APPLIED AND ENVIRONMENTAL MICROBIOLOGY, June 2002, p. 2814–2821
`0099-2240/02/$04.00⫹0 DOI: 10.1128/AEM.68.6.2814–2821.2002
`Copyright © 2002, American Society for Microbiology. All Rights Reserved.
`
`Vol. 68, No. 6
`
`Metabolic Engineering of Glycerol Production in
`Saccharomyces cerevisiae
`Karin M. Overkamp,1 Barbara M. Bakker,2 Peter Kötter,3 Marijke A. H. Luttik,1
`Johannes P. van Dijken,1 and Jack T. Pronk1*
`Kluyver Laboratory of Biotechnology, Delft University of Technology, NL-2628 BC Delft,1 and Molecular Cell Physiology,
`Free University Amsterdam, NL-1081 HV Amsterdam,2 The Netherlands, and Institut für Mikrobiologie,
`J. W. Goethe Universität Frankfurt, 60439 Frankfurt, Germany3
`
`Received 22 January 2002/Accepted 1 April 2002
`
`Inactivation of TPI1, the Saccharomyces cerevisiae structural gene encoding triose phosphate isomerase,
`completely eliminates growth on glucose as the sole carbon source. In tpi1-null mutants, intracellular accu-
`mulation of dihydroxyacetone phosphate might be prevented if the cytosolic NADH generated in glycolysis by
`glyceraldehyde-3-phosphate dehydrogenase were quantitatively used to reduce dihydroxyacetone phosphate to
`glycerol. We hypothesize that the growth defect of tpi1-null mutants is caused by mitochondrial reoxidation of
`cytosolic NADH, thus rendering it unavailable for dihydroxyacetone-phosphate reduction. To test this hypoth-
`esis, a tpi1⌬ nde1⌬ nde2⌬ gut2⌬ quadruple mutant was constructed. NDE1 and NDE2 encode isoenzymes of
`mitochondrial external NADH dehydrogenase; GUT2 encodes a key enzyme of the glycerol-3-phosphate shuttle.
`It has recently been demonstrated that these two systems are primarily responsible for mitochondrial oxidation
`of cytosolic NADH in S. cerevisiae. Consistent with the hypothesis, the quadruple mutant grew on glucose as the
`sole carbon source. The growth on glucose, which was accompanied by glycerol production, was inhibited at
`high-glucose concentrations. This inhibition was attributed to glucose repression of respiratory enzymes as, in
`the quadruple mutant, respiratory pyruvate dissimilation is essential for ATP synthesis and growth. Serial
`transfer of the quadruple mutant on high-glucose media yielded a spontaneous mutant with much higher
`specific growth rates in high-glucose media (up to 0.10 hⴚ1 at 100 g of glucose · literⴚ1). In aerated batch
`cultures grown on 400 g of glucose · literⴚ1, this engineered S. cerevisiae strain produced over 200 g of
`glycerol · literⴚ1, corresponding to a molar yield of glycerol on glucose close to unity.
`
`Glycerol is used to synthesize many products, ranging from
`cosmetics to lubricants. Its current annual production of ca.
`600,000 tonnes is mainly recovered as a by-product of soap
`manufacturing or produced from propylene (55). Alterna-
`tively, glycerol can be produced by microbial fermentation,
`using sustainable carbohydrate feedstocks. Although this may
`involve a wide range of microorganisms, including algae and
`bacteria, research has mostly focused on yeasts (1, 55; M. J.
`Taherzadeh, L. Adler, and G. Lidén, submitted for publica-
`tion). In yeasts, glycerol is produced by the reduction of dihy-
`droxyacetone phosphate (DHAP) to glycerol-3-phosphate
`(G3P), a reaction catalyzed by cytosolic NAD⫹-dependent
`G3P dehydrogenase. G3P is subsequently dephosphorylated by
`a glycerol-3-phosphatase (1). Especially during anaerobic
`growth, glycerol production serves as a redox sink to maintain
`the cytosolic redox balance (5, 50). Glycerol also functions as
`an osmolyte, thus enabling yeast growth at high osmolarity (1,
`36). Consistent with the latter role, the highest glycerol yields
`reported to date have been achieved with osmotolerant yeast
`strains (Table 1).
`The amounts of glycerol that are naturally produced by
`Saccharomyces cerevisiae as a response to anaerobiosis and/or
`osmotic stress are relatively small (42, 44, 51). Much effort has
`been invested in attempts to redirect sugar metabolism in this
`
`* Corresponding author. Mailing address: Kluyver Laboratory of
`Biotechnology, Delft University of Technology, Julianalaan 67, NL-
`2628 BC Delft, The Netherlands. Phone: 31 15 278 3214. Fax: 31 15
`278 2355. E-mail: j.t.pronk@tnw.tudelft.nl.
`
`yeast towards glycerol production. The first successful attempt
`was the sulfite process, devised by Neuberg and Reinfurth (34).
`In this process, sulfite added to fermenting S. cerevisiae cul-
`tures forms an adduct with acetaldehyde, thus making the
`latter compound unavailable as an electron acceptor for the
`reoxidation of glycolytic NADH. Instead, NADH is reoxidized
`by glycerol production. This early example of redirection of
`metabolic fluxes has been called metabolic engineering avant
`la lettre (13). Theoretically, the sulfite process can lead to the
`formation of equimolar amounts of glycerol, carbon dioxide,
`and sulfite-acetaldehyde adduct. However, the theoretical yield
`of glycerol on glucose of 0.51 g · g⫺1 has not been achieved, not
`even in modern adaptations of the sulfite process (Table 1).
`Moreover, the presence of by-products (ethanol, acetate, sul-
`fite-acetaldehyde adduct, and biomass) poses problems during
`glycerol recovery (1).
`Over the past decade, research on glycerol production by S.
`cerevisiae has shifted to true metabolic engineering, i.e., the
`application of recombinant DNA technology for a rational
`reprogramming of cellular metabolism (4). Several approaches
`were aimed at minimizing the reduction of acetaldehyde to
`ethanol, thus mimicking the sulfite process. Indeed, reduced
`expression of pyruvate decarboxylase (35) and deletion of al-
`cohol dehydrogenase genes (12) led to an increased production
`of glycerol, but the glycerol concentration did not exceed 5
`g · liter⫺1 (Table 1). Other strategies focused on overexpres-
`sion of the key enzymes of the glycerol pathway in S. cerevisiae
`(33, 39, 44). Overproduction of the GPD1-encoded cytosolic
`G3P dehydrogenase led to an increased production of glycerol
`
`2814
`
`

`

`VOL. 68, 2002
`
`GLYCEROL PRODUCTION IN SACCHAROMYCES CEREVISIAE
`
`2815
`
`TABLE 1. Some representative yeast processes used for glycerol production
`
`Strain and process
`
`Glycerol concn in
`broth (g · l⫺1)
`
`Yield of glycerol on
`glucose (g · g⫺1)
`
`Avg productivity
`(g · liter⫺1 · day⫺1)
`
`Reference
`or source
`
`S. cerevisiae
`Sulfite batch
`Sulfite; fed batch under vacuum
`GPD1 overproduction; batch
`ADH deletion; shake flask
`pdc2⌬ mutant; shake flask
`tpi1⌬ mutant; shake flask (0–24 h)
`with extra glucose (24–56 h)
`tpi1⌬ mutant; as above (0–44 h)
`tpi1⌬ nde1⌬ nde2⌬ gut2⌬ mutant, aerated batch
`
`Osmotolerant yeasts
`Candida magnoliae I2B; batch
`Saccharomyces strain LORRE Y8; fed batch
`Pichia farinosa; fed batch with solid glucose
`Candida glycerinogenes; batch
`
`a Total polyol concentration.
`
`45
`82
`25
`4.6
`2.9
`36
`43
`63
`219
`
`80a
`260
`300a
`127
`
`0.23
`0.25
`0.12
`0.26
`0.16
`0.46
`0.20
`0.44
`0.50
`
`0.32
`0.47
`0.46
`0.64
`
`9.0
`32.5
`4.3
`2.5
`2.1
`36
`5.3
`35
`57.6
`
`15.6
`29.8
`37.5
`40.6
`
`15
`27
`44
`12
`35
`9
`
`10
`This study
`
`46
`19
`53
`56
`
`(33, 44) (Table 1), whereas overproduction of glycerol phos-
`phatase had no effect on glycerol production (39, 44). Increas-
`ing glycerol export by deregulation of the Fps1p channel pro-
`tein increased glycerol production but, in combination with
`overproduction of G3P dehydrogenase, negatively affected
`growth (44).
`The highest glycerol yield and productivity reported to date
`for metabolically engineered S. cerevisiae were observed with a
`tpi1⌬ deletion mutant (9, 10) (Table 1). Apparently, in tpi1⌬
`mutants, which lack the glycolytic enzyme triose phosphate
`
`isomerase, accumulation of DHAP is prevented by its conver-
`sion to glycerol (Fig. 1). The maximum theoretical yield of this
`process is 1 mol of glycerol · mol of glucose⫺1 if all glucose is
`metabolized via glycolysis. However, S. cerevisiae tpi1⌬ mutants
`are unable to grow on glucose as the sole carbon source (9–11).
`Therefore, biomass was pregrown on glucose-ethanol mix-
`tures, followed by a bioconversion of glucose to glycerol (9,
`10). During the bioconversion phase, glycerol productivity de-
`creased strongly with time (Table 1).
`The biochemical mechanism responsible for the inability of
`
`FIG. 1. Hypothetical pathway (solid lines) for glucose dissimilation by triose phosphate isomerase-negative (tpi1⌬) S. cerevisiae. DHAP is
`detoxified by its reduction to G3P and subsequent dephosphorylation to glycerol. This requires that all NADH generated in the glyceraldehyde-
`3-phosphate dehydrogenase reaction be available for DHAP reduction. Oxidation of either cytosolic G3P or cytosolic NADH by other processes
`(dotted lines) will lead to accumulation of DHAP. Abbreviations: GAP, glyceraldehyde-3-phosphate; FBA, fructose-1,6-biphosphate aldolase;
`GPD, cytosolic G3P dehydrogenase; GPP, glycerol phosphatase; GUT, mitochondrial flavin adenine dinucleotide-dependent G3P dehydrogenase;
`NDE, external mitochondrial NADH dehydrogenase; PDC, pyruvate decarboxylase; ADH, alcohol dehydrogenase.
`
`

`

`2816
`
`OVERKAMP ET AL.
`
`APPL. ENVIRON. MICROBIOL.
`
`Strain
`
`CEN.PK113-7D
`CEN.PK122
`CEN.PK167-2B
`CEN.PK225-2C
`CEN.PK536-5B
`
`CEN.PK530-1B
`CEN.PK530-1C
`CEN.PK546-12B
`
`CEN.PK546-AHG
`
`TABLE 2. S. cerevisiae strains used in this study
`
`Genotypea
`
`Source or reference
`
`MATa
`MATa/MAT␣
`MATa nde1 (41–1659)::loxP-kanMX4-loxP nde2(51–100)::loxP-kanMX4-loxP
`MATa gut2(41–2010)::loxP-kanMX4-loxP
`MAT␣ tpi1(41–707)::loxP-kanMX4-loxP nde1(41–1659)::loxP-kanMX4-loxP nde2
`(51–100)::loxP-kanMX4-loxP
`MAT␣ tpi1(41–707)::loxP-kanMX4-loxP
`MATa tpi1(41–707)::loxP-kanMX4-loxP
`MATa tpi1 (41–707)::loxP-kanMX4-loxP nde1 (41–1659)::loxP-kanMX4-loxP
`nde2(51–100)::loxP-kanMX4-loxP gut2(41–2010)::loxP-kanMX4-loxP
`Spontaneous mutant of CEN.PK546-12B, selected for AHG concn
`
`31
`31
`31
`37
`This study
`
`This study
`This study
`This study
`
`This study
`
`a The numbers in parentheses indicate the deleted nucleotides (ATG ⫽ 1).
`
`S. cerevisiae tpi1⌬ mutants to grow on glucose as the sole
`carbon source is unknown. Since conversion of glucose to
`equimolar amounts of pyruvate and glycerol is neutral in terms
`of ATP (Fig. 1), growth of tpi1⌬ mutants should depend on the
`respiratory dissimilation of pyruvate. Glucose repression of the
`synthesis of key enzymes of respiratory glucose dissimilation
`(16, 50) may therefore prevent growth of tpi1⌬ mutants on
`glucose in batch cultures. However, tpi1⌬ mutants also failed
`to grow in aerobic, glucose-limited chemostat cultures,
`in
`which glucose repression is alleviated (11). As in batch cul-
`tures, growth on glucose in chemostat cultures required the
`inclusion of ethanol in the medium feed (11). This indicated
`that glucose repression of respiration is not the only factor that
`prevents growth of tpi1⌬ mutants on glucose as the sole carbon
`source.
`Complete conversion of DHAP to glycerol in a tpi1⌬ mutant
`requires that all NADH generated in the glyceraldehyde-3-
`phosphate dehydrogenase reaction be used to reduce DHAP
`to glycerol (Fig. 1). If other reactions were to compete for
`NADH with glycerol production, this would lead to accumu-
`lation of DHAP, which in turn can be converted to methylg-
`lyoxal, a cytotoxic compound that can inhibit growth (25, 32).
`We hypothesize that oxidation of cytosolic NADH by the mi-
`tochondrial respiratory chain is such a competing mechanism
`and thus contributes to the inability of tpi1⌬ mutants to grow
`on glucose as the sole carbon source (Fig. 1). Two mechanisms
`are involved in the mitochondrial oxidation of cytosolic NADH
`in S. cerevisiae: mitochondrial external NADH dehydrogenase
`and a G3P shuttle (29, 31, 37, 48). At low specific growth rates,
`either of these two systems is sufficient to sustain respiratory
`growth (37). The NDE1 and NDE2 genes encode two isoen-
`zymes of the external NADH dehydrogenase (31, 48), whereas
`the GUT2 gene encodes a mitochondrial respiratory-chain-
`linked G3P dehydrogenase, a key enzyme of the G3P shuttle
`(29, 45). The latter enzyme might also directly affect growth of
`tpi1⌬ mutants by reoxidizing G3P to DHAP (Fig. 1).
`The aim of this study is to test the hypothesis that mitochon-
`drial reoxidation of cytosolic NADH and/or G3P plays a role in
`the phenotype of tpi1⌬ mutants and to investigate whether
`metabolic engineering of mitochondrial respiration can be
`used to improve glycerol production by S. cerevisiae.
`
`MATERIALS AND METHODS
`
`Yeast strains and maintenance. The S. cerevisiae strains used and constructed
`in this study (Table 2) are prototrophic strains belonging to the CEN.PK family
`(49). Stock cultures were grown at 30°C in shake flasks on YPED medium (10 g
`
`of Bacto yeast extract · liter⫺1, 20 g of peptone · liter⫺1, 10 ml of ethanol · liter⫺1,
`and 0.5 g of glucose · liter⫺1) except the adaptation to high glucose concentration
`(AHG) strain, which was grown on 100 g of glucose · liter⫺1 in MMU medium
`(3.0 g of K2SO4 · liter⫺1, 3.0 g of KH2PO4 · liter⫺1, 3.0 g of urea · liter⫺1, 0.5 g of
`MgSO4 · 7H2O · per liter) with trace elements and vitamins prepared and ster-
`ilized as described previously (52). Urea was added to the medium after separate
`filter sterilization. When stationary phase was reached, 30% (vol/vol) sterile
`glycerol was added, and 2-ml aliquots were stored in sterile vials at ⫺80°C.
`Construction of null mutants. Standard techniques and media for genetic
`modification of S. cerevisiae were used (3). Deletions in TPI1 were obtained by
`the short flanking homology method (54), using pUG6 as a template (21). PCR
`amplification, yeast transformation, and verification of the correct gene deletion,
`as well as determination of the mating type, were carried out as described by
`Luttik et al. (31). The loxP-kanMX4-loxP cassette amplified by PCR was used to
`transform diploid strain CEN.PK122. After tetrad analysis, the G418R segregants
`were checked by diagnostic PCR for the correct integration of the kanMX
`cassette (Table 3).
`All further double, triple, or quadruple deletion strains were constructed by
`crossing of the corresponding single, double, or triple deletion strains. The
`resulting diploid strains were subsequently analyzed by tetrad analysis to obtain
`the respective segregants and further analyzed by diagnostic PCR to confirm the
`correct deletion of the corresponding genes. Strain CEN.PK536-5B (nde1⌬
`nde2⌬ tpi1⌬) resulted from crossing of strains CEN.PK167-2B (nde1⌬ nde2⌬)
`and CEN.PK530-1B (tpi1⌬). To obtain the quadruple deletion strain
`CEN.PK546-12B (nde1⌬ nde2⌬ tpi1⌬ gut2⌬), the strains CEN.PK536-5B (nde1⌬
`nde2⌬ tpi1⌬) and CEN.PK225-2C (gut2⌬), respectively, were crossed.
`Shake flask cultivation. Shake flask cultures were grown in an orbital incuba-
`tor (200 rpm, 30°C) in spherical flat-bottom flasks (500 ml) containing 100 ml of
`medium. Precultures were grown on MMU medium supplemented with 5 ml of
`ethanol · liter⫺1 and 1 g of glucose · liter⫺1. During the exponential growth
`phase, the optical density at 660 nm (OD660) was measured with an Amersham
`Pharmacia Novaspec II spectrophotometer. Exponential-phase cultures were
`used as the inoculum for further shake flask experiments. Prior to inoculation,
`cells were centrifuged (4,500 ⫻ g; 3 min) and washed aseptically with sterile
`
`TABLE 3. Oligonucleotides used for construction of the TPI1
`deletion cassette (S1/S2) and for diagnostic PCR (A1/K1) and (K2/
`A4) of the deletion strains.a
`Oligonucleotide
`DNA sequence
`TPII-S1................5⬘-ATG GCT AGA ACT TTC TTT GTC GGT
`GGT AAC TTT AAA TTA ACA GCT GAA
`GCT TCG TAC GC-3⬘
`TPII-S2................5⬘-TTA GTT TCT AGA GTT GAT GAT ATC
`AAC AAA TTC TGG CTT CGC ATA GGC
`CAC TAG TGG ATC TG-3⬘
`TPII-A1...............5⬘-CTT CTG CGG TAT CAC CCT AC-3⬘
`TPII-A4...............5⬘-CAA TGC AGT CTT CGG TAC AC-3⬘
`K1........................5⬘-GGA TGT ATG GGC TAA ATG TAC G-3⬘
`K2........................5⬘-GTT TCA TTT GAT GCT CGA TGA G-3⬘
`
`a Oligonucleotides used for the construction of the deletion cassettes and for
`diagnostic PCR of the strains deleted in NDE1, NDE2, and GUT2, respectively,
`were described previously (31, 37). Sequences complementary to the loxP-
`KanMX4-loxP cassette (pUG6) are underlined.
`
`

`

`VOL. 68, 2002
`
`GLYCEROL PRODUCTION IN SACCHAROMYCES CEREVISIAE
`
`2817
`
`FIG. 2. Growth on glucose of tpi1⌬ S. cerevisiae (E) and an isogenic
`tpi1⌬ nde1⌬ nde2⌬ gut2⌬ strain (F). Both strains were grown in shake
`flasks on MMU medium with 5 g of glucose · liter⫺1 as the sole carbon
`source. Three independent replicate cultures gave the same results.
`
`medium to remove residual substrates and metabolites. The washed cells were
`used to inoculate (initial OD660 of 0.1) fresh shake flasks containing MMU
`medium with 5 ml of ethanol · liter⫺1 and 1 g of glucose · liter⫺1. During the
`exponential growth phase, cells were harvested as described above, washed, and
`resuspended in the medium used for the final shake flask cultivation, which was
`MMU medium with glucose as the sole carbon source at a concentration of 1, 5,
`10, 50, or 100 g · liter⫺1. Growth was monitored by regular OD660 measurements.
`When OD660 was above 0.3, samples were diluted in MMU medium containing
`the same glucose concentration as the culture, to avoid changes in OD660 due to
`osmotic effects.
`Chemostat cultivation. Chemostat cultures were grown in 1-liter working-
`volume laboratory fermentors as described previously (31). This reference also
`describes procedures for gas analysis and determination of biomass dry weight.
`One-hundred-milliliter precultures in shake flasks, inoculated with 2 ml of a
`frozen stock culture, were grown to stationary phase on YPED medium. These
`precultures were used to inoculate a fermentor containing 1 liter of MMA
`medium [5.0 g of (NH4)2SO4 · liter⫺1, 3.0 g of KH2PO4 · liter⫺1, 0.5 g of
`MgSO4 · 7H2O per liter, vitamins, and trace elements as in the MMU medium
`described above] supplemented with 1 g of glucose · liter⫺1 and 10 ml of
`ethanol · liter⫺1. The continuous medium supply was initiated when a steep
`increase of the dissolved-oxygen concentration indicated depletion of the carbon
`source. The synthetic medium used for continuous cultivation was MMA me-
`dium supplemented with 7.5 g of glucose · liter⫺1 (60% of total carbon) and 83
`mM ethanol (40% of total carbon). The dilution rate was set at 0.05 h⫺1. For
`growth on glucose as a sole carbon source, the feed of the chemostat culture
`growing on the glucose-ethanol mixture was switched to the same medium
`without ethanol.
`Batch cultivation in fermentors. One milliliter of frozen stock culture of the
`glucose-adapted CEN.PK546-12B strain (CEN.PK546-AHG, where AHG is a
`designation for adapted to high glucose; see Results) was inoculated in shake
`flasks with double-strength MMU medium supplemented with 400 g of
`glucose · liter⫺1. After 4 days of incubation, 150 ml of this culture was transferred
`to a 2-liter laboratory fermentor (Applikon, Schiedam, The Netherlands). The
`fermentor was previously autoclaved (20 min at 110°C) while containing 1.2 liters
`of demineralized water with 600 g of glucose and 100 ␮l of silicone antifoam
`(BDH). Together with the inoculum, 45 ml of a solution containing 9 g of urea
`and 100 ml of 30-fold concentrated MM medium (MMU without urea) was
`added. This yielded an initial culture volume of 1.5 liters consisting of double-
`strength MMU medium and an initial glucose concentration of 400 g·liter⫺1. The
`
`FIG. 3. Effect of glucose concentration on the specific growth rate
`of isogenic S. cerevisiae tpi1⌬ (䊐) and tpi1⌬ nde1⌬ nde2⌬ gut2⌬ (E)
`strains and a tpi1⌬ nde1⌬ nde2⌬ gut2⌬-AHG (adapted to high glucose)
`strain (F) in shake flask cultures. All cultures were grown on MMU
`medium supplemented with different initial glucose concentrations.
`Data are presented as the average ⫾ mean deviation of at least two
`independent replicate cultures for each glucose concentration.
`
`pH was kept at 5.0 by an Applikon ADI 1030 biocontroller, via the automatic
`addition of 4 M KOH or 4 M H2SO4. The fermentor was aerated (0.75
`liter · min⫺1) and stirred at 900 rpm. Additional antifoam was added manually
`when foaming occurred. Biomass growth was monitored by OD660 measure-
`ments (when necessary, diluted in a 400-g 䡠 liter⫺1 glucose solution) and via
`biomass dry-weight measurements (31). Metabolite concentrations were deter-
`mined by high-performance liquid chromatography (30). High-biomass-density
`cultures were performed in the same way, but were inoculated with 200 ml of a
`stationary-phase fermentor culture grown as described above.
`
`RESULTS
`
`Growth on glucose in shake flask cultures. Previously pub-
`lished reports state that tpi1⌬ mutants of S. cerevisiae are
`unable to grow on glucose as the sole carbon source, both in
`shake flasks (9, 38) and in chemostat cultures (11). This was
`confirmed in shake flask cultures
`containing 5 g of
`glucose · liter⫺1, in which the specific growth rate was below
`0.01 h⫺1 (Fig. 2). To investigate whether this inability to grow
`on glucose was due to mitochondrial reoxidation of cytosolic
`NADH and/or G3P (Fig. 1), thus preventing the complete
`reduction of dihydroxyacetone phosphate, a tpi1⌬ nde1⌬
`nde2⌬ gut2⌬ mutant was constructed. Indeed, this quadruple
`deletion mutant was able to grow on glucose with a specific
`growth rate of 0.045 h⫺1 (Fig. 2). Yet, this specific growth rate
`is still substantially lower than that of the isogenic wild-type
`strain CEN.PK113-7D, which under these conditions exhibits a
`ca. eightfold-higher specific growth rate (data not shown).
`Specific growth rates of the tpi1⌬ nde1⌬ nde2⌬ gut2⌬ strain
`on glucose strongly depended on the glucose concentration in
`the medium (Fig. 3). At glucose concentrations of 50 g · liter⫺1
`and higher, the specific growth rate of the quadruple deletion
`
`

`

`2818
`
`OVERKAMP ET AL.
`
`APPL. ENVIRON. MICROBIOL.
`
`TABLE 4. Growth of the reference strain S. cerevisiae CEN.PK113-7D and the isogenic tpi1⌬ nde1⌬ nde2⌬ gut2⌬ mutant CEN.PK546-12B.a
`
`Strain (relevant genotype)
`
`qO2
`
`qCO2
`
`Glycerol yield
`Biomass yield
`(mol·mol glucose⫺1)
`(g·g glucose⫺1)
`1.4 ⫾ 0.1
`1.3 ⫾ 0.1
`⬍0.01
`0.47 ⫾ 0.01
`CEN PK113-7D (TPI1 NDE1 NDE2 GUT2)
`2.7 ⫾ 0.3
`2.2 ⫾ 0.2
`0.83 ⫾ 0.01
`0.20 ⫾ 0.02
`CEN.PK546-12B (tpi1⌬ nde1⌬ nde2⌬ gut2⌬)
`a Strains were grown in aerobic, glucose-limited chemostat cultures (30°C, pH 5.0, dilution rate ⫽ 0.05 h⫺1, 7.5 g of glucose·liter⫺1 in feed). Data are presented as
`the average ⫾ mean deviation of two independent chemostat experiments for each strain. Values for qO2 and qCO2 are in millimoles per gram per hour.
`
`Carbon recovery
`(%)
`
`95
`98
`
`strain was very low and similar to that of the tpi1⌬ single
`mutant. Conversion of glucose into equimolar amounts of glyc-
`erol and pyruvate is neutral in terms of ATP and redox me-
`tabolism. Therefore, growth of the tpi1⌬ mutants on glucose
`critically depends on respiration for ATP production. The syn-
`thesis of many enzymes involved in the respiratory dissimila-
`tion of pyruvate is subject to glucose catabolite repression (16,
`50), which may explain the glucose sensitivity of the tpi1⌬
`nde1⌬ nde2⌬ gut2⌬ strain.
`Glucose-limited chemostat cultivation. Analysis of fermen-
`tation products in glucose-grown shake flask cultures sug-
`gested that the tpi1⌬ nde1⌬ nde2⌬ gut2⌬ strain produced large
`amounts of glycerol. For a quantitative analysis of biomass and
`product yields, the tpi1⌬ nde1⌬ nde2⌬ gut2⌬ strain and the
`reference strain CEN.PK113-7D were grown in aerobic, glu-
`cose-limited chemostat cultures. In such cultures, glucose ca-
`tabolite repression can be alleviated by the low-residual-glu-
`cose concentrations. The biomass yield of the tpi1⌬ nde1⌬
`nde2⌬ gut2⌬ strain was more than twofold lower than the
`reference strain during growth at a dilution rate of 0.05 h⫺1
`(Table 4). This was partly due to the conversion of glucose into
`glycerol. The molar yield of glycerol on glucose was 0.83
`mol · mol⫺1. Under the same cultivation conditions, the iso-
`genic reference strain CEN.PK113-7D did not produce detect-
`able amounts of glycerol. The extensive production of glycerol
`is consistent with the metabolic scheme proposed in Fig. 1.
`This scheme predicts a glycerol yield of 1 mol · mol⫺1. The
`lower glycerol yield in the glucose-limited chemostat cultures
`can be explained from the assimilation of glucose-6-phosphate
`via metabolic pathways other than glycolysis, such as the pen-
`tose phosphate pathway, cell wall biosynthesis, and storage
`carbohydrate biosynthesis.
`Selection of a strain adapted to high-glucose concentra-
`tions. The glycerol yield (0.83 mol · mol glucose⫺1, i.e., 0.42
`g · g of glucose⫺1) in glucose-limited chemostat cultures of the
`tpi1⌬ nde1⌬ nde2⌬ gut2⌬ strain is among the highest reported
`for a growing, metabolically engineered S. cerevisiae strain,
`although it is matched by a bioconversion process with the
`nongrowing tpi1⌬ strain (Table 1). However, the glucose sen-
`sitivity of the quadruple mutant is a drawback for any large-
`scale application in glycerol production. The tpi1⌬ nde1⌬
`nde2⌬ gut2⌬ strain (CEN.PK 546-12B) was adapted to growth
`at high-glucose concentrations by serial transfer. This was
`started by transferring 10 ml of a steady-state glucose-limited
`chemostat culture to a shake flask containing MMA medium
`with 100 g of glucose · liter⫺1 (555 mM). This culture showed
`extremely slow growth accompanied by acetate production and
`a decrease of the culture pH to ca. 3. After 2 days, growth and
`acetate production ceased. After another 4 days, acetate con-
`sumption occurred and growth resumed. Eventually the low
`
`pH (ca. 2.5) caused by ammonia consumption abolished
`growth altogether, despite the presence of approximately 350
`mM residual glucose. From this culture, 1 ml was transferred
`to a new shake flask with the same medium. No acetate pro-
`duction was observed, but acidification again prevented com-
`plete consumption of glucose. Use of urea instead of ammo-
`nium salts as the sole nitrogen source can sometimes prevent
`acidification of yeast cultures (24). Therefore, 1 ml of culture
`was transferred to a shake flask with MMU medium containing
`100 g of glucose · liter⫺1. This indeed led to complete con-
`sumption of glucose. A total of 0.5 ml of culture was trans-
`ferred to the next flask with the same medium. After two
`further transfers, the culture was streaked onto agar plates
`with MMU medium containing 50 g of glucose · liter⫺1. The
`culture was purified by three subsequent transfers of a single
`colony to a new plate. Finally, a single colony was transferred
`to a shake flask with MMU medium containing 100 g of
`glucose · liter⫺1. This glucose-adapted culture of the tpi1⌬
`nde1⌬ nde2⌬ gut2⌬ strain was called CEN.PK546-AHG. Es-
`pecially at high-glucose concentrations, the selected mutant
`exhibited much higher specific growth rates than the original,
`nonadapted tpi1⌬ nde1⌬ nde2⌬ gut2⌬ strain (Fig. 3).
`Production of glycerol by engineered S. cerevisiae. Glycerol
`production by S. cerevisiae CEN.PK546-AHG was investigated
`in aerated batch cultures with 400 g of glucose · liter⫺1, which
`were inoculated from glucose-grown shake flasks at 0.3 g (dry
`weight) · liter⫺1. Despite the high initial glucose concentration,
`these batch cultures exhibited exponential growth (␮ ⫽ 0.027
`h⫺1). The slow growth was accompanied by the accumulation
`of glycerol to a final concentration of over 200 g · liter⫺1 (data
`not shown). Glycerol was the major fermentation product.
`Concentrations of acetate and ethanol remained below 1.5 and
`4 g · liter⫺1, respectively, and were completely consumed to-
`wards the end of fermentation (data not shown). Throughout
`the fermentation process, the glycerol yield on glucose was 1.0
`mol · mol⫺1. Due to the low initial biomass concentration and
`the long start-up period, the overall glycerol production rate
`was 1.0 g glycerol · liter⫺1·h⫺1. When the experiment was re-
`peated with an initial biomass dry weight of 3.9 g · liter⫺1 (Fig.
`4A), all glucose was consumed in 80 h and the final concen-
`tration of glycerol was 219 g · liter⫺1. The molar ratio of glyc-
`erol produced per glucose consumed was 0.99 mol · mol⫺1
`(Fig. 4B), and the overall glycerol production rate was 2.4 g
`glycerol · liter⫺1 · h⫺1. Glycerol consumption did not occur,
`even after glucose depletion. Most wild-type S. cerevisiae
`strains can consume glycerol, although it is a very poor carbon
`and energy source (43, 49). However, in the quadruple mutant,
`glycerol dissimilation is not possible as the deletion of the
`GUT2 gene and the absence of triose phosphate isomerase
`block the oxidation pathway.
`
`

`

`VOL. 68, 2002
`
`GLYCEROL PRODUCTION IN SACCHAROMYCES CEREVISIAE
`
`2819
`
`served for other glycolytic mutants of S. cerevisiae and K. lactis
`(17, 22). This has been attributed to the ability of K. lactis, but
`not S. cerevisiae, to couple the oxidation of cytosolic NADPH
`to the mitochondrial respiratory chain (20). This coupling en-
`ables glycolytic null mutants of K. lactis to use the pentose
`phosphate pathway, with its NADP⫹-linked dehydrogenases,
`as an alternative dissimilatory pathway (20, 26). In tpi1⌬ mu-
`tants, this partial bypass of fructose-1,6-bisphosphate aldolase
`(6) generates sufficient NADH to prevent the accumulation of
`DHAP.
`Even when glucose dissimilation cannot be rerouted via the
`pentose phosphate pathway, the metabolic network of S. cer-
`evisiae should, at least in theory, be sufficiently flexible to allow
`for growth on glucose in the absence of an active triose phos-
`phate isomerase. However, this demands that all NADH pro-
`duced in the glyceraldehyde-3-phosphate dehydrogenase reac-
`tion be used for reduction of DHAP to glycerol (Fig. 1). The
`inability of the tpi1⌬ mutant to grow on glucose as a sole
`carbon source and the partial restoration of growth in a qua-
`druple tpi1⌬ nde1⌬ nde2⌬ gut2⌬ strain (Fig. 2), indicates that
`mitochondrial oxidation of NADH and/or G3P competes for
`the NADH formed in glycolysis and thus contributes to the
`phenotype of tpi1⌬ mutants. This also provides an explanation
`for the restoration of growth by cofeeding ethanol or ethanol-
`formate mixtures to aerobic, glucose-limited chemostat cul-
`tures of tpi1⌬ mutants (11). In these cultures, the cytosolic,
`NAD⫹-dependent alcohol and formate dehydrogenases can
`provide sufficient NADH for complete reduction of DHAP,
`even in the presence of an active mitochondrial respiratory
`chain. Interestingly, it has recently been proposed that the
`nonviability of a tpi⌬ mutant of the bloodstream form of the
`parasite Trypanosoma brucei is due to a similar redox problem,
`caused in this case by the activity of a mitochondrial G3P
`oxidase (23).
`Özcan et al. (38) and Schulte et al. (47) isolated and char-
`acterized suppressor mutants of a S. cerevisiae tpi1⌬ mutant
`which had regained the ability to grow on glucose. Some of the
`suppressor mutants exhibited a strongly reduced activity of
`phosphoglycerate kinase. Reduced activity of phosphoglycer-
`ate kinase will cause a reduced flux through glyceraldehyde-3-
`phosphate dehydrogenase, which in turn affects the cytosolic
`NADH/NAD⫹ ratio. As a result of the different affinities of
`isolated mitochondria and NAD⫹-dependent G3P dehydroge-
`nases for NADH (2, 40), a changed cytosolic NADH/NAD⫹
`ratio may suppress the interference of mitochondrial respira-
`tion with redox metabolism in tpi1⌬ mutants (Fig. 1). Another
`class of suppressor mutants exhibited reduced glucose uptake
`and reduced glucose repression of respiratory enzymes (38,
`47). This appears to be more difficult to reconcile with our
`hypothesis. However, isolation of the suppressor mutants was
`done with glucose-containing complex medium (38). With such
`a medium, it is difficult to assess whether growth is due solely
`to glucose consumption or to consumption of other carbon
`sources present in the complex medium (14). To further study
`the phenotype of tpi1⌬ mutants, it would be of interest to
`isolate and characterize tpi1⌬ suppressor mutants that grow on
`glucose-containing synthetic medium.
`Although the additional deletion of the NDE1, NDE2, and
`GUT2 genes restored growth of tpi1⌬ S. cerevisiae (Fig. 2), the
`specific growth rate of the resulting quadruple mutant re-
`
`FIG. 4. Glycerol production in an aerated fermentor culture with
`an initial glucose concentration of 400 g · liter⫺1 by a spontaneous
`mutant of the tpi1⌬ nde1⌬ nde2⌬ gut2⌬ S. cerevisiae strain, isolated by
`serial transfer in high-glucose medium (AHG strain). An independent
`duplicate experiment gave identical results. (A) Concentrations of
`biomass (䊐), glucose (E), and glycerol (F). (B) The molar ratio be-
`tween glucose consumed and glycerol produced during the fermenta-
`tion.
`
`DISCUSSION
`Phenotype of tpi1⌬ mutants. In many organisms, null muta-
`tions in the structural gene encoding triose phosphate isomer-
`ase or mutations that lead to severely reduced activities of this
`enzyme result in growth deficiencies (7, 8, 23, 41). The com-
`plete inability of tpi⌬ mutants to grow on glucose as the sole
`carbon source, as found with S. cerevisiae (9, 11) (Fig. 2), is not
`representative for all yeasts. A Klu