See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/13190765

A method of graphically analyzing substrateinhibition kinetics

ARTICLE in BIOTECHNOLOGY AND BIOENGINEERING MARCH 1999
Impact Factor: 4.13 DOI: 10.1002/(SICI)1097-0290(19990220)62:43.3.CO;2-M Source: PubMed

CITATIONS

READS

15

144

5 AUTHORS, INCLUDING:
Jinsheng Wang
Natural Resources Canada
59 PUBLICATIONS 1,287 CITATIONS
SEE PROFILE

Available from: Jinsheng Wang

Retrieved on: 16 October 2015

A Method of Graphically Analyzing

Substrate-Inhibition Kinetics
Jinsheng Wang,1 Tetsuya Araki,1 Takahira Ogawa,2 Masayoshi Matsuoka,2
Hideo Fukuda2
1

Shiga Technology Center, Iwatani International Corporation,

Moriyama, 524 Japan
2
Department of Applied Microbial Technology, Kumamoto Institute of
Technology, Kumamoto, 860 Japan; telephone: +81-96-326-3111 (ext.
5136); fax: +81-96-326-3000; e-mail: ogawa@bio.kumamoto-it.ac.jp
Received 11 December 1997; accepted 31 July 1998

Abstract: A model of substrate inhibition for enzyme catalysis was extended to describe the kinetics of photosynthetic production of ethylene by a recombinant cyanobacterium, which exhibits light-inhibition behavior
similar to the substrate-inhibition behavior in enzyme reactions. To check the validity of the model against the
experimental data, the model equation, which contains
three kinetic parameters, was transformed so that a linear plot of the data could be made. The plot yielded reasonable linearity, and the parameter values could be estimated from the plot. The linear-plot approach was then
applied to other inhibition kinetics including substrate
inhibition of enzyme reactions and inhibitory growth of
bacteria, whose analyses would otherwise require nonlinear least-squares fits or data measured in constrained
ranges. Plots for three totally different systems all
showed reasonable linearity, which enabled visual validation of the assumed kinetics. Parameter values evaluated from the plots were compared with results of nonlinear least-squares fits. A normalized linear plot for all
the results discussed in this work is also presented,
where dimensionless rates as a function of dimensionless concentration lie in a straight line. The linear-plot
approach is expected to be complementary to nonlinear
least-squares fits and other currently used methods in
analyses of substrate-inhibition kinetics. 1999 John
Wiley & Sons, Inc. Biotechnol Bioeng 62: 402411, 1999.

Keywords: enzyme kinetics; photoinhibition; substrate

inhibition; inhibitory microbial growth

INTRODUCTION
Substrate inhibition is a subject discussed in most books
concerning enzyme kinetics. The inhibition makes the kinetic behavior complicated, featured by downward dependence of the reaction rate on substrate concentration after a
maximum rate, in comparison with noninhibitory kinetics.
When a nonproductive binding of substrate to enzyme is
considered, an additional parameter is required for description of the kinetics, and the formula is given as
v=

Vmax S
Ks + S + S2 Ki

Correspondence to: T. Ogawa

1999 John Wiley & Sons, Inc.

(1)

where [S] is the substrate concentration, and Vmax, Ks (this

notation will be used throughout this article for different
systems and mechanisms but similar kinetic patterns,
though in some cases it should be Km), and Ki are constants
independent of [S] (Dixon and Webb, 1964; Shuler and
Kargi, 1992; Blanch and Clark, 1996). The three parameters
in equation (1) can be evaluated by means of a nonlinear
least-squares fit (Cleland, 1963; Henderson, 1992). As Ki
, the equation reduces to the form of MichaelisMenten
kinetics (Michaelis and Menten, 1913), for which the LineweaverBurk plot (Lineweaver and Burk, 1934) was commonly used to determine the parameter values (Wilkinson,
1961; Henderson, 1992). Modifications of this plot are also
well known (Shuler and Kargi, 1992; Blanch and Clark,
1996). Though it had been pointed out that an unweighted
reciprocal plot could lead to poor estimates of the values
(Dowd and Riggs, 1965; Henderson, 1992; Ritchie and
Prvan 1996), the plot was applied widely and yielded a large
number of published data of the MichaelisMenten parameters (Henderson, 1992; Schnell and Mendoza, 1997). An
advantage of the reciprocal plot is that it enables visual
inspection of the model with respect to the data, and this is
an essential method of detecting abnormalities which might
invalidate the assumptions. If the plot deviates too much
from a linear relation, the equation may not be adequate for
the description. Besides, if the scatter of the data is too
great, the assay method may need to be improved (Henderson, 1992). Furthermore, for precisely measured data the
linear plot is believed to be adequate for the parameter
estimation (Tseng and Hsu, 1990). On the other hand, only
two parameters can be determined, i.e., from the intercept
and the slope, by the plot. It would be desirable if such a
plot could also be applied to the more complicated inhibition kinetics.
In this work, we present a method of producing a linear
plot for substrate-inhibition kinetics, based on our study of
ethylene production by a recombinant cyanobacterium. The
production is a photosynthesis process, which exhibits a
feature of light inhibition similar to that of substrateinhibition in enzyme reactions. We extended Eq. (1) to describe the specific production rate by substituting the sub-

CCC 0006-3592/99/040402-10

strate concentration with the light intensity. With the parameter values determined by a linear plot, the equation
gave a satisfactory description of the experimental results.
We then applied the linear-plot approach to other inhibition
kinetics including substrate inhibition of enzyme reaction
and inhibitory growth of bacteria, and produced a normalized plot for all the results applied. We hope that our results
will be useful for practical purposes and our approach be
complementary to the presently used nonlinear least-squares
fits and other methods.

MATERIALS AND METHODS

Cyanobacterial Strains
A cyanobacterial strain, Synechococcus sp. PCC7942 R2SPc, which lacked one of the cryptic plasmids (pUH24),
was used for a host. A strain named IEK2-2 was constructed
in which the host psbAI gene open reading frame (ORF) has
been replaced by a copy of the ethylene-forming enzyme
(EFE) gene ORF followed by a kanamycin resistance (Kmr)
gene (Sakai et al., 1997). The preparation methods of
IEK2-2 will be reported elsewhere.

Figure 1. Specific production rate of ethylene by a recombinant cyanobacterium IEK2-2 as a function of light intensity. The unit is nL (C2H4)/mL
(culture)/(cell concentration in OD)/min. OD denotes optical density,
which is a measure of cell concentration in the culture. 1 OD730 0.25 g
(dry weight cell)/L (culture).

light dependence did not change (results not shown). We

extended Eq. (1) to describe the intensity dependence as
= max

Growth conditions
Synechococcus cells were grown in BG-11 medium at 28C
with various light intensities by tungsten lamps (100 W)
ranging from 1 to 20 klux with a 1-liter Loux bottle under
200 mL/min 1% CO2 in air. Cell concentrations of Synechococcus were measured using a spectrophotometer and expressed in OD730. The conversion factor to the dry cell
weight was 0.25 g/L/OD730.
Assay of Ethylene-Forming Activity
For in vivo assays, the outlet gas from the each cultivation
vessel was collected directly into a test tube (16-mL volume) for a few minutes, and ethylene concentration was
measured by gas chromatography.

RESULTS AND DISCUSSION

Photosynthetic Production of Ethylene
The measured specific ethylene production rate at low cell
concentrations (the logarithmic phase of the cell growth) of
the recombinant cyanobacterium is shown Fig. 1 as a function of the intensity of incident light. The rate exhibits
clearly an inhibition effect of light intensity. Variations of
pH in the culture medium and growth history of the cells
resulted in changes in the production rate, but the pattern of

I0
Ks + I0 + I 20 Ki

(2)

where I0 is the light intensity, and max, Ks, and Ki are

constants independent of I0. The values of the constants can
be determined by nonlinear least-squares fit. However, we
want to check visually the applicability of the formula to the
data, and a linear plot similar to the LineweaverBurk type
will be convenient for this purpose.
As has been mentioned above, the LineweaverBurk plot
can determine only two parameters. If we reduce one unknown parameter in Eq. (2), we could then try to modify the
plot to determine the other two. Since the maximum production rate will be obtained at d/dI0 0 according to Eq.
(2), we can determine this light intensity as
I0 max KsKi.

(3)

Ki in Eq. (3) can thus be substituted by I0 max, which can be

estimated from the measured results. By taking this I0max as
2 klux from Fig. 1, we rearrange Eq. (2) into

1
1
Ks 1 I0
=
+
+
.
max max I0 4

(4)

A plot of 1/ against (1/I0) + (I0/4) should give a linear

relation, if Eq. (2) is applicable to the present system. As is
shown in Fig. 2, such a plot is practically linear. max and Ks
were determined as 2.7 nL/mL/OD730/min and 0.4 klux,
respectively, from the two intercepts. Hence we determined
the value of Ki as 10 klux from Eq. (3). The curve in Fig. 1
was calculated using these results, which represents the
measured data reasonably well.

WANG ET AL.: GRAPHICALLY ANALYZING SUBSTRATE-INHIBITION KINETICS

403

take the LineweaverBurk type transform to illustrate this

effect.
In the LineweaverBurk type plot, the independent variable is 1/I. The variance of this quantity depends on the
variances of I0 and I0 max as
1 I
1 I
2I0
dI0 +
dI0 max 3
dI0 max
I0
I0 max
I 0 max

d1 I =

(9)

where it is taken that dI0 0 since I0 is accurately controlled. To obtain better estimates of the parameters, the
variables can be weighted as proposed by Johansen and
Lumry (1961) for MicahelisMenten kinetics. Here we do
not repeat their theoretical considerations but apply the derived formulas directly.
A weighting factor can be given as
Wi =
Figure 2. Linear plot for validation of a model for photosynthetic production of ethylene by a recombinant cyanobacterium [Eq. (2)]. The plot is
based on Eq. (4) for the specific production rate of ethylene (Fig. 1).

Similaritiy between Substrate-Inhibition and

MichaelisMenten Kinetics
On the basis of the above discussion, Eq. (2) can be arranged to
=

max
1 + Ks1 I0 +

I0 I 20 max

max
1 + Ks/I

2
2
Ks0

Ii +
2

2max 0Ii 4

1
4i

(10)

i
2

where the subscript i relates to the ith data set (Ii , i), and
Ks0 and max 0 are preliminary estimates of Ks and max. is
an arbitrary constant independent of i, and 2(Ii ) and 2 (i)
are the variances of Ii and i. 2(Ii ) is equal to (dI)2 which
can be obtained as ((2I0/I30 max) dI0 max)2 from Eq. (9). Accordingly, the values of Ks and max by least-squares criterion are given as

W I I
,
K =
W
W
W
W
I I I
Wi

(5)

Wi

Wi

i
2

I0/I 20 max).

where I = 1/(1I0 +
This equation has the same
form of MichaelisMenten kinetics, which can be written as
v=

vmax
.
1 + Ks/[S

1/[(1/I0 + I0/I20 max)] Ks/max + 1/[max (1/I0 + I0/I20 max)]

(7)
or
max Ks//(1/I0 + I0/I20 max).

(8)

Effects of distortion of errors, due to the linear transforms,

on the estimation of the parameters were well discussed for
MichaelisMenten kinetics. For the present inhibition kinetics, there are additional errors due the estimate of I0 max. We

404

Wi

max =

(6)

Since Eq. (5) was originally for substrate inhibition, all

approaches for MichaelisMenten kinetics may be applied
to the inhibition kinetics. Here we discuss linear transforms
and error analyses, using our data of light-inhibited ethylene
production for illustration. The conclusions can then be applied directly to the substrate-inhibition kinetics.
Figure 2 is equivalent to a LineweaverBurk plot. We can
also plot the same data in Hanes or EadieHofstee ways by
rearranging Eq. (5) into

I
Wi

Wi

Ii 2
Wi

Wi
Ii

Wi
Ii

(11)

Wi
Ii i

(12)

In the case that the variance of I0max is negligible, 2 (Ii )

0 and Wi depends on 2 (i) only. Therefore, error in estimated I0max affects the estimates of Ks and max by introducing a term (K2so/2max 0(Ii )4)((2I0/I30 max) dI0 max)2 into the
weighting factor according to Eq. (10).
Although computation of Eqs. (11) and (12) is possible
with assumed 2(Ii ), it would be easier to let the computer
search an optimum I0 max first to make dI0 max 0. This
optimum I0 max should be near the one estimated by eye, and
result in maximum linearity of LineweaverBurk type plot.
The linearity can be reflected by the correlation coefficient

1 I 1 I1 1
R=
1I 1I 1 1
i

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 4, FEBRUARY 20, 1999

(13)

where (1/I) and (1/) are the means of 1/Ii and 1/i, respectively. This coefficient is readily determined if linear
regression is performed. Result of such a search for the
present data is listed in Table I, which shows that the maximum value of I0 max is from 2.18 to 2.20 klux. By taking
I0 max 2.19 klux, the parameter values are obtained from
linear least-squares fit of replotted data shown in Fig. 3a,
as max 2.88 nL/mL/OD730/min, Ks 0.516 klux and
Ki 9.29 klux. These values are based on unweighted
values of i, and can be improved further by weighting. By
taking dI0 max 0, we have 2(Ii ) 0 and Eq. (10) can be
reduced to
Wi =

2
1
4i

(14)

i
2

Since our data are not adequate for estimation of error distribution of i, we assume that (i) is proportional to i so
that
Wi g2i ,

(15)

where g is an arbitrary constant. By substituting this expression into equations (11) and (12), we obtain refined values
max 2.91, Ks 0.524 klux, and Ki 9.15 klux.
Alternatively, by assuming constant error of i, we obtain
Wi g4i ,

(16)

which gives max 2.93, Ks 0.533 klux, and Ki 9.00

klux.
Nonlinear least-squares fit of the same data, with the
above values as the initial values, gives the result of max
2.93, Ks 0.541 klux, and Ki 9.04 klux. Although our
data are not adequate, the above results show that after
weighting, with assumed error distributions, the linear plot
gives values very close to those from a nonlinear leastsquares fit. By contrast, differences in max and Ks between
unweighted plot with visually estimated I0 max and nonlinear
least-squares fit are within 10%, whereas the difference in
Ki is 26%. After the optimization of I0 max, the differences
become much less significant. Apparently, the optimization
of I0 max considerably reduces the error.
Table I. Correlation coefficient of linear (double reciprocal) plot for
photosynthetic ethylene production as a function of I0 max. R is the correlation coefficient defined by Eq. (13).
No.

I0 max (klux)

R (%)

1
2
3
4
5
6
7
8
9
10
11

2.00 (initialvalue)
1.90
2.10
2.20
2.30
2.25
2.22
2.21
2.19
2.18
2.17

99.903
99.821
99.958
99.976
99.948
99.968
99.974
99.975
99.976
99.976
99.975

Figure 3. Several linear plots for photosynthetic ethylene production rate

with optimized value of I0 max: (a) Lineweaver-Burk type; (b) Hanes type
[Eq. (7)]; (c) EadieHofstee type [Eq. (8)].

WANG ET AL.: GRAPHICALLY ANALYZING SUBSTRATE-INHIBITION KINETICS

405

Hanes and EadieHofstee type plots [Eqs. (7) and (8)] of

the data using the optimized I0 max are shown in Fig. 3b and
3c. The evaluated parameter values (by linear regression)
are listed in Table II together with the corresponding values
from unweighted and weighted double reciprocal plots and
direct nonlinear fit. Root mean squares of residuals RMS
( 1/n ni1 (i i)2; i and i, measured and fitted values of i; n, number of data) are also listed for comparison.
It can be seen that, for unweighted plots with optimized
I0 max, the latter two types yielded values closer to the
weighted or nonlinear fitting results, similar to the case of
the linear plots for MichaelisMenten kinetics.
The fact that the unweighted double reciprocal plot, with
visually estimated I0 max, yielded only rough estimates of the
parameters can largely be attributed to the inadequacy of
our data. It will be seen later that, with more data around the
maximum rate, better estimates can be obtained from an
unweighted plot. Moreover, the double reciprocal plot,
which does not have products of controlled and measured
variables, is probably the simplest one for checking data
against the model. When the applicability of the model is
justified and better estimates of the parameters are required,
an optimum I0 max can be searched near the estimated value
for improvement. Alternatively, the rough estimates of the
parameters from the unweighted plot can be used as initial
values for nonlinear least-squares fit.
In the following discussions, we will use unweighted
double reciprocal plot and compare the evaluated parameters with the corresponding results from nonlinear leastsquares fits.
Application of the Linear Plot to
Substrate-Inhibition Kinetics
Behavior of substrate inhibition can be useful in helping to
deduce the kinetic mechanism involved (Dixon and Webb,
1964; Tipton, 1992; Henderson, 1992), and the graphical
treatment of such data has been discussed in detail (Dixon
and Webb, 1964). However, for determination of the parameters of Eq. (1), only limited data, i.e., data obtained at
low concentration and high concentration ranges were considered. In brief, in the low concentration range the term
[S]2/Ki is neglected and the equation is written as
VmaxS
v=
,
Ks + S

(17)

which is the MichaelisMenten type and the Lineweaver

Burk plot can be used to determine Ks. On the other hand,
in the high concentration range the term Ks/[S] is neglected
and Eq. (1) becomes
v=

VmaxKi
,
Ki + S

(18)

and a plot of 1/v versus [S] is used to determine Vmax and Ki.
In this way, each plot uses only a constrained part of the
measured data. Furthermore, the value of Ks determined this
way would be affected greatly by errors of the data since it
has to be obtained from quite low concentration range
where the errors are more significant. As is well known, a
major argument against the LineweaverBurk plot is that
the errors in measured at low concentration range are
distorted more after the reciprocal transformation (Henderson, 1992). It has also been suggested that a double reciprocal plot based on Eq. (17) could be used first to obtain
Vmax first, and then the ratio Ks/Ki be determined from the
highest value of v/Vmax, which equals Ks/Ki/(2 + Ks/Ki) at
the substrate concentration Ks/Ki (Dixon and Webb, 1964).
The difficulty with this approach is that the true value of the
highest velocity is not always easily determined, when there
are no adequate data to show that a still higher velocity does
not exist. Even if the highest value is obtained, it has to be
very accurate to give reliable estimates of the K values from
this single point.
By comparison, our approach can produce a plot using all
available data with an estimate of the substrate concentration Ks/Ki, which is easier to obtain than v/Vmax or Ks/Ki/2
+ Ks/Ki), since it is not necessarily measured. As an example, we apply this method to nitrite oxidation by Nitrobacter winogradskyi (Boon and Laudelout, 1962), where
nitrite was an inhibitor, as is shown in Fig. 4a. The kinetics
was given as
v=

VmaxS
.
Ks + S1 + S Ki

(19)

Since dv/dS 0 yields [S]max KsKi , we can rearrange

the equation into

1
K s2
1
=
1+ 2
v Vmax
Smax

1
Ks
S
+
.
Vmax S S2max

(20)

Table II. Parameter values of photosynthetic ethylene production evaluated by several methods. RMS is the root mean square of the residuals of .

Unweighted double reciprocal plot with visually estimated I0 max

Unweighted double reciprocal plot with optimized I0 max
Weighted with assumption of proportionally increased error
Weighted with assumption of constant error
Unweighted Hanes type plot with optimized I0 max
Unweighted EadieHofstee type plot with optimized I0 max
Nonlinear least-squares fit

406

max
(nL/mL/OD730/min)

Ks
(klux)

Ki
(klux)

RMS
(nL/mL/OD730/min)

2.7
2.88
2.91
2.93
2.92
2.90
2.93

0.40
0.516
0.524
0.533
0.532
0.523
0.541

10
9.29
9.15
9.00
9.02
9.17
9.04

0.0371
0.0116
0.0100
0.00956
0.00955
0.0101
0.00879

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 4, FEBRUARY 20, 1999

nonlinear least-squares fit. No significant differences are

shown.
When Ki Ks, Eq. (19) reduces to Eq. (1). Such kinetics
was reported by D. Murry for hydrolysis of ethyl butyrate
by sheep liver carboxylesterase (Dixon and Webb, 1964).
The data are shown in Fig. 5a. To apply the linear plot, we
rearrange Eq. (20) into

1
Ks
1
=
1+
v Vmax
Ki

1
Vmax

1
Ks
S
+ 2
Vmax S Smax

1
Ks
S
+ 2
Vmax S Smax

(21)

and take [S]max as 3 103 M from the figure. A plot of 1/v

against (1/[S] + [S]/9) accordingly gave good linearity, as is
shown in Fig. 5b. The parameter values from this plot and

Figure 4. Application of the linear plot to nitrite oxidation by N. winogradskyi, where nitrite is an inhibitor. (a) Original data, where is the
oxidation rate. The curve represents the nonlinear least-squares fit. (b)
Linear plot based on Eq. (20).

Thus, a plot of (1/[S] + [S]/[S]2max) against 1/ should be

linear if Eq. (19) holds. The corresponding plot is shown in
Fig. 4b, where [S]max is taken from Fig. 4a as 20 (mM
nitrite). The plot shows reasonable linearity and Vmax, Ks,
and Ki are determined from linear regression (from now on
we will treat the plot in this way for comparison) and listed
in Table III, together with the corresponding values from

Table III. Comparison of parameter values evaluated from the linear plot
and the nonlinear least-squares fit for nitrite oxidation by N. winogradskyi.
RMS is the root mean square of the residuals of .

Linear plot (unweighted)

Nonlinear least-squares fit

Vmax
(%)

Ks
(mM)

Ki
(mM)

RMS
(%)

110.30
109.84

1.73
1.70

230.61
229.56

3.517
3.546

Figure 5. Application of the linear plot to inhibitory hydrolysis of ethyl

butyrate by sheep liver carboxylesterase. (a) Data presented by Dixon and
Webb. The curve represents nonlinear least-squares fit. (b) Linear plot
based on Eq. (21).

WANG ET AL.: GRAPHICALLY ANALYZING SUBSTRATE-INHIBITION KINETICS

407

the nonlinear least-squares fit of the data in Fig. 5a are listed

in Table IV.

Linear Plot for Substrate-Inhibited

Microbial Growth
Eq. (19) also applies to microbial growth, if it is written as
=

maxS
Ks + S1 + S Ki

(22)

where is the specific growth rate, and max, Ks, and Ki are
constants (Blanch and Clark, 1996). An example is shown
in Fig. 6a, where the specific growth rate of Candida utilis
is given as a function of acetate concentration (Cama and
Edwards, 1970). In a similar way, we plotted (1/[S] +
[S]/S2max)) against 1/. At first we set [S]max to 10 (g acetate/
L) corresponding to the maximum of the rate. However, this
value did not result in any linearity, as is shown by Fig. 6b.
Then we tried 6 (g acetate/L) which corresponds to the
second highest point, and saw improved result (not shown).
We searched further a better value around 6 (g/L). The plot
shown in Fig. 6c is the result of [S]max 5 (g/L), which
gives an approximately linear relationship. The values of
max, Ks, and Ki from this plot and the best fit values reported by Cama and Edwards are listed in Table V with
RMS. Again, no significant differences are shown.
In the case Ki Ks, Eq. (22) can be given as
=

maxS
Ks + S + S2 Ki

(23)

which is equivalent to Eq. (1). Such a pattern has been

reported for the inhibition of bacterial growth by n-pentane
(Uemura et al., 1969; Blanch and Clark, 1996), and the plot
based on Eq. (21) can be applied similarly.
It would be interesting to note that the plotted variables
for Eq. (20) and (21) are the same. Therefore, the linearity
itself does not suggest whether Eq. (20) or Eq. (21) is applicable. When this matter becomes a problem, maybe it is
better to assume first that Eq. (20) is operative. After determination of Ks and Ki, the criterion of whether Ki Ks
can be used for the judgment.
In comparison with the nonlinear least-squares fit, the
application of the above mentioned plot does not require

Table IV. Comparison of parameter values, evaluated from the linear

plot and the nonlinear least-squares fit, for inhibitory hydrolysis of ethyl
butyrate by sheep liver carboxylesterase. RMS is root mean square of the
residuals of .

Linear plot (unweighted)

Nonlinear least-squares fit

408

Vmax

Ks
(mM)

Ki
(mM)

RMS

0.933
0.930

0.853
0.842

10.55
10.63

0.00779
0.00777

initial estimates of Vmax, Ks, and Ki. On the other hand, the
value of [S]max can not be obtained with 100% accuracy,
although the maximum reaction rate or growth rate is always a focus of the measurements. Furthermore, the reciprocal-taking practice inevitably introduces additional errors.
Optimization of [S]max and weighting should be performed
if accurate values are required. However, although the parameter values from an unweighted plot are not likely the
ones which will give a best fit to the data, they can be
reasonable values for practical purposes, as can be seen
from the comparisons of Tables IIIV. The plot can be a
convenient method for visual inspection of the assumed
kinetics against experimental data. In addition, through the
plot it is possible to identify outliers which would affect the
results of nonlinear fitting (Henderson, 1992). For nonlinear
fits it is known that the same data can be fitted with greatly
different parameter values (Bates et al., 1987). Therefore a
linear plot may be used complementarily to such fits to
confirm the results, especially when the number of data is
not adequate for the latter, or the assumptions such as normal distribution of error, on which the validity of the nonlinear fit is based (Wilkinson, 1961; Cornish-Bowden,
1979; Henderson, 1992), are not justified. Our approach
may also be used complementarily to the plots for the limiting-condition data [Eqs. (17) and (18)]. Since it is well
known that those plots yield more reliable values for Vmax
than for K, one may use this Vmax value to plot Vmax/v
against (1/[S] + [S]/[S]2max) based on Eq. (21) to obtain Ks.
This way would be better than using the single point at the
highest velocity to determine Ks, as has been discussed earlier, since in the linear plot Ks is determined from all available points. Moreover, the linear plot can be used for comparison purposes, since for kinetics which follow Eq. (1)
precisely, parameter values determined by the different
methods should not differ appreciably.
Modification of a Direct Linear Plot
Next, we modify a direct linear plot used for the Michaelis
Menten kinetics to apply our method to the substrateinhibition kinetics. This plot has been discussed in detail,
and is believed to have several advantages over the leastsquares fit method (Cornish-Bowden, 1979, 1991; Henderson, 1992). The procedure is to plot v values onto a V
vertical axis and the corresponding negative S value onto a
Km horizontal axis. The corresponding points are then
jointed and the lines are extrapolated. Normally, for n pairs
of values, n(n 1)/2 intersections of the lines are obtained.
The coordinates of the intersections provide estimates of Ks
and Vmax, and the median values of these estimates are the
best-fit values. A disadvantage of this plot is that it cannot
be extended directly to fitting equations containing more
than two parameters, such as Eq. (1). However, we can
estimate [S]max first and rewrite the equation into
Vmax = vi + Ks

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 4, FEBRUARY 20, 1999

vi
1 1 Si + Si S2max

(24)

Table V. Comparison of parameter values, evaluated from the linear plot

and the nonlinear least-squares fit, for inhibitory growth of C. utilis. RMS
is root mean square of the residuals of .

Linear plot
Best fit values reported by
Cama and Edwards

max
(1/h)

Ks
(g/L)

Ki
(g/L)

RMS
(1/h)

0.626

1.434

17.44

0.0229

0.616

1.481

19.05

0.0229

where vi and [Si] are the ith pair of measured data. We can
then plot 1/(1/[Si] + [Si]/[S]2max) onto the horizontal axis and
vi onto the vertical axis and determine the values of Ks and
Vmax accordingly from the intersections. Figure 7 illustrates
the application of this method to our data of the specific
production rate of ethylene (Fig. 1). While Ks and max
determined from the median values of the intersections are
indeed close to the ones determined from our linear plot of
Fig. 2, the procedure would be laborious if there are more
pairs of data.
A Normalized Linear Plot for Various Systems
Finally, we produce a normalized plot for all of the data we
have discussed so far, using the linear-transform method.
Eqs. (20) and (21) can be rearranged respectively to
Vmax v

1+

Ks2
S2max

v
and

= Ks

1
S
+ 2
S Smax

1
Vmax v
S
= Ks
+ 2
.
v
S Smax

Figure 6. Application of the linear plot to inhibitory growth of C. utilis.

(a) Specific growth rate of C. utilis as a function of acetate concentration.
The curve represents a nonlinear fit, using parameters values (listed in
Table V) reported by Cama and Edwards. (b) A plot based on the extension
of Eq. (20) with [S]max 10 (g/L). (c) Linear plot based on the extension
of Eq. (20) with [S]max 5 (g/L).

(25)

(26)

Figure 7. Modification of a direct linear plot, for the inhibition kinetics

of photosynthetic production of ethylene.

WANG ET AL.: GRAPHICALLY ANALYZING SUBSTRATE-INHIBITION KINETICS

409

v
VmaxS
=
.
1 + S Ki Ks + S + S2 Ki

The terms on both sides of the equations are dimensionless.

We denote a dimensionless rate for the left-side terms and
a dimensionless concentration for the right-side term,
which is related to the substrate concentration or light intensity. Clearly, for kinetic data which can be described by
Eqs. (1), (2), (19), (22), and (23), the plots of versus
defined above should coincide on a line with unit slope and
zero intercept. Thus we can make a normalized linear plot
using the results of Figs. 2, 4b, 5b, and 6b, as is shown in
Fig. 8. Data measured up to larger , in the present case the
data for hydrolysis of ethyl butyrate by sheep liver carboxylesterase, will extend to higher part of the line. It might be
of interest to investigate the upper limit of such a plot,
which corresponds to the upper limit of for the inhibition
models to be valid, and see whether some characterizations
can be made.

A plot of v/[S] against [S] will then yield Vmax/Ks and

Vmax/KsKi, and hence /Ki. However, error in this estimate
would affect the subsequent analyses considerably. Such a
linear plot could be used at best for qualitative inspection of
the data, and we do not discuss more of it here.

Possible Application to a More General Case of

Substrate Inhibition

CONCLUSIONS

As has been mentioned at the beginning, Eq. (1) is based on

the consideration of a nonproductive binding of substrate to
enzyme. A more general case has been considered where
two identical active sides of enzyme exist. This kinetics can
be described by (Webb, 1963)
v=

VmaxS1 + S Ki
Ks + S + S2 Ki

(27)

where is a constant. Eq. (19) [and hence Eq. (1)] can be

regarded as a special case of 0. The formula also
applies to microbial growth if v is replaced by (Blanch
and Clark, 1996). Specifically, when > 1 the rate increases
with increasing substrate concentration and no maximum is
shown. Such kinetics would therefore not be easily distinguished from MichaelisMenten kinetics.
Eq. (27) can be written as

(28)

If the value of /Ki is known, the left side can be plotted as

a function of [S], and our linear-plot approach can be applied. The value of /Ki may be estimated from the data in
low concentration range, where the Ks >> [S] + [S]2/Ki and
Eq. (27) can be approximated by

v Vmax
=
1 + S .
S
K
K

s
i

(29)

The linear transformation of the inhibition model enabled us

to visually inspect the assumed kinetic pattern against measured data from a plot, for photosynthetic ethylene production by recombinant cyanobacteria, and several substrateinhibition kinetics of enzyme reaction and microbial
growth. The plot also yielded acceptable estimates of the
kinetic parameters. With a refined procedure, the plot can
give essentially the same parameter values as those from a
nonlinear least-squares fit. The linear plot, which uses the
value of KsKi estimated from the measured dependence of
velocity on the substrate concentration or light intensity,
may be used complementarily to nonlinear least-squares fit
and other methods, for validation of the assumed model, and
confirmation of evaluated parameter values. In the above
discussions, the substrate and mechanisms may differ
greatly, yet similar kinetics patterns have been observed.
Hence, the same approach may be extended to a considerably wider range of systems involving substrate or light
inhibition. Finally, a normalized linear plot is expected to
contain all of the results which follow the inhibition models.
By analyzing the similarities, new insights may be obtained.
The authors gratefully acknowledge the support of this work by
the Original Industrial Technology R&D Promotion Program
from the New Energy and Industrial Technology Development
Organization (NEDO) of Japan.

References

Figure 8.
work.

410

A normalized linear plot for all of the data discussed in this

Bates JHT, Bates DA, Mackillop W. 1987. On the difficulties of fitting the
double MichaelisMenten equation to kinetic data. J Theor Biol
125:237241.
Blanch HW, Clark DS. 1996. Biochemical engineering. New York: Marcel
Dekker, Inc.
Boon B, Laudelout H. 1962. Kinetics of nitrite oxidation by Nitrobacter
winogradskyi. Biochem J 85:440447.
Cama FJ, Edwards VH. 1970. Single-cell protein from chemical wastes:
The growth of Candida utilis on sodium acetate. J Ferment Technol
48:787793.

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 4, FEBRUARY 20, 1999

Cleland WW. 1963. Computer programmes for processing enzyme kinetic

data. Nature 198:463465.
Cornish-Bowden A. 1979. Fundamentals of enzyme kinetics. London: Butterworths.
Cornish-Bowden A. 1991. Parameter estimating procedures for the
MichaelisMenten model: Reply to Tseng and Hsu. J Theor Biol 153:
437440.
Dixon M, Webb EC. 1964. Enzymes. 2nd ed. London: Longmans.
Dowd JE, Riggs DS. 1965. A comparison of estimates of Michaelis
Menten kinetic constants from various linear transformations. J Biol
Chem 240:863869.
Henderson PJF. 1992. Statistical analysis of enzyme kinetic data, pp.
277316. In: Eisenthal R and Danson MJ, editors. Enzyme Assays.
New York: Oxford University Press.
Johansen G, Lumry R. 1961. Statistical analysis of enzymic steady-state
rate data. C R Trav Lab Carlsberg 32:185213.
Lineweaver H, Burk D. 1934. The determination of enzyme dissociation
constants. J Am Chem Soc 56:658666.
Michaelis L, Menten ML. 1913. Die kinetik der invertinwirkung. Biochem
Z 49:333369.
Ritchie RJ, Prvan T. 1996. Phyllotaxis as a dynamical self organizing
process, part I: The spiral modes resulting from time-periodic iterations. J Theor Biol 178:239254.

Sakai M, Ogawa T, Matsuoka M, Fukuda H. 1997. Photosynthetic conversion of carbon dioxide to ethylene by the recombinant cyanobacterium. Synechococcus sp. PCC 7942, which harbors a gene for the
ethylene-forming enzyme of Pseudomonas syringae. J Ferment Bioeng
84:417426.
Schnell S, Mendoza C. 1997. Closed form solution for time-dependent
enzyme kinetics. J Theor Biol 187:207212.
Shuler ML, Kargi F. 1992. Bioprocess engineering. Engelwood Cliffs, NJ:
Prentice Hall.
Tipton KF. 1992. Principles of enzyme assay and kinetic studies. pp. 158.
In: Eisenthal R, Danson MJ, editors. Enzyme Assays. New York:
Oxford University Press.
Tseng S, Hsu J-P. 1990. A comparison of the parameter estimating procedures for the MichaelisMenten model. J Theor Biol 145:457464.
Uemura N, Takahashi J, Ueda K. 1969. Fundamental studies on the cultivation of hydrocarbon-utilizing microorganisms. J Ferment Technol
(Jpn) 47:220228.
Webb JL. 1963. Enzyme and metabolic inhibitors. San Diego: Academic
Press.
Wilkinson GN. 1961. Statistical estimations in enzyme kinetics. Biochem
J 80:324332.

WANG ET AL.: GRAPHICALLY ANALYZING SUBSTRATE-INHIBITION KINETICS

411