Some of the work described in this chapter has been published
as:
Davey, C.L., Davey, H.M. and Kell, D.B. (1992) On the Dielectric
Properties of Cell Suspensions at High Volume Fractions. Bioelectrochem.
Bioenerg. 28, 219-340.
Davey, C.L., Davey, H.M., Kell, D.B. and Todd, R.W. (1993) An Introduction to the Dielectric Estimation of Cellular Biomass in Real Time, with Special Emphasis on Measurements at High Volume Fractions. Anal. Chim. Acta 279, 155-161.
In general there are two types of continuous culture, the chemostat
and the turbidostat (Anderson, 1956; Tempest, 1970). In the chemostat
the dilution rate is fixed by the experimenter and the growth
rate is nutrient-limited (Novick and Szilard, 1950; Herbert,
1958; Novick, 1958). In the turbidostat by comparison growth
is not nutrient-limited and the biomass level is constantly monitored
and only when it exceeds the setpoint does medium get pumped in
via a feed-back loop (Munson, 1970).
There are several advantages to the turbidostatic growth regime;
the volumetric productivity in biomass terms is higher than for
any other culture regime, the culture is stable at dilution rates
approaching mmax (Watson, 1972), and
of particular importance for the production of certain types of
improved strains the selection pressures in the turbidostat are
very high (Bryson and Szybalski, 1952). Usually selection will
be for the organism that will grow at the fastest rate in the
conditions provided (as the cells with slower growth rates will
tend to be washed out of the fermentor before they divide), but
conditions can be envisaged whereby other selection pressures
could be imposed. Since the turbidostat is by its very nature
a self-stabilising system it is possible to use selection for
organisms capable of growth in the presence of a toxic substance
(Aarnio et al., 1991).
Coupling of the selection pressures present in a turbidostat
with flow cytometric / cell sorter assays could enable rapid
isolation of strains exhibiting desired properties. For example,
mutants (or pre-existing subpopulations) selected in a turbidostat
would normally be assayed only for their final productivity of
the substance of interest when they were subsequently grown in
batch. Such assays are incapable of identifying any cells present
which produce high levels of the product much earlier than average
in the fermentation. If a non-lethal probe for the substance of
interest was developed then a cell sorter could be used to isolate
the cells showing the desired characteristics, these cells could
then be propagated to form a new "rapid-production"
strain.
Despite the advantages offered by a turbidostatically controlled
fermentation the method remains under-exploited in microbial physiology,
with the less-stable chemostat being the preferred option. There
are several problems associated with turbidostatic control that
may in part account for this (Martin and Hempfling, 1976). In
conventional turbidostats, as the name suggests, the biomass estimations
for feed-back control are determined by turbidity (optical density)
measurements (Myers and Clark, 1944). There are several disadvantages
associated with this method, the most important of which is that
optical density is only linear with biomass over a very narrow
range of low biomass concentrations, and for most organisms fails
to be linear at concentrations well below those which would be
of interest to industry. In addition such measurements are extremely
prone to sensor fouling by the microorganisms that they attempt
to measure due to biofilm formation on the walls of the fermentor
through which the optical density measurements are made (Anderson,
1953; Northrop, 1954; Watson, 1969). A third problem with turbidity
measurements is that it is not only biomass that is measured,
as necromass, particulate solids and gas bubbles will all
contribute to the optical density to some extent. For these reasons
novel methods for the on-line and real-time measurement of the
biomass content of industrial fermentations has long been an area
of interest (Harris and Kell, 1985; Kell et al., 1990;
Sonnleitner et al., 1992).
An instrument that monitors biomass via the radio-frequency
electrical capacitance of the cell suspension has been developed
and commercialised by Aber Instruments Ltd. (Science Park, Cefn
Llan, Aberystwyth, SY23 3AH, U.K.) and is called the Biomass Monitor
(Harris et al., 1987; Kell et al., 1987; Davey,
1993a,b). This chapter describes how the Biomass Monitor can be
used to measure biomass and how it can deal with most of the problems
encountered with turbidity measurements. In addition it is shown
that the method is linear to biomass concentrations far in excess
of those that would be found in most fermentations.
During the permittistatically controlled fermentations described herein, oscillations in the growth rate of the yeast cultures were observed, and these were investigated further. It is shown that whilst the detailed origin of these oscillations may not be easily explained, they appear to be linked to the central metabolic pathways of the cell.
When an electric field is applied to an ionic solution the ions
in that solution are forced to move (Figure 6.1). The positively
charged ions are pushed in the direction of the field while the
negatively charged ions are pushed in the opposite direction.
If cells are added to the ionic solution then the situation shown
in Figure 6.2 is seen. As before the ions in the solution are
forced to move, however as can be seen from the figure, many of
the ions both inside and outside of the cells can move only so
far before they bump into the cell's plasma membranes and are
prevented from moving any further. The result of this is that
there develops a charge separation or polarisation at the poles
of the cells. The extent of the field-induced charge separations
is measured by the capacitance (measured in Farads, F) of the
suspension. As the volume fraction of cells increases the amount
of membranes polarised increases and so the capacitance of the
suspension increases. Thus by the measurement of the capacitance
of the suspension, one can monitor its biomass content (Harris
et al., 1987; Davey, 1993a, b).
So far the field has only been shown going in one direction.
Of course one can reverse the field direction (Figure 6.3) and
if one does this then, although the polarity of the charge separations
has been reversed, their magnitude (and hence the capacitance
of the suspension) remains unchanged. One can also change the
rate at which the field changes direction. The number of
times the field changes direction per second is measured by its
frequency (Hz). Frequency has a marked effect on the capacitance
of a cell suspension, because a finite time is required for the
charge separations to be
Figure 6.1 : When an electric field is applied to an ionic
solution such as aqueous potassium chloride the ions in the solution
are forced to move. The positively charged potassium ions move
in the direction of the field, while the negatively charged chloride
ions move in the opposite direction.
Figure 6.2 : The effect of an applied electric field on
a cell suspension. The ions in the suspending medium and in the
cell cytoplasm can move only so far before they become trapped
by the cell membranes. The resulting charge separations can be
measured as the capacitance of the suspension.
Figure 6.3 : The effect of changing the direction of the
field. Although the charge separations are reversed by changing
the direction of the field, their magnitude, and hence
the capacitance of the suspension, remains unchanged.induced (Pethig,
1979; Foster and Schwan, 1986; Pethig and Kell, 1987). Figure
6.4 shows the typical polarisation induced across a cell as the
frequency of the electric field is increased (Figure 6.4AC).
Also shown is how the capacitance of the suspension changes with
frequency (Figure 6.4D). At low frequencies (Figure 6.4A) a lot
of ions have time to reach the plasma membranes and polarise them
before the field changes direction and moves the ions the opposite
way. Thus the capacitance of the cell suspension is high. At higher
frequencies (Figure 6.4B) fewer ions have time to reach the membranes
and so the extent of the induced transmembrane polarisation is
less and the capacitance of the suspension is also lower. At very
high frequencies (Figure 6.4C) very few ions have time to move
to, and polarise the membranes before the field changes direction,
and so the induced membrane polarisation is very small. At these
frequencies the cell's contribution to the capacitance of the
suspension is very small and one just measures the background
capacitance of the medium (which is mainly due to water dipoles).
From Figure 6.4D it is seen that the capacitance of the suspension goes from a high capacitance plateau at low frequencies (maximal cell polarisation) to a low capacitance plateau at high frequencies (minimal cell polarisation). This fall in the capacitance of a suspension due to the loss of induced membrane polarisation with increasing frequency is called the bdispersion (Pethig, 1979; Foster and Schwan, 1986; Pethig and Kell, 1987; Davey and Kell, 1994). The residual high-frequency capacitance due to the medium is called C¥ and the height of the low-frequency plateau above this is the DC (capacitance increment) of the bdispersion. The frequency when the fall in capacitance is half completed (i.e. the frequency when capacitance equals C¥+(DC / 2) is called the critical frequency (fc). The Cole-Cole a is related to the steepness of the fall in capacitance with increasing frequency.
As one is aiming to measure the biomass content of a cell suspension
then one needs to see what effect this has on the bdispersion
curve shown in Figure 6.4D. Figure
Figure 6.4 : Polarisation of the cell membranes involves
movement of ions and so it takes a finite time for charge separations
to be induced. At low frequencies (A) the cells are maximally
polarised, at higher frequencies (B) there are fewer charge
separations, and at much higher frequencies (C) the cells
are not polarised at all. The curve that describes this relationship
is known as the bdispersion (D).
For clarity in A-C the charges are shown for just
one field direction. The annotations are explained in the text.6.5
shows the bdispersion of hypothetical
cell suspensions with different biomass contents. The figure shows
that the fc is not changed by the biomass content, nor is C¥.
What does change as a function of biomass content is the magnitude
of DC, which increases with the biomass.
Thus the problem of measuring the biomass content of a cell
suspension reduces to one of measuring the magnitude of the DC
of the bdispersion .
One of the problems associated with using optical density for
the control of continuous cultures is that dead cells, gas bubbles
and non-biomass solids all interfere with the measurements so
their effects on the dielectric method of biomass estimation must
be considered. If the cells in a suspension have badly ruptured
plasma membranes (Figure 6.6) then the ions in the solution just
move through the holes in the membranes and fail to cause polarisation
of the membranes even at low frequencies. Thus dead cells do not
have a significant bdispersion
(i.e. a DC). If one uses the DC
of the bdispersion rather than
the more conventional turbidity measurements, to measure the biomass
content of a fermentor one can expect to measure only cells with
intact plasma membranes (i.e. living cells) as opposed to ruptured
(dead) cells (Stoicheva et al., 1989; Davey et al.,
1993b).
If non-biomass solids, oil droplets or gas bubbles are present
in the medium then their effect on the DC
of the bdispersion must be considered.
The ions in a growth medium either travel straight through the
non-biomass material if it is permeable to ions or just move round
it if it is not (Figure 6.7). In neither case are significant
charge polarisations induced (since there are no plasma membranes)
and so these materials will not produce a significant DC
term. Thus non-biomass materials are not expected to contribute
to the DC of the bdispersion
in a real fermentation medium.
Turbidity measurements often fail as a result of sensor fouling.
With the Biomass Monitor this problem is solved by use of electrolytic
cleaning pulses which may be applied to the electrodes in situ.
It should also be noted that under certain circumstances the electrodes
themselves can display a substantial, and frequency-
Figure 6.5 : The effect of increasing biomass concentrations
(in mg wet weight.ml1) on the bdispersion
of a typical cell suspension.
Figure 6.6 : When cells die their membranes rapidly break down, allowing ions to move freely. As no charge separations are induced there is a much reduced capacitance signal and hence the Biomass Monitor does not, unlike some techniques of biomass measurement, mistakenly record necromass as biomass.
Figure 6.7 : Charge separations are not induced at the surfaces of non-biomass materials as they lack a plasma membrane. Thus air-bubbles and particulate matter present in fermentation broth do not contribute significantly to the capacitance of the suspension.
dependant capacitance; artefacts of this type are minimised by
using a 4terminal electrode system such as that in the Aber
Instruments Biomass Monitor (Kell et al., 1987).
It was stated earlier that biomass measurement reduces to one
of the estimation of the DC of the
bdispersion. Thus one needs a
convenient means of measuring DC during
a fermentation. There are two ways of achieving this. Figure 6.4D
showed a bdispersion and marked
on it were two spot frequencies labelled flow and fhigh.
The capacitance at fhigh is approximately equal to C¥
whilst that at flow approximately equals (C¥
+ DC). Thus one can see that if one
measures the capacitance at fhigh and flow "simultaneously"
(or more practically in rapid succession), and then subtracts
the capacitance at fhigh (C¥)
from that at flow (DC + C¥)
one gets DC, and thus a measure of
biomass concentration. This is the principle of dual-frequency
biomass measurements. The second method of estimating DC,
and hence biomass concentration, uses the capacitance at flow
alone. At zero biomass concentration the capacitance at flow
equals C¥ (see Figure 6.5); thus
one can measure the capacitance of the medium at flow prior
to inoculation and then back off this capacitance to zero (i.e.
set C¥ to zero). This means that
any change in capacitance at flow during a fermentation
must reflect changes in DC and hence
biomass concentration. This is how single-frequency biomass measurements
work. For the present work dual-frequency measurements were used
throughout as these may be expected to be more stable to long-term
instrumental drift (Davey, 1993a) which occur within the time
scale (typically 2 months) of the continuous culture fermentations
that were carried out (since any changes that may occur will tend
to affect measurements at both frequencies to a similar degree).
Dual- and single-frequency measurements using a Biomass Monitor
(BM) have been successfully used to monitor the biomass concentrations
in a variety of systems (Kell et al., 1987). These include
bacterial and yeast cultures (e.g. (Harris et al., 1987;
Boulton et al., 1989; Ferris et al., 1990)), bacterial
biofilms (Markx and Kell, 1990), cultured cells (Markx et al.,
1991c, d), human blood (Beving et al., 1994), immobilised
cells (Salter et al., 1990) and filamentous cells in liquid
and solid substrate fermentations (Davey et al., 1991;
Penaloza et al., 1991; Fehrenbach et al., 1992;
Penaloza et al., 1992). Using a different instrument Mishima
and colleagues (Mishima et al., 1991a, b) have used 2terminal
capacitance measurements as a method for biomass estimation.
To relate the DC of the bdispersion
to the structure of the cells in the suspension giving rise to
it, one cannot work easily with capacitances. The reason for this
is that capacitance is a macroscopic measurement that depends
on the geometry of the electrodes used. For biomass measurements
this is fine because the electrode geometry remains constant.
However for physical calculations the need to adjust the capacitances
read to allow for the electrode geometry is an inconvenience.
Thus one requires a way of quoting capacitances so that they are
independent of the geometry of the electrodes. To do this one
converts capacitance to relative permittivity (e').
Just like capacitance, e' gives a measure
of the extent of the polarisations induced in a material by the
application of an electric field. The capacitance (C in Farads)
of a material is related to the equivalent relative permittivity
(e', unitless) by:
. . .(6.1)
where eo is the permittivity of free
space and is a constant with a value of 8.854 x 1012 F.m1
and k is known as the cell constant and has units of m1,
reflecting the geometry of the electrodes. The important point
is that electrodes of different geometries (different k values)
will record different capacitances for the same cell suspension,
but will all produce the same e' value.
From Equation 6.1 it is clear that the relative permittivity
of a material (e.g. a suspension) is equal to the capacitance
multiplied by a constant for a given electrode system of (k/eo).
Taking this into account then a bdispersion
plot like Figure 6.4 is reflected in the permittivity plot of
Figure 6.8. DC becomes De'
(dielectric increment) and C¥
becomes e'¥;
all that has happened is that the capacitance terms have been
multiplied by (k/eo) to convert them
to e' values. The fc is the same in
both plots. The formula (Schwan, 1957) that relates the magnitude
of De' (and hence the DC)
to the properties of the biological cells in the suspension is:
. . .(6.2)
where P (which is unitless) is the volume fraction of cells present
(i.e. the volume of material bounded by a plasma membrane, per
unit volume of suspension), r is the radius of the nominally spherical
cells from the cell centre to the plasma membrane (in metres)
and Cm is the plasma membrane capacitance per unit of membrane
area (F.m2). Cm gives a measure of the ability of the plasma
membrane to store charge and, for biological systems, typically
has a value of 0.01 F.m2.
For a given cell suspension, the mean radius and Cm are constant
and eo is a physical constant. Thus
a plot of De' versus cellular volume
fraction is a straight line of gradient (9rCm/4eo).
This relationship explains why DC is
linearly proportional to biomass concentration.
Equation 6.2 holds only for "low" volume fractions,
in which the electric field impinging on a given cell in the suspension
has not been distorted by the cells around it. At high volume
fractions this does not apply, and a plot of De'
versus volume fraction begins noticeably to plateau out
(at say P > 0.15, i.e. approximately 150 mg wet weight.ml1).
In most fermentations the linear approximation to the true relationship
between the permittivity increment and volume fraction (Equation
6.2) is adequate as the biomass concentrations encountered are
well below the critical level. Situations where the volume fraction
may be in the noticeably non-linear range include some
Figure 6.8. The data of Figure 6.4 converted to e' using Equation 6.1. The annotations on the figure are discussed in the text.
Escherichia coli fermentations (Riesenberg, 1991; Riesenberg
et al., 1991; Dubach and Märkl, 1992; O'Conner et
al., 1992), yeast pitching for fermentor inoculation in the
brewing industry, post-fermentation processing of bakers yeast,
and the latter part of some fed-batch fermentations. In order
that the dielectric method of biomass estimation be generally
applicable to all fermentations including those with high volume
fractions one needs to be able to model and compensate for this
loss of linearity.
Schwan and Morowitz (1962) suggested a modification to Equation
6.2 that allowed for this non-linearity at high volume fractions
(P):
. . .(6.3)
This is Equation 6.2 with the additional term 1/(1+(P/2))2 that
models the non-linearity at high values of P. This additional
factor depends only on the volume fraction of cells present and
is independent of the cell radius.
Although the modified equation is potentially of great use, prior
to the work described here, and published as (Davey et al.,
1992; Davey et al., 1993a), it had never been checked
thoroughly with real cells. By making careful and independent
measurements of P (by two different methods), r and De'
for a number of bacterial and yeast suspensions, it is shown herein
that this equation does indeed hold true over a wide range of
cell sizes and volume fractions. In addition it is shown that
the dielectric method of biomass estimation is a good method for
the control of the "turbidostat-type" continuous cultures
which, since they control the biomass via the permittivity
of the cell suspension, have been called the permittistat (Markx
et al., 1991a).
Three industrial strains of the yeast Saccharomyces cerevisiae
were used. Two of the strains were obtained as cell pastes; these
were Distillers Company Limited (DCL) "High Activity Baker's
Yeast" and DCL "Distillery Yeast". The third yeast
(BB11) was obtained as a pure isolate from the brewing industry.
BB11 was grown to a high biomass in a batch fermentation in a
medium consisting of 1.3% Ebroth and 5% malt extract (both
from Lab M) at a starting pH of 6.0. The non-baffled fermentor
had a working volume of 5 litres, the medium was not sparged with
air but was stirred gently (75 rev.min1) to keep the contents
in suspension. The temperature was held at 30oC throughout the
22 hour growth period; the pH was not controlled. After this growth
period the cells had reached stationary phase (5% budding) and
were harvested by centrifugation.
In addition the bacterium Micrococcus luteus Fleming strain
2665 was used. This was grown in batch culture in 5 litres of
Ebroth (pH 7.4). The medium was gently sparged with air
and the stirrer speed was set at 270 rev.min1 (with baffles
present). The temperature and pH were controlled at 30oC and 7.4
respectively and the foam was regulated with Sigma B silicone
antifoam. After 34 hours of growth the fermentor contents were
harvested as described for the BB11 yeast.
These organisms were chosen because they provided a broad range
of cell sizes and because both yeast and coccoid bacteria are
approximately spherical. The spherical nature of the organisms
meant that one could apply Equations 6.2 and 6.3 to the data with
confidence, as they were derived for spherical objects.
All of the cell types were washed and resuspended in the same
suspension buffer (SB). This contained 50mM KH2PO4 and 5 mM MgSO4.7H2O
and was adjusted to pH 7.0 with 5M KOH. SB was made up fresh for
each experiment, all reagents were from BDH and the water was
singly distilled in all glass apparatus.
The pellet of cells either harvested from batch culture or cut
from a block of paste, was suspended in at least three times its
own volume of SB. The resulting suspension was centrifuged at
1000 g (yeast) or 2750 g (M. luteus) for 10 minutes at
room temperature. The supernatant and any precipitated growth
medium components from the top of the pellet were discarded and
the cells were resuspended in fresh SB. These were then left at
room temperature for 45 minutes, with occasional mixing, to allow
the cells to equilibrate with the medium and to allow any limited
growth triggered by the resuspension of the cells to occur. The
cells were recentrifuged as described above and the supernatant
was discarded. A final wash was done in suspension buffer followed
by another centrifugation to obtain a pellet. The pellet was then
left slightly inclined for a while to allow any excess buffer
to run off. In the case of M. luteus a considerable amount
of buffer still remained and so for the very high volume fraction
measurements, a sample of the pellet was further centrifuged at
13000 rev.min1 in a bench Eppendorf-type centrifuge for
30 minutes and any supernatant was removed.
All of the cell suspensions were analysed using the Aber Instruments
Biomass Monitor. The machine was switched on at least 30 minutes
prior to use. The electrode used was a "Mexican Hat"
type and consisted of an epoxy resin body with a 1 ml sample
cavity, which contained four solid, 24carat gold electrode
pins. The electrode was thoroughly washed prior to each experimental
series with water then detergent then water followed by neat ethanol
then distilled water and finally by applying the electrolytic
cleaning pulses to the electrode while it contained dilute aqueous
KCl. After cleaning, the electrodes were left for at least 30
minutes in the KCl to stabilise prior to use. The electrode was
never allowed to dry out during the experiments (to prevent cells
or salts coating the pins) and it was well rinsed with distilled
water between each sample. The cell constant of the electrode
was calculated using Equation 6.4 by measuring the conductance
of 10 mM KCl (at 0.48 MHz) at a known temperature (and hence
a known conductivity).
. . .(6.4)
where k is the cell constant of the electrode and has units of
m1, s' is the conductivity of
the KCl and G is the conductance of the KCl measured using the
Biomass Monitor. The Biomass Monitor can be switched between two
sensitivity ranges known as High range and Low range, this switching
is reflected in the cell constant. The dielectric measurements
of the three yeast strains were all carried out in High range
(Cell Constant = 1.8 cm1) and the Micrococcus
luteus measurements were carried out in Low range (Cell Constant = 0.9 cm1).
The electrode was always filled to the same volume (1 ml)
during the experiment to prevent error due to any depth-dependency
of the cell constant.
Each cell suspension was scanned at ten frequencies between 0.2
and 10 MHz by Dr. C. Davey. The frequencies were chosen to
give an even spacing on a log frequency (Hertz) scale. The order
in which these frequencies were used in the scans was chosen at
random. For the two DCL yeasts the frequency scans were carried
out by manually adjusting the frequencies and noting the capacitance
and conductance values. A 1 second low pass filter was used to
remove any noise on the signal and the readings were taken once
the signal had stabilised at its new value. In these cases the
Biomass Monitor's capacitance at 0.4 MHz was zeroed in the
suspension buffer prior to the experiment. For the BB11 yeast
and the M. luteus the scans were carried out under full
computer control using a program called "MINISCAN",
written inhouse in Microsoft QuickBASIC v4.5 by Dr. C. Davey.
This time the Biomass Monitor was left in absolute capacitance
mode (i.e. the background capacitance of the medium was not backed
off to zero) and ten replicate readings at 0.1 second intervals
were taken at each frequency and averaged. The capacitance and
conductance data were converted to their equivalent permittivities
and conductivities using Equations 6.1 and 6.4 as they were recorded.
The equivalent conversions for the manually scanned yeast data
were done in a spreadsheet (VPPlanner) template.
For each cell suspension that was scanned as described above,
a polarisation control was done by me. This entailed adjusting
a 1 ml sample of the suspension buffer to the same conductance
(at 0.2 MHz.) as the cell suspension of interest. This was then
scanned in the same way as the cell suspensions. The polarisation
controls were then subtracted from their equivalent cell suspension
data sets to give the final dielectric data (Davey et al.,
1990b; Ferris et al., 1990). This process has the result
of setting e¥ to a value of approximately
zero.
For each cell type a series of dilutions in SB of the prepared
cell pellet was made and each was scanned as above. In the case
of the "solid" samples the paste was carefully pressed
into the 1 ml electrode cavity taking care not to bend the
electrode pins. As these samples were not easily amenable to volume
fraction measurements the dry weight of all of the suspensions
scanned were measured. Volume fraction measurements were then
done on the more dilute suspensions and a calibration curve of
these measurements versus their equivalent dry weights
was used for the estimation of the volume fractions of the more
concentrated suspensions.
For the "solid" suspensions the contents of the electrode
cavity were washed into a known volume of SB prior to dry weight
measurements. For the more dilute suspensions the electrode contents
were diluted in SB prior to dry weight analysis. A pre-weighed
25 mm diameter, 0.2 mm pore
size, Whatman filter (WCN type, cellulose nitrate, plain white)
was wetted with distilled water and placed under gentle vacuum.
A sample (typically 1 ml) of the diluted cell suspension
was then pipetted onto the filter and sucked "dry".
The cell pellet was then washed with 0.5 - 1 ml of distilled
water and again sucked "dry". The filter was then reweighed
to give the wet weight of cells per millilitre. The filters and
yeast were then dried overnight at 100oC before the dry weight
was determined.
For the more dilute suspensions of each cell type, volume fraction
measurements were performed using conductivity measurements made
using an EDT RE 387 Series 3 microprocessor conductivity meter.
The electrode cell had two platinum-blacked electrodes. Under
the conditions used here the operating frequency was 300 Hz
and the excitation voltage was 150 mV. The electrode had
been modified to work with a constant volume of 1.5 ml. The
cell constant of this arrangement as calculated for the Biomass
Monitor above was 1.1 cm1. All measurements were performed
with the machine's temperature compensation feature disabled owing
to the unknown temperature coefficients of the samples used. A
sample of the cell suspension was measured and then immediately
centrifuged for 5 minutes at 13000 rev.min1 in
an Eppendorf-type bench centrifuge, the conductivity of the resulting
supernatant was then measured. The volume fraction was then calculated
using the Bruggeman equation (Bruggeman, 1935):
. . .(6.5)
where s'l is the (low frequency) conductivity
of the cell suspension, s'o is the
conductivity of the suspending medium and P is the volume fraction
of cells present.
For the yeast samples volume fraction measurements were also
determined using calculations based on the numbers of cells per
millilitre (from haemocytometer counts) and the mean volumes of
the cells which were calculated by measuring the long and short
axes of the cells from photomicrographs as described in Chapter
2. The volumes of the cells were calculated using Equation 6.6.
. . .(6.6)
where V is the volume, a is the long semi-axis and b is the short
semi-axis of the cell. The volume fractions of the micrococci
samples were not calculated in this way as the haemocytometer
counts proved unreliable for such small cells and was complicated
by the fact that the cells tended to aggregate into clumps.
A yeast clone was isolated from baker's yeast obtained locally,
by repeated streaking onto agar plates followed by selection of
a single colony for growth in broth. The medium used was YPG which
contained (all w/v): glucose (BDH) 5%, yeast extract (Oxoid) 0.5%
and bacteriological peptone (Oxoid) 0.5%. The pH was set at 4.5
prior to autoclaving for 15 minutes at 121oC. Solid plates were
prepared by adding 1.5% Lab M agar to the same medium save that
the pH was set to 5.0. Plates and broth cultures were incubated
at 30oC.
Batch culture of Saccharomyces cerevisiae was carried
out in a 7 litre fermentor (FT Applikon Ltd., Station Drive, Bredon,
Tewkesbury, Gloucestershire) with a working volume of 5 litres.
The YPG medium described above was placed into the fermentor prior
to autoclaving for 20 minutes at 121oC. Silicone antifoam (Antifoam
B, Sigma, Poole, Dorset), diluted 1 in 10, was added as required
to control foaming. The fermentor was aerated with wetted air
and stirred at 700 rpm with a top-down propeller. The temperature
was controlled at 30oC by placing the fermentor in a water bath
and by means of a circulation pump which pumped warm water from
the water bath through pipes inside the fermentor.
Dielectric scans of the fermentor contents were carried out either
on-line in the fermentor using a long fermentor probe inserted
through the top of the fermentor or off-line in a Mexican hat
probe. Where off-line scans were used polarisation controls were
done using supernatant from fermentor samples diluted as appropriate
with distilled water to give conductances equivalent to those
in the cell suspensions (see above). Samples were withdrawn from
the fermentor at approximately 1-1.5 hour intervals for the determination
of pH, viability, budding index, wet weight, dry weight and total
counts. Flow cytometric determination of cell size and DNA content
was carried out as described in Chapters 2 and 3.
Wet and dry weights of the yeast samples were determined as described
above for the high volume fraction work, save that the samples
were dried using a Sartorius drying machine set at 80oC using
the auto-shutoff feature which dries the sample until no further
weight change is detected. Drying the yeast samples in this way
took about 10 minutes. The offline pH was measured using an Orion
Research (model 701A) pH meter that had been calibrated on the
day of the experiment with appropriate buffers.
Flow Cytometry was carried out using a Skatron Argus 100 flow
cytometer (Skatron Limited, PO Box 34, Newmarket, Suffolk) as
described by Steen and colleagues (Steen and Boye, 1980 or 1981;
Steen et al., 1982; Boye et al., 1983; Skarstad
et al., 1985; Steen and Lindmo, 1985) and elsewhere in
this thesis. The flow cytometer was set up as described in the
manufacturer's manual, save that an additional 0.1 mm
filter was placed in the sheath fluid line. The sheath fluid was
prepared from water that had been filtered through a 0.2 mm
filter using Millipore Milli-Q apparatus. 1 mM sodium azide was
added to the sheath fluid to prevent microbial growth in the sheath
fluid tank.
Cell size distributions of unfixed samples were estimated as
described in Chapter 2 (Davey et al., 1993c) based
on an initial calibration of the forward light scatter with latex
beads of known diameter and the application of the additional
calibration factor that describes the differences between light
scattering properties of beads and cells. The calibration and
plotting of the data were carried out using the SkatGraf package
described in Chapter 5.
Prior to DNA staining, the cells were fixed by sparging into
ethanol to give a final ethanol concentration of 70% (v/v). Cell
samples fixed in this way were then stored at 4oC for upto 3 months
prior to analysis by flow cytometry. Cells were removed from the
fixative by centrifugation and the pellet was washed twice in
50 mM phosphate buffer prior to resuspension in the staining solution
which contained 90 mg.ml-1 mithramycin
and 10 mg.ml-1 ethidium bromide in
25% aqueous ethanol. After 20 minutes incubation the samples were
examined using the B1 filter block supplied with the instrument
(optical characteristics: excitation 395-440; band stop 460; emission
>470).
Viability was estimated by staining the (dead) cells with methylene
blue. The methylene blue stain was prepared in three parts. Part
A was prepared by dissolving 0.23 g of methylene blue powder
(Sigma) in 20 ml of distilled water; part B was prepared by dissolving
27.22 g of KH2PO4 in 1 litre of distilled water; part C contained
0.284 g Na2HPO4 in 10 ml of distilled water. The stain
was then prepared from these three components by combining 20 ml
of A with 997 ml of B and 2.5 ml C. To ascertain the viability
of the sample the yeast suspension, methylene blue and phosphate
buffer (or fresh growth medium) were usually combined in the ratio
1 : 4 : 40, but for the very dilute samples
at the start of the batch fermentation the amount of cells added
was increased (and the amount of buffer was decreased accordingly).
The prepared samples were then examined immediately by light microscopy
at a magnification of x400. Cells appearing deep blue were scored
as dead while unstained cells and slightly coloured (grey) cells
were scored as alive (Davey et al., 1993b). At least 100
cells were scored for each sample and the percentage of viable
cells was calculated.
The percentage of cells with buds was determined by light microscopy
by scoring at least 100 cells. The cells examined were either
unstained or were the sample stained with methylene blue. Total
counts were determined by loading an unstained cell sample (diluted
as appropriate) into a haemocytometer and counting the cells within
a portion of the grid at a magnification of x400.
A yeast clone was isolated from bakers yeast by repeated streaking
onto agar, followed by growth in broth. The yeast was grown on
the YPG medium described above. For maintenance of the culture,
agar plates were prepared by the addition of 1.5% Lab M agar to
a variation of the medium which differed only in that it lacked
glucose. These plates were stored at 4oC for up to six weeks.
For permittistatic culture the yeast were grown in a 1 litre
fermentor (LH-Engineering) with a working volume of 750 ml,
the top of the fermentor was adapted to allow the insertion of
the Biomass Monitor's probe (standard 25 mm probe, cell constant
1.18 cm1). The temperature was controlled at 30oC with
a LH503 temperature controller and the pH was controlled
at 4.5 with a LH505 pH controller using 2M KOH and 2M HCl.
Filtered and wetted air was pumped through the fermentor at a
rate of approximately 1.5 volumes.min1, and the contents
of the fermentor were stirred at 450 rev.min1. The
Biomass Monitor was interfaced, via a set of amplifiers,
to a 386SX IBM-compatible computer containing a DT2811PGH
12bit analog / digital I/O board (Data Translation Ltd.
Wokingham, Berkshire, U.K.).
A program (PERMSTAT.EXE) was designed and written by me (in Microsoftâ
QuickBASIC v4.5) to run in the background under Microsoftâ
Windowsä. In brief, the program
set the measurement frequencies (0.4 MHz and 9.5 MHz) on
the Biomass Monitor, read back the resulting capacitances, calculated
the difference between the capacitances (DC),
and switched the medium pump on and off appropriately to control
the biomass level. Each time the pump was switched off the "pump-on
time" was recorded to a disk, thus allowing the volume of
medium required to correct the biomass concentration to be calculated
from the pump rate. Prior to commencing the permittistat experiments
the pump rate was determined by measuring the volume of liquid
pumped during a known time period. The pump rate was found to
be 21.5 ml.min1 and was constant over several repeat
measurements. At user-defined time intervals the low- and high-frequency
capacitances and conductances were also stored to a file. The
PERMSTAT program consists of some 1500 lines of code, and when
compiled occupies approximately 75 kilobytes of disk space.
Samples were removed from the permittistat periodically for the
determination of wet weight, dry weight, viability, budding index
and total counts using the methods described above. Flow cytometry
was carried out to determine the size and DNA content of the cells
in the sample as described above. Additional samples were removed
for determination of ethanol and glucose concentrations.
Samples for determination of ethanol concentration were placed
into Eppendorf tubes and centrifuged at 13000 rpm for 5 minutes.
The supernatants were removed and frozen until the day of assay.
Ethanol concentrations were then determined using Sigma Procedure
332-UV. This assay involves the enzyme alcohol dehydrogenase (ADH)
which catalyses the oxidation of alcohol to acetaldehyde with
the simultaneous reduction of NAD to NADH. This leads to an increase
in absorbance at 340 nm that is directly proportional to
the concentration of alcohol in the sample. The thawed supernatants
(diluted as appropriate) were added to the NAD-ADH single assay
vials together with glycine buffer (pH 9) and incubated at room
temperature for 10 minutes. The samples were then transferred
to plastic cuvettes and their absorbance at 340 nm was measured
versus a blank which contained distilled water in place
of the supernatant. Samples of known ethanol concentration were
measured in the same manner and the concentration of ethanol in
the supernatants was determined.
Samples for determination of glucose concentration were centrifuged
and stored as described for the ethanol determination above. Glucose
concentration was then determined using Sigma Procedure number
510. This assay is based on the conversion of glucose to gluconic
acid by glucose oxidase, a reaction that also produces hydrogen
peroxide. The hydrogen peroxide then oxidises colourless o-Dianisidine
to the brown oxidised form. The intensity of the brown coloration
after 45 minutes incubation at room temperature (which is proportional
to the glucose concentration in the original sample) is then measured
versus a blank at 450 nm. Samples of known glucose concentration
were measured in the same manner and the concentration of glucose
in the supernatants was determined.
Figure 6.9 compares the two methods of determining the volume
fraction (P) for the three yeast strains. It can be seen that
the two methods give very similar results. However, of the two
methods the systematic errors in the haemocytometer counts are
likely to be the more significant. Workers using conductimetry
to measure haematocrit, using methods and equations very similar
to those used here to measure P, have also encountered this problem
(Okada and Schwan, 1960). Several careful pieces of work have
shown that the errors in the haemocytometer measurements are large
compared with those from conductimetry. A further point is that
the conductivity-based measurement of volume fraction measures
only the protoplast inside the cell wall. The estimates of volume
fraction relying on the measurements of cell number and radius
also include the cell walls in the cell volumes. This may in part
explain why the slopes of the fitted lines on Figure 6.9 are all
slightly less than unity. For this reason the volume fractions
measured by conductivity were used for the dielectric calculations.
Figure 6.9 : Relationship between volume fractions estimated conductimetrically and microscopically. Measurements were made as described in the Materials and Methods section. The lines represent the best fits to the data: l DCL baker's yeast, n DCL distillery yeast, s BB11 yeast.
Figure 6.10 shows the relationship between De' and the volume fraction determined conductimetrically for each organism. Plotted on each graph are the best fits to the data for Equations 6.2 and 6.3. Since the values of De' and P were known the fitting process for both equations were carried out by iterating the value of 9rCm/4eo
until the optimum fit was achieved. Also shown on each graph is
a plot of Equation 6.2 with the 9rCm/4eo
term calculated from Equation 6.3 inserted into it. When this
is done Equation 6.2 becomes
. . .(6.7)
where De'lin is the linearised value
of De', and Ps
is the volume fraction calculated by conductimetry. In all cases
the better fit to the data is obtained from Equation 6.3.
Figure 6.11 is a plot of the linearised De'
divided by the measured Ps versus
the mean radius of the spheres of equivalent volume (via
Equation 6.6) for each cell type. It can be seen that, as one
would expect from Equation 6.2, the dielectric increment is proportional
to the (average) cell radius.
Figure 6.12 shows the growth curves in batch culture obtained
by dry weight, wet weight, DC and total
counts. It can be seen that there is good qualitative agreement
between the four methods. The data shown for DC
in this plot were obtained from offline scans, and are thus discontinuous
measurements, however the dielectric method has the advantage
over the other methods that it can be used to give an on-line
and real-time estimation of the biomass.
As can be seen from Figure 6.13 the viability determined by methylene
blue measurements was high (greater than 95%) for most of the
batch fermentation. As may be expected when the culture reached
stationary phase the viability began to decrease slightly. Also
shown in the figure is the percentage of cells with buds. The
graph clearly
Figure 6.10 : Relationship between the dielectric increment and the volume fraction of cells. Dielectric measurements were carried out as described in Materials and Methods and the volume fraction was measured conductimetrically. The lines constitute the best fit to Equations 6.2, 6.3 and 6.7. A Micrococcus luteus, B DCL baker's yeast, C DCL distillery yeast and D BB11 yeast.
Figure 6.10 continued.
Figure 6.10 continued.
Figure 6.10 continued.
Figure 6.11 : Relationship between the linearised dielectric
increment at a volume fraction of 1 and the cell radius. The line
constitutes the best linear least-squares fit to the data.
Figure 6.12 : Growth curves obtained during batch growth
of yeast. ¨ wet weight
(mg.ml1 / 5), + dry weight (mg.ml1),
$ DC
(pF), # number of cells
(millions / 15). There is good agreement between the
four methods of biomass estimation.
Figure 6.13 : Viability and % budding of the yeast during a batch fermentation. Measurements were made on methylene blue-stained cells as described in Materials and Methods. The viability remained high throughout the batch fermentation, while the number of budding cells rapidly increased and remained high during exponential growth before falling towards the initial level as the culture entered stationary phase.
shows that the number of budding cells rose rapidly from 61% in
the inoculum to a level of about 90%. The number of budding cells
remained at the higher level throughout the growth phase before
falling towards the initial level when the culture entered the
stationary phase at the end of the fermentation.
Figure 6.14 shows the DC of a culture
of baker's yeast during growth in permittistatic culture. The
cells were initially grown as a batch culture (i.e. no medium
inflow), and then at the point indicated by the arrow permittistatic
control of the biomass level was initiated. It can be seen that
the Biomass Monitor provided excellent control of the biomass
level in the fermentor, in this case for a period in excess of
two months.
One might expect the volume of medium pumped into the fermentor
per unit time to be constant during permittistatic culture, or
possibly one may expect it to increase as faster growing strains
(or mutants) take over. However, as can be seen in Figure 6.15,
this is not the case. This figure shows the pump activity during
a portion of the permittistat run represented in Figure 6.14.
Each point on the graph is integrated over a period of one hour.
Marked oscillations are apparent, with a prominent oscillation
of about 15 hours duration, and a second oscillation of about
100 hours duration. A fast Fourier transform of the data was done
using the MATLABä package (The
Mathworks Inc., Prime Park Way Natick, MA), and the resulting
graph is shown in Figure 6.16. Fourier transforms enable one to
identify any underlying regularities present in a data set, and
in this case the method has identified the two major periodicities
that were in fact visible in the original data (Figure 6.15).
Further attempts to analyse the data were made using a package
called "Chaos Data Analyzer" (Academic Software Library,
Raleigh, NC), which incorporates a suite of tools for the analysis
of time-series data. In the present case the data were analysed
using phase-space plots (Abraham and Shaw, 1992), calculation
of the Lyapunov exponent and of the correlation dimension (Petigen
et al., 1992). However the results
Figure 6.14 : The delta capacitance of a yeast culture
during growth in a permittistat. Initially the yeast were grown
in batch and then permittistatic control of the biomass concentration
was initiated at the time indicated by the arrow. The DC
was estimated from dual-frequency measurements as described in
the Materials and Methods and when it exceeded the set point (5
pF) medium was pumped in to dilute the yeast to the correct biomass.
Figure 6.15 : Time-dependent growth rate changes in a permittistat
culture. Measurements were made as described in Materials and
Methods. Each time the pump was switched on to control the biomass,
the length of time the pump was on was recorded to a file. The
volume of medium pumped per hour was then calculated. Prominent
oscillations in medium inflow rate were observed during the permittistatically-controlled
fermentation. The data presented are for the mid-part of the run
shown in Figure 6.14.
Figure 6.16 : A fast Fourier transform of the medium-inflow data shown in Figure 6.15. The Fourier transform shows two prominent peaks (at frequencies of 0.0092 and 0.0712 h1) representing underlying cycles of 108.7 and 14 hours duration.
were somewhat inconclusive with the pump activity data giving
results intermediate between those obtained from known, deterministically
chaotic data sets (e.g. the semi-logistic equation, (Markx and
Davey, 1991)) and from the example of white noise provided with
the package.
Attempts were made to gain an understanding of the processes
underlying the shorter of the two oscillations by analysing samples
from the permittistat every hour for a fifteen hour time period.
The pump activity during that fifteen hour period of the study
is shown in Figure 6.17, and may be summarised as a ten hour period
of medium input followed by a five hour period during which no
medium input occurred.
Figure 6.18 shows the percentage of budding cells present in
the permittistat during this time. It can be seen that the percentage
of cells with buds increases during the part of the oscillation
in which medium input is occurring and decreases towards the initial
level when medium input ceases. This would suggest some degree
of synchrony within the yeast culture although this was not evident
from the DNA histograms obtained by flow cytometry (not shown).
Also shown in Figure 6.18 is the data for the Cole-Cole a.
The Cole-Cole a was calculated
from offline dielectric scans of the yeast suspension by fitting
the data to the Cole-Cole equation:
. . .(6.8)
where e'w
is the permittivity of the suspension at a given frequency, e'¥
is the permittivity at a high frequency with respect to the b-dispersion,
De' is the dielectric increment of
the bdispersion, f is frequency
in Hertz, fc is the characteristic frequency of the bdispersion
(see Figure 6.4), and a is the Cole-Cole a.
GraFit v 2.0 (Erithacus Software Ltd., PO. Box 35, Staines U.K.)
was used to fit the equation to each data set following the subtraction
of its polarisation control. Figure 6.18 shows that changes in
Figure 6.17 : Short-term changes in flow rate in a permittistatic
culture. Measurements were made as described in the legend to
Figure 6.15. For a ten hour period a variable amount of medium
input was required to maintain the set-point, but over the next
five hours the biomass in the permittistat never exceeded the
set-point and so no medium input occurred.
Figure 6.18 : Changes in the percentage of budding cells during the oscillation shown in Figure 6.17 were reflected in the Cole-Cole a of the yeast suspension. Off-line scans of the yeast suspension were carried out in a Mexican hat electrode and for each scan a polarisation control was done (see the Methods). GraFit v 2.0 was used to fit the ColeCole equation (Equation 6.8) to the data. An increase in the value of both parameters was seen during the phase of the oscillation where medium input was occurring, and when medium input ceased the magnitude of both decreased.
the percentage of budding cells do appear to be reflected in the
curve for the Cole-Cole a. The
reasons for this relationship are unclear since the mechanisms
underlying the large Cole-Cole a values
of biological systems are not well understood (Markx et al.,
1991b; Davey and Kell, 1994), but may indeed reflect morphological
polydispersity.
Figure 6.19 shows the changes in glucose concentration occurring
in the permittistat during the oscillation. By comparison of this
figure with Figure 6.17 it can be clearly seen that during the
phase of the oscillation where medium is being pumped in to the
fermentor the glucose concentration increases, before falling
back towards the base level in the period when there is no pump
activity. Figure 6.20 shows the changes in ethanol concentration
over the same time. In contrast to the glucose concentration,
the ethanol concentration falls slightly as medium is pumped into
the fermentor, and rises when medium input ceases. However before
medium input restarts the ethanol concentration falls back to
the initial level; a possible explanation for this behaviour could
be that some portion of the population switches to growth on ethanol
as the carbon source when the glucose concentration falls. This
theory is supported by the fact that when anaerobic permittistatic
culture of yeast was carried out (Markx et al., 1991a)
no oscillations were observed.
Permittistatic culture of the Crabtree-negative yeast Kluyveromyces
marxianus was conducted in order to further investigate the
hypothesis that the oscillations were related to a respiro-fermentative
metabolic switch. Initially oscillations were not apparent in
these cultures, but when the permittistat had been running for
about 700 hours some oscillations in the growth rate were observed
although they were not as marked nor as regular as those seen
with the Saccharomyces cerevisiae (data not shown). Experiments
involving injection of ethanol into a permittistatic culture of
Saccharomyces cerevisiae also have induced oscillations
in the rate of medium influx (not shown). Although preliminary
results suggest that there is some link between the
Figure 6.19 : Changes in the glucose concentration during
the oscillation in Figure 6.17. The glucose concentration was
measured as described in the Methods. During the period where
medium inflow is occurring (0-10 hours approx.) the glucose concentration
steadily increases, before beginning to fall again when medium
input ceases.
Figure 6.20 : Short-term changes in ethanol concentration in the permittistat during the oscillation shown in Figure 6.17. The ethanol concentration was measured as described in the Methods. During the period of medium inflow (0-10 hours approx.) the ethanol concentration remained quite stable, however when medium input ceased an increase in the ethanol concentration was observed.
observed oscillations and a switch between growth on glucose and
on ethanol, further work would be required before any firm conclusions
may be made.
Figure 6.21 shows a permittistat run in which the setpoint was varied to show the utility of the permittistatic method for the stable maintenance of biomass at different concentrations. At the point labelled a the biomass monitor was switched from high-range to low-range, thus increasing the sensitivity of the instrument, manifested by an
approximate doubling of the delta capacitance. At points b,
c, d, and e the setpoint was changed. It
can be seen that the new setpoints are achieved rapidly and without
overshoot. At the point marked f, the fermentor became
contaminated with an acid-producing bacterium which affected the
conductance of the medium (see Figure 6.22). In fact the conductance
increased to such an extent that the biomass monitor was pushed
beyond its operational limits. This resulted in the capacitances
at both high and low frequencies becoming negative, and therefore
being read back into the computer as zeros. Thus it can be seen
that under certain conditions the biomass monitor can be useful
in identifying contamination of the fermentor.
Figure 6.23 shows the size distributions of yeast as determined
by flow cytometry from forward light scattering measurements (see
Chapter 2). Figure 6.23A is of yeast taken from a permittistatic
culture 4 volume changes after changing the setpoint. Figure 6.23B
shows yeast from the same permittistat after 6.8 volume changes
(3 days later). It can be seen that during this time the modal
diameter had increased, as had the range of cell sizes.
The measurement of biomass on-line and in real-time in fermentors
is a necessity if turbidostatically controlled continuous cultures
are to be used, and the development of sensing devices capable
of this task has long been an area of interest
Figure 6.21 : Delta capacitance of a continuous yeast culture
under permittistatic control. Dual frequency measurements were
made as described in Materials and Methods, and the DC
was calculated. The annotations are explained in the text.
Figure 6.22 : Conductance at 0.4 MHz (bottom line) and at 9.5 MHz (top line) of the yeast suspension during permittistatic culture. Contamination of the permittistat by an acid-producing bacterium resulted in a rapid increase in the conductance of the growth medium. At the point where the conductance began to fall the Biomass Monitor used was pushed beyond its operational limits of 6 mS.
A
Diameter mm
B
Diameter mm
Figure 6.23 : Cell size distributions of baker's yeast
grown in a permittistat. Cell size was estimated from the extent
of forward light scattering measured flow cytometrically. The
difference in the light-scattering properties of latex beads and
of cells was taken into account using the calibration factor described
in Chapter 2. The data presented are for samples taken after 4
volume changes (A) and after 6.8 volume changes (B)
following a change in the setpoint from 13 to 7 pF.(Harris
and Kell, 1985). An ideal sensing device for measuring the biomass
in a fermentor should be sensitive to changes in biomass, but
be insensitive to both non-biomass solids and necromass.
In addition the signal should be linear with biomass to concentrations
in excess of those that one would expect to find in fermentors.
It is shown herein that the Biomass Monitor fulfils these requirements.
The Biomass Monitor's signal is linear with biomass to concentrations
exceeding those found in most fermentations. At concentrations
where linearity is lost it is shown that this is predictable using
a simple equation that can be implemented as part of a linearising
routine. The Biomass Monitor is shown to be an excellent method
for the monitoring and control of biomass levels within the fermentor.
The work presented here indicates that it can be used for the
monitoring and control of biomass even at high volume fractions.
It is also shown that the dielectric increment is directly proportional
to the mean cell radius. This would allow one to argue that a
selection pressure for small cells may exist in a permittistat
as, for a given capacitance setpoint, the smaller the cells the
more cells there would be. However, such a selection pressure
will be offset by the fact that cells with high growth rate (which
would also be selected for in turbidostats) tend to be larger
(Frame and Hu, 1990). It was hoped that flow cytometric techniques
would enable one to see whether selection for cell size does occur
with time in permittistatically-controlled continuous cultures.
However, the large (and variable) degree of heterogeneity, coupled
with the oscillations observed in the current work make it unlikely
that such a relationship will be easy to observe.
Much of microbial physiology is based on measurements made at
"steady state" in continuous cultures yet oscillatory
behaviour of microbes has been reported in continuous cultures
by several groups of workers (Degn and Harrison, 1969; Harrison,
1970; Cunningham and Nisbet, 1983; Satroutdinov et al.,
1992). Steady state conditions are normally assumed to have been
reached once a rather arbitrary five volume changes have elapsed
following a change in dilution rate. In the permittistat cultures
described in this chapter oscillations were still apparent even
after 50 volume changes had elapsed.
Work characterising the oscillations in microbial cultures has
frequently concentrated on the macroscopic properties of the culture.
Flow cytometry has recently been used to study oscillations in
batch cultures of yeast (Münch et al., 1992a, b).
With flow cytometry it is possible to quantify the heterogeneity
of the population in multidimensional space. The more distributions
of properties one measures the less likely it is that one will
ever be able to consider a culture to be at steady state.
In conclusion, the Aber Instruments Biomass Monitor provides a reliable and convenient method for controlling the biomass concentration in a fermentor, even when high biomass levels are encountered. The use of the Biomass Monitor will enable turbidostat-type fermentor systems to be used without the problems commonly found when biomass concentration is controlled on the basis of optical density measurements. The permittistat also provides a useful vehicle for the study and induction of oscillations in growth rate during continuous culture.
Author: Hazel Davey
![[Mail Me]](/icons/mailbox.gif){kind=icon}