Here, we refer to the highest reproduction rate of cells in terms of the maximum specific growth rate lmax, which is expressed as an increase in net flux of biomass, Jgrowth-rate, per unit
Trang 1Super life – how and why ‘cell selection’ leads
to the fastest-growing eukaryote
Philip Groeneveld1, Adriaan H Stouthamer1and Hans V Westerhoff1,2,3
1 Department of Molecular Cell Physiology & Mathematical Biochemistry, Netherlands Institute for Systems Biology, Vrije Universiteit, Amsterdam, The Netherlands
2 The Manchester Centre for Integrative Systems Biology, Manchester Interdisciplinary Biocentre, School of Chemical Engineering
and Analytical Science, The University of Manchester, UK
3 Swammerdam Institute for Life Sciences, Netherlands Institute for Systems Biology, University of Amsterdam, The Netherlands
There is considerable interest in what determines the
rate at which reproductive growth occurs This issue
is most intriguing for the ‘maximum’ growth rate
(Jgrowth-max) of the fastest independently replicating
organism, relatives of which are used commercially as
‘living factories’ The fastest-dividing organisms are
micro-organisms, and we limit our analysis to
eukary-otic microbes, as they are most similar to cells of
higher organisms The cell-cycle time of one of the
fastest-growing eukaryotes (i.e a generation time of
70 min [1]) is still seven times longer than that of one
of the fastest-growing prokaryotes (i.e a generation
time of < 10 min [2,3]) One of the known
fastest-growing microbial eukaryotes is the non-pathogenic
industrial yeast Kluyveromyces marxianus, which GRAS
status (‘generally recognized as safe’) For these reasons,
this organism has been chosen as an efficient vehicle for single-cell protein production [4–7] In this context,
we do not consider early transient cleavage during fast embryonic growth of eukaryotes such as Xenopus laevis [8] Reproduction in terms of cell number by cleavage is much faster but the net biomass remain constant Here, we refer to the highest reproduction rate of cells in terms of the maximum specific growth rate (lmax), which is expressed as an increase in net flux of biomass, Jgrowth-rate, per unit of cell mass or total protein, and equals ‘ln 2’ divided by the genera-tion or cell-cycle time The quesgenera-tions posed in this study also address the minimum cell-cycle time The maximum (specific) growth rate refers to cellu-lar biosynthesis during which all nutrients are supplied
in excess (i.e substrate-saturated conditions relative to
Keywords
highest eukaryotic growth rate; modular
control analysis; pH-auxostat selection;
surface-to-volume ratio optimization;
systems biology
Correspondence
H V Westerhoff, The Manchester Centre
for Integrative Systems Biology, SCEAS,
The University of Manchester, Manchester
Interdisciplinary Biocentre (MIB), 131
Princess Street, Manchester M1 7ND, UK
Fax: +44 161 306 8918
Tel: +44 161 306 4407
E-mail: Hans.Westerhoff@manchester.ac.uk
(Received 20 December 2007, revised 26
October 2008, accepted 3 November 2008)
doi:10.1111/j.1742-4658.2008.06778.x
What is the highest possible replication rate for living organisms? The cellular growth rate is controlled by a variety of processes Therefore, it is unclear which metabolic process or group of processes should be activated
to increase growth rate An organism that is already growing fast may already have optimized through evolution all processes that could be opti-mized readily, but may be confronted with a more generic limitation Here
we introduce a method called ‘cell selection’ to select for highest growth rate, and show how such a cellular site of ‘growth control’ was identified
By applying pH-auxostat cultivation to the already fast-growing yeast Kluyveromyces marxianus for a sufficiently long time, we selected a strain with a 30% increased growth rate; its cell-cycle time decreased to 52 min, much below that reported to date for any eukaryote The increase in growth rate was accompanied by a 40% increase in cell surface at a fairly constant cell volume We show how the increase in growth rate can be explained by a dominant (80%) limitation of growth by the group of membrane processes (a 0.7% increase of specific growth rate to a 1% increase in membrane sur-face area) Simultaneous activation of membrane processes may be what is required to accelerate growth of the fastest-growing form of eukaryotic life
to growth rates that are even faster, and may be of potential interest for single-cell protein production in industrial ‘White’ biotechnology processes
Trang 2their transporter enzymes), and is therefore only
limited by the biological properties of the cell itself [1]
In so-called ‘rich’ media, substrate-saturated conditions
refer to the ample supply of undefined monomeric
nutrients in addition to provision of the main basic
carbon (C) and Gibbs free-energy (E) sources In
defined ‘mineral’ media, cells have to synthesize these
monomers from the basic C⁄ E sources together with
mineral salts and vitamins If these biosynthetic
path-ways are insufficiently active (due to shortcomings of
any possible metabolic process within or coupled to
these pathways), the lmax on mineral medium will be
lower than that on rich medium Under both
condi-tions, control of lmax is solely determined by the
bio-logical properties or ‘dynamic hardware’ of the cell
itself, properties that may comprise at least four main
metabolic processes, i.e catabolism, anabolism,
main-tenance and transport [9–11] Each of these metabolic
groups consists of a network of interacting metabolic
pathways through which substrates flow, and by which
products, including new cell material, are formed It is
unknown which individual process exercises the
stron-gest constraints on the flux into new cell material, and
hence ‘controls’ lmax[12–14] In baker’s yeast
(Saccha-romyces cerevisiae) for instance, the primary catabolic
pathway is glycolysis, and any of the components of
this pathway might have been expected to control
gly-colytic flux and growth It has been shown, however,
that the control of the glycolytic enzymes on the
glyco-lytic flux is rather small in this yeast [15–19] There is
substantial, but incomplete, evidence for a high control
of the glucose-uptake step on the yeast glycolysis
[16,19–21] Control by glucose transport has been
shown to be limited in Salmonella typhimurium [22] It
is a frequent observation that activation of single
aspects of cell metabolism fails to increase major fluxes
in the cell such as the growth rate [17,23,24] This has
been attributed to a shift of the limitation to the
second most rate-limiting step [25] Indeed, control
of fluxes is often distributed among several steps and
layers [26–31]
For biotechnologists, this is bad news, as further
increases of microbial productivity do not seem to be
as simple as over-expressing a single rate-limiting
enzyme Although solutions to this problem have been
devised in principle, they require over-expression of
large proportions of the of cell metabolism to the same
strictly related [32] or to rather diverse [33] extents
The enzymes that need to be over-expressed to the
same extent belong to a functional unit [34,35] or level
[36] of cell metabolism Intracellular chemistry appears
to be organized in terms of such modules, which often
correspond to operons or regulons [37] The cell itself
may modulate its fluxes by increasing the expression level of such a regulon as a whole, through a single transcription factor [38–40] Consequently, a new approach to bioengineering may be to first identify the natural regulons of the host organism and then modu-late their activities towards the desired effect [41] Changes in the morphology of micro-organisms may influence their physiology, because some cellular fluxes depend primarily on cell volume and others on the cell surface area [42] This distinction plays an important role in understanding why unicellular organisms are as small as they are With increased size, the surface-to-volume ratio decreases, and the supply rate of Gibbs energy and chemical substrates becomes insufficient for cytoplasm-based catabolic and anabolic processes [43] With regard to identification of what limits the growth rate of already fast-growing unicellular organisms, membrane-located processes (or outer wall transport [23]) are therefore possibly a major site of control The dynamic energy budget (DEB) model [44,45] reinforces this viewpoint It describes microbial growth as based
on cellular uptake capacity and volume The former
is considered proportional to the cell surface and is assumed to control growth proportionally Thus varia-tion in cell morphology may change the cell’s surface-to-volume ratio and hence its specific growth rate
In a similar vein, Hennaut et al [46] have specifi-cally shown for the three anabolic substrates arginine, lysine and uridine that the relative uptake rates decrease in proportion to the surface-to-volume ratio
in a series of isogenic multiploid strains of baker’s yeast growing on defined medium enriched with these monomers Transport of these substrates is catalyzed
by constitutive permeases [47–49] For substrates such
as methionine and leucine, for which transport is inducible [50], such a decrease was not observed The results of the study by Hennaut et al [46] imply that the cytoplasmic membrane in a haploid strain is satu-rated (or nearly satusatu-rated) with these constitutive permeases With increasing ploidy, the cell surface may become more and more limited for permease insertion,
as the increase in cell surface will fall short of the increase in cell mass or cell volume We surmise that if the major site of control on growth rate indeed resides
in the module of membrane processes, and if the acti-vity of these processes per cell increases with increasing membrane surface area per cell, selection for increased maximum growth rate (lmax) should yield strains with increased surface-to-volume ratios
Although substantial progress is being made with regard to understanding of the modular organization
of cell metabolism [35,51], it is not yet feasible to pre-dict how the module of membrane processes may
Trang 3be activated selectively However, as the issue is one
of growth rate, it may be possible to manipulate the
organism to do this by itself With this aim, the
organ-ism should be cultivated under conditions that select
for increased lmax For already fast-growing
micro-organisms, it is difficult to perform such an experiment
under well-defined fermenter conditions During batch
cultivation, there is only a limited time period during
which the cells are in steady state At higher cell
densi-ties, factors other than lmax are selected for The
best-known continuous culture system, the chemostat
[52,53], is not suitable because it is unstable at dilution
rates close to lmax Of the suitable continuous culture
systems, such as the turbidostat, or permittistat [54]
and the pH-auxostat [55,56], we here chose the latter
to select for increased lmax
A second uncertainty in our objective of selecting a
faster-growing variant of an already fast-growing
eukaryotic microbe is whether such a variant can exist
at all Because of the maximum limit of
diffu-sion-limited association, and because the complicated
chemistry of some biochemical reactions takes time,
there are maximum rates at which the processes
syn-thesizing new cell material can operate Making more
enzymes to catalyze these processes shifts but does not
eliminate the upper limit of growth rate, as the new
enzymes also have to be synthesized [57–59]
Conse-quently, due to limitations of chemistry, physics and
biocomplexity, there must be a ‘highest possible’
maxi-mum specific growth rate for living organisms, i.e a
‘lowest minimum’ cell-cycle time As additional
pro-cesses may well serve to enhance rates and efficiency,
this highest possible growth rate is unlikely to be
found in so-called ‘minimal organisms’, i.e organisms
with the smallest possible genome [60] or in extreme
thermophiles [61], because both are associated with
slow growth Yeasts from the genus Kluyveromyces,
however, constitute a case in point because of their
excellent (industrial) growth characteristics K
marxi-anus, in particular, already has a high specific growth
rate (approximately twice as high as baker’s yeast), a
high aerobic biomass yield (because of its
Crabtree-negative physiology [55,62]) and a high optimum
growth temperature (40C, which reduces the cooling
costs of large bioreactors) [15,63] Therefore, K
marxi-anus may be close to the true ‘absolute’ maximum
growth rate, perhaps even too close for any further
increase to occur on defined medium conditions
In this paper, we address four questions: (1) Can
one use the pH-auxostat to select for even
faster-grow-ing variants of fast-growfaster-grow-ing eukaryotic
micro-organ-isms? (2) Can an industrially useful yeast such as
K marxianus grow even faster than it already does?
(3) What is the fastest possible growth rate for eukary-otic life on defined medium? (4) To what extent does this indicate that the highest growth rate is controlled
by the surface-to-volume ratio? We report the selection
of a much faster-growing variant of K marxianus with
an almost proportional increase in surface-to-volume ratio We developed a bimodular control analysis to express growth control in quantitative terms for two separate cellular groups (functional modules) Control exerted by transport processes (module 1, including all membrane-located processes) and that exerted by intra-cellular metabolism (module 2, including all cyto-plasm-based processes) was defined and quantified
In the present post-genomic era, the methodologies presented here may offer a new integrative ‘top-down’ systems approach [10,64] for the identification of major sites of control on cellular growth rate
Results General characteristics of a pH-auxostat When continuous cultivation of microbial cultures at maximum specific growth rate (lmax) is required for and growth is accompanied by changes in pH, pH-auxostat or ‘phauxostat’ culturing may be the method of choice [56] The pH-auxostat is a continu-ous culture system in which, unlike the chemostat, the dilution rate can vary according to the properties of the micro-organism As in the chemostat, there is a continuous supply of growth medium and an equal continuous efflux of culture The two fluxes are mea-sured in terms of volume per unit time per unit volume
of the culture, i.e the dilution rate, D In the pH-auxo-stat, addition of fresh medium is coupled to pH con-trol of the medium in the culture vessel As the pH of the culture drifts from a given set point, fresh medium
is added to bring the pH back to the set point Thus, this system has an external control loop that keeps the
pH difference between the culture vessel and the reser-voir constant by adjusting the dilution rate When the difference in pH (DpH, defined as pH culture) pH reservoir, which in our set-up has a negative value), the biomass concentration is determined only by the buffering capacity of the inflowing medium (BCR, defined as the amount of acid or base required to change the pH of 1 L of the medium in the reservoir
to the pH of the medium in the culture vessel [56]), provided that a constant number of protons are pro-duced per unit biomass synthesized The rate of growth is independent of BCR and DpH, and depends only on the conditions under which the micro-organism is cultured and the properties of the
Trang 4micro-organism itself By computerized feedback
con-trol, the dilution rate is adjusted so as to make the pH
independent of time When all growth substrates are
supplied in sufficient excess and the BCRand DpH are
kept constant, a steady state at lmaxcan be maintained
at which the pH and biomass concentration in the
cul-ture vessel remain constant
Selection for highest cellular growth rate
(minimum cell-cycle time)
K marxianusCBS 6556 was cultured in a pH-auxostat
growing on defined mineral medium with all nutrients
(essential vitamins and mineral salts with ammonium
as the main nitrogen source) in excess and with glucose
as sole carbon and Gibbs free-energy source A few
hours after inoculation, the culture produced enough
protons to trigger the feed pump, supplying the culture
with fresh medium (pH 6.2), which kept the culture
pH constant at 4.5 From that point, a continuous
supply of fresh medium by the pH-controlled pump
and the removal of equal amounts of culture medium
by the fermenter overflow-outlet, keep all physiological
parameters constant in time The average maximum
specific growth rate (lmax) of the yeast population,
measured online as the culture’s dilution rate (D), was
0.6 h)1, (g new biomass per g biomass per hour) and
appeared to be constant during the initial 60 h of
culti-vation (Fig 1, Exp 1) After this long stable period
(which corresponds to 50 generations), the culture’s
dilution rate suddenly began to increase, i.e the
aver-age lmax increased from 0.6 to 0.8 h)1 within a period
of approximately 40 h This second steady state
remained constant for more than three subsequent
days The entire experiment was repeated more than
three times with essentially the same results Another
of these experiments is also shown in Fig 1 (Exp 2),
exhibiting an increase of lmax from 0.57 to 0.79 h)1
The ratios of the lmax values for the first steady state
to that for the second steady state in these long-term
pH-auxostat cultivations were 1.33 and 1.39 for the
two independent complete experiments
The highest cellular growth rate remains stable
outside the pH-auxostat
To exclude any culture contamination with other
micro-organisms, e.g with faster-growing prokaryotes,
liquid samples from the pH-auxostat were taken before
and after selection; both cultivars were identified as
K marxianus CBS 6556 at the Centraal Bureau voor
Schimmel (CBS Delft) Nutrient variation between
both steady states was excluded by changing vitamin
and mineral concentrations in control experiments; no effect on the pattern of dilution rate was observed In addition, plated colonies of the initial K marxianus CBS 6556 strain (obtained during the first steady state) and of the selected variant of our K marxianus CBS
6556 strain (obtained during the second steady state) were examined in subsequent regulated batch cultures (as shown in the inset to Fig 1) The selected popula-tion produced much more carbon dioxide over time, in agreement with its increased growth rate From the gas exchange, we calculated the lmaxvalue under batch conditions During the exponential growth phase, gas exchange (oxygen consumption and carbon dioxide production) should be directly proportional to the maximum specific growth rate The lmax for the selected cells remained 0.8 h)1, and that for the initial population was again 0.6 h)1 We also verified the sta-bility of the new growth characteristics by frequently re-plating a colony of the selected cells on defined medium agar plates The selected growth rate did not revert to the initial value after re-plating more than 10
A
B
–1 ·h
–1 )
Time (h)
µ max = 0.8 h–1
D = µmax Exp.1 Average cell-size (diameter) Exp.2
D = µmax Exp.2
µmax = 0.6 h –1
Fig 1 Specific growth rate and average cell size during two long-term pH-auxostat cultivations of K marxianus Right ordinate: steady-state dilution rate that equals the culture’s average maxi-mum specific growth rate (lmax) on defined mineral medium at optimal culture conditions (i.e saturated concentrations of all growth substrates, full air supply, pH 4.5 and temperature 40 C; open diamonds, l max for experiment 1; closed squares, l max for experiment 2) Left ordinate: change in average cell size (i.e mean cell diameter in lm as measured with a Coulter Counter particle size analyzer) during the second long-term pH-auxostat cultivation (closed circles, experiment 2) Inset: log of the percentage carbon dioxide production and oxygen consumption during regulated batch cultivation Two batch cultures were inoculated with either selected
or initial K marxianus colonies.
Trang 5times (see inset to Fig 1) As shown in the inset to
Fig 1, the re-plated colonies started in the
pH-auxo-stat at a lmaxof 0.8 h)1, whereas the lmaxof cells from
the initial steady state remained at 0.6 h)1 Clearly,
the characteristics gained in the pH-auxostat were
inherited over more than 400 generations (assuming at
least 40 generations per plate) in a non-pH-auxostat
environment Finally, we tested the stability of the
selected higher growth rate by glucose-limited
chemo-stat cultivation; 120 h of steady-chemo-state glucose-limited
growth at sub-maximum rate (Dss= l = 0.2 h)1) did
not cause the selected growth performance of
K marxianusto revert to the initial value (data shown
in [1]) Therefore, it was not likely that the increase in
lmax was caused only by an unknown, reversible
mechanism of cellular adaptation which occurred
under these rather special pH-auxostat conditions
Morphology changes during selection for highest
growth rate
As can be seen in Fig 1 (left ordinate), the average
cell-size distribution was constant during the initial
60 h of pH-auxostat cultivation However, during the
sudden increase in dilution rate, the average cell size
(measured as relative cell diameter) increased in
par-allel with the increase in cellular growth rate The
increase was accompanied by significant alterations in the average cell-size distribution (see Fig 2A) The dis-tribution of the average cell-size increased within 40 h
in parallel with the increase in specific growth rate each time after 60 h from the beginning of the steady pH-auxostat cultivation However, these data are just
a rough indication of the cell-size distribution in the yeast population and could not be used to calculate average relative changes in cell sizes Therefore, we used a phase-contrast microscope to measure the aver-age relative changes in cell sizes within the yeast popu-lation These microscopic observations accurately revealed the cause of the alterations in gross cell-size distribution: the average individual cell morphology changed considerably during the selection for highest growth rate The cell shape changed from spheroid (or ovoid) to an elongated cell form between the two suc-cessive pH-auxostat steady states (see Fig 2B,C) Although yeasts are usually regarded as discrete bud-ding ovoid cells, some genera exhibit dimorphism by producing mycelial or elongated growth forms under certain environmental conditions [65] Fig 2 clearly shows dimorphism of K marxianus, i.e transition from round ovoid cell morphology to elongated fila-ments when selection for the highest specific growth rate took place under defined optimal medium condi-tions in a pH-auxostat
Steady state 1
Steady state 2
Average cell-size distribution (µm)
l = n· d c, f (with n = 6)
length spheroid (ls) = diameter (ds)
Cell-size distribution and morphology during long-term pH-auxostat cultivation
A
B
C
6 cells·L
–1)
b: bud or daughter m: mother cell cell
Fig 2 Cell-size distribution (A) and morphology changes during long-term pH-auxostat cultivation at steady states 1 (B) and 2 (C) as deter-mined using a Coulter counter particle size analyzer and observed under a microscope.
Trang 6The cell surface-to-volume ratio increased during
selection for the highest growth rate
We estimated surface-to-volume ratios based on
micro-scopic observations The calculations did not show
significant geometric changes in cell volumes (v) during
the alteration of cell morphology (vspheroid= 0.113;
vcylinder= 0.104–0.120), nor in the biomass
concentra-tion of the culture (data not shown) We considered
the cell shape during the first steady state to
corre-spond to that of round ovoid spheres (spheroid s, as
shown in the inset to Fig 2B), and that during the
second steady state as elongated cells (cylinders, c,
or filaments, f, as shown in the inset to Fig 2C)
We estimated the length (l) of elongated cells to
be approximately six times their diameter (d) As
described in Doc S1, we estimated the volume and
surface area of the selected elongated cells in two
ways: by considering these cells as cylinders (c) with a
flat surface at both ends, or by considering these cells
as cylinder-like filaments (f) with both ends as half
spheroids Accordingly, the surface-to-volume (s⁄ v)
ratio of the two elongated cell types was estimated to
be between 1.44 and 1.50
Quantification of growth control by outer
membranes versus intracellular processes
Had the growth rate been precisely proportional to the
s⁄ v ratio, lmax would have increased by this same
factor i.e between 1.44 and 1.50 The experimentally
determined increase in lmax of 1.33–1.39 was not
far from this ratio 1.44–1.50, suggesting that much
growth control might well reside in membrane-located
processes
We then estimated how much growth control must
reside in the membrane processes to account for the
actual increase in maximum growth rate For this, we
needed to define a quantifier for the extent of growth
control by the outer membrane This quantifier is
called the control coefficient for growth control by
membrane processes, and is defined as the relative
increase in maximum growth rate for a 1% increase in
the activities of all membrane processes, or more
pre-cisely as:
CJ growthmax
m d ln Jgrowthmax
d ln m
where m refers to the activity of the membrane
pro-cesses Doc S2 shows that this control coefficient is
related to the ratio of the relative increase in growth
rate and the relative increase in surface-to-volume ratio
umby:
Clmax
m CJgrowthmax
m ¼ /mþ ð1 /mÞd ln lmax
d ln /m where Clmax
m refers to the control by membrane pro-cesses on the maximum ‘specific’ growth rate Inserting the experimental observations into this equation, and assuming that the surface-to-volume ratio is 10%, leads to an estimate for the control of maximum (spe-cific) growth rate by the membrane processes of 0.8 ± 0.1 This implies that the control by cytoplasmic processes must be 0.2 ± 0.1, i.e the control by mem-brane processes exerted on the maximum specific growth rate appears to be four times stronger than the control by cytoplasmic processes The response of the maximum specific growth rate to an increase in surface area should equal 0.8 ± 0.1 times that increase, whereas the response of the maximum non-specific growth rate (J) to such an increase should be 0.7 ± 0.1 times that increase (see Doc S2, Eqn 14)
Metabolic flow distribution as a function
of growth rate and glucose availability
We further verified our findings by determining the microbial physiology in the pH-auxostat and compar-ing the metabolic activity exhibited durcompar-ing selection for highest lmax with that in glucose-limited chemostat cultures By using both culture systems, we were able
to measure the metabolic flows of K marxianus as a function of the full range of growth rates under (defined) conditions of glucose limitation (chemostat) and substrate saturation (pH-auxostat) The specific cellular glucose and oxygen consumption rates, together with the specific protein and carbon dioxide production rates (qglu, qO2, qCO2 and qp, respectively) were determined for both culture systems Stable steady-state chemostat dilution rates (D = l) ranged from 0.05 to 0.55 h)1(Fig 3)
The specific glucose uptake rates (qglu) ranged from 0.7 to 6.0 mmolÆg)1Æh)1 for the lowest to the highest steady-state chemostat dilution rate, respectively In the pH-auxostat, stable glucose uptake rates started at approximately 7 mmolÆg)1Æh)1and increased in parallel with the increase in lmax to 9 mmolÆg)1Æh)1, i.e an increase of approximately 30% as already shown in Fig 1 for the pH-auxostat dilution rate The specific gas exchange rates (qCO2 and qO2) in the chemostat ranged from 2.3 to 16 mmolÆg)1Æh)1 During growth rate selec-tion in the pH-auxostat, these increased from 17–18 to
23 mmolÆg)1Æh)1; a 30% parallel increase All steady-state metabolic flows obtained before and after pH-auxostat selection were in line with the chemostat data, i.e all flows corresponded to linear extrapolations
Trang 7of the variation with specific growth rate observed in
the chemostat They remained fully coupled to the
spe-cific cellular growth rate of K marxianus The spespe-cific
proton production rate (qH þ), which equals the
ammo-nium uptake rate on defined medium (K marxianus
produced one proton per ammonium ion consumed),
was calculated for the auxostat system only The proton
stoichiometry was approximately 6 protons per gram
biomass As can be seen from Fig 3, much of the
ammonium consumed is incorporated into proteins
(qH þ⁄ qp= 1) The respiration quotient (RQ =
DCO2⁄ DO2) remained 1.0, indicating a fully oxidative
catabolism, confirming the Crabtree-negative
physiol-ogy of K marxianus, i.e no glucose fermentation to
ethanol (plus extra CO2production) even at ultra-high
glucose uptake rates (i.e under glucose saturation)
The carbon recovery during the sudden increase in
pH-auxostat dilution rate was and remained 100%
Both datasets indicate that no products other than
biomass and CO2were synthesized, confirming that the
specific flows of catabolism (qCO 2 and qO 2) and
ana-bolism (lmax) remained fully coupled during selection
for the highest possible growth rate under defined medium conditions
Discussion Background Our study focused on the highest possible growth rate
of microbial eukaryotes The main question was what limits cellular growth rate under nutrient-saturated defined medium conditions? As cellular metabolism is structurally organized into functional entities [66,67], our question was refined to what or which cellular reg-ulon, functional module, pathway or process step is insufficiently active and therefore responsible for growth limitation? As control of flux is distributed across various metabolic steps and hierarchical levels [68], it is difficult to find limitation in just one single pathway step; a ‘limitation’ may be an entire func-tional module of cell metabolism If such a controlling module could be localized at all, our further aim was
to quantify to what extent cellular growth rate was controlled by such a module
When K marxianus was cultivated under mineral nutrient-sufficient conditions with continuous selective pressure on the maximum specific growth rate, we observed a significant increase in lmax of approxi-mately 30% Although undefined rich media from cheap bulk waste streams are often used in industrial biotechnology, our approach of using mineral medium
is still relevant because less stable DNA vectors (pre-cious vectors in high-copy numbers) are readily lost after several generations of growth on rich media [15], due to the lack of selective markers K marxianus is used for the industrial production of commercially attractive proteins [4–6] To guarantee the sustainabil-ity of high-copy number vectors, more expensive selec-tive mineral media with substrate markers are used The drawback of using mineral medium is often a lower maximum growth rate, resulting in an increase
in cost-intensive bioreactor times To optimize the overall protein production process on mineral medium,
a higher maximum growth rate is called for Therefore, cell selection on mineral medium may help to increase the microbial productivity of industrial single-cell protein manufacturing As shown in our study, the pH-auxostat bioreactor can be used to select for cells with the potential to grow faster Stemmer [69] has developed a method for rapid evolution of a protein
in vitro by means of DNA shuffling Here, we showed
a rapid evolution of yeast cells in situ by pH-auxostat cultivation for more than 50 generations We call this
‘cellular selection’ Chemostats have also been shown
–1 · h
–1 )
Specific cellular growth rate (µ in g·g –1 · h –1 )
–1 ·g dry weight·h
qgas
qglu
qH+
Fig 3 Physiological properties of K marxianus under
glucose-lim-ited chemostat conditions (solid lines: 0 < l < 0.55 h)1) and during
substrate-saturated conditions in a long-term pH-auxostat (dashed
lines: 0.57 < l < 0.8 h)1) Specific carbon production rate (q CO 2 in
mmol per g dry weight per h; closed squares, chemostat; open
squares, pH-auxostat) Specific oxygen consumption rate (qO2 in
mmol per g dry weight per h; closed circles, chemostat; open
cir-cles, pH-auxostat) Specific glucose uptake rate (q glu in mmol per g
dry weight per h; closed diamonds, chemostat; open diamonds,
pH-auxostat) Specific protein biosynthesis rate (qpin g protein per
g dry weight per h; rightward-pointing open triangle, chemostat;
open triangle, pH-auxostat) Specific proton production rate (qHþ in
protons per gram dry weight per h; closed inverted triangle,
pH-auxostat; not determined using the chemostat).
Trang 8to be useful for the selection of microbes However,
selection is not for maximum growth rate alone;
selec-tion also occurs for an increased affinity for the
limit-ing substrate (1⁄ Ks), depending on the dilution rate
[51,52] In fact, selection in a chemostat often favors
the strain with the highest lmax⁄ Ksratio When
lower-ing bioreactor times on mineral medium, only the lmax
is of interest Therefore, selection using a pH-auxostat
is preferable to selection in a chemostat when an
increased growth rate on mineral medium is called for
However, the pH-auxostat can not be used for analysis
of growth-control on rich medium, as growth-associated
protons are not produced on rich medium, due to the
lack of sufficient proton-coupled ammonium uptake
(uptake of N-rich monomers instead of NH4+) In
addition to the pH-auxostat experiments presented
here, we used a CO2-auxostat to obtain steady-state
substrate-saturated (maximum) growth on rich
unde-fined medium [1] Under these conditions, we also
showed evolution towards a higher growth rate, but
this was not accompanied by an altered morphology
We think that moving from defined to rich medium
shifts the control from input processes to internal
pro-cesses, hence away from uptake This would explain
the observation but these results do not constitute
evidence for our contention that there is less control
by uptake when growth of K marxianus takes place in
rich medium Koebmann et al [70] found that the
prokaryotic growth rate is mainly controlled (> 70%)
by the demand for ATP By using Escherichia coli in
which intracellular ATP⁄ ADP levels could be
modu-lated, they showed that the majority of the control of
bacterial growth rate resides in anabolic reactions, i.e
cells growing on glucose-minimal medium are mostly
carbon-limited By quantifying the concomitant change
in the cell’s surface-to-volume ratio and maximum
growth rate, we showed that our results are consistent
with control of the growth rate of one of the
fastest-growing eukaryotes, K marxianus, mainly due to in
outer-membrane transport of carbon and⁄ or Gibbs
free-energy substrates
Highest eukaryotic growth rate
Our observation of microbial selection using an
auxo-stat also addressed the second issue of our study, i.e
whether one of the fastest-growing eukaryotes,
K marxianus, can grow even faster on defined mineral
medium The answer would appear to be yes The
average cell-cycle time of the faster-growing population
was 52 min, which is among the shortest steady-state
cell-cycle time of any eukaryotic organism on defined
glucose⁄ ammonium mineral medium, and is certainly
much shorter than that of the more minimalist Myco-plasma genitalium [60] Figure 1 shows that the steady-state pH-auxostat dilution rate (D) increased from one steady state to the new steady state, and lasted many generations
Importantly, an alternative scenario of a change in metabolism with an induced additional acid and CO2 production at constant specific growth rate is refuted
by our observations Here the special properties of the pH-auxostat [56] are important: at steady state, the dilution rate of the auxostat D equals the specific growth rate of the cells (l), and a change in the spe-cific rate of acid production at constant spespe-cific growth rate is reflected by a change in biomass density in the auxostat not by a change in dilution rate We did observe an increase in dilution rate from one steady state to a next, proving that there was an increase in specific growth rate In the case of a change in metab-olism at constant specific growth rate, enhanced acid production by K marxianus would have initiated with higher carbon dioxide production The semi-logarith-mic plot shown in the inset to Fig 1 would have shown an upward-shift in gas production with parallel straight slope indicating the same rate of the exponen-tional growth In addition to the theory and our obser-vations, there is ample evidence that this alternative scenario must be rejected For the two steady states,
we calculated 100% carbon recovery, indicating that
no carbon products (such as organic acids) were pro-duced other than biomass and CO2, confirming full oxidative metabolism of K marxianus during the entire experiment In Fig 3, all metabolic flows (including carbon dioxide production) are shown in terms of specific flow rates in mmolÆg)1 dry weight of biomass per hour, and all such flows were fully coupled to growth rate
Another alternative reason for the increase in dilu-tion rate, such as addidilu-tional wall growth inside the transparent fermenter vessel, was also rejected as no extreme amounts of biomass were observed If extreme amounts of biomass had been stacked inside the fer-menter vessel, wash-out of the entire culture would have take place Moreover, a fresh auxostat culture inoculated with the selected strain always started immediately at the elevated dilution rate, excluding an increase in dilution rate due to wall growth
The observed 30% increase in lmax from 0.6 to 0.8 h)1 (Fig 1) was irreversible in the sense that cells with the obtained higher growth rate did not return to the initial value upon injection into a fresh pH-auxo-stat This was also the case after frequent re-plating, batch cultivation or intermittent use of a glucose-lim-ited chemostat [1] at a sufficiently low dilution rate
Trang 9The initial value of lmax of 0.6 h)1 was the mean of
two experiments As can be seen in Fig 1, a difference
of 5% was observed between the initial lmax for the
two separate experiments As reported in Results, we
checked whether minor variations in additions of
vita-mins and essential minerals could have caused this, but
they did not affect the initial lmax Our working
expla-nation for this is the importance of the pre-induction
state of the cells; whether or not all enzymes of the
relevant pathways have already been induced depends
on the history of the cells put into culture As can also
be seen in Fig 1, after sufficient duration of
steady-state pH-auxostat cultivation enabling selection, this
apparent difference in lmax diminished, resulting in
one of the fastest possible growth rates for eukaryotic
life on mineral medium (0.8 h)1) This result answers
our third question: an increased specific growth rate
of 0.8 h)1 shown for K marxianus might be well the
fastest possible growth rate recorded thus far for
eukaryotic life on defined medium
Auxostat cultivation is not a standard microbial
method for selection of cells In contrast to serial batch
cultures or plate cultures, it allows long-term selection
for a higher growth rate under perfect steady-state
growth conditions [both in terms of the presence of
sufficient growth substrates (i.e S Kms, with Kmsas
the substrate concentration, S, at which the reaction
rate, V, is 0.5 Vmax) and absence of toxic products, P,
(i.e P > Kmp, with Kmpas the product concentration,
P, at which the reaction rate, V, is 0.5 Vmax)] The
intermittent stationary phases in serial batch cultures,
or the variations in substrates and products during
batch cultivation, would not provide the selective
steady-state conditions we required Chemostat
opera-tion would not specifically select for the highest
possi-ble growth rate either, but at most for a maximum
growth rate specifically obtained during
substrate-dependent, limiting cultivation in the chemostat, i.e at
which the chemostat method is unfortunately unstable
Therefore we used the pH-auxostat, in which cells
are able to grow as fast as they can under optimal
conditions for all medium components (all substrates
saturated and no toxic products)
Phenotypic adaptation or genetic evolution?
We tried to distinguish between adaptive and
non-adaptive [71] evolutionary changes in the traits of
K marxianusafter cultivation for more than 50
gener-ations at the highest growth rate Measurements of
cell-cycle times for individual cells within populations
of cells growing under steady-state conditions in
homogeneous environments revealed considerable
vari-ability Wheals and Lord [72] showed significant differ-ences in specific growth rates within a population of genetically identical (or very closely related) cells of
S cerevisiae Under pH-auxostat conditions, such clonal variability may, in principle, have been the basis
of selection for cells with a higher growth rate How-ever, the reason for this variability in the distribution
of cell-cycle times is still unclear Axelrod and Kuczek [73] have ascribed clonal differences in growth rate to potentially intrinsic, inheritable but non-genetic (epige-netic) differences between cells Variation in cell-cycle time has been ascribed to asymmetric partitioning of biosynthetic material (other than chromosomal), which may affect the rate at which cells traverse the cell cycle, or G1 in particular [74] Because of the epige-netic character of unequal partitioning, the value of the increased growth rate should return to the lower initial lmaxvalue in a non-selective environment due to the weaker cells (or relatively smaller daughter cells with lower growth rates) being retained in the popula-tion As shown in Fig 2A, the average relative sizes of the smallest two types of particles measured, assumed
to be daughter or mother cells, both increased to the same extent during selection Epigenetic selection for the biggest daughter cells at the time of separation due
to asymmetric partitioning may have occurred How-ever, in our additional experiments, the characteristics were inherited over more than 400 generations on plates subjected to repeated batch cultivation (as shown in the inset to Fig 1) and for 40 generations at
a rather low dilution rate (D = 0.2, i.e 25% of the selected lmax) in a glucose-limited chemostat [1] In view of the observed steadiness of the higher growth rate gained, it is unlikely that an increase due to the proposed epigenetic inheritance by asymmetric distri-bution of biosynthetic cell material (i.e on the basis of bud size at separation) was the cause of our observa-tions In closed systems such as batch culture and re-plating, newly born smaller cells did not reduce the higher average lmax value of the selected population Consequently, variation in daughter cell size could not account for the persistent increase in lmax
Simulation of the flow dynamics during the selection
in the pH-auxostat supported this reasoning Using from the hypothesis that our yeast population con-tained cells with a variety of growth rates normally distributed among the measured average growth rate
of 0.6 h)1(ranging from 0.5 to 0.7 h)1), and that these growth rates were inherited, produced a dilution pattern (see D in Fig 4A) that deviated from the experimental data shown in Fig 1 In this simulation, sub-populations of cells with higher maximum growth rates succeeded each other, and the sub-population
Trang 10with the highest lmax value ultimately predominates
over slower sub-populations Simulation of the dilution
rate (D) revealed a slow regular increase to a steady
state (see Fig 4A) without a pre-steady state of 60 h
at a lower value Therefore, the flow dynamics of the
selection of cells with a higher growth rate, distributed
around an average lmaxvalue, did not concur with the
experimental data as shown in Fig 1
Non-Mendelian inheritance of extra-genomic infor-mation by an ancestral RNA-sequence cache, as has been suggested by Lolle et al [75] for Arabidopsis thali-ana (and discussed in [76–78]), is also not a likely mechanism for the inheritance of a stable higher growth rate of K marxianus
Genetic variability as a more realistic explanation for our observed increase in growth rate was consid-ered In yeast, spontaneous mutations occur at a low frequency, approximately 10)4 to 10)8 per gene per generation [79] As the pH-auxostat population at the beginning of the first steady state derived from a single cell approximately 38 generations earlier, genetic heter-ogeneity must have existed Starting from 5000 yeast genes each undergoing spontaneous mutations at a rate of 10)7mutations per gene per generation, one in
50 cells should have been affected by a mutation after
38 generations (neglecting the effects of selection) If one key regulatory gene needs to be mutated to obtain
an increase in maximum growth rate, one in
250· 103cells should have a mutation in that gene If one in ten thousand mutations in that gene has a posi-tive effect, 4· 10)10cells out of a population should have such positive ‘upward’ mutation As shown in Fig 4B, simulation of the dynamics of the dilution rate and these numbers of wild-type and mutant cells concurred with the experimental data presented in Fig 1 In this simulation, we started with approxi-mately 2.5· 1011cells at a lmax of 0.6 h)1 and assumed that approximately 15 mutant cells (i.e a fraction of 0.6· 1010) were present at the start of the cultivation with an average lmax of 0.8 h)1 After 60 h
of pH-auxostat cultivation, the competition was com-pleted in favor of the faster-growing cells within a per-iod of approximately 40 h This corresponds to the experimental findings shown in Fig 1 Consequently, genetic diversity of K marxianus that arose as a result of
in spontaneous mutations at normal rates may well have been the cause of the increase in lmaxduring the long-term pH-auxostat cultivations Based on our calcula-tions of genetic variability and a simulation of the biore-actor flow dynamics, we conclude that the sudden increase in growth rate after 50 generations has been attained through mutation, rather than a slow epigenetic selection of faster-growing cells at the beginning of the culture
Why would such a faster-growing mutant selected
by our auxostat not already have appeared in nature
by natural selection? The answer could be that natu-ral surroundings change rapidly over time, causing the selection pressure to variate over time without allowing any selection of one particular microbial trait to occur Therefore, changing environments may
A
B
–1)
–1 )
Time, t (h)
Time, t (h)
Total amount of cells
–1) × 10
Fig 4 Simulations of the dilution rate (bold line, D) during
pH-auxo-stat cultivation (A) Each curve (dashed lines) represents the
num-ber of cells with one specific growth rate, indicated in the figure.
The simulation started with the assumption that the yeast
popula-tion contained cells with a variety of maximum growth rates, all
normally distributed around the measured average maximum
growth rate of 0.6 h)1, ranging from 0.5 to 0.7 h)1 All growth rates
ranging from 0.5 to 0.7 h)1with a standard deviation of 0.05 were
taken into account during simulation of D For clarity, only the
sub-populations of cells with growth rates of 0.59, 0.61, 0.64, 0.66,
0.69 and 0.71 h)1are visualized (B) Simulation of the dynamics of
the dilution rate (D) in the pH-auxostat during selection of cells
aris-ing by spontaneous mutation The simulation was started at time
t = 0 using approximately 2.5 · 1011 wild-type cells and 15
mutants with a lmaxof 0.6 and 0.8 h)1, respectively The dilution
rate (D, bold line) equals the steady-state average maximum
growth rate of the yeast population (l max ); open triangles, number
of cells with average lmax= 0.6 h)1; open squares, number of cells
with average lmax= 0.8 h)1.