Enzyme-Catalyzed Carbonyl Olefination by the E. coli Protein YfeX in the Absence of Phosphines

Article abstract of DOI:10.1002/cctc.201600227

The Wittig-type carbonyl olefination reaction has no biocatalytic equivalent. To build complex molecular scaffolds, however, C-C bond-forming reactions are pivotal for biobased economy and synthetic biology. The heme-containing E. coli protein YfeX was found to catalyze carbonyl olefination by reaction of benzaldehyde with ethyl diazoacetate under aerobic conditions in the absence of a triphenylphosphine oxophile. The reaction was performed in whole cells and showed a product formation of 440 mg L^-1 in 1 h. It was, moreover, shown that the reaction could be performed under Wittig-analogue conditions in the presence of triphenylphosphine or triphenylarsine.

Full text of DOI:10.1002/cctc.201600227

DOI: 10.1002/cctc.201600227
Communications
Enzyme-Catalyzed Carbonyl Olefination by the E. coli
Protein YfeX in the Absence of Phosphines
Martin J. Weissenborn+,*[a] Sebastian A. Lçw+,[a] Niels Borlinghaus,[a] Miriam Kuhn,[a]
Stefanie Kummer,[a] Fabian Rami,[b] Bernd Plietker,[b] and Bernhard Hauer*[a]
The Wittig-type carbonyl olefination reaction has no biocatalyt-
ic equivalent. To build complex molecular scaffolds, however,
CÀC bond-forming reactions are pivotal for biobased economy
and synthetic biology. The heme-containing E. coli protein YfeX
was found to catalyze carbonyl olefination by reaction of ben-
zaldehyde with ethyl diazoacetate under aerobic conditions in
the absence of a triphenylphosphine oxophile. The reaction
was performed in whole cells and showed a product formation
of 440 mgLÀ1 in 1 h. It was, moreover, shown that the reaction
could be performed under Wittig-analogue conditions in the
presence of triphenylphosphine or triphenylarsine.
ers impressively showed the use of heme-containing P450s—
which naturally catalyze hydroxylation reactions[14,15]—to cata-
lyze carbenoid insertion into C=C[16] and NÀH bonds
(Scheme 1).[17] Fasan et al. accomplished these reactions[18,19]
and SÀH insertions[20] with sperm-whale myoglobin. All these
reactions relied on electrophilic carbenes.
Although several chemical methods for CÀC bond formation
are known, biocatalytic alternatives are less developed.
Amongst the most established methods are the Grignard reac-
tion,[1] hydroformylation,[2] and the Wittig reaction.[3] A break-
through in metal-catalyzed CÀC bond formation was achieved
by the cross-coupling chemistry of Heck,[4] Negishi,[5] and
Suzuki,[6] which, today, are standard methods in chemistry lab-
oratories.
To construct molecular scaffolds by biocatalytic means, alter-
native CÀC bond-forming enzymes are required. Amongst the
CÀC bond-forming enzymes are aldolases,[7] cyanohydrin
lyases,[8] thiamin diphosphate dependent lyases,[9] methyltrans-
ferases,[10] and cyclases.[11] We recently reported the catalysis of
Friedel–Crafts-type[12] and Prins–ene-type reactions[11] by a mu-
tated squalene-hopene cyclase. The substrate scope, however,
of the lyases was very limited, and cyclases have thus far only
been shown to catalyze intramolecular reactions.
Scheme 1. a,b) Previously published insertion-type chemistry and c) the
herein-presented carbonyl olefination strategy.
A promising strategy for the development of novel enzymat-
ic reactions is through analysis of the transition states of chem-
ical reactions of interest and by comparing these transition
states to those existing in biocatalytic reactions. We focused
on reactions with nucleophilic carbon atoms, as they are re-
quired to perform CÀC bond-forming reactions. The in situ
formed iron carbenoid species in the abovementioned heme-
catalyzed reactions assembles into a structure similar to that of
a C=P phosphorane. The carbon atom of these phosphoranes
is used in Wittig-type reactions for nucleophilic attack at car-
bonyl groups and in this way performs carbonyl olefinations.
Another example of a transition state similar to the iron car-
benoid species is that formed in the Büchner–Curtius–Schlot-
terbeck reaction.[21,22] A CÀC bond is formed by nucleophilic
attack of a diazo group carrying a carbon atom at an aldehyde
moiety. The nitrogen atom is subsequently lost and an epoxide
or—after rearrangement—a ketone is formed.
CÀC bond formation reactions are particularly important in
synthetic biology, for which artificial reaction pathways are ge-
netically engineered into organisms and used for the produc-
tion of industrially relevant molecules.[13] Arnold and co-work-
[a] Dr. M. J. Weissenborn,+ Dr. S. A. Lçw,+ N. Borlinghaus, M. Kuhn, S. Kummer,
Prof. Dr. B. Hauer
Institute of Technical Biochemistry
University of Stuttgart
Allmandring 31, 70569 Stuttgart (Germany)
E-mail: bernhard.hauer@itb.uni-stuttgart.de
[b] F. Rami, Prof. Dr. B. Plietker
Institute of Organic Chemistry
University of Stuttgart
Thus far, however, no nucleophilic carbenoid reaction with
heme proteins is known. We reasoned that a specific active
site would be required to stabilize the electron-rich carbene.
We moreover postulate that it is possible to study in parallel
Pfaffenwaldring 55, 70569 Stuttgart (Germany)
[+] These authors contributed equally to this work.
Supporting Information for this article can be found under http://
dx.doi.org/10.1002/cctc.201600227.
ChemCatChem 2016, 8, 1636 – 1640
1636
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Communications
a great number of different proteins with various active sites,
prosthetic groups, cofactors, and protein scaffolds. This can be
achieved by employing a highly concentrated cell lysate. The
bacterium Escherichia coli (E. coli), for instance, contains 4400
different proteins representing all the major enzyme classes
and only 50% of its proteins have been characterized.[23] There-
fore, the lysate of this organism is an excellent starting point
to find novel non-natural enzyme reactivities.
Table 1. Triphenylphosphine-free carbonyl olefination reaction of ethyl
diazoacetate (2a) and benzaldehyde (1).[a]
Catalyst
Relative activity[b]
Alkene
TON
de[c]
[mm]
[%]
We wish to report first insight into enzyme-catalyzed car-
bonyl olefinations (Scheme 1c). The reaction was performed in
the presence of and in the absence of triphenylphosphine, and
whole-cell experiments were also performed. In particular, the
reaction performed in the absence of triphenylphosphine is of
high interest for the biocatalysis community.
hemin
hemin+BSA[d]
Hmp
Bfr
KatE
YfeX
YfeX+CO
YfeX[e]
1
23
29
73
30
137
3
1937
0.03Æ0.006
0.69Æ0.03
0.88Æ0.03
2.2Æ0.2
0.002
0.03
85
n.d.
0.044
0.110
0.045
0.205
0.004
2.905
48
66
53
47
n.d.
60
0.9Æ0.2
4.1Æ0.3
0.08Æ0.02
58.1Æ3.2
We started our investigations by preparing a concentrated
E. coli lysate. This lysate was dialyzed to remove remaining
siderophores and other inorganic substances to avoid back-
ground reactions. The lysate was afterwards treated with ethyl
diazoacetate (EDA, 2a) and benzaldehyde (1) in the presence
of a reducing agent (i.e., Na2S2O4). The reaction was performed
under aerobic conditions to focus on proteins that were stable
and could catalyze the reaction in the presence of oxygen. The
analysis was performed with sensitive GC–MS in the single ion
mode (SIM) for several different potential products. Remarka-
bly, we identified carbonyl olefination product 3 under these
conditions. To find the corresponding protein, the lysate was
fractioned by consecutive fast protein liquid chromatography
(FPLC), and the active fractions were analyzed by protein se-
quencing (data not shown). This way, four different heme-con-
taining proteins were identified: flavohemoprotein (Hmp), bac-
terioferritin (Bfr), hydroperoxidase II (KatE), and YfeX with
unknown natural function.
[a] The employed concentrations were 10 mm 1, 10 mm 2a, 10 mm
Na2S2O4, and 20 mm catalyst. The reactions were performed under aerobic
conditions at 308C and 180 rpm. KPi: potassium phosphate buffer.
[b] The formation of 0.03 mm 3 by using hemin was used as relative activi-
ty=1. [c] n.d.=not determined. [d] BSA=bovine serum albumin. [e] Opti-
mized reaction conditions: 1m benzaldehyde, 100 mm Na2S2O4, 100 mm
EDA, 20 mm catalyst, and 10 mm hemin.
The genes of these proteins were overexpressed, purified by
His-tag affinity chromatography, and subsequently tested for
activity in the carbonyl olefination reaction. Given that all four
proteins carry a heme domain, the reaction was performed by
using hemin as a catalyst as well. Free hemin showed forma-
tion of 0.03 mm alkene 3 and a diastereomeric excess (de) of
85% (E to Z).
Figure 1. Product concentration for the whole-cell conversion of ethyl diazo-
acetate and benzaldehyde into ethyl cinnamate (3) at an optical density
(l=600 nm) of 100 in M9-medium, 2 mm glucose, 40 mm Na2S2O4, 500 mm
benzaldehyde, 40 mm EDA, 10% ethanol, 1 h.
The transformations with the four purified proteins, howev-
er, showed a 137-fold reactivity increase (YfeX) relative to that
shown by free hemin (Table 1). Besides the improved formation
of alkene 3, an altered diastereomeric excess of 47% was also
observed. This indicated the influence of the active site on the
reaction selectivity. Focusing on the most active protein, we
optimized the reaction of YfeX further to obtain 58 mm prod-
uct, which corresponds to a turnover number (TON) of 2.9. Al-
though modest, these conversions represent the first enzymat-
ic activities for a carbonyl olefination reaction. With the finding
of initial activity, this enzyme can now be developed further by
directed evolution. This is comparable to the starting point of
de novo designed proteins.[24]
The scope of the carbonyl olefination and its versatility was
thereafter further evaluated. In previous work, FeII–porphyrin-
catalyzed carbonyl olefination was performed in the presence
of triphenylphosphine (PPh3) and in the absence of water and
oxygen.[25] While this paper was under review, Fasan and Tyagi
published a method for carbonyl olefination by using myoglo-
bin in the presence of an oxophile (phosphine or arsine).[26]
The formation of a stoichiometric amount of triphenylphos-
phine oxide in this reaction is a disadvantage that is also
known for the Wittig reaction.
However, we reasoned that the addition of triphenylphos-
phine could be used as a tool to alter the reaction mechanism
to influence the selectivity of the reaction. This would allow
the selectivity to be adjusted by adding differently stabilized
phosphines. In the presence of triphenylphosphine, the reac-
tion provided an altered diastereomeric excess of 67%
(Table 2). To demonstrate the influence of different substituted
Further improvement was achieved by performing the reac-
tion in whole cells. E. coli cells were transformed with the
pCA24N-YfeX plasmid and YfeX was overexpressed. The reac-
tion was performed in M9-medium for 1 h and yielded 2.5 mm
product, which corresponds to 440 mgLÀ1 hÀ1 (Figure 1).
ChemCatChem 2016, 8, 1636 – 1640
www.chemcatchem.org
1637
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Communications
Table 2. Carbonyl olefination reaction with benzaldehyde (1) in the
presence of different additives.[a]
Table 3. Catalytic activity of the YfeX mutants for the carbonyl olefination
reaction with benzaldehyde (1) and triphenylphosphine.[a]
Catalyst
Hemin addition
Alkene
TON
[mm]
[mm]
YfeX
–
–
–
–
–
5
5
10
20
50
127Æ15
130Æ6
22Æ3
6.4
6.5
1
1.5
0.5
11
YfeX (D143A)
YfeX (F248A)
YfeX (S234A)
YfeX (R232A)
YfeX (R232A)
YfeX (S234A)
YfeX (S234A)
YfeX (S234A)
YfeX (S234A)
Catalyst
X
Additive
de [%]
TON
29Æ4
hemin
YfeX
–
YfeX
–
YfeX
–
YfeX
–
YfeX[b]
hemin
YfeX
YfeX[b]
N2 (2a)
N2 (2a)
PPh3 (4a)
N2 (2a)
P(4-FC6H4)3 (5a)
N2 (2a)
P(4-MeOC6H4)3 (6a)
N2 (2a)
P(4-MeC6H4)3 (7a)
N2 (2a)
N2 (2a)
N2 (2a)
PPh3 (4b)
PPh3 (4b)
–
P(4-FC6H4)3 (5b)
–
P(4-MeOC6H4)3 (6b)
–
P(4-MeC6H4)3 (7b)
–
PPh3 (4b)
AsPh3 (8b)
AsPh3 (8b)
AsPh3 (8b)
69
67
65
67
68
80
82
82
82
77
99
94
92
6.6
7.7
–
6.9
–
10Æ2
216Æ16
321Æ61
310Æ23
346Æ27
277Æ29
16
15.5
17.3
13.9
1
–
[a] The employed concentrations were 10 mm 1, 5 mm 2a, 5 mm 4b,
5 mm Na2S2O4, and 20 mm catalyst. The reactions were performed under
aerobic conditions at 308C and 180 rpm for 1 h.
4.3
–
16
10.5
6.5
16
N2 (2a)
prosthetic heme group, we determined the protein loading by
using the Soret band at l=420 nm (Figure S7). Different
amounts of hemin were added to a YfeX-containing buffer so-
lution until no increase in the absorption at l=420 nm was
observed. Upon increasing the amount of hemin added to
above 10 mm, the hemin band at l=405 nm was observed,
which suggested a fully loaded protein. These measurements
showed a significantly lower loading of the YfeX protein with
the R232A and S234A variants. These variants were therefore
tested with additional hemin. In this way, the R232A variant
showed the formation of an increased amount of product
(216 mm alkene) in the presence of 5 mm hemin, and the S234A
mutant also showed the formation of an increased amount of
product (321 mm) in the presence of 5 mm hemin. The S234A
mutant showed the formation of 321 mm product with 5 mm
hemin added. This reactivity could not be improved by adding
higher amounts of hemin, which indicated that the reaction
was catalyzed in the active site of the protein (Table 3).
[a] The employed concentrations were 10 mm 1, 5 mm 2a, 5 mm 4a–7a,
5 mm 4b–8b, 5 mm Na2S2O4, and 20 mm catalyst. The reactions were per-
formed under aerobic conditions at 308C and 180 rpm. KPi: potassium
phosphate buffer. [b] Optimized reaction conditions: 1m benzaldehyde,
50 mm Na2S2O4, 50 mm 4b, 20 mm catalyst.
phosphines on the reaction selectivity, oxophiles 5b, 6b, and
7b were employed. They revealed altered de values of 67, 80,
and 82%.
To test if the enzyme environment had an influence on the
selectivity with this reaction setup, corresponding C=P phos-
phoranes 4a–7a were added instead of EDA (2a) and the tri-
phenylphosphine derivatives in the absence of YfeX. The ob-
served diastereomeric excess values of the phosphoranes were
the same as those for the phosphine derivatives, which sug-
gested that the selectivity was not influenced by the enzyme
in the presence of an oxophile.
In addition to olefin formation, an epoxide was also ob-
served as a side product. We found that the amount of epox-
ide formed in the presence of the enzyme was higher than
that formed in the presence of the hemin control, which sug-
gested an influence of the protein scaffold on this side reac-
tion (Figure S3 in the Supporting Information). An alternative
oxophile for the Wittig reaction is triphenylarsine (AsPh3);[27,28]
in this reaction setup with free hemin it showed a de of 99%.
In contrast to the phosphine experiments, the reaction with
YfeX showed altered selectivity (94%). This suggested that this
reaction proceeded in the active site of the YfeX protein.
To improve the reaction in the presence of the bulky triphe-
nylphosphine, a YfeX homology model[29] based on TyrA (from
Shewanella oneidensis PDB: 2IIZ)[30] was constructed (Figure S6).
Four amino acids (i.e., D143, F248, R232, and S234) were iden-
tified that represent potential steric hindrance or that lead to
side reactions owing to their polarity; thus they were ex-
changed for alanine. As expected, none of these variants
showed any influence on the selectivity. The F248, R232, and
S234 variants gave the products in lower amounts (Table 3).
However, as these residues might be important to bind the
On the basis of our results of the YfeX-catalyzed carbonyl
olefination in the presence and absence of phosphines we
suggest two different mechanisms:
1) In the presence of phosphine: the iron–carbenoid is
formed in the active site and subsequently undergoes nu-
cleophilic attack by triphenylphosphine, which is released
as phosphorane. The phosphorane then reacts with the al-
dehyde to form the alkene and O=PPh3 (Scheme 2, left).
2) In the absence of phosphine: the formed iron–carbenoid
acts as a nucleophile and attacks the aldehyde carbonyl
function. This results in an iron–oxetane intermediate. Sub-
sequent ring opening then leads to an FeIV =O species and
the alkene product (Scheme 2, right).
In summary, we showed two variations of biocatalytic car-
bonyl olefination reactions by using enzyme promiscuity
screening in E. coli. The isolated heme-containing E. coli protein
YfeX catalyzed the reaction under aerobic conditions with
modest turnovers, but this will serve as a starting point for a di-
rected evolution approach. The reaction was further improved
ChemCatChem 2016, 8, 1636 – 1640
www.chemcatchem.org
1638
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Communications
Scheme 2. Suggested reaction pathways for phosphine-free carbonyl olefination with a nucleophilic carbene (right) and in presence of triphenylphosphine
with an electrophilic carbene (left).
association European Federation of Pharmaceutical Industries
and Associations (EFPIA) under grant agreement no. 115360-2.
by performing it in whole cells, and this led to a product for-
mation of 440 mgLÀ1 hÀ1. The addition of a carbene-transfer-
ring phosphine influenced the reaction pathway significantly.
The addition of triarylphosphines led to an increase in the ac-
tivity and the ability to alter the E/Z selectivity by different sub-
stituents. This indicates that the reaction—in contrast to the
phosphine-free reactions—occurs outside of the active site.
Thus, the phosphine has to enter the active site of the enzyme
where it is transformed into the phosphorene, which can then
react outside of the active site. An interesting result is the al-
tered E/Z selectivity by adding triphenylarsine. Here, the reac-
tion seems to proceed at the metal–complex and the arsine
functions presumably as a reducing agent.
Keywords: biocatalysis
catalysis · olefination · Wittig reactions
·
CÀC bond formation
· enzyme
[1] V. Grignard, C. R. Acad. Sci. 1900, 130, 1322–1324.
[2] D. Evans, J. A. Osborn, G. Wilkinson, J. Chem. Soc. A 1968, 3133.
[3] G. Wittig, U. Schçllkopf, Chem. Ber. 1954, 87, 1318–1330.
[4] R. F. Heck, J. Am. Chem. Soc. 1968, 90, 5526–5531.
[5] E. Negishi, S. Baba, J. Chem. Soc. Chem. Commun. 1976, 596b.
[6] N. Miyaura, K. Yamada, A. Suzuki, Tetrahedron Lett. 1979, 20, 3437.
[7] B. L. Horecker, O. Tsolas, C. Y. Lai, Enzym. 1972, 7, 213–258.
[8] H. Wajant, F. Effenberger, Biol. Chem. 1996, 377, 611 –617.
[9] F. Jordan, Nat. Prod. Rep. 2003, 20, 184–201.
The herein reported results present an important contribu-
tion to our recent work on influencing biological systems
with chemical additives (chemBricks).[31,32] Besides the applica-
tion of phosphines and arsines as “chemBricks”, the phos-
phine-free reaction in particular offers the possibility to be ap-
plied in completely novel reaction pathways in whole-cell pro-
cesses, which would thereby avoid the formation of phosphine
oxide as a side product. A possible reaction pathway for the re-
action was postulated, but the mechanism has to be con-
firmed. This question is part of current investigations in our
laboratory.
[10] H. Stecher, M. Tengg, B. J. Ueberbacher, P. Remler, H. Schwab, H. Griengl,
M. Gruber-Khadjawi, Angew. Chem. Int. Ed. 2009, 48, 9546–9548;
Angew. Chem. 2009, 121, 9710–9712.
[11] S. C. Hammer, A. Marjanovic, J. M. Dominicus, B. M. Nestl, B. Hauer, Nat.
Chem. Biol. 2015, 11, 121–126.
[12] S. Hammer, J. Dominicus, P. SyrØn, Tetrahedron 2012, 68, 7624–7629.
[13] B. M. Nestl, S. C. Hammer, B. A. Nebel, B. Hauer, Angew. Chem. Int. Ed.
2014, 53, 3070–3095; Angew. Chem. 2014, 126, 3132–3158.
[14] H. Joo, Z. Lin, F. H. Arnold, Nature 1999, 399, 670–673.
[15] S. M. Hoffmann, M. J. Weissenborn, Ł. Gricman, S. Notonier, J. Pleiss, B.
Hauer, ChemCatChem 2016, DOI: 10.1002/cctc.201501397.
[16] P. S. Coelho, E. M. Brustad, A. Kannan, F. H. Arnold, Science 2013, 339,
307–310.
[17] Z. J. Wang, N. E. Peck, H. Renata, F. H. Arnold, Chem. Sci. 2014, 5, 598–
601.
[18] G. Sreenilayam, R. Fasan, Chem. Commun. 2015, 51, 1532–1534.
[19] M. Bordeaux, V. Tyagi, R. Fasan, Angew. Chem. Int. Ed. 2015, 54, 1744–
1748; Angew. Chem. 2015, 127, 1764–1768.
Acknowledgements
The authors thank Nico Kress and Dr. Janosch Klebensberger for
fruitful discussions, Max-Philipp Fischer for graphical support,
and Kerstin Bernecker for support in the laboratory. S.A.L. was fi-
nanced by the Innovative Medicines Initiative (IMI) Joint Under-
taking between the European Union and pharmaceutical industry
[20] V. Tyagi, R. Bonn, R. Fasan, Chem. Sci. 2015, 6, 2488–2494.
[21] E. Buchner, T. Curtius, Ber. Dtsch. Chem. Ges. 1885, 18, 2371–2377.
[22] F. Schlotterbeck, Ber. Dtsch. Chem. Ges. 1907, 40, 1826–1827.
[23] Y. Yamazaki, H. Niki, J. Kato, Methods Mol. Biol. 2008, 416, 385–389.
[24] H. A. Dailey, A. N. Septer, L. Daugherty, D. Thames, S. Gerdes, E. V. Stabb,
A. K. Dunn, T. A. Dailey, J. D. Phillips, MBio 2011, 2, e00248-11.
ChemCatChem 2016, 8, 1636 – 1640
www.chemcatchem.org
1639
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Communications
[25] G. A. Mirafzal, G. Cheng, L. K. Woo, J. Am. Chem. Soc. 2002, 124, 176–
177.
[30] C. Zubieta et al., Proteins 2007, 69, 234–243, see Supporting Informa-
tion.
[26] V. Tyagi, R. Fasan, Angew. Chem. Int. Ed. 2016, 55, 2512–2516; Angew.
Chem. 2016, 128, 2558–2562.
[31] S. A. Lçw, I. M. Lçw, M. J. Weissenborn, B. Hauer, ChemCatChem 2016, 8,
911–915.
[27] M. S. Holzwarth, I. Alt, B. Plietker, Angew. Chem. Int. Ed. 2012, 51, 5351–
5354; Angew. Chem. 2012, 124, 5447–5450.
[32] S. A. Lçw, B. M. Nestl, M. J. Weissenborn, F. Zepeck, B. Hauer, Org. Pro-
cess Res. Dev. 2015, 19, 1544–1547.
[28] P. Wang, C.-R. Liu, X.-L. Sun, S.-S. Chen, J.-F. Li, Z. Xie, Y. Tang, Chem.
Commun. 2012, 48, 290–292.
Received: February 25, 2016
[29] K. Arnold, L. Bordoli, J. Kopp, T. Schwede, Bioinformatics 2006, 22, 195–
201.
Published online on April 13, 2016
ChemCatChem 2016, 8, 1636 – 1640
www.chemcatchem.org
1640
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim