Combinatorial nucleoside-deletion-scanning mutagenesis of functional DNA

Article abstract of DOI:10.1002/anie.201208103

Compact and informative: Individual nucleosides are statistically mutated by replacement with a non-nucleosidic spacer unit Δ that encodes the "deletion" in a single DNA library. This efficient mutagenesis approach enables minimization of functional DNA,

Full text of DOI:10.1002/anie.201208103

Angewandte
Chemie
DOI: 10.1002/anie.201208103
Nucleic Acid Chemistry
Combinatorial Nucleoside-Deletion-Scanning Mutagenesis of
Functional DNA**
Biswajit Samanta and Claudia Hçbartner*
Artificial functional DNA molecules, such as deoxyribozymes
and aptamers, find increasing applications in research areas
ranging from biochemistry and molecular engineering to
experimental medicine.[1] Such functional single-stranded
oligonucleotides are generally identified by in vitro selection
from random-sequence DNA pools.[2] The size of the initially
randomized region is mostly arbitrarily chosen, and spans
from as few as 20 nucleotides (nt) to more than 200 nt.
Although DNA sequences with desired activities may initially
be discovered from pools with longer random regions, short
sequences are usually preferred for practical applications.
Computational analyses including sequence-alignment and
structure-prediction methods typically need large datasets to
reliably identify functional motifs. These approaches are often
limited by the absence of partial randomization and reselec-
tion data for reported functional sequences. Advances in this
direction are expected based on increased utilization of next-
generation sequencing techniques in nucleic acid in vitro
selection and screening approaches.[3] However, for the
analysis and optimization of known functional nucleic acids,
an experimentally simple and reliable method for minimiza-
tion and characterization of active sequences is currently not
available.
Herein we present combinatorial nucleoside-deletion-
scanning (NDS) mutagenesis as a new technique to efficiently
distinguish essential from non-essential nucleosides in the
catalytic region of deoxyribozymes, using a single synthetic
DNA library. We expect that this analysis will provide
minimal constructs for crystallization optimization, advance
mechanistic studies of DNA catalysis, and reveal preferred
regions for deoxyribozyme engineering. The method is based
on statistical “deletion” of individual nucleosides in a syn-
thetic DNA library, by replacement of original deoxyribonu-
cleosides with non-nucleosidic spacer units. Conceptually,
nucleoside-deletion-scanning mutagenesis is reminiscent of
alanine-scanning mutagenesis in proteins, which is used to
determine the catalytic or functional role of specific resi-
dues.[4] By replacing individual amino acids with alanine, the
side chain beyond the b carbon atom is eliminated, but the
main-chain conformation is retained. In distant analogy,
herein we eliminate nucleosides in DNA by replacing one
nucleoside per molecule with the acyclic spacer D (Figure 1).
Figure 1. Concept of combinatorial nucleoside-deletion-scanning
(NDS) mutagenesis. a) Comparison of a standard DNA nucleoside
with the cleavable non-nucleosidic spacer D (blue; the gray part
outlines the formally “deleted” portion of the nucleoside). Red arrows
indicate the preferred pathway for cleavage of the phosphodiester
backbone upon alkaline hydrolysis. b) Structure of phosphoramidite 1,
used for incorporation of D. c) Schematic representation of NDS;
parent nucleosides are statistically replaced by D. After separation of
inactive library members, the active fraction contains only those
mutants, in which nucleoside deletion does not interfere with catalysis.
d) Analysis of the interference pattern by denaturing polyacrylamide
gel electrophoresis (PAGE) reveals nucleosides that are sensitive to
replacement by D. DMT=4,4’-dimethoxytrityl, tBDMS=tert-butyldime-
thylsilyl, N=nucleobase.
This spacer moiety is a 1,2,4-butantriol derivative that
resembles an incomplete ribose residue, while the connectiv-
ity (i.e., six bonds between two phosphorous atoms) as well as
the polarity of a standard 3’,5’-connected phosphodiester
backbone are retained. The primary hydroxy group of D,
which formally corresponds to the 2’-OH of a ribonucleoside,
enables specific cleavage of the DNA backbone upon alkaline
hydrolysis and thereby serves as a chemical tag to encode the
nucleoside deletion in the combinatorial DNA library.[5]
Owing to the open-chain structure of D, the flexibility of the
backbone is increased. Thereby possible interference effects
from the 2’-OH group in a cyclic abasic site analogue are
avoided.
[*] B. Samanta, Dr. C. Hçbartner
Research Group Nucleic Acid Chemistry, Max Planck Institute for
Biophysical Chemistry
Am Fassberg 11, 37077 Gçttingen (Germany)
E-mail: claudia.hoebartner@mpibpc.mpg.de
Homepage: http://www.mpibpc.mpg.de/hoebartner
[**] We are grateful to the Max Planck Society and the MPI for
biophysical chemistry for the generous support of our research.
Members of the research group Nucleic Acid Chemistry are
gratefully acknowledged for stimulating discussions. We thank Uwe
Plessmann and Jꢀrgen Bienert from the chemistry and mass
spectrometry facilities at MPIbpc for recording NMR and MS
spectra.
The acyclic spacer D was easily introduced into DNA by
solid-phase synthesis using phosphoramidite 1, which was
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201208103.
Angew. Chem. Int. Ed. 2013, 52, 2995 –2999
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
2995

.
Angewandte
Communications
Scheme 1. Synthesis of phosphoramidite 1 for incorporation of D.
Reagents and conditions: a) DMT-Cl, pyridine, room temperature, 6 h,
99%; b) AD-mix-b, tBuOH, H2O, 08C, 24 h, 73%; c) tBDMS-Cl,
imidazole, CH2Cl2, room temperature, 16 h, 74%; d) (2-cyanoethyl)
N,N-diisopropyl chlorophosphoramidite, Me2NEt, CH2Cl2, RT, 45 min,
75%.
generated in four steps from 3-buten-1-ol (2, Scheme 1). First,
the primary hydroxy group was protected as a dimethoxytrityl
ether to give 3. Sharpless asymmetric dihydroxylation of the
alkene using AD-mix-b afforded the vicinal diol 4 in the
desired R configuration.[6] Silylation of the primary hydroxy
group of 4, followed by phosphitylation of the secondary
alcohol 5, yielded the phosphoramidite 1.
Figure 2. a) DNA-catalyzed synthesis of 2’,5’-branched nucleic acids.
b) 8LV13 deoxyribozyme for synthesis of 2’,5’-branched DNA.[8]
c) 9HR17 deoxyribozyme for connecting an RNA adaptor strand to
a DNA scaffold strand.[12] The first four nucleotides of each adaptor
strand form a base-paired (bp) region with the marked nucleotides of
loop B to generate a three-helix junction structure. LG=leaving group.
Upon mixing each of the four standard DNA nucleoside
phosphoramidites with 1, and using these mixtures in solid-
phase synthesis, combinatorial NDS libraries of deoxyribo-
zymes or any other functional DNA can be generated. The
next steps resemble the analysis performed in other interfer-
ence methods recently reported for characterization of func-
tional DNA, including CoMA and dNAIM.[7] First, active and
inactive library members are separated by electrophoresis or
affinity chromatography, for example, based on the reaction
catalyzed by a DNA enzyme of interest, or binding of a ligand
in the case of aptamers (Figure 1c). The readout of the
nucleoside deletion is then initiated by alkaline hydrolysis,
which leads to cleavage of the DNA strands at the deleted
nucleoside positions (i.e. at sites where D was incorporated),
and is followed by PAGE analysis and quantification of the
interference effect (Figure 1d). Missing hydrolysis bands in
the active fraction suggest that nucleosides at these positions
are essential for DNA function and must be retained. On the
other hand, hydrolysis bands that are present in the active
fraction reveal nucleosides that are dispensable for DNA
activity. Shortened DNA sequences are then generated and
functional assays are used to confirm the combinatorial
screening results. Initially positive variations are then com-
bined to look for synergistic improvements of activity, to
optimize the minimal functional sequence.
For our proof-of-concept demonstration of combinatorial
NDS mutagenesis of functional DNA, we chose two deoxy-
ribozymes that catalyze the synthesis of covalently branched
nucleic acids. These DNA enzymes activate the 2’-hydroxy
group of an internal branch-site nucleotide in the scaffold
strand substrate to form a new phosphodiester bond to the 5’-
terminus of the adaptor strand substrate (Figure 2a).[8]
Scaffold and adaptor strand can each be either DNA or
RNA, resulting in four possible combinations (note: a DNA
scaffold strand must have at least one internal ribonucleotide
that serves as a branch site). Deoxyribozymes have been
identified for each of these four combinations to generate
branched nucleic acids for various applications.[9] The most
prominent combination is 2’,5’-branched RNA, which resem-
bles the core structure of lariat RNA, the product generated
after the first step of RNA splicing.[10] RNA-DNA-branching
deoxyriboyzmes have been used to attach DNA constraints to
ribozymes for the regulation of RNA function, and the DNA-
catalyzed synthesis of 2’,5’-branched DNA has been consid-
ered as construction element for DNA nanotechnology.[11]
Herein, we study two deoxyribozymes that connect different
adaptor strands to the branch-site ribonucleotide of DNA
scaffold strands. The 8 LV13 deoxyribozyme uses 5’-adeny-
lated DNA as the adaptor substrate (Figure 2b),[8] and the
9HR17 deoxyribozyme uses 5’-triphosphorylated RNA to
attach an RNA adaptor to the DNA scaffold (Figure 2c).[12]
Both deoxyribozymes were selected in a structural context
that allows for the formation of a three-helix junction
structure and the formation of two single-stranded loop
regions. Loop A contains 33 nt, and loop B has 7 nt, which
together account for the 40 randomized nucleotides in the
DNA library used for in vitro selection.
We demonstrate that combinatorial NDS mutagenesis can
easily distinguish which of these 40 nt are required for the
activity of the deoxyribozyme, and which ones play no
essential role. The first step entailed the synthesis of the
respective DNA libraries, containing statistically distributed
D residues. To adjust the number of nucleoside deletions to
about one D per molecule, we studied the incorporation
efficiency of phosphoramidite 1 in competition with each of
the four standard DNA phosphoramidites (dN). Pentameric
model oligonucleotides were synthesized with different dN:1
ratios. After deprotection, the products were analyzed by
HPLC, and the product ratio was determined by integrating
the peak area of the UV absorbance trace.[13] The desired
incorporation ratio was achieved by using 30% 1 at a phos-
phoramidite concentration of 100 mm, and a coupling time of
four minutes. Using these conditions, deoxyribozyme libraries
were synthesized, for which the phosphoramidite mixtures
were employed only for loops A and B (the binding arms
were synthesized with standard DNA phosphoramidites
only). After deprotection and purification, the DNA libraries
2996 www.angewandte.org
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2013, 52, 2995 –2999

Angewandte
Chemie
were ligated to the respective adaptor strand substrates, and
then 32P-labeled on the 3’-end. The DNA-catalyzed ligation
reaction was performed with the respective scaffold strand
substrates. Active and non-ligated fractions were isolated by
electrophoresis (after various incubation times), and then
individually subjected to alkaline hydrolysis. The optimal
hydrolysis conditions (0.4m NaOH, 958C, 15 min) were
determined using a model substrate.[13] Owing to the higher
flexibility of the primary hydroxy group in D, harsher
hydrolysis conditions were necessary compared to CoMA
and dNAIM.[7] The nucleoside deletion mutagenesis results
for the 8LV13 deoxyribozyme are depicted in Figure 3, and
the results for 9HR17 are summarized in Figure 4. The
interference results are depicted as bar graphs, with the
nucleotide sequence on the x-axis and the interference value
(ratio of band intensities for each nucleotide in the full library,
lane I, and active fraction, lane II) as y-axis. Large interfer-
ence values indicate detrimental effects, while low values (less
than 2) mean that the deletion mutation is tolerated.
region could be shortened or potentially removed entirely.
Other stretches of non-essential nucleosides were found from
nt 15–20, and from nt 3–5. In loop B, all nucleosides except nt
34 and 35 showed large interference effects. To examine the
effect of removing non-essential nucleotides from the DNA
sequence, several shorter versions of the 8LV13 sequence
were then prepared, and their activity was determined in
kinetic ligation assays (Figure 3c,d). We found that loop A
could be shortened by deleting nt 24–32 (derivative M2)
without affecting the ligation activity; that is, M2 was as active
as the wt DNA. Interesting effects were observed for the
region of nt 15–20, which could not be entirely deleted, nor
replaced by an arbitrary sequence of the same length (see
Table S2 in the Supporting Information), even though the
NDS results suggested that individually replacing any one of
these six nucleotides with D had no effect on the catalytic
activity. The repetitive sequence element TTGGTT repre-
sents an interesting feature of the 8LV13 deoxyribozyme,
which could successfully be replaced by TGT, that is, deleting
one nucleotide of each of the three nucleotide pairs was
tolerated. These results demonstrate that combinatorial NDS
mutagenesis can identify critical nucleosides upon careful
analysis of the sequence context. The shortened 8LV13
variant M3 (with TGT) gave an almost fivefold increase in
kobs. The accelerating effect of this sequence change was also
observed in a second example, as depicted in the summary of
active 8LV13 mutants in Figure S18.[13] More-
A distinct pattern was observed in the hydrolysis lane of
the active 8LV13 library fraction (Figure 3a, lane II), clearly
indicating several missing bands (most prominently from nt
8–14 and nt 21–23 in loop A), suggesting that these regions
are essential for catalysis. In the 3’-region of loop A, from nt
24–33, all hydrolysis bands were present, indicating that these
nucleotides are not directly required, and suggesting that this
over, the DNA binding arm that hybridizes to
the scaffold strand in the 3’ direction of the
branch-site (P1) could be shortened from 9 to
5 nt by arbitrary mutation or deletion of four
nucleotides towards loop A. The shortest
8LV13 variant M6 with a 5 bp stem and
a 21-nt loop A is still about twofold faster
than the wt DNA which has a 9 bp-stem and
a 33 nt loop A. The kobs of this minimized
version of 8LV13 also compares highly favor-
ably with another related and previously
reported DNA-branching deoxyribozyme
(15HA9, which has a 5 bp binding arm and
contains 15 nt in loop A).[11c]
Combinatorial NDS mutagenesis can also
provide information on nucleoside require-
ments under different reaction conditions, for
example, using different metal-ion cofactors
for DNA-catalyzed reactions, as exemplified
herein for the 9HR17 deoxyribozyme. This
DNA enzyme was originally selected to
catalyze the ligation of an RNA adaptor to
a DNA scaffold in the presence of 40 mm
Mg2+ at pH 9.0, but the details of metal-ion
requirements were not analyzed.[12] For sev-
eral other deoxyribozymes that catalyze
RNA cleavage,[14] RNA ligation,[15] or the
formation of nucleopeptide linkages,[12] faster
reaction rates have been reported with Mn2+
rather than Mg2+. We found that 9HR17
performs 16-fold faster ligation of DNA and
RNA substrates when 20 mm Mn2+ is pro-
Figure 3. Combinatorial NDS mutagenesis for 8LV13-catalyzed DNA–DNA ligation.
a) Hydrolysis gel for 33 nt-loop A (full gel in Supporting Information) Hydrolysis samples:
I unselected library, II active fraction after ligation, 20 mm Mn2+, pH 7.5, 378C, 1.5 h.
III unligated fraction. b) Interference values for each nucleotide; green: nucleotides with
interference values less than 2, except nt 15–20 (see text for details), red: values over 2.
c) Selection of mutants and their kobs values. d) Kinetic plots for selected 8LV13 mutants.
e) Minimized 8LV13 DNA (M6). wt=wild type.
Angew. Chem. Int. Ed. 2013, 52, 2995 –2999
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 2997

.
Angewandte
Communications
zyme. These conclusions were confirmed by
kinetic assays of the ligation activity using
several 9HR17 derivatives (Figure 4d,e) in
which either the 5’ or the 3’ part of loop A
was truncated. For several mutants we found
that shortening the loop sequence was better
tolerated in presence of Mn2+ than Mg2+, as can
be deduced from the ratios of kobs values listed
in Table S4.[13] As in the case of 8 LV13 variants,
we found that the minimized version of 9HR17
(M6) did not require the 9 bp interaction with
the RNA substrate in P1. Indeed, by shortening
this region to 5 bp in the context of the
minimized loop (M7), the kobs improved about
twofold and compared well with the kobs of
wt 9HR17 (Figure 4d,e) demonstrating that
this shortened DNA is a highly efficient
deoxyribozyme.
In summary, we have developed combina-
torial NDS mutagenesis as a simple and
effective method to distinguish essential from
non-essential nucleosides in the catalytic loop
regions of deoxyribozymes. This new combina-
torial approach is less laborious compared to
systematic replacement of nucleotides by an
abasic site analogue or an alkyl linker in
individual DNA strands, which is practicable
only for a small number of nucleotides, such as
recently reported for the RNA-cleaving 10–23
and 8–17 deoxyribozymes.[16] For the DNA
catalysts analyzed in this study, we found that
a significantly smaller number of nucleotides
than originally present in the random pool is
sufficient to support efficient catalysis. This
finding is of general interest for future in vitro
selection experiments and may stimulate fur-
ther investigations on the correlation between
nucleic acid catalysis and random sequence
length.[17]
Figure 4. Combinatorial NDS mutagenesis of 9HR17-catalyzed DNA–RNA ligation.
a) Hydrolysis gel for 33 nt-loop A (full gel in Supporting Information). Hydrolysis
samples: I unselected library, II active fraction after ligation, 40 mm Mg2+, pH 9.0, 378C,
1.5 h. III unligated fraction. b,c) Interference values for ligation with Mg2+ (b) and
Mn2+ (c); nucleotides with interference values less than 2 are green, those with values
over 2 are red; purple indicates differences in presence of Mg2+ and Mn2+. d) Selection
of truncated mutants and their kobs values and final ligation yields (after 3 h reaction
time in 20 mm Mn2+). e) Kinetic plots for selected 9HR17 mutants. f) Schematic
diagram of minimized 9HR17 with 5 nt in P1 and essential nucleotides of loop A
highlighted. The green lines indicate several options for productively positioning the
core nucleotides.[13]
Combinatorial NDS mutagenesis repre-
sents a novel, resource- and time-efficient
approach for deoxyribozyme characterization
and engineering and will be easily adaptable to
other functional DNAs, including ligand-bind-
ing DNA aptamers. The most obvious practical
vided at pH 7.5, compared to the original conditions. There-
fore, we analyzed the NDS interference pattern of 9HR17 for
ligation with each of these divalent metal ions. At both ends of
loop A, at least 9 nt were not essential for catalysis with either
Mg2+ or Mn2+ (Figure 4b,c). A central region of six nucleo-
tides (nt 18–23) did not tolerate the deletion when the ligation
was performed in the presence of Mn2+. The critical region
was three nucleotides longer in the presence of Mg2+ (purple
nucleotides in Figure 4d). Since only a few nucleotides were
essential in the center of loop A, only one of the flanking
regions could be deleted while maintaining the proper
positioning of the central nucleotides. Moreover, our results
indicate that the predicted stem-loop in loop A (Figure 2c) is
not relevant for the active conformation of the deoxyribo-
advantage of NDS compared to CoMA[7a] lies in the faster
data collection, since only a single synthetic library is
required, and several functional DNAs can easily be analyzed
in parallel and under different reaction conditions. The results
of the NDS interference analysis narrow down the number of
nucleotides for structural analyses. Shorter sequences are less
prone to misfolding, which increases the probability for
crystallization. For NMR spectroscopy, minimized systems
are favorable for improved resonance separation and reduced
cost for isotope-labeling. Furthermore, by identifying nucle-
otides that cannot be replaced by D, NDS directly maps
crucial residues that are promising to reveal mechanistic
insights on DNA catalysis upon further analyses, for example,
via dNAIM,[7b] in which a large number of combinatorial
2998 www.angewandte.org
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2013, 52, 2995 –2999

Angewandte
Chemie
[5] See Supporting Information for a statement on the choice of the
libraries are synthesized using precious modified ribonucleo-
tides. In a practical sense, the NDS results directly suggest
permissive sites for insertion of regulatory elements or
replacement of non-essential nucleotides with external func-
tional units. The minimized deoxyribozymes offer ample
opportunities for further engineering, for example in combi-
nation with aptamers to build aptazymes for the development
of sensors and as construction elements in DNA nano-
technology.
butantriol-derived spacer D.
[6] a) K. B. Sharpless, W. Amberg, Y. L. Bennani, G. A. Crispino, J.
Hartung, K. S. Jeong, H. L. Kwong, K. Morikawa, Z. M. Wang, J.
Org. Chem. 1992, 57, 2768; b) H. C. Kolb, M. S. VanNieuwenhze,
K. B. Sharpless, Chem. Rev. 1994, 94, 2483.
[7] a) F. Wachowius, F. Javadi-Zarnaghi, C. Hçbartner, Angew.
Chem. 2010, 122, 8682; Angew. Chem. Int. Ed. 2010, 49, 8504;
b) F. Wachowius, C. Hçbartner, J. Am. Chem. Soc. 2011, 133,
14888.
[8] C. S. Lee, T. P. Mui, S. K. Silverman, Nucleic Acids Res. 2011, 39,
269.
[9] C. Hçbartner, P. I. Pradeepkumar in New strategies in chemical
synthesis and catalysis (Ed.: B. Pignataro), Wiley-VCH, Wein-
heim, 2012, p. 129.
Received: October 8, 2012
Revised: January 3, 2013
Published online: February 1, 2013
[10] a) R. L. Coppins, S. K. Silverman, Nat. Struct. Mol. Biol. 2004,
11, 270; b) S. K. Silverman, Acc. Chem. Res. 2009, 42, 1521.
[11] a) E. Zelin, S. K. Silverman, ChemBioChem 2007, 8, 1907; b) E.
Zelin, S. K. Silverman, Chem. Commun. 2009, 767; c) T. P. Mui,
S. K. Silverman, Org. Lett. 2008, 10, 4417.
[12] P. I. Pradeepkumar, C. Hçbartner, D. A. Baum, S. K. Silverman,
Angew. Chem. 2008, 120, 1777; Angew. Chem. Int. Ed. 2008, 47,
1753.
[13] See the Supporting Information for further details.
[14] a) R. P. G. Cruz, J. B. Withers, Y. Li, Chem. Biol. 2004, 11, 57;
b) A. R. Feldman, E. K. Leung, A. J. Bennet, D. Sen, Chem-
BioChem 2006, 7, 98.
Keywords: deoxyribozymes · DNA catalyst ·
interference analysis · mutagenesis · solid-phase synthesis
.
[1] a) I. Willner, B. Shlyahovsky, M. Zayats, B. Willner, Chem. Soc.
Rev. 2008, 37, 1153; b) M. Famulok, G. Mayer, Acc. Chem. Res.
2011, 44, 1349; c) B. Strehlitz, C. Reinemann, S. Linkorn, R.
Stoltenburg, Bioanal. Rev. 2012, 4, 1; d) S. K. Silverman, Angew.
Chem. 2010, 122, 7336; Angew. Chem. Int. Ed. 2010, 49, 7180;
e) M. Ye, J. Hu, M. Peng, J. Liu, H. Liu, X. Zhao, W. Tan, Int. J.
Mol. Sci. 2012, 13, 3341.
[15] R. L. Coppins, S. K. Silverman, J. Am. Chem. Soc. 2005, 127,
2900.
[16] B. Wang, L. Cao, W. Chiuman, Y. Li, Z. Xi, Biochemistry 2010,
49, 7553.
[17] An experimental study on that topic was published while this
manuscript was under review: T. E. Velez, J. Singh, Y. Xiao, E. C.
Allen, O. Y. Wong, M. Chandra, S. C. Kwon, S. K. Silverman,
ACS Comb. Sci. 2012, 14, 680.
[2] a) K. Schlosser, Y. Li, Chem. Biol. 2009, 16, 311; b) G. F. Joyce,
Annu. Rev. Biochem. 2004, 73, 791; c) G. F. Joyce, Angew. Chem.
2007, 119, 6540; Angew. Chem. Int. Ed. 2007, 46, 6420.
[3] a) J. Hoinka, E. Zotenko, A. Friedman, Z. E. Sauna, T. M.
Przytycka, Bioinformatics 2012, 28, i215; b) T. Schꢀtze, B.
Wilhelm, N. Greiner, H. Braun, F. Peter, M. Morl, V. A.
Erdmann, H. Lehrach, Z. Konthur, M. Menger, P. F. Arndt, J.
Glokler, PLoS One 2011, 6, e29604.
[4] a) B. C. Cunningham, J. A. Wells, Science 1989, 244, 1081;
b) K. L. Morrison, G. A. Weiss, Curr. Opin. Chem. Biol. 2001,
5, 302.
Angew. Chem. Int. Ed. 2013, 52, 2995 –2999
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 2999