Genetic Encoding and Enzymatic Deprotection of a Latent Thiol Side Chain to Enable New Protein Bioconjugation Applications

Article abstract of DOI:10.1002/anie.202102343

The thiol group of the cysteine side chain is arguably the most versatile chemical handle in proteins. To expand the scope of established and commercially available thiol bioconjugation reagents, we genetically encoded a second such functional moiety in f

Full text of DOI:10.1002/anie.202102343

A Journal of the Gesellschaft Deutscher Chemiker
Angewandte
International Edition Chemie
www.angewandte.org
Accepted Article
Title: Genetic encoding and enzymatic deprotection of a latent thiol
side chain to enable new protein bioconjugation applications
Authors: Marie Reille-Seroussi, Pascal Meyer-Ahrens, Annika Aust,
Anna-Lena Feldberg, and Henning D. Mootz
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Angew. Chem. Int. Ed. 10.1002/anie.202102343
Link to VoR: https://doi.org/10.1002/anie.202102343

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
Genetic encoding and enzymatic deprotection of a latent thiol
side chain to enable new protein bioconjugation applications
Marie Reille-Seroussi,[a]# Pascal Meyer-Ahrens,[a]# Annika Aust,[a]# Anna-Lena Feldberg, Henning D.
[a]
Mootz*[a]
[a]
Dr. M. Reille-Seroussi, M.Sc. P. Meyer-Ahrens, M.Sc. A. Aust, M.Sc. A.-L. Feldberg, Prof. Dr. H. D. Mootz
Institute of Biochemistry, University of Muenster,
Corrensstraße 36, 48149 Münster
E-mail: Henning.Mootz@uni-muenster.de, # = equal contribution
Supporting information for this article is given via a link at the end of the document.((Please delete this text if not appropriate))
Abstract: The thiol group of the cysteine side chain is arguably the
most versatile chemical handle in proteins. To expand the scope of
established and commercially available thiol bioconjugation reagents,
we genetically encoded a second such functional moiety in form of a
latent thiol group that can be unmasked under mild physiological
conditions. Phenylacetamidomethyl (Phacm) protected homocysteine
can also suffer from premature release or cellular conversion into
inactive forms.[8] Chemical deprotection of the thiazolidine-
protected cysteine moiety in thiaprolyl-lysine[5d] and the allyl-
protected selenocysteine analog[5e] require high concentrations of
methoxyamine at low pH and a palladium catalyst, respectively,
that will be problematic for many proteins.
(
HcP) was incorporated and its latent thiol group unmasked on
Thus, the genetic incorporation into proteins of a latent thiol or
selenol group, that can be efficiently deprotected under native,
mild and non-destructive conditions, has not been reported so far.
We envisioned that such a building block would represent a
formidable addition to the protein chemist’s toolbox to enable
novel chemical manipulation schemes with chemo- and
regioselectivity in presence of native cysteines and disulfides, and
that can be performed under mild conditions.
purified proteins using pencillin G acylase (PGA). The enzymatic
deprotection depends on steric accessibility, but can occur efficiently
within minutes on exposed positions in flexible sequences. The freshly
liberated thiol group does not require treatment with reducing agents.
We demonstrate the potential of this approach for protein modification
with conceptually new schemes for regioselective dual labeling, thiol
bioconjugation in presence of a preserved disulfide bond and
formation of a novel intramolecular thioether crosslink.
The cysteine thiol group has several unique properties that
establish its outstanding role in the set of the ubiquitous 20
proteinogenic amino acids.[1] For example, it serves as catalytic
residue in various enzymes, it forms a covalent disulfide linkage
in protein folding, it can be found both in the hydrophobic core and
exposed to the solvent owing to its amphipathic character, and it
is the target for many posttranslational modifications. The
underlying unique nucleophilic and redox properties are also the
Scheme 1. Concept of HcP incorporation and enzymatic deprotection. POI =
protein of interest.
[
2]
To address this challenge, we aimed to develop an
enzymatic deprotection scheme to liberate a latent thiol in a novel
non-canonical amino acid that is incorporated into proteins by the
genetic code expansion technology.[ Here we report on the
incorporation of phenylacetamidomethyl (Phacm) protected
homocysteine (HcP; 1) and its enzymatic deprotection to
homocysteine (Hcy) under physiological conditions using
penicillin G acylase (PGA; Scheme 1). HcP could be efficiently
deprotected within minutes using catalytic amounts of PGA, but
depending on structural accessibility in the protein. We
demonstrate the utility of this latent thiol group with a set of unique
new applications based on thiol chemistry, ranging from selective
dual-labeling and selective single labeling under preservation of
disulfide bonds to a novel intramolecular crosslinking approach.
basis for the importance of cysteine in chemical bioconjugation
and ligation reactions[ for protein modification and synthesis.
3]
Only the rare selenol group of selenocysteine is of similar
versatility.
9]
To exploit the unique reactivities beyond the opportunities
offered by a native or artificially introduced cysteine, several thiol-
and selenol-containing unnatural amino acids have been
[
4]
[5]
genetically incorporated in either free or protected form. The
even higher reactivity of the selenocysteine side chain allows for
chemoselective bioconjugation in the presence of thiol groups.[6]
However, a free selenol group can undergo undesired oxidation
to diselenide and mixed sulfide-selenide bonds during protein
biosynthesis, folding and purification. In order to more generally
enable a selective functionalization of an additional thiol or selenol
group, their incorporation in a protected form is highly desirable.
To date, all of the reported cases of such latent thiols or selenols
employed photo-chemical or chemical deprotection schemes that
require conditions potentially damaging to the protein of interest
Results and Discussion
(
POI). For example, UV irradiation to deprotect photocaged
Genetic incorporation and enzymatic deprotection of
Hcy(Phacm) (HcP)
cysteine, selenocysteine and homocysteine side chains can be
harmful to disulfides in the protein and transforms the liberated
thiol group into a reactive radical species.[ Photocage groups
7]
1
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
=
1.2 eq. HSiEt
3
in TFA/DCM. Yield = 46%. B) Coomassie-stained and α-His
We chose the Phacm group for protecting a latent thiol because
PGA is known to recognize and cleave its phenylacetyl (Pac)
portion and that the liberated thioaminal further decomposes to
reveal the Hcy core structure (Scheme 1). The heterodimeric PGA
enzyme is used on an industrial scale to remove the Pac group
western blot SDS-PAGE analysis of incorporation of 1. Ind.
expression. C) PGA-His -mediated deprotection of HcP side chain in protein 2.
Left panel: scheme of the reaction; right panel: Coomassie-stained SDS-PAGE
of purified proteins and deprotection reaction as indicated. D) ESI-MS analysis
of the reaction shown in C) with 0.01 eq. PGA-His
= induced
6
6
(1 h, 25 °C). E) Time-course
of the analysis shown in C & D).
from penicillin G; a key step in the production of semi-synthetic
penicillin variants.[
10]
PGA does not cleave regular peptide
bonds.[10-11] It is also utilized on short polypeptide substrates in
Having established incorporation of 1 via genetic code
expansion, we investigated removal of the Phacm group from the
model protein using PGA. We prepared purified PGA tagged with
a hexahistidine sequence at the C terminus of the β subunit by
recombinant expression in E. coli[ and tested the enzyme at
different concentrations. To our delight, incubation of the protein
substrate 2 (10 µM) with just 0.01 eq. PGA-His led to quantitative
6
deprotection, as confirmed by ESI-MS, and was virtually complete
already after 10 min (Figure 1D & E). Even with a lower amount
solid phase peptide synthesis (SPPS) to remove these groups
from orthogonally protected Lys(Pac) and Cys(Phacm).[11-12]
HcP (1) was synthesized according to the scheme in Figure
1A and obtained in good yield as a soluble TFA salt. To establish
14]
incorporation during protein translation by amber stop codon
suppression we found that a mutated Methanosarcina barkeri
pyrrolysine tRNA synthetase (PylRS*) evolved for lysine
derivatives with bromoalkyl chains[13] also accepted 1 as a
substrate. Escherichia coli BL21(DE3) cells were transformed
6
with two plasmids encoding the PylRS*/tRNA pair and a H -
diSUMO (small ubiquitin related modifier) model protein (2) with
an amber stop codon at the third position in front of the histidine
tag, effectively representing the second position after removal of
the start methionine. 1 was incorporated with good efficiency
of PGA-His
after 2 h (Figure S1). No protein degradation could be detected,
even at high PGA-His concentrations (10 µM ≙ 1 eq., tested up
6
(0.001 eq.) we still observed complete deprotection
6
to 24h), consistent with previous reports that PGA does not cleave
in the peptide backbone of proteins (Figure S2).[11] The ESI-MS
analysis also allowed us to observe the thioaminal intermediate,
further confirming the mechanism following PGA-mediated
removal of the Pac moiety (Figure S3). Of note, this and all
subsequent MS analyses with other proteins showed no
indication of a potential adduct formation with formaldehyde,
suggesting that no such undesired effects are caused by this side
product of the Phacm-release mechanism (Scheme 1), likely due
to the low concentration in the micromolar range.
(
yielding 13 mg/L expression culture of purified protein) when
added at 2 mM to the growth medium (Figure 1B). ESI-MS
analysis of the purified diSUMO(3HcP) protein (2) confirmed the
expected mass with no unprotected species detectable. This
result verified the incorporation of 1 and suggested that the
Phacm group remained stable during expression and purification.
Structural requirements for PGA-mediated HcP deprotection
We then studied the structural requirements for such efficient and
rapid PGA-mediated deprotection of HcP (1) in proteins. PGA’s
active site is at the bottom of a deep funnel that becomes
increasingly constricted and should be easily accessible only by
small molecules such as the native substrate penicillin G (Figure
A).[ Therefore, a clear dependency on the surrounding steric
15]
2
bulk was expected. The accessibility of the side chain of 1 should
depend on the globular fold around its position.[13a] In our initial
deprotection assays (Figure 1), 1 was located close to the
N terminus of an unstructured tail region, thus exhibiting nearly
maximal accessibility. We systematically tested other positions
with varying distance from the N terminus in diSUMO model
proteins (Figures 2B and S4-S6) and analyzed time-dependent
deprotection by ESI-MS as described above. Constructs with 1 in
a distance of 1, 4, or 8 aa from the N terminus and 20-29 aa away
from the globular fold of the first SUMO domain were
quantitatively deprotected under the chosen conditions (0.1 eq.
PGA-His
quantitatively deprotected with 0.01 eq. PGA-His
E, S3 and S6), suggesting that the position of 1 within an
6
, 4 h; see Figure 2B). These three constructs were even
6
in ≤ 1 h (Figures
1
unfolded peptide stretch is largely irrelevant for a rapid and
efficient deprotection as long as it is located with enough distance
to structured regions.
We then tested deprotection of 1 on the surface of the first
SUMO domain (substituting R61), representing the other extreme
in terms of structural accessibility. The globular SUMO fold
encompasses ~72 aa. In fact, no deprotection could be observed,
consistent with the notion that the SUMO domain would not fit
deep enough into the funnel to the active site of PGA. Finally, we
wondered about the minimally required distance of the 1 position
Figure 1. Synthesis, incorporation and deprotection of HcP (1). A) Scheme of
synthesis. a = 1.1 eq. tri-n-butylphosphine in DCM/H
2
O; b = 1.5 eq. N-hydroxy-
O/dioxane; d
methylphenylacetamide, 0.08 eq. TsOH in dioxane; c = LiOH in H
2
2
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
relative to a global protein domain for efficient deprotection to
occur. With 1 being only nine residues away from the globular
SUMO domain, the efficiency was lowered to about 65 % (0.1 eq.
PGA-His , 4 h). A shorter distance of only 5 residues led to a
6
further reduction of the efficiency (~40%; Figure 2B).
to reach the PGA active site from the outer rim of its substrate
funnel (of 25-30 Å distance) in a one-pass arrangement as shown
in Figure 2A (left panel). Even a two-pass arrangement of the
polypeptide chain in the funnel is possible and can lead to efficient
and rapid deprotection of 1, suggesting a significant increase of
possible protein substrates (Figure 2B, right panel). HcP (1) in
large globular structures and on flat surfaces is likely inaccessible
for PGA but may still be cleaved at a slower rate depending on
the architecture when applying longer incubation times and higher
PGA amounts.[13a]
Of note, the rates and efficiencies observed here for PGA-
mediated deprotection reveal an unexpectedly high catalytic
potential on protein substrates. They represent a dramatic
improvement over previously reported PGA deprotection
schemes of Lys(Pac) or Cys(Phacm) on both peptide[
12-13, 16]
and
protein[13a, 17] substrates, which were often incomplete and
required overnight incubation. Next to the structural
dependencies, it should be noted that PGA has typically been
used as a commercial sample from industrial production batches,
often in immobilized form. We recommend not to use such
samples for protein applications due to potential problems with
protein degradation and precipitation, likely due to their crude
nature.[13a]
Having established protocols for HcP(1) incorporation and
enzymatic deprotection to Hcy, we next aimed to develop new
protocols in protein chemistry and protein labeling.
Sequential and regioselective protein dual-labeling using
bioconjugation of thiol and latent thiol groups
We reasoned that the latent thiol group of 1 would allow for a
straight-forward sequential and regioselective dual-labeling
strategy of a protein of interest (POI). Following labeling of a
single Cys side chain, 1 would be converted to Hcy and the
liberated thiol group then be labeled in a second reaction.
Assuming no remaining reactive cysteine side chains after the first
labeling step, this scheme represents a quasi orthogonal labeling
of two thiol groups. Importantly, both labeling steps of this protocol
can employ well-established and efficient thiol bioconjugation,
which is appealing for its simplicity and availability of a plethora of
commercial reagents. Other reported strategies for regioselective
dual-labeling involve two different functional moieties, including
Figure 2. Deprotection efficiency of HcP (1) with PGA. A) One-pass (left panel)
and two-pass model (right panel) for a flexible polypeptide chain to reach
through the funnel into the active site of PGA. B) Deprotection yield of diSUMO
constructs with 1 at varying positions (t = 4 h). * = 1 on surface of globular part
of SUMO. C) Deprotection yield of Trx constructs with 1 at varying positions (t
[18]
more specialized types for bioorthogonal conjugation reactions.
Dual-labeling of a protein with an acceptor and a donor dye is
key to the Förster resonance energy transfer (FRET), a widely
used technique to study intramolecular distances and
conformational changes. We therefore tested the idea of
sequential and regioselective Cys and Hcy labeling with a
diSUMO FRET sensor. The design of the sensor was based on
the above-described model protein 2 with the latent thiol group of
=
1 h). D) Deprotection time-course of protein Trx-I with different amounts of
PGA-His
6
.
Similar findings were obtained using E. coli thioredoxin (Trx;
folded globular domain of ~95 aa) with an artificial N-terminal
extension as another model substrate protein (Figures 2C and S7).
Notably, quantitative deprotection was observed with 1 being
flanked by 20-23 aa on both sides as well as with 1 just 9 aa from
the globular domain. In contrast, a distance of only 2 aa led to
1
in the N-terminal region of the distal SUMO unit (Figure 3A).
This protein also contained a single cysteine (R61C) in the
proximal SUMO unit. In the first bioconjugation step, the thiol
group of the cysteine was labeled with the donor dye AlexaFluor
55 (AF555) as a maleimide reagent to give 2# (Figure 3A).
5
6
complete impairment (using 0.1 eq. PGA-His for 4h). A further
Following quenching with DTT (100 eq.) and subsequent removal
of excess quencher and dye by dialysis, we added PGA-SBP and
AF647 maleimide as the acceptor dye to simultaneously trigger
conversion of HcP to Hcy and labeling of the latter to give dually
labeled 2* (Figure 3A). PGA-SBP carries a streptavidin-binding
lowering of the PGA-His concentration to 0.01 eq. showed for
6
both model proteins that positions with a distance of only 9 aa to
the globular domain are deprotected less rapidly (Figure 2B-D).
Together, these data are supporting a model that about
6
peptide instead of the His tag and therefore could easily be
≥
8-10 aa are needed in unstructured or stretched conformation
3
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
2
h). Excitation of the gel was performed at 609 nm for AF647 and 535 nm for
removed together with the excess AF647 dye by Ni-NTA
purification of the His -tagged 2*. Notably, PGA and PGA-SBP
6
AF555.
have no cysteines and therefore do not cause any background
labeling in such one-pot reactions. The dual-labeling of 2* without
acceptor-acceptor or donor-donor conjugates was confirmed by
ESI-MS (Figure 3B). 2* exhibited intramolecular FRET (Figure
We then aimed to demonstrate the advantage of the new
regioselective dual-labeling protocol over the typically used
stochastic labeling of two cysteines in terms of the FRET
response. Whereas the stochastic labeling cannot avoid the
partial dual labeling with only acceptor and only donor dyes
(Figure 4A, right panel), no such unproductive species are formed
in our new approach and hence an increase in dynamic FRET
range would be expected (Figure 4A, left panel). To this end, we
designed a new nonribosomal peptide synthetase (NRPS) FRET
sensor to monitor the interaction of its adenylation (A) and
peptidyl-carrier-protein (PCP) domains, which together comprise
3C), which was lost by enzymatic cleavage into the AF647 and
AF555-labeled SUMO monomers 3 and 4, respectively, with the
SUMO protease SENP1 (Figure 3B bottom panel, Figures 3C &
D and S8).[18e] Together, these data confirmed the intended
regioselective conjugation of each SUMO monomer with just one
specific dye.
71 kDa. The in-solution analysis of the domain interaction is
important to understand the timing and dynamics of
conformational changes in these multi-domain biosynthetic
enzymes.[19] The only currently reported FRET sensor design for
these large, multidomain proteins utilized EGFP as a large
fluorescent protein donor dye, which is not optimal due to its
possible impact on domain mobility.[19] For a design based on two
small synthetic dyes, we introduced a single cysteine at position
N
N152C by site-directed mutagenesis in the large A subunit of the
A domain[
19a]
6
and fused a short tag (AGV(HcP)TEH ) containing
the new amino acid at the C-terminal end of the PCP domain to
give A-PCP(N152C/HcP-tag) (construct 5). For comparison,
cysteine was used at both positions to give the control construct
A-PCP(N152C/Cys-tag) (6) (Figure 4A). The Cys-HcP construct
5
was labeled with AF555 and AF647 maleimides according to
our new sequential protocol involving PGA-SBP to deprotect the
latent thiol group. Excess dye and PGA-SBP were separated off
by Ni-NTA chromatography. Successful regioselective dual-
labeling (91%) was confirmed by SDS-PAGE and ESI-MS
(
Figures S9A and S10). The Cys-Cys control construct 6 was
bioconjugated under the same conditions, however, by adding
AF555 and AF647 maleimides simultaneously as a mixture to
afford stochastic labeling of the two cysteines (Figure S9B).
Subsequently, both labeled proteins 5* and 6* were converted into
the catalytically active holo-forms by addition of the 4’-
[
20]
phosphopantetheinyl (Ppant) transferase Sfp and coenzyme A
(
Figure S10) to give the FRET sensors holo-5* and holo-6*.
The A-PCP FRET sensors were then biochemically and
biophysically characterized. The enzyme catalyzes L-Phe-AMP
formation from the substrates L-Phe and ATP as well as the
subsequent covalent binding of the amino acid as an L-Phe-
thioester on the Ppant prosthetic group. These chemical reactions
correlate with a shift of the conformational equilibrium to a
preferential binding of the PCP domain to the A domain (Figure
4
B).[19] Importantly, in the absence of substrates, the
regioselectively labeled sensor holo-5* showed a higher FRET
ratio (acceptor intensity over donor intensity = I /I ) compared to
a
d
holo-6*, consistent with the lack of the unproductive donor-donor
and acceptor-acceptor combinations (FRET ratios of 0.83 vs. 0.67,
respectively; Figure. 4C). Both FRET sensors reported on the
conformational change induced by the addition of substrates ATP
and L-Phe. Again, the regioselective design of holo-5* proved to
be more sensitive as it led to a greater change in the FRET ratio
Figure 3. Regioselective Cys and Hcy dual labeling of proteins. A) Scheme to
prepare diSUMO FRET sensor 2* and its proteolytic cleavage. B) ESI-MS
analysis of the unpurified reaction mixtures to analyze intermediates and
products. C) Fluorescence spectra of dually labeled 2* (2 µM) with and without
treatment with SENP1 (0.1 µM, 2 h) using excitation at 520 nm. C) SDS-PAGE
analysis of the proteolytic cleavage reaction of 2* (10 µM) with SENP1 (0.1 µM,
(
49% vs. 37% for holo-6*, respectively; Figure 4C), corresponding
to a higher dynamic range. Together, these findings clearly
demonstrate the advantage of regioselective over stochastic
4
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
labeling in FRET studies, and show the utility of our new labeling
protocol for such purposes.
in conjunction with the nascent nature of the Hcy thiol group
following deprotection by PGA, which should allow quantitative
bioconjugation without a preceding treatment with reductants like
TCEP or DTT, as is typically required for cysteine bioconjugation
following protein expression and purification. We hypothesized
that protein thiol bioconjugation on the latent thiol group of 1
should be possible with preservation of disulfide bonds and with
avoiding otherwise potentially harmful reducing conditions.
Figure 5. Latent thiol bioconjugation in presence of disulfide bond. A) Reaction
scheme involving a disulfide-linked anti-GFP nanobody (Nb) with subsequent
dual labeling after reducing the disulfide bond. B) ESI-MS analysis of unpurified
intermediates in the reaction scheme shown in A). C) Scheme of Nb-mediated
cell surface labeling of HeLa cells expressing EGFP. D) Confocal microscopy
images of transiently transfected HeLa cells to present EGFP as shown in C).
Note that non-transfected cells did not bind the nanobody. Scale bar = 25 µm.
Figure 4. Intramolecular multidomain NRPS FRET-sensor improved by
regioselective dual labeling. A) Schematic comparison of regioselective and
stochastic labeling. B) Scheme of conformational change of the NRPS di-
domain A-PCP-FRET sensor following substrate addition. C) Fluorescence
spectra of holo-5* or holo-6* (0.3 µM) to monitor FRET ratios with and without
substrates (using excitation at 520 nm). Measurements with substrates were
performed 30 min after adding 2 mM ATP and 2 mM L-Phe.
Nanobodies (Nbs) are single-domain antibody fragments
derived from heavy-chain-only antibodies with countless
applications in protein biochemistry, diagnostics and therapy.[
We noticed that when nanobodies are expressed into the
periplasm of E. coli, an extra single cysteine in a short appended
tag sequence tends to form a disulfide with a second monomer to
give a covalent homodimer. We therefore aimed to test our
Latent thiol bioconjugation in presence of intact cysteine
disulfide
In another novel strategy for protein modification we aimed to take
advantage of the fact that the latent thiol group of 1 does not
interfere with existing disulfide bonds in the protein during
expression and purification. This point is particularly noteworthy
21]
5
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
hypothesis by selectively bioconjugating the latent thiol group of
finding is consistent with our hypothesis that the cyclized protein
11 was resistant against TEV cleavage, and suggests that a near
quantitative intramolecular crosslink to 11 had occurred. Analysis
of the protein species by SDS-PAGE electrophoretically
separated cleaved 13 from 10/11 and corroborated an efficient
yield of >90% for the crosslink reaction (Figure S13).
1
in the presence of this disulfide bond between the two
nanobodies. In an additional extension, the disulfide could be
reduced to give the monomeric nanobody and the unpaired
cysteine be used for a second labeling reaction (Figure 5A).
To this end, we prepared a GFP-enhancer nanobody[22] with
a C-terminally appended cysteine and additionally fused a short
N-terminal tag to introduce the latent thiol group of 1. The resulting
protein 7 was purified as a disulfide-bridged dimer 7-S-S-7 (Figure
5
B) without any detectable free thiol of a monomeric species
Figure S11). Addition of PGA-His (0.02 eq., 1 h) to convert 1 into
Hcy and in situ labeling with biotin maleimide afforded the desired
(
6
#
#
conjugate 7 -S-S-7 with the intact disulfide bond in quantitative
yield, as confirmed by ESI-MS (Figure 5B, top panels).
Furthermore, to achieve subsequent dual-labeling, TCEP was
#
#
added to 7 -S-S-7 to reduce the disulfide and give the monomeric
#
species of 7 . Subsequent addition of Cy5 maleimide led to
bioconjugation of the unpaired cysteine side chain and furnished
the dually labeled nanobody 7* (Figure 5B, bottom panel and
Figure S12; note that under these conditions the other, internal
disulfide bond of the nanobody remained unaffected). Finally, we
proved the functional integrity of 7* (10 nM) by its specific binding
to EGFP presented on HeLa cells[ and by visualizing the biotin
moiety using iFluor405 conjugated streptavidin (Figure 5C and D).
23]
A latent thiol for enzyme-activated protein crosslinking
Finally, the latent thiol group of 1 inspired us to a novel strategy
to create stable protein crosslinks. We reasoned that as long as
the latent thiol group is kept protected, the thiol group of a free
cysteine could be chemically converted into a suitable electrophile.
Deprotection of the latent thiol by PGA then reveals the
nucleophilic reaction partner.
To test this idea, we chose to create an intramolecular
thioether covalent crosslink on ubiquitin as the model protein
(
Figure 6A). We incorporated 1 into an unstructured N-terminal
peptide extension, that also contained a TEV protease cleavage
site, and introduced single cysteine by site-directed
a
mutagenesis on the surface of Ub’s globular fold (K11C) to give
protein 8. The correct masses of this protein and the following
modification steps were confirmed by ESI-MS (Figure 6B).
Treatment of purified 8 with dibromoadipic amide converted
Cys11 into dehydroalanine (Dha11; protein 9).[24] Following
removal of excess 2,5-dibromoadipic amide we added PGA-His
6
(
(
0.01 eq.) to quantitatively deprotect the latent thiol of 1 to Hcy
protein 10). Overnight incubation was necessary to allow for the
slow spontaneous addition of the Hcy thiol to the Dha double bond
to give cyclized protein 11 with the novel protein crosslink.
However, since this reaction does not change the molecular mass
of the protein we were unable to distinguish 11 from 10 by ESI-
MS. To verify that the cyclization reaction had occurred, we
reasoned that the constrained conformation of the cyclized
peptide chain in 11 would likely twist the TEV cleavage site from
its stretched conformation necessary for recognition by TEV
protease and thereby partially or completely block TEV cleavage.
In contrast, the uncyclized structure in 9 and 10 should be
quantitatively cleaved (Figure 6A; right panels). Indeed, we found
9
to be quantitatively cleaved by TEV protease into 12 and 13. In
contrast, following incubation of the potential mixture of
uncyclized 10 and cyclized 11 with TEV we could only detect a
weak signal for cleavage product 13 and most of the signal for
Figure 6. Intramolecular protein crosslinking using the latent thiol group. A)
Scheme of the strategy involving the conversion of a single cysteine into an
electrophile. The structural representation of the Ub model protein is based on
pdb-file 1UBQ. Reaction conditions: i) dibromoadipic amide (10 eq., overnight,
10/11 remained unchanged (Figure 6B, bottom panel). This
6
37 °C); ii) PGA-His (0.01 eq.), TCEP (5 eq.); overnight); iii) TEV protease (0.1
6
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
eq., 3h). B) ESI-MS analysis of the reaction intermediates and products (12: Mcal.
Keywords: nanobody • genetic code expansion • penicillin G
=
1232.52 Da).
acylase • thiol bioconjugation • regioselective labeling
[1]
[2]
[3]
[4]
D. Fass, C. Thorpe, Chem Rev 2018, 118, 1169-1198.
Together, these results show that the latent thiol group in 1
S. B. Gunnoo, A. Madder, Chembiochem 2016, 17, 529-553.
S. B. Kent, Chem Soc Rev 2009, 38, 338-351.
can be used to efficiently introduce novel crosslinks into proteins.
Notably, the formed thioether crosslink is of minimal atom
economy, also due to the small size of the Hcy side chain. The
latter contrasts to the mostly long and bulky side chains
incorporated by the genetic code expansion technology for
related purposes.[25]
a) S. Virdee, P. B. Kapadnis, T. Elliott, K. Lang, J. Madrzak, D. P. Nguyen,
L. Riechmann, J. W. Chin, J Am Chem Soc 2011, 133, 10708-10711; b)
X. Li, T. Fekner, J. J. Ottesen, M. K. Chan, Angew Chem Int Ed Engl
2009, 48, 9184-9187; c) J. R. Frost, N. T. Jacob, L. J. Papa, A. E. Owens,
R. Fasan, ACS Chem Biol 2015, 10, 1805-1816; d) E. Brustad, M. L.
Bushey, A. Brock, J. Chittuluru, P. G. Schultz, Bioorg Med Chem Lett
2
008, 18, 6004-6006; e) T. Liu, Y. Wang, X. Luo, J. Li, S. A. Reed, H.
Xiao, T. S. Young, P. G. Schultz, Proc Natl Acad Sci U S A 2016, 113,
910-5915; f) M. Koh, H. Y. Cho, C. Yu, S. Choi, K. B. Lee, P. G. Schultz,
Conclusions
5
Bioconjug Chem 2019, 30, 2102-2105.
[5]
a) R. Uprety, J. Luo, J. Liu, Y. Naro, S. Samanta, A. Deiters,
Chembiochem 2014, 15, 1793-1799; b) N. Wu, A. Deiters, T. A. Cropp,
D. King, P. G. Schultz, J Am Chem Soc 2004, 126, 14306-14307; c) D.
P. Nguyen, M. Mahesh, S. J. Elsasser, S. M. Hancock, C. Uttamapinant,
J. W. Chin, J Am Chem Soc 2014, 136, 2240-2243; d) D. P. Nguyen, T.
Elliott, M. Holt, T. W. Muir, J. W. Chin, J Am Chem Soc 2011, 133, 11418-
We have reported the genetic encoding of the novel non-
canonical amino acid HcP (1) with a latent thiol group and its
enzymatic deprotection to Hcy. The PGA-mediated deprotection
was found to be surprisingly rapid and efficient at sterically well-
accessible positions in unstructured regions, but was poor or
impossible on flat surfaces of stably folded globular proteins.
Importantly, the deprotection is performed under mild,
physiological, non-denaturing and non-destructive conditions to
the protein of interest, representing a significant advancement in
terms of preparative utility over previously reported chemical or
photochemical deprotection schemes of latent thiol or selenol
groups. We have demonstrated the potential of the latent thiol
group in several novel approaches of selective chemical
modification of proteins, thereby significantly expanding the scope
of thiol bioconjugation. The variety of these approaches reflect the
versatility of thiol group chemistry. Protein thiol bioconjugation is
a classical and well-established technique, and is still highly
attractive due to its simplicity, chemoselectivity, efficiency and
wide-spread use. The synthesis of the required unnatural amino
acid 1 is easy to perform. The requirement for bioorthogonal
reactions is circumvented in our chemical modification schemes.
Bioorthogonal reactions require more sophisticated reagents of
more restricted availability, in particular for dual labeling,[ and
are of more restricted applicability in many regards, for example
as caused by the interference of cyclooctynes with free
cysteines.[ Our reported examples include a) two different
routes to regioselective and consecutive dual labeling of cysteine
and homocysteine (Hcy) as two quasi-orthogonal thiol groups, b)
selective thiol conjugation under preservation of disulfide bonds
that are otherwise sensitive to reducing conditions, and c) a
protein crosslinking strategy to introduce stable thioether bridges.
These protocols would not have been possible without the latent
thiol group. We believe that they present powerful new tools for
the protein chemist and biochemist and that the novel latent thiol
group will enable many other new and unique applications.
1
1421; e) J. Liu, F. Zheng, R. Cheng, S. Li, S. Rozovsky, Q. Wang, L.
Wang, J Am Chem Soc 2018, 140, 8807-8816; f) W. Ren, A. Ji, H. W. Ai,
Am Chem Soc 2015, 137, 2155-2158; g) R. Rakauskaite, G.
J
Urbanaviciute, A. Ruksenaite, Z. Liutkeviciute, R. Juskenas, V.
Masevicius, S. Klimasauskas, Chem Commun (Camb) 2015, 51, 8245-
8248; h) K. Yang, G. Li, P. Gong, W. Gui, L. Yuan, Z. Zhuang,
Chembiochem 2016, 17, 995-998.
[
[
[
6]
7]
8]
L. Johansson, C. Chen, J. O. Thorell, A. Fredriksson, S. Stone-Elander,
G. Gafvelin, E. S. Arner, Nat Methods 2004, 1, 61-66.
C. E. Hoyle, C. N. Bowman, Angew Chem Int Ed Engl 2010, 49, 1540-
1
573.
J. K. Böcker, W. Dorner, H. D. Mootz, Chem Commun (Camb) 2019, 55,
287-1290.
C. C. Liu, P. G. Schultz, Annu Rev Biochem 2010, 79, 413-444.
1
[9]
[10] K. Srirangan, V. Orr, L. Akawi, A. Westbrook, M. Moo-Young, C. P. Chou,
Biotechnol Adv 2013, 31, 1319-1332.
[
11] D. Kadereit, H. Waldmann, Chem Rev 2001, 101, 3367-3396.
12] M. Royo, J. Alsina, E. Giralt, U. Slomcyznska, F. Albericio, J. Chem. Soc.,
Perkin Trans. 1 1995, 1095-1102.
[
[
13] a) M. Reille-Seroussi, S. V. Mayer, W. Dorner, K. Lang, H. D. Mootz,
Chem Commun (Camb) 2019, 55, 4793-4796; b) M. Cigler, T. G. Muller,
D. Horn-Ghetko, M. K. von Wrisberg, M. Fottner, R. S. Goody, A. Itzen,
M. P. Muller, K. Lang, Angew Chem Int Ed Engl 2017, 56, 15737-15741.
18]
26]
[14] T. Cheng, M. Chen, H. Zheng, J. Wang, S. Yang, W. Jiang, Protein Expr
Purif 2006, 46, 107-113.
[
[
[
15] H. J. Duggleby, S. P. Tolley, C. P. Hill, E. J. Dodson, G. Dodson, P. C.
Moody, Nature 1995, 373, 264-268.
16] J. Tulla-Puche, M. Gongora-Benitez, N. Bayo-Puxan, A. M. Francesch,
C. Cuevas, F. Albericio, Angew Chem Int Ed Engl 2013, 52, 5726-5730.
17] a) Q. C. Wang, J. Fei, D. F. Cui, S. G. Zhu, L. G. Xu, Biopolymers 1986,
25 Suppl, S109-114; b) L. Zakova, D. Zyka, J. Jezek, I. Hanclova, M.
Sanda, A. M. Brzozowski, J. Jiracek, J Pept Sci 2007, 13, 334-341.
[18] a) E. M. Brustad, E. A. Lemke, P. G. Schultz, A. A. Deniz, J Am Chem
Soc 2008, 130, 17664-17665; b) B. Wu, Z. Wang, Y. Huang, W. R. Liu,
Chembiochem 2012, 13, 1405-1408; c) I. Nikic, T. Plass, O. Schraidt, J.
Szymanski, J. A. Briggs, C. Schultz, E. A. Lemke, Angew Chem Int Ed
Engl 2014; d) A. Sachdeva, K. Wang, T. Elliott, J. W. Chin, J Am Chem
Soc 2014, 136, 7785-7788; eL. J. Kost, H. D. Mootz, Chembiochem 2018,
Acknowledgements
19, 177-184.
We thank Kathrin Lang and Sheng Yang for providing plasmids
for the PylRS/tRNA system and PGA expression, respectively.
We gratefully acknowledge funding of this work by the Deutsche
Forschungsgemeinschaft (DFG; grants MO1073/8-1 and
SFB858/B14 to H. D. M.).
[19] a) J. Alfermann, X. Sun, F. Mayerthaler, T. E. Morrell, E. Dehling, G.
Volkmann, T. Komatsuzaki, H. Yang, H. D. Mootz, Nat Chem Biol 2017,
13, 1009-1015; b) F. Mayerthaler, A.-L. Feldberg, J. Alfermann, X. Sun,
W. Steinchen, H. Yang, H. D. Mootz, RSC Chem Biol 2021,
https://doi.org/10.1039/D1030CB00220H.
[20] R. H. Lambalot, A. M. Gehring, R. S. Flugel, P. Zuber, M. LaCelle, M. A.
Marahiel, R. Reid, C. Khosla, C. T. Walsh, Chem Biol 1996, 3, 923-936.
7
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
[21] R. W. Cheloha, T. J. Harmand, C. Wijne, T. U. Schwartz, H. L. Ploegh, J
Biol Chem 2020, 295, 15307-15327.
[
22] A. Kirchhofer, J. Helma, K. Schmidthals, C. Frauer, S. Cui, A. Karcher,
M. Pellis, S. Muyldermans, C. S. Casas-Delucchi, M. C. Cardoso, H.
Leonhardt, K. P. Hopfner, U. Rothbauer, Nat Struct Mol Biol 2010, 17,
1
33-138.
23] B. Jedlitzke, Z. Yilmaz, W. Dorner, H. D. Mootz, Angew Chem Int Ed Engl
020, 59, 1506-1510.
[
[
2
24] J. M. Chalker, S. B. Gunnoo, O. Boutureira, S. C. Gerstberger, M.
Fernandez-Gonzalez, G. J. Bernardes, L. Griffin, H. Hailu, C. J. Schofield,
B. G. Davis, Chem Sci 2011, 2, 1666-1676.
[25] a) V. Y. Berdan, P. C. Klauser, L. Wang, Bioorg Med Chem 2020, 29,
115896; b) H. Neumann, S. Y. Peak-Chew, J. W. Chin, Nat Chem Biol
2008, 4, 232-234; c) T. Kobayashi, C. Hoppmann, B. Yang, L. Wang, J
Am Chem Soc 2016, 138, 14832-14835.
[26] R. van Geel, G. J. Pruijn, F. L. van Delft, W. C. Boelens, Bioconjug Chem
2012, 23, 392-398.
8
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202102343
COMMUNICATION
Entry for the Table of Contents
Enzymatic deprotection of a latent thiol under mild and physiological conditions expands the scope of classic thiol bioconjugation.
New reaction schemes enable regioselective labeling and dual modifications for a variety of preparative applications and using the
established and commercially available plethora of thiol-directed reagents.
Institute and/or researcher Twitter usernames: ((optional))
9
This article is protected by copyright. All rights reserved.