Expanding metabolism for biosynthesis of nonnatural alcohols
Nature uses a limited set of metabolites to perform all of the biochemical reactions. To increase the metabolic capabilities of biological systems, we have expanded the natural metabolic network, using a nonnatural metabolic engineering approach. The branched-chain amino acid pathways are extended to produce abiotic longer chain keto acids and alcohols by engineering the chain elongation activity of 2-isopropylmalate synthase and altering the substrate specificity of downstream enzymes through rational protein design. When introduced into Escherichia coli, this nonnatural biosynthetic pathway produces various long-chain alcohols with carbon number ranging from 5 to 8. In particular, we demonstrate the feasibility of this approach by optimizing the biosynthesis of the 6-carbon alcohol, (S)-3-methyl-1-pentanol. This work demonstrates an approach to build artificial metabolism beyond the natural metabolic network. Nonnatural metabolites such as long chain alcohols are now included in the metabolite family of living systems. • metabolic engineering • protein engineering • chain elongation • long chain alcohols
Expanding metabolism for biosynthesis
of nonnatural alcohols
, Michael R. Sawaya
, David S. Eisenberg
, and James C. Liao
Chemical and Biomolecular Engineering and Chemistry and
Institute for Genomics and Proteomics, University of
California, Los Angeles, CA 90095
Edited by Frances H. Arnold, California Institute of Technology, Pasadena, CA, and approved November 4, 2008 (received for review July 24, 2008)
Nature uses a limited set of metabolites to perform all of the
biochemical reactions. To increase the metabolic capabilities of
biological systems, we have expanded the natural metabolic net-
work, using a nonnatural metabolic engineering approach. The
branched-chain amino acid pathways are extended to produce
abiotic longer chain keto acids and alcohols by engineering the
chain elongation activity of 2-isopropylmalate synthase and alter-
ing the substrate speciﬁcity of downstream enzymes through
rational protein design. When introduced into Escherichia coli, this
nonnatural biosynthetic pathway produces various long-chain al-
cohols with carbon number ranging from 5 to 8. In particular, we
demonstrate the feasibility of this approach by optimizing the
biosynthesis of the 6-carbon alcohol, (S)-3-methyl-1-pentanol. This
work demonstrates an approach to build artiﬁcial metabolism
beyond the natural metabolic network. Nonnatural metabolites
such as long chain alcohols are now included in the metabolite family
of living systems.
metabolic engineering 兩 protein engineering 兩 chain elongation 兩
long chain alcohols
ature uses a limited set of metabolites such as organ ic acids,
amino acids, nucleotides, lipids and sugars as building
blocks for biosynthesis. These chemicals support the biological
functions of all organ isms. So far, construction of artificial
biological systems (1–5) is limited by the existing metabolic
capabilities. By supplying living cells with chemically synthesized
nonnatural amino acids (6, 7) and sugars (8) as new building
blocks, it is possible to introduce novel physical and chemical
properties into biological entities. These efforts raise an inter-
esting question: Can we rewire metabolism in a bottom-up
fashion to produce nonnatural metabolites from simple carbon
source? If so, such engineered artificial metabolism should be
able to expand the chemical repertoire that liv ing systems can
use and produce. To begin to address this question, we developed
a strategy to produce 7-(C7) to 9-carbon (C9) 2-keto acids, which
can lead to useful nonnatural alcohols (C6–C8).
A liphatic alc ohols with carbon chain length ⬎5(C⬎ 5) are
attractive biofuel targets because they have higher energy den-
sit y, and lower water solubility [1-pentanol 23 g/L, 1-hexanol 6.2
g/L, 1-heptanol 1.2 g/L (9)] that could facilitate postproduction
purification from culture medium through an aqueous/organic
2-phase separation process. The only well-characterized mech-
an ism for aliphatic alcohol production is through the Ehrlich
pathway (10), which converts branched-chain amino acids into
alc ohols. The carbon number (up to 5) of the alcohols derived
f rom this type of pathway is limited by the carbon number in the
branched chain amino acid pathways (11). To overcome this
limit ation, existing metabolic networks need to be expanded.
This is a daunting task because a metabolic pathway usually
involves the c ollective function of multiple enzymes, which have
to be engineered by rational design (12) or directed evolution
(13, 14) to perform nonnatural activities.
Acet yl-CoA is a common chemical unit for carbon chain
elongation, such as reactions in tricarboxylic acid cycle, glyox y-
late cycle, mevalonate pathway, and leucine biosynthesis (15). To
ex plore the possibilit y of using acetyl-CoA related chemistry to
produce C6 alcohol, we have engineered a nonnatural metabolic
pathway (Fig. 1B, shaded region) into E. coli. First, we used the
existing metabolic capability of E. coli to synthesize (S)-2-keto-
3-methylvalerate, the 2-keto acid precursor of amino acid
isoleucine. The chemical structure of 2-keto-3-methylvalerate is
Author contributions: K.Z., M.R.S., and J.C.L. designed research; K.Z. performed research;
K.Z. and J.C.L. analyzed data; and K.Z., M.R.S., D.S.E., and J.C.L. wrote the paper.
The authors declare no conﬂict of interest.
This article is a PNAS Direct Submission.
To whom correspondence should be addressed at: 5531 Boelter Hall, 420 Westwood Plaza,
Los Angeles, CA 90095. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2008 by The National Academy of Sciences of the USA
Il C Il D
tdcB ilvG ilvM ilvC
Fig. 1. Pathway design. (A) Conversion of 2-keto acids to alcohols by a broad-
substrate range 2-keto-acid decarboxylase (KIVD) and an alcohol dehydrogenase
(ADH6). (B) Schematic representation of the biosynthetic pathway of 3-methyl-
1-pentanol. The engineered nonnatural metabolic pathway is shaded in laven-
der. Similar to 2-ketoisovalerate, 2-keto-3-methylvalerate can have one more
carbon added to its side chain by the leucine biosynthesis enzymes. (C) Synthetic
operons for gene expression. Overexpression of ThrABC, TdcB and IlvGMCD
drives the carbon ﬂux toward 2-keto-3-methylvalerate.
December 30, 2008
very similar to 2-ketoisovalerate (the 2-keto acid precursor of
L-valine), cont aining only one more methyl group on
the side chain. Because 2-ketoisovalerate is converted to 2-ke-
toisocaproate through a 3-step chain elongation cycle by 2-iso-
propylmalate synthase (LeuA), isopropylmalate isomerase c om-
plex (LeuC, LeuD) and 3-isopropylmalate dehydrogenase
(LeuB), we reasoned that LeuA, LeuB, LeuC and LeuD may be
promiscuous enough to allow 2-keto-3-methylvalerate go
through the same elongation cycle and produce a novel com-
pound 2-keto-4-methylhexanoate. A nalogous to the Ehrlich
pathway for production of fusel alc ohols (Fig. 1A), we speculated
that 2-keto-4-methylhexanoate could be converted to the cor-
responding aldehyde and then to a 6-carbon alcohol, (S)-3-
methyl-1-pent anol, by the broad-substrate-range 2-ketoisovaler-
ate decarboxylase (KIVD) from Lactococcus lactis (16) and
alc ohol dehydrogenase VI (ADH6) f rom Saccharomyces
Results and Discussion
Construction of a Nonnatural Metabolic Pathway for Biosynthesis of
)-3-Methyl-1-Pentanol. We constructed 3 synthetic operons (Fig.
1C) under the control of the P
lacO1 promoter: The first operon
is composed of 3 genes on a low copy plasmid in the transcrip-
tional order thrA-thrB-thrC; the second operon is composed of 5
genes on a medium copy plasmid in the transcriptional order
tdcB-ilvG-ilvM-ilvC-ilvD, and the third operon is composed of 6
genes on a high copy plasmid in the transcriptional order
leuA-leuB-leuC-leuD-KIVD-ADH6 (and a control operon with-
out leuABCD). Except for KIVD, ADH6 and ThrA (G433R
mut ant (18) insensitive to threonine feedback inhibition), all
other genes encode wild-type E. coli enz ymes. As a result of
overex pressing these 14 genes in a modified threonine-
hyperproduction strain (ATCC98082, ⌬ilvE, ⌬tyrB), 6.5 mg/L of
(S)-3-methyl-1-pentanol was produced from 20 g/L gluc ose
(Table 1, column 4), whereas a leucine-feedback insensitive
G462D mut ant (19) LeuA produced 40.8 mg/L of C6 alcohol
(Table 1, column 5). In contrast, without overexpression of
LeuABCD, no C6 alc ohol production was detected (Table 1,
c olumn 3).
Structure-Based Redesign of KIVD. Because KIVD and ADH6 are
promiscuous enzymes, they can also convert other intracellular
2-keto acids into alc ohols (Fig. 1B , Table 1). To reduce the
for mation of byproducts and drive the carbon flux toward the
t arget C6 alcohol, we examined the effect of engineering KIVD
with higher selectivit y toward 2-keto-4-methylhexanoate. The
protein sequence alignment shows that KIVD has 40% and 31%
sequence identities with Enterobacter cloacae indolepyr uvate
decarboxylase IPDC (20) and Zymomonas mobilis pyruvate
decarboxylase ZmPDC (21) respectively. A homology model for
the substrate-binding region of KIVD and IPDC was built based
on the crystal structures of ZmPDC (PDB: 1ZPD). As can be
seen from Fig. 2, 4 residues, Ser-286, Phe-381, Val-461, and Met
538, in c ombination with c ofactor thiamine diphosphate
(ThDP), delineate the keto-acid binding pocket of KIVD. No-
ticeably, the corresponding residues of ZmPDC have bulkier side
chains, Tyr-290, Trp-392, Ile-472, and Trp-551; and those of
IPDC have smaller ones, Thr-290, Ala-387, Val-467 and Leu-
542. These dif ferences can explain the substrate spectrum of
these 2-keto acid decarboxylases and suggests that substitution of
related amino acids might be able to change substrate specificit y.
A ZmPDC I472A variant was shown to be more active on
longer-chain keto acids other than pyruvate (22). The c orre-
sponding residue of KIVD, V461, was thus mutated to alanine.
Compared with the wild-type KIVD, the V461A mut ant pro-
duced 3 times more 3-methyl-1-pentanol (Table 1, column 6).
Further optimization was performed by mut ating either F381 or
M538 to smaller hydrophobic side chains such as leucine or
alan ine (Table 1). The F381L/V461A mutant was the best variant
obt ained and produced 384.3 mg/L of 3-methyl-1-pentanol.
Both wild-type and F381L/V461A KIVD were added an
N-ter minal 6xHis-tag, overexpressed and purified through Ni-
NTA columns. The kinetic parameters for activation of 2-ke-
toisovalerate (c ognate substrate) and 2-keto-4-methylhexanoate
Table 1. Production proﬁle of alcohols from the designed pathway, with different KIVD mutants
Alcohol titer (mg/L)
41.1±4.1 94.6±11.5 213.2±12.3 132.7±14.3 27.3±5.1 100.7±18.0 43.3±12.9 83.3±6.2
1179.1±76.5 936.2±42.7 81.8±19.1 49.6±12.9 5.3±2.9 37.3±8.1 16.1±3.3 8.0±1.1
ND 17.8±0.9 493.2±31.5 371.4±14.6 192.1±7.7 432.1±52.0 219.3±51.7 381. 7±36.3
54.1±5.5 63.4±14.8 205.2±9.4 264.5±9.9 142.9±10.5 246.2±38.0 122.8±33.6 68.0±6.7
131.6±2.6 384.7±91.3 726.4±5.9 687.5±16.9 898.7±11.6 750.5±149.4 826.8±144.4 963.1±48.3
ND ND 494.1±22.9 503.9±4.6 750.5±52.9 556.6±86.8 482.9±111.9 444.6±35.5
6.5±1.1 40.8±5.5 135.6±7.8 299.2±6.8 141.7±11.7 264.5±51.6 384.3±30.3
ND ND ND ND 17.4±0.3 ND 18.5±0.9 7.3±0.4
Note that the V461A/F381L mutant gives the highest titer of 3-methyl-1-pentanol. E. coli cultures were grown in M9 medium with 20 g/L glucose plus 0.1 mM
IPTG at 30 °C for 40 h. These products were identiﬁed by GC-MS and quantiﬁed by GC-FID. ND, not detectable.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0807157106 Zhang et al.
(nonnatural substrate) were determined using a coupled enzy-
matic assay (22). Compared with the wild-type KIVD, for the
smaller substrate, 2-ketoisovalerate, F381L/V461A KIVD has a
sign ificantly lower k
versus 38.3 s
) and higher K
mM versus 2.2 mM); for 2-keto-4-methylhexanoate, F381L/
V461A KIVD has a comparable k
versus 10.8 s
a slightly higher K
(0.22 mM versus 0.14 mM). Thus, the
specificit y constant k
of F381L/V461A KIVD toward 2-ke-
to-4-methylhexanoate is 40-fold higher than that toward 2-ke-
toisovalerate. In comparison, the specificity constant k
wild-type KIVD toward 2-keto-4-methylhexanoate is only 4-fold
higher than that toward 2-ketoisovalerate (Table 2). Such a
change in KIVD specificity distinguishably af fects the distribu-
tion profile of alcohol products (more long-chain alcohols and
less short- chain alcohols).
Enlarging the Binding Pocket of LeuA. Besides KIVD, the other key
enz y me deter min ing the carbon flux toward 3-methyl-1-
pent anol production is LeuA. LeuA catalyzes the condensation
of acetyl-CoA with 2-keto-3-methylvalerate, which is the first
step of the expanded metabolic pathway (Fig. 1B). LeuA also
c ompetes with KIVD for substrate 2-keto-3-methylvalerate, and
thus reduces the formation of side product 2-methyl-1-butanol.
Engineering KIVD with higher activit y toward 2-keto-3-
methylvalerate should help increase 3-methyl-1-pentanol pro-
duction. As inferred from the crystal structure of Mycobacterium
tuberculosis LeuA (23), residues His-167, Ser-216 and Asn-250
are within a radius of4Åofthe
-methyl g roup of bound
2-ketoisovalerate (Fig. 3A). Nonnatural substrate (S)-2-keto-3-
methylvalerate contains one more methyl g roup that would
cause steric hinderance with Ser-216, which could be relieved by
mut ating serine to the smallest amino acid glycine. Multiple
protein sequence alignment shows that E. coli LeuA shares 92%
and only 21% sequence identity with Salmonella typhimur ium
LeuA and M. tuberculosis LeuA respectively. Fortunately, the
binding pocket is well c onserved and the corresponding residues
of E. coli LeuA are His-97, Ser-139 and Asn-167 (Fig. 3B). The
G462D/S139G mut ant LeuA was cloned and produced 793.5
mg/L 3-methyl-1-pent anol (Table 3, column 3), tw ice the amount
by G462D LeuA.
Enz ymatic assay indicates that G462D LeuA has an extremely
) for (S)-2-keto-3-methylvalerate, which is
333-fold less than that for 2-ketoisovalerate (6.0 s
G462D LeuA has a comparable K
for both substrates (55
M), the low k
may be why a previous report
showed that 2-keto-3-methylvalerate is a strong inhibitor of
LeuA (24). However, the S139G mutation increases the k
7-fold for (S)-2-keto-3-methylvalerate to 0.12 s
Additional mut ations were then performed on His-97 and
Asn-167. Even though better mutant was not found for produc-
tion of 3-methyl-1-pentanol, interestingly, the G462D/S139G/
N167A triple mutant produced 51.9 mg/L 4-methyl-1-hexanol
(C7), and the G462D/S139G/H97A/N167A quadruple mutant
Fig. 2. Stereoview of active site of Z. mobilis pyruvate decarboxylase ZmPDC
(green) and the corresponding homology model of Enterobacter cloacae
indolepyruvate decarboxylase IPDC (cyan) and KIVD (purple), using ZmPDC as
the template. The multiple sequence alignment was performed with Clust-
alW. Residues Y290, W392, and W551 of ZmPDC restrict the size of the binding
pocket and prevent activating substrates larger than pyruvate. Residues F381,
V461, and M538 of KIVD were mutated to smaller hydrophobic residues such
as alanine and leucine to allow the enzyme accept substrates larger than
Table 2. Kinetic parameters of wild type and mutant KIVD
Wild type V461A/F381L
2-Ketoisovalerate 2.2± 0.9 38.3± 9.8 17 7.7± 1.8 2.7± 0.6 0.35
0.14± 0.01 10.8± 0.3 77 0.22± 0.02 3.0± 0.1 14
Fig. 3. Residues in the active site of LeuA. (A) Binding pocket of Mycobac-
terium tuberculosis LeuA (PDB: 1SR9) complexed with its natural substrate
2-ketoisovalerate (KIV, green). (S)-2-keto-3-methylvalerate has one more
methyl group (dark green sphere) that would cause steric conﬂict with Ser 216
(red sphere), His-167, and Asn-250. (B) Multiple sequence alignment of M.
tuberculosis, E. coli, and Salmonella typhimurium LeuA. The binding pocket is
conserved, and the corresponding residues of E. coli LeuA are His-97, Ser-139,
and Asn-167. These residues were subjected to site-speciﬁc mutagenesis.
Zhang et al. PNAS
December 30, 2008
produced 57.3 mg/L 4-methyl-1-hexanol (C7) and 22.0 mg/L
5-methyl-1-hept anol (C8).
Biosynthesis of a Repertoire of Nonnatural Alcohols. Because the
engineered LeuA has larger binding pockets, the chain elongation
activities may continue several more rounds by LeuA on the 2-keto
acids produced from the LeuABCD or other pathways (Fig. 4A).
For example, 2-ketobutyrate can be converted to 2-ketovalerate,
then to 2-ketocaproate, and finally to 2-ketoheptanoate by
LeuABCD. In parallel, 2-keto-3-methylvalerate can be converted
to 2-keto-4-methylhexanoate, then to 2-keto-5-methylhepatanoate
and to 2-keto-6-methyloctanoate. All these keto acids are substrates
of F381L/V461A KIVD. Upon decarboxylation, the corresponding
aldehydes are reduced to the corre sponding alcohols by ADH6.
Indeed, we observed accumulation of 5 other nonnatural alcohols:
1-pentanol, 1-hexanol, 4-methyl-1-pentanol, 4-methyl-1-hexanol
and 5-methyl-1-heptanol (Tables 1 and 3). The anteiso-methyl-
branched alcohols are all derived from the same chiral precursor,
(S)-2-keto-3-methylvalerate. The S-configuration of the stereo-
genic center in these alcohols remains unchanged during biosyn-
thesis as confirmed by chiral GC analysis of (S)-2-methyl-1-butanol
[see supporting information (SI)] and (S)-3-methyl-1-pentanol
(Fig. 4B). These enantiomerically pure alcohols may be useful chiral
synthons for chemical synthesis (25).
In this work, we have shown that by combining protein engi-
neering and metabolic engineering approaches, it is possible to
ex pand the intermediary metabolism of E. coli to produce
various C5 to C8 alcohols that are not readily produced by
Table 4. Kinetic parameters of wild type and mutant LeuA
2-Ketoisovalerate 182± 26.0± 0.3
33 104± 52.1± 0.1 20
55± 60.018± 0.001
0.33 144± 13 0.12± 0.02 0.83
Table 3. Alcohol production with different LeuA mutants
Alcohol titer (mg/L)
117.2±3.8 122.1±7.2 51.1±6.9 39.4±1.3 33.2±5.7 54.7±7.4
49. 6±2.2 70.0±9.0 155.2±12.3 165.1±18.6 208.1±8.3 230.4±39.1
178.5±5.5 174.1±13.1 25.2±4.2 30.6±2.6 28.6±2.4 17.9±6.3
37.4±2.3 69.4±8.8 37.3±7.9 16.4±2.6 81.8±2.6 12.2±1.9
901.3±28.6 867.2±20.8 594.7±40.2 661.3±21.2 740.5±28.2 613.5±43.9
204.7±16.5 169.8±36.5 29.9±4.4 17.3±0.5 14. 2±1.3 ND
70.5±4.6 48.5±18.0 202.4±1.1 123.2±12.2 ND 80.1±5.6
793.5±46.5 685.7±16.0 337.4±41.0 288.1±32.5 119.1±6.0 290.6±34.1
37.4±2.8 38.4±8.3 16.6±0.9 16.5±1.4 ND ND
ND ND ND 51.9±9.3 ND 57.3±7.8
ND ND ND ND ND 22.0±2.5
Note that the G462D/S139G mutant gives the highest titer of 3-methyl-1-pentanol). E. coli cultures were grown in M9 medium with 20 g/L glucose plus 0.1
mM IPTG at 30 °C for 40 h. These products were identiﬁed by GC-MS and quantiﬁed by GC-FID. ND, not detectable.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0807157106 Zhang et al.
microorgan isms. Because of their specific physical and chemical
properties, these long chain alcohols could be good candidates
as biofuels or renewable chemical reagents. For practical appli-
cations, further met abolic engineering (26) and enzyme engi-
neering (27) will be needed to increase the production yield and
rate of these compounds. Because the 2-keto acid precursors of
alc ohols can be converted to amino acids by aminotransferases,
we hereby also provide a biosynthetic way, instead of traditional
organ ic synthesis, to expand the repertoire of nonnatural amino
acids that have recently found broad applications (6, 28).
Materials and Methods
Vector Construction. All cloning procedures (see SI for cloning scheme) were
carried out in the E. coli strain XL10-gold (Stratagene). Oligos were synthe-
sized by Operon Biotechnologies (see SI for sequence details). PCRs were
performed with KOD polymerase (Novagen).
Fermentation Procedure. The aminotransferase genes, ilvE and tyrB,ofa
threonine-hyperproduction E. coli strain ATCC98082 were inactivated by P1
transduction (29). This modiﬁed strain was transformed with pZS㛭thrO,
pZAlac㛭tdcBilvGMCD and pZE㛭LeuABCDKA6 for alcohol production. Over-
night cultures incubated in LB medium were diluted 100-fold into 5 mL of M9
medium supplemented with 1⫻ trace metal mix A5 (11), 0.5% yeast extract
and 2% glucose in 125-ml conical ﬂasks. Antibiotics were added appropriately
(ampicillin 100 mg/L, spectinomycin 25 mg/L, kanamycin 25 mg/L). Cells were
grown to an optical density at 600 nm of ⬇1.0 at 37 °C, followed by adding 0.1
D-thiogalactoside (IPTG). Cultures were then transferred to a
30 °C shaker (250 rpm) and incubated for 40 h.
GC-MS Analysis. The GC-MS system is composed of model 6890N network GC
system (Agilent Technologies), a model 7883B injector and autosampler (Agi-
lent Technologies) and a model 5973 network mass selective detector (Agilent
Technologies). Samples were separated through a DB-5ms capillary column
(30 m, 0.25-mm internal diameter, 0.25-
m ﬁlm thickness; Agilent Technolo-
gies) with helium (1 mL䡠min
) as the carrier gas. Alcohols extracted by 200
of toluene from 1 mL of fermentation medium were directly injected for mass
GC–FID Analysis. Alcohol compounds were quantiﬁed by a gas chromatograph
equipped with ﬂame ionization detector. The system is composed of a model
5890A gas chromatograph (Hewlett Packard) and a model 7673A automatic
injector, sampler and controller (Hewlett Packard). Samples were separated
through A DB-FFAP capillary column (30 m, 0.32-mm internal diameter,
m ﬁlm thickness; Agilent Technologies). GC oven temperature was
initially placed at 40 °C for 2 min, increased with a gradient of 5 °C min
45 °C, and held for 4 min. Then it was increased with a gradient 15 °C min
until 230 °C and held for 4 min. Helium was used as the carrier gas. The
temperature of injector and detector was set at 225 °C. Alcohol standards
were purchased from either Sigma–Aldrich or TCI America.
For chiral GC analysis, samples were separated through a HP-CHIRAL 20ß
column (30 m, 0.32-mm internal diameter, 0.25-
m ﬁlm thickness; Agilent
Technologies). The racemic mixture of 3-methyl-1-pentanol could not be
directly resolved. However, after reaction with N-Methyl-N-[tert-butyldim-
ethyl-silyl]triﬂuoroacetimide (Pierce), the conjugated product could be re-
solved into 2 peaks. GC oven temperature was initially placed at 50 °C for 4
min, increased with a gradient of 10 °C min
until 90 °C, and held for 2 min.
And then it was increased with a gradient 2 °C min
until 130 °C and held for
2 min. Finally the temperature was increased with a gradient 35 °C min
235 °C and held for 2 min. Helium was used as the carrier gas. The temperature
of injector and detector was set at 225 °C.
Protein Expression and Purification. Both gene fragments encoding wild-type
and F381L/V461A KIVD were ampliﬁed from plasmid pZE㛭LeuABCDKA6,
using primers hiskivd㛭tevfwd and hiskivd㛭bamrev. After digestion with
BamHI, the gene fragments were inserted into expression plasmid pQE9
(Qiagen) to yield pQE㛭 hiskivd㛭wt and pQE㛭 hiskivd㛭FL. The ADH6 gene
fragment was ampliﬁed from yeast genomic DNA, using primers
hisadh㛭tevfwd and hisadh㛭bamrev, digested with BamHI and inserted into
pQE9 to generate pQE㛭hisadh6. Similarly, genes encoding G462D and
G462D/S139G LeuA were ampliﬁed from plasmid pZE㛭LeuABCDKA6, using
primers hisleua㛭tevfwd and hisleua㛭bamrev. After digestion with BamHI,
the PCR products were ligated into pQE9 to create pQE㛭hisleua㛭GD and
pQE㛭hisleua㛭GS. The resulting expression plasmids pQE㛭 hiskivd㛭 wt,
pQE㛭hiskivd㛭FL, pQE㛭hisadh6, pQE㛭hisleua㛭GD and pQE㛭hisleua㛭GS were
transformed into E. coli strain BL21(DE3) harboring pREP4 (Qiagen). Cells
were inoculated from an overnight preculture at 1/100 dilution and grown
in 200 mL of 2XYT rich medium containing 50 mg/L ampicillin and 25 mg/L
kanamycin. At an OD
of 0.6, recombinant proteins were expressed by
Fig. 4. Structures of biosynthesized alcohols. (A) Nonnatural alcohols produced and their corresponding metabolic pathways. (B) 3-methyl-1-pentanol is
S-isomer as conﬁrmed by chiral GC analysis after MTBSTFA (N-Methyl-N-[tert-butyldimethyl-silyl]triﬂuoroacetimide) derivatization.
Zhang et al. PNAS
December 30, 2008
induction of the cell cultures with 0.1 mM IPTG, followed by incubation at
30 °C overnight. Cell pellets were lysed by sonication in a buffer containing
250 mM NaCl, 2 mM DTT, 5 mM imidazole and 50 mM Tris pH 9.0. By
applying a stepwise gradient of imidazole (up to 250 mM), enzymes were
puriﬁed from crude cell lysates through Ni-NTA column chromatography.
The fractions of highest purity were pooled and buffer-exchanged using
Amicon Ultra centrifugal ﬁlters (Millipore). Storage buffer 1 [50
buffer (pH 8.0), 1 mM MgSO
, 20% glycerol] was used for LeuA and ADH6,
and storage buffer 2 [50
M Tris buffer (pH 8.0), 1 mM MgSO
, 0.2 mM
ThDP, 20% glycerol] was used for KIVD. The concentrated protein solutions
were aliquoted (100
L) into PCR tubes and ﬂash frozen at ⫺80 °C for long
Enzymatic Assay of KIVD. Substrate 2-ketoisovalerate (KIV) was purchased
from Sigma–Aldrich, and (S)-2-keto-4-methylhexanoate (KHV) was custom
synthesized by AsisChem Inc. Protein concentration was determined by mea-
suring UV absorbance at 280 nm. The decarboxylation activity of KIVD was
measured at 30 °C, using a coupled enzymatic assay method. Excess ADH6 was
used to reduce aldehyde into alcohol, and concomitantly, cofactor NADPH
was oxidized to NADP⫹. The assay mixture contained 0.2 mM NADPH, 0.1
ADH6 and 0.1–20 mM 2-keto acids in assay buffer (50 mM potassium phos-
phate buffer, pH 6.8, 1 mM MgSO
, 0.5 mM ThDP) with a total volume of 0.2
mL. The reactions were started by adding 2
L of KIVD (ﬁnal concentrations:
for KIV, 20 nM wild-type KIVD, 200 nM F381L/V461A KIVD; for KHV, 50 nM
both), and the consumption of NADPH was monitored at 340 nm (extinction
coefﬁcient, 6.22 mM
). Kinetic parameters (k
) were deter-
mined by ﬁtting initial velocity data to the Michaelis–Menten equation, using
Measurement of LeuA activity. The assay mixture contained 100 mM KCl, 2 mM
MgCl2, 1 mM acetyl-CoA and 100 mM Tris (pH 8.0) with a total volume of 100
L. G462D or G462D/S139G LeuA (each 100 nM) was reacted with 2-ke-
toisovalerate in a concentration range from 25
M to 1 mM for 10 min at 30 °C,
M G462D or 1.5
M G462D/S139G LeuA was reacted with (S)-2-keto-
3-methylvalerate in a concentration range from 50
M to 2 mM for 30 min at
30 °C. The reactions were stopped by adding 0.3 mL of ethanol. Then 0.2 mL
of a fresh 1 mM solution of 5,5⬘-Dithio-Bis (2 Nitrobenzoic Acid) in 100 mM Tris
buffer (pH 8.0) was added, and the yellow color product was measured at 412
nm. The values obtained were corrected for unspeciﬁc hydrolysis by subtract-
ing the absorbance of controlled samples without addition of 2-keto acids. A
molar extinction coefﬁcient of 13,600 M
was used in the ﬁnal calcu-
ACKNOWLEDGMENTS. This work was supported in part by the University of
California, Los Angeles Department of Energy Institute for Genomics and
1. Elowitz MB, Leibler S (2000) A synthetic oscillatory network of transcriptional regula-
tors. Nature 403:335–338.
2. Gardner TS, Cantor CR, Collins JJ (2000) Construction of a genetic toggle switch in
Escherichia coli. Nature 403:339 –342.
3. Basu S, Gerchman Y, Collins CH, Arnold FH, Weiss R (2005) A synthetic multicellular
system for programmed pattern formation. Nature 434:1130 –1134.
4. Fung E, et al. (2005) A synthetic gene-metabolic oscillator. Nature 435:118 –122.
5. Gibson DG, et al. (2008) Complete chemical synthesis, assembly, and cloning of a
Mycoplasma genitalium genome. Science 319:1215.
6. Wang L, Brock A, Herberich B, Schultz PG (2001) Expanding the genetic code of
Escherichia coli. Science 292:498 –500.
7. Yoo TH, Link AJ, Tirrell DA (2007) Evolution of a ﬂuorinated green ﬂuorescent protein.
Proc Natl Acad Sci USA 104:13887–13890.
8. Prescher JA, Dube DH, Bertozzi CR (2004) Chemical remodelling of cell surfaces in living
animals. Nature 430:873– 877.
9. Yalkowsky SH, He Y (2003) CRC Handbook of Aqueous Solubility Data (CRC, Boca
10. Ehrlich F (1907) U
ber die Bedingungen der Fuselo¨ lbildung und u¨ ber ihren Zusammen-
hang mit dem Eiweissaufbau der Hefe. Berichte der Deutschen Chemischen Gesell-
11. Atsumi S, Hanai T, Liao JC (2008) Non-fermentative pathways for synthesis of branched-
chain higher alcohols as biofuels. Nature 451:86 – 89.
12. Handel TM, Williams SA, DeGrado WF (1993) Metal ion-dependent modulation of the
dynamics of a designed protein. Science 261:879 – 885.
13. Atwell S, Ultsch M, De Vos AM, Wells JA (1997) Structural plasticity in a remodeled
protein–protein interface. Science 278:1125.
14. Arnold FH (2001) Combinatorial and computational challenges for biocatalyst design.
15. Berg JM, Tymoczko JL, Stryer L (2002) Biochemistry (W.H. Freeman & Co, New York), 5th
16. de la Plaza M, Fernandez de Palencia P, Pelaez C, Requena T (2004) Biochemical and
molecular characterization of alpha-ketoisovalerate decarboxylase, an enzyme in-
volved in the formation of aldehydes from amino acids by Lactococcus lactis. FEMS
Microbiol Lett 238:367–374.
17. Larroy C, Rosario Fernandez M, Gonzalez E, Pares X, Biosca JA (2003) Properties and
functional signiﬁcance of Saccharomyces cerevisiae ADHVI. Chem Biol Interact 143–
18. Shiio I, Nakamori S, Sano K (1971) US Patent 3,580,810.
19. Gusyatiner MM, Lunts MG, Kozlov YI, Ivanovskaya LV, Voroshilova EB (2002) US Patent
20. Schutz A, et al. (2003) Crystal structure of thiamindiphosphate-dependent indolepyru-
vate decarboxylase from Enterobacter cloacae, an enzyme involved in the biosynthesis
of the plant hormone indole-3-acetic acid. Eur J Biochem 270:2312–2321.
21. Dobritzsch D, Konig S, Schneider G, Lu G (1998) High resolution crystal structure of
pyruvate decarboxylase from Zymomonas mobilis. Implications for substrate activation
in pyruvate decarboxylases. J Biol Chem 273:20196 –20204.
22. Siegert P, et al. (2005) Exchanging the substrate speciﬁcities of pyruvate decarboxylase
from Zymomonas mobilis and benzoylformate decarboxylase from Pseudomonas
putida. Protein Eng Des Sel 18:345–357.
23. Koon N, Squire CJ, Baker EN (2004) Crystal structure of LeuA from Mycobacterium
tuberculosis, a key enzyme in leucine biosynthesis. Proc Natl Acad Sci USA 101:8295.
24. Wiegel J, Schlegel HG (1977) a-Isopropylmalate synthase from Alcaligenes eutrophus
H 16. Arch Microbio 112:239 –246.
25. Negishi E, Tan Z, Liang B, Novak T (2004) An efﬁcient and general route to reduced
polypropionates via Zr-catalyzed asymmetric CC bond formation. Proc Natl Acad Sci
26. Causey TB, Zhou S, Shanmugam KT, Ingram LO (2003) Engineering the metabolism of
Escherichia coli W3110 for the conversion of sugar to redox-neutral and oxidized
products: Homoacetate production. Proc Natl Acad Sci USA 100:825–832.
27. Bloom JD, et al. (2005) Evolving strategies for enzyme engineering. Curr Opin Struct
Biol 15:447– 452.
28. Dieterich DC, Link AJ, Graumann J, Tirrell DA, Schuman EM (2006) Selective iden-
tiﬁcation of newly synthesized proteins in mammalian cells using bioorthogonal
noncanonical amino acid tagging (BONCAT). Proc Natl Acad Sci USA 103:9482–
29. Baba T, et al. (2006) Construction of Escherichia coli K-12 in-frame, single-gene
knockout mutants: The Keio collection. Mol Syst Biol 2:10.1038.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0807157106 Zhang et al.