Re-engineering the discrimination between the oxidized
coenzymes NAD
+
and NADP
+
in clostridial glutamate
dehydrogenase and a thorough reappraisal of the
coenzyme specificity of the wild-type enzyme
Marina Capone*, David Scanlon, Joanna Griffin and Paul C. Engel
School of Biomolecular and Biomedical Science, Conway Institute, University College Dublin, Ireland
Introduction
The nicotinamide-nucleotide-dependent dehydrogenases
tend, in general, to be either NAD
+
-specific (and then
catabolic) or NADP(H)-specific (and accordingly ana-
bolic, except for those few enzymes such as glucose
6-phosphate dehydrogenase which provide NADPH
for biosynthesis) [1]. Crystallographic studies of arche-
typal NAD
+
-specific enzymes, such as alcohol and
lactate dehydrogenases [2,3], and archetypal NADPH-
specific dehydrogenases such as glutathione reductase
[4] have offered some degree of understanding of the
ways in which these enzymes achieve their coenzyme
specificity. This has been augmented by various
detailed studies of amino acid sequences [5,6], and has
been both tested and applied in some notably success-
ful examples of re-engineering of coenzyme specificity
[7–19]. As noted by Khouri et al. [17], however, the

were markedly diminished. This article reveals that the enzyme’s
discrimination in favour of NAD
+
and against NADP
+
had been greatly
underestimated and has indeed been abated by a factor of > 16 000 by the
mutagenesis. Initially, stopped-flow studies of the wild-type enzyme showed
a burst increase of A
340
with NADP
+
but not NAD
+
, with amplitude
depending on the concentration of the coenzyme, rather than enzyme.
Amplitude also varied with the commercial source of the NADP
+
. FPLC,
HPLC and mass spectrometry identified NAD
+
contamination ranging
from 0.04 to 0.37% in different commercial samples. It is now clear that
apparent rates of NADP
+
utilization mainly reflected the reduction of con-
taminating NAD
+
, creating an entirely false view of the initial coenzyme
specificity and also of the effects of mutagenesis. Purification of

es. The glutamate dehydrogenases are such a family
[20] and, as a result, have three different EC classifica-
tions,
EC 1.4.1.2, EC 1.4.1.3 and EC 1.4.1.4 for
NAD
+
-specific, dual-specific (particularly common in
higher animals and in Archaea), and NADP
+
-specific
members, respectively. Most of these are evolutionarily
and structurally related, in many cases quite closely,
despite the functional and classificational division
[21,22], and thus they provide a revealing example of
the way in which a single structural scaffold can be
adapted to produce remarkably different functional
outcomes.
Our own protein engineering experiments [23], based
on an analysis of the high-resolution structure of the
binary enzyme–NAD
+
complex of clostridial gluta-
mate dehydrogenase [24,25], were aimed initially at
facilitating productive binding of the phosphorylated
coenzyme by enlarging the potential binding pocket
and removing the negative charge, likely to repel the
2¢-phosphate of NADP(H), and replacing it with posi-
tive charge. Accordingly, mutants F238S, P262S and
F238S ⁄ P262S were created to provide more space, and
D263K to offer a more favourable electrostatic envi-

,
K
m
and k
cat
⁄ K
m
for the oxidized coenzymes are pre-
sented in Table 1. After replacement of Phe238 by
serine, NAD
+
was less effective as a coenzyme because
of moderate decreases in k
cat
(36, 14 and 33% at
pH 6.0, 7.0 and 8.0, respectively) and marked increases
in apparent K
m
( 10 fold at pH 6.0 and 7.0, 14 fold at
pH 8.0). NAD
+
is evidently bound very poorly to this
mutant. However, surprisingly, no improvement was
apparent with NADP
+
as the coenzyme. At pH 7.0
and 8.0, approximately five-fold decreases in k
cat
, and
increases of approximately four- and six-fold respec-

+
.
Turning to the third of the single mutants, D263K,
once again, there appeared to be a decrease in catalytic
efficiency with both NAD
+
and NADP
+
as coen-
zymes, though less marked than for the other muta-
tions, and at pH 8.0 with NADP
+
as coenzyme there
was little difference between the performance of the
wild-type and mutant enzymes (Table 1).
The results for these three mutants were extremely
puzzling; the mutations had been designed to facilitate
binding of the phosphorylated coenzyme, and with the
reduced coenzymes [23] there were indeed large shifts,
as expected, in the discrimination factor, 150 fold and
272 fold, respectively, for example, for P262S and
D263K at pH 8.0. In this study, only the double-
mutant F238S ⁄ P262S gave a result reasonably close to
expectation with the oxidized coenzymes (Table 1): at
both pH 7.0 and pH 8.0 the large discrimination factor
in favour of NAD
+
decreased to only 3–4 for this
mutant. Even in this case, however, the apparent
M. Capone et al. Coenzyme preference in glutamate dehydrogenase

NADP
+
as coenzyme, but not with NAD
+
. Two dif-
ferent phases were identified in the stopped-flow traces:
the first phase (Fig. 1A inset) consisted of the rapid
single exponential burst in A
340
, reaching an apparent
plateau within a few seconds. The small differences in
the height of this plateau for different concentrations
of enzyme (5–20 lm) are due to shifts in the baseline
as a result of the contribution of the enzyme itself to
A
340
; after correction for this baseline shift (Fig. 1B),
the burst amplitude was entirely independent of the
enzyme concentration. Over much longer periods
(Fig. 1A, main panel), the apparent plateau was
revealed as a very slow and initially linear second
phase of increase in absorbance, finally leading to the
expected reaction equilibrium in over 4 h.
Further analysis showed that the burst amplitude with
NADP
+
was dependent on the concentration of the
coenzyme itself, and not only on its concentration, but
also the commercial source. With 1 mm NADP
+

and their standard errors (± SE) were calculated by a nonlinear regression method [36] with
ENZPACK version 3.0 (Biosoft Ltd, Cambridge,
UK). The discrimination factor in the right-hand column, a measure of the preference for NAD
+
over NADP
+
, is calculated as the ratio of the
catalytic efficiency, k
cat
⁄ K
m
, for NAD
+
to that for NADP
+
. ND, not determined.
pH
NAD
+
NADP
+
Discrimination
factor
k
cat
(s
)1
) K
m
(mM)

F238S 7.0 17.6 ± 0.3 1.25 ± 0.46 14.1 0.125 ± 0.04 1.1 ± 0.7 0.113 125
P262S 7.0 23.0 ± 1.4 1.04 ± 0.13 23.0 0.457 ± 0.05 1.23 ± 0.22 0.371 62
F238S ⁄ P262S 7.0 3.31 ± 0.47 2.22 ± 0.21 1.49 0.242 ± 0.05 0.632 ± 0.27 0.382 3.9
D263K 7.0 12.6 ± 0.21 0.17 ± 0.02 72.4 0.40 ± 0.05 0.380 ± 0.002 1.05 69
Wild-type 8.0 51.6 ± 0.6 0.127 ± 0 .012 482 0.909 ± 0.19 0.336 ± 0.015 2.70 179
F238S 8.0 39.8 ± 1.7 1.83 ± 0.04 21.7 0.270 ± 0.05 2.04 ± 0.08 0.132 164
P262S 8.0 53.4 ± 2.6 0.90 ± .11 59.3 0.339 ± 0.06 1.04 ± 0.30 0.339 175
F238S ⁄ P262S 8.0 4.49 ± 0.34 6.28 ± 0.34 0.71 0.513 ± 0.193 2.24 ± 0.77 0.229 3.1
D263K 8.0 39.3 ± 1.9 0.376 ± 0.03 104 0.958 ± 0.05 0.336 ± 0.05 2.85 36.5
Coenzyme preference in glutamate dehydrogenase M. Capone et al.
2462 FEBS Journal 278 (2011) 2460–2468 ª 2011 The Authors Journal compilation ª 2011 FEBS
enriched with NAD
+
permitted identification of a peak
of the latter estimated at  0.37% in NADP
+
from
Roche,  0.15% in NADP
+
from Apollo Scientific and
 0.04% in NADP
+
from Sigma-Aldrich Ireland Ltd.
(Dublin, Ireland). The agreement of the rapid reaction
kinetics with the HPLC analysis suggested nearly total
conversion of NAD
+
into NADH in the reaction
observed. From the equilibrium constant for the oxida-
tive deamination of l-glutamate [27] it can be estimated

ial glutamate dehydrogenase is much tighter than the
binding of the oxidized coenzyme. Together with the
initial preference for the nonphosphorylated coenzyme,
this explains the potent inhibitory effect of such a
small NADH contamination.
Mass spectrometric identification of NAD
+
peak
Isolation and mass spectroscopic analysis confirmed
the identity of the contaminant. Comparison with the
spectrum of an authentic NAD
+
sample revealed
total similarity of the fragmentation pattern. The neg-
ative portion of the spectrum displayed a fragment at
m ⁄ z 540, along with a small amount of parental mol-
ecule (m ⁄ z 662.1). The signal at m ⁄ z 540 corresponds
to the ADP-ribose moiety of the coenzyme resulting
from splitting off the nicotinamide ring, suggesting
that the covalent bond between the nicotinamide and
the ribose of the coenzyme is particularly labile. The
signal of the parental molecule is also visible at
m ⁄ z 664.1 in the positive spectrum; the nicotinamide
ring, counterpart fragment of the ADP-ribose
(m ⁄ z 540) gives a signal at m ⁄ z 123.1, whereas ade-
nine is registered at m ⁄ z 136.1.
In view of these findings, the possibility of the
reverse contamination was also tested. However, the
0
0.2

and 40 m
ML-glutamate (concentrations after mixing) monitored at
340 nm over 250 min. The increase in absorbance was linear for
the first 30–40 min; on this basis, the value for the specific activity
was calculated as 1.84 nmolÆmin
)1
Æmg
)1
, rate 0.0015 s
)1
. The inset
shows stopped-flow traces observed over the first few seconds of
the forward reaction of 2.5 l
M (lowest trace), 5 lM, 7.5 lM, 10 lM
and 15 lM (highest trace) wild-type clostridial glutamate dehydroge-
nase with 1 m
M NADP
+
and 40 mML-glutamate (all concentrations
after mixing). A burst phase was detected in all cases. The almost
horizontal trace seen in each case after 2–3 s corresponds to the
slow, steady-state reaction monitored in the main panel. (B) Cor-
rection applied to the burst amplitudes calculated at different
enzyme concentrations in (A). The increase of enzyme concentra-
tion causes a significant baseline shift at 340 nm. A reference zero
baseline was first recorded by mixing 2 m
M NADP
+
and 80 mML-
glutamate with buffer in the stopped-flow; a set of individual base-

, it was necessary to reconsider the steady-
state results. First of all, the specific activity of clostrid-
ial glutamate dehydrogenase under standard assay con-
ditions (1 mm coenzyme, 40 mml-glutamate, 0.1 m
phosphate, pH 7.0) was re-determined with freshly
purified NADP
+
and using three different methods of
measurement, absorbance measurements in the stopped-
flow apparatus, conventional spectrophotometry and
fluorimetry. These three methods, respectively, yielded
values of 2.22, 2.33 and 2.78 · 10
)3
lmolÆmin
)1
Æmg
)1
(mean = 2.44 · 10
)3
lmolÆmin
)1
Æmg
)1
). This figure is
 11 000 times lower than the corresponding figure for
the preferred coenzyme NAD
+
. This remarkable dis-
crimination factor is nearly 40 times higher than the
300-fold factor reported by Syed et al. [26]. It is now

A detailed re-analysis of the steady-state properties
of wild-type glutamate dehydrogenase with freshly
purified NADP
+
was therefore carried out. The values
for K
m
and k
cat
given in Table 1 are 0.26 ± 0.01 mm
and 0.57 ± 0.06 s
)1
, respectively. The redetermination
with pure coenzyme gave a much higher value for the
K
m
of 3.2 ± 0.4 mm, 30-fold higher than the K
m
for NAD
+
. Moreover, k
cat
was calculated as
8.2 · 10
)3
±6· 10
)4
s
)1
,  2500 fold lower than the

in the other
chromatograms (indicated by the arrow).
The amounts of contaminant NAD
+
present
in the samples were calculated by peak
integration using the
MILLENNIUM software
package.
Fig. 3. Superposition of stopped-flow burst-kinetic traces for wild-
type clostridial glutamate dehydrogenase and different batches of
NADP
+
at identical concentrations. (Upper) Reaction with Grade I
NADP
+
from Roche. (Middle) Reaction with NADP
+
from Apollo
Scientific. (Lower) Reaction with freshly purified NADP
+
.
Coenzyme preference in glutamate dehydrogenase M. Capone et al.
2464 FEBS Journal 278 (2011) 2460–2468 ª 2011 The Authors Journal compilation ª 2011 FEBS
than the true value, underlining the impact of quite a
small level of contamination on these results.
Reassessment of the effects of the mutations on
coenzyme specificity
In view of the dramatically altered figure for the strin-
gency of coenzyme specificity in clostridial glutamate

fold. F238S, therefore, offers a 161-fold improvement
instead of a 1.5-fold deterioration in discrimination
factor. Even more dramatically, the modest apparent
21-fold improvement in the double-mutant should be a
16 200-fold improvement, entirely vindicating the ini-
tial thinking behind the mutagenesis.
Wider implications
Careful purification of nicotinamide coenzymes has
been recognized in the past as an important issue in
the study of dehydrogenases [29]: coenzymes were
often purified in research laboratories prior to use
[30,31] in order to avoid misleading kinetic anomalies,
but this routine has largely been abandoned in recent
years because the purity and stability of the best
commercial preparations have dramatically improved.
Although concern and worry persist over the purity of
reduced coenzymes, which are often contaminated by
the oxidized form, Grade I NAD
+
and NADP
+
are
generally utilized without further purification, even in
enzymatic studies of coenzyme specificity [7,32,33]. In
our own laboratory, because analytical HPLC only
revealed what we took to be trace, negligible contami-
nation, well below 0.5%, we have frequently proceeded
without further purification of the coenzyme. Other
authors, using sensitive detection with a dehydrogenase
coupled with the reduction of INT to a coloured for-

+
was utilized without further purification.
In this study, the extremely large effect of  0.3%
contamination of the NADP
+
by the favoured coen-
zyme NAD
+
is directly attributable to the very high
level of discrimination between the two coenzymes, so
that the 0.3% NAD
+
produces a rate far higher than
that for the 99.7% NADP
+
. Accordingly, with the
mutants, all those in which the discrimination has not
been largely abolished give grossly misleading results;
only the double-mutant, which approaches dual speci-
ficity status, gives a result remotely approaching the
truth, because in this situation the 0.3% contamination
is at last less dominant.
It may be argued that this problem is exceptional,
deriving from an extraordinarily high discrimination
factor of 80 000. However, because we ourselves
assumed until this study that clostridial glutamate
dehydrogenase, although NAD
+
specific, showed a far
Table 2. Summary of corrected kinetic parameters for the oxidative deamination. Values of K

(mM) k
cat
(s
)1
)
k
cat
⁄ K
m
(s
)1
ÆmM
)1
) K
m
(mM) k
cat
(s
)1
)
k
cat
⁄ K
m
(s
)1
ÆmM
)1
)
Wild-type 0.11 ± 0.01 22.7 ± 0.5 206 3.2 ± 0.4 8.2 E

+
lithium
salt grade I ( 100%) was obtained from Roche. Different
batches of NADP
+
were: NADP
+
disodium salt ( 98%),
from Roche; NADP
+
monosodium salt  97%, from
Sigma; NADP
+
monosodium salt > 98%, from Apollo
Scientific Ltd. All solutions of the above compounds were
freshly prepared in 100 mm phosphate buffer at pH 7.0, and
used in enzymatic assays within a few hours. Coenzyme
solutions were kept cold and their concentrations deter-
mined by measuring A
260
(e
NAD ðPÞþ
=18· 10
)3
m
)1
Æcm
)1
).
Examination of coenzyme specificity

steady-state analysis, and at lower concentrations where the
reaction in the above conditions was difficult to observe.
Coenzyme concentrations were kept at or above K
m
values
derived by steady-state analysis. A minimum of 5 lm
enzyme was used for the assays.
NADP
+
purification
Coenzymes were analysed with a Waters Controller 600
HPLC system on a reverse-phase column (SUPELCOSIL
LC-18-T, particle size 5 lm, 25 · 4.6 cm). The samples
were dissolved in 100 mm KH
2
PO
4
and 25–30 ng of each
was injected. The elution protocol was as advised by the
column manufacturers (Elution Protocol for nucleotides,
Supelco Catalogue). Solutions were adjusted to pH 6.0 to
prevent damage to the silica solid phase. Data were
acquired with a Waters Photodiode Array Detector 996
and chromatograms were monitored at 254 nm.
Where indicated, NADP
+
was purified on a preparative
scale (up to 9 mg) using a BioCAD Perseptive System FPLC
apparatus with a POROS 20 HQ column (4.60 · 100 mm), a
flow rate of 5 mLÆmin

with electrospray source.
Acknowledgements
MC was supported by a postgraduate scholarship from
the Irish Council for Science, Engineering and Tech-
nology. DS was supported through a Basic Science
Coenzyme preference in glutamate dehydrogenase M. Capone et al.
2466 FEBS Journal 278 (2011) 2460–2468 ª 2011 The Authors Journal compilation ª 2011 FEBS
research grant (SC2002 ⁄ 0502) to PCE from Enterprise
Ireland. PCE was supported during the writing of this
paper by a Fellowship grant (05 ⁄ FE1 ⁄ B857) from Sci-
ence Foundation Ireland. These sources of financial
support are gratefully acknowledged. We are also
grateful to Dr Dilip Rai of the School of Chemistry
and Chemical Biology at UCD who ran the mass spec-
trometric analysis of coenzyme samples and purified
contaminants.
References
1 Lehninger AL, Nelson DL & Cox MM (1993)
Principles of Biochemistry, 2nd edn, pp. 390–392.
Worth, New York.
2 Bra
¨
nde
´
n CI, Jo
¨
rnvall H, Eklund H & Furugren B
(1975) Alcohol dehydrogenases. In The Enzymes,
Vol. XI (Boyer PD, ed.), pp. 103–190. Academic Press,
New York.

119, 1014–1018.
11 Steen IH, Lien T, Madsen MS & Birkeland NK (2002)
Identification of cofactor discrimination sites in NAD-
isocitrate dehydrogenase from Pyrococcus furiosus. Arch
Microbiol 178, 297–300.
12 Zhang L, Ahvazi B, Szittner R, Vrielink A & Meighen
E (1999) Change of nucleotide specificity and enhance-
ment of catalytic efficiency in single point mutants of
Vibrio harveyi aldehyde dehydrogenase. Biochemistry
38, 11440–11447.
13 Chen R, Greer A & Dean AM (1996) Redesigning sec-
ondary structure to invert coenzyme specificity in iso-
propylmalate dehydrogenase from
Saccharomyces cerevisiae. Proc Natl Acad Sci U S A 93,
12171–12176.
14 Holmberg N, Ryde U & Buelow L (1999) Redesign of
the coenzyme specificity in l-lactate dehydrogenase
from Bacillus stearothermophilus using site-directed
mutagenesis and media engineering. Protein Eng 12
,
851–856.
15 Watanabe S, Kodaki T & Makino K (2005) Complete
reversal of coenzyme specificity of xylitol dehydrogenase
and increase of thermostability by the introduction of
structural zinc. J Biol Chem 280, 10340–10349.
16 Ehrensberg AH, Elling RA & Wilson DK (2006) Struc-
ture-guided engineering of xylitol dehydrogenase cosub-
strate specificity. Structure 14, 567–575.
17 Khouri AG, Fazelinia H, Chin JW, Pantazes RJ, Cirino
PC & Maranas CD (2009) Computational design of

determination of coenzyme specificity in clostridial
NAD
+
-dependent glutamate dehydrogenase. Enzyme
Res, in press.
24 Baker PJ, Britton KL, Engel PC, Farrants GW, Lilley
KS, Rice DW & Stillman TJ (1992) Subunit assembly
M. Capone et al. Coenzyme preference in glutamate dehydrogenase
FEBS Journal 278 (2011) 2460–2468 ª 2011 The Authors Journal compilation ª 2011 FEBS 2467
and active site location in the structure of glutamate
dehydrogenase. Proteins 12, 75–86.
25 Baker PJ, Britton KL, Rice DW, Rob A & Stillman TJ
(1992) Structural consequences of sequence patterns in
the fingerprint region of the nucleotide binding fold.
Implications for nucleotide specificity. J Mol Biol 228,
662–671.
26 Syed SE-H, Engel PC & Parker DM (1991) Functional
studies of a glutamate dehydrogenase with known
three-dimensional structure: steady-state kinetics of
the forward and reverse reactions catalysed by the
NAD
+
-dependent glutamate dehydrogenase of
Clostridium symbiosum. Biochim Biophys Acta 1115,
123–130.
27 Engel PC & Dalziel K (1967) The equilibrium constants
of the glutamate dehydrogenase system. Biochem J 105,
691–695.
28 Woodyer R, van der Donk WA & Zhao H (2003)
Relaxing the nicotinamide cofactor specificity of phos-

97–105.
36 Wilkinson GN (1961) Statistical estimations in enzyme
kinetics. Biochem J 80, 324–332.
Coenzyme preference in glutamate dehydrogenase M. Capone et al.
2468 FEBS Journal 278 (2011) 2460–2468 ª 2011 The Authors Journal compilation ª 2011 FEBS