Biokinetic Models for Representing

the Complete Inhibition of Microbial

Activity at High Substrate Concentrations

Gunaseelan Alagappan, Robert M. Cowan

Department of Environmental Sciences, 14 College Farm Road,
Cook College, Rutgers ± The State University of New Jersey,
New Brunswick, NJ 08901, USA; telephone: 732 932 8750;
fax: 732 932 8644; e-mail: [email protected]
Received 11 November 2000; accepted 18 June 2001

Abstract: This paper reintroduces the Wayman and and engineered systems has seen rapid growth since the
Tseng model for representing substrate inhibition effects introduction of inexpensive computers in the early 1980s.
on speci®c growth rate by further documenting its po-
tential predictive capabilities. It also introduces a modi- Examples of this increasing sophistication, to which this
®cation to this model in which an Andrews inhibition work is relevant, can be found in modeling of bioreactors
function is used in place of the Monod noninhibitory used in waste treatment, bioreactors used for the pro-
substrate function. This modi®cation better represents duction of chemicals, and contaminant fate and trans-
the relationship between speci®c growth rate and sub- port in the environment. The past e€orts of numerous
strate concentration for those substrates that show An-
drews type inhibition at lower substrate concentrations, researchers and practitioners have lead to substantial
rather than the Monod type noninhibitory behavior de- improvements in the ability to use information concern-
scribed in the model of Wayman and Tseng. Results ing the physical and chemical properties of contaminants,
from nonlinear, least squares regression analysis are contaminant mixtures, and the environment to predict
used to evaluate the ability of these models to empiri- the fate and transport of many organic and inorganic
cally represent experimental data (both new and from
the literature). The statistical goodness of ®t is evaluated chemicals (Rubin et al., 1997). The processes accounted
by comparing the regression results against those ob- for in these models include advection, dispersion, di€u-
tained using other empirical models. Finally, possible sion, adsorption, volatilization, other phase partitioning
mechanisms of toxicity responsible for the observed behavior, and the interaction between chemistry and/or
inhibition trends are used to further justify use of these geochemistry of the environment and the dominant mi-
empirical models. The dominant mechanism considered
to be relevant for conceptually explaining complete in- crobiology (i.e., available electron acceptors determine
hibition at high concentrations of solvents is the deteri- which types of microbial communities are dominant),
oration of cell membrane integrity. Literature citations and even an increased use of biodegradation reaction
are used to support this argument. This work should stoichiometry. However, despite this increasing sophis-
lead to improvements in the mathematical modeling of tication in the use of the physical and chemical knowl-
contaminant fate and transport in the environment and
in the simulation of microbial growth and organic com- edge and reaction stoichiometry, there has been little
pound biodegradation in engineered systems. ã 2001 progress made in using increased knowledge of biode-
John Wiley & Sons, Inc. Biotechnol Bioeng 75: 393±405, 2001. gradation kinetics for improving the mathematical
Keywords: substrate inhibition; kinetics; solvent toxicity; modeling of the microbially mediated reactions.
threshold concentration; membrane permeabilization; Typically, fate and transport models express biologi-
mathematical models
cally mediated transformations as ®rst order decay re-
actions wherein the rate is dependent only on the
INTRODUCTION
contaminant concentration and a rate coecient. Often
The use of sophisticated mathematical models to simu- this approach is taken because sparse data prevents the
late compound (pollutant) fate and transport in natural modeler from obtaining statistically relevant parameter
estimates for calibration of more complex models. This
Correspondence to: R. Cowan simpli®ed approach is usually required where ®eld data
Contract grant sponsor: New Jersey Department of Environmental is used. Where more sophisticated biokinetic models
Protection, Division of Science and Research; contract grant number: have been used, the models selected have generally been
48106000052 those of Monod (Eq. 1) and/or Andrews (Eq. 2). These
Contract grant sponsor: New Jersey Agricultural Experiment station
equations are empirically derived functions that are
Contract grant sponsor: U.S. Department of Defense, Advanced
Research Projects Agency, Oce of Naval Research, University Re- applicable to noninhibitory and asymptotic inhibitory
search Initiative; Contract grant number: N00014-92-J-1888, R&T conditions, respectively. The kinetic parameter values
code 40-41 ruri. used in applying the Monod and/or Andrews functions

ã 2001 John Wiley & Sons, Inc.

within a fate and transport model rarely are derived relevant kinetic parameter estimates, particularly when
from ®eld data. Instead, typically they are derived from there is a limited amount of data available (Aitken,
laboratory research data, and often the kinetic param- 1993); the discontinuous nature of some of these models
eter values used are taken from the literature. (Aitken, 1993); a lack of awareness that these alternate
models exist; doubt about the reality of the observed
^
S
l q^
S data that supports the use of these types of models; and
lˆ or q ˆ 1†
KS ‡ S KS ‡ S a lack of awareness that known mechanisms of toxicity
and inhibition could lead to the kinetic response pattern
^
S
l q^
S represented by their data. Furthermore, precedent sup-
lˆ or q ˆ 2† ports application of the Monod and Andrews functions.
KS ‡ S ‡ S2 =KI KS ‡ S ‡ S2 =KI
An exhaustive amount of literature is available in which
Although this approach is often e€ective and the application of these most popular models provides
Monod and Andrews functions often are appropriate for successful representation of experimental data, accept-
describing the e€ect of substrate (contaminant) concen- able simulation of observed process performance in a
tration on microbial activity, it is easy to conceive of wide variety of reactor types (Aitken, 1993; Jones et al.,
situations in which use of these functions would lead to 1973), and successful modeling of subsurface bioreme-
inaccurate, misleading, and nonconservative results. For diation (Schirmer, 1998; Schirmer et al., 1999). This
example, many solvents (including benzene, toluene, widespread application of the Monod and Andrews
ethylbenzene, and the xylenes) are known to be toxic and equations makes it easy to gain acceptance for their use
are capable of causing microbial cell death at concen- in additional situations.
trations below their solubility limit (Sikkema et al., The hypothesis behind this work is that existing ter-
1995). Therefore, in situations where a sucient quantity minal substrate inhibition models provide useful repre-
of solvent is present (e.g., a nonaqueous phase liquid, sentations of the e€ects that high concentrations of
NAPL, such as a pure solvent or gasoline in contact with certain organic compounds have on microbial activity
water) there is the potential for complete inhibition of when those compounds also serve as the electron donor
microbial activity, including that associated with biode- (i.e., substrate), and that a new terminal substrate con-
gradation of the inhibitory contaminant. Furthermore, centration inhibition model (an extension of an existing
the literature contains several examples of experimental model) is needed to obtain reasonable representations of
biodegradation kinetics data that show a clear trend of some experimental data. Thus, this work contains a re-
toxicity consistent with the complete loss of activity. view of the predictive capabilities of several threshold
The Andrews (1968) substrate inhibition model (and inhibition models, including the regression of these
others such as Aiba et al., 1968) describe inhibition be- models to experimental data. The data used include
havior in which the speci®c growth rate and speci®c examples found in the literature and new data that was
substrate removal rate asymptotically approach zero at collected as part of this work. The statistical results from
high substrate concentrations. These models cannot be the least squares regression of the models to the data are
used to e€ectively represent substrate inhibition that used as the basis for comparison of the various terminal
causes a complete loss of activity at a known ®nite substrate inhibition models to each other and to the
substrate concentration (Grady, 1990). Instead, these asymptotic Andrews inhibition model. In some cases the
situations require use of models that predict a more qualitative/visual appearance of the representation is
rapid decrease in activity and the loss of all activity as used together with the statistical analysis results to jus-
the aqueous concentration of the compound is at or tify selection of the best-®t equation. Finally, known
above some ®nite ``critical'' substrate concentration. mechanisms of microbial (solvent) toxicity are discussed
Fortunately, several inhibition models that are capable as they provide further support for the use of these
of ascribing the above-mentioned attribute are available empirical models.
(Han and Levenspiel, 1988; Luong, 1987; Wayman and
Tseng, 1976). However, it is rare to ®nd instances in the
literature where they have been applied. More com- METHODOLOGY
monly these models has been ignored, even when the
data presented would support their use. In these in- Pseudomonas putida Fl (Zylstra and Gibson, 1989) and
stances the data has been left either without a mathe- Burkholderia pickettii PKO1 (Kukor and Kaphammer,
matical representation or has been ®t with noninhibitory 1994) strains were obtained as frozen cultures from Dr.
or asymptotic inhibition models (Moletta et al., 1986; Gerben Zylstra and Dr. Jerome Kukor, respectively, at
Suwa et al., 1994). Rutgers University. They were transferred from frozen
Some of the most likely reasons why these terminal culture to R2A agar plates (APHA et al., 1995) and
concentration inhibition models have largely been ig- incubated overnight at 30°C. Single colonies were picked
nored are their greater number of kinetic parameters and transferred to mineral salts solution (200 mg/L
(Aitken, 1993); the diculty in obtaining statistically CaCl2; 15 mg/L FeCl3 á 6H2O; 1.5 mg/L CoCl2 á 6H2O;

394 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 75, NO. 4, NOVEMBER 20, 2001
1.5 mg/L ZnCl2; 0.5 mg/L CuCl2 á 2H2O; 0.15 mg/L This value was compared with the true growth yield
H3BO3; 0.03 mL concentrated HCl; 150 mg/L obtained for the same microbial population grown on a
MgSO4 á 7H2O; 4.5 mg/L MnSO4 á H2O; 0.5 mg/L low (noninhibitory) concentration of toluene. In all
Na2MoO4 á 2H2O; 4350 mg/L K2HPO4; 3,400 mg/L cases, the yield value observed for the inhibition ki-
KH2PO4; 1,012 mg/L NH4Cl; all of ACS reagent grade netics experiments was within 5% of the true growth
or better) agar (15 g/L of Difco) plates. These were in- yield, suggesting that toxicity did not lead to any
cubated at room temperature in a sealed vessel contain- substantial quantity of toluene undergoing only partial
ing a small beaker of toluene, i.e., toluene was available degradation.
from the vapor phase. Several colonies were picked from The experimentally derived data on speci®c growth
these plates and transferred to a 600 mL Pyrex bottles rate versus toluene concentration, and similar rate ver-
containing 400 mL of sterile mineral salts solution. This sus substrate concentration data collected from the lit-
culture was fed toluene to a known and measured con- erature, were used to verify the utility of a variety of
centration between 15 and 40 mg/L and grown to serve substrate inhibition models. Each substrate inhibition
as the inoculum for the experiments. The inoculum model was used to obtain best-®t parameter estimates
culture and all of the experimental cultures were grown for each data set. The statistical results from the least
at 20°C in an automated respirometer (Comput-Ox 244, squares regression of each model to each data set were
N-CON Systems, Inc., Athens, GA), allowing moni- then used to compare models. Regression analysis was
toring and collection of oxygen consumption data. Care performed using Scientist (MicroMath Inc., Salt Lake
was taken to harvest the inoculum for transfer to the City, UT), a Windows (Microsoft, Redmond, WA)-
experiments during late exponential/early stationary based modeling and curve-®tting program. When a
phase, before substantial biomass decay could occur. In continuous inhibition kinetics model was used, the full
all cases the residual toluene concentration in the inoc- data set was used for simultaneous regression based
ulum culture was estimated to be <1 mg/L. Inoculation estimation of all of the model parameters. However, for
was obtained by transferring 20 to 40 mL of the inoc- the discontinuous models (Wayman and Tseng, and the
ulum culture directly (i.e., without centrifugation or modi®ed Wayman and Tseng model proposed here) it
washing) to 400 mL of MSM. Thus, the maximum was necessary to ®rst obtain the (Monod or Andrews
transfer of residual substrate to the new reactor was 0.04 model) kinetic parameter estimates through regression
mg, or an equivalent of 0.1 mg/L. of the data collected in the lower concentration range
Experiments were initiated when a known quantity of (i.e., from 0 through Sh). The additional parameter es-
toluene, as pure compound, was injected through a timates were then obtained using the data falling in the
Te¯on-lined rubber septum into the culture, and oxygen toxicity inhibition region Sh £ S < Si, i.e., the range
uptake data collection started. Immediately preceding from Sh to the highest concentration at which growth
this step, a small sample of headspace and/or liquid was was observed. The analyses of Monod or Andrews re-
taken and analyzed by gas chromatography with ¯ame gion parameter values and the toxicity/inhibition pa-
ionization detector (GC/FID) and/or chemical oxygen rameter values were performed iteratively until an
demand (COD) analysis using HACH (Loveland, CO) optimum value was obtained for Sh. These optimum Sh
COD vials. Results of these analyses con®rmed that the values and the other stoichiometric and kinetic param-
initial substrate concentrations were close (‹5%) to the eters were de®ned as those providing the lowest sum of
target value. squares of the deviation (SSD) between the full data set
Oxygen uptake data were collected for the length of and the full discontinuous model.
time required to obtain complete removal of the added
toluene, except where no oxygen uptake was observed
due to toxicity. This data, along with knowledge of the RESULTS AND DISCUSSION
initial biomass and toluene concentrations, was used to
obtain estimates of speci®c growth rate and biomass Justi®cation of the Wayman and Tseng Model
yield. The speci®c growth rate, l, was calculated for Substrate Inhibition
through nonlinear regression analysis of the oxygen Wayman and Tseng (1976) developed their model to
uptake data collected before the oxygen transfer rate, illustrate that some inhibitory substrates elicit a sharp,
rather than the microbial growth rate, became limiting almost linear, drop in microbial activity when the sub-
(usually the ®rst 5 to 10 mg/L of oxygen uptake). Each strate concentration exceeds a characteristic threshold
value of l thus estimated was assigned as the speci®c concentration, Sh, thus leading to a complete lack of
growth rate for the corresponding initial toluene con- microbial activity at a terminal concentration, Si. To
centration. The yield value observed for each experi- show this mathematically, they developed the discon-
ment was calculated as the di€erence between the COD tinuous model given in Eqs. (3a), (3b), and (3c):
of the added toluene and the cumulative amount of
oxygen uptake at the time of complete toluene re- ^
S
l
lˆ when S  Sh 3a†
moval, divided by the initial toluene concentration. KS ‡ S

ALAGAPPAN AND COWAN: BIOKINETIC MODELS OF INHIBITION OF MICROBIAL ACTIVITY 395

 the Luong model simpli®ed through ®xing the power
^
S
l
lˆ iS Sh † when Sh < S <ˆ SI 3b† term value to unity (i.e., n = 1) provides an essentially
KS ‡ S identical ®t to that provided by the Wayman and Tseng
model (the resulting curves are virtually indistinguish-
lˆ0 when S  SI 3c† able with little di€erence in the r2 values); and, (iii) as
was reported by Wayman and Tseng (1976), the An-
Equation 3a is the Monod equation; the rectangular drews model provides a poor representation of the ob-
hyperbolic function used to empirically represent the served data. Thus, use of the model of Luong is the
microbial growth response below a de®ned threshold preferred means of representing the data sets of Asthana
concentration, Sh, where no inhibition is observed. A and Kortan because it is a continuous function, and in
discontinuity in the model response sets in when S = its full form it provided the best ®t (lowest r2 ) of the
Sh, and Eq. (3b) is used to represent microbial re- data set for all models tested. Furthermore, when the
sponse at concentrations from Sh to the point where Luong model was applied in its three-parameter form
all activity ceases, i.e., the terminal substrate concen- (where n is set equal to 1) it provided an equivalent
tration, Si. Eq. (3b), is a modi®ed Monod function in representation of the data as the four parameter, dis-
which the value i(S ) Sh) is subtracted from the continuous model of Wayman and Tseng.
Monod term value. The parameter i represents the An attempt is made to show relative ecacies of dif-
rate of increase in the extent of inhibition (essentially ferent models (primarily those of Luong and Wayman
the slope of the line connecting Sh and SI). Eq. (3c) and Tseng) to represent substrate inhibition data ob-
represents the occurrence of complete inhibition at tained in this study and some data found in the litera-
and above the terminal substrate concentration. ture. The goodness of ®t is calculated for each model
Wayman and Tseng demonstrated the utility of their with all the data sets for this purpose. The number of
model using the results of two sets of batch growth degrees of freedom for the Luong and Wayman and
data that had been collected by Asthana (growth on Tseng models were the same because both have four
methanol) and Kortan (growth on n-butanol). Un- parameters. Further, the F-values were calculated with
fortunately, neither of these data sets present a strong the goodness of ®t values for these two models, and the
argument for use of this discontinuous model because probability level at which one model is signi®cantly
the values of Sh are low and no extended zero order better is determined (see Edwards, 1970).
region is evident. A preferable model for ®tting these The published data of PilaÂt and Prokop (1975),
data, as demonstrated later, is the continuous function Velizarov and Beschkov (1998), Rudek M. (1992) in
introduced later by Luong (1987). Herbst and Wiesmann (1996), and toluene inhibition
The Luong (1987) function is presented as Eq. (4). of P. putida Fl shown in this study represent the type
Like the Wayman and Tseng model, it is an extension of of data for which the Wayman and Tseng model is
the Monod equation containing four parameters, and is clearly necessary. Figure 2 presents the PilaÂt and
a terminal substrate inhibition model involving a de®- Prokop data showing the e€ect of various initial
nite terminal high substrate concentration (Sm ) at which methanol (carbon and energy source) concentrations
growth stops. on the growth of the yeast, Candida boidinii 11 Bh at
30°C. A zero order noninhibitory growth region is
^
S
l

S
n prominent, followed by a region in which the activity
lˆ
1 when S < Sm 4a† begins to drop steadily toward complete inactivation.
KS ‡ S Sm
Clearly, the Wayman and Tseng model ®ts this data
much better than the other models considered (Luong
lˆ0 when S  Sm 4b† and Andrews).
Rudek's data (Fig. 3) shows the e€ect of oxygen
The fourth parameter, n, is a power term, the value of (electron acceptor) on dichloroethane-based growth
which determines the shape of the curve as S approaches rate, with the Wayman and Tseng, the Luong, and the
Sm . A reanalysis of the data used by Wayman and Tseng Andrews model ®ts. In this case, the data indicate that
are shown in Figures la (Asthana data) and b (Kortan oxygen is not inhibitory at concentrations high enough
data), including for each the best ®t model curves by that the speci®c growth rate might be considered zero
Wayman and Tseng model, Luong model curve with n order with respect to oxygen concentration (^ l region).
set equal to 1, Luong model curve (with n regressed and However, at dissolved oxygen concentrations greater
therefore not forced to be 1), and Andrews model. The than 13 mg/L, the speci®c growth rate is found to drop
parameter values for each model, and the respective rapidly. Unfortunately, there is no data in the range of
goodness of ®t (r2 ), are given in Table I. dissolved oxygen concentrations between 0 and 2 mg/L,
An examination of Figure 1 and the information given thereby the Ks value is poorly de®ned. The result is a
in Table I reveals that (i) regression analysis with the full virtually identical ®t by the Luong and the Wayman and
Luong model provides the best ®t for each data set; (ii) Tseng models (Table I).

396 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 75, NO. 4, NOVEMBER 20, 2001
ALAGAPPAN AND COWAN: BIOKINETIC MODELS OF INHIBITION OF MICROBIAL ACTIVITY

Table I. Kinetic parameters for the data sets ®tted with Luong, Wayman and Tseng, and Andrews models.

Data Model l
^ KS KI Sm or SI1 n i Sh r2 ´ 10)42
Asthana Andrews 0.31‹0.12 0.7‹0.1 12‹8.5 7.5
Luong 0.2‹0.01 0.03‹0.14 55.6‹4.0 0.84‹0.2 1.0
Luong (n = 1) 0.2‹0.01 0.08‹0.1 56.5‹2 1.0 (®xed) 0.8
Wayman & Tseng 0.21 0.094 56.0 0.004 2.2 1.2
Kortan Andrews 0.36‹0.04 0.01‹0.003 0.3‹0.07 4.0
Luong 0.31‹0.007 0.007‹0.007 1.16‹0.16 1.47‹0.3 0.43
Luong (n = 1) 0.3‹0.006 0.006‹0.0007 1.0 (®xed) 0.67
Wayman & Tseng 0.29‹0.008 0.006‹0.0007 0.95 0.31‹0.01 0.027‹0.02 0.60
PilaÂt and Prokop Andrews 0.19‹0.04 0.12‹0.07 3.5‹1.6 6.7
Luong 0.134 0.05‹0.01 7.97‹0.015 0.43‹0.06 1.8
Wayman & Tseng 0.14‹0.001 0.05‹0.003 8.1 0.03‹0.0015 3.67‹0.14 0.014
Rudek Andrews 1.4 0.6‹0.7 20‹4 300.0
Luong 1.39‹0.2 1.14‹0.66 40.3‹4.8 0.76‹0.32 18.0
Wayman & Tseng 1.0 0.089‹0.089 44.35 0.03‹0.002 12.9‹0.94 19.0
Velizarov and Beschkov Andrews 0.6‹0.2 0.027‹0.02 0.7‹0.4 37.5
Luong 0.5 0.027‹0.018 1.51‹0.13 0.68‹0.13 14.0
Wayman & Tseng 0.38‹0.006 0.0091‹0.001 1.59 0.39‹0.05 0.62‹0.06 3.6
This work, P. putida Fl Andrews 0.63‹0.5 11‹24 168‹216 125.0
Luong 0.4 6.8‹3.4 402‹1.3 0.48‹0.06 49.0
Wayman & Tseng 0.35‹0.01 0.6‹0.07 400 1.0E-3‹1E-4 280‹9 14.0

Note: Error values shown are ‹ one standard deviation or standard error as reported by Scientist.
1
The parameters Sm (Luong model) and SI (Wayman and Tseng model) are the concentrations at which growth/ activity completely ceases.
2 2
r ± Goodness of ®t.
397
 Figure 2. Methanol inhibition of C. boidinii 11 Bh, reported by PilaÂt
and Prokop (1975). Substrate inhibition models proposed by Andrews
± ± ± , Luong ± ± ± , and Wayman and Tseng ббР®t to
the data (d).

putida Fl (Fig. 5), where toluene serves as the sole car-

bon and energy source, also show the trend of a zero-
order e€ect of toluene concentration (^ l region, sug-
gesting it behaves as a noninhibitory substrate) followed
by a region where speci®c growth rate rapidly drops o€
to zero once a threshold concentration is exceeded.

Figure 1. Substrate inhibition data (s) of Asthana (a) and Kortan

(b) as given in Wayman and Tseng (1976). Predictions by Andrews
± ± ± , Wayman and Tseng - - - - , Luong with an n value of
1 ± ± ± ±, and the general Luong ббРmodels are shown.

Velizarov and Beschkov data (Fig. 4) shows the sub-

strate inhibition pattern of glucose on Gluconobacter
oxydans ®t with the three models. These data reveal
that the speci®c growth rate of G. oxydans follows a
Monod model pattern up to 0.62 M (112 g/L). Beyond
this concentration, the growth rate drops rapidly, clearly
suggesting that the discontinuous Wayman and Tseng
model is the appropriate model to represent this inhi- Figure 3. Speci®c growth rate on dichloroethane versus oxygen
concentration data collected by Rudek, as reported by Herbst and
bition trend. Wiesmann (1996). Substrate inhibition models proposed by Andrews
Data collected on the e€ect of high toluene concen- ± ± ± , Luong ± ± ± , and Wayman and Tseng ббР®t to
trations on the speci®c growth rate of Pseudomonas the data (d).

398 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 75, NO. 4, NOVEMBER 20, 2001
Figure 4. Dependence of speci®c growth rate of Gluconobacter oxy-
dans on initial glucose concentration, reported by Velizarov and Figure 5. E€ect of toluene concentration on the speci®c growth rate
Beschkov (1998). Substrate inhibition models proposed by Andrews of P. putida Fl. Substrate inhibition models proposed by Andrews
± ± ± , Luong ± ± ± ± , and Wayman and Tseng ббР®t to ± ± ± , Luong ± ± ± ± , and Wayman and Tseng ббР®t to
the data (j). the data (s).

Best-®t kinetic parameter values and the goodness of Justifying an Extension to the Wayman and Tseng
®t (r2) values for each of the numerical model regression Model for Substrate Inhibition
results (curves) presented in Figures 1±5 are given in While considering models for ®tting experimental data
Table I. Comparisons of the r2 values reveal that the for growth of Burkholderia pickettii PKOl and Pseudo-
Luong model provided a statistically comparable or monas mendocina KR (data not shown) on toluene as a
better ®t than that of Wayman and Tseng model for sole carbon and energy source, a unique inhibition trend
three data sets (Asthana, Kortan, and Rudek), whereas was observed for which no existing model proved ade-
the Wayman and Tseng model provided a much better quate. This situation involved the observation of an
®t for the other three (PilaÂt and Prokop, Velizarov and apparent Andrews type inhibition at moderate toluene
Beschkov, and our P. putida Fl data). The Andrews concentrations followed by a rapid loss of activity
model provided a comparatively poor ®t for every data leading to complete inactivity at higher concentrations.
set. From a close observation of the six data sets, we can Building on the work of Wayman and Tseng (1976), the
infer the conditions under which the Luong model is Andrews function was combined with a linearly de-
preferable and tends to provide a statistically better ®t creasing activity function to form a ®ve-parameter dis-
than the Wayman and Tseng model, and vice versa. continuous model referred to as the modi®ed Wayman
The Luong model is preferred when (i) the zero order and Tseng model. This new model is given as Eq. (5a),
region is not discernible, as in the case of Asthana and (5b), and (5c):
Kortan data sets, and (ii) when the SI/Sh ratio is high.
Based on the six data sets we have considered (Table II),
^
S
l
the Luong model gave a statistically better ®t than that lˆ when S  Sh 5a†
of the Wayman and Tseng model when the SI/Sh ratio is KS ‡ S ‡ S2 =KI
greater than 25. In the case of Rudek's data it is dicult
to judge which of these two models is better. There are ^
S
l
lˆ i
S Sh † when Sh < S < SI
no data available in the ®rst order region, leading to KS ‡ S ‡ S2 =KI
poor estimation of the half saturation concentration 5b†
with both models. The value suggested by the Wayman
and Tseng, however, falls within the typical range of
oxygen half saturation for heterotrophs, i.e., 0.01±0.2 l ˆ 0 when S  SI 5c†
mg/L (Grady et al., 1999), whereas the value based on
the Luong ®t is much higher. Hence, in this case the SI/ This modi®cation allows prediction of a slight drop
Sh ratio is not a good guideline to choose between the in activity before the threshold concentration. Figure 6
Luong and the Wayman and Tseng models. illustrates the e€ect of changing the KI (Andrews

ALAGAPPAN AND COWAN: BIOKINETIC MODELS OF INHIBITION OF MICROBIAL ACTIVITY 399

Table II. Correlation of Sm/Sh ratio to the relative suitability of the Luong or the Wayman and Tseng models.

Data Sm/Sh SSDL/SSDW&T1 P.L. (%)2 Better model

Kortan 35. 0.59 25 Luong
Asthana 25. 0.69 50 Luong
Rudek 3. 0.96 50 Comparable
Velizarov and Beschkov 2. 3.9 5 Wayman and Tseng
PilaÂt and Prokop 2. 125 0.5 Wayman and Tseng
This work, P. Putida F1 1. 3.5 2.5 Wayman and Tseng
1
Ratio of SSD for regression of the data to the Luong model to that for the Wayman and Tseng model. This value is the same as ratio of their
goodness of ®t or F-value, because the number of degrees of freedom for both the models is the same.
2
Probability level at which the better model is signi®cantly so.

The data showing the e€ect of 4-nitrophenol con-

centration on the speci®c growth rate of Pseudomonas
putida PNP1 (Fig. 8) collected by LoÈser et al. (1998) also
lends good support to the proposed model. The data
displays a clear discontinuity between the initial weakly
inhibited region up to 400 mg/L, and a sharp, approx-
imately linearly inhibited region beyond 400 mg/L,
suggesting acute toxicity. It must be acknowledged that
the phenomenological inhibition model proposed by
LoÈser in one of his earlier works (LoÈser and Ray, 1994)
also ®ts the data well. When the model was ®tted to their
data with the parameters reported (^ l = 0.615; KS =
0.145; kl = 1.25; k2 = 140; k3 = 560), a r2 of 0.0026
was obtained. This value is identical to that found using
the Wayman and Tseng model, but higher than the
value obtained using the modi®ed Wayman and Tseng
model (r2 = 0.0009). Although the LoÈser and Ray
model has the bene®t of being continuous, it gave a
statistically inferior ®t and appears to do a relatively
Figure 6. Predictions capability of the proposed modi®ed Wayman and poor job at distinguishing the point of transition be-
Tseng model illustrated for changes in the value of the parameter, KI. tween the two apparently distinct inhibition regions (as
compared with use of the threshold concentration
parameter, Sh, in the proposed model).
inhibition parameter) value on the behavior predicted by The data of Moletta et al. (1986) is presented in
the model. The e€ects of changing the value of the Figure 9, along with the model curves. It represents the
parameters i and Sh are identical to those presented for speci®c rate of methane production based on volatile
the original Wayman and Tseng model. The utility of suspended solids concentration for the methanogenic
this ®ve-parameter discontinuous model is demonstrated treatment of pea bleaching wastewater. In this case, the
using our data, as well as the reanalysis of some litera- proposed model gave a comparable ®t to the Luong
ture data that lend additional support for use of this model but a clearly better ®t than Andrews model. The
model (Table III). slight decrease in the speci®c rate of methane production
The speci®c growth rate versus toluene concentration from a high of 0.8 cm3/g VSS to 0.72 cm3/g VSS over the
data for Burkholderia pickettii PKO1 (Fig. 7) clearly acetate concentration range of 3.0 to 11.5 mg/L is con-
shows that the speci®c growth rate decreases slowly as sistent with the use of an asymptotic inhibition model.
the substrate concentration is increased until a threshold However, the rapid rate of decrease observed at higher
concentration is reached at 256 mg/L of toluene and concentrations is clear.
then a rapid loss of activity occurs. It is possible to see The data from Rozich and Castens (1986; Fig. 10) and
that the growth stops completely at about 420 mg/L of Suwa et al. (1994; Fig. 11) represent the e€ect of initial
toluene. This terminal concentration is well within the ammonium concentration on its oxidation rate. It is
solubility limit of toluene in water for the temperature at apparent that both data sets might be well represented
which this study was carried out (550 mg/L at 20°C). by the proposed model, but the sparse and/or widely
The proposed model gave a better ®t to this data com- variable data leaves this interpretation open to argu-
pared with the Luong or the Andrews models, both ment. The major problems with the Rozich and Castens
visually (Fig. 7) and statistically (Table III). data (Fig. 10) are the lack of measurements in the range

400 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 75, NO. 4, NOVEMBER 20, 2001
 Table III. Kinetic parameters for data sets ®t with the Modi®ed Wayman and Tseng (W&T) model.
ALAGAPPAN AND COWAN: BIOKINETIC MODELS OF INHIBITION OF MICROBIAL ACTIVITY

Data Model l
^ KS KI Sm or SI1 n i Sh r2´ 10)32
Rozich and Castens Andrews 1.2 2.45‹1.1 21‹4.9 40.0
Luong 1.43‹0.53 3.4‹2.1 35.3‹2.8 0.62‹0.37 25.0
W&T 0.72‹0.09 1.23‹0.75 0.065‹0.005 26‹0.044 28.0
Modi®ed W & T 1.2‹0.13 2.7‹1.16 26‹5.5 37.009 0.004 27.6‹0.28 29.0
Suwa et al. Andrews 0.73‹0.31 0.64‹0.43 2.72‹1.97 5.0
Luong Program cannot ®t this model to the data
W&T 0.34‹0.017 0.17‹0.03 0.029‹0.002 4.7‹0.47 0.96
Modi®ed W & T 0.44‹0.044 0.28‹0.06 14‹6 £16.4 0.02‹0.0001 6.5 0.7
0.03‹0.0003 9.5
0.05‹0.0007 12.5
0.15‹0.006 15
This work, B. Pickettii PKO1 Andrews 1.04‹0.67 25‹29 57‹48 6.9
Luong 0.5 4.8‹1.8 412‹6 0.7‹0.017 2.9
W&T 0.38‹0.02 2.4 6.5E-4‹1E-4 223‹27 3.3
Modi®ed W & T 0.44‹0.006 2.4‹0.2 426‹56 397 4.5E-4‹1E-4 259‹22 2.3
LoÈser et al. Andrews 0.73‹0.11 0.27‹0.14 348‹138 14.0
Luong Program cannot ®t this model to the data
W&T 0.54‹0.02 0.12‹0.03 2.5E-3‹8E-4 379‹42 2.6
Modi®ed W &T 0.63‹0.02 0.177‹0.02 1162‹228 2.4E-3‹3.7E-4 415‹13 0.9
Moletta et al Andrews 2.9 4.1 4.1 21.0
Luong 1.5 3.0‹0.3 15.8‹0.4 0.5‹0.06 3.0
W&T 1.12‹0.5 0.86‹0.08 0.11‹0.004 10.1‹0.14 5.8
Modi®ed W & T 1.16‹0.09 1.75 24‹9.4 17.3 0.1‹0.006 11.15‹0.16 3.4

Note: Error values shown are  one standard deviation or standard error as reported by Scientist
1
The parameters Sm (Luong model) and SI (Wayman and Tseng model) are the concentrations at which growth/activity completely ceases.
2 2
r ± Goodness of ®t.
401
Figure 7. E€ect of toluene concentration on the speci®c growth rate
Figure 9. E€ect of un-ionized acetic acid concentration on the
of Berkholderia pickettii PKO1. Best ®t of the data (s) with the
methane production rate from a pea bleaching wastewater (o), re-
Andrews ± ± ± , Luong - - - - , Wayman and Tseng ± ± ±, and
ported by Moletta et al. (1986). Best ®t of the data with the Andrews
modi®ed Wayman and Tseng ббРmodels.
± ± ± , Luong - - - -, Wayman and Tseng ± ± ± , and modi®ed
Wayman and Tseng ббРmodels.
of 2 to 20 mg/L NH‡ 4 -N, and the large range of vari-
ability observed for speci®c growth rates at concentra-
tions between 20 and 30 mg/L NH‡ 4 -N. The data of
envisage the range of inhibition patterns (Fig. 11) pre-
Suwa et al. lack measurements in that range of con- dicted by the proposed model. Under this premise, SI
centration (6±16 mg/L NH‡ 4 -N), where the activity ap-
would be less than or equal to 16.4. Further, Sh could be
parently dropped abruptly. However, it is possible to in the range of 6.5 (the last nonzero data reported) to
some concentration very close to 16.4 mM of ammoni-

Figure 8. Dependence of speci®c growth rate of Pseudomonas putida

PNP1 on 4-nitrophenol concentration, reported by LoÈser et al. (1998).
Best ®t of the data (s) with the Andrews ± ± ± , Loser and Ray - - - -, Figure 10. Data (s) collected by Rozich and Castens (1986) ®t with
Wayman and Tseng ± ± ±, and modi®ed Wayman and Tseng Andrews ± ± ± , Luong - - - , Wayman and Tseng ± ± ±, and
ббРmodels. modi®ed Wayman and Tseng бРmodels.

402 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 75, NO. 4, NOVEMBER 20, 2001
 of electrons across the membranes (dissipation of pro-
ton-motive force), e‚ux of potassium ions from the cell,
and similar e€ects are some of the associated events
identi®ed with toxicants like phenol or 4-chlorophenol
(Heipieper et al., 1991; Silver and Wendt, 1967). Many
researchers e€ectively correlate some physico-chemical
properties of a compound to its toxicity. For example,
Sikkema et al. (1994) showed that the octanol-water
partition coecient (KOW) for cyclic hydrocarbons
could serve as an index for the ability to partition onto
biological membranes. Additionally, Vaishnav and Lo-
pas (1985) demonstrated a parabolic relationship be-
tween log KOW and the experimental microbial EC50
values for 10 primary alcohols (in this context, EC50
refers to the concentration at which the biodegradation
rate is reduced to half the maximum observed rate).
The observed quantitative change in membrane
properties from exposure to di€erent concentrations of a
toxicant is of interest to this work. Findings by Skerlavaj
Figure 11. Andrews, Luong, Wayman and Tseng, and modi®ed et al. (1990) on the e€ect of bactenecins, a class of arg-
Wayman and Tseng models ®t to the ammonia oxidation data of Suwa
et al. (1994). Di€erent possible Sh values were speculated with the
inine-rich antibacterial peptides, on several gram-nega-
modi®ed Wayman and Tseng model. tive bacteria suggest that the decrease in bacterial
viability is causally related to the increase in membrane
permeability and the subsequent fall in respiration-
um sulfate, and the value of i would depend on the Sh linked proton-motive force, with the attendant loss of
and SI concentrations considered. cellular metabolites and macromolecular biosynthetic
Among the di€erent data considered in this section, ability. Smejtek (1987) showed a clear correlation be-
the LoÈser et al. data gives the best support for the tween pentachlorophenol (PCP) concentration, toxicity
proposed model, which is signi®cantly di€erent at the (drop in 14C assimilation), and a decrease in the elec-
5% probability level when compared with the Wayman trical resistance of the cell membrane of Selenastrum
and Tseng model as well as LoÈser and Ray model. With capricornutum, an algal species. Although at lower
the data of Suwa et al. and the toluene inhibition results concentrations there was no detectable e€ect of PCP on
on B. pickettii PKO1 from this study, the proposed activity or electrical resistance (i.e., zero order response
model is signi®cant only at a 50% probability level when to PCP), the same PCP concentration that was associ-
compared with the next best representing models ated with the ®rst detectable changes in membrane re-
(Wayman and Tseng, and Luong, respectively). With sistivity was the associated with the ®rst evidence of
the Moletta et al. and the Rozich and Castens data, the toxicity. This behavior is consistent with a Wayman and
Luong model provides equally good or slightly better Tseng-type relationship between speci®c activity and
representation when compared with the proposed substrate concentration.
model. Similar relationships have been reported to exist be-
tween the occurrence of toxicity and change in mem-
brane properties for other toxic compounds (antibiotics,
Mechanisms that Support the Use of Terminal for instance) that target the cell membrane. According
Substrate Inhibition Models
to Maftah et al. (1993), mesentericin Y105, a bacteriocin
One commonality among most of the substrates for produced by a Leuconostoc mesenteroides strain, dissi-
which the data support use of a terminal substrate in- pates the plasma membrane potential of Listeria mono-
hibition model is that they are organic solvents. Loss of cytogenes and inhibits the transport of leucine and
cell viability due to destruction of the plasma membrane glutamic acid. It also induces an e‚ux of preaccumu-
(i.e., a large increase in membrane permeability) in the lated amino acids from cells, possibly by inducing pore
presence of high concentrations of organic solvents is formation in the energy-transducing membrane. Loss of
well recognized by cell biologists (Flores et al., 1994). selective permeability of the cells seems to occur some-
Cell membrane toxicity has been of considerable interest times without structural damage, as in the case of
in the recent years. A deleterious e€ect on cell mem- chlorine dioxide action on E. coli, destroying the latter's
brane activity is shared by many hydrocarbons, both trans-membrane gradient (Berg et al., 1986). Mycotox-
aliphatic and aromatic, and is one of their chief mech- ins such as trichothecenes were found to have no e€ect
anisms of toxicity (Sikkema et al., 1995). Loss of on biosynthetic (enzyme related) mechanisms including
membrane integrity, resulting in increased passive ¯ux formation of proteins, DNA, or RNA, but deplete the

ALAGAPPAN AND COWAN: BIOKINETIC MODELS OF INHIBITION OF MICROBIAL ACTIVITY 403

cytoplasmic pool of soluble low mass precursors by sidering the use of asymptotic inhibition functions such
a€ecting the permeability of the cell membrane (Rottem as that of Andrews for use in representing substrate
et al., 1984). inhibition e€ects on microbial activity, one should also
These facts strongly indicate that microbial growth consider the use of these terminal substrate inhibition
inhibition need not always be enzyme related. Instead, models. This is particularly true when the substrates
when whole cells are involved, impact on cell organelles studied are organic solvents and/or compounds known
and membranes should also be given due consideration. to be toxic to some microbial communities.
Asymptotic inhibition models such as that of Andrews
cannot e€ectively model these whole cell e€ects, even The authors would like to thank the following people: Pro-
fessors Gerben Zylstra (Rutgers), Ronald Olson (University
empirically. Because the Andrews function is an identi-
of Michigan), and Jerome Kukor (Rutgers, formerly at
cal mathematical formulation to the enzyme inhibition University of Michigan) for supplying the pure culture tol-
function of Haldane (1930), it can be expected to be uene degrading bacteria; Professors David Kosson (Vander-
most capable of accounting for enzyme-related inhibi- bilt University, formerly at Rutgers) and Lily Young
tion. With both the Wayman and Tseng model and the (Rutgers) for providing funding through their DARPA
new model proposed in this work, the discontinuity grant; and Professor Peter Strom (Rutgers) for valuable
feedback on this manuscript. This is NJAES document
permits representation of the kind of growth inhibition D-07131-2000-02.
expected for substrates that a€ect membrane perme-
ability, or some other whole cell e€ect that may be in-
dependent of catabolic processing of the substrate. NOMENCLATURE
When the extent of membrane permeability or related
parameters associated with the concentration of the in- NAPL nonaqueous phase liquid
hibitory substrates is considered, the threshold substrate q speci®c substrate utilization rate (M/M T)
q maximum speci®c substrate utilization rate (M/M T)
concentration Sh can be viewed as analogous to the
KS half saturate coecient, concentration (M/L3 )
threshold permeability conferred to the membrane. This S substrate concentration (M/L3 )
concentration may coincide with the point where the KI Andrews inhibition coecient, concentration (M/L3 )
permeability increase starts to result in the loss of me- Sh threshold concentration below which the
tabolites, or distortion of trans-membrane ionic gradi- organism shows no or slight inhibition (M/L3 )
i inhibition constant (L3 /M T)
ents, as was shown by Smejtek (1987) to occur with
SI total inhibition concentration for the Way-
PCP. As it has been shown with some chemicals af- man and Tseng and modi®ed Wayman SI =
fecting membrane integrity, the inhibitory e€ect on cel- and Tseng models (SI ˆ Sh ‡ lh= i † (M/L)3
lular activity shows up only beyond a certain threshold Sm maximum substrate concentration above
concentration. This mechanistic activity threshold jus- which growth ceases (M/L3 )
n constant
ti®es the use of a discontinuous function in these em-
pirical models (Wayman and Tseng and its proposed Greek symbols
modi®cation). l speci®c growth rate (1/T)
l
^ maximum speci®c growth rate coecient (1/T)
lh speci®c growth rate corresponding to Sh (1/T)

CONCLUSIONS
References
Terminal inhibition concentration models are appro-
priate for situations in which microbial activity drops to Aiba S, Shoda M, Nagatani M. 1968. Kinetics of product inhibition in
zero at a ®nite substrate concentration, whereas the as- alcohol kinetics. Biotechnol Bioeng 10:845±864.
Aitken MD. 1993. Batch biological treatment of inhibitory substrates.
ymptotic Andrews function will predict activity beyond
J Environ Eng 119:855±869.
this. They should be relevant in modeling microbial Andrews JF. 1968. A mathematical model for the continuous culture
activity when high concentrations of hydrophobic of microorganisms utilizing inhibitory substrates. Biotechnol
chemicals are present, and in particular, when these are Bioeng 10:707±723.
present as a NAPL, provided partitioning of these APHA, AWWA, WEF. 1995. Microbiological examination. In: Eatori
chemicals between the NAPL and surrounding aqueous AD, Clesceri LS, Greenberg AE, editors, Standard methods for
the examination of water and wastewater, 19th ed. Washington,
phase could result in suciently high aqueous phase DC: American Public Health Association, American Water Works
concentrations. Solvents that were used as substrates Association, Water Environment Federation. p 9±32.
were seen to elicit inhibition trends that can be repre- Berg JD, Roberts PV, Matin A. 1986. E€ect of chlorine dioxide on
sented by terminal substrate inhibition models such as selected membrane functions of Eschericia coli. J Appl Bacteriol
those of Luong, Wayman and Tseng, or the modi®ca- 60:213±220.
Edwards VH. 1970. The in¯uence of high substrate concentrations on
tion of the Wayman and Tseng model introduced in this
microbial kinetics. Biotechnol Bioeng 12:679±712.
work. Studies on cell membrane alteration by solvents Flores MV, Voget CE, Ertola RJJ. 1994. Permeabilization of yeast
appear to corroborate the inhibition patterns observed cells (Kluyveromyces lactis) with organic solvents. Enzyme Microb
on their transformation. Therefore, in addition to con- Technol l6:340±346.

404 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 75, NO. 4, NOVEMBER 20, 2001
Grady CPL, Jr, 1990. Biodegradation of toxic organics: status and Rozich AF, Castens DJ. 1986. Inhibition kinetics of nitri®cation in
potential. J Environ Eng 116:805±828. continuous-¯ow reactors. J Water Pollut Control Fed 58:220±
Grady CP, Jr, Daigger GT, Lim HC. 1999. Biological wastewater 226.
treatment. 2nd, Marcel Dekker. Rubin H, Rathfelder K, Abriola LM. 1997. Modelling quasi-steady
Haldane JBS. 1930. Enzymes. London ed. New york: Longman. NAPL dissolution in fractured permeable media. J Environ Eng
Han K, Levenspiel O. 1988. Extended Monod kinetics for substrate, 123:205±216.
product, and cell inhibition. Biotechnol Bioeng 32:430±437. Rudek M. 1992. Mikrobieller abbau von 1,2-dichloroethan (microbial
Heipieper H-J, Keweloh H, Rehm H-J. 1991. In¯uence of phenols on degradation of 1,2-dichloroethane). Diplomarbeit, TU Berlin.
growth and membrane permeability of free and immobilized Esc- Schirmer M. 1998. Investigation of multi-scale biodegradation pro-
herichia coli. Appl Environ Microbiol 57:1213±1217. cesses: modelling approach. PhD thesis, University of Waterloo,
Herbst B, Wiesmann U. 1996. Kinetics and reaction engineering as- Ontario, Canada.
pects of the biodegradation of dichloromethane and dichloroe- Schirmer M, Butler BJ, Roy JW, Frind EO, Barker JF. 1999. A relative
thane. Water Res 30:1069±1076. least-squares technique to determine unique Monod kinetic pa-
Jones GL, Jansen F, McKay AJ. 1973. Substrate inhibition of growth rameters of BTEX compounds using batch experiments. J Con-
of bacterium NCIB 8250 by phenol. J Gen Microbiol 74:139± taminant Hydrol 37:69±86.
148. Sikkema J, de Bont JAM, Poolman B. 1994. Interactions of cyclic
Kukor JL, Kaphammer BA. 1994. Novel toluene-3-monooxygenase hydrocarbons with biological membranes. J Biol Chem 269:8022±
pathway cloned from Pseudomonas pickettii PKO1. J Bacteriol 8028.
176, 3749±3756. Sikkema J, de Bont JAM, Poolman B. 1995. Mechanisms of mem-
LoÈser C, Oubelli MA, Hertel T. 1998. Growth kinetics of the 4-ni- brane toxicity of hydrocarbons. Microbiol Rev 59:201±222.
trophenol degrading strain Pseudomonas putida PNP1. Acta Bio- Silver S, Wendt L. 1967. Mechanism of action of phenethyl alcohol:
technol 18:29±41. breakdown of the cellular permeability barrier. J Bacteriol 93:560±
LoÈser C, Ray P. 1994. Modelling the aerobic degradation of toluene in 566.
a closed chemostat with and without a head space outlet. Acta Skerlavaj B, Romeo D, Gennaro R. 1990. Rapid membrane permea-
Biotechnol 14:355±366. bilization and inhibition of vital functions of Gram-negative bac-
Luong JHT. 1987. Generalization of Monod kinetics for analysis of teria by bactenecins. Infection Immunity. 58:3724±3730.
growth data with substrate inhibition. Biotechnol Bioeng 29:242± Smejtek P. 1987. The physicochemical basis of the membrane toxicity
248. of pentachlorophenol: an overview. J Membrane Sci 33:249±268.
Maftah A, Renault D, Vignoles C, Hechard Y, Bressollier P, Ratinaud SuwaY, Imamura Y, Suzuki T, Tahiro T, Urushigawa Y. 1994. Am-
MH, Cenatiempo Y, Julien R. 1993. Membrane permeabilization monia-oxidizing bacteria with di€erent sensitivities to (NH4)2SO4
of Listeria monocytogenes and mitochondria by the bacteriocin in activated sludges. Water Res 28:1523±1532.
mesentericin Y105. J Bacteriol 175:3232±3235. Vaishnav DM, Lopas DM. 1985. Relationship between lipophilicity
Moletta R, Verrier D, Albagnac G. 1986. Dynamic modelling of an- and biodegradation inhibition of selected industrial chemicals. Dev
aerobic digestion. Water Res 20:427±434. Ind Microbiol 26:557±565.
Monod J. 1949. The growth of bacterial cultures. Ann Rev Microbiol Velizarov S, Beschkov V. 1998. Biotransformation of glucose to free
3:371±394. gluconic acid by Gluconobacter oxidans: substrate and product
PilaÂt P, Prokop A. 1975. The e€ect of methanol, formaldehyde, and inhibition situations. Process Biochem 33:527±534.
formic acid on growth of Candida boidinii 11 Bh. Biotechnol Wayman M, Tseng MC. 1976. Inhibition-threshold substrate concen-
Bioeng 16:1717±1728. trations. Biotechnol Bioeng 18:383±387.
Rottem S, Yagen B, Katznell A. 1984. E€ect of trichothecenes on Zylstra GJ, Gibson DT. 1989. Toluene degradation by Pseudomonas
growth and intracellular pool size of Mycoplasma gallisepticum. putida Fl. Nucleotide sequence of the todClC2BADE genes and
FEBS Lett 175:189±192. their expression in Escherichia coli. J Biol Chem 264:14940±14946.

ALAGAPPAN AND COWAN: BIOKINETIC MODELS OF INHIBITION OF MICROBIAL ACTIVITY 405