Angewandte
Research Articles
Chemie
efficiency, which may have been due to poor recognition by
HER2 Fab A121p-AzPhC mutant was recombinantly ex-
the PhSeRS enzyme; it might be possible to solve this
problem by evolving a more efficient and specific synthetase
mutant.
pressed in culture medium supplemented with 1 mM 22. The
yield of the periplasmically expressed anti-HER2 mutant was
1.5 mgLÀ1, whereas the yield of the wild-type protein was
2 mgLÀ1. Incorporation of p-AzPhC was confirmed by high-
resolution MS (Figure S9). Note that we observed reduction
of the azide to the corresponding amine, which is a known
problem for genetic incorporation of p-azido-l-phenylala-
nine.[41] When the resulting mutant protein was treated with
DBCO-mPEG20K in PBS buffer at 378C, the conjugation
efficiency was similar to that for a mutant containing p-azido-
l-phenylalanine (Figure 6b). To demonstrate the utility of
our approach, we conjugated the Fab-p-AzPhC mutant with
DBCO-Cy5 and confirmed the reaction by high-resolution
MS (Figure S10). The resulting conjugate successfully stained
HER2-positive cells, as indicated by flow cytometry (Fig-
ure 6c) and confocal fluorescence imaging. In contrast, the
conjugate did not stain HER2-negative cells (Figure 6d),
indicating that site-specific labeling with p-AzPhC did not
influence the activity of the conjugated protein. Similarly, the
bioorthogonal reactivity of p-AcPhC was confirmed by
conjugating the mutant protein with alkoxyamine-labeled
PEG molecules; the reactivity of the resulting conjugate was
essentially the same as that of a conjugate with a p-acetyl-l-
phenylalanine residue (Figure S11b). Our findings indicate
that mutant antibodies generated by the strategy described
herein can be expected to be useful for diagnostic and
therapeutic applications.
Next, we attempted to evolve a better aaRS mutant for
the desired ncAAs. To accomplish this, we screened a synthe-
tase library by simply feeding cells with the corresponding
small-molecule precursor. We were particularly interested in
thiol 22, a precursor for the biosynthesis of S-(4-azidophenyl)-
l-cysteine (p-AzPhC), which can be expected to be highly
useful for site-specific modifications of therapeutic proteins.
To evolve an efficient synthetase for p-AzPhC recognition,
we generated a library of MjTyrRS mutants by saturation
mutagenesis at residues Leu65, Phe108, Gln109, and Asp158
(with fixed Tyr32Trp, His70Gly, Leu162Glu and Asp286Arg
mutations) (Figure 5a); the codons of these positions were
randomized to NNK (N = A, T, G, or C; K = Tor G) to obtain
a synthetase library consisting of 1 106 diverse mutants. A
series of negative and positive selections were performed in
E. coli DH10B cells as previously described,[38] except that
instead of providing amino acids, we directly added the small-
molecule precursors to the screening plate (Figure 5b,c).
Although providing CysM facilitated the screening process,
we found that the well-established, routine screening proce-
dure worked well with the small-molecule precursors because
biosynthesis with the endogenous enzymes was sufficient.
Screening revealed
a novel synthetase, designated
pAzPhCRS, bearing Glu65, Thr108, Glu109, and Asp158
mutations (Figure 5d). This synthetase was cloned into pUltra
vector, and the amber suppression efficiency was evaluated by
means of a sfGFP Y151 amber mutant expression assay. As Conclusion
shown in Figure 5e, the normalized fluorescence for a sfGFP
mutant with pAzPhCRS was approximately 2 times that with
polyspecific PhSeRS in the presence of thiol 22 (1 mM). The
yields of purified mutant proteins were 39 and 70 mgLÀ1
(Figure 5i) for expression with PhSeRS and pAzPhCRS,
respectively. Because of the high expression efficiency of
pAzPhCRS, we next attempted to express double or triple
mutants from biosynthesized p-AzPhC. In addition to Y151,
we also selected G51 and D133 for amber codon mutations.
As shown in Figure 5 f, the yields of purified sfGFP containing
one, two, and three biosynthesized p-AzPhC residues were 70,
15, and 1 mgLÀ1, respectively (Figure 5i). Correct incorpo-
ration was confirmed by high-resolution MS (Figure 5g,h).
The reduced yields of protein with multiple mutations may
have resulted from competition with a release factor and
could in theory be improved by knocking out RF1[39] or using
an engineered strain.[40] These results suggest that the
structural and functional diversity of aromatic thiol analogues
could be further expanded by screening for suitable aaRS
mutants.
To our knowledge, the amino acids p-AzPhC and S-(4-
acetylphenyl)-L-cysteine (p-AcPhC), which were derived
from 22 and 7, have not previously been reported. To
characterize their reactivities (Figures 6a and S11a), we
genetically encoded p-AzPhC into the antigen-binding frag-
ment (Fab) of the anti-HER2 (human epidermal growth
factor receptor 2) antibody, which is clinically effective for
treating HER2-over-expressing breast cancer. The anti-
We have developed a highly efficient and versatile
strategy for coupling ncAA biosynthesis and genetic encod-
ing. By engineering the cysteine biosynthesis pathways and
aaRS mutants for GCE, we biosynthesized nearly 50 novel
ncAAs from economical commercially available or syntheti-
cally accessible aromatic thiol precursors and genetically
encoded the ncAAs into recombinant proteins in E. coli. The
aromatic thiols had a diverse array of structures and
functionalities, including bioorthogonal reactive groups such
as anilines, aromatic ketones, and aromatic azides. We showed
that the necessary aaRS mutants could be selected by using
the precursor directly, which eliminated the need for chemical
synthesis of the corresponding ncAAs. Thus, this biosynthetic
strategy can be expected to markedly reduce the cost and
synthetic burden of producing ncAA-containing recombinant
proteins for routine laboratory usage. Finally, we showed that
aromatic azides and aromatic ketones could be efficiently
incorporated into therapeutic proteins from simple precursors
and that their bioorthogonal reactivities were comparable to
those of p-azido-l-phenylalanine and p-acetyl-l-phenylala-
nine, two of the most commonly used ncAAs for site-specific
protein conjugation. Site-specifically labeled antibodies pro-
duced in this way can be expected to be highly useful for
developing next-generation, homogeneously modified pro-
tein therapeutics.
Angew. Chem. Int. Ed. 2021, 60, 10040 –10048
ꢀ 2021 Wiley-VCH GmbH