ARTICLE IN PRESS

Metabolic Engineering 11 (2009) 243–252

Contents lists available at ScienceDirect

Metabolic Engineering
journal homepage: www.elsevier.com/locate/ymben

Regular Article

Transcriptome analysis guided metabolic engineering of Bacillus subtilis

for riboflavin production
Shuobo Shi a,b,1, Tao Chen a,b,1, Zhigang Zhang c,d, Xun Chen a,b, Xueming Zhao a,b,!
a
Department of Biochemical Engineering, School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China
b
Key Laboratory of Systems Bioengineering, Ministry of Education, Tianjin University, Tianjin 300072, China
c
BioTechnology Institute, 1479 Gortner Avenue, University of Minnesota, St. Paul, MN 55108, USA
d
Department of Chemical Engineering and Materials Science, 421 Washington Avenue SE, University of Minnesota, Minneapolis, MN 55455, USA

a r t i c l e in fo abstract

Article history: A comparative transcriptome profiling between a riboflavin-producing Bacillus subtilis strain RH33 and
Received 24 December 2008 the wild-type strain B. subtilis 168 was performed, complemented with metabolite pool and nucleotide
Received in revised form sequence analysis, to rationally identify new targets for improving riboflavin production. The pur operon
20 April 2009
(purEKBCSQLFMNHD) together with other PurR-regulated genes (glyA, guaC, pbuG, xpt-pbuX, yqhZ-folD,
Accepted 5 May 2009
Available online 13 May 2009
and pbuO) was all down-regulated in RH33, which consequently limited the supply of the riboflavin
precursors. As 5-phospho-ribosyl-1(a)-pyrophosphate (PRPP) strongly inhibits the binding of PurR to its
Keywords: targets, it was inferred that the reduced expression of PurR-regulated genes might be caused by a low
Riboflavin PRPP pool, which was subsequently confirmed by metabolite analysis. Thus, we selected and co-
Bacillus subtilis
overexpressed prs and ywlF genes in RH33, which are involved in the biosynthetic pathway of PRPP from
Transcriptome
ribulose-5-phosphate. This co-amplification led to an elevated PRPP pool and thus the increased
PRPP pool
Metabolic engineering transcript abundances of PurR-regulated genes participated in riboflavin precursor biosynthesis. The
riboflavin titer was increased by 25% (up to 15 g l!1) in fed-batch fermentation.
& 2009 Elsevier Inc. All rights reserved.

1. Introduction subtilis, currently the most competitive riboflavin producer, has

In the biotechnology industry, classical strain improvement
has achieved a long history of success (Crueger and Crueger, 1984;
Peberdy, 1985), which relied on random mutagenesis and
selection. These methods are still very useful, especially with
the development of efficient strain selection methods (Gall et al.,
2008). However, unwanted changes in physiology and growth
retardation may occur alongside the desired improvements. Since
the introduction of metabolic engineering, a more rational
improvement approach emerges for microbial development
(Bailey, 1991; Stephanopoulos et al., 1998).
Riboflavin (vitamin B2) serves as a precursor for the synthesis
of the coenzymes flavin mononucleotide (FMN) and flavin adenine
dinucleotide (FAD), which are used as electron acceptors for many
oxidoreductases (Stahmann et al., 2000). As such, riboflavin is
required for a wide variety of cellular processes and is supple-
mented for feed and food fortification purposes in humans and
animals to maintain health. The Gram-positive bacterium Bacillus
! Corresponding author at: Department of Biochemical Engineering, School of
Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China.
Fax: +86 22 27406770.
E-mail address: xmzhao@tju.edu.cn (X. Zhao).
1
These authors contributed equally to this work.
been widely adopted in the commercial riboflavin production
processes (Knorr et al., 2007; Wu et al., 2007; Zamboni et al.,
2003). The biosynthesis of riboflavin in B. subtilis occurs through
seven enzymatic steps starting from GTP and ribulose-5-phos-
phate, which is shown in Fig. 1. The riboflavin producer B. subtilis
was initially developed using the ‘‘classical’’ strain development
approach that relied on iterative cycles of random mutagenesis
and selection to create genetic diversity and identify improved
riboflavin mutants (Perkins et al., 1991; Stahmann et al., 2000).
Then, a number of conceivable strategies were carried out to
construct high-level riboflavin-producing B. subtilis strains. These
include enhancement of both gene dosages and transcriptional
level of riboflavin operon in the mutants (Perkins et al., 1999),
constitutive expression of the key gene (ribA) in riboflavin
biosynthetic pathway (Humbelin
+ et al., 1999), enhancing energy
generation and reducing maintenance metabolism (Zamboni
et al., 2003), increasing precursor supply by modulating carbon
flow through pentose phosphate pathway (Zamboni et al., 2004a;
Zhu et al., 2006), and deregulating gapB expression by ccpN
knockout based on screening B. subtilis transposon mutants
(Tännler et al., 2008).
However, these strategies were not based on a comprehensive
analysis of the complex microbial metabolism and regulation,
which would further facilitate the successful and efficient

1096-7176/$ - see front matter & 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.ymben.2009.05.002
 ARTICLE IN PRESS
244 S. Shi et al. / Metabolic Engineering 11 (2009) 243–252

Fig. 1. Schematic overview of expression profiles of genes involved in relevant pathways of riboflavin production. The numbers are the ratios of the comparative expression
levels in B. subtilis strain RH33 vs. 168. Blue lettering indicates downregulation, red indicates up-regulation, and black indicates without notable changes. The genes that
were underlined were selected for overexpression. G6P, glucose-6-phosphate; F6P, fructose-6-phosphate; GAP, D-glyceraldehyde 3-phosphate; PGA, 3-phosphoglycerate;
PYR, pyruvic acid; AcoA, acetyl coenzyme A; Ser, serine; Gly, glycine; 10-formyl-THF, 10-formyl tetrahydrofolate; gluconate-6-P, 6-phospho-D-gluconate; Ru-5-P, ribulose-5-
phosphate; Ribo-5-P, ribose-5-phosphate; PRPP, phosphoribosylpyrophosphate; Gln, glutamine; X5P, xylulose 5-phosphate; E4P, D-erythrose 4-phosphate; IMP, inosine
monophosphate; XMP, xanthosine monophosphate; GMP, guanosine mono-phosphate; GTP, guanosine tri-phosphate; DARPP, 2,5-diamino-6-ribosylamino-4(3H)-
pyrimidinone-50 -phosphate; ARPP, 5-amino-6-(50 -phosphoribosylamino)uracil; ArPP, 5-amino-6-(50 -phosphoribitylamino)uracil; ArP, 4-(1-D-ribitylamino)-5-amino-2,6-
dihydroxypyrimidine; DHPB, 3,4-dihydroxy-2-butanone 4-phosphate; DRL, 6,7-dimethyl-8-ribityl-lumazine; FAD, flavin adenine dinucleotide (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.).

microbial metabolic engineering (Nielsen and Olsson, 2002). Table 1

Recently, as a holistic and discovery-driven approach, transcrip- Strains and plasmids used in this study.
tome analysis has been successfully applied in metabolic
Strain or Description of genotype Source
engineering for rationally probing the complex gene and meta- plasmid
bolic regulatory networks. Novel target genes have been identified
to optimize microbial production strains (Choi et al., 2003; Harris Strains
et al., 2009; Izallalen et al., 2008; Jaluria et al., 2007; Park et al., B. subtilis Wild-type BGSC
168
2007; Peebles et al., 2009; Sindelar and Wendisch, 2007; Wierckx B. subtilis Emr, Cmr, containing multiple riboflavin operons Laboratory
et al., 2008). RH33 stock
Here we took a strategy to increase riboflavin production in a B. subtilis Emr, Cmr, Spcr, containing a P43-prs (CDS) fragment This study
riboflavin-producing B. subtilis strain based on comparative RH33-Prs integrated in the chromosome of RH33
B. subtilis Emr, Cmr, Spcr, containing a P43-prs-ywlF (CDS) This study
transcriptome analysis between a riboflavin high-producer RH33
RH33-PY fragment integrated in the chromosome of RH33
and wild type 168, integrated with DNA sequencing and E. coli Top10 Host strain for constructing plasmids Laboratory
metabolite pool measurement. Our integrated approach allowed stock
us to understand genome-wide transcriptional differences under- Plasmids
lying strain performance, and to identify potential metabolic pUC18 Ampr BGSC
bottlenecks for riboflavin production. We rationally overexpressed pSG1192 Ampr, Spcr BGSC
two genes simultaneously to activate precursor purine biosynth- pHPL10 pHP13, containing a P43 promoter Laboratory
stock
esis pathways by modulating global regulator activity via
pRPU10 Ampr, containing a prs (CDS) fragment This study
metabolite pool manipulation. This strategy was capable of pRPU12 Ampr, containing a P43-prs (CDS) fragment This study
elevating riboflavin titer by 25% in B. subtilis RH33 (up to 15 g l!1) pRPU13 Ampr, Spcr, containing a P43-prs (CDS) fragment This study
in fed-batch fermentation. pRPU14 Ampr, containing a P43-prs-ywlF (CDS) fragment This study
pRPU15 Ampr, r
Spc , containing a P43-prs-ywlF (CDS) fragment This study

BGSC, Bacillus Genetic Stock Center (http://www.bgsc.org/).

2. Materials and methods

2.1. Strains and plasmids

Bacterial strains and plasmids used in this work are listed in
Table 1. Two different riboflavin-producing mutants B. subtilis
RH13 and B. subtilis J617 were derived from B. subtilis 168 by
multiple rounds of selection with azaguanine (Azr), decoyinine
(Dcr) and roseoflavin (RoFr) for resistance mutations that
deregulate the riboflavin biosynthetic pathway. An integration
vector pRB63 was constructed by inserting a native rib operon into
pDG364 (BGSC, http://www.bgsc.org/). Then the vector pRB63
was integrated into the B. subtilis RH13 chromosome at the native
rib operon locus by single crossover and selected at 5 mg ml!1
chloramphenicol. Subsequently, the copy number of the
integrated pRB63 was increased by selecting colonies that grew
at higher chloramphenicol concentrations. Colonies that were able
to grow up with 40 mg ml!1 chloramphenicol were isolated, and
one was denoted as B. subtilis RH13H[pRB63]n. Another
 ARTICLE IN PRESS
S. Shi et al. / Metabolic Engineering 11 (2009) 243–252
integration vector pRB62 carried erythromycin resistant was
integrated into the B. subtilis J617 chromosome at the native
thrC locus by double crossover. The vector was constructed by
insertion of a 4.4-kb P43-modified rib operon (substituting the
native ribP1 promoter and RFN regulatory element with the
constitutive P43 promoter) into pDG1664 (BGSC, http://
www.bgsc.org/). The resulting strain was denoted as B. subtilis
J617H[pRB62]. Subsequently, B. subtilis RH13H[pRB63]n and B.
subtilis J617H[pRB62] were fused by protoplast fusion. The fusants
were selected in solid regeneration medium containing
40 mg ml!1 chloramphenicol and 5 mg ml!1 erythromycin. One
clone was denoted as B. subtilis RH33, which was used as the
parental strain in our study.
Plasmids were constructed by the following procedures. Firstly,
polymerase chain reaction (PCR) was used to amplify structural
genes of prs and ywlF with the following primers: prs-up 50 -
GGGGCCCGGGCCAGAGCGAGACAAGTAAA, prs-down 50 -GGG-
GGAGCTCGCTAGCTCCTATTACAAACAATACCCA, ywlF-up 50 -GGG-
GCCCGGGGCTAGCGGCTGCGCGGTCAATA, ywlF-down 50 -GGGG-
GAGCTCGCGGCCGCTTGTTTCAATTCCGCTTGGTC, based on the
published B. subtilis 168 genome sequence (Kunst et al., 1997).
The prs PCR product was ligated into pUC18 using XmaI and SacI
restriction sites to construct pRPU10. The constitutively expressed
P43 promoter (Wang and Doi, 1984) was obtained from the vector
pHPL10 (unpublished work) after digested with BamHI and XmaI.
It was then ligated in the BamHI and XmaI restriction sites of
pRPU10 to get plasmid pRPU12. The ywlF PCR product was
digested with NheI and SacI, and cloned into NheI-SacI-sites of
pRPU12 to construct pRPU14. To facilitate further research, a
spectinomycin resistance gene (spc) from pSG1192 was inserted at
the SalI site of pRPU12 and pRPU14 to give plasmids pRPU13 and
pRPU15, respectively. All DNA manipulations were carried out as
described previously (Sambrook and Russell, 2001). LB medium
supplemented with corresponding antibiotics was used as the
standard medium for plasmid construction in Escherichia coli. The
plasmids were transformed into competent cells of B. subtilis
RH33 as described previously (Anagnostopoulos and Spizizen,
1961).
2.2. Growth conditions
For transcriptome and quantitative RT-PCR analyses, 5-phos-
pho-ribosyl-1(a)-pyrophosphate (PRPP) pool, and riboflavin mea-
surements, a preculture of B. subtilis RH33 and B. subtilis 168 were
inoculated from a fresh LB agar plate and cultured to exponential
growth period in LBG medium (LB medium with 1% glucose) at
37 1C. The intracellular metabolites, PRPP, ribose-5-phosphate and
ribulose-5-phosphate, were determined during exponential
growth period in minimal medium with 20 g l!1 glucose, 2 g l!1
(NH4)2SO4, 13.1 g l!1 KH2PO4, 6 g l!1 K2HPO4, 1.2 g l!1 NaC6H5O7 "
2H2O, 10 mg l!1 MgSO4 " 7H2O and supplemented with trypto-
phan, phenylalanine and tyrosine (25 mg l!1 each).
Fed-batch fermentation was performed in B. subtilis strains for
riboflavin overproduction. The strain was revived by growing on
LB agar slants. The seed cultures of the revived B. subtilis strains
were prepared in 500-ml shake flasks at 41 1C with 240 rpm. The
seed medium contained 20 g l!1 glucose, 5 g l!1 yeast extract,
10 g l!1 NaNO3, 5 g l!1 NH4NO3, 1 g l!1 KH2PO4, 1.65 g l!1 K2HPO4,
1.5 g l!1 MgSO4 " 7H2O, 0.03 g l!1 FeCl2, 0.04 g l!1 MnCl2 " 4H2O,
0.04 g l!1 ZnCl2 " 7H2O. After 14 h incubation at 41 1C, 150 ml of the
seed culture was transferred into in a 5-l bioreactor (Bao Xing,
China) containing 2350 ml fermentation medium. The initial
medium of fed-batch fermentations contained 20 g l!1 glucose,
5 g l!1 yeast extract, 10 g l!1 NaNO3, 5 g l!1 NH4NO3, 4 g l!1
KH2PO4, 7.5 g l!1 K2HPO4, 1.5 g l!1 MgSO4 " 7H2O, 0.03 g l!1 FeCl2,
245
0.03 g l!1 ZnCl2 " 7H2O, 0.04 g l!1 MnCl2 " 4H2O. The cells were
grown in the batch mode for 6 h after inoculation, and then the
feeding process was continued from 6 to 48 h. The feed solution
contained 650 g l!1 glucose, 10 g l!1 yeast extract, 10 g l!1 NaNO3,
5 g l!1 NH4NO3, 5 g l!1 KH2PO4, 5 g l!1 K2HPO4, and 2 g l!1
MgSO4 " 7H2O. The residual glucose concentration in the fermen-
tor was maintained no more than 10 g l!1 by controlling the
glucose feed rate manually. The cultivation was carried out for
48 h at the agitation rate of 900 rpm, with an aeration rate of
1.1 vvm and temperature of 41 1C. The head pressure is maintained
at 0.5 atm. pH was maintained at 6.8 with 1 M H2SO4 and 10%
ammonia. All the experiments were carried out independently in
biological triplicates, and the reported results were the average of
three replicate experiments.
2.3. Transcriptome analysis
About 25 ml of the culture was used for preparation of total
RNA, which was extracted using RiboPureTM-Bacteria Kit (Ambion,
UK). Total RNA was stored at !70 1C until use. The quantity and
quality of RNA was analyzed by UV spectrophotometry and
denaturing formaldehyde agarose gel electrophoresis, respec-
tively. Probe sets on the B. subtilis Genome Array (Antisense)
were designed based on the wild-type B. subtilis 168 genome
sequence data of Kunst et al. (1997) by Affymetrix (Santa Clara,
CA, USA). An aliquot of 10 mg B. subtilis total RNA was used to
synthesize first-strand cDNA with random primers and Super-
Script II reverse transcriptase. Then the cDNA was fragmented to
50–200 bp and labeled by biotin. After hybridization at 45 1C for
16 h at 60 rpm, the microarray slides were washed and stained on
Affymetrix Fluidics Station 450. The scanned images were
obtained with the GeneChip Scanner 3000 (Affymetrix) and were
analyzed using the default setting of GeneChip Operating Soft-
ware (GCOS 1.4). Then a LOWESS normalization method was
performed to normalize the different arrays using dChip software.
The differentially expressed genes were identified through over-
lapping gene analysis of biological duplicate experiments using a
2-fold change as an empirical criterion.
2.4. Quantitative RT-PCR
To validate transcriptome results, relative abundances of
selective transcripts were measured by quantitative reverse
transcription-PCR, which was carried out by an iCycler (Bio-Rad,
USA) using the QuantiTect SYBR Green RT-PCR kit (Qiagen,
Germany) according to the manufacturer’s instructions. In brief,
100 ng of DNA-free total RNA was used in a total reaction volume
of 50 ml with 0.4 mM of each primer (Supplementary Table 1). The
fold change of each transcript in each sample relative to the
control sample was measured in triplicates, normalized to internal
control gene gapA and calculated according to the comparative Ct
method (Livak and Schmittgen, 2001).
2.5. Metabolite concentration measurement
Immediately upon harvest, 5 ml of cell sample was added to
15 ml of quenching fluid containing 70 mM HEPES in 60% (v/v)
aqueous methanol. The samples were centrifuged to separate the
quenching fluid and the remaining cell pellets were resuspended in
35% (v/v) perchloric acid. After one freeze–thaw cycle the sample
was neutralized by 5 M K2CO3. The precipitate was removed by
another centrifugation, and resulting supernatants were stored at
!80 1C until analysis. Measurement of intracellular PRPP concen-
trations was carried out using the methods described by Kornberg
et al. (1955). The ribose-5-phosphate and ribulose-5-phosphate
 ARTICLE IN PRESS
246 S. Shi et al. / Metabolic Engineering 11 (2009) 243–252
were measured using a Luminescence Spectrophotometer (Hoque
et al., 2005).
2.6. Enzyme activity assay in crude extracts
The crude cell extract was prepared as described previously
(Fisher and Mangasanik, 1984). For enzyme assay, strains were
cultivated at 37 1C in minimal medium. Quantification of PRPP
synthetase activity, ribose-5-phosphate isomerase activity, and
transketolase activity were performed as described (Hove-Jensen
and Maigaard, 1993; Iida et al., 1993; Sakuma et al., 1991). Specific
activity was expressed as nmol min!1 mg!1 protein. Protein
content was determined according to the Bradford method.
2.7. Analytical methods
Cell growth (OD600 nm) was monitored with a spectrophot-
ometer. Cell growth rate was calculated by log-linear regression
analysis of OD600 nm versus time. Glucose concentration was
determined using a glucose analyzer (Model-SBA40, Shandong,
China). For riboflavin measurements, the samples were diluted
with 0.05 M NaOH to the linear range of the spectrophotometer,
and centrifuged at 12,000 rpm for 2 min to remove the cells. Then
the OD444 nm was immediately measured. The culture broth was
centrifuged and the resulting supernatant was used for measure-
ment of the concentrations of extracellular metabolites (acetate,
acetoin, etc.) by using a HPLC system (Agilent, HP1100, USA)
equipped with a UV–visible light monitor (Bai et al., 2004).
2.8. Nucleotide sequencing
Selected genes involved in riboflavin biosynthesis and cellular
central metabolism were sequenced for both regulatory and
coding regions, using the traditional Sanger method by Beijing
AuGCT biotechnology Co., Ltd., China.
3. Results and discussion
3.1. Comparative transcriptome analysis
Transcriptome profiles of a riboflavin-producing B. subtilis RH33
was compared to wild-type B. subtilis 168 on DNA microarray using
mid-exponential phase cell samples (OD600 nm#0.90) grown in LBG
medium. In B. subtilis RH33, this condition corresponded to the
state of high riboflavin production, which was coupled to cell
growth. The specific growth rate of RH33 and 168 were 1.5770.15
and 1.6170.15 h!1, respectively. During mid-exponential phase, the
specific growth rates were nearly constant, without discernible
accumulation of byproducts such as acetate (data not shown). Our
past experiences showed that there was no significant difference in
transcriptional profiles at different time-points of mid-exponential
phase.
The microarray experiment was very reproducible. The Pearson
correlation coefficient between the biological duplicates was
about 0.95; the overlapping of differentially expressed genes
identified using an empirical 2-fold criterion from each experi-
ment was more than 90%. It was found that 619 genes showed
significant variations at the genome-wide transcriptional level
between RH33 and 168, representing about 15% of B. subtilis
genome. Quantitative RT-PCR experiment was performed in
parallel to validate microarray data (Table 2). The average fold
change of ribA and narG transcripts, were identified to be 159 and
2.99 in RH33 relative to 168, respectively. This was consistent
with microarray data, since both genes were significantly up-
Table 2
Validation of microarray data by quantitative RT-PCR.
Genes Quantitative RT-PCR Microarray
purF 0.2570.01 0.3870.01
purD 0.1170.01 0.5070.01
glyA 0.4270.02 0.6870.03
folD 0.4470.02 0.6070.04
prs 0.9570.02 1.0070.03
ywlF 0.8770.01 0.9070.06
ribA 15970.37 13.2770.08
narG 2.9970.05 6.4870.06
Fold change of each transcript in B. subtilis RH33 relative to 168 was reported.
regulated to 13.27 and 6.48-fold, respectively. The genes prs and
ywlF were identified not to be differentially expressed (1.00 and
0.90) by microarray, which was confirmed by quantitative RT-PCR
(0.95 and 0.87). Furthermore, by quantitative RT-PCR, the mean
expression levels of purF, purD, glyA, and folD were identified to be
0.25, 0.11, 0.42, and 0.44, respectively. All the four genes were
identified as significantly down-regulated (0.38, 0.50, 0.68, and
0.60) in B. subtilis RH33 by transcriptome analysis. Therefore, the
correlation between quantitative RT-PCR and microarray was
good (Table 2), suggesting the validity of the microarray gene
expression measurements.
The differentially expressed genes fell into nearly all functional
categories (Supplementary Table 2). The largest group with
altered transcriptional levels in the B. subtilis RH33 was a group
of 83 genes encoding for the transport/binding proteins and
lipoproteins, all of which are associated with the cell membrane.
Forty-eight affected genes belonged to the group involved in
metabolism of amino acids and related molecules, especially
aspartate, cysteine, glutamate, histidine, leucine, threonine,
tyrosine, and serine biosynthesis genes. Other large functional
groups changed significantly included 24 genes associated with
motility, 23 genes associated with carbohydrate metabolism
(notably myoinositol and acetoin metabolism), 23 genes asso-
ciated with metabolism of nucleotides and nucleosides (mainly
purine biosynthetic genes), 32 genes associated with transcription
regulation and 27 genes associated with phage-related function.
Therefore, it seems that complex transcriptional regulation
mechanism(s) underpinned the strain performance differences,
directly or indirectly affecting riboflavin production. Some of
these transcriptional effects will be discussed in detail as follows.
3.2. Nucleotide sequencing of selected genes
Since B. subtilis RH33 was developed mostly by repeated
random mutagenesis and selection, mutations that affect gene
function were also expected to occur. The identification of
mutational changes is necessary to fully understand riboflavin
synthesis processes in RH33. However, the unknown random
mutations may affect the activity of gene product, but may not
affect gene expression per se and thus such mutations may not be
detected by microarray. To identify possible beneficial mutations
for riboflavin production in RH33, we selected 67 genes for DNA
sequencing, including both regulatory and coding regions. These
genes participated in riboflavin biosynthesis, precursor supply
and cell central metabolism, which were simply grouped into six
functional categories (Table 3): glycolysis and TCA cycle, pentose
phosphate pathway, purine pathway and other PurR-regulated
genes, riboflavin biosynthesis and transport, glycine biosynthesis,
and nitrogen metabolism. The nucleotide sequences of six genes
were identified to contain mutations in RH33, including ribC (the
riboflavin kinase and FAD synthase gene), tkt (the transketolase
 ARTICLE IN PRESS
S. Shi et al. / Metabolic Engineering 11 (2009) 243–252 247

Table 3
Relative expression changes and nucleotide sequence mutations of selected genes directly or indirectly participated in riboflavin biosynthesis in RH33.

Gene DFold Gene product Mutations

Glycolysis and TCA cycle

pycA 1.32 Pyruvate carboxylase No
pdhA 0.68 Pyruvate dehydrogenase (E1 alpha subunit) No
pdhB 0.72 Pyruvate dehydrogenase (E1 beta subunit) No
pdhC 0.88 Pyruvate dehydrogenase (dihydrolipoamide acetyltransferase E2 subunit) No
pdhD 0.88 Pyruvate dehydrogenase/2-oxoglutarate dehydrogenase (dihydrolipoamide dehydrogenase E3 subunit) No
gapB 1.08 Glyceraldehyde-3-phosphate dehydrogenase No
citZ 1.96 Citrate synthase II (major) No
citG 0.47 Fumarate hydratase No
pta 0.77 Phosphotransacetylase No
ldh (lctE) 1.17 L-lactate dehydrogenase Gly29Gly
ackA 0.80 Acetate kinase No
alsS 0.82 Alpha-acetolactate synthase No
alsD 0.76 Alpha-acetolactate decarboxylase No
ptsG 1.07 PTS glucose-specific enzyme IICBA component No

Pentose phosphate pathway

zwf 1.44 Glucose-6-phosphate 1-dehydrogenase No
gntZ 0.29 6-Phosphogluconate dehydrogenase No
yqjI 1.11 Similar to 6-phosphogluconate dehydrogenase No
tkt 0.89 Transketolase 17delA
rpe 0.90 Ribulose-5-phosphate 3-epimerase Val4Asp
ywlF 0.90 Ribose-5-phosphate isomerase B No
prs 1.00 Phosphoribosylpyrophosphate synthetase No

Nitrogen metabolism
yweB (rocG) 4.82 Glutamate dehydrogenase No
glnA 0.81 Glutamine synthetase Glu65Lys
gltC 0.82 Transcriptional activator of the glutamate synthase operon No
gltA 0.46 Glutamate synthase (large subunit) (EC: 1.4.1.13) No
gltB 0.37 Glutamate synthase (small subunit) (EC: 1.4.1.13) No
nasF 1.62 Uroporphyrin-III C-methyltransferase No
nasE 1.59 Assimilatory nitrite reductase (subunit) No
nasD 1.34 Assimilatory nitrite reductase (subunit) No
nasC 1.31 Assimilatory nitrate reductase (catalytic subunit) No
nasB 1.81 Assimilatory nitrate reductase (electron transfer subunit) No
narG 6.48 Nitrate reductase (alpha subunit) No
narH 13.59 Nitrate reductase (beta subunit) No
narJ 16.35 Nitrate reductase (protein J) No
narI 19.27 Nitrate reductase (gamma subunit) No

Glycine biosynthesis
serA 0.37 Phosphoglycerate dehydrogenase No
serC 0.71 Phosphoserine aminotransferase No
glyA 0.68 Serine hydroxymethyltransferase No
folD 0.6 Methylenetetrahydrofolate dehydrogenase/methenyltetrahydrofolate cyclohydrolase No

Purine biosynthesis pathway and other PurR-regulated genes

purE 0.47 Phosphoribosylaminoimidazole carboxylase I (EC: 4.1.1.21) No
purK 0.48 Phosphoribosylaminoimidazole carboxylase II (EC: 4.1.1.21) No
purB 0.28 Adenylosuccinate lyase (EC: 4.3.2.2) No
purC 0.21 Phosphoribosylaminoimidazole succinocarboxamide synthetase (EC: 6.3.2.6) No
yexA (purS) 0.23 Required for phosphoribosylformylglycinamidine synthetase activity No
purL 0.24 Phosphoribosylformylglycinamidine synthetase II (EC: 6.3.5.3) No
purQ 0.24 Phosphoribosylformylglycinamidine synthetase I (EC: 6.3.5.3) No
purF 0.38 Glutamine phosphoribosylpyrophosphate amidotransferase (EC: 2.4.2.14) No
purM 0.30 Phosphoribosylaminoimidazole synthetase (EC: 6.3.3.1) No
purN 0.37 Phosphoribosylglycinamide formyltransferase (EC: 2.1.2.2) No
purH 0.50 Phosphoribosylaminoimidazole carboxy formyl formyltransferase/inosine-monophosphate cyclohydrolase No
purD 0.50 Phosphoribosylglycinamide synthetase No
purR 0.98 Transcriptional repressor of the purine operon No
purA 0.55 Adenylosuccinate synthetase No
guaB 0.88 Inosine-monophosphate dehydrogenase No
guaC (yumD) 0.23 GMP reductase Leu305Ile
yebB (pbuG) 0.42 Hypoxanthine/guanine permease No
Xpt 0.36 Xanthine phosphoribosyltransferase No
pbuX 0.3 Xanthine permease No
yqhZ 0.74 Probable transcription termination No
ytiP (pbuO) 0.24 Unknown No

Riboflavin biosynthesis and transport

ribA 13.27 GTP cyclohydrolase II and 3,4-dihydroxy-2-butanone 4-phosphate synthase (EC: 3.5.4.25) No
ribB 13.19 Riboflavin synthase alpha chain [EC: 2.5.1.9] No
ribC 3.26 Riboflavin kinase and FAD synthase (EC: 2.7.1.26 and EC: 2.7.7.2) Gly199Asp
ribG 21.74 Diaminohydroxyphosphoribosylaminopyrimidine deaminase/5-amino-6-(5-phosphoribosylamino)uracil No
reductase [EC: 3.5.4.26 1.1.1.193]
ribH 10.62 Riboflavin synthase (beta subunit) (EC: 2.5.1.9) No
ribT (ribD) 7.13 Reductase No
ypaA 4.70 A possible transporter of riboflavin No
 ARTICLE IN PRESS
248 S. Shi et al. / Metabolic Engineering 11 (2009) 243–252
gene), guaC (GMP reductase), ldh (L-lactate dehydrogenase), glnA
(the glutamine synthetase gene), and rpe (the ribulose-5-
phosphate 3-epimerase gene). The mutations in the ribC and tkt
were considered as beneficial mutations for riboflavin
biosynthesis as discussed below, while causal links to riboflavin
production were not obvious for the other mutations except the
synonymous mutation in gene ldh, Gly(GGG)-29-Gly(GGA).
3.3. Transcriptional regulation and nucleotide mutation contributed
to the riboflavin overproducing traits in B. subtilis RH33
First, in the riboflavin biosynthesis metabolism, the transcrip-
tion of the rib genes, which are organized in an operon and encode
the riboflavin biosynthesis pathway in B. subtilis, were strongly
up-regulated in RH33 (Table 3; Fig. 1). The expression of the rib
operon genes were significantly up-regulated by 13.2 (ribA), 13.2
(ribB), 10.6 (ribH), and 21.7-fold (ribG), reflecting their over-
expression and direct contribution to riboflavin overproducing
phenotype. Since rib operon genes are directly involved in
riboflavin biosynthesis (Perkins et al., 1999), it is likely that the
up-regulation of rib operon genes positively contributed to the
riboflavin production.
In B. subtilis, the rib operon has an untranslated regulatory
leader region of 300 base pairs in front of its first gene, ribG. The
leader sequence, a conserved regulatory element called RFN
element, is responsible for negative regulation of the operon by
FMN binding, which prevents the expression of the downstream
genes by transcriptional or translational mechanisms (Nudler and
Mironov, 2004). Revealed by DNA sequencing, there are no
mutations in the RFN element in RH33. The up-regulation of the
bifunctional flavokinase/FAD synthase gene ribC (3.3-fold), which
converts riboflavin to FMN and FAD, is not beneficial for high
riboflavin accumulation. However, DNA sequencing indicated that
ribC of RH33 had a Gly-199-Asp mutation, which was coincident
with the report of Bresler et al. (1973) and could lead up to 95%
drop in enzyme activity. Thus, the up-regulation of rib operon in
RH33 may be partly attributed to diminished feedback inhibition
of FMN resulting from the low flavokinase activity of the mutated
ribC. In addition to the riboflavin biosynthesis genes, the RFN
element was also observed in the upstream of ypaA (Nudler and
Mironov, 2004) which was predicted to encode a transporter of
riboflavin. In RH33, ypaA gene was up-regulated by 4.7-fold. Based
on these data, it was likely that the FMN concentration was
maintained at a low level, reflected from the highly up-regulation
of rib operon, and ypaA gene that all have the RFN element (Lee
et al., 2001; Vogl et al., 2007). The mutation in ribC, which leads
up to 95% drop in enzyme activity (Bresler et al., 1973), may
directly contribute to the low FMN pool.
Second, transcriptome profiling revealed that the mRNA
expression level of the genes involved in the central carbon
metabolism had small variations between wild-type strain 168
and the riboflavin overproducing strain RH33 (Table 3; Fig. 1).
These genes may be highly expressed in both strains, as
qualitatively reflected from their high absolute fluorescence
intensities in one-color DNA microarray hybridization. Though
riboflavin was over-produced, few genes involved in central
metabolism showed differential expression in strain RH33. It
may indicate that precursor supply due to the carbon flux through
central metabolism was sufficient in RH33 under riboflavin
overproduction condition.
Alteration of carbon flow into the pentose phosphate pathway
will affect the riboflavin production because the pentose phos-
phate pathway is a major source for the supply of ribose
(or ribulose-5-phosphate), which is the starting point of riboflavin
biosynthesis (Sauer et al., 1996; Zhu et al., 2006). As shown in Fig.
1, glucose-6-phosphate enters the pentose phosphate pathway
after being oxidated to gluconate-6-phosphate by glucose-6-
phosphate dehydrogenase (zwf) and then is subsequently oxidized
to ribulose-5-phosphate by gluconate-6-phosphate dehydrogen-
ase. It was found that the transcription of gntZ, which was
considered encoding the major 6-P-gluconate dehydrogenase, was
significantly down-regulate (0.29-fold) in B. subtilis RH33. How-
ever, according to a recent report (Zamboni et al., 2004b), its
isoenzyme gene yqjI was shown to have the major 6-P-gluconate
dehydrogenase activity in B. subtilis, which was moderately up-
regulated in RH33 (1.1-fold) to compensate the enzyme activity
and carbon flux. It should be noted that the expression of the
transketolase gene tkt did not alter significantly in magnitude
(0.89-fold). However, there was a frame-shift mutation in its
coding region. It had been shown that the B. subtilis RH33 grew
very slowly in minimal medium supplement with xylose or
arabinose as the sole carbon source (unpublished data), which
may be related to a dramatic lose of the transketolase activity.
Indeed, B. subtilis RH33 showed only 4% transketolase activity
compared with that of B. subtilis 168 (177 nmol min!1 mg!1
protein). The low activity of transketolase may decrease the shunt
of pentose back into glycolysis, allowing cells to redirect more
carbon flux through the oxidative pentose pathway to the purine
nucleotide biosynthesis pathway, which provided the precursors
of riboflavin production (Kamada et al., 2001). In short, tkt
contributed to riboflavin overproduction trait of RH33, not at the
level of transcriptional regulation, but through frame-shift
mutation to reduce enzyme activity of gene tkt.
3.4. Identification of metabolic bottleneck for increasing riboflavin
production in B. subtilis RH33
Purine nucleotides, involving in biosynthesis of riboflavin by
providing the immediate precursor GTP, are synthesized by purine
genes (purEKBCSQLFMNHD). Unexpectedly, the purine genes
together with glycine biosynthesis genes (serA, serC, glyA, folD)
(Saxild et al., 2001) were all greatly down-regulated in RH33
(Table 3; Fig. 1). The down-regulation of purF, purD, glyA, and folD
were also revealed by quantitative RT-PCR experiment (Table 2).
Thus, the down-regulated purine and glycine (another building
block for riboflavin) biosynthesis genes would limit the supply of
the precursors for riboflavin overproduction (Heefner et al., 1988;
Jimenez et al., 2005; Mateos et al., 2006; Schlupen et al., 2003;
Stahmann et al., 2000).
Nonetheless, it was interesting to note that the down-
regulated purine, glycine, and other genes involved in purine
metabolism (guaC, pbuG, xpt-pbuX, yqh, and pbuO) were all
negatively regulated by global regulator PurR in B. subtilis (Saxild
et al., 2001). The nucleotide sequencing results indicated that
there were no mutations in the regulatory region (PRPP binding
motif) and coding region of the purR gene. Among all the PurR-
regulated genes only guaC had a point mutation (Table 3). It was
reported that the defective GMP reductase (guaC) that reduces
conversion of GMP to IMP had been identified in a B. subtilis
riboflavin overproducer (Nygaard, 1993). However, the mutation
of guaC in B. subtilis RH33 was neutral (from Leu to Ile) and
probably had no effect on protein function.
Since no mutations were found in the regulatory region and
coding region of the PurR-regulated genes (including purR) except
guaC, the down-regulation of those genes may be due to the
repression by PurR. It binds to the operator sites (PurBox) of its
target genes and inhibits their transcription. According to the
current model, the PRPP can bind PurR to prevent its binding to
DNA and lead to derepression of PurR-regulated genes (Weng
et al., 1995). Therefore, it can be deduced that the low expression
 ARTICLE IN PRESS
S. Shi et al. / Metabolic Engineering 11 (2009) 243–252 249

of PurR-regulated genes might be caused by a small cellular PRPP
pool. Subsequently, metabolite analysis was carried out and tested
this hypothesis (Table 4). The PRPP pool in B. subtilis RH33 was
only 24% that of B. subtilis 168 measured at the conditions as the
microarray experiments. In addition, PRPP itself is an important
metabolite precursor for purine biosynthesis as it is required in
the de novo and salvage pathways of purine biosynthesis (Jimenez
et al., 2008). It was proposed that enhancing PRPP pool would
derepress the expression of pur operon and glycine biosynthesis
genes, thus increasing the precursors supply and leading to
increased production of riboflavin in B. subtilis RH33. At present,
there is only one report employing the strategy of modulating the
PRPP pool for riboflavin production (Jimenez et al., 2008) in a
hypothesis-driven approach.
The selected prs gene was overexpressed in B. subtilis RH33
under the control of the P43 promoter to construct the strain
B. subtilis RH33-Prs. The enzyme activities of PRPP synthetase
were 3.070.7 and 10.172.0 nmol min!1 mg!1 protein for B.
subtilis RH33 and B. subtilis RH33-Prs, respectively, indicating that
PRPP synthetase was overexpressed successfully.
To assess the effects of the manipulation on the physiological
parameters, batch cultivations were carried out in minimal
medium. The results displayed only minor variability in the
specific cell growth rate, specific glucose uptake rate, specific
organic acids secretion rate, and specific riboflavin production rate
between B. subtilis RH33 and B. subtilis RH33-Prs (Table 6).
Unexpectedly, the PRPP concentration was almost constant in the
mutant (Table 4). These indicated that overexpression of prs gene
alone had little effect on the supply of PRPP and on specific
riboflavin production rate.

3.5. Effect of prs overexpression on riboflavin production

PRPP is an important metabolite required for purine, pyrimi-
dine, tryptophan, and histidine biosynthesis. As discussed above,
the low concentration of PRPP may be the bottleneck for further
increasing riboflavin production in B. subtilis RH33. The formation
of PRPP is catalyzed by the enzyme PRPP synthetase, which is
encoded by prs gene. The expression level of prs was not notably
altered. In addition, the enzyme activity assay showed that the
enzyme activity of RPPP synthetase was almost the same between
B. subtilis RH33 and B. subtilis 168 (Table 5). Therefore, it was
reasoned that amplification of prs gene was expected to increase
the supply of PRPP and consequently result in increased riboflavin
production.
3.6. Enhancement of riboflavin production by co-expression of prs
and ywlF genes
As the strain B. subtilis RH33-Prs did not show the expected
traits in improving PRPP supply, to diagnose the problem(s),
further measurements of precursor concentrations of the meta-
bolites were carried out to elucidate the limitations. Compared to
RH33, ribulose-5-phosphate had similar levels but ribose-
5-phosphate had significantly lower concentrations in RH33-
Prs (Table 4). The insufficient ribose-5-phosphate pool was
identified as the limiting factor for the indistinctive effect on
increasing RPPP concentration and riboflavin production in B.
subtilis RH33-Prs.

Table 4
Intracellular metabolite concentration in cell extracts of B. subtilis.

Metabolite [nmol mg!1 (dry wt)] 168 RH33 RH33-Prs RH33-PY

PRPP 2.5070.36a 0.8870.31a 0.9170.32a 4.1470.60a

3.6270.45b 0.8770.38b 0.9870.39b 5.7770.65b
Ribose-5-P 1.1170.29a 1.2870.33a 0.5570.25a 1.9870.40a
Ribulose-5-P 4.3071.54a 4.3671.52a 4.1871.46a 4.2071.52a

a
Sampled during exponential growth period cultivated in minimal medium.
b
Sampled during exponential growth period cultivated in LBG medium.

Table 5
Specific enzyme activities in cell extracts of B. subtilis, sampled during exponential growth period.

Enzyme activities (nmol min!1 mg!1 protein) 168 RH33 RH33-Prs RH33-PY

PRPP synthetase 4.170.8 3.070.7 10.172.0 10.772.0

Ribose-5-phosphate isomerase B 28.772.5 22.372.5 27.472.5 61.475.5

Table 6
Metabolic characterization of B. subtilis strains during exponential growth period (OD600 0.3–0.6) cultivated in minimal medium.

Parameter 168 RH33 RH33-Prs RH33-PY

!1 !1
Specific glucose uptake rate (mmol g CDW h ) 4.570.4 4.270.4 4.170.4 4.270.4
Specific growth rate (h!1) 0.3170.03 0.2770.02 0.2870.03 0.2770.03
Specific organic acidsa secretion rate (mmol g!1 CDW h!1)
1.6370.31 1.5170.28 1.4770.22 1.3070.22
Specific riboflavin production rate (mmol g!1 CDW h!1) NDb 0.04570.002 0.04770.003 0.05370.003

Results represent the mean values with standard deviations from three independent measurements.
a
Organic acid including acetate, lactate, and pyruvate.
b
Not detected.
 ARTICLE IN PRESS
250 S. Shi et al. / Metabolic Engineering 11 (2009) 243–252

As shown in Fig. 1, ribose-5-phosphate can be generated from 10

ribulose-5-phosphate catalyzed by ywlF encoding ribose-5-phos- 9
phate isomerase B. We decided to co-overexpress the prs and 8
ywlF genes simultaneously. In this study, the plasmid containing
7
both prs gene and ywlF gene under the control of the P43

Glucose (g/L)
promoter was introduced into the prs locus of the B. subtilis 6
RH33 chromosome to construct B. subtilis RH33-PY strain. Both 5
PRPP synthetase and ribose-5-phosphate isomerase B enzyme 4
activities were significantly higher than strain RH33, as shown in 3
Table 5, indicating the successful overexpression of prs and ywlF
2
genes.
When both prs and ywlF were overexpressed, the specific cell 1
0
growth rate and glucose uptake rate were almost the same 10
compared with RH33 (Table 6). Meanwhile, RH33-PY had a 9
slightly lower organic acid secretion. The productivity of riboflavin 8
was significantly increased from 0.04570.002 to 0.0537
7
0.003 mmol g!1 CDW h!1. The concentration of intermediate me-

Biomass (0D600)
tabolites PRPP and ribose-5-phosphate were much higher in B. 6
subtilis RH33-PY compared to RH33 strain (Table 4), which 5
implied that the higher yield of riboflavin was attributable to 4
the sufficient supply of PRPP. Although the overexpression of 3
ribose-5-phosphate isomerase B caused more carbon flux to
2
redirect from ribulose-5-phosphate to ribose-5-phosphate, the
concentrations of ribulose-5-phosphate in these strains were 1
almost constant, which indicated the abundant precursor supply 0
provided by central carbon metabolism in B. subtilis RH33 as 250
Riboflavin Concentration (mg/L)

mentioned above. 200

PRPP has dual roles in riboflavin production: one is used as a
precursor, and the other is to bind PurR and depress PurR-
150
regulated genes. To confirm that indeed the increased PRPP pool
resulted in up-regulation of PurR controlled genes in RH33-PY
strain, a quantitative RT-PCR experiment was conducted to 100
measure the transcript abundances of PurR-regulated genes
relative to parental strain RH33. The fold change of the selected 50
PurR-regulated genes, including purE, purD, guaC, purA, glyA, folD,
pbuG, xpt, pbuX, ytiP, and yqhZ, were shown in Fig. 2. As expected, 0
expression of those PurR-regulated genes, including the purine 0 2 4 6 8 10
pathway genes (purE, purD) and the glycine biosynthesis pathway Time (h)
genes (glyA, folD), were induced in RH33-PY (Fig. 2), which was in
Fig. 3. Time-course profiles of biomass, glucose, and riboflavin concentration of B.
accordance with our hypothesis. In B. subtilis RH33-PY, it is
subtilis RH33 (closed triangles), RH33-Prs (closed squares) and RH33-PY (closed
suggested that the derepressed PurR-regulated genes activated circles) in LBG medium during shake flask batch cultivations.
the pathways that supplying precursors for riboflavin
biosynthesis, which in turn led to the increase of riboflavin
Fig. 3 shows the time-course profiles of riboflavin production,
production.
biomass, and glucose concentration exhibited by RH33, RH33-
Prs, and RH33-PY grown in shake flask with LBG medium. The
7 cell growth and glucose uptake profiles of RH33-Prs and RH33-
PY were almost the same as the parental strain RH33. Little effect
on riboflavin production was observed in RH33-Prs compared
Relative gene expression (RH33-RY/RH33)

6
with RH33. However, the RH33-PY mutant exhibited a riboflavin
titer of 3273% higher than RH33 (Fig. 3). Thus, it indicated
5
that the metabolic engineering strategy identified by trans-
criptome analysis successfully improved the trait of riboflavin
4 overproduction in RH33 on a shake flask scale.

3
3.7. Riboflavin production in fed-batch fermentation
2
To investigate the potential of the strain B. subtilis RH33-PY in a
larger scale fermentor, we tested the performance of strain B.
1
subtilis RH33-PY in fed-batch fermentation using the feeding
profile as described in Materials and methods. The cell growth
0
rates were also very similar among the parental strain RH33 and
pur E pur D pur A gly A fol D guaC pbu G xpt pbuX ytiP yqhZ
two engineered strains RH33-Prs and RH33-PY (Fig. 4A).
Genes Compared to parental strain RH33, an average of 25% increase in
Fig. 2. The fold change of selected PurR-regulated transcripts in B. subtilis RH33- riboflavin titer, up to 15 g l!1, was achieved in RH33-PY (Fig. 4B),
PY, relative to RH33. Values represent mean and s.d. (n ¼ 3). while the improvement was only about 6% in strain RH33-Prs
 ARTICLE IN PRESS
S. Shi et al. / Metabolic Engineering 11 (2009) 243–252 251

analysis. By co-overexpressing prs and ywlF in RH33, PRPP

140 pool level was significantly increased. This had dual effects:
one was to derepress the purine genes negatively regulated
by PurR (Fig. 2), and the other was to provide more precursors
120
for purine biosynthesis pathways. The strategy developed in
this work allowed for a 25% increase in riboflavin titer, up to
100 15 g l!1, compared to the parental strain during a 5-l fed-batch
Biomass (OD600)

fermentation.
80 An increasing number of publications have demonstrated how
transcriptome analysis has been successfully applied to under-
60 stand and engineering metabolism of the organisms. This trend
will be expected to continue in the post-genomic era. Unfortu-
40 nately, elucidating and application of the mechanism(s) under-
lying the improved performance of the strains developed by
20 random mutagenesis and selection are still limited, partly due to
the complexities of metabolic and regulatory networks encoded in
0 the even simplest bacterial genome. The whole-genome gene
0 6 12 18 24 30 36 42 48 expression analysis, integrated with nucleotide sequencing and
Feed time (h) metabolite measurements, represents a powerful approach to
exploring and exploiting of genomic data, thus greatly facilitating
the identification of novel targets for strain improvement. This
18 systems strategy should be in principle applied for rational
microbial development by deducing the global physiology and
16 regulation of an overproducer strain that is obtained by
Riboflavin concentration (g/L)

14 combinatorial approaches.

12
10 Acknowledgments

8 This research was supported by the National Natural Science

6 Foundation of China (NSFC-20536040), the National Project of Key
4 of Science and Technology of Tianjin (05YFGZGX04500) and
2 Programme of Introducing Talents of Discipline to Universities
0 help with the microarray experiment. The microarray data
0 6 12 18 24 30 36
Feed time (h)
Fig. 4. Time-course profiles of biomass (A) and riboflavin concentration (B) of B.
subtilis RH33 (closed triangles), RH33-Prs (closed squares) and RH33-PY (closed
circles) during fed-batch cultivations in a 5-l bioreactor.
!1
(12.7 g l ). Therefore, in fed-batch fermentation, modulating
PRPP pool by co-overexpressing prs and ywlF successfully
improved riboflavin production in B. subtilis while causing little
effect on cell growth. These data also suggested that increasing
PRPP pool may be a general strategy to improve riboflavin
production. The alterations of riboflavin production in the three
strains have the same trend in the medium tested. Another
successful example for enhancing riboflavin production by
manipulation of the PRPP pool was achieved in Ashbya gossypii
(Jimenez et al., 2008).
In this study, based on known regulatory and metabolic
information and new insights generated from transcriptome
analysis, combined with nucleotide sequencing and intracellular
metabolite concentration measurement, metabolic bottleneck and
novel target genes were identified and co-amplified to further
enhance production of riboflavin in B. subtilis. Microarray analysis
identified that purine and glycine biosynthesis genes were
significantly down-regulated in the riboflavin-producing strain
RH-33. These genes were repressed by the global regulator PurR
whose activity was modulated by small molecular PRPP, which is
also the precursor for purine synthesis. It was inferred that the
reduced expression of PurR-regulated genes might be caused by a
low PRPP pool, which was subsequently confirmed by metabolite
Fundamental Research (2007CB707802), the Development Project
(B06006). CapitalBio Co. Ltd., in Beijing was acknowledged for
42 48 described in this publication had been deposited in the public
database (Gene Expression Omnibus, GEO) with accession
number of GSE12873.
Appendix A. Supplementary material
Supplementary data associated with this article can be found
in the online version at doi:10.1016/j.ymben.2009.05.002.
References
Anagnostopoulos, C., Spizizen, J., 1961. Requirements for transformation in Bacillus
subtilis. J. Bacteriol. 81, 741–746.
Bai, D.M., Zhao, X.M., Li, X.G., Xu, S.M., 2004. Strain improvement of Rhizopus oryzae
for over-production of L(+)-lactic acid and metabolic flux analysis of mutants.
Biochem. Eng. J. 18, 41–48.
Bailey, J.E., 1991. Toward a science of metabolic engineering. Science 252,
1668–1675.
Bresler, S.E., Glazunov, E.A., Chernik, T.P., Schevchenko, T.N., Perumov, D.A., 1973.
Study of riboflavin biosynthesis operon in Bacillus subtilis. Flavinmononucleo-
tide and flavinadeninedinucleotide as effectors of the riboflavin operon.
Genetika 9, 84–92.
Choi, J.H., Lee, S.J., Lee, S.J., Lee, S.Y., 2003. Enhanced production of insulin-like
growth factor I fusion protein in Escherichia coli by coexpression of the down-
regulated genes identified by transcriptome profiling. Appl. Environ. Microbiol.
69, 4737–4742.
Crueger, W., Crueger, A., 1984. Biotechnology: A Textbook of Industrial Micro-
biology. Sinauer Associates, Sunderland, MA.
Fisher, S.H., Mangasanik, B., 1984. Synthesis of oxaloacetate in Bacillus subtilis
mutants lacking the 2-ketoglutarate dehydrogenase enzymatic complex. J.
Bacteriol. 158, 55–62.
 ARTICLE IN PRESS
252 S. Shi et al. / Metabolic Engineering 11 (2009) 243–252
Gall, S., Lynch, M.D., Sandoval, N.R., Gill, R.T., 2008. Parallel mapping of genotypes
to phenotypes contributing to overall biological fitness. Metab. Eng. 10,
382–393.
Harris, D.M., Westerlaken, I., Schipper, D., van der Krogt, Z.A., Gombert, A.K.,
Sutherland, J., Raamsdonk, L.M., van den Berg, M.A., Bovenberg, R.A.L., Pronk,
J.T., Daran, J.M., 2009. Engineering of Penicillium chrysogenum for fermentative
production of a novel carbamoylated cephem antibiotic precursor. Metab. Eng.
11, 125–137.
Heefner, D.L., Weaver, C.A., Yarus, M.J., Burdzinski, L.A., Gyure, D.C., Foster, E.W.,
1988. Riboflavin producing strains of microorganisms, method for selecting,
and method for fermentation. Patent WO 88/09822.
Hoque, M.A., Ushiyama, H., Tomita, M., Shimizu, K., 2005. Dynamic responses of
the intracellular metabolite concentrations of the wild type and pykA mutant
Escherichia coli against pulse addition of glucose or NH3 under those limiting
continuous cultures. Biochem. Eng. J. 26, 38–49.
Hove-Jensen, B., Maigaard, M., 1993. Escherichia coli rpiA gene encoding ribose
phosphate isomerase A. J. Bacteriol. 175, 5628–5635.
Humbelin,
+ M., Griesser, V., Keller, T., Schurter, W., Haiker, M., Hohmann, H.P., Ritz,
H., Richter, G., Bacher, A., Van Loon, A.P.G.M., 1999. GTP cyclohydrolase II and
3,4-dihydroxy-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.
Iida, A., Teshiba, S., Mizobuchi, K., 1993. Identification and characterization of the
tktB gene encoding a second transketolase in Escherichia coli K-12. J. Bacteriol.
175, 5375–5383.
Izallalen, M., Mahadevan, R., Burgard, A., Postier, B., Didonato, R., Sun, J., Schilling,
C.H., Lovley, D.R., 2008. Geobacter sulfurreducens strain engineered for
increased rates of respiration. Metab. Eng. 10, 267–275.
Jaluria, P., Betenbaugh, M., Konstantopoulos, K., Frank, B., Shiloach, J., 2007.
Application of microarrays to identify and characterize genes involved in
attachment dependence in HeLa cells. Metab. Eng. 9, 241–251.
Jimenez, A., Santos, M.A., Pompejus, M., Revuelta1, J.L., 2005. Metabolic engineer-
ing of the purine pathway for riboflavin production in Ashbya gossypii. Appl.
Environ. Microbiol. 71, 5743–5751.
Jimenez, A., Santos, M.A., Revuelta, J.L., 2008. Phosphoribosyl pyrophosphate
synthetase activity affects growth and riboflavin production in Ashbya gossypii.
BMC Biotechnol. 8, 67–78.
Kamada, N., Yasuhara, A., Takano, Y., Nakano, T., Ikeda, M., 2001. Effect of
transketolase modifications on carbon flow to the purine-nucleotide pathway
in Corynebacterium ammoniagenes. Appl. Microbiol. Biotechnol. 56, 710–717.
Knorr, B., Schlieker, H., Hohmann, H.P., Weuster-Botz, D., 2007. Scale-down and
parallel operation of the riboflavin production process with Bacillus subtilis.
Biochem. Eng. J. 33, 263–274.
Kornberg, A., Lieberman, I., Simms, E.S., 1955. Enzymatic synthesis and properties
of 5-phosphoribosylpyrophosphate. J. Biol. Chem. 215, 389–402.
Kunst, F., Ogasawara, N., Moszer, I., Albertini, A.M., Alloni, G., Azevedo, V., Bertero,
M.G., Bessieres, P., Bolotin, A., Borchert, S., et al., 1997. The complete genome
sequence of the Gram-positive bacterium Bacillus subtilis. Nature 390, 249–256.
Lee, J.M., Zhang, S., Saha, S., Anna, S.S., Jiang, C., Perkins, J., 2001. RNA expression
analysis using an antisense Bacillus subtilis genome array. J. Bacteriol. 183,
7371–7380.
Livak, K.J., Schmittgen, T.D., 2001. Analysis of relative gene expression data using real-
time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25, 402–408.
Mateos, L., Jimenez, A., Revuelta, J.L., Santos, M.A., 2006. Purine biosynthesis,
riboflavin production, and trophic-phase span are controlled by a myb-related
transcription factor in the fungus Ashbya gossypii. Appl. Environ. Microbiol. 72,
5052–5060.
Nielsen, J., Olsson, L., 2002. An expanded role for microbial physiology in metabolic
engineering and functional genomics: moving towards systems biology. FEMS
Yeast Res. 2, 175–181.
Nudler, E., Mironov, A.S., 2004. The riboswitch control of bacterial metabolism.
Trends Biochem. Sci. 29, 11–17.
Nygaard, P., 1993. Purine and pyrimidine salvage pathways. In: Sonenshein, A.L.,
Hoch, J.A., Losick, R. (Eds.), Bacillus subtilis and Other Gram-Positive Bacteria:
Biochemistry, Physiology, and Molecular Genetics. American Society for
Microbiology, Washington, DC, pp. 359–378.
Park, J.H., Lee, K.H., Kim, T.Y., Lee, S.Y., 2007. Metabolic engineering of Escherichia
coli for the production of L-valine based on transcriptome analysis and in silico
gene knockout simulation. Proc. Natl. Acad. Sci. 104, 7797–7802.
Peberdy, J.F., 1985. Biology of penicillins. In: Demain, A.L., Solomon, N.A. (Eds.),
Biology of Industrial Microorganisms. Benjamin-Cummings, Menlo Park, pp.
407–431.
Peebles, C.A.M., Hughes, E.H., Shanks, J.V., San, K.Y., 2009. Transcriptional response
of the terpenoid indole alkaloid pathway to the overexpression of ORCA3 along
with jasmonic acid elicitation of Catharanthus roseus hairy roots over time.
Metab. Eng. 11, 76–86.
Perkins, J.B., Pero, J.G., Sloma, A., 1991. Riboflavin overproducing strains of bacteria.
European Patent 0405370.
Perkins, J.B., Sloma, A., Pero, J.G., Hatch, R.T., Hermann, T., Erdenberger, T., 1999.
Bacterial strains which overproduce riboflavin. US Patent 5925538.
Sakuma, R., Nishina, T., Yamanaka, H., Kamatani, N., Nishioka, K., Maeda, M., Tsuji,
A., 1991. Phosphoribosylpyrophosphate synthetase in human erythrocytes:
assay and kinetic studies using high-performance liquid chromatography. Clin.
Chim. Acta 203, 143–152.
Sambrook, J., Russell, D.W., 2001. Molecular Cloning: A Laboratory Manual, third
ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Sauer, U., Hatzimanikatis, V., Hohmann, H.P., Manneberg, M., Van Loon, A.P.G.M.,
Bailey, J.E., 1996. Physiology and metabolic fluxes of the wild-type
and riboflavin-producing Bacillus subtilis. Appl. Environ. Microbiol. 62,
3687–3696.
Saxild, H.H., Brunstedt, K., Nielsen, K.I., Jarmer, H., Nygaard, P., 2001. Definition of
the Bacillus subtilis PurR Operator Using Genetic and Bioinformatic Tools and
Expansion of the PurR Regulon with glyA, guaC, pbuG, xpt-pbuX, yqhZ-folD, and
pbuO. J. Bacteriol. 183, 6175–6183.
Schlupen, C., Santos, M.A., Weber, U., de Graaf, A., Revuelta, J.L., Stahmann, K.P.,
2003. Disruption of the SHM2 gene, encoding one of two serine hydro-
xymethyltransferase isoenzymes, reduces the flux from glycine to serine in
Ashbya gossypii. Biochem. J. 369, 263–273.
Sindelar, G., Wendisch, V.F., 2007. Improving lysine production by Corynebacterium
glutamicum through DNA microarray-based identification of novel target
genes. Appl. Microbiol. Biotechnol. 76, 677–689.
Stahmann, K.P., Revuelta, J.L., Seulberger, H., 2000. Three biotechnical processes
using Ashbya gossypii, Candida famata, or Bacillus subtilis compete with
chemical riboflavin production. Appl. Microbiol. Biotechnol. 53, 509–516.
Stephanopoulos, G., Aristidou, A.A., Nielsen, J., 1998. Metabolic Engineering:
Principles and Methodologies. Academic Press, San Diego.
Tännler, S., Zamboni, N., Kiraly, C., Aymerich, S., Sauer, U., 2008. Screening of
Bacillus subtilis transposon mutants with altered riboflavin production. Metab.
Eng. 10, 216–226.
Vogl, C., Grill, S., Schilling, O., Stulke, J., Mack, M., Stolz, J., 2007. Characterization of
riboflavin (vitamin B2) transport proteins from Bacillus subtilis and Coryne-
bacterium glutamicum. J. Bacteriol. 189, 7367–7375.
Wang, P.Z., Doi, R.H., 1984. Overlapping promoters transcribed by Bacillus subtilis
sigma 55 and sigma 37 RNA polymerase holoenzymes during growth and
stationary phases. J. Biol. Chem. 259, 8619–8625.
Weng, M.L., Nagy, P.L., Zalkin, H., 1995. Identification of the Bacillus subtilis pur
operon repressor. Proc. Natl. Acad. Sci. 92, 7455–7459.
Wierckx, N.J.P., Ballerstedt, H., de Bont, J.A.M., de Winde, J.H., Ruijssenaars, H.J.,
Wery, J., 2008. Transcriptome analysis of a phenol-producing Pseudomonas
putida S12 construct: genetic and physiological basis for improved production.
J. Bacteriol. 190, 2822–2830.
Wu, Q.L., Chen, T., Gan, Y., Chen, X., Zhao, X.M., 2007. Optimization of riboflavin
production by recombinant Bacillus subtilis RH44 using statistical designs.
Appl. Microbiol. Biotechnol. 76, 783–794.
Zamboni, N., Mouncey, N., Hohmann, H.P., Sauer, U., 2003. Reducing maintenance
metabolism by metabolic engineering of respiration improves riboflavin
production by Bacillus subtilis. Metab. Eng. 5, 49–55.
Zamboni, N., Maaheimo, H., Szyperski, T., Hohmann, H.P., Sauer, U., 2004a. The
phosphoenolpyruvate carboxykinase also catalyzes C3 carboxylation at the
interface of glycolysis and the TCA cycle of Bacillus subtilis. Metab. Eng. 6,
277–284.
Zamboni, N., Fischer, E., Laudert, D., Aymerich, S., Hohmann, H.P., Sauer, U., 2004b.
The Bacillus subtilis yqjI gene encodes the NADP(+)-dependent 6-P-gluconate
dehydrogenase in the pentose phosphate pathway. J. Bacteriol. 186,
4528–4534.
Zhu, Y.B., Chen, X., Chen, T., Shi, S.B., Zhao, X.M., 2006. Over-expression of glucose
dehydrogenase improves cell growth and riboflavin production in Bacillus
subtilis. Biotechnol. Lett. 28, 1667–1672.