistent with TCA cycle enzyme activities being increased during growth on acetate (McCullough et al., 1977) and the increased expression of TCA cycle genes in ethanol relative to glucose (David et al., 2006). Expression of acuN encoding enolase, is dependent on AcuK and AcuM during growth on gluconeogenic carbon sources (Hynes et al., 2007) and, consistent with this, the CCGN7CCG motif is present in the 5′ region of enolase genes in Aspergillus spp. (Table S2). The motif was also found upstream of other glycolytic/gluconeogenic genes in A. nidulans (Table S2) and preliminary RNA sequencing data indicate that maintenance of transcription of these genes upon shifting from glucose to proline or acetate is AcuK and AcuM dependent. The expression of genes for the electron transport chain; succinate dehydrogenase (E.C., cytochrome c, ubiquinol-cytochrome c reductase (complex II – E.C. and cytochrome c oxidase (complex IV – E.C.; are increased relative to glucose during growth on ethanol in A. nidulans (David et al., 2006). All of these genes contain at least one upstream CCGN7CCG motif indicating a role for AcuK and AcuM in controlling their expression (not shown).

Studies of possible sources of induction

Our results indicated that increased expression of gluconeogenic enzymes does not result from regulation at the level of DNA binding by AcuK and AcuM. It is likely therefore that the response results from signalling by one or more metabolites produced from gluconeogenic carbon sources. In previous studies with an AcuF–LacZ reporter it was found that acetate and proline induction was not additive, metabolism of acetate via acetyl-CoA and acetyl-carnitine entry into mitochondria was required and proline and glutamate induction required metabolism via NAD-dependent glutamate dehydrogenase (GdhB) activity to 2-oxoglutarate (Hynes et al., 2002). Furthermore induction by both proline and acetate was increased in an acuF mutant, predicted to result in oxaloacetate accumulation, and both succinate and malate could result in induction (Hynes et al., 2007). These data suggested that malate and/or oxaloacetate produced by metabolism of acetate by the glyoxylate cycle or by proline degradation to 2-oxoglutarate and metabolism via the TCA cycle resulted in induction. Relevant pathways are shown in Fig. 5.

Figure 5.

Relevant pathways of carbon metabolism. Metabolism of acetate via acetyl-CoA and the glyoxylate cycle results in succinate and malate formation while proline is metabolized via glutamate to 2-oxoglutarate. Enzyme abbreviations: PKI, pyruvate kinase; PYC, pyruvate carboxylase; PCK, PEP carboxykinase; ACL, ATP citrate lyase; MDH, malate dehydrogenase; MLS, malate synthase; ICL, isocitrate lyase; CS, citrate synthase; FUM, fumarase; SDH, succinate dehydrogenase; ACO, aconitase; IDH, isocitrate dehydrogenase; GDH, NAD-dependent glutamate dehydrogenase; KGD, 2-oxoglutarate dehydrogenase. Note that fumarase is proposed to be localized to both cytosol and mitochondria and to favour malate formation, as shown for S. cerevisiae (Pines et al., 1996; Sass et al., 2001). Open arrows indicate metabolite entry and exit from mitochondria. AcuL is the predicted succinate-fumarate mitochondrial transporter (Flipphi et al., 2009).

The effect of loss of the specific enzymes of the glyoxylate cycle on induction was not previously studied. Induction of the acuF–lacZ reporter by acetate, but not by proline, was found to be lost in both acuD (isocitrate lyase) and acuE (malate synthase) loss of function mutants (results not shown) showing that malate production from acetyl-CoA and glyoxylate is required.

Because malate dehydrogenase (MDH, l-malate:NAD oxidoreductase, E.C. catalyses the NAD/NADH-dependent interconversion of oxaloacetate and malate, the importance of cytoplasmic MDH in gluconeogenesis has been investigated. There are at least two isozymes of MDH in most eukaryotic cells. The mitochondrial isozyme functions in the TCA cycle; while the cytoplasmic isozyme, by producing cytoplasmic oxaloacetate, the substrate for PCK, is potentially required for gluconeogenesis. In A. nidulans, MDH activity is present in both cytoplasmic and mitochondrial fractions of glucose-grown cells at almost equal levels (Osmani and Scrutton, 1983). Total MDH activity increases two- to threefold during growth on acetate, but not in mutants that cannot utilize acetate (Kelly, 1980). Three predicted genes encoding MDH are present in the A. nidulans genome (Flipphi et al., 2009). AN6717 (mdhA) encodes a protein with a mitochondrial targeting sequence and is predicted to be involved in the TCA cycle. AN5031 (mdhB) is predicted to encode a protein lacking MTS or PTS targeting sequences while AN6499 (mdhC) lacks a MTS sequence. Inspection of mdhC EST sequences predicts that alternative splicing of a 3′ intron occurs. When the intron is removed the C-terminal sequence is VEFAQSPPPKL, predicted to result in peroxisomal localization (Neuberger et al., 2003). When the intron is retained the C-terminal sequence is VEFAQSPPPK which is not predicted to act as a PTS sequence. In Penicillium chrysogenum, the product of Pc12g04750, a predicted MDH, has been found in the peroxisomal proteome (Kiel et al., 2009). Therefore MdhC is predicted to be both cytoplasmic and peroxisomal.

The role of mdhB and mdhC has been investigated by the creation of deletion strains. Deletion of mdhA was not attempted because of the likely essential function in the TCA cycle. The mdhBΔ strain did not exhibit any significant growth phenotypes, either by itself or in combination with mdhCΔ, on glucose or gluconeogenic carbon sources (Fig. 6A). Close orthologues of mdhB are not present in other Aspergillus spp., with the exception of A. niger, nor in other euascomycete genomes and the level of expression is low and unaffected during growth in ethanol medium (David et al., 2006). Therefore the function of this predicted gene product is unclear.

Figure 6.

Effects of deletion of mdh genes on gluconeogenesis. A. Growth of mdh deletion strains on carbon sources. Growth was for 2 days. Carbon sources were added to minimal medium with 10 mM ammonium tartrate (glucose) or ammonium chloride as the sole nitrogen source Concentrations were as follows: glucose, 1%, proline and acetate, 50 mM and ethanol, 0.5%. B. Northern blot analysis of mdhC expression. RNA was extracted from mycelium grown for 16 h in 1% glucose medium and then transferred to fresh medium containing 1% glucose, 50 mM proline or 50 mM acetate medium for 6 h. mdhC RNA was detected by hybridization with a labelled PCR product corresponding to +412 to +869, while hybridization with a histone H3 gene probe was used as a loading control. C. Expression of the acuF (−2364 to +440)–lacZ reporter, in the indicated mdhBΔ and mdhCΔ backgrounds. Growth of mycelium for assays was as for Fig. 4. Means of results and standard errors based on three replicates are shown. D. Northern blot analysis of the effects of the mdhCΔ on acuF expression detected by hybridization with a labelled PCR product corresponding to +146 to +2199. E. Induction of the acuF–lacZ reporter in an mdhCΔ background. Mycelium was grown for 16 h in 1% glucose medium and then transferred to fresh medium containing 1% glucose, 1% glucose containing chloramphenicol (4 mg ml−1), or minimal medium with malate, fumarate or succinate (all 20 mM) added as the sole carbon source for 6 h.

Deletion of mdhC resulted in no gross colony defects and no effects on sexual development. However, consistently, colonies had a brownish appearance resulting from reduced conidiation and quantification showed a 50% reduction in the production of conidia (results not shown). Growth on acetate, oleate or proline of mdhCΔ was not affected (Fig. 6A and not shown) indicating that neither cytoplasmic nor peroxisomal MDH activity is essential either for the glyoxylate cycle or for gluconeogenesis. Growth on ethanol was partially reduced however. This may result from cytoplasmic MDH playing a role in re-oxidization of NADH produced by alcohol and acetaldehyde dehydrogenases during growth on ethanol, as proposed for S. cerevisiae (Bakker et al., 2001). Microarray analysis has shown that expression is elevated more than twofold on ethanol compared with glucose (David et al., 2006).

Increased expression of mdhC on acetate and proline was AcuK- and AcuM-dependent (Fig. 6B). This is consistent with the presence of predicted AcuK/M binding sites in the 5′ of mdhC (Table S1) and upstream of predicted orthologues in other fungi (Table S3). The expression of acuF was studied in the mdh gene deletion backgrounds to determine the effects of blocking the inter-conversion of oxaloacetate and malate in the cytoplasm. A strain with the acuF–lacZ reporter was crossed to mdhBΔ and mdhCΔ strains to generate appropriate strains (Fig. 6C). Induction by acetate and proline was not affected by mdhBΔ but was higher in the mdhCΔ background suggesting that accumulation of cytoplasmic malate resulted in increased induction. Surprisingly, levels on glucose were greatly reduced in the mdhCΔ background. This suggested that, during growth on glucose, endogenous induction of acuF–lacZ transcription occurred via conversion of oxaloacetate (formed from pyruvate by pyruvate carboxylase and from citrate by ATP-citrate lyase) to cytoplasmic malate by MdhC (see Fig. 5). This pathway for malate production in the cytosol has been established for S. cerevisiae and Aspergillus spp. (Pines et al., 1996; Goldberg et al., 2006). Stable β-galactosidase accumulation during the 16 h pre-growth period on glucose would account for the levels observed. Northern blot analysis of acuF expression showed higher levels of induction observed in the mdhCΔ background (Fig. 6D). In an mdhCΔ background induction by malate, as well as fumarate and succinate, which can be converted to malate, was observed (Fig. 6E). Overall these data were consistent with malate accumulation resulting in induction.

Regulation of the alternative oxidase

We have investigated the regulation and role of the alternative oxidase (AOX) because orthologues of acuK and acuM in N. crassa and Podospora anserina have been found to regulate the expression of a gene encoding this enzyme (Chae et al., 2007; Sellem et al., 2009). Mitochondrial AOX catalyses the transfer of electrons from ubiquinol to oxygen without the formation of ATP, thereby by-passing the final steps of the electron transport chain. AOX is induced by inhibitors of complex III or complex IV (antimycin A and cyanide respectively), by inhibitors of mitochondrial protein synthesis such as chloramphenicol and by mutations in genes for components of complex III or complex IV (Turner and Rowlands, 1976; Li et al., 1996; Veiga et al., 2000; 2003; Millenaar and Lambers, 2003; Tanton et al., 2003; Helmerhorst et al., 2005; Magnani et al., 2007; Sellem et al., 2009).

Some fungi encode two AOX isoforms – one constitutive and one inducible (Li et al., 1996; Huh and Kang, 2001; Tanton et al., 2003). However A. nidulans has a single AOX-encoding gene (AN2099 – designated aodA) and this was deleted by gene replacement. This resulted in sensitivity to antimycin A in the presence of glucose. Growth and development of aodAΔ containing strains and, importantly, utilization of the gluconeogenic carbon sources acetate and proline was not detectably different from wild type (Fig. 7A). Induction of acuF RNA by acetate and proline was not affected by aodA (Fig. 7B). Therefore AOX does not play an essential role in gluconeogenic growth. In some fungi AOX expression is induced by sources of reactive oxygen species, e.g. in A. fumigatus and C. albicans (Huh and Kang, 2001; Magnani et al., 2007). However it has been reported that aodA expression is not induced by oxidative stress in A. nidulans (Pusztahelyi et al., 2011) and the aodAΔ did not result in altered sensitivity to the oxidative stress agents, hydrogen peroxide or menadione (results not shown).

Figure 7.

Regulation of aodA, encoding AOX, and the effects of deletion of the gene. A. The aodAΔ results in sensitivity to antimycin but does not affect growth on gluconeogenic carbon sources. Strains were point inoculated on acetate and proline plates (as described in Fig. 6) and plates containing antimycin A (added at the indicated concentration to 1% glucose minimal medium with 10 mM ammonium tartrate as the nitrogen source). Plates were incubated for 2 days. In addition 2 µl of dense conidial suspensions were spotted on to antimycin (30 µg ml−1) and incubated for 1 and 2 days. B. The aodAΔ does not affect acuF expression. RNA was extracted from mycelia grown in glucose minimal medium with ammonium tartrate as the nitrogen source and transferred to the indicated carbon sources for 6 h and subject to Northern blotting sequentially with PCR generated probes specific to aodA (co-ordinates: +12 to +1104) and acuF (as for Fig. 6). Ethidium bromide staining of rRNA as loading controls is shown. C. Regulation of expression of aodA by AcuK and AcuM. RNA was extracted from mycelia grown in glucose minimal medium with ammonium tartrate as the nitrogen source and transferred to the indicated carbon sources for 6 h and subject to Northern blotting sequentially with probes specific to aodA, acuF and the histone H3 gene. D. Expression of aodA is not observed with carbon starvation. RNA was extracted from mycelia grown in glucose minimal medium with ammonium tartrate as the nitrogen source for 16 h (glucose) and also from mycelium transferred to medium lacking an added carbon source for 6 h (C starvation).

Expression of aodA was readily detected in the presence of glucose, unlike acuF, and was increased by acetate and proline (Fig. 7C); while carbon starvation for 6 h resulted in loss of detectable aodA expression (Fig. 7D). Microarray analysis has shown that aodA is expressed in glucose-grown cells and the level is approximately threefold higher with ethanol (David et al., 2006). Some strains of N. crassa produce AOX transcripts under non-inducing conditions but no corresponding AOX protein or activity implying translational control (Tanton et al., 2003). Whether this also occurs for A. nidulans is not known.

In P. anserina, antimycin was found to result in induction of PCK and FBP RNA (Sellem et al., 2009) and chloramphenicol, which is an inducer of AOX expression in other fungi, was found to induce AcuF–LacZ expression (Fig. 6E). Therefore inhibition of NADH oxidation via mitochondrial electron transport can induce expression of both AOX and gluconeogenic enzymes. The expression of aodA was abolished by deletion of either acuK or acuM (Fig. 7D). Consistent with this an equivalent level of antimycin A sensitivity was observed in acuKΔ and acuMΔ compared with the aodAΔ (Fig. 7A). Therefore AcuK and AcuM are absolutely required for aodA expression as for N. crassa and P. anserina (Chae et al., 2007; Sellem et al., 2009). ChIP analysis showed very strong binding of AcuK and AcuM to a 5′ upstream region of aodA containing the CCGN7CCG motif (Fig. 3). This motif was conserved in the 5′ of AOX genes in Aspergillus spp. (Table S1) and was also present in other fungal spp. with the sole exception of C. albicans (Table S3).

The effects of targeted mutations on acuK and acuM function

A strategy was devised for the replacement of acuK and acuM with altered versions of the genes. This involved transformation of acuKΔ and acuMΔ strains with linear fragments generated by PCR from plasmid constructs containing the bar selectable marker downstream of the altered genes and selection for glufosinate resistance. This could result in either gene replacement by a distal cross-over or no gene replacement when a proximal cross-over occurred. These events could be distinguished by scoring for loss of the selectable marker used for gene deletion –riboB+ for acuK and pyrG+ for acuM. The strategy is diagrammed in Fig. 8A and B. This method proved to be efficient with about 50% of transformants giving the desired gene replacements. Replacement of the deletions with the respective wild-type gene yielded strains with wild-type phenotypes indicating that the presence of the inserted downstream bar gene did not affect function. This procedure was used to generate the epitope-tagged versions of the genes used for the Western and ChIP analyses described above.

Figure 8.

Generation and analysis of variants of acuM and acuK. A and B. The strategy for construction of acuK and acuM variants respectively. Strains containing the gene replacement deletions of acuK and acuM (acuKΔ riboB2 and acuMΔ pyrG89) were transformed with linear PCR fragments of plasmid constructs containing variant sequences (designated by asterisks) and the bar marker and transformants selected for glufosinate resistance. Desired gene replacements, resulting from cross-overs at positions 1 and 3, result in replacement of riboB+ giving riboflavin auxotrophy (acuK) or pyrG+ giving uracil/uridine auxotrophy (acuM). Co-ordinates are relative to the position of the start codons of the genes. C. Variants generated and summary of phenotypes. Plasmid construct numbers are detailed in Supporting information. The proteins are diagrammed with DNA-binding domains (DBD), a weakly conserved central sequence (shaded box) and the conserved C-terminal regions shown. Co-ordinates refer to amino acids. Thick lines indicate regions deleted and dots indicate positions of amino acid substitutions. Replacement with sequences from the S. cerevisiae genes RDS2, ERT1 and GSM1 are indicated with predicted encoded amino acid co-ordinates. Function refers to growth tests on gluconeogenic carbon sources (carbon) with +, ± and − representing wild type, partial and no growth respectively and on antimycin (30 µg ml−1) with R and S representing resistance and sensitivity respectively.

A summary of the results for the gene replacements made is shown in Fig. 8C. Deletion of the DBDs and two small and two large deletions in the highly conserved C-terminal region resulted in complete loss of function. A small deletion of a less conserved region (acuMΔ205–239 and acuKΔ245–286) had no effect. In N. crassa amino-acid substitutions in the proposed PAS domains of the acuK and acuM orthologues were found to abolish AOX function as assayed by antimycin sensitivity (Chae et al., 2007). The corresponding mutations (acuMG418A and acuKG535A) were found to lead to antimycin sensitivity but growth on gluconeogenic carbon sources was only slightly affected (Fig. 9A). However the double mutant strain was much more affected, thereby providing evidence for a functional interaction between the AcuK and AcuM PAS domain sequences.

Figure 9.

Growth tests of selected acuM and acuK variants. A. Effects of glycine to alanine substitutions in the PAS domains. Growth on gluconeogenic carbon sources is more affected in the double mutant. B. Effects of switching DNA-binding domains. The proteins are diagrammed showing the predicted heterodimers. Only the strain with each monomer containing the AcuK DBD is partially functional. C. The C-terminal region of S. cerevisiae Rds2 can substitute for AcuM sequences but Rds2 cannot replace AcuM. Strains were generated as described in Fig. 8 and double mutants obtained by crossing. The designation of each variant is shown as the predicted protein product. Growth was for 2 days. Glucose, acetate, proline, ethanol and antimycin containing media were as described in Figs 6 and 7. Quinate (0.25%), butyrate (10 mM), oleate (10 mM) and glutamate (50 mM) were added as sole carbon sources to minimal medium.

Because the N-terminal DBD regions of both AcuK and AcuM are essential, it was of interest to investigate the effects of swapping DBDs between the proteins (Fig. 9B). When the DBDs were swapped between the proteins no function was observed. This indicated that the protein context of the DBDs is essential for binding. When the AcuK DBD was present in both proteins partial function for growth on gluconeogenic carbon sources was observed; while complete loss of function was observed when the AcuM DBD was present in both proteins (Fig. 9B).

Three S. cerevisiae genes encode proteins related to AcuK and AcuM. Rds2, which has high similarity to AcuM (Fig. S1), has been shown to regulate some gluconeogenic genes (Soontorngun et al., 2007). Replacement of acuM with the complete RDS2-coding sequence resulted in complete loss of function (Fig. 8). However a gene fusion in which RDS2 sequences corresponding to aa 256–446 replaced acuM sequences corresponding to the conserved C-terminus (aa 325–522) was functional for carbon source utilization and this was acuK dependent (Fig. 9C). Therefore the function of the AcuM C-terminus is conserved in Rds2. The products of the ERT1 and GSM1 genes have been reported to be involved in controlling gluconeogenic gene expression (Turcotte et al., 2010). Ert1 is similar to AcuK while Gsm1, although clearly related to AcuK, is more divergent (Fig. S1). Replacement of acuK with the coding regions of these genes did not result in function and gene fusions replacing acuK sequences corresponding to the C-terminus (aa 438–701) with corresponding C-terminal sequences from either Ert1 or Gsm1 were also non-functional (Fig. 8). Crosses between relevant strains showed that combining RDS2 and ERT1 gene replacements or acuMRDS2 and acuK–ERT fusion genes did not result in function.

Yeast two-hybrid analysis of interaction between the C-terminal regions of AcuK and AcuM

The above EMSA and ChIP result indicated that AcuK and AcuM bind DNA as a heterodimer, and that the N-terminal DBD is sufficient to determine this (at least in vitro). The AcuK and AcuM C-terminal regions contain sequences related to PAS domains which are commonly involved in protein–protein interactions (Partch and Gardner, 2010; Henry and Crosson, 2011). Therefore we used yeast two-hybrid analysis to determine whether AcuK and AcuM C-terminal sequences, excluding the N-terminal DBDs, could also interact. Strong interactions were observed between all combinations of AcuK144–701 and AcuM126–522 (Fig. 10). Interestingly, on both SD medium (Fig. 10B) and rich YPD medium (not shown), growth of strains transformed with only a single plasmid type was reduced compared with the recipient and also to strains containing both plasmids. This indicated that overexpression of these AcuK and AcuM fusion proteins was deleterious; possibly because of interactions with one or more endogenous S. cerevisiae proteins or by inappropriate DNA binding. These results showed that the C-terminal regions strongly interact and may contribute to the stable interaction between the full-length proteins and DNA.

Figure 10.

Yeast two-hybrid analysis of interactions between the C-terminal regions of AcuK and AcuM. A. Key to S. cerevisiae strains spotted on to plates (5 µl of a dense suspension). The no plasmid strain was the AH109 recipient for plasmid transformation. AD and DBD designate inserts into the plasmids pGADT7 (GAL4 activation domain) and pGBKT7 (GAL4 DNA-binding domain) respectively of cDNA sequences encoding aa 126–522 of AcuM and 144–701 of AcuK. B. Growth on non-selective medium. C. Formation of blue colonies on X-Gal containing medium (SD + trp + leu + his). D. Histidine prototrophy resulting from interaction between sequences fused to the GAL4 activation domain and the GAL4 DNA-binding domain. trp, tryptophan; leu, leucine; his, histidine and all media contained adenine.


The AcuK and AcuM transcription factors are each essential for the growth of A. nidulans on gluconeogenic carbon sources and are required for the induction of PCK. We have now found that these proteins bind interdependently to DNA, presumably as heterodimers, both in vitro and in vivo. In agreement with in vitro results for the N. crassa orthologues (Chae et al., 2007), the binding site consensus is direct repeats of CCG separated by seven bases. Two motifs in opposite orientation in the 5′ region of acuF have been found to act synergistically in induction. AcuK and AcuM bind in vivo to a sequence containing the CCGN7CCG motif upstream of acuG, encoding FBP, an enzyme specific for gluconeogenesis, and the promoters of many genes that respond to growth on gluconeogenic carbon sources contain the motif.

Zn cluster proteins commonly bind as dimers to CCG triplets separated by a fixed number of bases with each DBD contacting the triplet (MacPherson et al., 2006). Binding can be symmetrical to inverted or everted CGG repeats or asymmetrically to direct repeats as found for Hap1 in S. cerevisiae (King et al., 1999) and here for AcuK and AcuM. In Aspergillus spp. two AmyR subunits bind cooperatively to a CGGN8A/CGG motif but a single triplet can be bound with low affinity, at least in vitro (Tani et al., 2001). Binding specificity is usually determined by linker regions of variable length usually immediately C-terminal to the Cys6 sequence (MacPherson et al., 2006). It is noteworthy that all AcuK orthologues have a highly conserved sequence of 11 amino acids extending from the sixth cysteine while there are only three conserved amino acids in the equivalent position in AcuM orthologues (see Fig. S1). Conservation found in the CCGN7CCG motifs upstream of orthologous genes in Aspergillus spp. suggests that binding affinities are also determined by flanking sequences as well as by the spacer sequence (see Table S1).

Dimerization of Zn cluster proteins is usually determined by heptad repeats forming a coiled coil C-terminal to the linker region (MacPherson et al., 2006). Both AcuK and AcuM and their homologues lack heptad repeats. However, in pull down and size-exclusion chromatography experiments using expressed tagged proteins containing just the N-terminal sequences used in DNA binding experiments, it has been shown that the N. crassa orthologues preferentially form heterodimers (Chae and Nargang, 2009). Therefore the N-terminal sequences containing the DBDs are sufficient for dimerization. It remains to be determined whether pre-existing heterodimers bind to DNA or dimerization occurs on the DNA in vivo. The partial function of the putative heterodimer with each monomer containing the AcuK DBD (Fig. 9B) might indicate that the AcuK DBD initially contacts the binding site. The AcuM DBD may then bind to allow the formation of a stably bound heterodimer. A weaker interaction might be possible when an AcuK DBD is present in AcuM. This implies directionality of binding to the DNA.

AcuK and AcuM lack middle homology domain sequences found in many Zn cluster proteins (Schjerling and Holmberg, 1996; MacPherson et al., 2006). In general this domain is thought to modulate activity by masking C-terminal activation domains in the absence of effector molecules. Activation occurs by effector molecule binding resulting in a conformational change that enables the function of C-terminal activation domains. Instead AcuK and AcuM have C-terminal sequences, highly conserved in orthologues, comprising approximately 40% of the proteins, which are related to each other but clearly form two distinct classes (Fig. S1). Within these sequences are PAS domains, found in a diverse range of proteins, including transcription factors (Partch and Gardner, 2010) and sensor histidine kinases, (Henry and Crosson, 2011) and see Fig. S1. Although divergent in primary sequence they have a conserved secondary structure. Important for their potential function in AcuK and AcuM, these domains are frequently involved in protein–protein interactions, including dimerization and interaction with co-activators, and can bind a wide range of ligands (Partch and Gardner, 2010; Henry and Crosson, 2011). We have shown strong yeast two-hybrid interactions between all combinations of the C-terminal sequences of AcuK and AcuM (Fig. 10). It is therefore likely that AcuK and AcuM bind to DNA as a heterodimer determined by their N-terminal sequences and this allows a strong association of the C-terminal sequences. There is strong evidence that the heterodimerization of the PAS domain regions to form a hybrid structure is essential for function. Even small deletions of sequences in either protein abolish function (Fig. 8) and, furthermore, the acuMG418A and acuKG535A mutations in the PAS domains have weak effects by themselves but a more drastic loss of function in the double mutant (Fig. 9A). In P. anserina, single amino acid substitutions of residues in the C-terminal region of AcuK and AcuM orthologues result in deregulated expression of PCK, FBP as well as AOX genes (Sellem et al., 2009). These amino acids are conserved in A. nidulans (see Fig. S1). On the basis of these observations we propose that dimerization of the AcuK and AcuM C-terminal regions results in a structure essential for gene activation.

Common mechanisms used to control the activity of transcription factors are one or more of regulated expression, nuclear localization, ligand binding and post-translational modification such as phosphorylation. AcuK and AcuM nuclear localization is not affected by the carbon source and protein expression is only moderately increased in the presence of gluconeogenic carbon sources. Furthermore, ChIP analysis shows that AcuK and AcuM bind together in the 5′ regions of the acuF, acuG and aodA genes independently of the carbon source. Therefore activation by these transcription factors is determined by signalling to the AcuK/AcuM heterodimer in response to the carbon source. Preliminary analysis of ChIP Seq data indicates that virtually all binding sites genome-wide are occupied by both AcuK and AcuM during growth on glucose, proline or acetate. The formation of active homodimers or heterodimers with different DNA binding sequences or affinities by two or more transcription factors is a common strategy to increase the diversity and flexibility of regulation of sets of downstream genes (e.g. Akache et al., 2004). Our evidence strongly indicates that this scenario does not apply for AcuK and AcuM i.e. binding and activation only occurs as the heterodimer.

A major question then is what is the signal for AcuK/AcuM activation? Proline and glutamate metabolism to 2-oxoglutarate and acetate metabolism via acetyl-CoA and the glyoxylate cycle are required for induction of PCK expression (Hynes et al., 2002; 2007). Induction is increased in an acuF mutant background indicating that oxaloacetate accumulation enhances expression (Hynes et al., 2007). However cytoplasmic MDH activity can convert oxaloacetate to malate and therefore we examined the effects of deleting mdhC to block this activity. Induction by acetate and proline of acuF expression was higher in the mdhCΔ background suggesting that malate rather than oxaloacetate results in induction. Support for this proposal is provided by the expression level of acuF–lacZ on glucose being more than 20-fold reduced by the mdhCΔ, a result we attribute to growth on glucose resulting in endogenous induction by conversion of oxaloacetate to cytoplasmic malate by MdhC. While this may be partly attributed to the accumulation of β-galactosidase activity in the reporter assays, the question of glucose repression of acuF expression is raised. In the presence of glucose, proline can act as the sole nitrogen source by conversion to 2-oxoglutarate and it was previously shown that these conditions result in increased expression of acuF–lacZ and acuF RNA above the levels observed on glucose with ammonium as the nitrogen source (Hynes et al., 2002; 2007). Therefore acuF can be expressed under inducing conditions even in the presence of glucose; although these levels are not as high as when glucose is absent. We suspect that there is a mechanism whereby growth on glucose can result in a repression mechanism which can be partially overcome when inducer accumulates. It should be noted that for acuG (encoding FBP) standard CreA mediated glucose repression of basal levels occurs with AcuK/M-mediated induction occurring on gluconeogenic carbon sources (Hynes et al., 2007). Expression of aodA is observed on glucose, but not under carbon starvation conditions, and is absolutely dependent on AcuK and AcuM. The ChIP analysis indicates a high level of binding by AcuK and AcuM to the 5′ of aodA which might allow significant expression even in the absence of added gluconeogenic carbon sources.

Overall our results suggest that malate accumulation results in induction. How does this explain induction of expression of AOX genes by inhibitors or mutations affecting the respiratory chain and by inhibitors of mitochondrial protein synthesis mediated by AcuK and AcuM orthologues in N. crassa and P. anserina (Chae et al., 2007; Sellem et al., 2009)? Antimycin was observed to induce PCK and FBP in P. anserina (Sellem et al., 2009) and we have observed chloramphenicol induction of acuF–lacZ expression. A probable explanation is that reduced activity of the electron transport chain results in accumulation of NADH leading to lower malate dehydrogenase activity and therefore malate accumulation during growth on glucose. Organic acids are one class of the many ligands of PAS domains (Henry and Crosson, 2011). In particular malate is a ligand for the PAS domain of the DcuS sensor histidine kinase widely distributed in bacteria (Golby et al., 1999). As noted above the conserved C-terminal domains of both AcuK and AcuM are required for activity and are capable of forming a heterodimer. Therefore it is suggested that this forms an activation domain with the activity modulated by the binding of malate to one or both of the PAS domains. It is also possible that a second ligand such as succinate also binds to one of the PAS domains. Detailed studies on purified proteins are required.

The phenotype of the mdhCΔ surprisingly revealed that cytoplasmic and peroxisomal MDH activity is neither required for the glyoxylate cycle nor required for gluconeogenesis. Clearly in A. nidulans mitochondrial MDH activity converting malate to oxaloacetate is sufficient both for the glyoxylate cycle and for gluconeogenesis indicating that oxaloacetate synthesized in the mitochondrion can enter the cytoplasm to act as the substrate for PCK. This contrasts with S. cerevisiae where cytoplasmic Mdh2 is required for growth on acetate and ethanol (Minard and McAlister-Henn, 1991; Steffan and McAlister-Henn, 1992; Steffan et al., 1992) and there is evidence for physical interaction between Mdh2, Pck1 and Fbp1 via an N-terminal 12-amino-acid extension forming a cytoplasmic complex channelling gluconeogenesis (Gibson and McAlister-Henn, 2003). MdhC lacks an equivalent N-terminal sequence.

The observation that AcuK and AcuM constitutively bind to DNA raises the question of whether they can also act to reduce the expression of some genes. For example this could occur by interfering with the binding or activity of other transcription factors. In A. fumigatus deletion of acuM has been found to result, not only in an inability to grow on gluconeogenic carbon sources, but also defects in iron metabolism and pathogenicity (Liu et al., 2010). The A. fumigatus acuMΔ was defective in growth on iron-deficient medium in contrast to the A. nidulans acuMΔ (Liu et al., 2010). The sreA gene, which encodes a repressor of genes required for iron acquisition under iron-sufficient conditions (Schrettl et al., 2008), was derepressed in the acuMΔ mutant under iron-deficient conditions. A CCGN7CCG motif is present at −74 (relative to the start codon) in the sreA gene of A. fumigatus and its close relative Neosartorya fischeri but is not present in the 5′ sequences of the A. niger, A. nidulans or A. oryzae genes. This raises the interesting speculation that binding of the AcuK/AcuM heterodimer results in repression of sreA in A. fumigatus but not in other Aspergillus species. Because many of the TCA cycle functions involve iron containing proteins (e.g. cytochrome c), the relationship between iron availability and AcuK and AcuM regulation is of future interest.

The S. cerevisiae proteins, Gsm1 and Rsd2, are orthologues of AcuK and AcuM – although the DBDs are more divergent (Fig. S1). For Rds2, binding to genes related to gluconeogenesis has been observed in ChIP studies with microarrays (Soontorngun et al., 2007). Rds2 is phosphorylated by Snf1 and, for most sites, binding is stronger when ethanol is the carbon source. Furthermore, in experiments with expressed Rds2 DBD protein, in vitro binding to the same CSRE sites as Cat8 in the PCK and FBP promoters has been observed (Soontorngun et al., 2007). Therefore it appears that Rds2 has an overlapping binding spectrum to Cat8 and this is Snf1 dependent. However in ChIP experiments Rds2 binding to the PCK promoter is detected in glucose-grown cells and is only partially dependent on Cat8 and Snf1 (Soontorngun et al., 2007). The promoter of PCK1 has two CCGN7CCG motifs (Table S3) and furthermore a two-hybrid interaction between Rds2 and Gsm1 has been found (Ito et al., 2001). Therefore for PCK1, Gsm1 and Rds2 may bind together to the CCGN7CCG motifs to contribute to regulation. For FBP1, binding of Rds2 in vivo is almost completely dependent on Snf1 and Cat8 (Soontorngun et al., 2007) and a CCGN7CCG motif is present as well as CSRE sites (Table S1). Gsm1/Rds2 binding dependent on Snf1 phosphorylated Cat8 at many promoters is likely. Our observation that the conserved C-terminal region of Rds2 can substitute for that of AcuM (Fig. 9C) supports conservation of function. The Gsm1 C-terminal region could not functionally replace that of AcuK – perhaps due to a more divergent sequence. Replacement of AcuK- and AcuM-coding sequences with those of Gsm1 and Rds2 did not permit growth on gluconeogenic carbon sources. This may be due, in part, to divergent DBDs: although we did not establish that these proteins were expressed. Ert1 is related to AcuK and has been proposed to also affect gluconeogenic metabolism (Turcotte et al., 2010). This protein has been found to bind to a CCGN8CCG motif in vitro (Ho et al., 2006). It should be noted that Rds2, Gsm1 and Ert1 are proposed to regulate the expression of Hap4, the activating subunit of the Hap complex which controls TCA cycle and respiratory gene expression (Turcotte et al., 2010). Hap4 does not have an orthologue in A. nidulans. Overall the situation for the regulation of gluconeogenesis in S. cerevisiae appears to be complex with many unresolved questions (Turcotte et al., 2010). C. albicans can grow on amino acids as well as sources of acetyl-CoA as gluconeogenic carbon sources (Fleck et al., 2011) and amino acids can result in PCK induction (Leuker et al., 1997). In heterologous studies the expression of C. albicans FBP, ICL and MLS but not PCK was found to be lost in the S. cerevisiae cat8Δ (Eschrich et al., 2002). C. albicans PCK has a CCGN7CCG motif (Table S3) and therefore this result might be explained by regulation by S. cerevisiae Rds2 and Gsm1 independently of Cat8. A puzzling observation for C. albicans, however, is that deletion of both copies of CAT8 neither affects growth on acetate, ethanol or fatty acids nor affects expression of ICL or FBP in response to acetate (Ramirez and Lorenz, 2009).

It is clear that AcuK and AcuM are essential for the reprogramming of metabolism necessary for adaptation to growth on carbon sources metabolized via TCA cycle intermediates. They are essential for expression of PCK and upregulation of FBP, the unique enzymes of gluconeogenesis, as well as for expression of the reversible enzymes of glycolysis/gluconeogenesis during growth on gluconeogenic carbon sources. There is also strong evidence that they upregulate the expression of TCA cycle enzymes and the respiratory enzymes allowing an increased flux to generate sufficient carbon precursors for the synthesis of sugars. During gluconeogenesis, AOX might play a role in partially bypassing the final steps of the electron transport chain thereby maintaining a high flux through the cycle and reducing excessive ATP and reactive oxygen formation. Given the complete dependence on AcuK and AcuM for aodA expression, it might have been expected that this would contribute to glucononeogenesis. However we detected no effect of the aodAΔ on growth on gluconeogenic carbon sources. It should be noted that loss of function mutations in either acuK or acuM abolish the expression of NADP-dependent malic enzyme encoded by maeA (McCullough and Roberts, 1974; Armitt et al., 1976) and CCGN7CCG motifs are conserved in the upstream region of this gene in Aspergillus spp. (Table S1). Studies on the role of this enzyme activity will be the subject of a future communication. An overview of the transcriptional control of the utilization of gluconeogenic carbon sources by A. nidulans is presented in Fig. 11. In the absence of repressing glycolytic carbon sources, specific transcription factors activate the expression of pathways leading to TCA cycle intermediates with AcuK and AcuM controlling gluconeogenesis and the TCA cycle, independent of the particular carbon source.

Figure 11.

Overview of the transcriptional regulation of the utilization of gluconeogenic carbon sources in A. nidulans. AcuK and AcuM are required for the expression of genes specific for gluconeogenesis and increase the expression of TCA cycle genes. Pathway-specific transcription factors activate the expression of genes required for the production of TCA cycle intermediates in response to induction by specific carbon sources. FacB, in response to acetate, controls the production of acetyl-CoA and its entry via acetyl-carnitine into the mitochondria as well as the glyoxylate cycle (Todd et al., 1998; Hynes et al., 2011). FarA and FarB, in response to fatty acids, controls β-oxidation, acetyl-CoA entry via acetyl-carnitine into the mitochondria as well as the glyoxylate cycle (Hynes et al., 2006). AlcR activates the conversion of ethanol via acetaldehyde to acetate (Felenbok et al., 2001), AmdR activates the conversion of GABA to succinate and 2-oxoglutarate (Andrianopoulos and Hynes, 1990), QutA activates conversion of quinate to succinate and acetyl-CoA (Wheeler et al., 1996) and PrnA activates the conversion of proline to 2-oxoglutarate in response to proline (Gómez et al., 2002). Not shown is carbon catabolite repression mediated by the C2H2 zinc finger repressor, CreA, which represses the expression of the enzymes of the specific pathways and/or the pathway-specific activators when strong glycolytic carbon sources such as glucose are present (Dowzer and Kelly, 1991; Kulmburg et al., 1993; Cubero and Scazzocchio, 1994).

Consistent with the presence of AcuK and AcuM orthologues, the CCGN7CCG motif was found upstream of many relevant genes analysed in a range of fungal species (Table S3). Notably this was found for the basidiomycete, Cryptococcus neoformans, and the chytrid, S. punctatus. The motif was also found in the 5′ of genes for the key gluconeogenic enzymes, PCK and FBP, in the Mucor, Rhizopus oryzae (not shown). Clearly there is strong circumstantial evidence that the role of AcuK and AcuM in gluconeogenic carbon metabolism has been highly conserved over more than 500 million years of fungal evolution (Stajich et al., 2009). Presumably the genes result from an extremely ancient gene duplication and have been maintained through fungal evolution. The importance of alternative carbon metabolism in fungal development and pathogenesis (Lorenz and Fink, 2002; Wilson and Talbot, 2009; Fleck et al., 2011; Kronstad et al., 2012) suggests that the study of the role of these transcription factors in a wide range of fungi will be fruitful.

Experimental procedures

Media and growth conditions and molecular and genetic techniques

Media and conditions for growth of A. nidulans were as described previously (Cove, 1966; Todd et al., 2007). Antimycin A (Sigma A8674) and chloramphenicol (Boehringer–Mannheim) were dissolved as stock solutions in methanol. Mycelia for enzyme assays, RNA and DNA preparations were grown in 100 ml of medium in 250 ml Ehlenmeyer flasks at 37°C. β-Galactosidase assays were carried out as described (Davis et al., 1988). A. nidulans crosses were as previously described (Todd et al., 2007). Standard methods for DNA manipulations, RNA isolation, nucleic acid blotting and hybridization have been described (Sambrook and Russell, 2001; Hynes et al., 2006). A list of oligonucleotide primers is presented in Supporting information (Table S4).

A. nidulans strains and transformation

All strains were derived from the original Glasgow strain and contained the veA1 mutation. The genotypes of key strains are provided in Supporting information. Preparation of protoplasts and transformation selection markers were as described previously (Andrianopoulos and Hynes, 1988; Nayak et al., 2006). In many cases recipient strains contained the nkuAΔ to promote homologous integration events (Nayak et al., 2006). DNA from transformants was analysed by Southern blotting to confirm predicted integration events. In addition crossing was used to confirm segregation of markers and phenotypes as single genes and, where appropriate, double mutant strains were confirmed by out-crosses. DNA sequencing of plasmid constructs was used where necessary to confirm predicted junctions and sequence changes.

Details of plasmid constructions and selection of transformants are provided in Supporting information.

DNA sequence analysis

Aspergillus spp. sequences were obtained from the genome sequences available at or Other fungal sequences were derived from specific genome sequences available at either the Broad Institute (, NCBI ( or the Fungal Orthogroups site ( Analysis of sequences in promoters used tools available at (van Helden, 2003).

Western blotting

Strains were grown for 16 h at 37°C in 1% glucose minimal liquid medium and then transferred to minimal medium containing 1% glucose, 50 mM proline, 50 mM acetate with 10 mM ammonium chloride as the nitrogen source for a further 6 h. Total protein extraction and Western blotting were as previously described (Todd et al., 2005). Fifty micrograms of total protein from the samples was separated by 8% SDS-PAGE. Myc was detected with anti-Myc (9B11 from Cell Signalling) as the primary antibody and anti-mouse IgG-horseradish peroxidase (Promega Corp.) as the secondary antibody. Anti-α-tubulin was used to detect α-tubulin as a loading control as previously described (Murray and Hynes, 2010). Signals were detected using Fujifilm image reader LAS-3000 (Berthold Australia Pty).


Strains were grown on coverslips for 16 h at 37°C in 1% glucose minimal liquid medium and then transferred to minimal medium containing 1% glucose, 50 mM proline, 50 mM acetate with 10 mM ammonium chloride as the nitrogen source for a further 4 h. Fixation and microscopy were as previously described (Szewczyk et al., 2001).

Expression of MBP fusion proteins and EMSA

MBP–AcuK(1–123) (pYS6598) was expressed in the E. coli Rosetta strain (Novagen) and MBP (pMAL-c2) and MBP–AcuM(1–124) (pYS6600) in E. coli BL21-DE3. Extracts were prepared from IPTG induced cultures as previously describe (Hynes et al., 2006). Protein concentrations were determined using the DC protein assay kit (Bio-Rad) and induced expression of proteins monitored by 10% SDS-PAGE analysis. EMSA was performed using total crude protein extracted in binding buffer without purification.

Probes derived from the acuF 5′ sequence were used. Three different truncation fragments and a CCG deletion mutation of site A were generated. The fragments F2, F3 and F4 were prepared by PCR using a plasmid OD002 as a template. Primers sets used were pacuF-5′/pacuF-3′-b for F2, pacuF-5′/pacuF-3′-c for F3 and pacuF-5′-b/pacuF-3′-b for F4 (Table S4). OD002 contains −468 to +1718 of the acuF gene. PCR fragments were cloned into pBluescript SK+ at the EcoRV site to give rise to pYS7206 (F2), pYS7208 (F3) and pYS7209 (F4). Inverse PCR with primers pacuF-CCGd-5′WT/pacuF-CCGd-3′ was performed on pYS7206 and pYS7209 to create a deletion of CCG (−72 to −70) of site A to give pYS7207 (F2Amut) and pYS7210 (F4Amut) respectively. To create EMSA probe fragments the plasmids pYS7206 (F2), pYS7208 (F3) and pYS7209 (F4) were digested with EcoRI and HindIII, labelled by end filling with 1 unit of DNA polymerase I Klenow fragment and 20 µCi of [α-32P]-dATP. The unincorporated [α-32P]-dATP was removed using SigmaSpinTM Post Reaction Clean Up Columns (Sigma). In addition, three double-stranded oligonucleotide probes were used. Foligo1, Foligo2 and Foligo1mut. Annealing of oligonucleotides (see Table S4) and labelling by end-filling were as described (Hynes et al., 2006). Binding reactions were carried out as previously described with the addition of 1 µg of poly(dI-dC) per binding reaction (Todd et al., 1998; Hynes et al., 2006). Electrophoresis of samples and autoradiography was as described (Hynes et al., 2006).

Chromatin preparation and immunoprecipitation (ChIP)

Mycelia was cross-linked with 1% formaldehyde for 20 min at room temperature with gentle mixing, harvested and press-dried with paper towel. Approximately 50 mg of pressed-dried mycelia was lyophilized and subsequently lysed twice for 5 min (or until most mycelia clumps were disrupted) in 1 ml of ice-cold FA lysis buffer (Aparicio et al., 2004) using Zirconia beads and a Bullet Blender® Homogenizer with a speed setting of 10. The lysate was separated from the beads and cross-linked chromatin was pelleted at 4°C by centrifugation at 14 000 r.p.m. for 15 min. The pellet was resuspended and sonicated in 0.5 ml of ice-cold FA lysis buffer using a Misonix Sonicator 3000 (Cole-Parmer Instrument Company, Vernon Hills, IL, USA) equipped with a horn and a rotating device. Sonication was performed in an ice-water bath, in pulses of 10 s with an output setting of 6 and with 10 s cooling time in between for a total time of 20 min. The supernatant containing chromatin fragments with an average size of ∼ 200 bp was recovered by centrifugation at 4°C for 30 min. Chromatin immuno-precipitation was performed as described previously (Fan et al., 2008) using 50 µl of the respective chromatin supernatant (dilute to 0.5 ml volume with FA lysis buffer) with 2 µl of anti-HA (F7, Santa Cruz Biotechnology) or 2 µl of anti-Myc (9E10, Santa Cruz Biotechnology). Quantitative real-time PCR was performed using the following gene-specific primers: acuF (CCCGATCCTTCATTCTTGTT and AATAGACCCGAACCCCAGAT); acuG (TTCTCCCGCGACAATTATTC and TTCCCCGCGATCATATTCTA); aodA (GTCGATAGGGCTACGCAAAC and CCACTAACTGGGCCATGACT).

Yeast two-hybrid analysis

The Matchmaker GAL4 Two-Hybrid System 3 kit from Clontech was used. cDNA sequences encoding aa 126–522 of AcuM and 144–701 of AcuK were each inserted into the vectors pGBKT7 and pGADT7 in-frame with sequences encoding the GAL4 DBD and sequences encoding the GAL4 activation domain respectively (see Supporting information). These were transformed into the S. cerevisiae strain AH109 (genotype: MATa, trp1-901, leu2-3, ura3-52, his3-200, gal4Δ, gal80Δ, LYS2::GAL1UAS-GAL1TATA-HIS3, GAL2UAS-GAL2TATAADE2, URA3::MEL1UAS-MEL1TATA–lacZ) according to the Clontech Yeast Protocols Handbook, selecting for tryptophan prototrophy (inserts in pGBKT7), leucine prototrophy (pGADT7) or both. Interactions in transformants were screened for by the appearance of blue colonies on X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) containing medium and by histidine prototrophy. Standard SD and YPD media were used.


This work was supported by a grant from the Australian Research Council and a postdoctoral fellowship to K.H.W. from the Croucher Foundation (Hong Kong). Andrea Prynych is thanked for generating the yeast transformants.