OUP user menu

Transcriptional responses of Saccharomyces cerevisiae to preferred and nonpreferred nitrogen sources in glucose-limited chemostat cultures

Viktor M. Boer, Siew Leng Tai, Zeynep Vuralhan, Yalun Arifin, Michael C. Walsh, Matthew D.W. Piper, Johannes H. De Winde, Jack T. Pronk, Jean-Marc Daran
DOI: http://dx.doi.org/10.1111/j.1567-1364.2007.00220.x 604-620 First published online: 1 June 2007


Aerobic, glucose-limited chemostat cultures of Saccharomyces cerevisiae grown with six different nitrogen sources were subjected to transcriptome analysis. The use of chemostats enabled an analysis of nitrogen-source-dependent transcriptional regulation at a fixed specific growth rate. A selection of preferred (ammonium and asparagine) and nonpreferred (leucine, phenylalanine, methionine and proline) nitrogen sources was investigated. For each nitrogen source, distinct sets of genes were induced or repressed relative to the other five nitrogen sources. In total, 131 such ‘signature transcripts’ were identified in this study. In addition to signature transcripts, genes were identified that showed a transcriptional coresponse to two or more of the six nitrogen sources. For example, 33 genes were transcriptionally upregulated in leucine-grown, phenylalanine-grown and methionine-grown cultures; this was partly attributed to the involvement of common enzymes in the dissimilation of these amino acids. In addition to specific transcriptional responses elicited by individual nitrogen sources, their impact on global regulatory mechanisms such as nitrogen catabolite repression (NCR) were monitored. NCR-sensitive gene expression in the chemostat cultures showed that ammonium and asparagine were ‘rich’ nitrogen sources. By this criterion, leucine, proline and methionine were ‘poor’ nitrogen sources, and phenylalanine showed an ‘intermediate’ NCR response.

Key words
  • Saccharomyces cerevisiae
  • chemostat
  • transcriptome
  • amino acid
  • nitrogen catabolite repression


Its ability to use a broad range of nitrogen sources (Large, 1986; ter Schure, 1998; Barnett, 2003) has made Saccharomyces cerevisiae a popular model for studying nitrogen-source-dependent regulation. In S. cerevisiae, growth and gene expression strongly depend on the identity and concentration of the nitrogen source. Furthermore, carbon skeletons derived from several amino acids are converted to fusel alcohols (Ehrlich, 1907), which are key contributors to flavor in beer and wine fermentations (Henschke & Jiranek, 1993; Mauricio, 2001; Hernandez-Orte, 2002; Bartowsky & Henschke, 2004).

On the basis of the specific growth rate in glucose-containing media, nitrogen sources are classified as ‘good’ (‘rich’, ‘preferred’) or ‘poor’ (‘nonpreferred’) (Castor, 1953; Watson, 1976; Magasanik & Kaiser, 2002). Good nitrogen sources are generally easily converted into glutamate and glutamine (Watson, 1976; Henschke & Jiranek, 1993; Hofman-Bang, 1999), with glutamine, asparagine and ammonium being prominent examples (Watson, 1976; Cooper, 1982). Leucine and phenylalanine are considered to be ‘average’ nitrogen sources, whereas proline, methionine, γ-aminobutyrate and allantoin exemplify typical poor nitrogen sources (Watson, 1976; Cooper, 1982; Henschke & Jiranek, 1993; Hofman-Bang, 1999; Magasanik & Kaiser, 2002).

In mixed-substrate cultures, S. cerevisiae exhibits a sequential use of good, average and poor nitrogen sources (Castor, 1953; Henschke & Jiranek, 1993). This sequential use is controlled by a transcriptional regulation mechanism known as nitrogen catabolite repression (NCR) (Messenguy & Cooper, 1977; Cooper & Sumrada, 1983; Cooper, 2002; Magasanik & Kaiser, 2002). When a good nitrogen source is present, NCR shuts down pathways for the use of nonpreferred nitrogen sources (Cooper, 1982, 2002; Cooper & Sumrada, 1983; Magasanik & Kaiser, 2002). NCR is mediated by upstream activating sequences that contain the GATAA motif and a four-member family of GATA-binding transcription factors (Fig. 1): Gln3p, Gat1p, Dal80p, and Gzf3p (Cunningham & Cooper, 1991; Blinder & Magasanik, 1995; Stanbrough, 1995; Cunningham, 1996; Magasanik & Kaiser, 2002). In the presence of a rich nitrogen source, Gln3p and Gat1p form cytosolic complexes with Ure2p, which prevents activation of NCR-sensitive transcription (Fig. 1). In the absence or at limiting concentrations of a rich nitrogen source, Gln3p and Gat1p are dephosphorylated and targeted to the nucleus, where they activate transcription of NCR-sensitive genes. Tor kinases play a key role in the signal transduction pathway that couples nitrogen status to Gln3p and Gat1p phosphorylation (Beck & Hall, 1999; Bertram, 2000; Shamji, 2000; Kulkarni, 2001; Crespo & Hall, 2002; Crespo, 2002, 2004).


Mechanism of NCR. Tor kinases directly and/or indirectly regulate the localization of the transcription factors Gln3p and Gat1p in response to glutamine and glutamate concentrations, respectively (Beck & Hall, 1999; Bertram, 2000; Crespo, 2004). Black lines indicate processes active during preferred nitrogen source conditions, and white lines indicate processes active during nonpreferred nitrogen source conditions. Interactions in the nucleus indicate regulation of transcription; interactions outside the nucleus indicate allosteric regulation. Adapted from Coffman & Cooper (1997).

With amino acids as the nitrogen source, NCR is complemented by specific regulation mechanisms. Most extracellular amino acids are sensed via the trans-plasma-membrane SSY1-PTR3-SSY5 (SPS) complex (Forsberg, 2001). Metabolism of some amino acids is regulated by specific transcriptional regulators, such as Leu3p for leucine (Friden & Schimmel, 1987), and Aro80p for aromatic amino acids (Iraqui, 1999).

Research on nitrogen-source-dependent transcriptional regulation has mainly been performed in batch cultures (Watson, 1976; Cooper, 1982; Henschke & Jiranek, 1993; Stanbrough & Magasanik, 1995). In batch cultures, all nutrients are at least initially present in excess, and parameters such as dissolved oxygen concentration, metabolite concentrations and pH change over time. Furthermore, different nitrogen sources support different specific growth rates. Various physiologic parameters, including, for example, amino acid pools (Watson, 1976) and protein and RNA levels (Ertugay & Hamamci, 1997; Regenberg, 2006), are strongly influenced by specific growth rate. It has been demonstrated that specific growth rate has a strong impact on transcriptional regulation (Regenberg, 2006). Consequently, shake flask cultures do not allow for discrimination between direct effects of nitrogen sources on transcriptional regulation and indirect, growth-rate-dependent phenomena.

Chemostat cultivation is a useful tool for genome-wide transcription studies (Piper, 2002; Boer, 2003; Saldanha, 2004; Bro, 2004; Tai, 2005; Usaite, 2006), primarily because it allows maintaining a specific growth rate and other important culture parameters to be kept constant. This elimination of specific growth rate as an experimental variable should facilitate analysis of transcriptional responses to different nitrogen sources. For example, transcriptional analysis in chemostats should enable a separation of the role of Tor kinases in NCR (Shamji, 2000) from their involvement in cell growth control (Thomas & Hall, 1997; Powers & Walter, 1999; Schmelzle & Hall, 2000; Jorgensen, 2004).

The present study was aimed at investigating unique and shared transcriptional responses of S. cerevisiae to different nitrogen sources, and establishing whether the nitrogen-source dependence of NCR is preserved in cultures grown at a fixed specific growth rate. To this end, glucose-limited chemostat cultures of S. cerevisiae grown on six different nitrogen sources (ammonium, asparagine, phenylalanine, leucine, methionine or proline) were subjected to transcriptome analysis.

Materials and methods

Strain and growth conditions

The prototrophic S. cerevisiae strain CEN.PK113-7D (MATa) (van Dijken, 2000; Daran-Lapujade, 2003) was grown at 30 °C in 1.0-L working volume chemostats as described previously (van den Berg, 1996). Cultures were fed with a synthetic medium (Verduyn, 1992) that supported glucose-limited growth. The glucose concentration in the reservoir medium was 7.5 g L−1. Different nitrogen sources were used, at the following concentrations: (NH4)2SO4, 5.0 g L−1; l-asparagine, 5.0 g L−1; l-phenylalanine, 5.0 g L−1; l-leucine, 10.0 g L−1; l-proline, 8.8 g L−1; and l-methionine, 11.3 g L−1. When an amino acid served as nitrogen source, the synthetic medium was supplemented with 6.6 g L−l K2SO4. The dilution rate was set at 0.10 h−1. The pH was measured online and kept constant at 5.0 by the automatic addition of 2 M KOH with the use of an Applikon ADI 1030 biocontroller. The stirrer speed was 800 r.p.m., and the airflow was 0.5 L min−1. The dissolved oxygen tension was measured online with an Ingold model 341003002 probe, and was above 50% air saturation. A condenser connected to a cryostat set at 2°C cooled the off-gas, and oxygen and carbon dioxide were measured offline. Steady-state samples for arrays were taken after 10–14 volume changes to minimize the impact of the evolutionary adaptation that occurs after long-term cultivation (Ferea, 1999; Jansen, 2004). Dry weight, metabolite profiles and gas profiles were constant over at least three volume changes prior to sampling for RNA extraction.

Analytical methods

Culture supernatants were obtained after centrifugation of chemostat samples and were analyzed by an HPLC set-up fitted with an AMINEX HPX-87 H ion exclusion column using 5 mM H2SO4 as the eluent. Culture dry weights were determined via filtration as described by Postma (1989). Total nitrogen and ammonium were determined with the use of DRLANGE cuvette tests (Düsseldorf, Germany).

Microarray analysis

Sampling of cells from chemostats, probe preparation and hybridization to Affymetrix GeneChip Microarrays was performed as described by Piper (2002). The results for each growth condition were derived from three independent chemostat cultures.

Data acquisition and statistical analysis

Acquisition and quantification of array images and data filtering were performed using the affymetrix software packages: microarray suite v5.0, microdb v3.0, and data mining tool v3.0. For comparison, all arrays were globally scaled to a target value of 150 using the average signal from all gene features, using microarray suite v5.0. To eliminate insignificant fold changes, absent genes (value below 12) were set to 12. The full dataset of the global transcriptome response of each culture condition can be downloaded from the Gene Expression Omnibus (GEO) database http://www.ncbi.nlm.nih.gov/geo/ (Barrett, 2006) under the accession series number GSE6405. For the identification of changed transcripts between conditions, Microsoft Excel running the software package significance analysis of microarrays (sam; v1.12) (Tusher, 2001) was used on all probe sets that were present in at least one condition. The statistical significance of observed differences was assessed by SAM, using an expected false discovery rate (FDR) of 1% and a twofold-change cut-off. For visualization of significantly changed genes, treeview v1.60 was used. For clustering of genes with similar transcriptional profiles with NCR ‘marker’ genes, genespring v6.02 was used. A confidence level of 0.80 was applied for all nine analyses (one analysis for each ‘marker’ gene), and genes were selected that met this criterion in at least five of the nine analyses. To calculate the significance of the differences in NCR strength between conditions, a Students' t-test was used with n−1 degrees of freedom to calculate the P-values. To represent the variation in the triplicate measurements, the average coefficient of variation was calculated as described previously for all conditions (Boer, 2003).

Results and discussion

Physiology of S. cerevisiae during growth on ammonium or amino acids

To verify our choice of good, intermediate and poor nitrogen sources, S. cerevisiae CEN.PK113-7D was grown in shake flasks on glucose with the six nitrogen sources selected for this study (Table 1). The specific growth rate was highest on asparagine, followed by ammonium. Phenylalanine and leucine gave significantly lower rates, and proline and methionine supported the lowest specific growth rates. These results are consistent with earlier studies in other S. cerevisiae strains (Castor, 1953; Watson, 1976; Cooper, 1982).

View this table:

Specific growth rates in batch cultures and physiologic parameters of chemostat cultures used in this study

Nitrogen sourceBatch μ (h−1)ChemostatResidual NH4+ (mM)
YsxqglucoseEmbedded ImageEmbedded ImageRQCarbon recovery (%)Residual N (mM)
Ammonium0.37 ± 0.010.49 ± 0.011.1 ± 0.02.8 ± 0.32.8 ± 0.31.0 ± 0.098 ± 358 ± 158.2 ± 1.3
Asparagine0.45 ± 0.000.63 ± 0.020.9 ± 0.02.7 ± 0.03.3 ± 0.01.2 ± 0.0100 ± 243 ± 215.7 ± 2.3
Phenylalanine0.28 ± 0.010.31 ± 0.031.7 ± 0.26.7 ± 0.76.7 ± 0.71.0 ± 0.0101 ± 428 ± 2< 1
Leucine0.26 ± 0.000.28 ± 0.032.0 ± 0.28.0 ± 1.07.7 ± 1.01.0 ± 0.093 ± 169 ± 0BD
Methionine0.20 ± 0.010.34 ± 0.011.6 ± 0.05.6 ± 0.15.5 ± 0.11.0 ± 0.095 ± 273 ± 2BD
Proline0.20 ± 0.010.79 ± 0.010.7 ± 0.03.0 ± 0.23.0 ± 0.21.0 ± 0.197 ± 547 ± 1BD
  • * Specific growth rate.

  • Biomass yield on glucose (gram biomass per gram of glucose consumed).

  • Millimoles of glucose consumed per gram of biomass per hour.

  • § Millimoles of oxygen consumed per gram of biomass per hour.

  • Millimoles of carbon dioxide produced per gram of biomass per hour.

  • RQ, respiratory quotient (Embedded Image).

  • ** BD, below limit of detection.

  • Unless indicated otherwise, data represent the average and SD of data from two independent batch cultivations or three independent steady-state chemostat cultivations.

To study quantitative physiology and genome-wide transcriptional regulation, S. cerevisiae CEN.PK113-7D was grown in aerobic glucose-limited chemostat cultures with either ammonium sulfate or one of the five amino acids as the nitrogen source. Under all conditions, glucose metabolism was fully respiratory, as no ethanol was detected. Consistent with the absence of alcoholic fermentation, all cultures except for those grown on asparagine as the nitrogen source showed a respiratory quotient (RQ) of 1.

Despite the absence of alcoholic fermentation, the biomass yield on glucose was remarkably different for the six nitrogen sources (Table 1). On ammonium, the biomass yield was 0.49 g (g glucose)−1, which is a well-established value for these conditions (Verduyn, 1992; Luttik, 1998; Van Maris, 2001). During growth on proline and asparagine, the biomass yield on glucose was much higher [0.79 and 0.63 g (g glucose)−1, respectively] (Table 1). This increased biomass yield is probably due to the assimilation of the carbon skeletons of these amino acids into biomass and/or their use as an additional energy source. The high RQ of the asparagine-grown cultures (Table 1) is consistent with the fact that complete oxidation of oxaloacetate, a degradation product from asparagine, requires only 2.5 O2 for the production of four CO2. Conversely, the biomass yields on glucose of cultures grown on phenylalanine, leucine and methionine were much lower than the biomass yield found in ammonium-grown cultures (Table 1). This reduced biomass yield, and the increased respiration rates in these cultures (Table 1), imply a high-energy requirement associated with the use of these compounds as nitrogen sources. More precisely, this energy requirement might be directly linked to the formation of long-chain acids originating from the carbon skeleton of these three amino acids (Vuralhan, 2005), resulting in an uncoupling effect due to the ATP requirement for the export of these acids (Hazelwood, 2006).

In all cultures, the nitrogen source was present in excess (Table 1). In the ammonium-grown and asparagine-grown cultures, free ammonium was detected in the culture supernatants (Table 1). In ammonium-grown cultures, this is merely the excess ammonium due to the medium design. In asparagine-grown cultures, ammonium is formed by a direct deamination reaction in the first step of asparagine degradation, which can take place both inside and outside the cell (Kim, 1988; Sinclair, 1994).

Microarray reproducibility and global transcriptome responses

To control reproducibility and data quality, three independent chemostat cultivations were performed for each growth condition. The average coefficient of variation for the triplicates of the six nitrogen sources was 0.21 or lower (Table 2). As an additional check, the transcript levels of ACT1 and PDA1, two commonly accepted Northern blot loading standards (Ng & Abelson, 1980; Wenzel, 1995), were analyzed. Levels of these transcripts were not notably different for cultures grown on the six different nitrogen sources (Table 2).

View this table:

Summary of microarray experiment quality parameters for each growth condition

Nitrogen sourceACVACT1PDA1
Ammonium0.182738 ± 392352 ± 30
Asparagine0.142418 ± 122317 ± 19
Phenylalanine0.212917 ± 575441 ± 39
Leucine0.142149 ± 204413 ± 64
Methionine0.112393 ± 143490 ± 33
Proline0.122295 ± 128469 ± 53
  • * Represents the average of the coefficient of variation (SD divided by the mean) for all genes except the 900 genes with the lowest mean expression.

  • Encoding actin; average signal and SD from probe set ‘5392_at’ composed of 16 probe pairs found within 400 nucleotides of the 3′-end of the ORF.

  • Encoding pyruvate dehydrogenase; average signal and SD from probe set ‘5526_at’ composed of 16 probe pairs found within 400 nucleotides of the 3′-end of the ORF.

To analyze the transcriptome data, pairwise comparisons were performed between the transcriptome data for the six nitrogen sources (Fig. 2). In total, 30 pairwise comparisons were carried out (reciprocal pairwise comparisons, e.g. ‘ammonium-grown cultures vs. asparagine-grown cultures’ and ‘asparagine-grown cultures vs. ammonium-grown cultures’, were counted as two comparisons). One thousand four hundred and forty-five genes (24% of the genome) showed a different transcript level in at least one comparison (threshold of twofold change with a 1% FDR), whereas 434 genes (representing 7% of the genome) did not yield a detectable transcript level on any of the six nitrogen sources. The remaining 4205 transcripts differed by less than twofold in all pairwise comparisons (Fig. 2).


Pairwise comparisons of transcriptome data used in this study (red denotes upregulation and green denotes downregulation on the nitrogen source per column). aTranscripts that are consistently upregulated or downregulated on each nitrogen source when compared with the other sources.

Specific transcriptional response to growth on one of the six nitrogen sources tested is extremely limited

Of the 1445 genes that showed differential expression between at least two cultivation conditions, only 131 genes were identified that yielded a unique response to one of the six nitrogen sources tested. Each of the six nitrogen sources tested yielded discrete sets of signature transcripts, although the number of signature transcripts differed between nitrogen sources (Figs 2 and 3).


Signature transcripts and the identities of the genes that were specifically upregulated or downregulated on each of the nitrogen sources. The three independent transcriptome datasets were averaged and then compared. The gene coverage of overrepresented sequences from the promoters of coregulated genes is also presented. Elements were counted as being present in a gene promoter if they occurred at least once. aRedundant nucleotides are indicated by: m, A or C; k, G or T. bRelative to 6451 ORF upstream promoters in the yeast genome according to Regulatory Sequence Analysis Tools (RSAT). cNS, no significant enrichment of motifs was identified from RSAT.

During growth on ammonium as the sole nitrogen source, two genes were consistently upregulated. RPL7B encodes a ribosomal protein and YLR211C a protein with unknown function. The levels of three transcripts were lower in ammonium-grown cultures than in cultures grown on the five amino acids (Fig. 3). CAR1 encodes the first enzyme in the arginine catabolic pathway, ABP1 encodes an actin-binding protein, and DIP5 encodes an amino acid permease that transports Glu, Asp, Gln, Asn, Ser, Ala and Gly (Regenberg, 1999). The strongly reduced transcript level (average of 52-fold difference between ammonium-grown and amino acid-grown cultures) suggests that DIP5 is induced by a wide range of amino acids, even when these are not substrates for Dip5p. Repression of DIP5 by ammonium ions is less likely in view of the high transcript levels in asparagine cultures, in which high concentrations of residual ammonium were observed. DIP5 has recently been shown to be regulated by Grr1p, which is also involved in the transcriptional regulation of genes encoding the SPS-dependent permeases AGP1, BAP2, BAP3, GNP1 and TAT1 (Eckert-Boulet, 2005).

The 12 genes that showed an elevated transcript level in asparagine-grown cultures were scattered over several functional categories, and most of them lacked a clearly established biochemical function. Of the 10 genes that yielded a reduced transcript level on asparagine, six (DAL3, GAP1, GDH3, PUT1, MEP1 and MEP2) are involved in nitrogen metabolism and have been demonstrated to be NCR-sensitive. Indeed, a motif containing the GATAA sequence was overrepresented in this set of genes (Fig. 3).

During growth on phenylalanine, two genes (ARO80 and MIP6) showed elevated transcript levels relative to cultures grown on the other five nitrogen sources: ARO80 encodes a positive regulator of ARO9 and ARO10, both involved in phenylalanine degradation (Iraqui, 1999). The three genes with reduced transcript levels could not be directly linked to phenylalanine metabolism.

The five genes with specifically elevated transcript levels in leucine-grown cultures were scattered over several functional categories. Most of the 17 genes that were specifically downregulated on leucine were related to either metabolism or cell rescue (Fig. 3). Of these, LEU1 and LEU2 are directly linked to leucine metabolism, and ISU2 is linked to the iron–sulfur assembly cluster on LEU1. In silico promoter analysis of these 17 genes showed a fivefold overrepresentation of a promoter element that harbored the CCCCT sequence known to be bound by the ‘general’ stress-related transcription factors Msn2p and Msn4p.

The 17 genes that yielded increased transcript levels with methionine as the nitrogen source belonged to a variety of functional categories and did not yield overrepresented promoter elements. Of these genes, SPE1, encoding ornithine decarboxylase and YFR055W, could be linked to methionine metabolism. YFR055W exhibits a strong sequence similarity to genes encoding β-cystathionases. Of the 34 downregulated transcripts, 21 were related to metabolism, five to transport, and four to cell rescue and cell cycle; four transcripts were from genes with unknown function. Of the 21 genes involved in metabolism, 14 were directly related to methionine metabolism. In fact, all genes encoding enzymes involved in the synthesis of methionine from extracellular sulfate were downregulated (MET3, MET14, MET16, MET5/ECM17, MET10, MET25/MET17, and MET6), as well as the transcriptional regulators of methionine and cysteine metabolism, MET28 and MET32. In addition, permeases involved in transport of sulfur-containing compounds (SUL1, SUL2 and MMP1) were downregulated. In silico promoter analysis of the downregulated genes showed enrichment of two promoter elements corresponding to the binding sites for the Cbf1p–Met4p–Met28p complex and the Met31p–Met32p transcriptional regulators. Together, these transcriptional regulation complexes control sulfur fixation and biosynthesis of the sulfur-containing amino acids (Thomas & Surdin-Kerjan, 1997). In addition to functioning as a nitrogen source, methionine is a preferred sulfur source for S. cerevisiae, as suggested by this study. The observed downregulation of the entire biosynthetic pathway from extracellular sulfate (including sulfate transport) to intracellular methionine prevents the unnecessary expenditure of two ATP and four NADPH per methionine synthesized (Thomas & Surdin-Kerjan, 1997).

Growth on proline yielded 26 genes with specifically elevated transcript levels (Fig. 3). Remarkably, 17 of these were members of the 23-member seripauperin protein family, characterized by the lack of a serine-rich C-terminus. It has been postulated that seripauperins contribute to the composition of cell wall mannoproteins and stress resistance (Marguet, 1988; Viswanathan, 1994; Robyr, 2002). However, these results have to be interpreted with care, as a high degree of sequence similarity of the members of the seripauperin family may have led to some cross-hybridization on the arrays. The nine remaining genes that responded uniquely to proline are mainly involved in transport and detoxification. The case of MCH5 is interesting. Recently, MCH5 has been identified as a riboflavin/vitamin B2 transporter (Reihl & Stolz, 2005). The assimilation of proline as nitrogen source depends on the availability of riboflavin-containing cofactor FAD, which is essential for oxidation of proline by Put1, the proline oxidase. The very tight expression regulation upon proline addition could reflect the fact that the transcription regulator Put3 governs its transcription. Among the 26 genes specifically elevated during growth on proline, four overrepresented promoter elements were found, all of which reside in half or more of the genes. The identity of possible binding proteins remains to be elucidated.

Biosynthetic genes for phenylalanine, asparagine and proline were not significantly repressed in the corresponding cultures, which may be attributed to cross-pathway control and shared biosynthetic pathways for amino acid families. Of the leucine biosynthesis pathway genes, only LEU1 and LEU2 were repressed in leucine-grown cultures, probably via the regulator of leucine biosynthesis, Leu3p (Kohlhaw, 2003). The other genes of the leucine biosynthetic pathway are predominantly controlled by other regulators under this condition, and hence not reduced (Friden & Schimmel, 1987; Hinnebusch, 1992).

One of the proposed applications of DNA microarrays in biotechnology is to analyze large-scale industrial fermentations. It has been proposed that the identification of signature transcripts, i.e. transcripts that show a unique response to change in a single environmental or nutritional parameter, may contribute to the development of small, cost-effective diagnostic arrays (Boer, 2003; Wu, 2004; Tai, 2005). The signature transcripts described in this study complement other sets of chemostat-based signature transcripts (ter Linde, 1999; Piper, 2002; Boer, 2003; Daran-Lapujade, 2004; Tai, 2005), which may facilitate the interpretation of studies with complex industrial media and dynamic cultivation conditions, such as transcriptome analysis of wine fermentations (Marks, 2003; Rossignol, 2003). However, only five transcripts were specifically assigned to growth on phenylalanine, and it is doubtful whether any of these five ‘phenylalanine signature transcripts’ would retain this status if tyrosine or tryptophan was included in the analysis.

The present study illustrates that, in some respects, a ‘signature-transcript-based approach’ may be too simplistic. Depending on the cell's nutritional status and the identity of the amino acid, it may be used as a nitrogen, carbon or, as illustrated by the methionine case, sulfur source. Consequently, meaningful analysis of industrial fermentation with microarrays is likely to require interpretation of complex sets of genes, rather than simple ‘indicator’ genes for all relevant process parameters and nutrients.

Genes with transcriptional coresponses to different nitrogen sources

Because of the involvement of common genes in transport, assimilation and dissimilation of different nitrogen sources, some of the genes involved can be expected to exhibit transcriptional coresponses to different nitrogen sources. To identify coresponsive genes, transcriptomes of cultures grown on the five amino acids were subjected to pairwise comparisons with those of ammonium-grown cultures. Ammonium-grown cultures were used as a reference because ammonium is the only non-amino acid nitrogen source used in the present study. The 1001 genes (16% of the genome) that were found to exhibit transcriptional coresponses were clustered on the basis of nitrogen-source-dependent expression patterns (Fig. 4; supplementary Table S1). The complete overview of the coresponsive clusters is provided in supplementary Table S1. Some of the most remarkable groups are discussed in detail below.


Overall Eisen diagram and identities of the genes that were cooperatively upregulated or downregulated during growth on different nitrogen sources as compared to ammonium. The three independent transcriptome datasets were averaged and then compared. Red denotes a positive fold-change as compared to ammonium, and green squares denote a negative fold-change as compared to ammonium-grown cultures. A list of all genes is given in supplementary Table S1.

Genes showing similar upregulation on nonpreferred nitrogen sources (proline, phenylalanine, leucine and methionine)

A group of 23 genes was upregulated during growth on phenylalanine, leucine, methionine and proline (Fig. 4c; supplementary Table S1). This matches the classification of phenylalanine, leucine, methionine and proline as nonpreferred nitrogen sources, and of asparagine and ammonium as preferred nitrogen sources. The formation of ammonium by asparagine-grown cultures (Table 1) may also contribute to the similar transcriptional response of these genes in asparagine-grown and ammonium-grown cultures. Of these 23 genes, six are involved in metabolism of the nonpreferred nitrogen sources allantoin and urea (DAL1, DAL2, DAL5, DUR1, DUR2, DUR3), and five encode transporters for nitrogen-containing compounds (GAP1, PTR2, MEP2, MEP3 and OPT2). This group also includes the GATA factor-encoding genes DAL80 and GAT1, and the NAD-dependent and NADP-dependent glutamate dehydrogenases encoded by GDH2 and GDH3, respectively. Of these 23 genes, 14 genes are established NCR targets. An over-representation of the GATAAG motif was found within these 23 transcripts. This motif was present in 39% of this gene cluster (occurring at least twice in the promoter region) compared to 3% in the whole genome (not shown).

To analyze in more depth NCR-mediated regulation in cultures grown at a fixed specific growth rate, two additional and complementary approaches were used.

The first was a supervised approach, where the transcript levels of established GATA factor-regulated genes [as reported in the Yeast Proteome Database (Costanzo, 2001)] were used to indicate NCR strength (Table 3). To minimize secondary effects, genes from this list were excluded when: (1) regulation by only one GATA factor was reported; (2) interfering regulatory mechanisms (such as Gcn4p or Put3p) could have an effect; and (3) genes had a low expression level (average below 80 U). This left us with nine NCR-sensitive ‘marker’ genes (Table 3). The average normalized signals of these nine transcripts were used to illustrate the trend over the six conditions (Fig. 5, black bars).

View this table:

Established and newly identified NCR-sensitive genes

GeneGATA-factor regulatedAdditional documented transcriptional regulatorsN-lim sig. trans.Similar NCR profile
UGA1Dal82p, Uga3p
PUT1Put3p, Pho4p
PUT2Put3p, Pho4p
DAL1Gcn4p, Dal82p
GDH1Gcn4p, Hap2p, Leu3p, Spt3p
GDH3Msn2p, Msn4p
GLN1Gcn4p, Pho2p
ENA1Cin5p, Crz1p, Hal9p, Rim101p, Skn7p, Yap6p, Nrg1p
ALD2Msn2p, Msn4p
ARG1Gcn2p, Arg80p, Gal11p, Hfi1p, Spt3p
  • References can be obtained from supplementary Table S2

  • * Regulated by these GATA-factors as documented by YPD (http://www.incyte.com) on 4 March 2005.

  • Directly regulated by these transcription factors as given in YPD (http://www.incyte.com) on 4 March 2005.

  • N-lim signature transcript (32 genes) as given in Tai (2005).

  • § Genes were identified that displayed the same transcript profile as at least five of the marker genes under all conditions. Genes that contained at least two GATAAG sequences in their promoter were included in this table (27 genes). For statistical criteria, see ‘Materials and methods’.

  • These nine genes were selected as ‘marker genes’ to analyze NCR-sensitive regulation. For criteria, see text.

  • Nonunique probe set.

  • ** Not on array.

  • †† Low signal.

  • ‡‡ Below limit of detection.


Representation of the NCR-sensitive transcript levels. Black bars, average normalized levels of established GATA factor-regulated genes; gray bars, average normalized levels of nitrogen limitation signature transcripts (see text for details). Numbers represent P-values obtained with a two-tailed, unequal variance Students' t-test. On the right of the diagonal are given the P-values of the Students' t-test of established GATA factor-sensitive transcript levels between conditions; on the left of the diagonal are given P-values of the Students' t-test of nitrogen limitation signature transcripts between conditions (see text for details).

In a second approach, we used the transcript levels of 32 ORFs previously identified as being specifically upregulated by ammonium limitation (Tai, 2005) to indicate NCR strength. Ammonium is regarded as a preferred nitrogen source; however, when ammonium is limiting, this induces relief of NCR. The average normalized signals of these 32 transcripts were used to illustrate the trend over the six conditions (Fig. 5, gray bars). The normalized transcript levels in the two comparisons described above were also subjected to a Students' t-test to assess significant differences between the conditions. P-values below 0.01 were considered to be significant (Fig. 5). The two approaches reveal the same trend of NCR-sensitive transcript abundance over the six conditions. The average transcript levels in both approaches indicate that during growth on ammonium and asparagine, NCR-sensitive transcript abundance is lower than with the other conditions (P<0.01), is intermediate during growth on phenylalanine, and is highest during growth on leucine, methionine and proline on both criteria (Fig. 5).

In order to find as yet unidentified GATA factor-regulated genes, we used the transcript profile of the nine ‘marker’ genes to identify genes with a similar transcript profile (see ‘Materials and methods’). Of the 76 genes thus identified, 27 (36%) contained a GATAAG sequence (Cunningham & Cooper, 1991; Magasanik & Kaiser, 2002) in their promoter (at least twice), in comparison to only 3% genome-wide. The 27 genes are included in Table 3. Fifteen of these genes have not yet been reported to be regulated by GATA factors. However, the promoters of five of these, ARG1, CPS1, DCG1, IDP1, and OPT2, have recently been identified in chromatin imunoprecipitation studies as beong bound (P<0.005) by Gln3p, Gat1p, or both (Harbison, 2004).

The research described here provides the unique opportunity to quantify NCR strength at a fixed growth rate, thus distinguishing GATA factor-regulated gene expression from growth rate-dependent regulation. Corroborating previous results obtained in shake flasks, asparagine and ammonium were found to elicit the strongest NCR. This can be partly explained by the presence of ammonium in the medium, as ammonium has been reported to be a strong effector of NCR regulation (ter Schure, 1995). However, other signals, such as glutamate and glutamine concentrations, are likely to play important roles as well (Table 4) (Stanbrough, 1995; Beck & Hall, 1999; Bertram, 2000).

View this table:

Average mRNA levels with SDs (Affymetrix signal units) of genes involved in central nitrogen metabolism

Nitrogen source
GDH11374 ± 154710 ± 941502 ± 2201514 ± 2151428 ± 231366 ± 28
GDH3222 ± 58108 ± 111429 ± 2621169 ± 1161229 ± 102531 ± 56
GDH2306 ± 9494 ± 38635 ± 162615 ± 261101 ± 77901 ± 54
GLN11254 ± 185698 ± 973578 ± 5952490 ± 4533053 ± 1102430 ± 8
GLT1606 ± 104336 ± 51432 ± 52592 ± 50696 ± 12765 ± 104

Glutamate and glutamine metabolism

α-Ketoglutarate represents a major split between carbon (and energy) and nitrogen metabolism, as it can be converted to succinyl-CoA and processed by the tricarboxylic acid (TCA) cycle, or alternatively, aminated to form glutamate, the most important nitrogen donor in the cell. Hence, the synthesis of α-ketoglutarate is regulated by both nitrogen-specific and carbon-specific factors. The latter is the heme activator protein complex Hap2,3,4,5p, an activator of respiratory gene expression. Most genes encoding enzymes in the TCA cycle, both upstream and downstream of α-ketoglutarate, are under its control. The first reactions in the TCA cycle upstream of α-ketoglutarate are also regulated by the retrograde regulator proteins Rtg1p and Rtg3p. Retrograde regulation is activated in response to conditions of low glutamate (e.g. due to mitochondrial dysfunction) (Liu & Butow, 1999). CIT2 mRNA has been the main measure of retrograde regulation, as up to now, only Rtg1p and Rtg3p have shown to regulate CIT2 transcription.

The retrograde proteins Rtg1p and Rtg3p are influenced by several signaling molecules and regulatory kinases, including glutamate, Tor, Mks1p and Lst8p [for review see Liu & Butow (2006)]. Dilova (2004) have shown that CIT2 expression is not affected upon repression of Tor (by addition of rapamycin) when glutamate or proline are used as nitrogen source, but CIT2 expression becomes Tor-dependent in the presence of ammonium. In addition, these experiments show that CIT2 mRNA is elevated during growth on ammonium in comparison to growth on glutamate or proline, even in the absence of rapamycin.

In our experiments, essentially no ammonium was present in the four conditions, growth on phenylalanine, leucine, methionine, and proline. Hence, CIT2 expression in these conditions can be used as a measure of Tor-independent retrograde regulation. We found high levels of CIT2 mRNA in three of the conditions, growth on phenylalanine, leucine, and methionine (Fig. 6). This implies that these three amino acids are poor sources of glutamate. In all three cases, glutamate is synthesized by transamination reactions between the respective amino acid and α-ketoglutarate. In the case of phenylalanine, competition for α-ketoglutarate, which is synthesized in the mitochondria, might be even less favorable for glutamate synthesis, as the two known transaminases, Aro8p and Aro9p, both reside outside the mitochondria (Huh, 2003).


Expression intensity of the retrograde-sensitive genes. Data represent the average and SD of data from three independent steady-state chemostat cultivations.

In comparison, proline is a direct source of glutamate, leading to a low level of CIT2 expression. During growth on ammonium, where the level of CIT2 expression is also low, glutamate is directly synthesized by an NADPH-dependent reaction with α-ketoglutarate, which is energetically a much more favorable reaction with physiologic concentrations of the substrates as compared to the transamination reactions (Tewari, 1998). Asparagine is a source of both ammonium and aspartate; the latter can be used to transaminate α-ketoglutarate, both inside and outside the mitochondria. CIT2 expression on growth on asparagine is low. Figure 6 shows, in addition to the expression level of CIT2, that of DLD3, another retrograde-regulated gene, and those of CIT1, ACO1, IDH1 and IDH2, other genes regulated by Rtg1p and Rtg3p. For this latter group, the same trend can be observed as for CIT2, although it is less pronounced, probably due to Hap2,3,4,5p regulation under these conditions (Fig. 6).

Coregulation of amino acid transport

Thirteen genes were upregulated relative to ammonium-grown cultures in all amino acid-grown cultures except for those on proline (Fig. 4b; supplementary Table S1). Six of these are involved in transport: AGP1, a broad-substrate-specificity permease; GNP1, a high-affinity glutamine permease; BAP3, a branched-chain amino acid permease; TAT1, a tyrosine and tryptophan permease; TAT2, encoding a high-affinity tryptophan permease; and SIT1, a heavy metal ion transporter. These genes, except for SIT1, form a subset of related amino acid permeases that is induced by the availability of amino acids, except for proline and arginine (Nelissen, 1997; Forsberg, 2001), and is dependent on the SPS (Ssy1p, Ptr3p, Ssy5p) sensor complex. Instead, a specific transporter, Put4p, takes up proline. Proline's high specificity in transport may result from the fact that proline is not required in large quantities in yeast proteins, and is not utilized as a nitrogen source under anaerobic conditions, and that proline degradation increases the levels of reactive oxygen species, which have toxic effects on the cell (Nomura & Takagi, 2004). In addition, proline (like arginine) is a very abundant amino acid in must and wort (Henschke & Jiranek, 1993; Hernandez-Orte, 2002), and shared signaling pathways could interfere with selective use of other amino acids.

Regulation of the formation of fusel alcohols/acids, the Erhlich pathway

A substantial number of genes (33 genes) exhibited transcriptional coresponses on phenylalanine, leucine, and methionine (Fig. 4a; supplementary Table S1). Of these genes, ARO9, encoding an aromatic amino acid aminotransferase, and ARO10, encoding a broad-substrate-specificity 2-oxoacid decarboxylase, showed the strongest upregulation, which is consistent with the common catabolism of these amino acids. Indeed, after these three amino acids are transaminated, the resulting 2-oxoacids cannot be used for anabolic purposes (except to reform the corresponding amino acids), but are instead decarboxylated via the Ehrlich pathway, yielding fusel alcohols and acids (Ehrlich, 1907; Vuralhan, 2005). On the basis of the common upregulation of the ARO10 gene on phenylalanine, leucine, and methionine, we recently performed a detailed analysis of the Aro10p-dependent decarboxylase activity (Vuralhan, 2005). Indeed, Aro10p was shown to be responsible for a broad-substrate-specificity decarboxylase activity involved in the catabolism of all three amino acids (Vuralhan, 2005). Similarly, the upregulation of PDR12 on all three amino acids (Fig. 4) has recently been correlated with ATP-driven transport of fusel acids derived from the catabolism of aromatic, branched chain amino acids and methionine (Hazelwood, 2006).

Together with passive diffusion of these compounds into the cells, this transport process is likely to contribute to uncoupling of the plasma membrane proton motive force, thus explaining the reduced biomass yields of cultures grown on methionine, leucine, and phenylalanine (Table 1). The coresponse of ARO9, encoding the aromatic amino acid transaminase, suggests that it may act as a broad-substrate transaminase. Further research is required to investigate Aro9p substrate specificity.

This study provides a clear example of how transcriptome analysis can be used to guide functional analysis studies. As demonstrated elsewhere, ARO10 (Vuralhan, 2003, 2005) and PDR12 (Hazelwood, 2006) encode the biocatalysts that are involved in the Ehrlich pathway. The transcriptional information provided here suggests that this pathway is common to all fusel alcohol/acids forming nitrogen sources.

However, this dataset did not allow us to identify the gene(s) encoding the dehydrogenase activity that, from the functional characterization of alcohol dehydrogenase, would be encoded by ADH6 or ADH7. No ADH genes were differentially expressed.

The regulatory mechanism by which the Ehrlich pathway-encoding genes are coregulated remains unclear. The data presented here show that ARO80, the transcriptional activator of ARO10 and ARO9, exhibits differential expression on phenylalanine only (Fig. 3). This suggests that a more general regulatory mechanism participates in the regulation of the Ehrlich pathway genes on the different nitrogen sources.

The transcriptome analysis of S. cerevisae reported here reveals the complexity of the genome-wide expression response to nitrogen sources. The notion of a transcript that specifically responds to a unique nitrogen source seems erroneous. Similar experiments using additional nitrogen sources should reduce even further this small number of signature genes. Chemostat-based transcriptomics allowed us to study the impact of the nitrogen source irrespective of its impact on the growth rate and related indirect effects. On that basis, this study also provided insights into the nitrogen catabolite repression and central nitrogen metabolism of S. cerevisiae grown on poor or good nitrogen sources.

Furthermore, transcriptome analysis proved to be a valuable tool for gene function discovery and functional characterization, as the reported data have been used to identify two major components of the Ehrlich pathway, the broad-substrate-specificity 2-oxoacid decarboxylase Aro10p and the weak acid transporter Pdr12p involved in the export of fusel acids formed from methionine, and aromatic and branched chain amino acids (Vuralhan, 2005; Hazelwood, 2006).


The full dataset can be downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/) (Barrett, 2006) under the accession series number GSE6405.


The research group of J.T. Pronk is part of the Kluyver Centre for Genomics of Industrial Fermentation, which is supported by the Netherlands Genomics Initiative. This work was financially supported by the board of the Delft University of Technology, DSM and the Dutch Ministry of Economic Affairs (NWO-CW project 99 601).


View Abstract