Biosynthesis of riboflavin
Screening for an improved GTP cyclohydrolase II mutant
Martin Lehmann
1
, Simone Degen
1
, Hans-Peter Hohmann
1
, Markus Wyss
1
, Adelbert Bacher
2
and
Nicholas Schramek
2
1 DSM Nutritional Products Ltd., Basel, Switzerland
2 Lehrstuhl fu
¨
r Biochemie, Technische Universita
¨
tMu
¨
nchen, Lichtenbergstr, Garching, Germany
Introduction
More than 3000 metric tons of vitamin B
2
(riboflavin;
6) are produced per year for use in human nutrition,
animal husbandry and as a food colorant. In recent
years, efficient fermentation processes have replaced
chemical synthesis for manufacturing the vitamin [1,2].

¨
tMu
¨
nchen,
Lichtenbergstr. 4, D-85747 Garching,
Germany
Tel: +49 089 289 13336
Fax: +49 089 289 13363
E-Mail:
(Received 16 March 2009, Revised 24 May
2009, accepted 28 May 2009)
doi:10.1111/j.1742-4658.2009.07118.x
GTP cyclohydrolase II catalyzes the first dedicated step in the biosynthesis
of riboflavin and appears to be a limiting factor for the production of the
vitamin by recombinant Bacillus subtilis overproducer strains. Using error-
prone PCR amplification, we generated a library of the B. subtilis ribA
gene selectively mutated in the GTP cyclohydrolase II domain. The ratio of
the GTP cyclohydrolase II to 3,4-dihydroxy-2-butanone synthase activities
of the mutant proteins was measured. A mutant designated Construct E,
carrying seven point mutations, showed a two-fold increase in GTP cyclo-
hydrolase II activity and a four-fold increase in the K
m
value with GTP as
the substrate. Using the analog 2-amino-5-formylamino-6-ribosylamino-
4(3H)-pyrimidinone 5¢-triphosphate as the substrate, the mutant showed a
rate enhancement by a factor of about two and an increase in the K
m
value
by a factor of about 5. A series of UV absorption spectra obtained in
stopped-flow experiments using the wild-type and mutant enzymes revealed

system of intermediate 8 is then hydrolytically released,
and the reaction is terminated by hydrolysis of the
phosphodiester bond between the covalently bound
intermediate 9 and the protein.
This article describes studies directed at an increase
in the overall reaction rate of GTP cyclohydrolase II.
Results
Whereas most enzymes of the riboflavin biosynthetic
pathway have low catalytic rates, the activity of GTP
cyclohydrolase II appears to be rate limiting for the
overall productivity of a recombinant B. subtilis strain
[14]. Both initial steps of the convergent riboflavin bio-
synthetic pathway are catalyzed in B. subtilis by the
bifunctional RibA protein comprising a GTP cyclohy-
drolase II and a 3,4-dihydroxy-2-butanone 4-phosphate
domain on the same subunit. In order to increase selec-
tively the GTP cyclohydrolase II activity, the gene seg-
ment specifying the cognate protein domain was
subjected to in vitro mutagenesis by error-prone PCR
(on average, two to five mutations per gene), and the
resulting amplificates were ligated to the gene segment
specifying the 3,4-dihydroxy-2-butanone 4-phosphate
domain (that had not been subjected to mutagenesis).
The resulting, mutated genes were ligated into the
expression plasmid pQE60 and transformed into an
Escherichia coli strain carrying a ribA
)
mutation
(Fig. 3). Growth occurred only if the mutated ribA
A

Numbering of the amino acid residues included the 14
amino acids of the His-tag. The original start methio-
nine of RibA became amino acid residue 15. The
neutral mutation, T203S, of A#1 G2, was removed,
and mutation Y210C, which was found in another
mutant of the library, was introduced in return. By
SDS gel chromatography it became apparent that the
Fig. 2. Hypothetical reaction mechanism of GTP cyclohydrolase II
[13].
mrgshhhhhhgidh
Fig. 3. Generation of the cyclohydrolase II mutant library. The DHB
synthase and cyclohydrolase II domains were separately amplified
by PCR to permit the integration of random mutations only into the
cyclohydrolase II domain. Afterwards, the two PCR products were
combined by a third PCR, digested by EcoRI and BamHI, and trans-
formed into the cyclohydrolase II-deficient E. coli strain Rib7
(pREP4).
M. Lehmann et al. Biosynthesis of riboflavin – generation of an improved cyclohydrolase II
FEBS Journal 276 (2009) 4119–4129 ª 2009 The Authors Journal compilation ª 2009 FEBS 4121
mutation reduced the susceptibility to proteolytic
cleveage into two typical fragments of the RibA wild-
type protein. The new mutant (Y210C, A290T, A296T)
was used as template for a second cycle of mutagenesis
and selection (10 000 mutants). It afforded 351 novel
candidate strains. After rescreening, 10 mutant proteins
were purified and characterized, and their effective
mutations were determined. The best combinations of
the newly found mutations resulted in Constructs C
(Y210C, A290T, Q293R, A296T, K322R, M361I) and
E (Y210C, A290T, Q293R, A296T, K322R, F339Y,

values (Table 1).
It appears likely that the hydrolytic opening of the
imidazole ring of GTP has the highest Gibbs free
energy barrier of all partial reactions in the GTP
cyclohydrolase II trajectory. However, the comparative
steady-state analysis using the natural substrate, GTP,
and the ring-opened reaction intermediate 10, suggests
that the increase in the overall rate constants observed
with the mutated proteins is not caused by a lowering
of that free energy barrier.
Previously, we studied GTP cyclohydrolase II of
E. coli using presteady-state kinetic analysis [13]. Unex-
pectedly, those experiments had suggested a relatively
slow formation of the phosphoguanosyl derivative 7
under release of pyrophosphate. That covalently bound
moiety appeared to undergo rapid hydrolytic release of
formate from the imidazole ring and ⁄ or hydrolytic
cleavage of the phosphodiester bond. It was, in fact, a
Fig. 4. Steady-state kinetics of GTP cyclohydrolase II from Bacil-
lus subtilis, using GTP as the substrate. Symbols represent the
experimental data. Lines represent the Michaelis–Menten approxi-
mation (—, wild-type; —, Construct E; ÆÆÆÆ, Construct C).
Fig. 5. Steady-state kinetics of GTP cyclohydrolase II from Bacil-
lus subtilis using 2-amino-5-formylamino-6-ribosylamino-4(3H)-pyri-
midinone 5¢-triphosphate (Compound 10) as the substrate. Symbols
represent the experimental data (310 nm). Lines represent the
Michaelis–Menten approximation (—, wild-type; - - -, Construct E).
Table 1. Kinetic properties of different GTP cyclohydrolase II
proteins from Bacillus subtilis.
GTP as

the photometric signal by stray light. An in-depth
kinetic analysis of the single-turnover data was not
possible under these experimental conditions. Never-
theless, the data enabled a comparison to be made
of the different B. subtilis proteins as well as a
comparison with the E. coli protein.
Figure 6 shows a single-turnover experiment with
wild-type GTP cyclohydrolase II of B. subtilis that was
performed using an enzyme ⁄ substrate ratio of 1 : 0.7.
The reaction was characterized by a decrease in absor-
bance at 252 nm and an apparently synchronous
increase in absorbance at 292 nm. The superposition
of spectra taken from the series showed an apparent
isosbestic point at 278 nm, which suggests an apparent
0.160
0.120
0.080
0.040
0.0
240 260 280 300 320 340 360 380 400
Wavelength (nm)
Absorbance
0.1
0.5
1
10
50
Time (s)
5
Fig. 6. Optical spectra from a stopped-flow experiment with wild-

1 : 2.5. Progression curves are shown at 295 nm, a
wavelength where the absorption is dominated by
the nascent 2,5,6-triaminopyrimidinone motif present
in the hypothetical covalent intermediate 9 and the
product 10, with only a minor contribution (by the
substrate, GTP 1) to the absorbance. Differentials of
the absorbance at 295 nm (dA
295
⁄ dt) are shown in the
frames on the right side of the Figure. Whereas the
curves for wild-type and mutant proteins were similar
under conditions where there was a slight excess of
protein over substrate, the similarity broke down
under presteady-state conditions with an excess of
AB
C
D
Fig. 8. Numerical simulation of stopped-flow data of wild-type GTP cyclohydrolase II from Bacillus subtilis (A, B) and Construct E (C, D),
using GTP as the substrate. The enzyme solution was mixed with substrate solution at molar ratios of 1 : 0.7 (s), 1 : 1.3 (
) and 1 : 2.5 ( ).
Symbols represent the experimental data and lines represent the numerical simulation using the kinetic constants in Table 2. Data sets were
analyzed using the program
DYNAFIT [24].
Table 2. Single-turnover rate constants of different GTP cyclo-
hydrolase II proteins from Bacillus subtilis using GTP as the
substrate.
Wild type Construct E
k2 ⁄ k1 (l
M) 4.36 ± 0.18 25.6 ± 0.8
k3 (min

reaction steps catalyzed by the respective enzyme.
Specifically, GTP cyclohydrolase I catalyzes the ring
opening of GTP, followed by hydrolytic deformyla-
tion, Amadori re-arrangement and ring closure
resulting in the production of dihydroneopterin
triphosphate, which serves as the first committed
precursor in the biosynthesis of tetrahydrofolate and
tetrahydrobiopterin (for review see Refs. [17]). The
recently discovered MptA protein produces the 2¢,3¢-
cyclophosphate of dihydroneopterin that is believed to
serve as a precursor for the biosynthesis of tetrahydro-
methanopterin, a one-carbon transfer cofactor of
methanogenic coenzymes [18]. GTP cyclohydrolase II,
the subject of this article, is believed to catalyze the
release of phosphate from GTP, which is conducive to
the formation of a covalent guanylyl adduct that can
be resolved by a sequence of ring opening, deformyla-
tion and ⁄ or phosphodiester cleavage resulting in the
production of 2, the first committed intermediate in
the biosynthesis of riboflavin. The covalent adduct
remains to be confirmed by direct evidence, but the
recently reported 3D structure suggests that Arg128 is
the acceptor of the phosphodiester linkage in the
E. coli protein [19]. GTP cyclohydrolase III of Archae-
bacteria catalyzes ring opening without accessory
reactions [20,21]. The resulting 2-amino-5-formylami-
no-6-ribosylamino-4(3H)-pyrimidinone 5¢-triphosphos-
phate is believed to serve as the first committed
intermediate in the biosynthesis of riboflavin and of
the deazaflavin-type cofactor F420.

corruption of the optical readout by the stray light
contribution, it was not possible to perform a detailed
deconvolution of the stopped-flow kinetic da ta, in
analogy to the earlier study performed with the E. coli
enzy me; despite this shortcoming, the data suggest
that the kinetic profile of the B. subtilis enzyme is
indeed similar to that of the E. coli enzyme, with the
formation of the covalent adduc t as a relatively slow
initial step.
Even without the opportunity to conduct a compre-
hens ive data deconvolution, stopped- flow analysis
under single-turnover conditions, as well as prest eady-
state conditions, afforded useful information for
comp arison of the wild-type protein with the mutant
Cons truct E. Specifically, absorbance progression
curves of the two pr oteins were similar under strict
single-turnover conditions conducted with a molar
excess of enzyme over substrate (Figs 7 and 8). By
contrast, the kinetic differences became progressively
larger when the substrate was proffered in excess ( pre-
steady-state conditions with a protein ⁄ substrate ratio
up to a value of 1 : 2.5; Fig. 8). Clearly, under these
conditions, the mutan t protein generated product at a
higher overall rate than did the wild-type protein.
Moreover, these observations are perfectly in line with
the stea dy-state analysis, indicati ng an approximately
two-fold increased k
cat
of Construct E compared with
the wild-type protein. These findings are best

complex would be irrelevant for the catalytic rate
under saturating conditions.
There is precedent for enzymes with substrate release
as the rate-limiting step for the overall reaction. Never-
theless, it came as a surprise that the extensive
enzyme-evolution process conducted in this study
failed to increase the rate constant significantly for any
of the catalytical partial reactions sensus strictiori
(resulting in chemical modification of the reactant),
although the overall reaction velocity was increased
(via accelerated product release, as described earlier).
For the practical purpose of improving the produc-
tivity of a riboflavin-overproducing strain by intro-
ducing the improved GTP cyclohydrolase II domain
into a riboflavin-producing strain, it is irrelevant
whether the rate acceleration is caused by enhanced
substrate conversion or by enhanced product-release
rates. The increase in K
m
accompanying the increase
in V
max
can be tolerated in the technical application
because the cellular GTP concentrations are well
above the K
m
, even for Construct E; moreover, this
enzyme would be working under substrate-saturating
conditions in the in vivo situation.
Based on the recently reported X-ray structure of the

pQE60-ribANhis (Table 1) was used in which ribA from
B. subtilis was cloned between the EcoR1 and the BamH1
sites of pQE60 (Table 1). The gene itself was slightly modi-
fied by the addition of the DNA sequence motif 5¢-GAA
TTCattaa
agaggagaaattaact ATG AGA GGA TCT CAC
CAT CAC CAT CAC CAT GGG ATC GAT CAT-3¢ in
front of the start codon. The modified ORF features an
Nde1 site at the start codon and specifies a RibA protein
carrying an N-terminal 6· His tag. In order to introduce
mutations exclusively into the cyclohydrolase II domain of
ribA and not into the 3,4-dihydroxy-2-butanone (DHB)
synthase domain, error-prone PCR was performed [(using
the oligonucleotides ribA3S and ribA4AS (Table 4) as
primers] only on the DNA fragment coding for the cyclo-
hydrolase II domain. Reaction mixtures for error-prone
PCR contained 5 mm MgCl
2
, 0.7 mm MnCl
2
, 0.2 mm
Fig. 9. Structure of wild-type GTP cyclohydrolase II from Escheri-
chia coli [22]. The positions that are homologous to the mutations
in Construct E are marked in yellow.
Biosynthesis of riboflavin – generation of an improved cyclohydrolase II M. Lehmann et al.
4126 FEBS Journal 276 (2009) 4119–4129 ª 2009 The Authors Journal compilation ª 2009 FEBS
nucleotide triphosphates, 10 ng of template DNA, 2 lm of
each primer and 2.5 U of Taq polymerase in 50 lL of the
1· buffer supplied with the polymerase. The reaction condi-
tions were as follows: step 1, 3 min, 95 °C; step 2, 30 s,

)1
of
kanamycin and 100 lgÆmL
)1
of ampicillin) and were grown
overnight. Dimethylsulfoxide (15 lL per well) was added,
and the plates were stored at )80 ° C. For further rounds of
mutagenesis, the original ribA gene was replaced with the
improved mutants, as selected.
Bacterial culture
Aliquots (5 lL) from each well of a master plate were trans-
ferred into a deep-well plate filled with 250 lL of LB medium
(supplemented with 25 lgÆmL
)1
of kanamycin and
100 lgÆmL
)1
of ampicillin) per well. The plates were incu-
bated overnight (37 °C, 250 r.p.m.) on a rotary shaker. The
next morning, LB medium (1.2 mL) supplemented with
25 lgÆmL
)1
of kanamycin and 100 lgÆmL
)1
of ampicillin
was added to each well. The plates were incubated at 30 °C
on a rotary shaker at 250 r.p.m. After 6 h, isopropyl thio-b-
d-galactoside was added to a final concentration of 0.5 mm.
The plates were incubated for another 16 h at 30 °C with
shaking (250 r.p.m.). At the end of this incubation period,

and 35 gÆL a-naphthol was added together with 50 lLof
saturated creatine solution. The mixture was incubated at
20 °C for periods of 60 to 120 min, and the absorbance at
525 nm was determined [23].
Protein purification
Frozen E. coli cell mass (25 g) was thawed in 60 mL of
50 mm Tris–HCl (pH 8.0), containing 0.3 m sodium chlo-
ride and 10 mm magnesium chloride. The cells were
Table 3. Microorganisms and plasmids used in this study.
Strain or plasmid
Genotype or relevant
characteristics
Reference
or source
Escherichia coli Rib7 thi leu pro lac ara xyl
endA recA hsd r
-
m
-
pheS supE44 rib
[15]
Plasmids
pREP4 Low-copy-number plasmid
expressing lacI
Quiagen Inc.
pQE60 Expression plasmid for
E. coli
Quiagen Inc.
pQE60-ribA-Nhis pQE60 with an N-terminally
tagged ribA from

2
,2mm dithiothreitol and pro-
tein in a total volume of 400 lL. Experiments were per-
formed at 30 °C. The reaction was initiated by the addition
of GTP to a predetermined concentration (0.017–1.7 mm).
The assay was monitored photometrically at 310 nm. Reac-
tion rates were calculated using an absorption coefficient of
7.43 mm
)1
Æcm
)1
for 2,5-diamino-6-ribosylamino-4(3H)-pyri-
midinone 5¢-phosphate .
Stopped-flow kinetic experiments
Experiments were performed using an SFM4 ⁄ QS apparatus
from Bio-Logic (Claix, France) equipped with a linear
array of three mixers and four independent syringes. The
content of a 1.5-mm light path quartz cuvette behind the
last mixer was monitored using a Tidas diode array spec-
trophotometer (200–610 nm) equipped with a 15 W deute-
rium lamp as the light source (J&M Analytische Meß- und
Regeltechnik, Aalen, Germany). The reaction buffer con-
tained 50 mm Tris–HCl (pH 8.5), 100 mm NaCl, 10 mm
MgCl
2
and 2 mm dithiothreitol. The enzyme solution was
mixed with substrate solution at a temperature of 35 °C
and a total flow rate of 4 mLÆs
)1
. The calculated dead time

(Mosc) 34, 2883–2892.
9 Volk R & Bacher A (1990) Studies on the 4-carbon pre-
cursor in the biosynthesis of riboflavin. Purification and
properties of L-3,4-dihydroxy-2-butanone-4-phosphate
synthase. J Biol Chem 265, 19479–19485.
10 Wacker H, Harvey RA, Winestock CH & Plaut GW
(1964) 4-(1¢-d-ribitylamino)-5-amino-2,6-dihydroxypyr-
imidine, the second product of the riboflavin synthetase
reaction. J Biol Chem 239, 3493–3497.
11 Kaiser J, Schramek N, Eberhardt S, Puttmer S, Schus-
ter M & Bacher A (2002) Biosynthesis of vitamin B2.
Eur J Biochem 269, 5264–5270.
12 Ritz H, Schramek N, Bracher A, Herz S, Eisenreich W,
Richter G & Bacher A (2001) Biosynthesis of riboflavin:
studies on the mechanism of GTP cyclohydrolase II.
J Biol Chem 276, 22273–22277.
13 Schramek N, Bracher A & Bacher A (2001) Biosynthe-
sis of riboflavin. Single turnover kinetic analysis of
GTP cyclohydrolase II. J Biol Chem 276, 44157–44162.
14 Hu
¨
mbelin M, Griesser V, Keller T, Schurter W, Haiker
M, Hohmann HP, Ritz H, Richter G, Bacher A & Van
Loon APGM (1999) GTP cyclohydrolase II and 3,4-di-
hydroxy-2-butanone 4-phosphate synthase are rate-
limiting enzymes in riboflavin synthesis of an industrial
Bacillus subtilis strain used for riboflavin production.
J Ind Microbiol Biotechnol 22 , 1–7.
15 Richter G, Ritz H, Katzenmeier G, Volk R, Kohnle A,
Lottspeich F, Allendorf D & Bacher A (1993) Biosyn-

istry 47, 230–242.
22 Ren J, Kotaka M, Lockyer M, Lamb HK, Hawkins
AR & Stammers DK (2005) GTP cyclohydrolase II
structure and mechanism. J Biol Chem 280, 36912–
36919.
23 Picollelli MA, Viitanen PV & Jordan DB (2000)
Spectrophotometric determination of 3, 4-dihydroxy-2-
butanone-4-phosphate synthase activity. Anal Biochem
287, 347–349.
24 Kuzmic P (1996) Program DYNAFIT for the analysis
of enzyme kinetic data: application to HIV proteinase.
Anal Biochem 237, 260–273.
M. Lehmann et al. Biosynthesis of riboflavin – generation of an improved cyclohydrolase II
FEBS Journal 276 (2009) 4119–4129 ª 2009 The Authors Journal compilation ª 2009 FEBS 4129