Reversible Inhibitors Arrest ClpP in a Defined Conformational State that Can Be Revoked by ClpX Association

Article abstract of DOI:10.1002/anie.201507266

Caseinolytic protease P (ClpP) is an important regulator of Staphylococcus aureus pathogenesis. A high-throughput screening for inhibitors of ClpP peptidase activity led to the identification of the first non-covalent binder for this enzyme class. Co-crystallization of the small molecule with S. aureus ClpP revealed a novel binding mode: Because of the rotation of the conserved residue proline 125, ClpP is locked in a defined conformational state, which results in distortion of the catalytic triad and inhibition of the peptidase activity. Based on these structural insights, the molecule was optimized by rational design and virtual screening, resulting in derivatives exceeding the potency of previous ClpP inhibitors. Strikingly, the conformational lock is overturned by binding of ClpX, an associated chaperone that enables proteolysis by substrate unfolding in the ClpXP complex. Thus, regulation of inhibitor binding by associated chaperones is an unexpected mechanism important for ClpP drug development.

Full text of DOI:10.1002/anie.201507266

.
Angewandte
Communications
International Edition: DOI: 10.1002/anie.201507266
German Edition: DOI: 10.1002/ange.201507266
Drug Development
Reversible Inhibitors Arrest ClpP in a Defined Conformational State
that Can Be Revoked by ClpX Association
Axel Pahl, Markus Lakemeyer, Marie-Theres Vielberg, Mathias W. Hackl, Jan Vomacka,
Vadim S. Korotkov, Martin L. Stein, Christian Fetzer, Katrin Lorenz-Baath, Klaus Richter,
Herbert Waldmann, Michael Groll,* and Stephan A. Sieber*
Abstract: Caseinolytic protease P (ClpP) is an important
regulator of Staphylococcus aureus pathogenesis. A high-
throughput screening for inhibitors of ClpP peptidase activity
led to the identification of the first non-covalent binder for this
enzyme class. Co-crystallization of the small molecule with
S. aureus ClpP revealed a novel binding mode: Because of the
rotation of the conserved residue proline 125, ClpP is locked in
a defined conformational state, which results in distortion of
the catalytic triad and inhibition of the peptidase activity. Based
on these structural insights, the molecule was optimized by
rational design and virtual screening, resulting in derivatives
exceeding the potency of previous ClpP inhibitors. Strikingly,
the conformational lock is overturned by binding of ClpX, an
associated chaperone that enables proteolysis by substrate
unfolding in the ClpXP complex. Thus, regulation of inhibitor
binding by associated chaperones is an unexpected mechanism
important for ClpP drug development.
Figure 1. SaClpP in its a) extended (PDB ID: 3V5E), b) compact (PDB
ID: 4EMM), and c) compressed (PDB ID: 3QWD) state. The E helices
are shown in blue. d) Structure of AV145. e) Analytical ultracentrifuga-
tion of wild-type ClpP and the heptameric[2] ClpP mutant R171A.
DMSO served as a control. Sedimentation coefficient maxima are
shown as the meanÆstandard deviation.
C
aseinolytic protease P (ClpP), a member of the serine
hydrolase enzyme family, is a major regulator of bacterial cell
homeostasis.[1] The enzymatic complex consists of two adja-
cent heptameric rings that are connected by central a-
helices E, forming a tetradecameric barrel (Figure 1a–c).
The E helix is linked to strand b9, which forms crucial
hydrogen bonds across the heptamer interface. Different
conformations of ClpP have been observed. Whereas an
extended E helix (Figure 1a, PDB ID: 3V5E)[2] is important
for peptidolytic activity and tetradecamer stability, a kink in
this helix leads to either compact (Figure 1b, PDB ID:
4EMM)[3] or compressed (Figure 1c, PDB ID: 3QWD)[4]
states with misaligned catalytic triads. This distortion of
catalytic residues induces a rotation of the conserved Pro125
(in strand b9), which in turn pulls at helix E, resulting in its
collapse. ClpP requires associated chaperones, such as
hexameric ClpX, to unfold and digest larger protein sub-
strates.[1] Moreover, ClpX binding is believed to induce
conformational selection of the active and extended state.[5]
Thus far, only b-lactones and phenyl esters have been
reported as specific ClpP inhibitors in whole proteome
studies.[6] Both compound classes covalently acylate the
active-site Ser98.[7] Covalent, irreversible binding is beneficial
for proteome-labeling experiments, mechanistic studies, and
for achieving a prolonged target residence time. The in vi-
tro[5,6] and in vivo[8] application of irreversible ClpP inhibitors,
however, has been limited by the low stability of their
electrophilic motifs owing to hydrolysis.[9] Thus far, all
attempts to design non-covalent inhibitors have failed.[7a]
Surprisingly, given the importance of ClpP inactivation and
the need for rational design, only one complex crystal
structure of the protease with a nonspecific chloromethyl
ketone (CMK) peptide ligand is available (PDB ID: 2FZS).[10]
Herein, we report the first co-crystal structure of a specific
ClpP inhibitor with a novel, reversible mode of action, which
was further exploited to synthetically optimize the ligand by
rational design and in silico screening.
[*] Dr. A. Pahl,[+] M. Lakemeyer,[+] M.-T. Vielberg,[+] M. W. Hackl,
J. Vomacka, Dr. V. S. Korotkov, Dr. M. L. Stein, C. Fetzer,
Dr. K. Lorenz-Baath, Dr. K. Richter, Prof. Dr. M. Groll,
Prof. Dr. S. A. Sieber
Center for Integrated Protein Science at the Department of
Chemistry, Technische Universität München
Lichtenbergstrasse 4, 85747 Garching (Germany)
E-mail: michael.groll@tum.de
stephan.sieber@tum.de
Prof. Dr. H. Waldmann
Department of Chemistry and Chemical Biology
Technische Universität Dortmund
Otto-Hahn-Strasse 6, 44221 Dortmund (Germany)
[+] These authors contributed equally to this work.
Supporting information and ORCID(s) from the author(s) for this
article are available on the WWW under http://dx.doi.org/10.1002/
anie.201507266.
15892
ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 15892 –15896

Angewandte
Chemie
A previous high-through-
put screen (HTS) with about
140000 compounds from the
COMAS library (MPI Dort-
mund) did not reveal a single
non-covalently binding SaClpP
inhibitor with an IC50 < 2 mm.[6a]
However, the low activity of
ClpP with the standard Suc-
Leu-Tyr-AMC
substrate
requires at least 1 mm ClpP to
record significant turnover, so
that the potencies of the best
inhibitors are still in the micro-
molar concentration range. We
therefore reinvestigated our
previous HTS results by low-
ering the selection criteria to an
IC50 ꢀ 10 mm.[6a] Four com-
pounds of the HTS fulfilled
this prerequisite (Supporting
Information, Figure S1). Mass-
spectrometric analysis of the
intact proteins revealed that
AV 145 binds non-covalently
(Figure 1d, Figure S1), and
medicinal-chemistry considera-
tions made this compound the
most promising candidate for
further analysis. Compound
AV 145 lacks any reactive
groups and consists of three
characteristic
a pyrazolopyridine as well as
2-(thiophen-2-yl)oxazole
heterocycles,
a
moiety. Time-dependent incu-
bation of SaClpP with AV 145
did not change the IC50 (Fig-
ure S2), supporting the fact
that the compound is a reversi-
ble inhibitor. Furthermore,
analytical ultracentrifugation
demonstrated that AV 145,
unlike most phenyl esters and
lactones, did not induce disso-
ciation of ClpP into heptamers
(Figure 1e).[6a,7b]
Figure 2. a) Binding of AV145 to SaClpP (PDB ID: 5DL1). b) Structural superposition of SaClpP:AV145
with the extended, compact, or compressed form of SaClpP. Structural changes between the extended and
co-crystal structure are indicated by colored arrows. c) Detailed exploration of the structural changes of a-
helix E and strand b9 compared to the extended form. d) Analysis of the specific enzyme–inhibitor
contacts and overlay with the EcClpP:Z-LY-CMK structure (PDB ID: 2FZS), illustrating the non-covalent
inhibition mechanism of AV145. Color code: SaClpP:AV145 marine, AV145 orange, SaClpP:extended/
To gain insights into the
mechanism of inhibition, we
co-crystallized AV 145 with
SaClpP and solved the complex compact/compressed and EcClpP:Z-LY-CMK gray; heteroatoms: O red, N blue, S yellow.
structure by molecular replace-
ment at 3 Š resolution (Rfree
=
27.4%, Table S1, PDB ID: 5DL1). The asymmetric unit
contains a ClpP tetradecamer with the compound bound to
every subunit. Despite variations in occupancy, model build-
ing into the averaged 2FoÀFc electron-density map allowed
the unambiguous positioning of AV 145 near to the active site
between a-helix E and strand b9 (Figure 2a). Surprisingly,
overlays with structures of the apo enzyme in its extended,
compact, and compressed form revealed that the inhibitor
binds in a non-substrate-like mode and induces an unprece-
dented conformational state (Figure 2b). Although helix E is
almost completely aligned as in the extended, active enzyme,
binding of AV 145 displaces Ile143 and Arg147 because of
Angew. Chem. Int. Ed. 2015, 54, 15892 –15896
ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 15893

.
Angewandte
Communications
a steric clash (Figure 2c) leading to a shift of about 1.4 Š in
the center of the helix (green arrow in Figure 2b). The
majority of the structural alterations, however, are found in
strand b9 (purple arrow in Figure 2b). Formation of a b-sheet
with the subunit in the opposite ring is impeded, as can be
seen by the partial disorder of the two glycine residues 127
and 128. Notably, b9 is mainly defined by a large inhibitor-
induced flip of Pro125 of about 1808. The residue still retains
its trans peptide bond and resides at a similar position as in the
inactive form of the protease (Figure 2c). One direct con-
sequence of this relocation is a shift of His123 out of the
catalytic triad, leading to a distortion of the active site, which
explains the inhibiting effect of AV 145. The induced struc-
tural rearrangement of Pro125 also causes a major change to
Gln124. Hereby, Gln124 is arrested in a position where it is
able to form two hydrogen bonds between its peptide
backbone and AV 145, likely providing important protein–
inhibitor interactions. Moreover, an extended network of van
der Waals contacts with Phe102, Ile122, His142, Ile143,
Thr146, Leu150, and Leu154 as well as Thr169 stabilizes
compound binding (Figure 2d). Overall, this novel mode of
action explains why AV 145 is a non-covalent SaClpP inhib-
itor. Superposition of our structure with the Z-LY-CMK
bound EcClpP structure (PDB ID: 2FZS)[10] reveals differ-
ences (Figure 2d). Unlike CMK, AV 145 does not influence
the catalytic center of SaClpP directly, but rather transmits its
function through neighboring residues. Such a type of indirect
binding has not been observed for this class of enzymes thus
far.
The co-crystal structure of SaClpP in complex with AV 145
provides key information for improving inhibitor potency.
The two constituent parts of the molecule, namely the
pyrazolopyridine (acid part) and 2-(thiophen-2-yl)oxazole
(amine part) moieties, were substituted with different chem-
ical groups to explore their structure–activity relationships
(SARs; Figure 3a). Compounds were designed according to
Figure 3. SAR studies. a) The amine part of AV145 is intolerant to modifications whereas the acid part can be modified to improve ClpP inhibition.
b) Synthesis of AV280: Reagents and conditions: i) 3,4-dimethylbenzyl bromide, K2CO3, DMF, RT, 18 h; ii) LiOH, THF/H2O (9:1), RT, 18 h;
iii) neat, 1208C, 16 h; iv) NaN3, DMF, 608C, 16 h; v) PPh3, THF/H2O (9:1), RT, 16 h; vi) TBTU, NMM, DMF, 08C to RT, 18 h. Structures and IC50
values for derivatives with c) N-aryl substituents and d) para-substituted aryl groups that were inspired by virtual screening. IC50 values are derived
from at least two biological experiments with three technical replicates per concentration and are shown as mean and 95% confidence interval.
See the Supporting Information for further details.
15894 www.angewandte.org
ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 15892 –15896

Angewandte
Chemie
rational considerations or resulted from a virtual screen (see
the Supporting Information). Alterations within the amine
part, for example, by introduction of thiophene substituents
and the replacement of the thiophene by a phenyl or oxazole
group, resulted in inactive compounds (Figures S3, S4).
However, modifications within the acid part were much
more tolerated and led to improved inhibitors. Based on this,
we explored the possibility of incorporating aromatic groups
for p-stacking with His142 (part of the strand b9) through the
design of eleven derivatives with N-aryl substituents (Fig-
ure 3c).
The synthesis of these derivatives, exemplified for AV 280
in Figure 3b, involved N-alkylation of pyrazolopyridine
1 with 3,4-dimethylbenzyl bromide, followed by saponifica-
tion of the ester with lithium hydroxide. Preparation of the
oxazole amine part 3 started with the condensation of the
appropriate aryl amide with dichloroacetone at 1208C. The
resulting chloromethyl-substituted oxazole 2 was converted
into amine 3 in two steps. Standard amide coupling of the acid
with the amine resulted in the final products.
AV 191 and AV 280, which bear a benzyl and a 3,4-
dimethylbenzyl substituent, respectively, showed a sixfold
increase in ClpP inhibition compared to AV 145, with IC50
values of 1.7 and 1.5 mm, respectively (Figure 3c). Based on
results from the virtual screening, we next replaced the
pyrazolopyridine moiety by aryl rings with a para urea or
a para carbamate substituent, resulting in AV 286 with an IC50
value of 0.9 mm (Figure 3d). Notably, this value shows that
a stoichiometric amount of AV 286 (relative to the monomer
concentration of the protease) is sufficient to inhibit 50% of
SaClpP. An increase in substrate concentration did not
influence inhibitor binding (Figure S5).
Figure 4. In situ target validation in living S. aureus cells with AV321.
a) Chemical structure of the activity-based photoprobe AV321. b) Fluo-
rescence SDS-PAGE analysis of the soluble fraction. c) Volcano plot
representation of gel-free quantitative ABPP experiments measured on
the Orbitrap XL. Data are derived from three biological replicates for
DMSO and AV321 (5 mm), respectively. The log2-fold enrichment
values were z-score-normalized, and log10(p values) were calculated
using a two-sided one-sample Student’s t-test. Protein enriched in
addition to SaClpP: Protein ID Q2G0P0: 50S ribosomal protein L1. See
the Supporting Information for further details and additional analysis.
With potent and reversible ClpP inhibitors at hand, we
investigated whether the best compound of this series exhibits
cell permeability and target selectivity in situ. Therefore, an
AV 286-related activity-based protein profiling[11] photoprobe
(AV 321) equipped with a diazirine photocrosslinker and an
alkyne tag was prepared (Figure 4a).[12] Pleasingly, the probe
retained a low IC50 value (2 mm) for SaClpP peptidase
inhibition (Figure S6). Living S. aureus cells were incubated
with AV 321, irradiated with UV light to form a covalent link
between target protein and diazirine, lysed, and clicked to
a functionalized azide tag (fluorescent dye or biotin) via the
alkyne; the labeled proteome was then analyzed by fluores-
cence SDS-PAGE analysis (Figure 4b) or mass spectrometry
(MS; Figure 4c). A dominant fluorescent protein band at the
molecular weight of ClpP appeared on SDS-PAGE (Fig-
ure 4b). To confirm ClpP binding in situ we performed
quantitative gel-free MS by isotope labeling.[13] MS-based
target enrichment was visualized by volcano plots (Figure 4c,
Figure S7). In all runs, ClpP was highly enriched, and only
a few putative off-targets (significance level: p ꢀ 0.05), such as
the 50S ribosomal protein L1 (Protein ID ID: Q2G0P0), were
detected. We next examined whether cell-permeable inhib-
itors reduced the production of a-hemolysin (hla), a predom-
inant S. aureus toxin regulated by ClpP. Surprisingly, the
general level of hla was high, suggesting that inhibition of
intracellular proteolysis was not as efficient as the reduction
of in vitro peptidase activity.
To explore this finding in more detail, we tested the most
potent peptidase inhibitors in a ClpXP protease assay with
fluorescent GFP–SsrA (tagged for ClpXP degradation) as
a substrate. Indeed, all inhibitors were largely inactive in this
assay (Figure S8), demonstrating that ClpX overrides the
inhibitor-induced conformational lock. To investigate the
mechanistic basis of this unexpected behavior, we utilized
a previously introduced activator of ClpP.[14] This compound
displaces ClpX in protease assays and was thus used as
a surrogate for the chaperone (Figure S9). The molecule
abolished AV 286 inhibition of ClpP in casein and peptidase
assays in a concentration-dependent manner, which validates
a conformational selection of the active enzyme state by ClpX
or its surrogates (Figure S9).[15] Our observation is in line with
a recent study that showed a conformational switch to the
active and extended state upon binding of acyldepsipeptides,
which mimic ClpX binding.[5] Thus, regulation by ClpX and
other proteolytic activators represents an important and thus
far neglected parameter for future ClpP inhibitor develop-
ment.
In conclusion, AV 145 bound to ClpP arrested the enzyme
in an unprecedented inactive conformational state by a sig-
nificant rotation of Pro125 and an associated misalignment of
the catalytic triad architecture. This binding mode was
explored for the design of new compounds, which resulted
in AV 286, the first non-covalent inhibitor of ClpP with an
Angew. Chem. Int. Ed. 2015, 54, 15892 –15896
ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 15895

.
Angewandte
Communications
IC50 value of < 1 mm. Importantly, ClpX revoked non-cova-
lent binding of the examined inhibitors to ClpP, a counter-
intuitive finding when compared to related complex proteo-
lytic systems, such as the proteasome, where stable proteolytic
inhibitors for individual subunits have been described.[16] Our
results therefore strongly imply that regulators such as the
ATP-dependent ClpX orchestrate and modulate various
distinct conformational stages in the hydrolytic chamber of
ClpP. For the identification of non-covalent ClpP inhibitors
that are also active in situ, it is thus important not to solely
focus on the peptidase activity of the isolated ClpP complex,
as this, although easy to monitor, does not accurately reflect
the situation in live cells. We hypothesize that sustained ClpP
inhibition may either be achieved by increasing the small-
molecule-induced conformational lock by, for example, the
incorporation of large ligands that block the proteolytic
channel, or by inhibitor discovery with the intricate ClpXP
proteolytic complex. Both would certainly represent attrac-
tive starting points for the future optimization of anti-
virulence compounds against S. aureus and its antibiotic-
resistant strains.
[1] a) T. A. Baker, R. T. Sauer, Biochim. Biophys. Acta Mol. Cell
Res. 2012, 1823, 15 – 28; b) J. Ortega, S. K. Singh, T. Ishikawa,
M. R. Maurizi, A. C. Steven, Mol. Cell 2000, 6, 1515 – 1521;
c) J. A. Alexopoulos, A. Guarne, J. Ortega, J. Struct. Biol. 2012,
179, 202 – 210.
[2] M. Gersch, A. List, M. Groll, S. A. Sieber, J. Biol. Chem. 2012,
287, 9484 – 9494.
[3] F. Ye, J. Zhang, H. Liu, R. Hilgenfeld, R. Zhang, X. Kong, L. Li,
J. Lu, X. Zhang, D. Li, H. Jiang, C. G. Yang, C. Luo, J. Biol.
Chem. 2013, 288, 17643 – 17653.
[4] S. R. Geiger, T. Bottcher, S. A. Sieber, P. Cramer, Angew. Chem.
Int. Ed. 2011, 50, 5749 – 5752; Angew. Chem. 2011, 123, 5867 –
5871.
[5] M. Gersch, K. Famulla, M. Dahmen, C. Gobl, I. Malik, K.
Richter, V. S. Korotkov, P. Sass, H. Rubsamen-Schaeff, T. Madl,
H. Brotz-Oesterhelt, S. A. Sieber, Nat. Commun. 2015, 6, 6320.
[6] a) M. W. Hackl, M. Lakemeyer, M. Dahmen, M. Glaser, A. Pahl,
K. Lorenz-Baath, T. Menzel, S. Sievers, T. Bottcher, I. Antes, H.
Waldmann, S. A. Sieber, J. Am. Chem. Soc. 2015, 137, 8475 –
8483; b) T. Bçttcher, S. A. Sieber, Angew. Chem. Int. Ed. 2008,
47, 4600 – 4603; Angew. Chem. 2008, 120, 4677 – 4680.
[7] a) M. Gersch, F. Gut, V. S. Korotkov, J. Lehmann, T. Bottcher,
M. Rusch, C. Hedberg, H. Waldmann, G. Klebe, S. A. Sieber,
Angew. Chem. Int. Ed. 2013, 52, 3009 – 3014; Angew. Chem.
2013, 125, 3083 – 3088; b) M. Gersch, R. Kolb, F. Alte, M. Groll,
S. A. Sieber, J. Am. Chem. Soc. 2014, 136, 1360 – 1366.
[8] F. Weinandy, K. Lorenz-Baath, V. S. Korotkov, T. Bottcher, S.
Sethi, T. Chakraborty, S. A. Sieber, ChemMedChem 2014, 9,
710 – 713.
Acknowledgements
The work was funded by the Deutsche Forschungsgemein-
schaft (SFB1035), the Bundesministerium für Bildung und
Forschung (FKZ: 031A131), the European Communityꢀs
Seventh Framework Programme (FP7/2007–2013) under
BioStruct-X (283570), and the Center for Integrated Protein
Science (CIPSM). M.L. was supported by the German
National Academic Foundation. We thank the staff of the
beamline X06SA at the Paul Scherrer Institute, Swiss Light
Source (Villingen, Switzerland) for help with data collection.
Furthermore, we would like to acknowledge Heike Hofmann,
Ernst Bernges, Katja Bäumel, Burghard Cordes, and Mona
Wolff for technical assistance. We are grateful to Elena
Kunold for help with the analysis of quantitative mass
spectrometry data, Pavel Kielkowski for synthesis of the
ADEP fragment, and Megan Wright for critical revision of
the manuscript.
[9] A. Cole et al., Cancer Cell 2015, 27, 864 – 876.
[10] A. Szyk, M. R. Maurizi, J. Struct. Biol. 2006, 156, 165 – 174.
[11] a) M. J. Evans, B. F. Cravatt, Chem. Rev. 2006, 106, 3279 – 3301;
b) P. P. Geurink, L. M. Prely, G. A. van der Marel, R. Bischoff,
H. S. Overkleeft, Top. Curr. Chem. 2012, 308-319, 85 – 113.
[12] a) Z. Li, P. Hao, L. Li, C. Y. J. Tan, X. Cheng, G. Y. J. Chen, S. K.
Sze, H.-M. Shen, S. Q. Yao, Angew. Chem. Int. Ed. 2013, 52,
8551 – 8556; Angew. Chem. 2013, 125, 8713 – 8718; b) V. V.
Rostovtsev, J. G. Green, V. V. Fokin, K. B. Sharpless, Angew.
Chem. Int. Ed. 2002, 41, 2596 – 2599; Angew. Chem. 2002, 114,
2708 – 2711; c) C. W. Tornøe, C. Christensen, M. Meldal, J. Org.
Chem. 2002, 67, 3057 – 3064; d) R. Huisgen, Proc. Chem. Soc.
1961, 357 – 396.
[13] P. J. Boersema, R. Raijmakers, S. Lemeer, S. Mohammed, A. J.
Heck, Nat. Protoc. 2009, 4, 484 – 494.
[14] D. W. Carney, C. L. Compton, K. R. Schmitz, J. P. Stevens, R. T.
Sauer, J. K. Sello, ChemBioChem 2014, 15, 2216 – 2220.
[15] M. Merdanovic, T. Mçnig, M. Ehrmann, M. Kaiser, ACS Chem.
Biol. 2013, 8, 19 – 26.
Keywords: caseinolytic protease · conformational selection ·
non-covalent inhibition · protein crystallography ·
structural rearrangement
[16] E. M. Huber, M. Groll, Angew. Chem. Int. Ed. 2012, 51, 8708 –
8720; Angew. Chem. 2012, 124, 8838 – 8850.
Received: August 4, 2015
Revised: September 15, 2015
Published online: November 13, 2015
How to cite: Angew. Chem. Int. Ed. 2015, 54, 15892–15896
Angew. Chem. 2015, 127, 16121–16126
15896 www.angewandte.org
ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 15892 –15896