Fed-Batch Xylitol Production with Two
Recombinant Saccharomyces cerevisiae
Strains Expressing XYL1 at Different
Levels, Using Glucose as a Cosubstrate:
A Comparison of Production Parameters
and Strain Stability
Nina Q. Meinander, Bärbel Hahn-Hägerdal
Applied Microbiology, Lund Institute of Technology/University of Lund,
P.O. Box 124, S-22 100 Lund, Sweden; telephone: 46 46 2228428; fax: 46 46
2224203; e-mail: Barbel.Hahn-Hagerdal@tmb.lth.se
Received 15 April 1996; accepted 24 October 1996
Abstract: Xylitol production with two recombinant Sac-
charomyces cerevisiae strains expressing the XYL1 gene,
coding for xylose reductase (XR), at different levels, the
‘low XR strain’ at 0.51 U/mg and the ‘high XR strain’ at
10.8 U/mg, was compared in batch and fed-batch culture.
Xylose was not consumed in the presence of high glucose
concentrations, because both sugars are transported by
the glucose transport system, which has a higher affinity
for glucose than for xylose. When glucose was fed gradu-
ally to the culture, high concentrations were avoided, and
xylose was converted to xylitol with a specific productiv-
ity of 0.10 g g21 h21 attained with the low XR strain and
0.19 g g21 h21 with the high XR strain, indicating that
factors other than the XR-activity control the rate of xy-
lose conversion.
The overproduction of XR put a substantial protein bur-
den on the high XR strain, contributing to a 50% decrease
in specific growth rate and reduced biomass yield com-
pared with the low XR strain. Despite the use of selective
medium, the stability of the high XR strain was poor in
long fed-batch and chemostat cultures, whereas the low
XR strain was stable. The high XR strain lost its XR activity
almost completely in some fed-batch cultures and in
chemostat culture. In chemostat cultivation, part of the
population lost the plasmid harboring the XR gene. This
was due to the fact that leucine was released into the
broth from plasmid containing cells, which enabled some
cells to grow without the plasmid containing the LEU2
auxotrophic complementation selection marker. Further-
more, isolation and analysis of plasmids from a popula-
tion that had lost its XR activity, showed that in addition
to the original plasmid, a rearranged form of the plasmid,
retaining the selection marker but not the expression of
active XR, was present. However, these observations
Correspondence to: B. Hahn-Hägerdal
Contract grant sponsor: The Nordic Industrial Fund
Contract grant sponsor: The National Swedish Board for Techni-
cal Development
Contract grant sponsor: The Knut and Alice Wallenberg Foun-
dation
Contract grant sponsor: The Swedish National Science Research
Council
 1997 John Wiley & Sons, Inc.
could only partly explain the decrease in XR activity. 
1997 John Wiley & Sons, Inc. Biotechnol Bioeng 54: 391–399,
1997.
Keywords: xylitol production; recombinant Saccharo-
myces cerevisiae; protein burden; plasmid stability
INTRODUCTION
Biotechnical xylitol production can be achieved using
xylose metabolizing yeasts such as Candida sp. (Chen
and Gong, 1985; Gong et al., 1981; Horitsu et al., 1992;
Lu et al., 1995; Meyrial et al., 1991), Pachysolen tan-
nophilus (Thonart et al., 1987; Watson et al., 1984) or
Debaromyces hansenii (Roseiro et al., 1991) that excrete
xylitol as a metabolic byproduct. Genetic engineering
techniques have enabled the construction of a recombi-
nant Saccharomyces cerevisiae strain that converts xy-
lose to xylitol with stoichiometric 1:1 yield (Hallborn et
al., 1991). The recombinant S. cerevisiae expresses the
xylose reductase (XR) encoding XYL1 gene from Pichia
stipitis. The XR reduces xylose to xylitol using NADPH
or NADH as cofactor (Rizzi et al., 1988; Verduyn et
al., 1985).
Xylitol cannot be metabolized by the recombinant S.
cerevisiae; therefore, xylose conversion requires
the simultaneous metabolism of a cosubstrate such
as glucose, ethanol or acetate, for regeneration
of NAD(P)H and generation of maintenance energy
(Hallborn et al., 1994). The genetically engineered S.
cerevisiae strain has been physiologically characterized
in semi-continuous and chemostat cultivation using glu-
cose as cosubstrate (Thestrup and Hahn-Hägerdal,
1995; Meinander et al., 1996). The key fermentation
variables and the xylitol production efficiency in fed-
batch cultivation using ethanol as cosubstrate have been
investigated (Meinander et al., 1994).
Recently, a novel plasmid construct, which increased
the XR activity of recombinant XYL1 expressing S.
CCC 0006-3592/97/040391-09
cerevisiae 20-fold, was described (Walfridsson et al., to
appear). The xylitol production of the previously men-
tioned strain with low XR activity and the strain with
20-fold higher activity, immobilized in Ca-alginate has
been compared in a continuous packed bed reactor
(Roca et al., 1996). Under these conditions the xylitol
productivity of the high XR strain was not markedly
higher than that of the low XR strain. However, the
specific productivity of the low XR strain was much
lower in immobilized conditions compared with
previous results using free cells. During the contin-
uous cultivations with immobilized cells, a marked
instability in the XR activity of the high XR strain
was observed.
In the present study, the xylitol production of the low
and high XR strains in fed-batch fermentation with free
cells using glucose as cosubstrate was compared, and
the growth kinetics of the two strains in batch cultivation
were determined. The key fermentation variables and
xylitol production efficiency in fed-batch fermentation
with glucose as cosubstrate were compared with pre-
viously described fed-batch fermentation with ethanol
as cosubstrate (Meinander et al., 1994). The stability of
XR activity in the high XR strain was determined both in
fed-batch and chemostat cultivation, and investigations
into some possible reasons for instability were con-
ducted.
MATERIALS AND METHODS
Microorganisms
The recombinant XYL1 expressing Saccharomyces cere-
visiae strains H475 and S641 are based on the host strain
H158: GPY55-15Ba leu2-3, leu2-112, ura3-52, trp1-289,
his4-519, prb1, cir1 obtained from Gregg Payne, Berke-
ley University, California. Strain H475 contains the plas-
mid pUA103 (Hallborn et al., 1991), and strain S641
contains the plasmid pM2 (Walfridsson et al., to ap-
pear), which was previously called pMW410 (Walfrids-
son et al., 1992; Roca et al., 1996). In the pUA103 and
pM2 plasmids the expression of the XYL1 gene is con-
trolled by the constitutive phosphoglycerate kinase pro-
moter (Mellor et al., 1983), and selection is achieved
through the LEU2 selection marker. The plasmids are
identical, except for an empty ADH promoter-termina-
tor present on the plasmid pM2 (Walfridsson et al., to
appear). The XR activities of H475 and S641 have been
reported to be 0.51 U/mg and 10.8 U/mg, respectively
(Hallborn et al., 1991; Walfridsson et al, to appear). In
the following, H475 will be denoted ‘‘low XR strain’’
and S641 ‘‘high XR strain.’’
Escherichia coli DH5a [F2 F80dlacZ DM15 D(lac-
ZYA-argF ) U169 deoR recA1 endA1 hsdR17(rK2 mK1)
sup E44 x2 thi1 gyrA69 relA1] (GIBCO BRL, Gaithers-
392 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 54, NO. 4, MAY 20, 1997
burg, Maryland) was used for subcloning plasmids iso-
lated from the high XR strain.
Medium
Yeast cultivations were performed in a mineral medium
(Verduyn et al., 1992), supplemented with 0.05 g L21
uracil, histidine, and tryptophane. In anaerobic batch
and continuous fermentations, 0.01 g L21 ergosterol
and 0.42 g L21 polyoxyethylenesorbitan mono-oleate
(Tween 80, Sigma) were added. Inocula for fermentor
cultivations were grown in mineral medium containing
20 g L21 glucose. In the fermentor cultivations, sugar
concentrations as specified in the Fermentations section
were used.
Fermentations
Fermentations were conducted at 308C with pH main-
tained at 5.5 by addition of 100 g L21 KOH. A new
inoculum plate was prepared for each experiment by
transferring a small aliquot of frozen cell suspension to
an SC-leu plate (Sherman et al., 1983) and incubating
at 308C for 2–3 days. The inoculum was grown in two
stages, first 50 mL medium in a 250 mL baffled flask
was inoculated from the plate and incubated in a Gallen-
kamp INR-200 (Leicester, UK) orbital incubator, 140
rev min21 at 308C for 24 h. This culture was then trans-
ferred to 200–500 mL medium in a 1 L baffled flask
and incubated under the above described conditions for
24 h. The inoculum was centrifuged in a Beckman J2-
21 (Geneva, Switzerland) centrifuge at 4440 g for 15
min, washed once with 0.1M phosphate buffer pH 5.5
and resuspended in buffer.
Batch and continuous fermentations were performed
in a New Brunswick Bioflo III fermentor (New Bruns-
wick Scientific Co., Edison, New Jersey) under anaero-
bic conditions (N2 flushing). The stirring speed was 200
rev min21. The batch cultivations were performed in a
total volume of 1.5 L, and the initial sugar content was
20 g L21 glucose and 10 g L21 xylose. The continuous
chemostat cultivations were conducted as previously de-
scribed (Meinander et al., 1996).
Fed-batch fermentations were performed in the Bio-
flo III fermentor or in a Chemoferm LMS-500 (Belach
Bioteknik, Stockholm, Sweden) fermentor. The cultures
were started as batch cultivations with 20–40 g L21 glu-
cose as carbon source and grown overnight with vigor-
ous aeration. After the batch growth phase, xylose was
added in concentrations between 110 and 160 g L21 as
specified in the results section, the aeration conditions
were set to oxygen limitation (airflow of 600 mL min21)
or anaerobic (N2 flushing), and a glucose feed (0.9–2 g
h21) was started.
Determinations of mmax for strains transformed with
plasmids derived from chemostat cultivations were done
in magnetically stirred stoppered flasks with intermit-
tent adjustment of pH to 5.5.
Analyses
The concentrations of glucose, xylose, xylitol, glycerol,
acetate and ethanol in the fermentation broth were de-
termined by column liquid chromatography (CLC). A
Shimadzu CLC system (Kyoto, Japan) equipped with a
Biorad HPX87-H cation exchange column (Richmond,
CA) for separation of the analytes, and a Shimadzu
RID6A refractive index detector was used. The mobile
phase was 5 mM H2SO4 pumped at 0.6 mL min21 and
the separation temperature 458C.
The biomass concentration was analyzed by dry
weight determination. A known volume of fermentation
broth was filtered through a 0.45 mm preweighed mem-
brane, which was dried in a microwave oven and
weighed. The biomass concentration was also monitored
by optical density (OD) measurements at 620 nm. How-
ever, when antifoam was added into the fermentor these
measurements were disturbed, so that only approximate
results were obtained. The amino-acids in the fermenta-
tion broth were analyzed by separation on an ion-
exchange CLC column and detection by fluorescence
after post-column derivatization with OPA (Barkholt
and Jensen, 1989).
Specific XR activity was measured from cell lysates.
The cells were centrifuged in the same manner as the
inoculum, washed with 0.1 mM phosphate buffer, pH
7, resuspended in phosphate buffer containing 0.5 mM
EDTA, 0.5 mM dithiothreitol and 1 mM phenylmethyl-
sulphonylfluoride (PMSF) and disintegrated by vor-
texing with glass beads (0.5 mm diameter). The lysate
was centrifuged and the XR activity measured from the
supernatant as previously described (Hallborn et al.,
1991). The protein concentration was determined using
a commercial Coomassie Protein Assay Reagent
(Pierce, Rockford, Illinois) (Bradford, 1976). The activi-
ties are expressed as units (U) per mg protein, one unit
being equivalent to the amount of enzyme needed to
reduce one mmol xylose per minute.
The fraction of plasmid containing cells in the fermen-
tation broth at the end of the fermentations, was esti-
mated by plating diluted samples of the broth on unse-
lective SC1 and selective SC-leu plates (Sherman et al.,
1983). The cells retaining the recombinant plasmid with
the selection marker (the LEU2 gene) can grow on both
types of plates, while cells that have lost the plasmid and
selection marker cannot grow on the selective plates.
Nucleic Acid Manipulations and Transformations
Plasmids were prepared from yeast cells according to
the quick method described by Strathern and Higgins
(1991). Standard techniques for electro-transformation
of E. coli and preparation of plasmids from E. coli were
MEINANDER AND HAHN-HÄGERDAL: XYLITOL PRODUCTION BY RECOMBINANT S. CEREVISIAE
used (Sambrook et al., 1989). Yeast transformation was
performed by the LiAc method (Schiestl and Gietz,
1989). Restriction enzymes were purchased from Boeh-
ringer Mannheim Scandinavia AB, Bromma, Sweden.
RESULTS
Batch Cultivations
To investigate the growth kinetics and product pattern
for the strains with different expression levels of XR,
anaerobic batch fermentations with 20 g L21 glucose
and 10 g L21 xylose were performed (Fig. 1). The mmax
estimated from optical density measurements was 0.19
h21 for the low XR strain and 0.098 h21 for the high XR
strain. The final OD was 8.0 for the low XR strain and
5.1 for the high XR strain. Neither strain consumed
xylose or produced xylitol until the glucose concentra-
tion had decreased to about 10 g L21. When the glucose
was exhausted, the consumption of xylose stopped due
to the lack of a reducing power generating cosubstrate.
The ethanol and acetate present in the fermentation
broth could not be metabolized as cosubstrates under
anaerobic conditions. The decrease in ethanol concen-
tration apparent at the end of the cultivation was due
to evaporation or stripping of ethanol from the broth
by the nitrogen gas flowing through the fermentor. The
strains showed similar product patterns, where glycerol
was produced concommittantly with growth, and small
amounts of acetate were formed at the end of the growth
phase. Similar xylitol concentrations were reached with
both strains at the end of the fermentations. The only
apparent differences between the strains were the
slower growth and the lower biomass yield of the high
XR strain.
Fed-Batch Xylitol Production
In the batch fermentations it was evident that glucose
and xylose were not simultaneously consumed, when
the glucose concentration in the fermentor was high. In
order to facilitate simultaneous xylose conversion and
glucose consumption, fed-batch fermentations were
conducted, where the glucose was gradually fed to the
fermentor to avoid high glucose concentrations.
Anaerobic Conditions
One of the advantages of using glucose as cosubstrate
compared with ethanol is that the fermentation can be
conducted anaerobically. Two anaerobic fed-batch fer-
mentations with varying glucose feed rates were carried
out with each strain. Examples of the concentration
profiles of substrates and main products in one fermen-
tation with each strain are shown in Figure 2. Addition-
ally, acetate and glycerol were formed as byproducts of
393
Figure 1. Substrate consumption and product formation in batch cultivation of the low XR strain (A) and high XR strain (B). Glucose (s),
ethanol (d), xylose (n), xylitol (m), glycerol (h), acetate (j) and OD620 (---).

the glucose metabolism (data not shown). The glucose
feed rate, xylitol recovery, yield of xylitol on glucose,
mean volumetric and specific xylitol productivities, and
mean biomass and residual glucose concentrations in
the different fermentations are summarized in Table I.
The values are corrected for the effects of dilution and
sampling on the broth volume. Since xylitol is not me-
tabolized by the recombinant strains, all consumed
xylose should be recovered as xylitol. However, less
accurate estimation of the pumping rate and less concen-
trated feed solutions giving large dilution effects in the
fermentations with higher feed rate, led to poor accuracy
in the xylitol recoveries, compared with the fermenta-
tions with lower feed rates, where feed solutions were
highly concentrated and feed rates were measured by
weight.
In all fermentations the low XR strain reached a
higher initial biomass concentration during the batch
growth phase, compared with the high XR strain. Dur-
ing the fed-batch xylitol production phase, the biomass
concentrations stayed approximately constant, indicat-
ing that only slow growth occurred, balancing out the
dilution effect of the feed. The yield of xylitol on glucose
(around 1 g g21) was similar for both strains. With the
high XR strain a 1.9 times higher specific xylitol produc-
tivity was achieved, compared with the low XR strain.
In spite of the differences in specific productivity, the
volumetric productivity was similar with both strains,
or even slightly higher with the low XR strain, due to
its higher biomass yield. The glucose feed rate had a
direct influence on the residual glucose concentration
in the fermentor. In the fermentation shown in Figure
2B, the feed rate was lowered after elevated residual
glucose concentrations were detected. This immediately
led to a decrease in the glucose concentration. The im-
portance of keeping the residual glucose concentration

Figure 2. Concentration profiles for anaerobic fed-batch xylitol production with the low XR strain (A) and high XR strain (B). The dashed
line indicates the time of changing the glucose feed rate. In addition to consumption and production, the concentrations are subject to dilution
by the glucose feed streams, A: 2.2 mL h21, ,95 h; 1.6 mL h21, .95 h, B: 2.2 mL h21, ,95 h; 1.3 mL h21, .95 h. Glucose (s), ethanol (d)
xylose (n), xylitol (m), and biomass (1).

394 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 54, NO. 4, MAY 20, 1997
Table I. The overall xylitol recovery from consumed xylose (Yxol/xyl), xylitol yield on consumed glucose, mean productivities, mean biomass,
and mean residual glucose concentrations obtained with the low and high XR strains with different glucose feed rates under anaerobic and
oxygen limited conditions.

Feed rate Yxol/xyla Yxol/glu Qxol qxol Cbiomass Cresidual glu

Strain Oxygenation (g h21) (%) (g g21) (g L21h21) (g g21h21) (g L21 d.w.) (g L21)

Low XR anaerobic 0.9 98.6 0.97 0.43 0.075 5.5 1.96

Low XR anaerobic 1 86 0.91 0.43 0.10 4.3 0
2 0.20 0.01 0.004 3.1 5.02
Low XR oxygen limited 0.99 100 0.82 0.57 0.078 7.3 0.12
High XR anaerobic 0.88 97.5 0.94 0.39 0.11 3.5 3.59
High XR anaerobic 1 88 1.20 0.42 0.19 2.4 1.7
2 0.44 0 0 1.3 31.5
High XR oxygen limited 1 105 0.13 0.001 0.0003 4.1 3.1
a
Overall yield calculated for the entire fermentation.

low was exemplified by the fermentations with higher
feed rates. When the feed rate was increased from 1
to 2 g h21, the maximal glucose consumption rate of the
yeast was exceeded, the residual glucose concentration
increased sharply and xylitol production stopped (Ta-
ble I).
Oxygen Limited Conditions
In order to investigate whether limited aeration would
increase growth and thus increase volumetric productiv-
ities, or enable more effective regeneration of NADH
through TCA cycle activity and thus increase the xylitol
yield on glucose, fermentations similar to the above but
sparingly aerated (0.5 vvm) were performed with both
strains. The concentration profiles of substrates and
products for the fermentation with the low XR strain,
are shown in Figure 3. The initial glucose concentration
was 40 g L21, so that a higher initial biomass concentra-
tion (7.3 g L21 d.w.) compared with the anaerobic fer-
Figure 3. Concentration profiles for oxygen limited fed-batch xylitol
production with the low XR strain. In addition to consumption and
production, the concentrations are subject to dilution by the glucose
feed stream, 2.3 mL h21. Glucose (s), ethanol (d) xylose (n), xylitol
(m) and biomass (1).
mentations was reached. The glucose feed was contin-
ued for a longer period of time, than in the anaerobic
fermentations, allowing more than 90% of the initially
added xylose to be converted into xylitol. The glucose
was almost completely consumed at all times, the resid-
ual glucose concentration never exceeding 0.3 g L21. As
in the anaerobic fermentations, no increase in biomass
concentration during the xylitol production phase was
observed. Comparison of the mean productivities with
those obtained with the same strain under anaerobic
conditions (Table I), shows that the specific productivity
is in the same range, while the higher biomass concentra-
tion in the oxygen limited fermentation led to a higher
volumetric productivity. The xylitol yield on glucose was
somewhat lower than under anaerobic conditions. In
the oxygen limited fermentation with the high XR strain
very low xylitol productivities were observed (Table I).
Strain Stability
To maintain the XYL1 containing plasmid in the cells,
selective medium was used in all fermentations. The
selective medium lacks leucine, which makes retention
of the plasmid containing the LEU2 selection marker
essential for growth. The stability of the heterologous
enzyme activity was evaluated by comparing the in vitro
XR activity of the inoculum and the fermentor culture.
The fraction of cells retaining the plasmid (the selection
marker) was estimated by plating samples of the cultures
on selective and unselective plates. These analyses were
performed in the anaerobic fermentations with high
feed rates and in the oxygen limited fermentation with
the high XR strain. Xylose conversion with the high XR
strain was also investigated in glucose limited chemostat
cultivations with 15 g L21 xylose in the medium. At
constant dilution rate, xylitol productivity decreased
from 0.095 to 0.016 g g21 h21 (285%) in 100 hours. The
continuous fermentation was repeated with similar re-
sults.
The XR activities and fractions of plasmid containing
cells are summarized in Table II. A decreasing XR activ-

MEINANDER AND HAHN-HÄGERDAL: XYLITOL PRODUCTION BY RECOMBINANT S. CEREVISIAE 395

Table II. XR activities and fraction of plasmid containing cells of samples obtained from fed-batch and chemostat cultivations at the indicated
sampling times, compared with XR activities of the inoculum.

Fermentation Activity of inoculum Activity of culture Sampling time Fraction of plasmid-

conditions Strain (U/mg) (U/mg) (h) containing cells (%)

Anaerobic fed-batch Low XR 0.3a 0.2 114 100

Anaerobic fed-batch High XR 6.5a 3.6 114 100
Oxygen lim. fed-batch High XR 4.0 0.2 167 80
4.0 0.01 241 100
Chemostat High XR 4.3 0.3 314 55
Chemostat High XR 6.5a 0.02 1052 n.d.
a
The activity of the actual inoculum was not measured, but the measurement was made on shake flask cultures grown similarly to the inoculum.
n.d. 5 not determined.

ity was observed for the high XR strain in all cases. The 7 out of the 8 plasmids had identical cleavage patterns
poor xylitol productivity observed in the oxygen limited with the original pM2, while one showed quite different
fed-batch fermentation (Table I) therefore appears to patterns for all restriction enzymes. The size of pM2
be due to a rapid loss of enzyme activity. With the is 12.5 kb, and the size of the dissimilar plasmid was
low XR strain, only a slight decrease in activity was estimated to around 10.5 kb on the basis of the mean
observed. This is in agreement with the previously re- of the sums of the fragment sizes obtained from the
ported good stability of the low XR strain during 1050 different digestions. The size of the XYL1 gene is 954 bp.
hours of chemostat cultivation (Meinander et al., 1996). However, the cleavage patterns show that the dissimilar
In the present study, the high XR strain lost almost all plasmid was not a result of the XYL1 gene simply being
of its XR activity under similar conditions. The fraction removed, but point to extensive recombination events
of plasmid-containing cells at the end of the fermenta- having taken place, both adding and subtracting some
tions indicated that the plasmid was usually not lost to cleavage sites from the plasmid, leaving hardly any of
the same extent as the activity decreased. In many cases the fragments intact.
no plasmid loss was detected; however, in the chemostat To test whether the isolated plasmids had intact XR
cultivation it was apparent that a part of the population genes, the yeast host strain H158 was transformed with
had lost the plasmid (Table II). each plasmid, by the LiAc method. The resulting strains
Analysis of the amino acid content of the fermenta- were grown in shake flasks, the cells were disintegrated,
tion broth at 290 h and 314 h in the 314 h long chemostat and the XR activities of the lysates were measured. The
cultivation, showed that leucine was present in the broth strains transformed with the plasmids showing identical
in concentrations of 8.8 and 4.7 mg/L, respectively. In restriction cleavage patterns with pM2 all exhibited XR
addition to histidine and tryptophane, which were added activities between 4.25 and 5.94 U/mg, with one excep-
to the medium, several other amino-acids, for example
asparagine, serine, glutamine, proline, glycine, alanine,
cysteine, valine, isoleucine and lysine were also detected
in the broth in varying concentrations. This shows that
auxotrophic complementation of amino acids is not a
suitable selection method in long cultivations, where
substantial amounts of amino acids may be released
through excretion or cell lysis.
To further investigate the reason for the decrease in
XR activity, which could only partly be attributed to
loss of the plasmid (Table II), cells taken from the end of
the 314 h long chemostat cultivation were disintegrated,
and plasmids were prepared from the lysate. To isolate
individual plasmids, E. coli DH5a was transformed with
the plasmid preparation by electroporation.
After several transformation attempts, in which only
a few transformants were obtained each time, approxi-
mately 50 Escherichia coli clones were collected, of
which 8 were randomly chosen for further investigation.
Figure 4. Cleavage patterns of pM2 (C1), pM2 treated in the same
The 8 plasmids were prepared from E. coli and analyzed way as the plasmid preparation from the chemostat culture (C2), and
by restriction cleavage with EcoRI (Fig. 4), HindIII, eight plasmids isolated from cells harvested from chemostat culture,
NcoI and SalI and their combinations. It was found that digested with EcoRI.

396 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 54, NO. 4, MAY 20, 1997
tion of 16.5 U/mg. The strain transformed with the dis-
similar plasmid had no XR activity. In magnetically
stirred stoppered flasks the mmax of the strain containing
the dissimilar plasmid was about 50% lower than that
of the original strain.
DISCUSSION
A 20-fold enhanced XR expression level resulted in
a less than twofold increase in xylitol productivity,
indicating that transport of xylose into the cell or
generation of reduced cofactors for the reduction of
xylose are more important rate controlling factors for
xylitol production, than the specific XR activity. The
result confirms similar conclusions reached indirectly
in physiological studies of the low XR strain (Thestrup
and Hahn-Hägerdal, 1995; Meinander et al., 1996) and
with a XYL1 expressing mutant unable to produce
glycerol. This strain had a three times higher xylitol
productivity (0.38 g g21 h21) due to increased availa-
bility of reduced cofactors for the reduction of xylose
(Lidén et al., 1997).
The specific xylitol productivity of the strains varied
with residual glucose concentration and biomass con-
centration, so that the highest specific productivities
were reached at low biomass and low residual glucose
concentrations. Under these conditions the specific pro-
ductivity of the high XR strain (0.19 g g21 h21) ap-
proached the highest values reported for xylitol produc-
tion by xylose metabolizing yeasts (0.28 g g21 h21,
(Meyrial et al., 1991), 0.45 g g21 h21 (Lu et al., 1995)).
The specific productivities of the low XR strain were
higher compared with fed-batch cultivation using etha-
nol as cosubstrate (Meinander et al., 1994) and in the
same range as previously reported in semi-continuous
cultivation with glucose as cosubstrate (Thestrup and
Hahn-Hägerdal, 1995). Due to the quite low biomass
concentrations reached on the mineral medium, the vol-
umetric productivities (0.4–0.6 g L21 h21), were some-
what lower than the best result previously obtained with
ethanol as cosubstrate (1 g L21 h21 (Meinander et al.,
1994)) and considerably lower than the best result re-
ported for xylose metabolizing yeasts (2.67 g L21 h21,
(Horitsu et al., 1992)). In a continuous immobilized cell
bioreactor, with a high concentration of immobilized
biomass, volumetric productivities of 3.4 g L21 h21 and
5.8 g L21 h21 have been achieved with the low and high
XR strains, respectively (Roca et al., 1996).
The batch fermentations showed that xylose was not
converted to xylitol in the presence of high glucose
concentrations, and in fed-batch fermentations xylitol
production stopped when glucose accumulated in the
broth (Table I). Xylose and glucose are both transported
by the facilitated glucose transport system in S. cere-
visiae (Leao and van Uden, 1982; Busturia and Lagunas,
1986; van Zyl et al., 1993), which has about 100-fold
higher affinity for glucose than xylose (Busturia and
MEINANDER AND HAHN-HÄGERDAL: XYLITOL PRODUCTION BY RECOMBINANT S. CEREVISIAE
Lagunas, 1986). In the presence of high glucose concen-
trations the transport system is saturated with glucose,
permitting no xylose transport. Therefore, if glucose is
to be used as cosubstrate for xylitol production, it must
be fed gradually. The residual glucose concentration
was quite low with a feed of 1 g h21, but obviously too
high with 2 g h21 under anaerobic conditions (Table I).
In oxygen limited fermentation with 1 g h21 feed, the
residual glucose concentration was near zero through-
out the fermentation, which might have limited the xyli-
tol production rate. Therefore, large scale fed-batch
xylitol production with glucose as cosubstrate would
require on-line feed rate control.
The advantages of using glucose as cosubstrate com-
pared with ethanol are higher specific productivity,
lower cosubstrate cost, and lower processing cost be-
cause aeration is not required. A drawback is the re-
quirement of stringent feed-rate control. With ethanol,
less stringent feed-rate control is required, but the co-
substrate is more expensive, and acetic acid excreted as
a byproduct of ethanol metabolism is toxic to the yeast.
The specific activity of purified XR from P. stipitis
has been reported to be 23.5 U/mg, 47.8 U/mg and 43.2
U/mg (Verduyn et al., 1985; Rizzi et al., 1988; Webb
and Lee, 1992). Using these values, XR can be calculated
to constitute about 1–2% of the total cell protein in the
low XR strain and about 23–46% in the high XR strain.
The high expression of recombinant enzyme in the high
XR strain almost halved the growth rate (from 0.19 to
0.098 h21) and lowered the biomass yield. The effect of
protein burden has been suggested to be due to the
energy cost of the synthesis of the high amount of recom-
binant protein, leaving less energy available for growth
(Koch, 1983; Bailey, 1993).
Recently, a simple quantitative model for the effect
of overproduction of glycolytic enzymes on growth rate
and metabolic flux in Zymomonas mobilis was devel-
oped (Snoep et al., 1995). The model was confirmed by
experimental data, showing that the overexpression of
one enzyme diluted all other enzyme activities in the
cell, leading to reduced flux and therefore to reduced
growth rate. In addition to the poor volumetric produc-
tivity caused by low growth rate and poor biomass yield,
a strain with high protein burden is easily outcompeted
by mutant strains with lower protein burden allowing
higher growth rate.
The two strains showed marked differences in stabil-
ity during long fermentations (Table II). The low XR
strain retained its XR activity quite well in fed-batch
fermentation and has previously been stable during long
chemostat cultivations (Meinander et al., 1996) and dur-
ing fed-batch fermentations in rich medium (Meinander
et al., 1994). The high XR strain, on the other hand,
lost most of its XR activity during the two chemostat
cultivations, and during the longer fed-batch fermenta-
tion (Table II). The high XR strain has also been found
to be considerably less stable than the low XR strain in
397
continuous fermentation with immobilized cells, where
both plasmid loss and loss of XR activity were observed
(Roca et al., 1996). The plasmid loss can be explained by
leucine release either from dead lysed cells, or plasmid
containing cells producing more leucine than needed
for biosynthesis. However, the decrease in XR activity
cannot be quantitatively accounted for by plasmid loss.
In addition, the plasmid can never be totally lost from
a culture for this reason, because a certain number of
leucine producing cells is required to supply the cells
lacking the plasmid with leucine.
Other possible explanations for the loss of XR activity
include: (1) spontaneous integration of the LEU2 gene
into the genome through homologous recombination,
(2) recombinations or point mutations in the plasmid
silencing the XR gene, or leading to the expression of
an inactive protein, (3) segregational instability leading
to highly varying plasmid copy numbers in individual
cells (Bugeja et al., 1989) and, (4) effects on the PGK
promoter leading to decreased or no expression of
XYL1.
In the present study, only point (2) was investigated,
and a rearranged plasmid was isolated from a culture
having lost most of its original XR activity. However,
the rearranged plasmid did not appear to be widespread
in the population, since it was only recovered as one
out of eight plasmids isolated. For a mutant strain to
dominate, it must have a growth advantage over the
original strain. Determinations of mmax showed that the
rearranged plasmid did not confer such advantage to
the strain. Another type of rearrangement yielding a
plasmid that was not recovered through the used isola-
tion method, or any of the other points listed above,
could produce a strain with growth advantage, and thus
be the true explanation for the loss of activity. Since the
original plasmid could be recovered from the culture,
a likely explanation could be segregational instability,
where cells containing very few copies of the plasmid,
and therefore having a low expression of XYL1 would
dominate the culture.
The fact remains that the high XR strain was less
stable than the low XR strain. This could be due to the
protein burden effect, which implies that the high XR
strain with a low growth rate is easily outcompeted by
a mutated strain with lower protein burden. The present
results demonstrate the importance of optimizing,
rather than maximizing, the expression level of recombi-
nant enzymes, in the development of metabolically engi-
neered production organisms.
Anne Blicher, Department of Biochemistry and Nutrition,
Technical University of Denmark, is thanked for performing
the amino acid analysis.
References
Bailey, J. E. 1993. Host-vector interactions in Escherichia coli. Adv.
Biochem. Eng. 48: 29–52.
398 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 54, NO. 4, MAY 20, 1997
Barkholt, V., Jensen, A. L. 1989. Amino acid analysis: Determination
of cysteine plus half-cysteine in proteins after hydrochloric acid
hydrolysis with a disulfide compound as additive. Anal. Biochem.
177: 318–322.
Bradford, M. M. 1976. A rapid and sensitive method for the quantita-
tion of microgram quantities of protein utilizing the principle of
protein-dye binding. Anal. Biochem. 72: 248–254.
Bugeja, V. C., Kleinman, M. J., Stanbury, P. F., Gingold, E. B. 1989.
The segregation of the 2m-based yeast plasmid pJDB248 breaks
down under conditions of slow, glucose limited growth. J. Gen.
Microbiol. 135: 2891–2897.
Busturia, A., Lagunas, R. 1986. Catabolite inactivation of the glucose
transport system in Saccharomyces cerevisiae. J. Gen. Microbiol.
132: 379–385.
Chen, L. F., Gong, C.-S. 1985. Fermentation of sugarcane bagasse
hemicellulose hydrolysate to xylitol by hydrolysate-acclimatized
yeast. J. Food. Sci. 50: 226–228.
Gong, C.-S., Chen, L. F., Tsao, G. T. 1981. Quantitative production
of xylitol from D-xylose by a high-xylitol producing yeast mutant
Candida tropicalis HXP2. Biotechnol. Lett. 3: 130–135.
Hallborn, J., Gorwa, M.-F., Meinander, N., Penttilä, M., Keränen, S.,
Hahn-Hägerdal, B. 1994. The influence of cosubstrate and aeration
on xylitol formation by recombinant Saccharomyces cerevisiae ex-
pressing the XYL1 gene. Appl. Microbiol. Biotechnol. 42: 326–333.
Hallborn, J., Walfridsson, M., Airaksinen, U., Ojamo, H., Hahn-Häg-
erdal, B., Penttilä, M., Keränen, S. 1991. Xylitol production by
recombinant Saccharomyces cerevisiae. Bio/Technol. 9: 1090–1095.
Horitsu, H., Yahashi, Y., Takamizawa, K., Kawai, K., Suzuki, T.,
Watanabe, N. 1992. Production of xylitol from D-xylose by Candida
tropicals: Optimization of production rate. Biotechnol. Bioeng.
40: 1085–1091.
Koch, A. L. 1983. The protein burden of lac operon products. J. Mol.
Evol. 19: 455–462.
Leao, C., van Uden, N. 1982. Effects of ethanol and other alkanols
on the glucose transport system of Saccharomyces cerevisiae. Bio-
technol. Bioeng. 24: 2601–2604.
Lidén, G., Walfridsson, M., Ansell, R., Anderlund, M., Adler, L.,
Hahn-Hägerdal, B. 1997. A glycerol-3-phosphate dehydrogenase-
deficient mutant of Saccharomyces cerevisiae expressing the heterol-
ogous XYL-1 gene. Appl. Environ. Microbiol. 62: 3894–3896.
Lu, J., Tsai, L. B., Gong, S. C., Tsao, G. T. 1995. Effect of nitrogen
sources on xylitol production from D-xylose by Candida Sp. L-102.
Biotechnol. Lett. 17: 167–170.
Meinander, N., Hahn-Hägerdal, B., Linko, M., Linko, P., Ojamo,
H. 1994. Fed-batch xylitol production with recombinant XYL1-
expressing Saccharomyces cerevisiae using ethanol as a co-substrate.
Appl. Microbiol. Biotechnol. 42: 334–339.
Meinander, N., Zacchi, G., Hahn-Hägerdal, B. 1996. A heterologous
reductase affects the redox balance of recombinant Saccharomyces
cerevisiae. Microbiol. 142: 165–172.
Mellor, J., Dobson, M. J., Roberts, N. A., Tuite, M. F., Emtage, J. S.,
White, S., Lowe, P. A., Patel, P. A., Kingsman, A. J., Kingsman,
S. M. 1983. Efficient synthesis of enzymatically active calf chymosin
in Saccharomyces cerevisiae. Gene 24: 1–14.
Meyrial, V., Delgenes, J. P., Moletta, R., Navarro, J. M. 1991. Xylitol
production from D-xylose by Candida guillermondii: Fermentation
behaviour. Biotechnol. Lett. 13: 281–286.
Rizzi, M., Erlemann, P., Bui-Thahn, N.-A., Dellweg, H. 1988. Xylose
fermentation by yeasts. 4. Purification and kinetic studies of xylose
reductase from Pichia stipitis. Appl. Microbiol. Biotechnol. 29:
148–154.
Roca, E., Meinander, N., Hahn-Hägerdal, B. 1996. Xylitol production
by immobilized recombinant Saccharomyces cerevisiae in a continu-
ous packed-bed bioreactor. Biotechnol. Bioeng. 51: 317–326.
Roseiro, J. C., Peito, M. A., Gı́rio, F. M., Amaral-Collaco, M. T. 1991.
The effects of oxygen transfer coefficients and substrate concentra-
tion on the xylose fermentation by Debaromyces hansenii. Arch.
Microbiol. 156: 484–490.
Sambrook, J., Fritsch, E. F., Maniatis, T. 1989. Molecular cloning: A
laboratory manual. Cold Spring Harbor Laboratory, Cold Spring
Harbor, New York.
Schiestl, R. H., Gietz, D. 1989. High efficiency transformation of intact
yeast cells by single stranded nucleic acids as carrier. Curr. Genet.
16: 339–346.
Sherman, F., Fink, G., Hicks, J. B. 1983. Methods in yeast genetics:
A laboratory manual. Cold Spring Harbor Laboratory, Cold Spring
Harbor, New York.
Snoep, J. L., Yomano, L. P., Westerhoff, H. V., Ingram, L. O. 1995.
Protein burden in Zymomonas mobilis: Negative flux and growth
control due to overproduction of glycolytic enzymes. Microbiol.
141: 2329–2337.
Strathern, J. N., Higgins, D. R. 1991. Recovery of plasmids from yeast
into Escherichia coli: Shuttle vectors, pp. 319–329. In: C. Guthrie
and G. R. Fink (eds.), Guide to yeast genetics and molecular biol-
ogy. Academic Press, San Diego.
Thestrup, H. N., Hahn-Hägerdal, B. 1995. Xylitol formation and re-
duction equivalent generation during anaerobic xylose conversion
with glucose as cosubstrate in recombinant Saccharomyces cere-
visiae expressing the XYL1 gene. Appl. Environ. Microbiol. 61:
2043–2045.
Thonart, P., Guerreiro, M., Foucart, M., Paquot, M. 1987. Bioconver-
sion of xylose into xylitol by Pachysolen tannophilus. Med. Fac.
Landbouww. Rijksuniv. Gent. 52: 1517–1529.
MEINANDER AND HAHN-HÄGERDAL: XYLITOL PRODUCTION BY RECOMBINANT S. CEREVISIAE
Walfridsson, M., Bao, X., Anderlund, M., Hahn-Hägerdal, B. Expres-
sion of different levels of enzymes from the Pichia stipitis XYL1
and XYL2 genes in Saccharomyces cerevisiae and their effects on
product formation during xylose utilisation, to appear.
Walfridsson, M., Penttilä, M., Hahn-Hägerdal, B., Keränen, S. 1992.
Recombinant Saccharomyces cerevisiae expressing xylose reductase
and xylitol dehydrogenase under control of different promoters.
ECB6 Sixth European Congress on Biotechnology. Firenze, Italy.
van Zyl, C., Prior, B. A., Kilian, S. G., Brandt, E. V. 1993. Role of
D-ribose as a cometabolite in D-xylose metabolism by Saccharo-
myces cerevisiae. Appl. Environ. Microbiol. 59: 1487–1494.
Watson, N. E., Prior, B. A., du Preez, J. C., Lategan, P. M. 1984. Oxy-
gen requirements for D-xylose fermentation to ethanol and polyols
by Pachysolen tannophilus. Enzyme Microb. Technol. 6: 447–450.
Webb, S. R., Lee, H. 1992. Characterization of xylose reductase from
the yeast Pichia stiptis: Evidence for functional thiol and histidyl
groups. J. Gen. Microbiol. 138: 1857–1863.
Verduyn, C., Postma, E., Scheffers, W. A., van Dijken, J. P. 1992.
Effect of benzoic acid on metabolic fluxes in yeasts: A continuous
culture study on the regulation of respiration and alcoholic fermen-
tation. Yeast 8: 501–517.
Verduyn, C., Van Kleef, R., Frank, J., Schreuder, H., Van Dijken,
J. P., Scheffers, W. A. 1985. Properties of the NAD(P)H-dependent
xylose reductase from the xylose fermenting yeast Pichia stipitis.
Biochem. J. 226: 669–677.
399