J. Biochem. Biophys.

Methods 62 (2005) 13 – 24
www.elsevier.com/locate/jbbm

The comparison of the estimation of enzyme kinetic

parameters by fitting reaction curve to the integrated
Michaelis–Menten rate equations of different
predictor variables
Fei Liao*, Xiao-Yun Zhu, Yong-Mei Wang, Yu-Ping Zuo
Biochemistry Department, Chongqing University of Medical Sciences, Chongqing 400016, PR China
Received 28 November 2003; received in revised form 26 May 2004; accepted 25 June 2004

Abstract

The estimation of enzyme kinetic parameters by nonlinear fitting reaction curve to the integrated
Michaelis–Menten rate equation ln(S 0/S)+(S 0
S)/K m=(V m/K m)t was investigated and compared
to that by fitting to (S 0
S)/t=V m
K m[ln(S 0/S)/t] (Atkins GL, Nimmo IA. The reliability of
Michaelis-Menten constants and maximum velocities estimated by using the integrated Michaelis-
Menten equation. Biochem J 1973;135:779-84) with uricase as the model. Uricase reaction curve
was simulated with random absorbance error of 0.001 at 0.075 mmol/l uric acid. Experimental
reaction curve was monitored by absorbance at 293 nm. For both CV and deviation b20% by
simulation, K m from 5 to 100 Amol/l was estimated with Eq. (1) while K m from 5 to 50 Amol/l was
estimated with Eq. (2). The background absorbance and the error in the lag time of steady-state
reaction resulted in negative K m with Eq. (2), but did not affect K m estimated with Eq. (1). Both
equations gave better estimation of V m. The computation time and the goodness of fit with Eq. (1)
were 40-fold greater than those with Eq. (2). By experimentation, Eq. (1) yielded K m consistent with
the Lineweaver–Burk plot analysis, but Eq. (2) gave many negative parameters. Apparent K m by Eq.
(1) linearly increased, while V m were constant, vs. xanthine concentrations, and the inhibition

Abbreviations: K m, Michaelis–Menten constant; V m, maximum reaction rate; A b, background absorbance;

CV, variation coefficient.
* Tel.: +86 23 68485073; fax: +86 23 68485111.
E-mail addresses: liaofeish@vip.sina.com, liaofeish@yahoo.com (F. Liao).

0165-022X/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbbm.2004.06.010
14 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24

constant was consistent with the Lineweaver–Burk plot analysis. These results suggested that the
integrated rate equation that uses the predictor variable of reaction time was reliable for the
estimation of enzyme kinetic parameters and applicable for the characterization of enzyme
inhibitors.
D 2004 Elsevier B.V. All rights reserved.

Keywords: The integrated method; Michaelis–Menten constant; Reaction curve; Predictor variable; Nonlinear
fitting; Uricase

1. Introduction

The characterization of inhibitor by its inhibition type and inhibition constant is

important for the discovery of inhibitor as drug, and it requires fast assay of both the
Michaelis–Menten constant (K m) and the maximum reaction rate (V m) of an enzyme
[1, 2]. It is common to estimate the apparent kinetic parameters of enzymes in the
presence of inhibitor by Lineweaver–Burk plot, but it requires substrate concentration
distributed from that well below the K m in the absence of inhibitor to that far above
this K m [1–3]. The technique of higher sensitivity, usually with more cost and/or labor,
is required when K m is too low, while the solubility of substrate is a problem when K m
is too high. And its costs on both the target enzyme and substrate are quite large. Thus
alternative methods are needed for fast estimating kinetic parameters and screening
inhibitors.
Fitting an enzyme reaction curve to the integrated Michaelis–Menten rate equation, the
integrated method, is an alternative way to estimate enzyme kinetic parameters [4–6].
Using the integrated method, only one reaction curve is required to determine both V m and
K m at a substrate concentration that usually enables common techniques to monitor the
reaction. Thus both the labor and the cost for the characterization of inhibitor are
effectively reduced. The integrated Michaelis–Menten rate equation of enzyme acting on
single substrate with neither reverse reaction nor product inhibition is Eq. (1),
lnðS0 =S Þ þ ðS0
S Þ=Km ¼ ðVm =Km Þ  t ð1Þ
and it can be transformed into Eq. (2), a linear form with respect to kinetic parameters.
ðS0
S Þ=t ¼ Vm
Km  ½lnðS0 =S Þ=t  ð2Þ
For the classical integrated method, Eq. (2) is used to estimate K m and V m, and more than
85% consumption of the substrate at initial concentration above 1.4-fold of the K m is
required [4–6]. But in fact, this classical integrated method is too sensitive to common
errors to give reliable kinetic parameters by either simulation [6] or experimentation [7].
Moreover, the classical integrated method does not estimate the equivalent of initial
substrate concentration [8]. These results suggested the validity of Eq. (2) requires more
thorough investigation and consequently it is seldom used in routine practice.
Recently, it has been found that with K m as a preset constant, the integrated method
with Eq. (1) is reliable for the estimation of both initial substrate concentration [9] and
enzyme V m over a wide range of substrate concentrations [10–12]. And this integrated
 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24 15

method showed enhanced precision and strong resistance to common errors. These
properties suggested some differences between Eq. (1) and (2) for parameter estimation,
and nonlinear fitting of an enzyme reaction curve to Eq. (1) may give reliable estimation of
both K m and V m. But to date there have been no investigations on the reliability of Eq. (1)
for the simultaneous estimation of enzyme K m and V m. The uricase reaction follows single
substrate Michaelis–Menten kinetics with neither reverse reaction nor product inhibition.
Its reaction can be monitored by absorbance at 293 nm, with xanthine as its competitive
inhibitor [8,13]. Here with uricase as the model, nonlinear fitting reaction curve to Eq. (1),
in comparison to Eq. (2), was investigated by both simulation and experimentation for the
estimation of K m and V m and for the characterization of its inhibitor xanthine.

2. Materials and methods

2.1. Simulation of uricase reaction curve

The initial concentration of uric acid was at 75 Amol/l with absorptivity (e) of 11.5
(mmol/ld cm)
1, and V m/K m was at 2.5 min
1 to calculate the error-free reaction curve
with more than 95% consumption of substrate. Uniform absorbance error of 0.001 was
randomly inserted into the reaction curve to give the simulated curve [9]. Random
insertion of 10% relative error to any one point with absorbance above 0.200 gave the
simulated reaction curve with one outlier.

2.2. The monitor of experimental reaction curve

The reaction mixture of 1.20 ml contained 1.00 ml buffer, 75 Amol/l uric acid (ICN),
and final 0.01 U/ml Candida utilis uricase (EC 1.7.3.3, ICN 101203) without inhibitor, or
0.02 U/ml in the presence of xanthine below 20 Amol/l, respectively [8]. The reaction
curve by absorbance at 293 nm was monitored 15 s after the addition of enzyme with the
interval of 15 s, and the recording was lasted for 10 min to give residual uric acid
b2.0 Amol/l.

2.3. Data processing for fitting to Eq. (1)

There were at least five points with absorbance change above 0.002 for analysis. Any
one point with the reaction time t i,0 and absorbance A i,0 was assigned as the first datum for
analysis to transform reaction time (t i ) into a new series (t iV) according to t iV=t i
t i,0. The
background absorbance was taken as a constant of A b and the total absorbance (A i ) at any
t i equal to the absorbance of uric acid plus A b. Thus Eq. (1) was transformed into Eq. (3)
with an intercept a to account for the effects of lag time of steady state reaction and A b.



ln Ai;0
Ab =ðAi
Ab Þ þ Ai;0
Ai =ðe  Km Þ ¼ a þ ðVm =Km Þ  ti V ð3Þ

The reaction curve was fitted to Eq. (3) after data transformation according to the left in
Eq. (3) with instantaneous rates as weighting factors. A b was treated as nonlinear
16 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24

parameter by the decrement of 0.001 from that 0.002 below the lowest absorbance under
analysis to
0.020, and K m was adjusted by the step of 2% K m from 40% to 200% of the
preset K m or the apparent K m (from Lineweaver–Burk plot by experimentation) [9–11].
The combination of A b, K m and V m/K m for the highest F value was the estimate of
parameters, and V m/K m times the corresponding K m yielded V m.

2.4. Data processing for fitting to Eq. (2)

Eq. (2) was transformed into Eq. (4),

Ai;0
Ai =ðe  ti VÞ ¼ Vm
Km  ln Ai;0
Ab =ðAi
Ab Þ =ti V ð4Þ

data were transformed according to right part of Eq. (4) with (ln((A i,0
A b)/(A i
A b))/t iV) as
the predictor variable to fit the reaction curve, and instantaneous rate was taken as the
weighting factor while A b as the nonlinear parameter [4–6]. The goodness of fit was
examined by F test, the final parameters were those that gave the highest F value.

2.5. Programming

The program in Qbasic 4.5 was essentially the same as before when Eq. (3) was used
[9–11]. When Eq. (4) was used to fit the reaction curve, data were transformed according
to the left and right part of Eq. (4), respectively, to nonlinearly fit the reaction curve by
stepwise downregulation of A b within the same range [6].

2.6. Statistic analysis of the results

The results were xFS.D. The averages were compared by Student’s t-test, and the
effects of initial concentrations of substrate on K m estimation were analyzed by F test,
respectively, with Pb0.05 as the confidence limit. The inhibition of xanthine on C. utilis
uricase was characterized by re-plot of the apparent kinetic parameters. The linearity of
regression analysis was examined by its determination coefficient.

3. Results

3.1. The analysis of simulation reaction curve

By simulation with both equations, higher initial substrate concentration and lower
residual substrate concentration gave better estimation of kinetic parameters. The
estimation of V m by two equations usually showed the deviation and variation coefficient
(CV) below 10%. With both equations, the concentration of residual substrate necessary
for reliable estimation of the preset K m by Eq. (3) not only depended on the initial
concentration of substrate, but also depended on the preset K m itself (Fig. 1). There was
difference between the ranges of K m able to be estimated with reliability by two equations.
With initial substrate concentration at 70 Amol/l and residual substrate below one-fifth of
 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24 17

Fig. 1. The highest residual uric acid (S f, Amol/l) to yield K m with both deviation and variation below 20% at
given initial concentrations of uric acid (S 0) by Eq. (4).

the preset K m, the analysis of reaction curve by Eq. (3) yielded both deviation and CV
below 15% for preset K m from 5 to 100 Amol/l, while the analysis by Eq. (4) yielded both
deviation and CV below 20% only for K m from 5 to 50 Amol/l. The analysis of reaction
curve by Eq. (4) yielded both deviation and CV far above 30% in K m preset at 100 Amol/l.
For K m below initial substrate concentration and residual substrate below one-eighth of the
preset K m, both the relative deviation and CV in K m by fitting to Eq. (4) were
approximately twice of those obtained by fitting to Eq. (3), and these differences were
more pronounced at lower initial concentrations of substrate. With residual substrate
concentration below one-tenth of the initial one that was above 5-fold of the K m, both
equations yielded deviation and CV far above 30% for K m of 1 Amol/l, and K m of 4 Amol/l
served as the lower limit for estimation by both equations. The highest F value for fitting
reaction curve to Eq. (3) was always 100-fold higher than that for fitting the same reaction
curve to Eq. (4). The computation time with Eq. (3), however, was 80-fold longer than that
with Eq. (4).
The effects of A b on K m estimation by these two equations were different. Simultaneous
insertion of artificial A b of 0.100 into reaction curve did not alter the K m estimated by Eq.
(3), and A b for best fitting usually showed no difference from the preset one. But this A b
led to about 30% negative K m by Eq. (4), even when initial substrate was above twice the
K m and residual substrate was below 7% of its initial value. With residual substrate
increased to 20% of its initial one that was comparable to the K m, the insertion of A b
comparable to initial absorbance of substrate led to negative values to nearly half of the
K m, V m and A b by Eq. (4). But with Eq. (3) to analyze the same reaction curve, there still
were no negative values of kinetic parameters except the increase of deviation and CV up
to around 30% for K m, and around 15% for V m, respectively.
The error in the lag time to reach steady state reaction also differently altered the
estimation of kinetic parameters by these two equations. Insertion of artificial lag time of
up to 2 min into the reaction time did not change kinetic parameters estimated by Eq. (3)
even when reaction time was not transformed by presetting that at the first datum for
analysis to zero. Without the transformation of reaction time, the error of less than 0.2 min
in the lag time required that there were initial substrate concentrations above twice the K m
and residual substrate below 5% of the initial one to give both deviation and CV below
18 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24

20% in K m estimated by Eq. (4). The error larger than 0.2 min in the lag time partially
resulted in negative K m. Larger error in the lag time of reaction and higher activity of
enzyme always led to more negative K m estimated by Eq. (4), but the appearance of
negative V m was relatively rare. If reaction time was transformed by presetting that at the
first datum for analysis to zero, there were also no effects of the error in the lag time of
reaction on the estimation of kinetic parameters by Eq. (4).
The estimation of kinetic parameters by both equations was very sensitive to outlier in
the reaction curve. Insertion of one outlier usually resulted in 50% negative K m, and 30%
negative V m by Eq. (4) with the highest F value below 1.0. Usually A b for best fit to Eq.
(4) also showed negative values. When Eq. (3) was used, the same outlier reduced the
highest F value to below 5% of that without this outlier, and the deviation and CV for both
K m and V m were doubled. But there were no negative values in either K m or V m, while
there was negative A b for best fitting.

3.2. The analysis of experimental reaction curve without inhibitor

Lineweaver–Burk plot analysis yielded K m of (13.1F1.3) Amol/l for C. utilis uricase

(uric acid 5–40 Amol/l, n=3, determination coefficients N0.97). The point 30 s after the
initiation of reaction was taken as the first datum for analysis by Eq. (3) after the
transformation of reaction time. Under this condition, the analysis by Eq. (3) of reaction
curves with initial substrate N14 Amol/l while residual substrate b2.0 Amol/l still yielded
no negative K m. When residual substrate was b2.2 Amol/l, the decrease of initial substrate
concentration from 70 to 14 Amol/l resulted in no changes in either the mean or standard
error of K m (Table 1, PN0.1 by F test). The average of K m was (12.8F1.0) Amol/l,
consistent with that by Lineweaver–Burk plot ( PN0.1 by t-test). When the initial
concentration of uric acid was slightly below 14 Amol/l while residual substrate was
around 2.5 Amol/l, there were deviation and CV above 20% in K m by Eq. (3), but there still
were no negative kinetic parameters. For all these cases, CV in V m by Eq. (3) were usually
below 10%.
However, the analysis of all the reaction curves by Eq. (4) after the transformation of
reaction time yielded negative values for nearly one-fifth of the K m, and there were also

Table 1
The comparison of K m (xFS.D.) estimated by two equations
Sf S0 60–70 50–60 45–50 40–45 35–40 25–35 14–25
b2.2 Eq. (3) 12.2F1.1 12.2F0.9 12.3F0.7 12.2F0.8 12.4F0.9 11.9F0.5 11.2F1.5
Eq. (4)
34.7F42.2 11.8F2.3 10.7F3.5 12.4F2.4 13.6F3.8 14.1F2.6 19.6F10.4
(n) (11) (10) (11) (10) (12) (13) (22)
2.3–4.5 Eq. (3) 14.4F1.8 16.2F1.9 15.3F2.9 15.3F2.9 16.2F2.6 14.7F1.1 17.8F5.5
Eq. (4)
33.1F47.8 16.9F5.7 22.9F4.8 12.8F3.3 20.4F1.0 17.8F7.1 11.5F6.7
(n) (3) (3) (3) (2) (2) (3) (8)
Both initial substrate concentration (S 0) at the specified first datum under analysis and the residual substrate
concentration (S f) were in micromolar, n is the number of reaction curves as the combination of different S f to
different S 0. By Eq. (4), there were negative K m in the groups when variation coefficients were above 25% for
more than 11 combined reaction curves. All results were from the analysis of 15 independent reaction curves after
the transformation of reaction time.
 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24 19

negative V m and A b. With different initial concentrations of uric acid and the same residual
substrate b2.2 Amol/l at the end point, the analysis of reaction curve by Eq. (4) yielded
higher S.D. in K m than that by Eq. (3) ( Pb0.05 by F test). Both the computation time and
the highest F values for fitting to Eq. (4) were always less than 3% of those for fitting to
Eq. (3). Without the transformation of reaction time, the analysis by Eq. (3) of all reaction
curves resulted in no change of K m and V m. But without the transformation of reaction
time, the analysis by Eq. (4) yielded negative K m up to 70% with reaction curves from 15 s
after the initiation of reaction to the end point of residual uric acid b2.2 Amol/l, and the
analysis of reaction curve from 45 s after the initiation of reaction to the same end point
yielded negative K m close to 30%. The percentage of negative V m at the same time was
usually below 15%. If the reaction time was transformed before the analysis by Eq. (4),
taking the point 15 s after the initiation of reaction as the first datum for analysis gave 30%
negative K m, and taking the point 45 s after the initiation of reaction as the first datum for
analysis gave less than 4% negative K m. But for these cases negative V m was very rare.
There was no negative K m if the point 90 s after the initiation of reaction was taken as the
first datum for analysis by Eq. (4) after the transformation of reaction time. The analysis of
reaction curve by Eq. (4) with the point 15 s after the initiation of reaction as the first
datum yielded higher S.D. in K m vs. that with the point 45 s after the initiation of reaction
as the first datum, when residual substrate at the end point was b2.2 Amol/l ( PN0.1
by F test).

3.3. The inhibition of xanthine on uricase characterized by fitting to Eq. (3)

The concentrations of residual uric acid were b2.2 Amol/l while initial uric acid
concentrations at the point 30 s after the initiation of reaction were N65 Amol/l for the
estimation of apparent kinetic parameters by fitting reaction curve to Eq. (3) in the
presence of xanthine below 20 Amol/l (apparent K mb58 Amol/l). Under these conditions,
CV for the estimation of apparent K m was usually b12%, and that for the estimation of V m
was usually b7% (n=4). By Lineweaver–Burk plot analysis with substrate concentrations
from 10 to 100 Amol/l and final uricase at 0.006 U/ml, the estimation of the apparent K m
and V m showed CV below 10% (n=3). There were no differences between the K m
estimated by nonlinearly fitting to Eq. (3) and that by Lineweaver–Burk plot analysis for
all the tested concentrations of xanthine. Both methods yielded V m without significant
changes, but yielded apparent K m of linear increase to xanthine concentrations (Table 2).
By either method, there was no difference between the intercept of the response plot of the

Table 2
Regression analysis of the response of apparent kinetic parameters to xanthine concentrations (C x) from 0 to
20 Amol/l
Methods Regression equation R2 S P
Lineweaver–Burk K m=13.9+2.05C x N0.985 1.5 b0.05
Plot V m=4.6–4.0810
4C x b0.653 0.07 N0.1
The integrated K m=13.4+2.14C x N0.992 1.4 b0.01
Method with Eq. (3) V m=13.5+1.1110
4C x b0.672 0.07 N0.1
R 2 is the determination coefficient, S the standard error of estimate, and P the probability, respectively.
20 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24

apparent K m to xanthine concentration from that without xanthine. The analysis of the
response of the apparent K m estimated by Eq. (3) to xanthine yielded the inhibition
constant of (5.8F1.4) Amol/l (n=3), and that by Lineweaver–Burk plot was (6.6F0.8)
Amol/l (n=3). These two inhibition constants showed no differences in both the S.D.
( PN0.1 by F-test) and the average ( PN0.1 by t-test).

4. Discussion

The characterization of the inhibition type of enzyme inhibitor helps the rational design
of inhibitors. Apparent K m of enzyme was the most important parameter for the
characterization of inhibitor. Usually the estimation of K m shows the lowest reliability, and
both the deviation and CV below 20% are taken as the criteria for reliable estimation of
enzyme kinetic parameters [14]. It was suggested that Eq. (2) was reliable for the
estimation of enzyme kinetic parameters [3–5]. But this classical integrated method was
seldom practiced while it gave negative K m to phosphodiesterase in our laboratory [7]. The
analysis of simulated uricase reaction curves by Eq. (4) yielded many negative kinetic
parameters, even through special strategies was used to reduce common errors. In contrast,
Eq. (3) gave reliable estimation of kinetic parameters by both simulation and
experimentation, and was applicable to characterize enzyme inhibitor. Thus there must
be some inherent differences between Eqs. (1) and (2) for parameter estimation, and the
integrated method with Eq. (1) may be reliable for the estimation of kinetic parameters.
In fact, lower statistic validity of Eq. (2) had already been mentioned before, but there
was no trial of Eq. (1) [4,5]. Results in this report showed that Eq. (1) fitted uricase
reaction curve better than Eq. (2) by both experimentation and simulation, suggesting
higher statistic validity of Eq. (1). It is putative that the predictor variable in the equation
for parameter estimation by common least square fitting should possess no errors and wide
distribution [15]. Eqs. (1) and (2) show obvious difference in the sensitivity of their
predictor variables to common errors. Thus it is likely that different properties of the
predictor variables result in different statistic validities of Eqs. (1) and (2), and
consequently lead to their different reliabilities for the estimation of kinetic parameters.
The predictor variable in Eq. (2) contains random and systematic errors. The integrated
method only applies to steady state reaction. The uncertainty in the lag time of steady state
reaction results in systematic error in reaction time. The predictor variable in Eq. (2) is a
function of reaction time, hence both the intercept and the slope, i.e. V m and K m,
respectively, in Eq. (4) are subject to alteration by this systematic error in reaction time.
The simulation results both here and reported [5] showed this error greatly distorts kinetic
parameters estimated by Eq. (2). Assigning the reaction time at the first datum for analysis
to zero to transform reaction time reduces this systematic error. This is a rather special
treatment for the analysis of enzyme reaction curves, and the results from both simulation
and experimentation showed it effectively improved parameter estimation by Eq. (2). But
by experimentation as long as 90 s is needed to eliminate the effects of this systematic
error on the estimation of kinetic parameter by Eq. (2). Substrate concentration at the first
datum for analysis by Eq. (2) should be above 1.4-fold of enzyme K m and residual
substrate concentration should be low enough for reliable estimation of kinetic parameters
 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24 21

[4]. There should be lower enzyme activity to give initial substrate concentration higher
than the K m after so lone a lag time, and surely lower enzyme activity requires much
longer reaction time to give the desired residual concentration of substrate, which requires
higher stability of the enzyme. Therefore, only the enzyme with lower K m and much
higher stability might be desirable for analysis by Eq. (2). But there is inherent lower limit
for K m by the integrated method, and these factors may make the range of K m suitable for
estimation by Eq. (2) too limited to practice even through reaction time was transformed to
reduce its systematic error. On the other hand, the predictor variable in Eq. (2) is also a
function of substrate concentration. Both the systematic error in the equivalent of initial
substrate concentration and the random error in specified substrate concentration
propagated from experimentation will greatly alter the predictor variable in Eq. (2).
Indeed both simulation results here and reported showed the error in the equivalent of
initial substrate concentration greatly distorts kinetic parameters estimated by Eq. (2) [6].
There might be lower statistic validity to treat the equivalent of initial substrate
concentration as a nonlinear parameter with Eq. (2) because it is included both in the
predictor variable and in the dependent variable. This may be why simulation results
indicated that treating the equivalent of initial substrate concentration as a nonlinear
parameter is by no means satisfactory with Eq. (2) for the estimation of kinetic parameters
[6]. Experimentation with uricase showed at initial substrate concentration above 4-fold of
the K m and residual substrate below 5% of its initial one, Eq. (2) still gives negative K m
even through the equivalent of initial substrate concentration is treated as a nonlinear
parameter besides reaction time is transformed to reduce the systematic error. Therefore, it
is possible that the predictor variable in Eq. (2) is inherently sensitive to common errors,
and this property may reduce the reliability of kinetic parameters estimated by Eq. (2).
Furthermore, there is inherently narrow distribution of the predictor variable in Eq. (2).
The expansion of the distribution of the predictor variable in Eq. (2) requires higher
percentage of substrate consumption within shorter reaction duration, i.e. it requires higher
activity of the enzyme and lower initial substrate concentration. But lower initial
concentration of substrate violates the estimation of kinetic parameters by Eq. (2), and
parameter estimation is much more sensitive to the error in the lag time of reaction at
higher activity of enzyme. By experimentation, the lag time to reach steady state reaction
is so long that higher enzyme activity will surely result in too lower initial substrate
concentration at the first datum under analysis. These contradiction lead to no practical
ways to expand the distribution of the predictor variable in Eq. (2) for reliable estimation
of parameters. Thus it may be concluded that Eq. (2) possesses too lower statistic validity
to give reliable estimation of parameters because its predictor variable does not meet the
statistic requirements for parameter estimation by least square fitting.
In contrast to Eq. (2), the predictor variable in Eq. (1) is reaction time per se that is
essentially free from random error except the systematic error propagated from that in the
lag time to reach steady state reaction. Lower activity of the enzyme with stability can
expand the distribution of the predictor variable of reaction time as wide as desired. The
use of reaction time per se as the predictor variable makes only the intercept in Eq. (3), an
equivalent of Eq. (1), subject to alteration by the systematic error in reaction time. But no
desired parameters are related to this intercept in Eq. (3), hence there will be no effects of
the error in the lag time to reach steady state reaction on parameter estimation by Eq. (1)
22 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24

even the reaction time was not transformed. Therefore Eq. (1) meets the fundamental
statistic requirements for parameter estimation by least square fitting, and it should give
parameters with reliability higher than that by Eq. (2). Indeed, both simulation and
experimentation clearly indicated the reliability of Eq. (3) for the estimation of kinetic
parameters and its resistance to the error in the lag time of steady state reaction. And
experimentation also showed the estimation of kinetic parameters with Eq. (3) always gave
better estimation of kinetic parameters than Eq. (4) did. Moreover, the equivalent of initial
substrate concentration appears only in the dependent variable in Eq. (1) after data
transformation, which provides statistic validity to reduce the effects of the errors in initial
substrate concentration on the estimation of kinetic parameter. This is an advantage over
Lineweaver–Burk plot that is sensitive to the errors in the preset substrate concentrations.
And it also enables the estimation of the equivalent of initial substrate concentration for
kinetic substrate assay [12]. All these results indicated Eq. (1) is statistically valid for
parameter estimation by the integrated method, and it can give reliable kinetic parameters
with resistance to common errors. This may be the intrinsic basis of the reliability of the
integrated method with Eq. (1).
Experimentation with uricase showed if the residual substrate concentration is low
enough, K m slightly below the initial substrate concentration is estimated with consistency
to Lineweaver–Burk plot by fitting reaction curve to Eq. (3). Thus initial substrate
concentrations within much wide range may be used for the estimation of K m of specified
enzyme, and it enables the adjustment of initial substrate concentration within the
specified range for monitoring the reaction curve by common techniques. This is another
advantage over Lineweaver–Burk plot because for most cases neither substrate solubility
nor the sensitivity of common techniques for quantification is a problem. More important
is that with given initial concentration of substrate it provides the possibility for the
estimation of kinetic parameters of enzyme in the presence of competitive inhibitor, which
is the most common type of inhibitor for drug development. Simulation showed that the
major interference for kinetic parameter estimation by Eq. (1) is the outlier in reaction
curve. Automated recording is reliable to give reaction curve with few outliers. Xanthine is
a putative competitive inhibitor of uricase [8,13]. By manual operation with given initial
substrate concentration, the integrated method with Eq. (3) yielded competitive inhibition
of xanthine on uricase, and neither the average nor the precision of the inhibition constant
showed difference from that by the Lineweaver–Burk plot analysis (Table 2). C. utilis
uricase shows K m comparable to that of soybean root nodule uricase for uric acid [13]. The
inhibition constant of xanthine on C. utilis uricase was comparable to that on soybean root
nodule uricase. These results further support the reliability and applicability of the
integrated method with Eq. (1) to estimate enzyme kinetic parameters and to characterize
the inhibitor of specified enzymes. However, Eq. (1) requires much longer computation
time for fitting to the reaction curve. It is possible that the computation resource was too
limited to use Eq. (1) for the estimation of kinetic parameters at the first time to deal with
the integrated method [4–6]. This property may be the principle cause why Eq. (1) can
only be practiced for the estimation kinetic parameters providing the ready availability of
computation resource.
It was found that with reasonable initial concentration of substrate and residual
substrate below one-sixth of the apparent K m, the integrated method with Eq. (1) also
 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24 23

estimated Michaelis–Menten constant of uricase from Candida species in the presence of

xanthine (data not published). With reduced glutathione above 20-fold of NBD-Cl, this
integrated method also yielded consistent K m of mouse liver a-glutathione-S-transferase
for NBD-Cl when initial NBD-Cl concentration was slightly above its K m [11]. These
results further supported the reliability of this integrated method with Eq. (1) for the
estimation of kinetic parameters. Thus it may be concluded that the predictor variable of
reaction time per se in Eq. (1) provides its statistic validity for parameter estimation by the
integrated method, and consequently it ensures the reliability and applicability of this
integrated method for inhibitor characterization in routine practice. The integrated rate
equation of enzyme reaction with the predictor variable of reaction time can be work out
for most cases, and this new integrated method may show some universality. The
optimization of the reaction condition is necessary to ensure the stability of enzyme and
the removal of product inhibition benefit the estimation of enzyme kinetic parameter. The
background or the maximum of product concentration should be treated as a nonlinear
parameter to ensure the resistance of this new integrated method to the systematic error in
initial substrate concentration [12]. It is reasonable to use enzyme activity that ensures the
desired initial concentration of substrate after the necessary lag time for the achievement of
steady state reaction. With these strategies, nonlinear fitting enzyme reaction curves to the
integrated rate equation with the predictor variable of reaction time may be an alternative
method to estimate enzyme kinetic parameters with higher efficiency, lower cost, higher
resistance to common errors and ease to practice, and this method may be useful for the
characterization and the screen of inhibitors of specified target enzymes.

Acknowledgement

This work was supported by a grant from the National Natural Science Foundation of
China (No. 30200266). The linguistic revision by Dr. Ruan X.Z. of the Center for
Nephrology, Royal Free and University College Medical School, University College
London is acknowledged.

References

[1] Bronson DD, Daniels DM, Dixon JT, Redick CC, Haaland PD. Virtual kinetics: using statistical
experimental design for rapid analysis of enzyme inhibitor mechanisms. Biochem Pharmacol 1995;
50:823 – 31.
[2] Kakkar T, Pak Y, Mayersohn M. Evaluation of a minimal experimental design for determination of enzyme
kinetic parameters and inhibition mechanism. J Pharmacol Exp Ther 2000;293:861 – 9.
[3] Lopez-Fidalgo J, Wong WK. Design issue for the Michaelis–Menten Model. J Therm Biol 2001;
215(1):1 – 15.
[4] Atkins GL, Nimmo IA. The reliability of Michaelis–Menten constants and maximum velocities estimated by
using the integrated Michaelis–Menten equation. Biochem J 1973;135:779 – 84.
[5] Orsi BA, Tipton KF. Kinetic analysis of progress curves. Methods Enzymol 1979;63:159 – 83.
[6] Newman PFJ, Atkins GL, Nimmo IA. The effects of systematic error on the accuracy of Michaelis constant
and maximum velocities estimated by using the integrated Michaelis–Menten equation. Biochem J
1974;143:779 – 81.
24 F. Liao et al. / J. Biochem. Biophys. Methods 62 (2005) 13–24

[7] Liao F, Zhang X-P, Zhou Q-X, Yang J-Q, Zeng Z-C, Chen W-Y, et al. Ion-exchange high-performance liquid
chromatographic analysis of the reaction mixture of rabbit liver phosphodiesterase. J Chongqing Univ Med
Sci 2001;26:20 – 2.
[8] Hamilton SD, Pardue HL. Kinetic method having a linear range for substrate concentration that exceed
Michaelis–Menten constants. Clin Chem 1982;28:2359 – 65.
[9] Liao F, Tian K-C, Yang X, Zhou Q-X, Zeng Z-C, Zuo Y-P. Kinetic substrate quantification by nonlinear
fitting the enzyme reaction curve to the integrated Michaelis–Menten equation. Anal Bioanal Chem
2003;375:756 – 62.
[10] Liao F, Liu W-L, Zhou Q-X, Zeng Z-C, Zuo Y-P. Assay of serum arylesterase activity by fitting reaction
curve to an integrated rate equation. Clin Chim Acta 2001;314:67 – 76.
[11] Liao F, Li J-C, Kang G-F, Zeng Z-C, Zuo Y-P. Measurement of mouse liver glutathione S-transferase activity
by the integrated method. J Med Coll PLA 2003;18:295 – 300.
[12] Meyer-Almes FJ, Auer M. Enzyme inhibition assay using fluorescence correlation spectroscopy: a new
algorithm for the derivation of K(cat)/K(M) and K(i) values at substrate concentration much lower than the
Michaelis constant. Biochemistry 2000;39:13261 – 8.
[13] Kahn K, Tipton PA. Kinetic mechanism and cofactor content of soybean root nodule urate oxidase.
Biochemistry 1997;36:4731 – 8.
[14] Cleland WW. Statistical analysis of enzyme kinetic data. Methods Enzymol 1979;63:103 – 38.
[15] Draper NR, Smith H. Applied regression analysis (2nd ed.), Wiley, New York; 1981. p. 7, and p. 122–4.