ARTICLE

Thermodynamic Electron Equivalents

Model for Bacterial Yield Prediction:
Modifications and Comparative Evaluations
Perry L. McCarty
Silas H. Palmer Professor Emeritus, Department of Civil and Environmental Engineering,
Stanford University, Stanford, California 94305-4020; telephone: (650) 723-4131;
fax: (650) 725-3164; e-mail: pmccarty@stanford.edu
Received 2 March 2006; accepted 11 October 2006
Published online 6 November 2006 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/bit.21250

ABSTRACT: Modifications are made to an earlier thermodynamic model (TEEM1) for prediction of maximum
microbial yields from aerobic and anaerobic as well as
heterotrophic and autotrophic growth. The revised model
(TEEM2) corrects for lower yields found with aerobic
oxidations of organic compounds where an oxygenase is
involved and with growth on single-carbon (C1) compounds. TEEM1 and TEEM2 are based on energy release
and consumption as determined from the reduction potential or Gibbs free energy of -reaction reduction equations
together with losses of energy during energy transfer. Energy
transfer efficiency is a key parameter needed to make
predictions with TEEM2, and was determined through
evaluations with extensive data sets on aerobic heterotrophic
yield available in the literature. For compounds following
normal catabolic pathways, the best-fit value for energy
transfer efficiency was 0.37, which permitted accurate
predictions of growth with a precision of 15%20% as
determined by standard deviation. Using the same energy
transfer efficiency, a similar precision, but somewhat less
accuracy was found for organic compounds where oxidation
involves an oxygenase (estimates 8% too high) and for C1
compounds (estimates 17% too high). In spite of the somewhat lower accuracy, the TEEM2 modifications resulted in
improved predictions over TEEM1 and the comparison
models.
Biotechnol. Bioeng. 2007;97: 377388.
! 2006 Wiley Periodicals, Inc.
KEYWORDS: bacterial yields; thermodynamics; autotrophic; heterotrophic; oxygenase; C1 compounds

Introduction

and anaerobic as well as aerobic growth was presented

(McCarty, 1965). This model, using oxygen equivalents,
considered energy for cell synthesis from various carbon
sources, energy available from substrate transformation, and
efficiency of energy transfer. The model was modified later
(McCarty, 1971) to change from oxygen equivalents to
electron equivalents, which was felt more fundamental as
well as useful for constructing balanced stoichiometric
equations for growth and substrate utilization (McCarty,
1975), and a further modification was made (Rittmann and
McCarty, 2001) to change the way yields are calculated when
nitrogen sources other than ammonia are used. This earlier
method and its modifications will be referred to here for
simplification as the thermodynamic electron equivalents
model one (TEEM1).
Recently, VanBriesen and colleagues have noted limitations of TEEM1 in its application for aerobic heterotrophic
growth and proposed modifications to address these
limitations. One limitation noted was for reactions that
involve the use of an oxygenase as an initial step in
substrate oxidation, resulting in some wasted energy for
these reactions (VanBriesen, 2001; Yuan and VanBriesen,
2002). More recently, Xiao and VanBriesen (2006) developed a significantly modified procedure to address inaccurate
predictions observed for certain compounds with low
degrees of reduction, which they suggested was caused by
carbon-limited rather than energy-limited growth. An
alternative explanation explored here is that these errors
are associated primarily with single carbon (C1) compounds
that follow the serene or the ribulose monophosphate
pathway in synthesis.
In this article, modifications are made to TEEM1 to
address the limitations noted for aerobic heterotrophic
reactions that involve an oxygenase or a C1 compound. The

In 1965, a thermodynamic model for predicting maximal

bacterial yields from autotrophic as well as heterotrophic,
Correspondence to: P.L. McCarty

! 2006 Wiley Periodicals, Inc.

Biotechnology and Bioengineering, Vol. 97, No. 2, June 1, 2007

377

model is then compared for accuracy and variance in

predictions with modifications proposed by VanBriesen and
colleagues (VanBriesen, 2001; Xiao and VanBriesen, 2006;
Yuan and VanBriesen, 2002).

TEEM1 Development
TEEM1 has been presented in detail (Rittmann and
McCarty, 2001). The model is based upon the use of
half-reaction reductive equations for electron donors and
acceptors as well as for cell synthesis, and the associated
Gibbs free energies or reduction potentials for the half
reactions. Use of such half reactions is illustrated in the
upper portion of Figure 1. The methods for developing half
reactions and computing the half-reaction reduction
potentials (DGo0 , kJ/eeq) are provided by Rittmann and
McCarty (2001). Values for the compounds evaluated in this
article are given in the Appendix, and were developed using
the standard free energy of formation for each compound
(DGof ) also listed in the Appendix. Half reactions for an
electron donor and an electron acceptor can be combined to
produce an energy reaction with its associated Gibbs free
energy (DGr). Half reactions for electron donor and cell
synthesis can be combined to produce the synthesis reaction,
from which the Gibbs free energy for synthesis is derived
(DGs). An overall reaction for cell growth is obtained by
combining in proper proportion the energy reaction and
synthesis reactions. This proportion depends upon the
energy transfer efficiency (e) and is represented by A, a value
that is obtained by consideration of the energy released by
the energy reaction and that required for synthesis:
DGs
A"
"DGr

(1)

A in essence gives the ratio of the fraction of a donor

associated with energy production to that associated with
cell synthesis. DGr is determined from the half reaction
reduction potentials for the electron acceptor (DGa)
and that of the electron donor (DGd), as well as from the
activity of reactants and products, {i}, and their respective
stoichiometric coefficients, vi:
DGr DGor RT

vi lnfig

(2)

DGor DGoa " DGod

(3)

where R is the universal gas constant, T is absolute

temperature, and DGoa , DGod , DGor are the reduction
potentials for the electron acceptor half-reaction, the
electron donor half-reaction, and the overall energy
reaction, respectively. Since most microbial reactions of
interest occur at pH 7, the reduction potential is modified so
that the activity {i} for {H} is taken as 10"7. For this case, a

378

prime is added to the superscript (DGoj ). This will be

referred to here as the standard case since it is the one used
throughout this article and in the models under comparison.
In many cases for predicting bacterial yields, especially with
aerobic heterotrophic reactions, there is little difference
0
between DGrand DGor , but this is often not the case in
some anaerobic and autotrophic reactions. For these cases,
the correction indicated by the right hand term on the right
side in Equation 2 must be applied. Since all reactions
0
evaluated here were aerobic, DGoa equals the half reaction
reduction potential for oxygen of "78.72 kJ/eeq.
DGs in TEEM1 consists of two energy terms, one for the
conversion of the electron donor to an intermediate
compound, (DGic), and another for conversion of the
intermediate to cells (DGpc):
DGs

DGic DGpc

"n
"

(4)

Energy may be required to convert the cell carbon source to

the intermediate compound (DGic > 0), in which case n 1,
or it may be obtained from the conversion itself when
(DGic < 0), in which case n "1. In TEEM1, the
intermediate compound was taken to be pyruvate as this
compounds half-reaction reduction potential (35.09 kJ/
eeq) was originally thought to be close to that of acetylCoA, which was considered the main intermediate in
synthesis. However, it is now known that the difference
between the two is quite significant, and so it is suggested
that the reduction potential for acetyl-CoA itself be
used, which is "0.32 V (Madigan et al., 1997) or 30.9 kJ/
eeq. Thus,
DGic DGin " DGd 30:9 " DGd

(5)

This does not make a great deal of difference in calculated

yields, but is felt to be a better theoretical choice.
DGpc was estimated from reported values of ATP in moles
required for cell synthesis, and with an assumed cell relative
composition of C5H7O2N was set equal to 18.8 kJ/eeq when
ammonia serves as the source for cell synthesis. This value, as
well as the cell synthesis half reaction, differs when other
forms of nitrogen are used for synthesis (Rittmann and
McCarty, 2001). Since cell yields reported here all resulted
with ammonia present, the model comparisons here use
only the 18.8 kJ/eeq value.
The true or maximum cell yield (Y) is determined from A.
Yield can be expressed in electron equivalent (eeq) units as
the fraction of donor electron equivalents converted
for synthesis f os , while the remaining fraction is used for
energy ( f oe ):
A

feo o
1
A
; fo
; and
; f
1A e
1A
fso s

fso feo 1 (6)

Commonly, cell yield is represented in other units. That

often used by others developing thermodynamic models for

Biotechnology and Bioengineering, Vol. 97, No. 2, June 1, 2007

DOI 10.1002/bit

Figure 1. Example of yield calculations for aerobic oxidation of acetate with e 0.37. Here, either TEEM1 or TEEM2 could be used since neither an oxygenase nor a C1
compound is involved.

McCarty: TEEM Modifications for Bacterial Yield Predictions

Biotechnology and Bioengineering. DOI 10.1002/bit

379

yield estimation is YC/C, which is yield expressed as moles of

cell carbon formed per mole carbon of the donor substrate.
This term only applies for heterotrophic reactions. For
autotrophic reactions, it is common to express yields in
moles of cell carbon per mole of electron donor used (YC/M).
When using electron equivalents, fos serves whether the
reaction is heterotrophic or autotrophic. As illustrated in
Figure 1, fos is also useful for constructing overall balanced
stoichiometric equations for cell growth and substrate
utilization. VanBriesen (2002) has shown that YC/C can
readily be obtained by multiplying fos by the ratio of the
degree of reduction of carbon in the donor (gd) to that in
cells (gx):
YC=C

! "
gd o
f
gx s

(7)

Here, the degree of reduction of carbon in an organic

compound is equal to p, the number of electron equivalents
in a mole of the compound as determined from the reaction reduction equation, divided by the number of
carbon atoms in the compound (C). Combining Equations
1, 6, and 7, the carbon yield for heterotrophic growth is
given by,
YC=C

gd
DGr
g x DGr " DGs ="

(8)

Modifications in TEEM2
Two basic modifications are made to TEEM1 to develop
TEEM2, the first addresses aerobic heterotrophic reactions
involving oxygenases, and the second addresses aerobic
heterotrophic oxidation of C1 organic compounds.
Oxygenase Modification
Modifications to TEEM1 to address organic electron donor
oxidation reactions involving oxygenases were developed
by VanBriesen (2001) and Yuan and VanBriesen (2002).
Oxygenases are generally used by aerobic microorganisms in
the initial oxidative steps for hydrocarbons, aromatic
compounds, and ammonia to convert them to forms that
can be used for energy. Oxygenase reactions generally
require the input of energy and reducing power in the
form of NADH. The above authors made corrections for
the energy losses involved in oxygenase reactions by
considering in detail the biochemical steps involved in
the oxidation process. This is appropriate for obtaining
information on energy losses. A somewhat simpler approach
was incorporated into TEEM2 by assuming that the energy
loss involved in each step involving an oxygenase is equal to
the standard energy associated with NADH oxidation. This
is represented by the difference between that of the oxygen
reduction equation ("78.72 kJ/eeq) and that of the NADH/

380

NAD reduction equation (30.88 kJ/eeq), which equals

"109.6 kJ/eeq or "219.2 kJ/mol NADH. Thus, the energy
loss in kJ per mole of donor is qDGxy, where q is the number
of times an oxygenase is used in the complete oxidation
of the compound and DGxy equals "219.2 kJ/mol. The
energy available for synthesis from the energy reaction then
becomes:
! "
q
DGr DGa " DGd "
DGxy
p

(9)

Modification for C1 Compounds and Oxalate

Xiao and VanBriesen (2006) noted that yield estimates with
TEEM1 for certain compounds with low degrees of
reduction are not accurate. They proposed modifications
that involve a significantly different method for making
carbon and energy balances. While their modification
improves estimates for compounds with a low degree of
reduction, it offers little change for compounds with high
degrees of reduction, some of which are also associated with
large prediction inaccuracies. Here, a different approach is
proposed to address such inaccuracies. It is hypothesized
that the deficiency noted is not a function of the
compounds degree of reduction, but rather due to the
different synthesis pathways taken for C1 compounds such
as formate and methanol. The aerobic synthesis pathway for
these compounds usually involves first conversion to
formaldehyde, which then is converted into cellular
material, either through the ribose monophosphate pathway
or the serine pathway (Madigan et al., 1997). Energy is lost in
the conversion to formaldehyde, and this needs consideration. Also, additional energy is lost with compounds
following the serine pathway, as this requires the input of
NADH. Consequently, yields are even less for organisms
using both C1 compounds and the serine pathway.
For example, the yields for Type II methanotrophs
that synthesize cells from methane through the serine
pathway are much less than that of Type I methanotrophs
that use the ribose monophosphate pathway (Madigan et al.,
1997).
The modification for C1 compounds proposed here
addresses only the loss of energy from conversion of
substrate to formaldehyde. A modification similar to that for
oxygenases can be used additionally for those using the
serine pathway. The energy loss through conversion to
formaldehyde (DGfa) is provided here through its incorporation into DGs:
DGs

DGfa " DGd DGin " DGfa DGpc

"m
"n
"

(10)

For C1 compounds, DGfa equals the half-reaction Gibbs free

energy for formaldehyde (46.53 kJ/eeq), while for other
compounds, DGfa is zero. The Gibbs free energy for a

Biotechnology and Bioengineering, Vol. 97, No. 2, June 1, 2007

DOI 10.1002/bit

reaction is either multiplied or divided by the energy transfer

efficiency, depending upon whether the reaction provides
energy to the organism or consumes it, respectively. The
exponents m and n are used to make this adjustment. This is
accomplished as follows: m equals 1 for C1 compounds, and
equals n for all others; the exponent n is similar as before
and equals 1 if m n and (DGin " DGd) > 0, otherwise it
equals "1. An example calculation is provided in Figure 1
and the summary equations for TEEM2 are contained in
Figure 2.
Oxalate is another compound that presents an exception
to the general pathway for synthesis and energy production
followed by most organic compounds. Oxalate is one of the
most oxidized organic compounds, having a degree of
reduction of 1.0. Many organisms that can use oxalate for
energy and synthesis also use formate (Blackmore and
Quayle, 1970). Oxalate is a high-energy compound with
DGo0 of 52.1 kJ/eeq. However, in energy metabolism, much
of this energy is lost as oxalate is first transformed into
formate, which then is used for energy. The synthesis with
oxalate is also unique and may involve the glyoxylate or
serine pathways (Blackmore and Quayle, 1970; Kornberg,
1966). The modification in TEEM2 proposed here for this
compound is to assume it has the same DGo0 as formate of
39.19 kJ/eeq and follows the same synthesis pathway as
formate.

TEEM2 Calibration and Comparisons

The effort made initially was to obtain a better calibration on
the energy transfer efficiency value that might be used in
TEEM2 for prediction purposes for aerobic heterotrophic
growth. The large sets of yield information provided by
others (Heijnen and van Dijken, 1992; VanBriesen, 2001;
Xiao and VanBriesen, 2006; Yuan and VanBriesen, 2002)
were used to calibrate e and to compare the predictive
capability of TEEM2 with other models.
In order to make comparisons between models, the error
defined as the difference between a reported value and a
predicted value by each model is calculated:
Error

Reported Value " Predicted Value

Reported Value

(11)

The average error and the standard deviation of the error

was determined for each data set and model under
comparison. The average error provides an indication of
a models accuracy in prediction, while the standard
deviation of the error provides a measure of the variance
or precision of the prediction.
One difficulty in making true comparisons between the
different models is that different assumptions are often
made in values chosen for the models. The question then is
whether the differences noted result from the basic structure
of a model itself or from values assumed for the model. For
example, in the comparison set, the authors generally

assume that standard conditions (unit activities) exist for all

reactants and products, except for [H], which is assumed
to equal 10"7 molar. However, different investigators tend
to use different forms for the half-reactions, thus producing
somewhat different half-reaction reduction potentials. It
would be better to use the true activities for relevant species,
but these often are not known. An analysis was made of the
impact of these differences in the models tested and this
difference was found not to have a significant effect on the
comparative analysis made here. However, impacts might
be quite different in overall reactions that yield relatively
little energy, such as some anaerobic reactions. Thus, an
understanding of this issue is necessary with general use of
the predictive equations. Another difference is in the cell
formulations used. Different empirical formula have been
reported by different investigators with different organisms
fed on different substrates (Rittmann and McCarty, 2001).
An average or typical formulation of C5H7O2N
(CH1.4O0.4N0.2), which has a degree of reduction of 4.0,
is the formulation used for this article. In the comparison
approach, Xiao and VanBriesen (2006) used CH2O0.6N0.2
which has a degree of reduction for carbon of 4.2. This
difference is relatively small and did not impact the
comparisons made here greatly, although it could affect
somewhat the best-fit value for the energy transfer efficiency.
In the following, model calibrations are made through the
selection for a given data set of the best fit value for the
energy transfer efficiency, e, a value that brings the average
error for a data set close to zero. Xiao and VanBriesen (2006)
estimated the value for e for heterotrophic aerobic oxidation
from known ATP production for 10 different substrates. An
analysis of these data gives an average value of 0.39 with 95%
confidence limits of 0.380.40. They also evaluated e from
six measurements of yield from aerobic oxidation of acetate,
yielding an average value of 0.41 with 95% confidence limits
of 0.350.47. While they selected a value of 0.41 to use in
their equations, this is not necessarily the optimum value to
use because the true value possibly lies elsewhere within the
confidence limits. In addition, models are simplifications of
reality and as such have inherent errors that are dependent
to some degree on the assumptions made as already noted.
Thus, for a predictive model, the best value to use for e for
good predictions may be somewhat different than the true
value. One of the purposes of this study was to find
which value of e for TEEM2 provides the best value for
prediction of yields for aerobic heterotrophic growth
with organic compounds following normal catabolic
processes. The accuracy and precision are then determined
for TEEM2 model predictions for aerobic heterotrophic
reactions involving an oxygenase and then for C1
compounds. Results are compared with those of other
model predictions.

Energy Transfer Efficiency Evaluation

The extensive YC/C values from Heijnen and van Dijken
(1992) for aerobic heterotrophic growth with both pure and

McCarty: TEEM Modifications for Bacterial Yield Predictions

Biotechnology and Bioengineering. DOI 10.1002/bit

381

Figure 2.

382

Summary of equations for TEEM2.

Biotechnology and Bioengineering, Vol. 97, No. 2, June 1, 2007

DOI 10.1002/bit

mixed cultures were used to determine the best-fit energy

transfer efficiency (e ) for the TEEM model. Here, all C1
compounds and compounds known to involve oxygenases
were excluded so that they would not bias the selection of e.
The best-fit value for e so found was then used in the
comparative evaluation of TEEM2 for organic compounds
involving oxygenases and for C1 compounds. Table I
contains a summary of this analysis for pure cultures and
Table II that for mixed cultures. Two separate evaluations
were made. First, the value for e that would produce the
reported YC/C for each substrate using TEEM1 was
determined and the results are listed as e implied in Tables
I and II. This required solving for e using a combination of

Table II. Mixed culture e evaluation.

Reported
YC/C&

Substrate
2"

0.375
0.365
0.400
0.510
0.610
0.510
0.410
0.560
0.670
0.480
0.445
0.530
0.575

Malate
Citrate3"
Succinate2"
Gluconate"
Glucose
Lactate"
Acetate"
Mannitol
Glycerol
Propionate"
Acetone
Ethanol
Propanol
Average
Std. Dev.
Number

Calculated YC/C e 0.37

e Implied

fs

YC/CCmol/Cmol

Error

0.40
0.40
0.40
0.43
0.46
0.42
0.38
0.40
0.45
0.38
0.30
0.31
0.34
0.39
0.05
13.00

0.46
0.45
0.42
0.48
0.49
0.45
0.39
0.48
0.48
0.40
0.42
0.45
0.43

0.343
0.340
0.365
0.441
0.486
0.450
0.394
0.520
0.555
0.463
0.566
0.668
0.644

0.09
0.07
0.09
0.14
0.20
0.12
0.04
0.07
0.17
0.04
"0.27
"0.26
"0.12
0.03
0.15
13.00

&

Yield data from Heijnen and van Dijken (1992).

Table I. Pure culture e evaluation.

Calculated YC/C e 0.37
Substrate
Pseudomonas oxalaticus
Glyoxylate"
Tartrate2"
Malonate2"
Citrate3"
Succinate2"
Acetate"
Fructose
Glycerol
Ethanol
Candida utilus
Citrate3"
Pyruvate"
Succinate2"
Gluconate"
Glucose
Xylose
Acetate"
Glycerol
Acetoin
23 Butanediol
Ethanol
Paracoccus denitrificans
Malate2Succinate2GluconateMannitol
Thiobacillus acidophilus
l-Malate2PyruvateGlucose
Glycerol
Average
Std. Dev.
Number
&

Reported
e Implied
YC/C&

fs

YC/C
Cmol/Cmol

Equations 4 and 8, yielding:

Error

0.220
0.280
0.238
0.390
0.385
0.406
0.505
0.569
0.558

0.32
0.35
0.32
0.42
0.39
0.38
0.38
0.38
0.32

0.53
0.48
0.43
0.45
0.42
0.39
0.49
0.48
0.45

0.263
0.301
0.285
0.340
0.365
0.394
0.486
0.555
0.668

"0.20
"0.08
"0.20
0.13
0.05
0.03
0.04
0.02
"0.20

0.441
0.434
0.448
0.559
0.595
0.490
0.455
0.692
0.424
0.446
0.617

0.48
0.42
0.45
0.47
0.45
0.37
0.42
0.46
0.29
0.28
0.35

0.45
0.46
0.42
0.48
0.49
0.49
0.39
0.48
0.45
0.45
0.45

0.340
0.384
0.365
0.441
0.486
0.486
0.394
0.555
0.568
0.618
0.668

0.23
0.12
0.18
0.21
0.18
0.01
0.13
0.20
"0.34
"0.39
"0.08

0.420
0.480
0.510
0.620

0.45
0.48
0.43
0.44

0.46
0.42
0.48
0.48

0.343
0.365
0.441
0.520

0.18
0.24
0.14
0.16

0.250
0.320
0.400
0.550

0.29
0.32
0.31
0.37
0.38
0.06
28

0.46
0.46
0.49
0.48

0.343
0.384
0.486
0.555

"0.37
"0.20
"0.22
"0.01
0.00
0.19
28

Yield data from Heijnen and van Dijken (1992).

If DGic < 1; then n "1; and

"

DG
&
%'
#$ pc
DGr 1 " g d g x YC=C " DGic

!0:5

(12)

If DGic > 1; then n 1; and

"

!0:5
DGpc DGic
&
%'
#$
DGr 1 " g d g x YC=C

(13)

The average values so found for pure cultures and mixed

cultures are listed at the bottoms of Tables I and II. The
averages (0.38 and 0.39) and standard deviations (0.06 and
0.05) are similar and not statistically different.
In the second approach, a comparison was made between
YC/C reported and YC/C predicted for different values of e.
The error between the two values in each case was then
calculated using Equation 11. A sum of the errors was then
made for the combined set of pure culture and mixed
culture values. The best-fit e was determined to be that value
resulting in a minimum sum of errors. The results are
provided in the last three columns in Tables I and II. Here,
the best-fit e was 0.37 with a standard deviation error of 19%
for the pure cultures and 15% for the mixed cultures. Thus,
using this approach, the best value for e is somewhat lower
than the values of 0.38 and 0.39 found with the first analysis.
This results largely because of the effect of inversion as well
as the non-linear nature of the relationships. Since the
model is being used to predict yield (YC/C) the best-fit
value of 0.37 was selected for e. This then is the

McCarty: TEEM Modifications for Bacterial Yield Predictions

Biotechnology and Bioengineering. DOI 10.1002/bit

383

transfer efficiency selected for use in the comparison study

for the special compounds covered by the TEEM2
modifications.

Comparison with Oxygenase Reactions

Yield and thermodynamic data for aerobic heterotrophic
reactions involving oxygenases are summarized in Table III
along with model predictions for maximum yields by
VanBriesen (2001) and Yuan and VanBriesen (2002). These
authors followed the detailed biochemical pathways for
oxidation of the substrates in arriving at their predictions.
Also included in the Table are the predictions using TEEM2
and the energy transfer efficiency of 0.37 found from
the calibration study. TEEM2 provides significantly better
accuracy ("0.08 vs. "0.16), while the precisions of the two
different approaches are similar. As a further comparison,
the same set of data was analyzed using the unmodified
TEEM1 approach, yielding a larger error of "0.15 and
standard deviation of 0.23. The simplified oxygenase
modification included in TEEM2 thus appears to be good
for making predictions of yields for reactions where an
oxygenase is involved, providing that the number of
oxygenase reactions per mole of substrate is known.

Evaluation of the C1 Compound Modification

The inaccurate prediction by TEEM1 of maximum yields for
organisms growing on certain organic compounds, such as
formate, was noted by Xiao and VanBriesen (2006). They
suggested that the problem was in the manner in which
TEEM1 addressed carbon balances, especially with compounds having a degree of reduction different from that
Table III. Model comparison for reactions involving oxygenases using
e 0.37 with TEEM2.

TEEM2
Reported
YC/C
p m

Substrate
Methane
Benzene-theoretical
Benzene-average
Toluene-theoretical
Toluene-average
Phenol-theoretical
Phenol-average
Naphthalene-average
Naphthalene-maximum
Phenanthrene-average
Phenanthrene-maximum
NTA
EDTA
Average
Std. Dev.
Number

384

0.55
0.54
0.43
0.55
0.47
0.50
0.36
0.47
0.52
0.32
0.56
0.27
0.27

8
30
30
36
36
28
28
48
48
66
66
18
34

1
1
1
1
1
1
1
1
1
1
1
1
4

fs
0.281
0.390
0.390
0.386
0.386
0.405
0.405
0.389
0.389
0.390
0.390
0.467
0.425

Yuan and
VanBriesen
(2002)

YC/C Error YC/C Error

0.56
0.49
0.49
0.50
0.50
0.47
0.47
0.47
0.47
0.46
0.46
0.35
0.36

"0.02
0.10
"0.13
0.10
"0.06
0.06
"0.31
0.01
0.10
"0.44
0.18
"0.30
"0.34
"0.08
0.20
13

0.63
0.54
0.54
0.61
0.61
0.53
0.53
0.51
0.51
0.48
0.48
0.35
0.28

"0.15
0.00
"0.26
"0.11
"0.30
"0.06
"0.47
"0.09
0.02
"0.50
0.14
"0.30
"0.04
"0.16
0.19
13

of cells, and proposed a modification in an attempt to

address this problem. Their approach in essence divided
the synthesis side of the conversion of substrate to an
intermediate compound into two steps, one of which
converts the substrate carbon 100% into the intermediate
compound carbon. This resulted in either an excess or
deficiency of electrons, depending upon the degree of
reduction of substrate compared with that of cells. The
second step involved adding or subtracting sufficient
substrate to produce an electron balance in the synthesis
reactions. They then chose to apply the efficiency factor
differently to each of the two steps. While not presented as
such in their article, the overall balance they used for the
synthesis energy requirements can be represented in
equation form as follows:
DGs

DGin " g d =g x DGd

g " gd
"" x
DGd
"n
gx

DGpc
"

(14)

The value of n was taken to be 1 when the numerator above

en was positive, and "1 when it was negative. Generally,
the numerator became negative for compounds with low
degree of reduction and positive for those with a high degree
of reduction. Calculated yields with this approach tend to be
lower than that determined with TEEM1 for compounds
with low degree of reduction such as formate, but about
the same for those with high degree of reduction such as
methanol.
Rather than carbon balance difference, it was hypothesized for TEEM2 that errors are primarily with C1
compounds because of energy losses in the synthesis
pathway for these compounds. In order to evaluate the
two different hypotheses, calculated maximum yields by the
two approaches were compared with reported values.
Table IV contains a comparison with the data presented
by Xiao and VanBriesen (2006). Some of their measured
yield data are the same as values in Tables I or II from
Heijnen and van Dijken, (1992) (as noted in the Table IV),
and some came from other sources. In TEEM2, e was
assumed to equal the predetermined value of 0.37. Here, it is
seen that the two approaches produce about the same
accuracy and precision in yield estimation. The differences
when considering the broader range of substrates are not
statistically significant.
A further analysis was made using only the data from the
two articles for C1 compounds to determine how the two
different approaches compare in this case. The results in
Table V indicate that the TEEM2 predictions for this set of
compounds are much more accurate than the predictions by
Xiao and VanBriesen (2006), with average error of "0.17
versus "0.35. An analysis of these data using the original
TEEM1 model with e of 0.37 resulted in an error of "0.39
with standard deviation of 0.29 (data not shown). Thus
the modified TEEM2 method for C1 compounds represents

Biotechnology and Bioengineering, Vol. 97, No. 2, June 1, 2007

DOI 10.1002/bit

Table IV. Model comparisons for compounds examined by Xiao and

VanBriesen (2006) with different gs using e 0.37 with TEEM2.
Xiao and
VanBriesen
(2006)

TEEM2
gs

Reported
YC/C

fs

YC/C

Error

YC/C

Error

1.00
2.00
2.00
2.50
2.67
3.00
3.00
3.00
3.33
3.50
3.66
4.00
4.00
4.00
4.00
4.00
4.66
5.00

0.086
0.162
0.220*
0.280*
0.238*
0.333
0.368
0.348
0.377
0.385
0.535
0.447
0.535
0.505
0.510*
0.470
0.596
0.660

0.402
0.402
0.527
0.482
0.427
0.483
0.453
0.457
0.461
0.417
0.482
0.394
0.486
0.486
0.450
0.507
0.476
0.427

0.101
0.201
0.263
0.301
0.285
0.362
0.340
0.343
0.384
0.365
0.441
0.394
0.486
0.486
0.450
0.507
0.555
0.534

"0.17
"0.24
"0.20
"0.08
"0.20
"0.09
0.08
0.01
"0.02
0.05
0.18
0.12
0.09
0.04
0.12
"0.08
0.07
0.19

0.107
0.216
0.247
0.297
0.268
0.337
0.334
0.342
0.397
0.383
0.464
0.446
0.501
0.501
0.480
0.524
0.578
0.616

"0.24
"0.33
"0.12
"0.06
"0.13
"0.01
0.09
0.02
"0.05
0.01
0.13
0.00
0.06
0.01
0.06
"0.11
0.03
0.07

6.00
6.00

0.552
0.558*

0.375 0.563 "0.02 0.728 "0.32

0.445 0.668 "0.20 0.692 "0.24
"0.02
"0.06
0.13
0.14
20
20

Substrate
2"

Oxalate
FormateGlyoxylate"
Tartrate2"
Malonate2"
Iminodiacetate
Citrate3"
Malate2"
Pyruvate"
Succinate2"
Gluconate"
Acetate"
Glucose
Fructose
Lactate"
Formaldehyde
Glycerol
Ethylenediamine
(ED)
Methanol
Ethanol
Average
Std. Dev.
Number

&
Data from Heijnen and van Dijken (1992) that is also listed in Tables I
or II.

a significant improvement over TEEM1. However, the

predictions are much higher for formate than measured,
thus leading to the larger than desired negative average
values with TEEM2. Additional corrections for the reduced

Table V. Model comparisons for C1 compounds using e 0.37 with

TEEM2.
Xiao and
VanBriesen
(2006)

TEEM2
Substrate
FormateFormateFormateFormaldehyde
Methanol
Methanol
Methanol
Average
Std. Dev
Number

gs

Reported
YC/C&

fos

YC/C

Error

YC/C

Error

2.00
2.00
2.00
4.00
6.00
6.00
6.00

0.162
0.120
0.180
0.470
0.540
0.540
0.552

0.402
0.402
0.402
0.507
0.375
0.375
0.375

0.201
0.201
0.201
0.507
0.563
0.563
0.563

"0.24
"0.68
"0.12
"0.08
"0.04
"0.04
"0.02
"0.17
0.23
7

0.216
0.216
0.216
0.524
0.728
0.728
0.728

"0.33
"0.80
"0.20
"0.11
"0.35
"0.35
"0.32
"0.35
0.23
7

&
Data from Heijnen and van Dijken (1992) and Xiao and VanBriesen
(2006), with one reported YC/C value of 0.100 for formate not included as
value appears to be erroneous (correspondence with VanBriesen).

yields for organisms following the serine pathway were not

made here, and this may partially explain the less accurate
predictions with formate. In any event, TEEM2 with an
energy transfers efficiency of 0.37 appears to provide a good
estimate of maximum yield in general for a range of
compounds under aerobic conditions, regardless of the
degree of carbon reduction. In summary, the estimates
for C1 compounds, covering a wide range of degrees of
reduction, are better than that provided by the degree of
reduction model by Xiao and VanBriesen (2006).

Discussion
TEEM2 was found to be as good as or better than the
comparison models for predicting maximum aerobic
bacterial yields. The comparisons made here were aerobic
growth because of the extensive data bank that was here
available. TEEM2 is based upon an electron equivalents
balance, with yields reported as fraction of substrate or
electron donor electron equivalents converted for synthesis.
Other models tend to report yields in moles of cell carbon
per mole of substrate carbon for organic electron donors or
in moles cell carbon per mole of electron donor for
autotrophic reactions. Conversions to such units is readily
achieved when using electron equivalents. The comparison
models also use electron equivalents for determining
reaction energies so that use for determining yield in
electron equivalents would not be a difficult transition to
make. Using electron equivalents has the advantage that
conversion factors are not needed in the models, which often
leads to confusion. An additional advantage of using
electron equivalents is that stoichiometric equations for the
overall microorganism reactions for growth and energy
production can be more directly produced.
With the large data set of aerobic heterotrophic yield
values analyzed here, the best energy transfer efficiency
found for use in TEEM2 for aerobic growth was 0.37. With
this value, predicted yields were within 13%23% of the
measured yields. Good accuracy and about the same
variation was found for organic reactions involving
oxygenases. The TEEM2 modification to address single
carbon compounds was also found to be quite accurate for
all cases using the 0.37 energy transfer efficiency, except for
formate, where yield predictions were on average 35% too
high. This larger error was affected mainly by one especially
low formate yield measurement, which may be the result of a
measurement error or to formate catabolism following a
lower energy biochemical pathway than assumed here. For
other C1 compounds, the accuracy was good and much
better than with the comparison model, especially for
methanol, which has a high degree of reduction. These
results tend to support the hypothesis proposed here that
yield measurements that are much lower than predicted by
TEEM1 are the result of energy inefficient biochemical
pathways taken in transformations of C1 compounds, rather

McCarty: TEEM Modifications for Bacterial Yield Predictions

Biotechnology and Bioengineering. DOI 10.1002/bit

385

than to a compounds degree of reduction as proposed by

Xiao and VanBriesen (2006).
In comparing the precision of model predictions one
needs to be cautious since there is also much variation in
reported yields for a given substrate. The question arises as
to how much variation in predictions is the result of
variations between organisms, variations in ability of single
organisms to use different compounds efficiently, or the
result of different experimental protocols used in making
measurements. In order to address this, there are at least
three different yield reports in Tables IV for aerobic growth
on each of 11 different organic electron donors. The average
and standard deviation of reported yields with each
compound was determined and the relative error computed
for each. The relative errors for the nine compounds was
then pooled to determine an overall relative error and found
to be 13%. Thus, a significant portion of the model error of
13%23% can be attributed to normal differences in
reported yield measurements.
Table I lists implied energy transfer efficiencies with
different compounds for four different organisms. The
energy transfer efficiency for Candida utilus varied from a
low of about 28% for oxidations of acetoin and 2,3butanediol to a high of 48% for citrate. The energy transfer
efficiency for Thiobacillus acidophilus ranged from 29 to 37%
with an average of 32% while that of Paracoccus denitrificans
was much higher with a range of 43 to 48% and an average of
46%. Thus we see differences in energy transfer efficiency
with different compounds for the same organism as well as
quite different average energy transfer efficiencies for the
same compound with different organisms.
The important lesson for predictive models for bacterial
yield from reaction energetics is that there are many factors
that affect energy transfer efficiency that cannot be captured
in simple models. What can be captured is general
relationships, such as likely yields or variations in predicted
yields. The results of this analysis suggest that in using
TEEM2 for predictions for aerobic heterotrophic growth
and using an energy transfer efficiency of 0.37, yield
predictions within plus or minus 20%25% of normally
measured yields might be expected.
Rittmann and McCarty (2001) indicate that under
optimum conditions, transfer efficiencies of 55%70% are
typical and an e value of 0.60 is frequently employed. The
analysis made here, however, suggests that an e of 0.37 is
better for use in predictions of yield with aerobic
heterotrophic growth. This difference needs to be addressed.
In the original development of TEEM1 (McCarty, 1965),
upper limits on energy transfer efficiencies of 0.50.6 were
reported for autotrophic growth and of 0.450.65 for
anaerobic heterotrophic growth. However, for the 24
aerobic heterotrophic yield measurements evaluated, values
typically ranged between 0.20 and 0.40 and averaged only
0.29, a value even lower than that of 0.37 found here.
The lower energy transfer efficiencies for aerobic
heterotrophic growth were hypothesized (McCarty, 1965)
to result from failure to consider maintenance energy

386

requirements in making yield measurements and also the

result of energy losses associated with the production of
soluble microbial products (smp). To address this issue,
reported oxygen uptake measurements associated with
aerobic mixed culture growth on organic compounds were
evaluated. Here, foe was determined from oxygen uptake
measurements that included corrections for cell decay.
Then, fos was taken to equal 1 " foe (Eq. 6), which in effect
includes donor conversion to both particulate cell material
and smp. In this manner, e was found to average a much
higher 0.57 for the seven measurements presented that did
not include C1 compounds. In a subsequent evaluation
(McCarty, 1971), analysis was made of mixed culture data by
Burkhead and McKinney (1969) who simultaneously
measured particulate cell production, smp, and oxygen
uptake, while making corrections for cell decay. Assuming fos
included both the particulate fraction and smp, energy
transfer efficiency here for the 11 non-C1 compounds
studied averaged 0.58 with a standard deviation of 0.15.
These results led to the recommended use of an expected
range of 0.40.7 for bacterial yields for aerobic as well as
anaerobic conditions (McCarty, 1971).
This analysis suggests that smp as well as particulate
bacterial cell production need consideration in yield
predictions. Unfortunately, there are few reports of
simultaneous measurements of all three components, that
is particulate cell production, smp formation, and oxygen
consumption. More good mass balance data are needed.
Also, efforts are normally not made to correct cell yield for
maintenance energy requirements or the related cell decay
rate. Thus, many yield reports may represent net yield rather
than the maximum yield for which TEEM was developed.
Thus, while an energy transfer efficiency of 0.37 was found to
be the best-fit value for the range of data evaluated here,
caution is required in using this value for determinations of
fos and foe and the construction of balanced stoichiometric
equations. For this, perhaps higher values within the range
of 0.40.7 as originally proposed would be a better choice.
Better experimental data that allows good mass balances of
all materials involved is obviously needed to further the
understanding of energy relationships in bacterial growth
and to make useful growth predictions.
Also, unfortunately, the data readily available for
anaerobic autotrophic and heterotrophic growth was much
too limited here to evaluate model accuracy and precision
any more than provided in the original publications on
TEEM1. In order to do this, good information on reaction
products is needed so that reaction Gibbs free energy can be
determined well. In many cases, Equation 2 will need to be
used in order to obtain a sufficiently accurate determination
of reaction energy. True yields with different compounds
and fermentation pathways can vary widely under anaerobic
conditions because of the great differences in energy
available by the different pathways. It is here that
thermodynamic predictions can provide the most useful
information. Clearly, more efforts towards evaluating
anaerobic growth yields are also needed.

Biotechnology and Bioengineering, Vol. 97, No. 2, June 1, 2007

DOI 10.1002/bit

Appendix. Thermodynamic properties of compounds at 258C.

Substance
"

Acetate
Acetoin
Acetone
Acetyl-CoA
Alanine
n-Alkanes
Benzene
Benzoate
Butane
2-3 butanediol
n-Butanol
ButyrateCitrate3"
Dihydroxy-acetone
EDTA
Ethanol
Ethylene glycol
Ethylenediamine
Formaldehyde
Formate"
Fructose
Gluconate"
Glucose
Glycerol
Glycine
Glyoxylate"
Iminodiacetate
Lactate"
Lactose
Malate2"
Malonate2"
Mannitol
Methane
Methanol
NADH
Naphthalene
NTA
Oxalate2"
Phenanthrene
Phenol
Phenylalanine
n-Propanol
Propionate"
Pyruvate"
Succinate2"
Sucrose
Tartrate2"
Toluene
Xylose
NH4
CO2
H
H (pH 7)
H2CO3
H2O
HCO"
3

DGof (kJ/mol)

Reference

aq.
aq.
aq.

"369.41
"280
"161.17

aq.
aq.
aq.
aq.
g
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
g
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
aq.
g
aq.
aq.
aq.
l
aq.

"371.54
60
133.9
"245.6
"15.707
"322
"171.84
"352.63
"1168.34
"450
"1209.15
"181.75
"323.21
"10.05
"130.54
"351.0
"915.38
"1128.3
"917.22
"488.52
"370.788
"658.1
"655.2
"517,81
"1515.24
"845.08
"700
"942.61
"50.75
"175.39

State

219.97
"954.79
"674.9
310.99
"47.5
"207.1
"175.81
"361.08
"474.63
"690.23
"1551.85
"1010
127
1077
"79.37
"394.36
0
"39.87
"623.16
"237.18
"586.85

c
a
a

a
f
d
e
c
a
a
a
c
b
a
a
b
a
a
a
a
a
a
a
a
b
a
a
a
a
a
a
a
a
b
b
a
b
f
a
a
a
a
a
a
c
f
g
a

Oxidized form
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
NAD
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2
CO2

DGo (kJ/mol)

p eeq/mol

C mol/mol

g degree of reduction

27.40
33.59
29.62
30.88
31.37
27.48
28.34
27.34
27.56
32.25
29.26
27.72
33.08
41.67
33.66
31.18
39.01
29.80
46.53
39.19
41.27
40.21
41.35
38.88
39.80
51.30
40.56
32.29
42.09
34.17
29.78
39.89
23.53
36.84
30.88
27.80
33.97
52.10
27.62
29.50
29.42
29.94
27.63
35.09
29.09
42.00
40.24
27.85
41.35

8
20
16

2
4
3

4.00
5.00
5.33

12
92
30
30
26
22
24
20
18
12
34
12
10
10
4
2
24
22
24
14
6
4
12
12
48
12
8
26
8
6
2
48
18
2
66
28
40
18
14
10
14
48
10
36
20

3
15
6
7
4
4
4
4
6
3
10
2
2
2
1
1
6
6
6
3
2
2
4
3
12
4
3
6
1
1

4.00
6.13
5.00
4.29
6.50
5.50
6.00
5.00
3.00
4.00
3.40
6.00
5.00
5.00
4.00
2.00
4.00
3.67
4.00
4.67
3.00
2.00
3.00
4.00
4.00
3.00
2.67
4.33
8.00
6.00

10
6
2
14
6
9
3
3
3
4
12
4
7
5

4.80
3.00
1.00
4.71
4.67
4.44
6.00
4.67
3.33
3.50
4.00
2.50
5.14
4.00

a
a
a
a
a
a

Thauer et al. (1977).

Yuan and VanBriesen (2002).
Heijnen and van Dijken (1992).
d
Madigan et al. (1997).
e
Weast and Astle (1980).
f
Sawyer et al. (2003).
a
Xylose assumed to have same reduction potential of 41.35 kJ/eeq as glucose.
b
c

McCarty: TEEM Modifications for Bacterial Yield Predictions

Biotechnology and Bioengineering. DOI 10.1002/bit

387

References
Blackmore MA, Quayle JR. 1970. Microbial growth on oxalate by a route
not involving glyoxylate carboligase. Biochem J 118:5359.
Burkhead CE, McKinney RE. 1969. Energy concepts of aerobic microbial
metabolism. J Sanitary Eng Div, Proc Amer Soc Civil Eng 95 SA2:253
268.
Heijnen JJ, van Dijken JP. 1992. In search of a thermodynamic description
of biomass yields for the chemotrophic growth of microorganisms.
Biotechnol Bioeng 39(8):833858.
Kornberg HL. 1966. Anaplerotic sequences and their role in metabolism. In:
Campbell PN, Greville GD, editors. London: Academic Press. p 131.
Madigan MT, Martinko JM, Parker J. 1997. Brock biology of microorganisms. Upper Saddle River, NJ: Prentice Hall. 986 p.
McCarty PL. 1965. Thermodynamics of biological synthesis and growth. Int
J Air Water Pollut 9:621639.
McCarty PL. 1971. Energetics and bacterial growth. In: Faust SD, Hunter JV,
editors. Organic compounds in aquatic environments. New York:
Marcel Dekker, Inc. p 495531.

388

McCarty PL. 1975. Stoichiometry of biological reactions. Prog Water

Technol 7:157172.
Rittmann BE, McCarty PL. 2001. Environmental biotechnology: Principles
and applications. New York: McGraw-Hill. 754 p.
Sawyer CN, McCarty PL, Parkin GF. 2003. Chemistry for environmental
engineering and science, 5th Edn. New York: McGraw-Hill. 752 p.
Thauer RK, Jungermann K, Decker K. 1977. Energy conservation in
chemotrophic anaerobic bacteria. Bacteriol Rev 41(1):100180.
VanBriesen JM. 2001. Thermodynamic yield predictions for biodegradation
through oxygenase activation reactions. Biodegradation 12(4):265.
VanBriesen JM. 2002. Evaluation of methods to predict bacterial yield using
thermodynamics. Biodegradation 13(3):171.
Weast RC, Astle MJ. 1980. CRC handbook of chemistry and physics. Boca
Raton, FL: CRC Press, Inc.
Xiao J, VanBriesen JM. 2006. Expanded thermodynamic model for microbial true yield prediction. Biotechnol Bioeng 93(1):110121.
Yuan ZW, VanBriesen JM. 2002. Yield prediction and stoichiometry of
multi-step biodegradation reactions involving oxygenation. Biotechnol
Bioeng 80(1):100.

Biotechnology and Bioengineering, Vol. 97, No. 2, June 1, 2007

DOI 10.1002/bit