of 10
1
D
ual
-
Function Enzym
e
C
atalys
is
for
Enantioselectiv
e Carbon
Nitrogen
Bond Formation
Zhen Liu
1
, Carla Calv
ó
-
Tusell
2
, Andrew Z. Zhou
1
, Kai Chen
1
,3
*, Marc Garcia
-
Borràs
2
* and Frances
H. Arnold
1
*
Chiral amines can be made by insertion of a
carbene
into an
N
H
bond
using
two
-
catalyst
system
s
that combine a transition metal carbene
-
transfer catalyst
and
a chiral proton
-
transfer catalyst t
o
enforce stereocontrol
.
H
aem
proteins
can
effect
carbene
N
H
insertion, but
asymmetric
protonation in an
active
site
replete with
proton sources
is challenging
.
Here we
describe
engineered cytochrome P450 enzymes
that catalyze
carbene N
H inserti
on
to prepare biologically
relevant
α
-
amino lactones
with
high activity and enantioselectivity
(up to
3
2
,
1
00 total
turnovers,
>99% yield and 98% e.e.)
.
T
he
se
enzymes
serve
as dual
-
function catalysts
,
induc
ing
carbene
transfer
and
promot
ing
the
subsequent
proton transfer
with
excellent stereoselectivity
in
a single active site
. Computational studies uncover the detailed mechanism of this
new
-
to
-
nature
enzym
atic
reaction
and explain
how
active
-
site
residues
accelerate th
is
transformation and
provide
stereocontrol
.
A
mines
are
ubiquitous
in bioactive molecules and functional materials
1,2
,
and
the
develop
ment of
efficient and selective
methods
for
C
N bond
construct
ion
remains
one of the central themes of modern
organic chemistry and biochemistry
3
5
. Among the numerous
ways
to construct C
N bonds, carbene
insertion into
N
H
bonds
6
10
benefits from
the
high reactivity
of carbene species
and excellent
functional group compatibility
to
rapidly bui
ld
complex
nitrogen
-
containing
molecules
.
In
the last
several years
,
empowered by
directed evolution,
metallo
-
haem
-
dependent
enzymes
(cytochromes P450,
cytochromes
c
and globins
, for example
)
have exhibited
an impressive ability to
catalyz
e
non
-
natural
carbene
-
and nitrene
-
transfer reactions
with high efficiency and selectivity.
Specifically,
haem proteins
have been
engineered to perform
carbene N
H insertion reactions
with catalytic efficiency far
exceed
ing
their
small
-
molecule
counterparts
(up to thousands of total turnover number
s
(TTN))
11
14
.
However,
compared to
cyclopropanation
15
,
C
H insertion
16
and
many o
ther carbene
transfer reactions
also
catalyzed by
haem proteins
17,18
,
N
H insertion reactions are
still underdeveloped
,
especially
with
respect to
high
stereo
c
ontrol
.
In
small
-
molecule
cataly
sis
, a
common
strategy
for asymmetric N
H insertion
is
to
employ
a
transition
-
metal catalyst for carbene transfer
along with
a separate chiral
proton
-
transfer catalyst
(PTC)
for stereoinduction
(
Fig
.
1a
)
19,20
.
T
he
carbene precurs
or first react
s
to form
a
metal
carbene
species
,
which
can be
trapped by
the amine substrate
through
nucleophilic attack
,
generating
an
ylide
intermediate
.
T
he
asymmetric
protonation
of
the
ylide
is
then
guided by
a
chiral
PTC
, such as
a
chiral
phosphoric acid
19
or
amino
-
thiourea
20
;
other
prot
on
sources
need to
be
strictly
avoided
to ensure high
asymmetric induction
.
C
omputational studies by Shaik and coworkers
21
have
revealed
a similar
mechanism
for
haem
protein
-
catalyzed N
H insertion
reactions
.
Thus,
the
challenge
in
achieving high
enantioselectivity originates from the difficulty
in
precisely
controlling the protonation of
highly
1
Division of Chemistry and Chemical Engineering
210
-
41
, California Institute of Technology,
Pasadena, CA, USA.
2
Institut de Química Computacional i Catàlisi and Departament de Química,
Universitat de Girona, Girona, Catalonia, Spain.
3
Current address:
Innovative Genomics Institute,
University of California, B
erkeley, CA, USA
.
*e
-
mail: chenkaikc1990@gmail.com;
marcgbq@gmail.com; frances@cheme.caltech.edu
2
reactive
ylide intermediates in
an environment
where
the
enzyme as well as
water
molecules
are
potential proton sources
.
T
he
sole
biocatalytic system
for asymmetric N
H insertion
reported
to date
by Fasan
13
us
ed
2
-
diazopropanate benzyl esters and primary anilines
with
engineered myoglobin
s
as
the catalyst
and achieved
generally
low to moderate
enantiocontrol
.
B
y drawing an analogy to th
e
dual
-
catalyst
strategy
used in small
-
molecule catalysis
, we
envisioned that a
highly
enantioselective
biocatalytic
N
H insertion reaction
would be
possible if
the
enzym
e can
perform
two
di
stinct
functions
(
Fig. 1b
)
: 1)
generat
e the
carbene species
,
which
trigger
s
nucleophilic attack
from amines
,
and
2) induc
e
a
selective proton
-
transfer
event
immediately
after ylide formation in the
enzyme
active site.
Fig.
1
|
Asymmetric carbene N
H insertion using small
-
molecule catalysts and
enzymes. a
, The
proposed
reaction
mechanism for asymmetric
N
H insertion catalyzed by a transition
-
metal catalyst
and a chiral PTC.
b
,
H
aem
protein serves as a dual
-
function catalyst for e
nantioselective
carbene
N
H
insertion
.
Results and discussion
We
commenced
this
investigation
of
enzymatic
carbene N
H insertion
by
focus
ing
on
the
reaction
between
lactone diazo
1
and
N
-
methyl aniline
2a
.
This transformation is of particular interest
as
it is
expected to afford a biologically relevant
α
-
amino lactone product
3a
22
. In addition,
lactone
-
based
carbenes are usually associated with undesired
β
-
hydride
elimination processes, and
no
small
-
molecule
catalysts have been shown to
afford
precise stereocontrol
using
this type of carbene
23,24
.
O
ur previous
work
, however,
s
howed
that
engineered h
aem
proteins
could dramatically accelerate
the desired
lactone
carbene
-
transfer process
while circumventing
undesired
side
reactions
25,26
.
We thus
focused on
discoverin
g
an enzyme that
not only facilitate
s
the transfer of the lactone
-
carbene species
to amines
but
als
o
impose
s
stereocontrol in the subsequent proton
transfer
step to
deliver
an enantio
-
enriched product.
To
this end
, we screened
a
collection
of
40
h
aem
protein
variants
,
previously evolved
for different
carbene and nitrene
transformations
, in
the
form of
whole
Escherichia coli
(
E. coli
)
cell catalysts
(
Fig
.
2a
)
(See
Supporting Information
(SI)
for details)
.
While
most of the variants only exhibited low levels
of activity (<
5
% conversion)
,
a
n
FAD domain
-
truncated
P411
variant
(serine
-
ligated P450)
L7
(
well
C10
in Fig
.
2a
)
,
generated in
previous
engineering efforts for
lactone
carbene
C
H insertion
26
,
catalyze
d
3
this
N
H insertion
reaction with
81
%
yield
and 9
4
%
enantiomeric excess (
e.e.
).
Fu
rther
evaluation
of
the
enzyme
lin
e
age
for
lactone carbene C
H insertion
showed that
both
L6
and
L7
were
superior
biocatalysts for
this
amination
reaction
compared to
other
variants
in the same lineage
(
L1
L5
) (
Fig
.
2b
).
R
estoring
L7
to
a full
-
length
P411
(
L7_FL
)
by re
-
attaching its
native
P450
reductase domain
further improve
d
the
catalytic
performance of
th
e
enzyme
,
which
genera
ted
product
3a
in 92%
yield
an
d 95%
e.e
.
The improvement may be due to
the increased stability of
the
full
-
length protei
n
27
.
I
n going from
L5
to
L6
,
a single mutation
A264S
,
had a dramatic
impact on
th
e
N
H insertion
reaction
, especially
with respect to
enantioselectivity
(from
21%
to 92%
e.e.
) (
Fig. 2b
). This
observation
indicates that the
amino acid
residue at site 264, which is located above the
haem
cofactor,
play
s
an important role in both promoting the lactone carbene
-
transfer process and exerting control o
ver
the
enantio
-
determining
step
. W
e
then
prepared
five
variants based on
the full
-
length version of
L6
,
L6_F
L
,
with different mutations at site 264 and evaluated their performance
i
n
the
N
H insertion
reaction
(
Fig. 2c
). Interestingly, mutation of serine to
smaller amino acids
(A
or
G) led to comparable
activity but much
lower
selectivity
, suggesting that the hydrophilic side chain of serine may be involved
in controlling the enantioselectivity
.
However,
protic
residues
of larger size
(D, T
and C) at site 264 are
detrimental to both the yield and
stereocontrol
.
Fig.
2
|
Screen for enzymatic N
H insertion with
a
h
aem
protein
collection
and identif
ication of
A
264S as the key mutation for
achieving
high activity and
selectivity. a
, Initial screening was
4
performed with 40 h
aem
protein variants, which led to the discovery of
L7
(
in well C10
)
, which
originated from
a
previous lactone carbene C
H insertion project
.
b
, Rescreening of
the
lactone carbene
C
H insertion lineage. Variants with
the
A264S mutation were found to be excellent catalysts for N
H
insertion.
c
,
Mutagenesis
studies showed that
replacing
S
264
with
other amino acids led to
low
selectivities and
diminished
yields.
Residue
A264 in t
he
active site
of P411
variant
E10
(
PDB
ID
:5UCW)
is
highlighted.
L7_FL
and
L6_FL
are
L7
and
L6
restored
to their respective full
-
length
P411 enzymes.
We performed computational studies t
o understand the molecular basis of this enzymatic
transformation
and elucidate the role of S264 in promoting asymmetric carbene N
H insertion
.
Extensive molecular dynamics (MD) simulations considering the
haem
domain of
variant
L6
with the
lactone
-
carbene bound showed that the lactone
mainly
explores a single orientation in
the active site of
L6
. This is stabilized by
the
S
264 side chain that establishes persistent H
-
bond
ing
interactions with the
lactone ester group (Fig.
3a
).
Within this orientatio
n, only one face of the
electrophilic
carbene can
be
attacked by the nucleophilic amine substrate,
enantioselectively
yielding a reactive ylide intermediate
(s
ee
Fig. S14
for
the complete reaction mechanism studied by
density functional theory (DFT)
model
calculations
). Simulations performed with
variant
L7
describe similar behavior,
whereas
equivalent
modelling for
L5
variant show that, in the absence of
S
264, the lactone explores multiple conformations
(
see
Fig. S5
).
Furthermore
, H
-
bond
ing
interactions involving the carbonyl group of the carbene were
found to enhance its electrophilicity
(Fig. S15)
,
in line with what
was
disclosed in previous work
28,29
.
Consequently, in addition to controlling the carbene orientat
ion, S264 also contribut
es
to
i
ncreasing
the
reactivity
of the carbene species
.
We then modelled the approach of amine substrate
2a
to the carbene and the subsequent formation
of the ylide intermediate in the active site of
L6
.
Amine substrate binding in a near
-
attack conformation
for the
N
-
nucleophilic addition to
the lactone carbene is
stabilized
by hydrophobic interactions between
the substrate aromatic ring and
active site
hydrophobic residues
(L75, V328,
L
437, P329)
, as observed
from MD trajectories
(see Fig. 3b
, Fig. S6
and
S7
).
The
simulations also
show
that the H
-
bond between
the lactone ester group and
S264
is maintained
when the ylide is formed
.
In addition, according to
the
simulations,
only a
few
water molecules are present in the
active site
pocket
,
which
are precisely
funneled through
two
water channel
s, one
formed
by Y263 and T438 from the top face of the ylide
lactone ring
and
the other
one
guided
by the anionic
haem
carboxylates
near the ylide amine group
(Fig.
3c
, Fig. S8
and
S9
)
.
DFT
calculations
show
that, once dissociated from the iron, the ylide intermediate
can
rapidly
react
with these water molecules
to
achieve
the stereo
selective
proton rearrangement
in the
enzyme active site prior
to
product release
: a water molecule protonates
the ylide at the C position
from
the top face
(
pro
-
S
face
)
of the lactone ring
while a second
water
deprotonates the amine group (Fig
. 3d
and
Fig. S18
).
The low activation barriers calculated indicate that this proton transfer step can take place
immediately once the ylide
dissociates from the iron.
Collectively, the enantio
selective
formation of the
ylide and the precise placement of water molecules in th
e active site for proton transfer
enable the
enzyme to control the selectivity
of
thi
s
N
H insertion reaction.
5
Fig.
3
|
Computational modelling elucidates the o
rigins of
enantioselectivity
of
carbene
transfer
into
N
H
bonds
catalyzed by
P411
-
L6
.
a
, Representative snapshot from MD simulations describing
the conformations explored by the lactone carbene in
P411
-
L6
.
The
(N
Fe
C
1
C
2
) dihedral
angle
measured along the MD trajectory describes the relative orientation explored by the carbene (see SI for
additional replicas).
b
, Overlay of
three
representative snapshots obtained from constrained
-
MD
simulations exploring near attack conformations f
or the
N
-
nucleophilic attack of
2a
to the lactone
carbene in
L6
.
c
, Overlay of
three
representative snapshots from constrained
-
MD simulations exploring
L6
active site arrangement when ylide
2a
is formed. Displayed water molecules are drawn from 25
random structures across the 100 ns MD trajectory.
d
, DFT
-
optimized model transition states (TS) for
stereo
selective
ylide proton transfer. These models were built based on the arrangement of water
mo
lecules around the ylide intermediate in
the
L6
active site observed from MD simulations (see SI for
details)
.
K
ey distances are given in
Å
.
W
e
envisioned
th
at the
dual
-
function catalytic cavity of
L7_FL
could promote lactone
-
carbene
transfer to
other
amine substrates with
stereocontrol.
Indeed, u
sing
lactone diazo compound
1
as the
carbene precursor, a variety of amine nucleophiles
could undergo the desired
N
H insertion reaction
under our standard whole
-
cell reaction conditions (OD
600
= 30 in M9
-
N buffer)
, as summarized in
Fig.
4
.
Secondary anilines bearing an
N
-
alkyl
group
(
3
b
g
) were
well tolerated
in this transformation,
giving
good yields and excellent
enantioselectiv
it
ies.
N
-
heterocycles such as indoline (
3f
) and
tetrahydroquinoline (
3g
), which are structur
al motifs
common
ly found
in bioactive molecules, also
served as
competent
substrates.
S
teric hinderance
from
nitrogen substituents
did not
significantly
interfere with
enzyme
performance
, giving
>70% yields and
98% e.e.
(
3b
and
3c
).
L7
_
FL
also
6
displayed
high
activit
y
toward primary
aniline
s
(
3h
m
) bearing various substitution patterns on the
aromatic ring
,
including
a sterically hindered
substrate
with
two
ortho
-
substitu
ent
s
(
3m
)
. Arguably,
al
iphatic amines are
significantly
more
challenging
than anilines
for
asymmetric
N
H insertion
reactions due to
their increased Lewis basicity at
the
nitrogen
atom
17
. To our delight,
L7
_
F
L
w
as
able
to
accept
these amines
,
primary
(
3n
p
)
or secondary
(
3
q
)
,
for the desired transformation
s, furnishing
the corresponding
α
-
amino lactone products with
good
activities and
enantioselectivit
ies
.
Fig.
4
|
Enantioselective carbene N
H insertion of secondary, primary anilines and aliphatic
amines.
The experiments were performed using
E. coli
(OD
600
= 30) that expresse
d
the
L7_FL
enzyme
with 10 mM
diazo
1
and 10 mM amine
(
2b
q
)
at room temperature under anaerobic conditions. See
SI
for details.
To demonstrate the utility of
this
biocatalytic
platform
,
we
pushed the limit of
the
enzyme’s catalytic
capability
by using
much
low
er
enzyme
loading
s
.
Under
the
standard conditions using whole
-
cell
catalysts at
OD
600
= 30, the TTNs of
the
reactions are typically in the range of 1
,
000
2
,
000
(
Fig.
4
)
.
Gratifyingly
, l
owering
the
amount
of whole
-
cell
catalyst did not
result in
a
substantial drop in product
formation
,
giving T
T
Ns of
9
,
640
(
OD
600
=
5)
and 3
2
,
1
00
(
OD
600
=
1)
and
showing that
the
enzym
atic
platform
is particularly
robust
for
this
asymmetric
amination chemistry
(
Fig.
5
a
)
.
In addition
, the
enzymatic reactions
are
readily scalable
(
Fig
.
5
b
).
Using
N
-
methyl
-
p
-
toluidine
(
2d
) as
the amine
source
,
th
e
N
H insertion
reaction
was
performed at gram scale
, delivering product
3d
in 97%
isolat
ed
yield
and 96% e.e.
Products
3e
and
3l
were prepared at 1
-
mmol scale
,
with
crystal structures
determined by
X
-
ray crystallography.
Finally,
this
enzymatic
amination strategy
was
applied to
the
formal
synthesis