Journal of the American Chemical Society
Communication
interaction that is also observed in the Ser E(Aex1). Modeling of a
hypothetical Thr E(Aex1) that maintains the Asp300 hydrogen
bond reveals a strong steric clash between the Thr methyl group
and the backbone carbonyl of Gly298 (Figure S3), accounting for
the lack of E(Aex1) observed by UV−vis spectroscopy.
UV−vis spectroscopy was used to probe the basis for
enhanced activity through directed evolution. Addition of 20
mM Thr to the engineered proteins from each generation
resulted in a clear trend (Figures 4b and S4): the 412 nm peak
decreased while a new absorbance band at 350 nm with a broad
shoulder out to 550 nm appeared, consistent with accumulation
of the electrophilic aminoacrylate species E(A-A). A similar
phenomenon was observed during directed evolution for
independent TrpB function;8 however, comparison of the Ser-
and Thr-bound Pf TrpB2B9 spectra indicates that E(A-A) with
excess of Thr to nucleophile (Chart 1). Reactions were run to
intermediate yield to determine the TTN with a given
nucleophile. The identity of the products was established with
a separate preparative-scale reaction using 100 mM nucleophile,
1.0 M Thr, and 0.02−0.13 mol % of Pf TrpB2B9 catalyst. We
observed good reactivity with the 2-methyl and 6-methylindole
substrates, but decreased TTNs compared to indole demonstrate
that the active site is sensitive to steric perturbations. To probe
the role of electronic effects on the C−C bond-forming step, we
tested activity with 4-fluoro- and 5-fluoroindole, which are more
closely isosteric with indole but have decreased electron density
in the π-system. We observed product formation with each
substrate and 3.4-fold lower TTN with the 4-fluoro substrate,
which is more electron-deficient at C-3 than the 5-fluoro
substituent. Increased steric constraints of the active site were
again clear, as we did not observe any activity with the 5-chloro-,
5-bromo-, and 6-hydroxyindoles, which do undergo reaction
with Ser.8
A productive reaction with 220 turnovers was observed with 7-
azaindole, which is a substantially weaker nucleophile than
indole. Interestingly, a second product was detected in the
reaction with 7-azaindole that is consistent with N-alkylation.
This regioselectivity is well-known with indazole, which we
found reacts exclusively to form an N-alkylated product.
Surprisingly, we did not observe product formation with
indoline, which is a stronger nucleophile than both indazole
and indole and reacts significantly faster in the reaction with Ser.
Future studies will be needed to ascertain the origin of this effect.
Lastly, we tested for S-alkylation using thiophenol and measured
1300 turnovers.
In conclusion, we have discovered a new, non-natural
enzymatic route for the production of (2S,3S)-β-MeTrp. This
activity lies in the β-subunit of TrpS, which was subsequently
engineered for increased activity with Thr as the amino acid
donor. Development of the resultant catalyst, Pf TrpB2B9, was
greatly facilitated by previous efforts to replace the native two-
enzyme system with a single stand-alone catalyst.8 The
engineered enzymes have high thermal stability and expression
in E. coli, enhancing their utility as practical catalysts. This
enzymatic route to β-MeTrp is dramatically shorter than
previous synthetic routes and also the native 3-enzyme pathway
to this natural metabolite.9,11 This study highlights the ability of
engineered biocatalysts to efficiently produce complicated
molecules from simple precursors and offers a simple and
expandable route for the production of β-methyl ncAA
analogues.
The engineered Pf TrpB2B9 enzyme has several desirable
features as a catalyst for β-MeTrp production. It is robustly
expressed in E. coli and can be prepared in a moderately pure
form as a heat-treated lysate (Figure S6), and its thermal stability
permits high reaction temperatures, routinely up to 75 °C, which
greatly increases the solubility of hydrophobic substrates.
However, in reactions with 1 equiv each of indole and Thr, we
observed only 44% conversion to product, corresponding to
2220 turnovers (Figure S7). A critical clue explaining this low
conversion came from UV−vis spectroscopy, which revealed that
addition of Thr to Pf TrpB2B9 results in a time-dependent
increase in absorbance at 320 nm, while the remainder of the
spectrum remains constant (Figure S8). We attribute this to α-
ketobutyrate production from the well-described deamination
reaction that results when a nucleophile does not add into C-β,
and E(Ain) is reformed through transimination.7 The precise
mechanism of deamination is unknown, but the net effect is an
abortive reaction wherein the aminoacrylate hydrolyzes to form
α-ketobutyrate and ammonium. We therefore added additional
equivalents of Thr to the β-substitution reaction with indole
(Figure S7), which enabled >99% conversion of indole to β-
MeTrp and up to 8200 total turnovers to the desired product
(Chart 1) with >99% ee and de.
With these reaction conditions, we next characterized the
substrate scope of the reaction with Thr. Previously, we screened
a small panel of indole-like nucleophiles for reaction with
PfTrpB0B2 and Ser.8 We expanded on this panel of nucleophiles
and tested for activity with PfTrpB2B9 using a 10-fold molar
Chart 1. Production of β-MeTrp Analogues
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge on the
■
S
Additional figures, experimental details, and character-
AUTHOR INFORMATION
■
Corresponding Authors
Author Contributions
a
Total turnover numbers are given in parentheses. Reactions
†M.H. and P.vR. contributed equally.
b
conducted in triplicate. Circles indicate the site of alkylation. Single
reaction conducted at 25 °C with 12.5 mM DTT. See Supporting
Notes
The authors declare no competing financial interest.
C
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX