Scientific Publications - Work Done by Microbiology Reader Bioscreen C
 

A cybernetic modelling  approach for cell biology Case study of the Central Nitrogen Metabolism in Saccharomyces cerevisiae Een cybernetische model aanpak voor cel biologie Het centraal stikstof metabolisme in Saccharomyces cerevisiae als voorbeeld (met een samenvatting in het Nederlands) PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de Rector Magnificus Prof. Dr. H.O. Voorma, ingevolge het besluit van het College voor Promoties in het openbaar te verdedigen op maandag 10 januari 2000 des middags om 16.15 uur door Natal Adriaan Wilhelm van Riel Geboren op 22 februari 1973 te Gilze

Chapter 6 Physiological study of a DGLT1 (GOGAT) mutant of Saccharomyces cerevisiae Natal A.W. van Riel¹, José M. Guillamon³, Marco L.F. Giuseppin² and C. Theo Verrips^1,2 1 Department of Molecular Cell Biology, Institute of Biomembranes Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands 2 Unilever Research Vlaardingen Olivier van Noortlaan 120, 3133 AT Vlaardingen, The Netherlands 3 Departament de Bioquimica i Biotecnollogia Universitat Rovira i Vigili, Av. Ramón y  Cajal 70, 43005 Tarragona, Spain

Chapter 6 120 Abstract Initiated by the results of previously developed mathematical models, the Central Nitrogen Metabolism in S. cerevisiae has been further investigated. Different models have indicated that the glutamate synthase (GOGAT) pathway plays a more important physiological role in yeast than is generally assumed. This pathway could give the cell the possibility to deal with rapid fluctuations in glutamine availability. The responses of both a GOGAT negative (Dglt1) mutant and a wild-type strain after pulsing different nitrogen sources and different concentrations to glutamine limited continuous cultures have been determined. In this work the data of the pulse experiments have been embedded in a broader physiological interpretation and have been discussed in the context of the Dynamic Optimal Metabolic Control modelling framework. This framework is focused on the holistic function of the various aspects of Central Nitrogen Metabolism. Improved understanding of the physiological function of the different parallel pathways is necessary if the flux in Central Nitrogen Metabolism of S. cerevisiae needs to be redirected for Metabolic Engineering applications. New data have been presented on growth and especially the effect of the nitrogen / carbon ratio in the medium. As previously suggested by the model, GOGAT appeared to be associated to the mitochondria, which is a crucial aspect for its in vivo function. The change in elemental biomass composition after a nitrogen pulse has been determined. Within 2 hours the nitrogen content of the biomass increased by 14%, information which is especially relevant for the models. It has been hypothesised that the low biomass yield and high by-product formation for the mutant are related to a redox imbalance. 6.1 Introduction Glutamine is a preferred nitrogen source for the yeast Saccharomyces cerevisiae. Glutamine represses the metabolism of other nitrogen sources which can be present in rich media used for production processes. This makes the study of glutamine catabolism particularly relevant. Glutamine catabolism is poorly known (Chapter 1), in contrast to that of glucose as the most preferred carbon source. For growth on glutamine, glutamate synthase (GOGAT) and glutaminases are the possible pathways for glutamine degradation and glutamate production. In the latter case, the ammonia produced can be used by NADPH-dependent Glutamate DeHydrogenase, also producing glutamate. From glutamate and glutamine the cell can synthesise all other nitrogen containing compounds, e.g. proteins, nucleotides and lipids. There is controversy about the relevance and function of the parallel pathways for glutamate biosynthesis. A good characterisation of the dynamic properties of the Central Nitrogen Metabolism is essential when this metabolism is to be engineered. Metabolic redundancy is common in central metabolism and if not of all parallel pathways the function is known, this will cause unexpected problems when a related metabolic flux needs to be redirected. Information on the function of such pathways usually can only be obtained in well defined physiological studies, often employing mutant strains (e.g. Luttik et al. 1998; Ter Schure et al., 1998). Both for GOGAT and the glutaminase pathway no such data have been reported.

Physiology of GOGAT negative S. cerevisiae   121 The physiological study as presented, was driven by previous mathematical models of the CNM in yeast (Van Riel et al., 1998, 2000). Mathematical models provide highly efficient and compact frameworks to structure available information. Besides serving as engineering tools they help to focus experimental effort and fuel new hypotheses. With the (dynamic) computer simulation models several knowledge gaps for the CNM in S. cerevisiae have been identified and some striking predictions and suggestions were made (Van Riel et al., 1998, 2000). The biologically most important prediction was the proposed important role of GOGAT for growth on glutamine, especially when the availability of nitrogen changes quickly. For a further physiological characterisation of the CNM in yeast, with special attention on the function of GOGAT, a GOGAT negative (Dglt1) mutant strain (VWk274 LEU+) has been studied in glutamine limited continuous cultures, with pulses of different nitrogen sources and of different size (Guillamon et al., 1999). As reference, the same experiments were done with a wild-type strain (CEN.PK-113D, also called VWk43). A large number of intracellular and extracellular metabolites were analysed during the steady-state and after the pulses. The responses of some genes coding for enzymes of the CNM have also been studied. As suggested in the discussion of the kinetic model (Van Riel et al., 1998), the redox state in the cell was taken into account and the NAD(H) and NADP(H) concentrations have been determined as well as the reduced and oxidised form of glutathione (GSH and GSSG respectively). From the mathematical models it was also clear that within two hours after the pulses the synthesis of nitrogen containing compounds, other than glutamine and glutamate, should increase. Van Riel et al. (1998) suggested that a nitrogen containing compound derived from glutamate, such as glutathione, could be used for storage of excess nitrogen after nitrogen pulses. The results of the pulse experiments showed that GOGAT indeed plays an important role in CNM, especially for glutamate synthesis when glutamine is the sole nitrogen source (Guillamon et al., 1999). As mentioned in Chapter 5, the Dglt1 mutant had several unexpected phenotypes. The most striking was the completely changed redox state of the cell. The almost complete lack of reduced equivalents (NAD(P)H) also reduced ethanol formation, a drastic effect for S. cerevisiae. Instead, large amounts of acetaldehyde were present in the supernatant. Guillamon et al. (1999) suggested that maintaining a correct redox equilibrium in the cell could be one of the main roles of GOGAT in yeast. In this chapter the data of the pulse experiments have been embedded in a broader physiological study. New data are presented on growth of the wild-type and the GOGAT negative mutant strain on different nitrogen sources and different initial concentrations. In continuous cultures the effect of the ratio of nitrogen versus carbon in the feed has been studied. The consistency of the steady-state data is checked with elemental mass balances. The pulse experiment data are discussed on the basis of a classification of the different dynamic responses. This allows the interpretation of the responses in the context of the Dynamic Optimal Metabolic Control concept (Giuseppin and Van Riel, 2000; Van Riel et al., 2000). Since the substrate limited continuous cultures resulted in well defined physiological steady-states, disturbances of the homeostasis could be studied. With substrate pulses of various concentrations it was possible to observe some of the qualitative different responses as hypothesised within the DOMC framework. The developed kinetic model contained an unexpected structure of the CNM (Van Riel et al., 1998). This model structure indicated that GOGAT was in, or associated to the

Chapter 6 122 mitochondrial membrane and that it could use mitochondrial a-ketoglutarate to produce glutamate. This was a first suggestion of a protein connecting the TCA-cycle to CNM. More recently this idea was confirmed by the appearance of a GOGAT structure prediction which shows several transmembrane domains (Chapter 5). In the current work the localisation of GOGAT has been studied by fractionation of the mitochondria. For a more consistent dataset also the change in the elemental composition of the biomass has been determined after one of the pulses. These data complemented the results of the changes in nitrogen contained in free amino acid pools and glutathione (Guillamon et al., 1999). 6.2 Materials and methods 6.2.1 Strains A GOGAT negative mutant was constructed in the diploid strain CEN.PK219 carrying a leu2 marker. The gene was deleted using the method of PCR-targeting with short-flanking homology (Wach et al., 1994). The DNA fragments for homologous integration were generated by PCR using a loxP-KanMX-loxP cassette with kanamycin resistance as dominant marker (Guldener et al., 1996). After tetrad analyses of the resulting heterozygous deletion strain, the correct integration was verified by diagnostic PCR and subsequently the Kan^R gene was removed by expressing the cre recombinase (verified by PCR). In the resulting strain VWk274 (MAT a, leu2-3,112, glt1(41,6000)::loxP), the leu2 mutations were removed by transformation with a wild-type LEU2 gene derived from YDpLEU (Berben et al., 1991) yielding strain VWk274 LEU+. 6.2.2 Growth conditions Microtitre batch fermentations The growth rates of VWk43 (wild-type) and VWpk274 LEU+ (Dglt1 mutant) for different nitrogen sources and with different concentrations have been determined in microtitre plates in combination with an automatic Optical Density reader at 600 nm (Bioscreen C, Labsystems). The pre-cultures were grown overnight in a defined, minimal medium with glutamine as sole nitrogen source (same medium as for continuous culture experiments, see below). In the 2´100 well microtitre plates, 395 ml medium was inoculated with 5 ml pre-culture. While shaking (temperature 30°C), every 10 minutes the OD at 600 nm was determined during 47.5 hours. (Every strain was inoculated four times on the same medium.) All media were based on the minimal medium for the continuous culture experiments as described below. The nitrogen sources were glutamine, glutamate, proline and ammonium. For all substrates the same amount of nitrogen was included in the media (glutamine contains two nitrogen atoms). The concentrations used were 41.1, 10, 5, 1 or 0.2 N-mol×l^-1. The specific growth rates m [h^-1] have been determined from the OD readings in the time period from 2 to 12 hours after inoculation when all cultures were exponentially growing (after an initial, short lag phase). Linear regression was applied on the ln(OD₆₀₀) data. Continuous culture fermentations Continuous culture experiments with pulses of glutamine and glutamate were repeatedly carried out in different fermenters at different sites. Initial experiments (of which some data will be reported) were done in 0.5 l

Physiology of GOGAT negative S. cerevisiae   123 fermentors (Sixfors, INFORS AG) with 450 ml working volume. For the pulse experiments 2.0 l fermentors with a working volume of 1.0 or 1.5 litre have been used (Fourfors fermenter, INFORS AG and BiofloIII fermentor, New Brunswick Scientific) connected to a computer controller unit running with Wizcon (PC Soft International) or Advanced Fermentation Software (New Brunswick Scientific). The inocula were grown in shake-flasks at 30°C and 135 - 180 rpm in YPD medium (yeast extract, peptone and 2% glucose) or the defined minimal medium used for the continuous cultures (see below). After overnight growth, 50 - 80 ml preculture was used to inoculate the fermenters. S. cerevisiae strains VWk43 (wild-type) and VWk274 LEU+ (Dglt1) were grown aerobically at 30°C during batch fermentation. The pH in the fermentor was automatically controlled at 5.0 by addition of 2M or 3M KOH. The oxygen tension was kept above 20% air saturation to assure aerobic growth. The stirrer speed was set between 300 and 600 rpm. Carbon dioxide and oxygen concentrations in the exhaust gas of the fermentor were measured on line by either an Uras3G CO₂ analyser and a Magnos4G O₂ analyser (Hartmann and Braun) or by a mass spectrometer (Prima 600, ThIS Gas Analyses Systems BV). With the mass spectrometer the ethanol concentration in the exhaust gas of the fermenter was also measured. After batch growth, the continuous culture was started when less than 100 ppm of ethanol was present in the off-gas and the Respiration Coefficient (RQ, the ratio of CO₂ Evolution Rate (CER [mmol×l^-1×h^-1]) and Oxygen Uptake Rate (OUR [mmol×l^-1×h^-1])) was below 0.9, usually after overnight growth. The continuous feed was started at a dilution rate of 0.1 h^{- 1}. The volume of the fermentor adjusted itself to the working volume by withdrawing the excess volume. The substrate flow was controlled using a feedback controller, based on weight. The first steady-state was assumed to be reached after 5 times the dilution time (50 h). The growth conditions were essentially as previously described (e.g. Ter Schure et al, 1995). The in the continuous cultures used minimal medium was based on the medium according to Egli (e.g. Sierkstra et al., 1992) with 20 g×l^-1 glucose, but ammonium chloride as nitrogen source was replaced by different concentrations of glutamine, ranging from 2.0 to 5.5 g×l^-1. The actual glucose and glutamine concentrations were determined enzymatically (see below). During dissolving and solution storage glutamine degrades into equimolar amounts of ammonia, especially at higher temperatures (such as room temperature) and in phosphate buffers (Khan and Elia, 1991). No associated glutamate is formed, most likely pyroglutamate (a cyclised amino acid) is the other degradation product. Most of the pulse experiments were done with (approximately) 2.5 g×l^-1 glutamine in the feed. Struktol antifoam agent was added to the medium at a concentration of 100 ml×l^-1 (0.01%). Above a certain critical growth rate m_crit, S. cerevisiae grows respiro-fermentatively and produces ethanol. The m_crit in carbon limited cultures was determined in so-called accelerostat experiments in which the dilution rate is continuously increased. An acceleration of 0.01 h^-2 has been used. Four pulses (20 mM or 10 mM glutamine or 40 mM or 20 mM glutamate) were added to the glutamine limited steady-state cultures. Approximately 40 ml of a sterilised, concentrated stock solution was injected aseptically (within 10 seconds). After each pulse samples were taken and the culture was allowed to reach steady-state before a new experiment was done

Chapter 6 124 (at least 24 hours). The glutamine pulses were half the concentration of the glutamate ones so as to add the same amount of nitrogen. 6.2.3 Sampling and sample preparation Samples for the determination of intracellular metabolites and cofactors were taken aseptically from the fermentor, quenched immediately and extracted to assay metabolites with high turnover rates. Essentially the method developed by Gonzalez et al. (1997) was used. An aliquot of the culture was dropped into 5 volumes of 60% methanol diluted with Hepes buffer (10 mM, pH 7.5) kept at -40ºC. The mixture was centrifuged at low temperatures (the temperature of the suspension should be below -20ºC after centrifugation). The supernatant was poured and 3 ml of boiling 75% ethanol diluted with Hepes buffer (70 mM, pH 7.5) was added to the pellet and put in an 80ºC waterbath for 3 minutes. The ethanol was evaporated under a nitrogen flow and the pellet was resuspended in 1.5 ml of milliQ water. This solution was centrifuged at 5000 rpm for 10 min (4ºC) to clean the supernatant from fines. Supernatant was taken and used for the metabolite assays. To determine extracellular metabolites and amino acids, culture samples were filtered through a 0.45 mm pore size filter and frozen. Preparation of cell-free extracts for mRNA isolation was performed as described by Sierkstra et al. (1992). In steady-state experiments approximately 10 ml of culture was sampled for the determination of intracellular metabolites and cofactors. For extracellular compounds about 15 ml was sampled. 15 ml culture samples were used for biomass determination. For the pulse experiments samples were taken at -20, -10, 0, 2, 5, 10, 20, 30, 60, 120 minutes after the pulse and often also overnight. 2 or 5 ml of these samples were quenched, 1 ml was immediately put at -80°C or frozen in liquid nitrogen for RNA extraction and 1 ml was filtered to get the supernatant. Before the pulse and 2 hours after the pulse, samples of at least 5 ml were taken for biomass determination. For enzyme activity analyses cell lysate was obtained by mixing in a Vibrax with glass beads (425-600 µm diameter) for 2 times 5 minutes (in 0.1 M potassium phosphate buffer pH 7.0). The lysate was spun down. For the fractionating of intracellular organelles, sheroplasts, prepared using Zymolyase, were broken in a glass Dounce homogeniser. 6.2.4 Analyses Biomass determination For biomass concentration the culture Dry Cell Weight (DCW) was determined (e.g. Sierkstra et al., 1992). Approximately 10 ml of culture was pipetted into preweighted and predried glass tubes, the samples were centrifuged, washed with water, centrifuged again and dried in the oven overnight at 100°C. The tubes were cooled down in an exsiccator before being weighed again. Elemental biomass composition The elemental biomass composition was determined with a CHNO analyser from washed and dried biomass (dried overnight at 100°C and cooled down in an exsiccator). Extracellular metabolites Carbon compounds (acetaldehyde, acetic acid, ethanol, glycerol and pyruvic acid) were measured by means of HPLC (Shimazu, Aminex HPX-87H column,

Physiology of GOGAT negative S. cerevisiae   125 Biorad, temperature 60°C, H₂SO₄ solution of pH 2.0 as eluent). To determine the glucose concentration in the feed and the supernatant a Cobas Mira S autoanalyser (Hoffmann-La Roche) was used with the glucose kit from IntruChemie, comprising hexokinase and glucose-6-phosphate-dehydrogenase. Free amino acids were measured by means of HPLC using the AccQ-tag system (Waters / Millipore, USA) equipped with a reversed-phase C¹⁸ column (temperature 37°C). Amino acids were derivatized with 6-aminoquinolyl-N- hydroxysuccinimidyl carbamate (AQC). The separation was performed using a nonlinear gradient of 1% to 17% acetonitryl in a 130 mmol×l^-1 sodium acetate buffer (AccQ.TAGTM eluent). Amino acid derivates were detected by a fluorescence detector; excitation at 245 nm and emission at 395 nm. Glutamate and glutamine were also measured using the L- glutamic acid determination kit comprising NAD-dependent glutamate dehydrogenase (Boehringer Manheim cat. no. 139 092). In a two stage reaction, NAD⁺ reduction during deamination of glutamate is stoichiometrically linked to a reoxidation reaction producing formazan, measured at 492 nm. Asparginase (Boehringer Manheim), having a glutaminase side-activity, was used to transform glutamine into glutamate. Intracellular metabolites Intracellular carbon metabolites and free amino acids were measured as described above. a-Ketoglutarate was measured enzymatically using glutamate dehydrogenase as described by Bergmeyer et al. (1974) (implemented on Cobas Mira S). NADH oxidation was measured as a decrease in absorbance at 340 nm. Total glutathione and its oxidised form (GSSG) were determined in the intracellular samples as described by Griffith (1980), using a Cobas Fara autoanalyser (Hoffmann-La Roche). The method for both determinations is almost the same: during the GSSG determination GSH is masked by vinylpyridine. The concentration of reduced glutathione (GSH) was calculated from these results. The GSH concentration will be reported to be approximately 20 times larger than GSSG in the wild-type (41.0 vs. 2.0 mmol×gX^-1, Table 6.2). However, before reconciliation of the datasets some experiments resulted in much larger concentrations of GSSG (up to 23 mmol×gX^-1) (Guillamon et al., 1999). Redox cofactors were measured fluorometrically as described by Bergmeyer (1974) using the reaction catalysed by alcohol dehydrogenase for determination of NAD, glucose-6- phosphate-dehydrogenase for NADP, glycerol-3-phosphate dehydrogenase for NADH and glutathione reductase for NADPH. Labelling of oligonucleotides and Northern blot analysis For the detection of the ACT1, GAP1, GDH1, GLN1 and H2A/H2B mRNA labelled oligonucleotides were used and Northern analyses were performed as described previously (Sierkstra et al., 1992; Ter Schure et al., 1995). Northern blots for the detection of the GAP1 and GDH1 mRNA levels were probed with ACT1 as internal control for the amount of RNA blotted and for the detection of GLN1 mRNA levels H2A/H2B was used as an internal control. The quantitative results were obtained by calculating the intensity ratio between the gene of interest and the reference gene with the maximum expression level observed defined as 100%. GOGAT activity The GOGAT enzyme activity [mAbs×(mg protein)^-1×min^-1] was determined according to Roon et al. (1974). NADH oxidation was measured as the decrease in absorbance at 340 nm during 30 minutes. The enzyme activity was corrected for a blank

Chapter 6 126 reaction carried out in the absence of glutamine or a-ketoglutarate. Protein concentrations [g×l^-1] were determined according to Bradford (1976) with Bovine Serum Albumin (BSA) as standard. Fractionating of cell extracts Mitochondria were isolated at pH 6.0 and purified by density gradient centrifugation according to De Kroon et al. (1999). Spheroplasts, prepared using Zymolyase, were broken in a glass Dounce homogeniser. Unbroken spheroplasts, nuclei and debris were removed by centrifugation. The crude mitochondria were loaded onto sucrose step gradients or Nycodenz gradients. After ultracentrifugation the purified mitochondria were collected and frozen in liquid nitrogen. GOGAT enzyme activity was determined as described above, with protein concentrations corrected for BSA present in buffer. Calculation of CER and OUR The carbon dioxide and oxygen concentrations in the dried exhaust gas (cooled in the headspace to 4°C) of the fermenter were measured online (Section 6.2.2). The gas flow rate of the exhaust gas was measured. Then the CER [mmol×l^- 1 ×h^-1] and OUR [mmol×l^-1×h^-1] can be calculated according to ( ) 1 2 , 2 , ) ( - - = m in CO in gas out CO out gas V p q p q CER (6.1) ( ) 1 2 , 2 , ) ( - - = m out O out gas in O in gas V p q p q OUR (6.2) in which q_gas represents the measured gas flow rate [h^-1], V_m is the molar volume at atmospheric pressure and room temperature (22.4 L), p_CO2 and p_O2 stand for the measured volume fraction of CO₂ and O₂ [%]. The gas volume changes were corrected by o u t in N N 2 2 % % or o u t o u t in in CO O CO O 2 2 2 2 % % 100 % % 100 - - - - (6.3) for Prima 600 mass spectrometer and the Uras3G CO₂ and Magnos4G O₂ analyser respectively. The oxygen and CO₂ in the outflowing culture were neglected. 6.2.5 Mass balances Mass balances were calculated at a macroscopic scale for the fermenter. The recovery of the nitrogen or carbon, in principle, should be 100% for a reliable dataset. Since nitrogen is the most relevant compound in this work, all concentrations were transformed to N-moles. The change in time of the residual (extracellular) concentration of a metabolite x_ex [N- mmol×l^-1] was written as a differential equation: ) ( , ex N N feed ex ex x X D Dx outflow culture nflow i feed x dt dx + - = - = = & (6.4) x_ex: residual (extracellular) concentration [N-mmol×l^-1]

Physiology of GOGAT negative S. cerevisiae   127 D: dilution rate [h^-1] ⁷⁾ x_feed,N: concentration in the feed in [N-mmol×l^-1] X_N: biomass concentration in [N-mmol×l^-1] To setup the macroscopic mass balances (also called ‘black box’ description of the biomass) the elemental biomass composition needs to be known, especially the nitrogen content. 6.3 Growth rate of wild-type and mutant on different nitrogen sources The specific growth rate was determined for different nitrogen sources present at different concentrations. As could be expected, for growth on glutamine the growth rate of the GOGAT mutant was lower than of the wild-type (Table 6.1). There was no clear influence of the glutamine concentration in the medium on the growth rate although the growth rate varied considerably for the wild-type. We report also a noticeable difference in the growth rate between the wild-type and the GOGAT negative mutant for growth on ammonia. High glutamate concentrations in the medium seemed to hamper growth, both for the wild-type and the mutant. The results suggested that the mutant might grow slightly faster than the                                                                 7) To calculate the effective dilution rate also the added base (for constant pH) needs to be taken into account. Table 6.1 Specific growth rates m of wild-type VWk43 and Dglt1 mutant VWk274 LEU+ in microtitre plates. The initial concentrations of the nitrogen sources are indicated. glutamine [N-mol×l^-1] wild-type [h^-1] Dglt1 [h^-1] difference [%] glutamate [N-mol×l^-1] wild-type [h^-1] Dglt1 [h^-1] difference [%] 0.2 0.094 0.083 -11.3 0.2 0.100 0.085 -15.1 1.0 0.110 0.088 -19.5 1.0 0.100 0.095 -6.4 5.0 0.115 0.087 -24.1 5.0 0.098 0.100 +1.8 10.0 0.109 0.084 -22.9 10.0 0.061 0.067 +8.3 41.1 0.103 0.086 -16.7 41.1 no growth no growth average 0.106 0.086 -19.2 average 0.090 0.087 -3.3 proline [N-mol×l^-1] wild-type [h^-1] Dglt1 [h^-1] difference [%] ammonia [N-mol×l^-1] wild-type [h^-1] Dglt1 [h^-1] difference [%] 0.2 0.094 0.083 -12.1 0.2 0.098 0.083 -14.9 1.0 0.099 0.082 -17.2 1.0 0.091 0.085 -6.8 5.0 0.090 0.084 -6.7 5.0 n.d. n.d. 10.0 0.092 0.086 -6.2 10.0 0.104 0.088 -14.8 41.1 0.092 0.087 -5.4 41.1 0.109 0.088 -18.8 average 0.093 0.084 -9.7 average 0.100 0.086 -14.0 n.d.: not determined

Chapter 6 128 wild-type for higher glutamate concentrations. According to our knowledge, this is the first time actual data of the specific growth rate of a GOGAT negative strain have been reported. The activity of glutamate synthase was determined in cell lysates of wild-type VWk43 and in the Dglt1 mutant VKk274 LEU+ (Table 6.4). The activity in the wild-type was 6.5. [mAbs×(mg protein)^-1×min^-1] and in the Dglt1 mutant a residual activity of 0.9 was determined. The origin of this activity, i.e. NADH oxidation in a reaction with glutamine and/or glutamate, is unknown. Cogoni et al. (1995) reported a zero residual activity in their GOGAT null mutant. 6.4 Steady-state growth of wild-type and Dglt1 mutant 6.4.1 Nitrogen limitation In nitrogen limited chemostat cultures the cells will immediately start to consume the 0 2 4 6 8 10 12 0.00 0.02 0.04 0.06 0.08 0.10 N/C [mol/mol] 0 1 2 3 4 5 6 C-limitation N-limitation ethanol ‘double’ limitation Fig. 6.1 Biomass concentrations (DCW [g×l^-1]) for different glutamine / glucose ratios in the feed of continuous cultures (D = 0.1 h^-1) for wild-type and Dglt1 mutant. u biomass wild-type VWk43 [g×l^-1], n biomass Dglt1 mutant VWk274 LEU+ [g×l^-1], s residual glutamine wild-type [mM], 5 residual glucose wild-type [mM], Q residual glutamine Dglt1 mutant [mM] and l residual glucose Dglt1 mutant [mM]. 2 g×l^-1 glutamine and 20 g×l^-1 glucose results in a N/C of 0.04 (N-mol×C-mol^-1) and 3.8 g×l^-1 glutamine and 20 g×l^-1 glucose yields a ratio of 0.078. The circles indicate the average values from Table 6.2.

Physiology of GOGAT negative S. cerevisiae   129 available nitrogen after a nitrogen pulse. As discussed in the previous chapter, a nitrogen limited continuous culture is not a trivial situation for S. cerevisiae. True, single nitrogen limitation does not exist in S. cerevisiae when growing on glucose as carbon and energy source in purely respirative continuous cultures (Larsson et al., 1993; Larsson et al., 1997). The residual concentrations of both glutamine, the nitrogen source used, and glucose are close to zero or undetectable. 6.4.2 Biomass yield S. cerevisiae strains VWk274 LEU+ (Dglt1 mutant) and VWK43 (wild-type) were grown in aerobic continuous cultures with a dilution rate D = 0.1 h^-1 for different nitrogen / carbon ratios in the feed. The theoretical value for the shift from nitrogen to carbon limitation could be calculated from the elemental biomass composition. The composition as determined by Lange et al. (1999) for aerobic glucose limited growth (D = 0.1 h^-1) of VWk43 was used: C₁H_1.77O_0.62N_0.14 (resulting in a Molar Weight of 25.7 g×C-mol^-1). A biomass yield of 10.1 g×l^-1 (Fig. 6.1) for 37.2 mM glutamine and 107.8 mM glucose in the feed and with 3.9 mM residual glutamine, resulted in a yield of 0.57 C-mol×C-mol^-1 for the wild-type under glucose limitation, with glutamine as nitrogen source and the biomass composition according to Lange et al. (1999). Assuming a constant biomass nitrogen content of 0.14 [N-mol×C-mol^-1], the limitation was expected to switch at a N/C feed ratio of 0.08 [N-mol×C-mol^-1]. The experimental results obtained, are collected in Fig. 6.1. For increasing concentrations of glutamine in the feed, the biomass concentration increased. Up to a N/C ratio of approximately 0.08 N-mol×C-mol^-1 no residual glutamine was detectable in the fermentation broth for the steady-states. In this range sometimes small amounts of residual glucose could be detected, but without a (clear) tendency. For the lowest ratio used for the wild- type (2.1 g×l^-1 glutamine and 20.6 g×l^-1 glucose, resulting in a N/C » 0.04) the culture was in a respiro-fermentative state as some residual ethanol was detected (data not shown). The evolution of the different biomass yields with respect to the concentration of glutamine in the feed is shown in Fig. 6.2. The biomass yield on glutamine Y_X/gln was not constant and decreased for higher concentrations of glutamine in the feed. The biomass yield on glucose Y_X/glc initially increased and reached a plateau for a N/C ratio in the feed of approximately 0.08 N-mol×C-mol^-1. Also when the nitrogen source was limiting (i.e. residual glutamine was undetectable) almost all glucose was consumed. Since it could not be (efficiently) used for biomass synthesis, the yield on glucose decreased for decreasing N/C ratios. The changing biomass yield on glutamine for the whole range of ratios applied, indicated a changing biomass composition. Likely the nitrogen content in biomass was low for low N/C ratios and initially rather constant (when Y_X/gln did not change much, Fig. 6.2). The carbon content for low N/C ratios could be higher due to an increased pool of storage carbohydrates. For higher N/C feed ratios an increased N/C ratio in the biomass resulted in a decreasing yield on glutamine (probably due to an increasing nitrogen content and at the same time decreasing synthesis of storage carbohydrates). At even higher N/C feed ratios, when glucose became the only limiting substrate (and residual glutamine was present), the yield on glutamine Y_X/gln continued to decrease, possibly because glutamine was used as an additional energy and carbon source.

Chapter 6 130 Based on the biomass composition of the wild-type, the macro-chemical-equation for aerobic growth on glutamine (C₅H₁₀O₃N₂) and glucose (C₆H₁₂O₆) and without the formation of any by-products can be written as: with yield coefficients Y_glc/X, Y_OX, Y_gln/X and Y_CX [mol×mol^-1] relating biomass to glucose, oxygen (as electron donor), glutamine and CO₂ production respectively. The formation of water has been omitted from the equation. The theoretical Y_gln/X which should be obtained when all nitrogen in the feed is incorporated into biomass was 14.3 C-mole biomass per mole glutamine (= 2.52 g×g^-1) with glucose as main carbon source. Based on the theoretical yield on glutamine the expected yield on glucose Y_glc/X was calculated according to ( ) ( ) ex feed ex feed X n g l X glc glc glc n gl n gl Y Y ] [ ] [ ] [ ] [ / / - - = (6.6) - C₅H₁₀O₃N₂ - (Y_glc/X)^-1 C₆H₁₂O₆ - (Y_OX)^-1 O₂ + Y_gln/X C₁H_1.77O_0.62N_0.14 + (Y_CX)^-1 CO₂ = 0 (6.5) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.00 0.02 0.04 0.06 0.08 0.10 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 ethanol C-limitation N-limitation ‘double’ limitation N/C [mol/mol] Fig. 6.2 Biomass yields for different glutamine / glucose ratios in the feed of continuous cultures (D = 0.1 h^-1) for wild-type and Dglt1 mutant. u Y_gln/X wild-type VWk43 [g×g^-1], n Y_glc/X wild-type [g×g^-1], s Y_gln/X Dglt1 mutant VWk274 LEU+ [g×g^-1], and l Y_glc/X Dglt1 mutant [g×g^-1]. The solid line is the theoretical yield on glucose for both wild-type and mutant, assuming a constant biomass composition. The dashed lines are a suggestive interpolation of the data. 2 g×l^-1 glutamine and 20 g×l^-1 glucose results in a N/C of 0.04 (N-mol×C-mol^-1) and 3.8 g×l^-1 glutamine and 20 g×l^-1 glucose yields a ratio of 0.078. The circles indicate the average values from Table 6.2.

Physiology of GOGAT negative S. cerevisiae   131 This relation has been included in Fig. 6.2 (solid line). The theoretical yield was only slightly lower than the experimental values for the wild-type and the correlation was correct, which in fact was unexpected. As just discussed, the biomass was expected to change with a changing N/C feed ratio. For the calculation of the theoretical yield a constant biomass composition was assumed. Nevertheless it correlated well with the experimental values. This could only be explained by assuming that during nitrogen limitation the uncoupling fluxes were proportional to the excess glucose flux while the storage carbohydrate synthesis fluxes and the varying biomass composition led to the offset between theoretical and experimental yield values. In conclusion, purely glucose limiting conditions seemed to start at a nitrogen / carbon feed ratio above approximately 0.08 when residual glutamine became detectable, in agreement with the expected value. Below ratios of approximately 0.06 N-mol×C-mol^-1 the yield on glutamine was more constant, indicating that (mainly) the glutamine feed concentration determined the biomass concentration in the fermenter. Below a N/C of approximately 0.04, growth was respiro-fermentative. A hypothesis to explain the ethanol formation at a certain limiting nitrogen level could be derived from the common explanation of respiro-fermentative growth in S. cerevisiae (the Crabtree effect). A limited capacity of the respiratory chain (reoxidising NADH while generating ATP) is assumed. When the flux through glycolysis exceeds the capacity of the oxidative phosphorylation, then the excess of reduced equivalents is reoxidised by producing ethanol (Sonnleiter and Käppeli, 1986). The growth rate at which the Crabtree effect becomes visible is called the critical growth rate m_crit. A decreasing nitrogen content 0.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.00 0.02 0.04 0.06 0.08 0.10 N/C [mol/mol] ‘double’ limitation C-limitation N-limitation respiro-fermentative growth (ethanol formation) respiratory growth Fig. 6.3 Hypothetical relation between nitrogen / carbon ratio in the feed and the critical growth rate m crit for onset of respiro-fermentative growth. The dashed line indicates the profile of m_crit . The m_crit in carbon limited cultures has been determined in accelerostat experiments.

Chapter 6 132 of the cell for an increasing nitrogen limitation will reduce the levels of the nitrogen containing macromolecules in the cell. Severe nitrogen limitation will affect the protein machinery of the cell, i.e. the amount of enzymes available. It can be hypothesised that in this situation: 1) only most essential pathways are maintained and 2) the maximum capacity of those pathways will decrease with a decreasing N/C feed ratio (until the cells are starved). Therefore, also the capacity of the respiratory chain will decrease with a decreasing N/C feed ratio and this results in a decreasing m_crit. Based on the experimental findings it is suggested that for a N/C feed ratio of approximately 0.04 mol×mol^-1, the critical growth rate has been reduced to 0.1 h^-1, which was the dilution rate of the continuous culture experiments. This hypothesis is visualised in Fig. 6.3. The shape of the curve is purely speculative. The different physiological windows for varying N/C feed ratios seem to be comparable for the wild-type VWk43 and the Dglt1 mutant VWk274 LEU+ (Fig. 6.1 and 6.2). However, in all the steady-states, the biomass yield of the Dglt1 mutant was significantly lower than of the wild-type. The theoretical yield on glucose has been calculated to be the same for the mutant and the wild-type (Fig. 6.2). The lower experimental yield indicates that for the mutant it is not correct to assume that no by-products were formed (Eq 6.5). In the second part of this chapter the dynamic responses of the glutamine limited wild- type and mutant cells to different nitrogen pulses will be investigated. For these pulse experiments a nitrogen / carbon feed ratio of approximately 0.05 N-mol×C-mol^-1 was aimed at (a combination of 20.0 g×l^-1 and approximately 2.5 g×l^-1 glutamine has been used), which resulted in glutamine limited, completely oxidative growth. Although the actual feed ratios were prone to (significant) variance, as mentioned before, the experiments based on a resembling feed ratio have been taken as replicate experiments. Average steady-state data are shown in Table 6.2 and will be used as reference for further evaluation and discussion. In the reference steady-state the biomass yield of the Dglt1 mutant was 25 % lower than for the wild-type (5.9 vs. 7.9 gl^-1). 6.4.3 Biomass composition As already suggested by the biomass yields for the different nitrogen / carbon feed ratios, the nitrogen / carbon ratio in the biomass for a glutamine limited culture was lower than for a carbon limited culture. For the reference N/C feed ratio the biomass composition of wild- type VWK43 for glutamine limited, aerobic growth (D = 0.1 h^-1) was C₁H_1.72O_0.56N_0.12 (resulting in a MW of 24.3 g×C-mol^-1). The nitrogen content was 9.3% lower than reported for carbon (glucose) single limitation at the same growth rate (Lange et al., 1999). The biomass composition of the Dglt1 mutant VWk274 LEU+ under the same growth conditions was: C₁H_1.71O_0.53N_0.10 (MW = 23.6 g×C-mol^-1). Compared to the glutamine limited wild-type, the nitrogen content of the GOGAT mutant was 15.1% lower. With a decreasing nitrogen content, the degree of reduction of the biomass increased from 3.93 for the glucose limited grown wild-type to 4.24 and 4.35 for the glutamine limited grown wild-type and Dglt1 mutant respectively. The low biomass nitrogen content made the lower experimental biomass yields of the mutant for glutamine limitation (Fig. 6.1 and 6.2, Table 6.2) even more striking. These results indicated that, while the nitrogen from the feed was completely consumed, a limited capacity of a certain pathway prevented complete metabolism of the nitrogen substrate. This bottleneck resulted in overflow metabolism and

Physiology of GOGAT negative S. cerevisiae   133 therefore secretion of organic nitrogen. This is a hypothesis in analogy to the previously discussed explanation for the Crabtree effect. Also during respiro-fermentative growth all carbon from the feed is consumed, but the glycolytic flux is too high for complete oxidation and carbon by-products (e.g. ethanol) are secreted. Based on the data of the feed (Table 6.2) and the biomass composition, the theoretical biomass yields have been calculated. For the wild-type, on average, 40.5 mM of nitrogen was available in the feed (as a combination of glutamine and ammonia, the latter resulted from glutamine degradation). The theoretical (average) biomass yield was 8.2 g×l^-1 for the Table 6.2 Steady-state concentrations in glutamine limited continuous cultures of wild-type and Dglt1 mutant (D = 0.1 h^-1, N/C = 0.05). Averages of n = 9 and n = 4 experiments for wild-type and mutant respectively. wild-type VWk43 Dglt1 mutant VWk274 LEU+ average std average std glutamine in feed [mM] 16.9 1.8 17.6 2.0 ammonia in feed [mM] 6.6 2.5 5.7 - glucose in feed [mM] 113.6 3.0 111.5 0.8 DCW [g×l^-1] 7.9 0.5 5.9 0.3 N-content biomass [N-mole×C-mole^-1] 0.120 - 0.099 - MW biomass [g×mole^-1] 24.3 0.5 23.6 0.2 residual glutamine [mM] 0.0 0.0 0.0 - residual ammonia [mM] 0.0 0.0 0.0 0.0 residual glucose [mM] 0.6 0.5 0.4 0.0 OUR [mmol×l^-1×h^-1] 34.0 34 CER [mmol×l^-1×h^-1] 37.0 38.0 nitrogen recovery [%] 97.0 70.0 carbon recovery [%] 96.0 98.2 Intracellular glutamine [mmol×gX^-1] 29.0 9.8 93.5 48.1 glutamate [mmol×gX^-1] 320.8 122.0 102.0 30.8 ammonia [mmol×gX^-1] 175.8 97.11 14.8 7.9 a-ketoglutarate [mmol×gX^-1] 9.4 1.9 5.2 1.5 NAD [mmol×gX^-1] 2.70 0.05 2.70 0.15 NADH [mmol×gX^-1] 1.24 0.06 0.08 0.06 NADP [mmol×gX^-1] 0.27 0.02 0.27 0.15 NADPH [mmol×gX^-1] 0.30 0.04 0.10 0.10 GSSG [mmol×gX^-1] 2.0 0.3 5.5 1.9 GSH [mmol×gX^-1] 41.0 5.1 25.5 3.8 aspartate / aspargine¹⁾ [mmol×gX^-1] 29.0 11.0 10.8 2.6 serine ¹⁾ [mmol×gX^-1] 10.1 3.3 7.1 3.0 alanine ¹⁾ [mmol×gX^-1] 34.8 14.7 9.9 4.7 1) Only the amino acids with a free pool size larger than 10 mmol×gX^-1 have been included.

Chapter 6 134 wild-type. The actual biomass yield for the wild-type was only 4 % lower. The feed for the Dglt1 mutant contained on average 41 mM nitrogen, resulting in a theoretical yield of 9.8 g×l^-1. The actual biomass yield for the mutant was 40 % lower than this theoretical value. 6.4.4 Residual concentrations As expected, based on the biomass yield, in the fermentation broth of the continuous cultures of the wild-type with a nitrogen / carbon feed ratio around the reference value (0.05) no residual glutamine and ammonia could be detected. Neither were other amino acids, possibly secreted by the cells, present. Strikingly, for the mutant residual organic nitrogen could not be detected either, i.e. in the compounds analysed thus far. The extracellular nitrogen has to be contained in unidentified compounds and this should appear in the nitrogen balance (next section). In all cultures there was some residual glucose present, but no consistent value could be determined for the different experiments (indicated by the high standard deviation, Table 6.2). However, high concentrations of acetaldehyde (approximately 45 ±13 mM) were determined in the steady-state cultures of the Dglt1 mutant whereas the wild-type did not exceed typical ‘background’ values for these fermentation conditions ( approximately 2 mM, results not included in Table 6.2). Based on the yield (difference) and the CO₂ production for the mutant (Table 6.2) a total residual carbon concentration of 129 mM could be expected of which 70% is included in the measured residual acetaldehyde (a C2 molecule). Careful analysis of the different fluxes of acetaldehyde production by the mutant for the experiments around the reference feed ratio of 0.05 [mol×mol^-1] suggested that this production was (negatively) correlated to the N/C ratio. The amount of extracellular acetaldehyde increased rapidly for lower N/C ratios (results not shown). Based on linear extrapolation of the data, the onset of acetaldehyde overflow metabolism is expected to lie below a N/C ratio of 0.06. Furthermore, the acetaldehyde production appears to be extremely sensitive to small changes in the N/C ratio. Roughly estimated, the sensitivity at the reference ratio was calculated to be -3 mol acetaldehyde per (N-mol×C-mol^-1). This explained the high standard deviation obtained for the average of the different experiments around the reference ratio of 0.05. If the sensitivity of the acetaldehyde flux in the mutant is indicative for the metabolic state of the cells under these physiological conditions, than this also could explain the high standard deviations for the other measurements for slightly different N/C ratios (Table 6.2). 6.4.5 Mass balances The carbon balance for the steady-state continuous cultures was calculated as: inflow - outflow = D× (5 gln_feed + 6 glc_feed )× X^-1 - 1000 D× (MW_X)^-1 - ( D× (5 gln_ex + 6 glc_ex + 2 EtOH_ex + 2 acetal + ...) + 1 CER + 2 r_EtOH  )× X^-1 (6.7) with 1000 D× (MW_X)^-1 the biomass outflow in C-mmol, MW_X molecular weight of 1 C-mole biomass [g×C-mol^-1] and CER and r_EtOH the measurable carbon in the exhaust gas. The (organic) nitrogen balance equation reads:

Physiology of GOGAT negative S. cerevisiae   135 inflow - outflow = D (2 gln_feed + 1 NH₄ + feed   + …) × X^-1 - 1000 D× N_X× (MW_X)^-1 - D× (2 gln_ex + 1 NH₄ + ex + ...)× X^-1 (6.8) with 1000 D× N_X× (MW_X)^-1 the biomass outflow (N-mmol), N_X nitrogen content of biomass [N-mol×C-mol^-1], resulting from the biomass formulas. As shown in Table 6.2 the carbon recovery was very acceptable for both wild-type and mutant (96% and 98% respectively) and also 97% recovery of nitrogen for the wild-type was good, indicating no unknown by-products were produced. Due to the unidentified nitrogen containing compound(s) produced by the mutant, only 70% of the nitrogen inflow could be recovered. In total 10.6 mM of unidentified residual nitrogen should be present. Flux Calculations The substrate consumption and product secretion rates in the glutamine limited reference steady-state of the wild-type VWk43 and the Dglt1 mutant VWk274 LEU+ were calculated with Eq. 1.6 and the data from Table 6.2. The results showed the extent of the secretion of acetaldehyde in the mutant (Table 6.3). The lower efficiency of the mutant with respect to both glutamine and glucose metabolism was confirmed. 6.4.6 Intracellular metabolites The concentrations of the intracellular carbohydrates, which were determined in the reference steady-states with standard HPLC, were in the same range for the wild-type and Dglt1 mutant. Except for acetaldehyde, which had very low levels in the wild-type, but reached up to 0.7 mmol×gX^-1 in the mutant. a-Ketoglutarate plays an essential role in the interaction between carbon and nitrogen metabolism. The steady-state a-ketoglutarate levels in the wild-type and GOGAT mutant were approximately the same, but are 3 - 4 times lower than the concentrations reported by Ter Schure et al. (1998) for a different strain. Significant differences in the steady-state levels of intracellular free ammonia and several amino acids were measured (Table 6.2). Glutamine in the mutant is approximately 3 times higher than for the wild-type (93.5 vs. 29.0 mmol×gX^-1). On the other hand glutamate and ammonia are lower in the mutant, glutamate 3 times (102.0 vs. 320.8 mmol×gX^-1) and ammonia even more than 11 times lower (14.8 vs. 175.8 mmol×gX^-1). The levels of glutamine and glutamate were much higher than reported by Ter Schure et al. (1998). In general the pool size of the free amino acids is larger in the wild-type than in the mutant (often a factor 2 to 4). For the glutathione levels Table 6.3 Transport fluxes in glutamine limited continuous cultures of wild-type and Dglt1 mutant (D = 0.1 h^-1, N/C = 0.05) wild-type VWk43 Dglt1 mutant VWk274 LEU+ glutamine uptake [mmol×gX^-1×h^-1] -0.22 -0.30 ammonia uptake [mmol×gX^-1×h^-1] -0.08 -0.09 glucose uptake [mmol×gX^-1×h^-1] -1.44 -1.90 acetaldehyde secretion [mmol×gX^-1×h^-1] 0.02 0.77

Chapter 6 136 (glutathione containing 3 nitrogen in reduced form (GSH) and 6 in oxidised form (GSSG)) the difference between wild-type and mutant in the total nitrogen stored in glutathione is smaller (135.0 vs. 109.5 mmol×gX^-1). Striking are some of the large standard deviations between the different experiments. Both from the elemental biomass composition and from the intracellular pool concentrations it is clear that the Dglt1 mutant has a lower nitrogen content, in free pools, but probably also bound in proteins and perhaps lipids. 6.4.7 Redox state In the wild-type the NADP⁺ and NADPH concentrations were approximately 15% of the NAD⁺ and NADH concentrations (Table 6.2). The oxidised redox equivalents showed similar levels during the steady-state for wild-type and Dglt1 mutant. In contrast, Guillamon et al. (1999) showed that the NADH and NADPH levels in all the samples analysed of the mutant were very close to zero (while in the wild-type the NADH concentration was approximately 50% of the NAD⁺ concentration). As expected this striking difference between wild-type and mutant also appeared in the ratio of reduced vs. oxidised glutathione, although less pronounced. For the wild-type GSH / GSSG = 20.5 [mol×mol^-1] and only 4.6 for the mutant. It is very likely that the incomplete metabolism of the nitrogen contained in glutamine for the Dglt1 mutant is directly related to the unbalanced redox state. Verduyn et al. (1991) were able to relate the biomass yield to the redox state in the yeast Hansenula polymorpha. An increasing drain on intracellular NADH by increasing the NH₄ + NH₄ + gln glu gln NH₄ + TCA glc aKG ATP ADP NAD NADH p r o t e i n NADH NAD pyr aKG GSH / G S S G EtOH acetaldehyde EtOH NADH NAD NADH NAD NAD NADH NADH NAD NADH NAD GOGAT NADPH NADP Acetyl-CoA 3-P glycerate serine NADH NAD ADP ATP O₂ NH₄ + ? Fig. 6.5 Central Nitrogen Metabolism in Saccharomyces cerevisiae, a hypothesis for aerobic growth on glutamine and glucose. ?: indicates unknown transporter of a-ketoglutarate.

Physiology of GOGAT negative S. cerevisiae   137 exogenous flux of H₂O₂ resulted in decreased biomass yields for a catalase negative mutant growing on hydrogen peroxide / glucose mixtures. The destruction process of H₂O₂ requires reduction equivalents which must be provided by the respiratory chain (H₂O₂ partly replaced oxygen as electron acceptor). Extra energy is required for the higher rate of reoxidation and the extra ATP is obtained from glucose metabolism, resulting in a lower biomass yield. The almost undetectable NADH pool explains why the Dglt1 mutant cells secrete acetaldehyde (toxic for the cell) as overflow metabolite instead of ethanol. The Alcohol DeHydrogenase enzyme (present both in cytosol and mitochondria) cannot oxidise acetaldehyde to produce ethanol. The lack of reduced equivalents suggests the presence of either an extra drain for NADH and NADPH or the lack of an important supply compared to the wild-type. This is remarkable because, based on the genetic difference between wild-type and mutant, there is no NADH consumption in the CNM of the mutant, whereas in the wild-type the reaction catalysed by GOGAT needs NADH as cofactor. 1) As suggested in Chapter 5, it cannot be excluded that glutaminase activity results from cycles of synthesis and degradation of amino acids derived from glutamine and degraded to glutamate. An option would be to assume that these reactions also require reduced equivalents and are less efficient than GOGAT. No data on possible cofactors of glutaminase activity have been reported so far. Of the possibilities mentioned in Chapter 5, alanine and asparagine synthesis are redox neutral, formation of arginine and trypophane indeed need NADPH, but histidine synthesis yields NADH (Table 5.3). This consideration does not support the hypothesis of the need of reduced cofactors for glutaminase activity. Another more complex hypothesis is captured in Fig. 6.5. Flux Analysis for growth on glutamine as nitrogen source showed a net flux of a-ketoglutarate from CNM to the TCA- cycle due to the a-ketoglutarate released by the synthesis of many amino acids with glutamate as nitrogen donor (Van Riel et al., 1998, 2000). Strikingly, a transporter in the mitochondrial membranes for a-ketoglutarate is not known (to our knowledge). The influx of a-ketoglutarate may lower the activity of the TCA-cycle or even partly reverse the flux. As a result the TCA-cycle needs less pyruvate and in a more extreme case the cycle could produce pyruvate or secrete a-ketoglutarate and other cycle intermediates as reported by Albers et al. (1996) for (anaerobic batch) growth on glutamate. As typical for yeast, this different flux towards the TCA-cycle does not result in a feedback regulation lowering the rate of glycolysis or glucose uptake. Carbohydrates tend to accumulate at the level of pyruvate, but can be decarboxylated by Pyruvate DeCarboxylase, yielding acetaldehyde, which normally is oxidised to ethanol to be secreted. However, the changed fluxes through the TCA-cycle directly influence the redox state. Less NADH is produced and there might even be a net consumption. When there are not enough reducing equivalents, then ethanol cannot be formed and again acetaldehyde is secreted. That a lower or possibly reversed activity of the TCA-cycle during growth on amino acids reduces the formation of NADH was supported by the lower glycerol production reported by Albers et al. (1996). A positive role of GOGAT could be suggested in restoring the redox balance. 2) An option which cannot be excluded is that in vivo the reaction catalysed by GOGAT operates in the opposite direction from what has been described (and is based on in vitro studies of the enzyme). Then NADH is produced under glutamate degradation. For growth on glutamine, this means the glutaminases and NADPH-GDH have to carry an

Chapter 6 138 extra flux which provides both glutamate for biosynthesis and cycling of glutamate for NADH (re)generation. The extra flux through NADPH-GDH consumes another reducing equivalent, which is mainly produced in the Pentose Phosphate Pathway (PPP). Because of the excess glucose uptake and glycolytic flux, there is no reason to assume that the PPP could not provide this extra NADPH. The concerted action of GOGAT (in reverse direction) and NADPH-GDH would be a way to convert NADPH to NADH. Yeast does not possess a transhydrogenase which directly transfers NADPH into NADH or vice versa. The combination of NADPH- and NAD-GDH is also able to perform the same function (Chapter 5). 3) A second option for a role of GOGAT in maintaining the redox balance could be its activity in a redox shuttle between cytosol and mitochondria. Most of the reactions for amino acid production are oxidative and since part of the amino acid synthesis takes place in the mitochondria, the concentration of the local mitochondrial NADH pool is probably not as low as the average measured in the cell extracts (Table 6.2). For this hypothesis it is assumed that GOGAT is associated to the mitochondrial membrane, as previously also suggested by the kinetic model (Van Riel et al., 1998). Furthermore it is assumed that the cofactor binding sites are directed towards the mitochondrial matrix such that GOGAT utilises mitochondrial NADH (Fig. 6.5). The concerted action of GOGAT and NAD-GDH then could cycle glutamate while producing NADH in the cytosol, maintaining a correct redox balance. This shuttle mechanism is similar to other shuttles between cytosol and mitochondria, such as the glycerol 3-phosphate shuttle (Larsson et al., 1998) and the glutamate-malate shuttle. In Fig. 6.5 it is assumed GOGAT utilises cytosolic a- ketoglutarate. Another possibility would be to assume that GOGAT withdraws a- Table 6.4 GOGAT activity in different cell fractions. Activity in milli absorbance units of NADH oxidation per minute per mg protein. Protein determined with method of Bradford (1976). Strain GLT1 fraction GOGAT activity [mAbs ×mg^-1×min^-1] - vibrax glass bead lysate 0.9 + vibrax glass bead lysate 5.5 + lysate, total cell homogenate 20.7 + sup. after centrifugation of total lysate 0.2 + raw mitochondria 1.6 + purified mitochondria by Nycodenz gradient 5.2 + purified mitochondria by sucrose gradient 18.0

Physiology of GOGAT negative S. cerevisiae   139 ketoglutarate from the mitochondrial pool. This suggests that GOGAT is also related to transport of a-ketoglutarate between cytosol and mitochondrion (Van Riel et al., 1998). It is clear that the localisation of GOGAT is essential for its physiological function. The localisation of GOGAT was determined by cell fractionation and purification by density gradient centrifugation. The validated method of De Kroon et al. (1999) is relatively mild, allowing the detection of loosely bounded proteins. The results of the activity measurements in the different fractions can be found in Table 6.4. The results did not exclude GOGAT activity in the cytosolic fraction, but the increasing specific activity during purification strongly suggested that GOGAT was associated to the mitochondria. The lower activity after the Nycodenz gradient compared to the sucrose gradient indicated that most of the GOGAT protein was washed off by the Nycodenz environment (or that its activity was inhibited). A weak association with the mitochondria does not exclude that the protein is differently localised depending on the physiological conditions. Most likely, GOGAT is associated to the mitochondrial outer membrane. Whether this would allow the GOGAT cofactor binding sites to oxidise mitochondrial NADH is not clear. The (initial) experimental confirmation of the previous model suggestions is very important for further improvement of the mathematical models. 6.5 Response to glutamine and glutamate pulses 6.5.1 Biomass after pulses Biomass yield The wild-type and Dglt1 mutant cultures were pulsed with different glutamine and glutamate concentrations. In general the changes in the biomass concentration within 2 hours after the pulses were below 10%, except for the 40 mM glutamate pulse to the mutant for which the biomass concentrations after 2 hours was 27% lower than the steady-state value (results not shown). There was no (obvious) pattern when the biomass concentrations at t = 120 minutes after the different pulses were compared. 0.092 0.096 0.100 0.104 0.108 0.112 0.116 0.120 0 3 0 60 90 120 time [min] Fig. 6.6 Evolution of the N/C content in the biomass of the Dglt1 mutant VWk274 LEU+, after a 20 mM glutamine pulse to an aerobically growing, glutamine limited chemostat (D = 0.1 h^-1) with N/C feed ratio of 0.05 [N-mol×C-mol^-1]. The error bars indicate the standard deviation of 2 experiments.

Chapter 6 140 Biomass composition Sierkstra et al. (1994) found that a 100 mM pulse of NH₄ + did not result in a dramatic response of S. cerevisiae except for a rapid change in the free amino acids pool. It was concluded that S. cerevisiae is not able to accelerate growth upon addition of an ammonium pulse, in contrast to the response to a glucose pulse of a carbon limited continuous culture. This has been one of the fundamental assumptions for the two mathematical models of the CNM in S. cerevisiae developed so far (Van Riel et al., 1998, 2000). Besides for the steady-states, the elemental biomass composition has been determined during 2 hours after a 20 mM glutamine pulse to the Dglt1 mutant (Fig. 6.6). Within 10-20 minutes already a clear increase in the nitrogen content was visible. After 2 hours the biomass composition of the Dglt1 mutant was C₁ H_1.72 O_0.54 N_0.11, i.e. the nitrogen content had increased by 14% compared to the steady-state. While the nitrogen content of the biomass increased, the degree of reduction slightly decreased from 4.35 in steady- state to 4.30 two hours after the pulse. 6.5.2 Substrate uptake after pulses It is assumed that the kinetics of the transporters for the extracellular substrates glutamine and glutamate (indicated as S_ex) can be described by first-order Michaelis-Menten kinetics: ex S ex a x m S K S V + = f (6.9) The initial uptake rates f_ini after the pulses were determined following the concentrations in the supernatant of glutamine and glutamate after the pulses to derepressed (glutamine limited) continuous cultures of the Dglt1 mutant and wild-type (a typical example is shown in Fig 6.7). The initial uptake rates, based on the average data and corrected for the dilution by the feed, can be found in Table 6.5. With the combination of two pulse sizes A 0 2 4 6 8 10 12 0 30 60 90 120 time [min] B 0 2 4 6 8 10 0 30 60 9 0 120 time [min] 0.0 0.2 0.4 0.6 0.8 1.0 Fig. 6.7. Extracellular glutamine (u) and ammonium (n) after 10 mM glutamine pulses to (A) wild- type and (B) Dglt1 mutant in aerobic, glutamine limited chemostats (D = 0.1 h^-1) with N/C feed ratio of 0.05 [N-mol×C-mol^-1] The error bars indicate the standard deviation of three experiments. The steep dashed line indicates the initial uptake rate (Table 6.5) and the more flat dashed line the wash-out profile.

Physiology of GOGAT negative S. cerevisiae   141 and two resulting uptake rates, just enough information is available to calculate the apparent transporter capacities V_max and the substrate affinities K_S of the first-order Michaelis-Menten kinetics (Table 6.5). The results are only rough indications of the possible parameter values of the lumped capacities and affinities for this physiological situation. Nevertheless, it is clear that the uptake of the pulsed glutamine and glutamate is dominated by high capacity, low affinity transporters. On the other hand, the very low residual glutamine concentration in steady-state is transported by a permease with a high affinity. Such a combination of transporters has been reported before for other substrates, usually with a low capacity for the high affinity transporter. The most logical explanation for the differences in initial uptake between wild-type and mutant would be an inhibition by the intracellular concentrations. The higher levels of intracellular glutamine in the mutant could cause (stronger) inhibition of the uptake after the pulses. The parameters can be determined more accurately by applying non-linear regression to the original datasets of the individual experiments. The information of Table 6.5 can be used for kinetic modelling and DOMC models, in which the maximum rates are used as upper boundaries for the uptake fluxes. Together with the glutamine pulses inevitable also a small amount of ammonia was pulsed due to the degradation of glutamine during preparation. The ammonium was simultaneously consumed with the glutamine (e.g. Fig 6.7). 6.5.3 Relative importance of different pathways Response types The responses of intracellular amino acids, redox cofactors, glutathione, carbohydrates and mRNA have been analysed (Guillamon et al., 1999). Several qualitative response types appear to be very general. Of the intracellular amino acids, redox cofactors and glutathione, a total of 66 good measurable responses have been analysed qualitatively. Besides 2, all the other 64 responses could be grouped into 10 classes (Fig. 6.8). Most responses clearly suggest control of homeostasis, as defined in the DOMC model framework. Within two hours after the pulse the tendency is towards the steady- state values. Classes C, G, and I (Fig. 6.8) seem to reach a (temporary) new homeostatic value. A few responses show a continuing drift after a pulse (class H). The compounds in class J hardly respond and only fluctuate a bit. Class A to C are often observed in the intracellular concentrations of the pulsed substrates. The small temporary decrease after an initial fast increase (B and C) is typical. Such ‘fast dynamics’ are also present in the Table 6.5 Apparent substrate uptake kinetics after pulse addition to aerobic, glutamine limited chemostats (D = 0.1 h^-1) with N/C feed ratio of 0.05 [N-mol×C-mol^-1]. Wild-type VWk43 Dglt1 mutant VWk274 f ini [mmol×l^-1×h^-1] V_max [mmol×l^-1×h^-1] K_S [mM] f ini [mmol×l^-1×h^-1] V_max [mmol×l^-1×h^-1] K_S [mM] glutamine 10 mM 75 16 glutamine 20 mM 105 1.6×10² 1×10¹ 32 8×10¹ 4×10¹ glutamate 20 mM 19 18 glutamate 40 mM 1) - - 30 9×10¹ 8×10¹ 1) No clear uptake pattern, no relevant f_ini could be calculated.

Chapter 6 142 initial response of classes D, E (the largest class) and G, and to some extent F. Classes D, E and F are related to compounds which are significantly disturbed after the pulses, but which could be regarded as strongly controlled towards homeostasis. These same response types have been observed before in experimental data for different strains (Ter Schure et al., 1998) and model simulations (Van Riel et al., 1998, 2000). The actual responses are summarised in Table 6.6. The minimum and maximum deviation from the steady-state value have also been indicated. Based on the proposed classification the actual responses can be more easily discussed. The pulsed glutamine and glutamate was taken-up by the cells. The concentration of the pulsed nitrogen source increased in the cell, following a response A for glutamine pulses to the wild-type strain and type B for the glutamate pulses to the same strain. This increase was much faster (abrupt) for glutamate than for glutamine in case of the wild- type, although maximum uptake for glutamate was lower. This indicates also the intracellular turnover rate of glutamine increased instantaneously after the glutamine pulse, whereas this not occurred for glutamate. The uptake patterns for the Dglt1 mutant are less consistent. In the mutant the intracellular accumulation of glutamine and 0 30 60 90 120 time [min] A B C D E F G H I J Fig. 6.8 After glutamine and glutamate pulses to both wild-type and Dglt1 mutant the responses of the intracellular amino acids, redox cofactors and glutathione, have been grouped into 10 classes (numbers behind the characters indicate size of the class): A (5), B (9), C (7), D (7), E (15), F (7), G (9), H (2), I (3), J (6). 2 responses could not be classified. - Class A: (strong) accumulation and gradual decrease, simple dynamics - Class B and C: a characteristic small, temporary decrease after an initial fast increase - Classes D and E: significant fluctuations around steady-state value. - Classes C, G, and I: move to new homeostatic reference value. - Class H: continuing drift - Class J: almost no response, only small fluctuations

Physiology of GOGAT negative S. cerevisiae   143 glutamate, respectively, occurred with a similar rate. Next the responses are discussed based on the substrate type. Glutamine pulses As expected, the wild-type had a higher metabolic capacity for glutamine catabolism than the GOGAT negative mutant. In the wild-type GOGAT carried a significant flux after the glutamine pulses, in agreement with previous model predictions (Van Riel et al., 1998, 2000). This resulted in a higher glutamine consumption rate and more glutamate production than with the glutaminases alone. Within 2 hours after the 10 mM glutamine pulses, both the wild-type and the mutant had consumed approximately 6 mM glutamine. After the 20 mM glutamine pulses the wild-type consumed 16 mM and the mutant 13 mM within 2 hours. After the pulses the glutamine uptake by the mutant slowed-down and aborted earlier than for the wild-type, probably due to the earlier onset of catabolic repression. In agreement with this lower uptake rate, the maximum intracellular glutamine concentrations after the glutamine pulses in the wild- type were 35 to 45% higher than in the Dglt1 mutant (370 vs. 200 mmol×gX^-1 after 10 mM pulse and 770 vs. 500 mmol×gX^-1 after 20 mM glutamine pulse respectively). Also intracellular metabolism of glutamine was slower in the GOGAT negative mutant, especially after the 20 mM pulse (intracellular glutamine followed a class C response pattern). The intracellular glutamate concentration almost remained at the (low) level of 100 mmol×gX^-1 in the mutant and doubled in the wild-type from 250 to 500 mmol×gX^-1 after the 20 mM glutamine pulse (Table 6.6). For the 10 mM glutamine pulses the responses of wild-type and mutant were similar (type F), but the initial decrease in glutamate was relatively larger in the mutant. The glutaminases could not carry the same flux for glutamate production as the combination of GOGAT and glutaminases in the wild-type. This was probably due to the limited capacity of the glutaminases (which was smaller than the combined capacity of the two pathways in the wild-type). Product inhibition of the glutaminases was excluded since both the glutamate and ammonium levels in the mutant were lower than in the wild-type, in the steady-state as well as after the glutamine pulses. In the mutant glutamine degradation resulted in an equimolar production of glutamate and ammonia. Based on the decrease in glutamate compared to the 50% increase in ammonia after the 10 mM glutamine pulse, it was concluded that glutamate consumption was higher than the assimilation of ammonia. This suggested the operation of NAD-GDH, resulting in glutamate consumption and ammonia production. After the 20 mM pulse to the mutant it was the opposite: both ammonia and glutamate increased (type C and E respectively), but glutamate approximately 40 mmol×gX^-1 and ammonia only 26 mmol×gX^-1, suggesting NADPH-GDH was active. (Despite the increase of intracellular ammonium in the mutant after the glutamine pulses the levels remained much lower than in the wild-type.) The DOMC model predicted that in the wild-type, besides the flux through GOGAT, also the flux through GDA should increase after the glutamine pulses. The observed net decrease of the intracellular ammonium concentration in the wild-type after the 10 mM glutamine pulse (response G) and the lack of a clear increase after the 20 mM pulses (response E), indicated an increased ammonium consumption. NADPH-GDH was an obvious candidate, itself yielding extra glutamate. It is remarkable that the variations in the ammonium pool and glutamate pool hardly affected the a-ketoglutarate pool with a size (buffer capacity) of on average 7 mmol×gX^-1, which is approximately 20 times smaller than the fluctuations observed in ammonium and

Chapter 6 144 glutamate. In the wild-type, a-ketoglutarate has been predicted to decrease 2 mmol×gX^-1 by the DOMC model and 5 mmol×gX^-1 by the kinetic model (Van Riel et al., 1998, 2000) because of an increased flux through both GOGAT and NADPH-GDH. On the other hand, because of the lack of GOGAT and the possible operation of NAD-GDH (as observed) in Table 6.6 Response type of intracellular compounds after the pulses to glutamine limited continuous cultures of wild-type and Dglt1 mutant (D = 0.1 h^-1, N/C in feed 0.05 N-mol×C-mol^-1). The values indicate the minimum and maximum values [mmol×gX^-1] reached, relative to steady- state values in Table 6.2. wild-type VWk43 Dglt1 mutant VWk274 10mM gln 20mM gln 20mM glu 40mM glu 10mM gln 20mM gln 20mM glu 40mM glu Glutamine A -0 +370 A -0 +773 I -0 +136 C -0 +74 B -24 +198 C -0 +501 I -1 +109 A -0 +150 Glutamate F -98 +37 C -0 +255 B -0 +507 B -0 +583 F -75 +0 E -5 +40 B -0 +476 A -0 +1090 Ammonia G -63 +54 E -61 +42 E -0 +37 E -80 +14 B -0 +19 C -0 +26 C -1 +7 A -0 +23 a-keto- glutarate G -1.9 +0.0 un-defined B -0.7 +4.8 E -2.3 +2.3 G -1.2 +0.6 G -3.6 +0.0 J -0.9 +2.8 J -1.7 +0.0 NAD C -0.3 +1.1 n.d. I/J -0.1 +0.3 n.d. D -1.4 +0.3 D -1.1 +0.0 C -0.1 +1.6 B -0.0 +1.9 NADH E -0.3 +0.3 n.d. F -0.9 +0.0 n.d. G -0.0 +0.1 D -0.1 +0.2 G -0.1 +0.1 E -0.2 +0.1 NADP J -0.0 +0.2 n.d. J -0.1 +0.0 n.d. E -0.2 +0.0 E -0.2 +0.0 E -0.1 +0.1 E -0.0 +0.1 NADPH J -0.1 +0.3 n.d. J -0.2 +0.2 n.d. G -0.1 +0.0 E -0.2 +0.1 G -0.1 +0.0 un-defined GSSG E -0.5 +0.9 H -0.1 +1.4 F -0.3 +0.4 F -0.5 +0.0 D -3.0 +0.2 G -3.0 +0.0 D -1.4 +1.6 B -0.0 +2.1 GSH F -0.0 +13.7 H -1.6 +22.0 F -2.0 +11.1 F -6.7 +0.5 D -12.1 +16.4 D -7.0 +0.4 E -0.8 +5.1 B -0.0 +7.4 n.d.: not determined

Physiology of GOGAT negative S. cerevisiae   145 the mutant a smaller decrease, or even an increase of a-ketoglutarate could be expected after the glutamine pulses. The wild-type showed a decrease of 1.9 mmol×gX^-1 after the 10 mM pulse and the mutant a decrease of 1.2 and 3.6 mmol×gX^-1 after a 10 and 20 mM glutamine pulse respectively. The response in a-ketoglutarate after the 20 mM glutamine pulse to the wild-type showed no clear pattern (Table 6.6). This could be related to the different response of glutamate (class C, reaching an increase of 255 mmol×gX^-1) for this case as compared to the other glutamine pulses to both wild-type and mutant. Besides glutamine and glutamate (and ammonia) 12 other amino acids were determined by HPLC. Most of the amino acid pools in the wild-type VWk43 showed a significant increase one hour after the glutamine pulses (up to 10-fold higher concentrations), as previously reported by Sierkstra et al. (1994). In general the increase was larger for the 20 mM glutamine pulse. On the contrary, the mutant data revealed smaller increases (even decreases in some cases). Arganine, serine and alanine showed most significant increases (both relatively and absolutely): 15.8 mmol×gX^-1 (+84%), 18.7 (+170%) and 19.7 (+200%) respectively. For the stoichiometric models of yeast, Van Gulik and Heijnen (1995) and Giuseppin and Van Riel (2000) have included serine synthesis with 3-phosphoglycerate (G3P) and glutamate as precursors and NAD⁺ as cofactor, in agreement with Jones and Fink  (1982). According to Stephanopoulos et al. (1998) and Voet and Voet (1995) the precursor of serine is glyceraldehyde 3-phosphate (GAP) in a reaction consuming NADPH and producing NADH. Alanine is synthesised from pyruvate and glutamate, without requiring a cofactor (Jones and Fink, 1982; Van Gulik and Heijnen, 1995; Giuseppin and Van Riel, 2000). The stoichiometry for arginine synthesis is more complex, but effectively NADPH is used (stoichiometries of 1, 2 and 4 molecules NADPH per arginine can be found in literature, Van Gulik and Heijnen, 1995; Giuseppin and Van Riel, 2000; Stephanopoulos et al., 1998). None of these synthesis reactions requires the scarce NADH and most precursors are derived from glycolysis, of which the flux is not limiting during nitrogen limited growth on glucose. Especially serine synthesis could be relevant for NADH (re)generation in the mutant. Glutamate pulses The increase in intracellular glutamate after the 20 and 40 mM glutamate pulse to the wild-type and the 20 mM pulse to the mutant was similar (approximately 500 mmol×gX^-1, type B). This result is also in agreement with a similar glutamate uptake rate in wild type and the mutant after the 20 mM glutamate pulse (Table 6.5). After the 40 mM pulse the mutant accumulated approximately 500 mmol×gX^-1 within 10 minutes after the pulse, comparable to the other responses of intracellular glutamate, but continued to accumulate and reached 1160 mmol×gX^-1 after 1 hour (following a class A response). For the wild-type, the response type of intracellular glutamate after the 40 mM pulse was of class B, but without a tendency back to the original homeostatic level within 2 hours after the pulse. A similar pattern with fluctuations was observed for the concentration of the extracellular glutamate after it was added, which hampered the calculation of first order uptake kinetics (Table 6.5). Because of the similar increase of the intracellular glutamine in both strains after the glutamate pulses, it was concluded that glutamine synthetase (GS) was active (and not affected by GOGAT deletion). After the glutamate pulses, the intracellular ammonia concentration in the wild-type varied (class E response). GS consumes ammonium and

Chapter 6 146 likely ammonium is generated by NAD-GDH. The mutant showed increased intracellular ammonia concentrations after the glutamate pulses (type C and A responses), but less abrupt than observed for the glutamine pulses. Possibly because the rates of GS and NAD-GDH were almost equal (balanced). For both the wild-type and the mutant an increased a-ketoglutarate could be expected because of the NAD-GDH activity. In the wild-type the intracellular a-ketoglutarate increased after the 20 mM glutamate pulse (type B response) and only fluctuated after the 40 mM pulse (type E). a-ketoglutarate in the Dglt1 mutant increased after 30 minutes for the 20 mM glutamate pulse, but slightly decreased after the 40 mM pulse (similar as after the glutamine pulses). The increase in the intracellular pools of the other free amino acids after the glutamate pulses to the wild-type (maximal 300%) was smaller than for the glutamine pulses. Strikingly, the 20 mM pulse resulted in stronger accumulation than the 40 mM glutamate pulse. Opposite to the glutamine pulses, the glutamate pulses produced also an important net increase in most of the amino acids pools of the Dglt1 mutant after the glutamate pulses, although serine and especially alanine showed the most important increases. 6.5.4 Redox balance Redox cofactors play an important role in the CNM. The cofactor and glutathione evolution after both the glutamine and glutamate pulses was completely different (Table 6.6). There was no clear trend in the ratios of reduced versus oxidised redox cofactors  after the pulses, neither for the GSH/GSSG ratio. In general, the redox state in the GOGAT negative mutant (undetectable reduced redox cofactors in the steady-state) varied more than for the wild-type. Guillamon et al.(1999) suggested a physiological role of the GOGAT enzyme in maintaining a steady redox state in the cell. The role of the glutathione pool as storage of excess nitrogen was also supported: both GSH and GSSG increased, especially in the wild-type after the largest glutamine pulse (22 mmol×gX^-1 within 2 hours). The intracellular NAD⁺ in the wild-type experiment showed an important increase after the 10 mM glutamine pulse (type C response, Fig. 6.9A), possibly due to an increased flux through GOGAT, but this NAD⁺ production did not correspond with a NADH decrease. NADH varied (type E response) and the ratio NAD⁺/NADH increased. These data suggested that the sum of NAD⁺/NADH increased in the first minutes. After both glutamine pulses to the mutant the NAD⁺ concentration varied, reaching a minimum value A 0 1 2 3 4 0 30 60 90 120 time [min] B 0 1 2 3 4 0 30 60 90 120 time [min] Fig. 6.9 Intracellular NAD (s) and NADH (5) after 10 mM glutamine pulses to (A) the wild-type and (B) the Dglt1 mutant in aerobic, glutamine limited chemostats (D = 0.1 h^-1, N/C feed ratio 0.05 [N-mol×C-mol^-1]).

Physiology of GOGAT negative S. cerevisiae   147 at 20 minutes after the pulse (type D responses, Fig. 6.9B), possibly because there was no reduction by GOGAT. NADH remained undetectable. The Dglt1 mutant showed a strong increase in the NAD⁺ concentration after the glutamate pulses (NADH undetectable, results not shown). This could be due to an increased production, but also a decreased consumption of NAD⁺. The first could be caused by a (partly) reversed flux in the TCA-cycle (Fig. 6.5) or several anabolic reactions for amino acid synthesis. A decreased NAD⁺ reduction could be caused by a decreased NAD-GDH reaction. However, NAD-GDH was assumed to be not active in the steady- state and therefore could not decrease, unless the reaction was reversed because of the redox state. In the wild-type, the NADH concentration dropped during the first twenty minutes after the 20 mM glutamate pulse (and recovered the steady-state value later on, following a class F response), whereas NAD⁺ slightly increased, indicating NADH consumption increased or production decreased. For the wild-type the GSH/GSSG ratio increased in time after the glutamate pulses, which could be one of the causes of a decreased NADH concentration. The results suggest that after the pulses the idea of the nicotinamide adenine dinucleotides as conserved moieties is disputable, at least at the global cell level. Fig 6.10 mRNA levels after pulse disturbances. u GAP1, n GDH1 and s GLN1. (A): 10 mM glutamine pulse wild-type, (B): 20 mM glutamate pulse wild-type, ©: 10 mM glutamine pulse Dglt1 mutant, (D): 20 mM glutamate pulse Dglt1 mutant. Northern blot intensities have been normalised to internal control and with the maximum expression level observed as 100%.

Chapter 6 148 6.5.5 Stress After the glutamine pulses, the glutamate concentration increased in the medium and after the glutamate pulses, glutamine accumulated, both for the wild-type and the mutant (results not shown). This was due to secretion of the yeast cells and not just the result from the feed which continued to be added whereas uptake was catabolically repressed. The excretion was smaller for the small pulses. (As mentioned before, the increase of ammonium, immediately after the glutamine pulses was caused by the ammonium present in the pulsed solution due to the breakdown of glutamine during preparation, Fig. 6.7.) This secretion of nitrogen containing compounds by both mutant and wild-type cells, which were glutamine limited just before the pulses, was unexpected and indicates that the control of uptake is not strong or fast enough to deal with the surplus of nitrogen. In the context of the previously introduced cybernetic modelling framework (Chapter 1 and 3), the secretion of, in principle, valuable compounds indicates a stress response. For the mutant this dynamic overflow response added to the unidentified overflow metabolite(s) in steady-state (section 6.4.4). Despite (or due to) the stress responses of the yeast cells after the pulses, the homeostatic state can be recovered by the Nitrogen Catabolite Repression mechanism. 6.5.6 Nitrogen Catabolic Repression Nitrogen Catabolic Repression was studied by Northern analysis of the mRNA of two genes in CNM known to be NCR sensitive and GAP1 was included as a representative of the (general) amino acids transporters, also (highly) sensitive to NCR. GAP1 showed that 20 minutes after the glutamine and glutamate pulses the mRNA could no longer be detected (Fig. 6.10), neither in the wild-type nor in the mutant. GAP1 essentially remained repressed during the period for which transcription was monitored. Repression of GAP1 was faster after the glutamine pulses than for the glutamate pulses. No difference between wild-type and mutant was observed. GDH1 expression (encoding the NADPH-dependent GDH) also decreased after the pulses, but this repression was not as strong and fast as for GAP1. The expression of the gene for Glutamine Synthetase, GLN1, increased during the first 30 minutes (in the mutant) and 60 minutes (in the wild-type) after the glutamine pulse and decreased later on. GLN1 was never completely repressed at any time studied, in contrast to the results of Ter Schure et al. (1998). This was a striking result because in the discussion so far it was assumed that GS was not active after the glutamine pulses. It was also not likely that the GLN1 gene product would be or become active due to the high glutamine concentrations in the cell. After the glutamate pulse, GLN1 expression decreased in the wild-type (but no complete repression) and increased in the firsts minutes in the mutant. As discussed, the GLN1 gene product could be expected to carry a flux after the glutamate pulses, in contrast to steady-state growth on glutamine. Based on the work of Ter Schure et al. (1998), both the intracellular ammonia and glutamine concentration can trigger catabolic repression. Glutamine and ammonia increased after all pulses. For the wild-type it has been predicted that the general transcription activator Gln3p is completely inactivated (Van Riel et al., 1998). A strong catabolic repression could also be expected for the mutant. This is confirmed by the repression of GAP1, of which Gln3p is the dominant transcription activator. Although Gln3p also activates transcription of GLN1 and likely GDH1, in general, these genes were not completely repressed after the pulses, but showed a dynamic response.

Physiology of GOGAT negative S. cerevisiae   149 6.6 Discussion and conclusions Initiated by the model results (Van Riel et al., 1998, 2000), the dynamics of the CNM in S. cerevisiae were further studied by determining the dynamic responses of nitrogen derepressed cells to different substrate pulses. The results of these physiological studies were not straightforward. The analysis of the large dataset, which resulted from the dynamic experiments with different strains and different pulses, was complex. The introduction of a classification of dynamic responses facilitated the qualitative discussion. However, mathematical models, such as previously developed, are indispensable tools for further interpretation. Interpretation of dynamic biological data is usually limited to the identification of trends. Fluctuations (oscillations) are often not discussed and unexpected, relative fast changes in data are often classified as outliers or experimental variation. The mathematical models (Van Riel et al., 1998, 2000) have predicted the type of dynamics observed in the replicate experiments presented here and by Guillamon et al. (1999). This indicates that real system characteristics are observed. For example, in many responses analysed there was a temporary decrease twenty minutes after the pulse (response types B and C, Fig. 6.8). This pattern was also found by Ter Schure et al. (1998) in other strains and described or predicted by the models (Fig. 2.4, 2.5 and 3.4 - 3.7 ). From the mRNA analyses it can be concluded that the response after 20 minutes coincides with complete effective NCR and the onset of other regulation (induction) at the transcription level (Fig. 6.10). As predicted by the models (Fig. 2.5d, 2.8d, and 3.5, 3.7), after 20 - 30 minutes also a plateau was visible in all uptake patterns of the pulsed substrates due to a repressed uptake (Fig. 6.7). Guillamon et al. (1999) suggested that the accumulated intracellular amino acids could be used by the cell to increase protein synthesis. This synthesis must start about 10 to 20 minutes after the intracellular amino acids pools increased. Such increased protein synthesis is in agreement with the increasing nitrogen content of the biomass 10 to 20 minutes after the pulse (Fig. 6.6). These data are especially important for a quantitative physiological approach. To our knowledge this is the first report which includes the effect of substrate pulses on the elemental biomass composition. Despite the effort so far, for a completely quantitative approach also the molecular biomass composition of wild-type and mutant for glutamine limited growth, before and after the pulses should be determined. The standard deviation between the different experiments is much higher than that of multiple measurements of a single sample. Although internal procedures have been standardised, slightly different culture conditions and sample handling, by different experimenters, resulted in significant variances. As suggested by the acetaldehyde production by the mutant in steady-state, these variances could be caused by the slightly different N/C feed ratios used. Data accuracy as reported by Lange et al. (1999) for different steady-states has not been obtained, especially not for the dynamic response data. For the results presented, relatively small changes in the N/C feed ratio and in the amount of nitrogen pulsed, resulted in qualitatively different responses. Unexpected phenotypes of a mutation in a single gene were revealed. A linear paradigm and similar model is not

Chapter 6 150 sufficient to interpret such data. The Dynamic Optimal Metabolic Control framework (Giuseppin and Van Riel, 2000; Van Riel et al., 2000) is more suitable to explain why different pulse sizes can result in qualitatively different responses. In this concept the cell is regarded as an optimally controlled system with strategies. The responses of the cell to substrate pulses are determined by the dynamic balance between the different postulated strategies. Besides as the base for the incorporation of regulation in metabolic models, the DOMC concept is also valuable for the biological interpretation of the (fast) dynamic responses. A fluctuating, varying or oscillating pattern was observed in many analysed compounds (response classes D, E and F) and an unbalanced state in the cells as a consequence of the pulses could be suggested. In analogy with manmade, controlled systems, such fluctuations (‘overshoot’ in engineering terminology) could indicate non- optimally tuned control. The substrate pulses perturb the cellular homeostasis and the cell reacts to stabilise the intracellular balances. The CNM structure allows quick looping or cycling to increase or decrease a certain product. The relatively fast dynamics observed, relate to the (internal) control of a flexible system, which attempts to rebalance after a perturbation. The responses to large substrate pulses are probably dominated by complete catabolic repression, such as observed by Ter Schure et al. (1998). Small pulses could trigger more subtle regulatory mechanisms. Likely, first the flexibility of the metabolic network is exploited before a more adaptive and definite mechanism, such as NCR, is initiated. This could explain why the smaller pulses often resulted in the most profound intracellular responses of metabolites and cofactors. In agreement with the DOMC model of CNM in yeast (Van Riel et al., 2000), the experimental data confirmed that the regulation of the genes of the CNM is not a pure repression because of high glutamine and ammonium levels in the cell. The observed responses are the result of a combination of the different regulators involved and their dynamic balance. The results yield a stronger experimental base for the DOMC framework. Based on the results of Ter Schure et al. (1998) for different strains and larger substrate pulses, the kinetic model (Van Riel et al., 1998) contained Gln3p as the key transcription regulator of NCR. From the results presented here, it is clear that this concept needs to be revised. For most genes in CNM also other transcription factors besides Gln3p are known (e.g. Ter Schure et al., 1999). These might prevent complete repression or even result in stronger induction after the pulses as observed for GDH1 and GLN1 respectively (Fig. 6.10). In the next chapter the balance between the different known activators and repressors will be investigated by a mathematical model. The mass balance analyses showed a consistent dataset for the wild-type in steady-state and revealed a 30% shortage in the nitrogen balance for the Dglt1 mutant. Under glutamine limited steady-state growth, an unidentified, nitrogen-rich compound was secreted by the mutant. The comparison of a wild-type strain and GOGAT negative mutant has confirmed the important role of GOGAT in the CNM of S. cerevisiae, as previously predicted by the models (Van Riel et al., 1998, 2000). GOGAT helps the cell to maintain homeostasis and to deal with metabolic fluctuations. The cofactor and glutathione analyses showed that GOGAT must be important in the control of the redox state, instead of being subject to it. Possibly, the lower biomass yields of the GOGAT mutant are related to the redox imbalance, as shown before for H. Polymorpha (Verduyn et al., 1991). According to this hypothesis, rebalancing the redox state by exogenous