Engineered biosynthesis of β
-
alkyl tryptophan analogs
Christina E. Boville
+
, Remkes A. Scheele
+
, Philipp Koch, Sabine Brinkmann
-
Chen, and Andrew R.
Buller*, Frances H. Arnold*
Abstract:
Non
-
canonical amino acids (ncAAs) with dual
stereocenters at the α and β positions are valuable precursors to
natural products and therapeutics. Despite the potential applications
of such bioactive β
-
branched ncAAs, their availability is limited due to
the
inefficiency of the multi
-
step methods used to prepare them. Here
we report a stereoselective biocatalytic synthesis of β
-
branched
tryptophan analogs using an engineered variant of
Pyrococcus
furiosus
tryptophan synthase (
Pf
TrpB),
Pf
TrpB
7E6
.
Pf
TrpB
7E6
is
the first
biocatalyst to synthesize bulky β
-
branched tryptophan analogs in a
single step, with demonstrated access to 27 ncAAs. The molecular
basis for the efficient catalysis and broad substrate tolerance of
Pf
TrpB
7E6
was explored through X
-
ray crystallog
raphy and UV
-
visible
light spectroscopy, which revealed that a combination of active
-
site
and remote mutations increase the abundance and persistence of a
key reactive intermediate.
Pf
TrpB
7E6
provides an operationally simple
and environmentally benign plat
form for preparation of β
-
branched
tryptophan building blocks.
Amino acids are nature’s premier synthetic building blocks for
bioactive molecules. Alongside the standard proteinogenic amino
acids are diverse non
-
canonical amino acids (ncAAs) that are
struc
turally similar but are not ribosomally incorporated into
proteins. Due to the presence of functional groups that confer
novel chemical and biological properties,
[1]
ncAAs can be found in
natural products and 12% of the 200 top
-
grossing
pharmaceuticals.
[2,3]
Of interest are β
-
branched ncAAs, which
possess a chiral center at the β
-
position
in addition to the standard
chirality at the α
-
position of an amino
acid
(
Figure 1a)
. The two
adjacent stereocenters impose conformational constraints that
affect the biochemical properties of both the amino acids
themselves and the molecules they compose.
[4
–
7]
These
properties make β
-
branched ncAAs frequent components of
useful natural products, biochemical probes, and therapeutics
(Figure 1b)
.
[8
–
13]
Despite their broad utility, most β
-
branched
ncAAs are not readily available due to the challenge of forming
two adjacent stereocenters while
tolerating the reactive functional
groups present in amino acids.
[14
–
18]
For example, traditional
organic synthesis of (2
S
, 3
S
)
-
β
-
methyltryptophan (β
-
MeTrp)
requires
multiple steps that incorporate protecting groups,
hazardous reagents, and expensive metal catalysts.
[19,20]
To take
full advanta
ge of these bioactive molecules, an improved
methodology is needed to synthesize β
-
branched ncAAs.
α
α
α
β
β
β
Figure 1.
Representative β
-
branched amino acids. (a) Examples of β
-
branched
ncAAs. (b) Examples of products derived from
β
-
branched tryptophan analogs
(red).
Enzymes offer an efficient and sustainable alternative to
chemical synthesis and are routinely used to generate
enantiopure amino acids from simple materials without the need
for protecting groups.
[21]
Although several classes of enzymes
have been employed in this pursuit, those using the pyridoxal
phosphate cofactor (PLP, vitamin B6) are among the most
prominent.
[22]
The most common biocatalytic route to an amino
acid requires a fully assembled carbon skeleton and a PLP
-
dependent transaminase which is used to set the
stereochemistry. However, as with traditional organic
methodologies, the enzymatic synt
hesis of β
-
branched ncAAs is
often confounded by the presence of a second stereocenter. The
capacity to incorporate biocatalytic C
–
C bond
-
forming steps
en
route
to diverse β
-
branched ncAAs would therefore be a powerful
synthetic tool.
Few β
-
branched ncAA
synthases have been reported, and
even more rare are enzymes that produce branches larger than a
methyl group. We previously engineered the β
-
subunit of the PLP
-
dependent enzyme tryptophan synthase from the
thermophilic
archaeon
Pyrococcus furiosus
(
Pf
Trp
B) as a stand
-
alone ncAA
synthase able to generate tryptophan (Trp) analogs from serine
(Ser) and the corresponding substituted indole
(Figure 2a)
.
[23
–
25]
Further engineering of
Pf
TrpB for imp
roved C
–
C bond formation
with indole analogs and threonine (Thr) led to
Pf
TrpB
2B9
(eight
mutations from wild
-
type
Pf
TrpB), which exhibited a >1,000
-
fold
improvement in (2
S
, 3
S
)
-
β
-
methyltryptophan (β
-
MeTrp)
production relative to wild type
(Figure 2b)
.
[26,27]
While the
reactive amino
-
acrylate intermediate (E(A
-
A))
(Figure 3a)
readily
forms with Thr, mechanistic analysis showed that competing
hydrolysis of (E(A
-
A))
resulted in abortive deaminat
ion that
consumed the amino acid substrate
(Figure 3b),
[28,29]
limiting the
enzyme’s yield (typically < 50%) with a single equivalent of Thr.
Further,
Pf
TrpB
2B9
accepted only Ser and Thr as substrates since
larger β
-
alkyl substrates were unable to efficiently form E(A
-
A).
[a]
Dr.
C. E. Boville, R. A. Scheele, P. Koch, Dr. S. Brinkmann
-
Chen,
Prof. F. H. Arnold
Division of Chemistry and Chemical Engineering 210
-
41, California
Institute of Technology, 1200 East California Boulevard, Pasadena,
California 91125, United State
s
Email: frances@cheme.caltech.edu
[b]
Prof. A. R. Buller
Department of Chemistry, University of Wisconsin, 1101 University
Avenue, Madison, WI 53706, United States
Email: arbuller@wisc.edu
[
+]
These authors contributed equally to this work
.
Figure 2.
Synthesis of Trp and Trp analogs by
Pf
TrpB.
To surmount these challenges, we sought to identify
mutations that would facilitate formation of E(A
-
A) with the more
challenging (2
S
, 3
R
)
-
β
-
ethylserine (β
-
EtSer) and (2
S,
3
R
)
-
β
-
propylserine (β
-
PrSer)
substrates
while
simultaneously
decreasing E(A
-
A) hydrolysis
(Figure 2c)
. The latter is essential,
as incre
asingly bulky alkyl chains are thought to hinder
nucleophilic attack. Increased E(A
-
A) persistence will allow more
time for the intrinsically slower addition reaction to occur while
reducing the amount of starting material lost to competing
hydrolysis
(Fig
ure 3b)
.
We chose
Pf
TrpB
2B9
as our engineering starting point to
increase production of β
-
EtTrp. While an active catalyst with Thr,
Pf
TrpB
2B9
was sluggish with β
-
EtSer (80 total turnovers, TTN) and
gave too little signal for high
-
throughput screening.
[23]
We
speculated th
at active
-
site mutations would promote the formation
of E(A
-
A) with larger, sterically demanding β
-
substituents and
used a structure
-
guided approach to improve activity with β
-
EtSer. Modeling β
-
EtSer into the
Pf
TrpB
2B9
active site as E(A
-
A)
(PDB: 5VM5)
[30]
suggested a steric clash with L161
(Figure 4a)
.
Hypothesizing that this constraint could be reduced by mutating
L161 to a residue with a smaller side chain
(Figure 4b)
, we
expressed and assayed variants
Pf
TrpB
2B9
L161V, L161A, and
L161G. We found that L161V and L161A increased the TTN 14
-
fold and 10
-
fold, respectively, whereas L161G decreased activity
by a factor of 2.6
(Figure 4c)
. As our long
-
term interest
is to
produce a catalyst that accommodates a wider range of β
-
alkyl
chains, we selected
Pf
TrpB
2B9
L161A as the parent enzyme for
directed evolution, with the rationale that the smaller sidechain of
alanine would minimize steric clashes with bulkier substr
ates.
We then introduced random mutations into the
Pf
TrpB
2B9
L161A gene and screened for enhanced β
-
EtTrp synthesis
(Table
1)
at 290 nm under saturating substrate conditions.
[23]
Screening
made u
se of starting materials containing a mixture of
diastereomers, however only the (2
S
,3
R
) diastereomer
underwent a productive reaction. High
-
throughput screening of
352 variants yielded
Pf
TrpB
0E3
(L91P), which displayed a 43
-
fold
increase in TTN for β
-
EtTrp
(Figure 4d)
.
Pf
TrpB
0E3
was then used
as the parent for a second round of random mutagenesis, yielding
variant
Pf
TrpB
8C8
(V173E), which improved β
-
EtTrp yields by 54
-
fold relative to
Pf
TrpB
2B9
. At this juncture, a third round of random
mutagenesis failed
to yield further improvements after screening
880 variants. Although the accumulated mutations increased
activity, we speculated that further improvements were hindered
by deleterious mutations that reduced enzyme stability.
[31]
We
therefore recombined mutations in TrpB
8C8
,
allowing a 50%
chance for each residue to retain the mutation or revert to wild
type. Recombination included all residues except those which
were crucial for starting activity with Ser (T2
92S), Thr (F95L), and
β
-
EtSer (L161A and L91P)
(Table S1)
. Recombination also
included F274L, which was previously identified as an activating
mutation.
[23]
Recombined variants were assayed for β
-
EtTrp
production at 290 nm, which revealed that I68V and T321A were
non
-
essential
, but that F274L was beneficial, yielding variant
Pf
TrpB
7E6
. Though
Pf
TrpB
7E6
did not show improved stability
(Table S2)
, recombination did enhance activity, with a 58
-
fold
Figure 3.
The putative catalytic cycle for
Pf
TrpB synthesizing
β
-
MeTrp. (a)
Catalysis initiates as E(Ain) with the mobile COMM domain predominantly in the
open conformation (blue). With the addition of Thr, the COMM domain
undergoes rigid body motion, transitioning to a partially closed position
through
E(Aex
1
) (red) followed by full closure with formation of the reactive E(A
-
A)
intermediate (green). E(A
-
A) is then attacked by indole and undergoes an
addition reaction to form
β
-
MeTrp. (b) E(A
-
A) may also undergo a kinetically
competing hydrolysis
reaction to generate
α
-
keto acids, observable at 320 nm.
This deamination reaction consumes an equivalent of the amino acid substrate.
improvement relative to
Pf
TrpB
2B9
(Figure 4d)
. An additional
round of recombination sampled other previously identified
activating mutations (Q38R, M139L, N166D, S335N) and allowed
for reversion of L91P. This process produced a variant (
Pf
TrpB
2G8
,
see
Tabl
e 1
) that
lacked the L91P mutation and had only slightly
lower activity than
Pf
TrpB
7E6
. Although subsequent work showed
that
Pf
TrpB
2G8
is also a proficient enzyme (vide infra), the parent
Pf
TrpB
7E6
was selected for mechanistic characterization as it is a
comparatively simple catalyst with excellent activity and more
amenable to crystallization.
Figure 4.
Engineering
Pf
TrpB for β
-
EtTrp synthesis. (a) β
-
EtSer as E(A
-
A)
(yellow) modeled in the
Pf
TrpB
2B9
(PDB: 5VM5, gray) active site. Spheres
represent the Van der Waals radii and highlight a clash with L161 (green). (b)
As in (a), but with the mutation L161A shown (purple). (c) β
-
EtTrp production by
Pf
TrpB
2B9
with L161V, L161A, or L161G mutations. (d)
β
-
EtTrp production by
engineered
Pf
TrpB variants. Bars represent the average of all data points, with
individual reactions shown as circles. At minimum, reactions were performed in
duplicate.
We sought to identify which newly evolved properties of
Pf
TrpB e
nabled increased TTNs with challenging β
-
branched
substrates. As described above, the activity and substrate scope
of the parent enzyme,
Pf
TrpB
2B9
,
were limited by low steady
-
state
population
(abundance)
and
subsequent
breakdown
(persistence) of the reacti
ve E(A
-
A) intermediate.
[27]
To assess
the abundance of E(A
-
A), we capitalized on the intrinsic
spectroscopic properties of the PLP cofactor to visualize the
steady
-
state distribution of intermediates throughout the catalytic
cycle
(Figure 3a)
.
[32]
With the addition of β
-
EtSer to
Pf
TrpB
7E6
, the
internal aldimine peak (E(Ain), 412 nm) decreased and E(A
-
A)
(350 nm) became the major species
(Figure 5a)
. This is a notable
change, as when β
-
EtSer was added to
Pf
TrpB
2B9
, E(Aex
1
)
accumulated and no E(A
-
A) was observed
(Fig 5a)
. To assess
the persistence of E(A
-
A), we assayed the deamination rate and
coupling efficiency of
Pf
TrpB
7E6
. In the presence of both Thr and
β
-
EtSer,
Pf
TrpB
7E6
displayed up to a 4
-
fold decrease in the
deamination reaction relative to
Pf
TrpB
2B9
(Table S3)
. We then
probed the enzyme’s coupling efficiency under reaction
conditions with high catalyst loading and equimolar substrate
equivalents, where product formatio
n is limited only by the
consumption of starting material through the competing
deamination reaction. We observed an increase in product
formation from 5% with
Pf
TrpB
2B9
to 96% with
Pf
TrpB
7E6
when
β
-
EtSer was the substrate
(Figure 5b)
. Collectively, these
data
indicate that increased product formation was achieved by
incorporating mutations that facilitate the formation of E(A
-
A) and
increase its lifetime in the active site.
During directed evolution,
Pf
TrpB was altered by the
introduction of nine mutation
s. Although
Pf
TrpB
7E6
has only a
single mutation in the active site
(Figure S1)
, mutations governing
enzyme activity are scattered throughout the protein.
[23,33]
Remote
mutations may be affecting the enzyme’s conformational
dynamics, which have been previously shown to be linked to the
catalytic cycle of
Pf
TrpB
(Figure 3a)
.
[30,33]
In its resting state,
Pf
TrpB binds PLP via the catalytic lysine (K82) as E(Ain) with the
mobile communication (COMM) domain in a predomin
antly open
conformation. Addition of an amino acid
substrate induces
formation of the external aldimine (E(Aex
1
)),
which is
accompanied by partial closure of the COMM domain.
Dehydration to form the electrophilic E(A
-
A) species occurs when
TrpB populates
a fully closed conformation, where it remains until
product is formed.
[28,29]
To examine the state of the
Pf
TrpB
7E6
active site and its connection to the COMM domain
conformational cycling, we determined the X
-
ray crystal
structures of
Pf
TrpB
7E6
in the E(Ain) state as
well as with β
-
EtSer
bound in the active site as E(A
-
A).
Figure 5.
Directed evolution stabilizes E(A
-
A) and improves coupling efficiency.
(a) The steady
-
state population of
Pf
TrpB as determined by UV
-
visible light
spectroscopy. In the absence of substrate, the predominant population of
Pf
TrpB
7E6
(black) is E(Ain). β
-
EtSer
-
bound
Pf
TrpB
2B9
(orange) accumulates
E(Aex
1
) and
Pf
TrpB
7E6
(green) forms E(A
-
A). All spectra are normalized
to the
absorbance value of E(Ain) at 412 nm. (b) Variant coupling efficiency with β
-
EtSer. Bars represent the average of all data points, with individual reactions
shown as circles. At minimum, reactions were performed in duplicate.
Earlier
Pf
TrpB variants, including
Pf
TrpB
2B9
, were nearly
identical to wild
-
type
Pf
TrpB (PDB: 5DVZ) in the open state. Here,
the 2.26
-
Å structure of
Pf
TrpB
7E6
(PDB: 6CUV) shows distinct
preorganization toward a more closed conformation. Specifically,
Table 1.
Engineering
Pf
TrpB
through directed evolution for improved β
-
EtTrp production. Engineering began with
Pf
TrpB
2B9
(
Pf
TrpB
I16V, E17G,
I68V, F95L, F274S, T292S, T321A, and V384A) with 80 TTN. All reactions
were performed in at least duplicate with 0.1% catalyst loading for 24 hours
at 75 °C.
Variant
Mutations
Added
Mutations
Removed
Average
TTN
[a]
Pf
TrpB
2B9
L161A
L161A
N/A
800
[b]
Pf
TrpB
0E3
L91P
N/A
3400
[b]
Pf
TrpB
8C8
V173E
N/A
4300
[c]
Pf
TrpB
7E6
F274L
I68V, T321A
4600
[c]
Pf
TrpB
2G8
M139L, N166D, S335N
L91P
3800
[a] Site
-
directed mutagenesis. [b] Random mutagenesis. [c] Recombination.
in half of the
protomers, the COMM domain has shifted into a
distinct partially
-
closed conformation that was previously
associated with substrate binding
(Figure 6a)
. While many
residues may contribute to the stabilization of this state, we
hypothesize that the mutation
L91P destabilizes open states; this
residue lies on an α
-
helix immediately prior to the COMM domain
in the sequence and causes a kink in the helix that shifts the
structure toward more closed states
(Figure 6b)
.
We next soaked
Pf
TrpB
7E6
with β
-
EtSer and o
btained a 1.75
-
Å structure with β
-
EtSer bound as E(A
-
A) in two protomers (PDB:
6CUZ). As expected, the COMM domain underwent rigid
-
body
motion to the closed conformation
(Figure 6a)
where the steric
complementarity between the longer β
-
alkyl chain and L161
A
becomes apparent. Notably, the L161A mutation does not appear
to induce significant alterations elsewhere in the active site
(Figure 6c).
When indole is modeled into the active site, there is
space to accommodate even longer β
-
branched substituents as
we
ll as a range of indole nucleophiles
(Figure 6d)
.
Figure 6.
Substrate binding and conformational changes in
Pf
TrpB. (a) The
COMM domain of
Pf
TrpB undergoes rigid body motion that is linked to the
catalytic cycle. In the absence of substrate, wild
-
type
Pf
TrpB (PDB: 5DVZ, gray)
is in the open conformation, while
Pf
TrpB
7E6
(PDB: 6CUV, green) assumes a
partially closed conformation. When
β
-
EtSer is bound to
Pf
TrpB
7E6
as E(A
-
A)
(PDB: 6CUV, orange), the COMM domain undergoes a rigid body shift to a
closed confo
rmation. (b) The mutation L91P introduces a kink in the α
-
helix
adjacent to the COMM domain. (c)
β
-
EtSer bound to
Pf
TrpB
7E6
as E(A
-
A) is
shown
with F
o
-
F
c
map contoured at 2.0
σ
(green). The gamma carbon of the
amino
-
acrylate is not well resolved.
Hydrogen bonds are shown as red dashes.
(d) Indole (yellow) modeled in the active site of
Pf
TrpB
7E6
with
β
-
EtSer as E(A
-
A). The green dash represents the bond
-
forming atoms in indole and
β
-
EtSer.
As our goal was to evolve a versatile β
-
branched ncAA
synth
ase, we next explored the substrate scope of
Pf
TrpB
7E6
. We
hypothesized that, if improvements in activity came through
increased stability of E(A
-
A), the same mutations should increase
activity with multiple amino acid substrates. Indeed, we found that
although we screened for β
-
EtTrp synthesis, the TTN for β
-
MeTrp
and (2
S,
3
S
)
-
β
-
propyltryptophan (β
-
PrTrp) synthesis were
simultaneously improved 3.6
-
fold and 36
-
fold, respectively
(Figure 7a)
. Consistent with our previous observations, directed
evolution i
mproved the enzyme’s coupling efficiency
(Figure 7b)
and amino
-
acrylate persistence
(Figure 7c
-
d)
with all three acid
substrates.
Next, we revisited our earlier hypothesis that the
L161A mutation would be more beneficial than L161V by reducing
steric clash
es with larger substrates. We observed that although
Pf
TrpB
7E6
L161V is viable for synthesis of β
-
MeTrp and β
-
EtTrp,
the TTN for β
-
PrTrp formation was reduced 5
-
fold
(Figure S2a)
.
In addition,
Pf
TrpB
7E6
retained the robust Trp activity that is the
hallmark
of the wild
-
type enzyme
(Figure S2b)
, demonstrating
that the L161A mutation was successful in accommodating
bulkier substrates, allowing catalysis with four different amino acid
substrates.
Figure 7.
Pf
TrpB engineering grants access to a range of β
-
branched
tryptophan analogs. (a) TTN of
Pf
TrpB for β
-
MeTrp (orange), β
-
EtTrp (blue),
and β
-
PrTrp (green). (b) Variant coupling efficiency with Thr (orange), β
-
EtSer
(blue), and β
-
PrSer (green). Bars represent the average of all data points, with
individual reactions shown as circles. At minimum, r
eactions were performed in
duplicate. (c) The steady
-
state population of
Pf
TrpB with Thr as determined by
UV
-
visible light spectroscopy. In the absence of substrate, the predominant
population of
Pf
TrpB
7E6
(black) is E(Ain). With the addition of Thr to
Pf
T
rpB
2B9
(orange) has a mixed population of E(Aex
1
) and E(A
-
A), while
Pf
TrpB
7E6
(green)
is predominantly E(A
-
A). (d) β
-
PrSer
-
bound
Pf
TrpB
2B9
(orange) remains as
E(Ain) while
Pf
TrpB
7E6
(green) predominantly forms E(A
-
A).
However, activity was not observed wit
h all β
-
alkyl
substrates and reactions with (2
S
)
-
β
-
isopropylserine (β
-
iPrSer)
showed only trace activity. To understand why catalysis did not
proceed with this bulkier sidechain, we soaked β
-
iPrSer into
Pf
TrpB
7E6
crystals and obtained a 1.77
-
Å structure (PDB: 6CUT),
which shows the catalytically unreactive (2
S
, 3
S
) diastereomer of
β
-
iPrSer bound as E(Aex
1
)
(Figure S3)
. Though (2
S
, 3
S
)
-
β
-
iPrSer
can form E(Aex
1
), dehydration across the C
α
–
C
β
bond requires a
rotameri
c shift of the side chain that we hypothesize is hindered
by steric interactions with an adjacent loop.
[34]
Further work is
needed to understand
whether the poor activity of
Pf
TrpB
7E6
with
(2
S,
3
R
)
-
β
-
iPrSer reflects inhibition by an isomeric analog,
increased allylic strain of the amino
-
acrylate, or a combination of
effects.
In addition to acting on multiple amino acid substrates, we
hypothesized that
Pf
TrpB
7E6
would retain the wild
-
type enzyme’s
breadth of reactivity with indole analogs.
[23
–
25]
We performed
analytical biotransformations with 11 representative nucleophiles
with three β
-
branched amino acid substrates,
yielding 27
tryptophan analogs, 20 of which are previously unreported
(Table
2)
. Each reaction was analyzed by liquid
-
chrom
atography/
mass
spectrometry (LCMS) and TTN were
calculated by comparing
product and substrate absorption at the isosbestic wavelength
(Table S4)
. Happily, we found that substituted indole analogs