Discovery of small-molecule interleukin-2 inhibitors from a DNA-encoded chemical library

Article abstract of DOI:10.1002/chem.201200952

Libraries of chemical compounds individually coupled to encoding DNA tags (DNA-encoded chemical libraries) hold promise to facilitate exceptionally efficient ligand discovery. We constructed a high-quality DNAencoded chemical library comprising 30 000 drug-like compounds; this was screened in 170 different affinity capture experiments. High-throughput sequencing allowed the evaluation of 120 million DNA codes for a systematic analysis of selection strategies and statistically robust identification of binding molecules. Selections performed against the tumor-associated antigen carbonic anhydrase IX (CA IX) and the pro-inflammatory cytokine interleukin-2 (IL-2) yielded potent inhibitors with exquisite target specificity. The binding mode of the revealed pharmacophore against IL-2 was confirmed by molecular docking. Our findings suggest that DNA-encoded chemical libraries allow the facile identification of drug-like ligands principally to any protein of choice, including molecules capable of disrupting high-affinity protein-protein interactions.

Full text of DOI:10.1002/chem.201200952

FULL PAPER
DOI: 10.1002/chem.201200952
Discovery of Small-Molecule Interleukin-2 Inhibitors from a DNA-Encoded
Chemical Library
Markus Leimbacher,[a] Yixin Zhang,[a] Luca Mannocci,[b] Michael Stravs,[a]
Tim Geppert,[a] Jçrg Scheuermann,[a] Gisbert Schneider,[a] and Dario Neri*[a]
Abstract: Libraries of chemical com-
pounds individually coupled to encod-
ing DNA tags (DNA-encoded chemical
libraries) hold promise to facilitate ex-
ceptionally efficient ligand discovery.
We constructed a high-quality DNA-
encoded chemical library comprising
30000 drug-like compounds; this was
screened in 170 different affinity cap-
ture experiments. High-throughput se-
quencing allowed the evaluation of
120 million DNA codes for a systemat-
ic analysis of selection strategies and
statistically robust identification of
binding molecules. Selections per-
formed against the tumor-associated
tokine interleukin-2 (IL-2) yielded
potent inhibitors with exquisite target
specificity. The binding mode of the re-
vealed pharmacophore against IL-2
was confirmed by molecular docking.
Our findings suggest that DNA-encod-
ed chemical libraries allow the facile
identification of drug-like ligands prin-
cipally to any protein of choice, includ-
ing molecules capable of disrupting
high-affinity protein–protein interac-
tions.
antigen
carbonic
anhydrase IX
(CA IX) and the pro-inflammatory cy-
Keywords: DNA-encoded chemical
libraries · drug discovery · high-
throughput screening · inhibitors ·
ligands
Introduction
istics (e.g., molecular weight >500 Da, clogP>5, polar sur-
face area >140 ꢀ2) and can thus display suboptimal ligand
efficiency and drug-like properties.[17–19] The quality assur-
ance (e.g., purity, reaction yields, equimolarity among li-
brary members) of libraries synthesized by multiple chemi-
cal reactions also remains difficult.
We constructed a DNA-encoded chemical library based
on the split-and-pool assembly of only two sets of building
blocks (Scheme 1a); this allows both the synthesis of drug-
like molecules and thorough quality control during library
construction. The resulting library comprising 30000 DNA-
encoded compounds (DEL30000 library) was systematically
tested in affinity selections against streptavidin and human
carbonic anhydrase IX (CA IX), in order to assess the quali-
ty of library synthesis and investigate biopanning strategies.
Subsequently, optimized protocols were applied in an affini-
ty screening campaign against commercially available pro-
teins; these led to the discovery of potent, highly selective
and biologically active ligands against human interleukin-2
(IL-2).
There is a need for innovative technologies that overcome
the limitations in time, costs and logistics associated with
the conventional high-throughput screening of chemical li-
braries for hit discovery programs.[1–3] To this aim, the tag-
ging of chemical compounds with unique DNA fragments
serving as identification barcodes facilitates the construction
of encoded chemical libraries.[4–12] In affinity selection ex-
periments, the library is exposed to a target protein of
choice, which is immobilized on a solid support.[13,14] After
equilibration, nonbinders are removed by washing whereas
binders are eluted separately upon protein denaturation.
The recovered DNA codes are amplified by PCR, subjected
to Illumina high-throughput sequencing and decoded by
computational algorithms.[15,16] The most abundant com-
pounds are selected for resynthesis and hit validation.
Encoded combinatorial repertoires are typically assem-
bled by multiple split-and-pool procedures; this leads to mo-
lecular entities with unfavorable physicochemical character-
[a] M. Leimbacher, Dr. Y. Zhang, M. Stravs, Dr. T. Geppert,
Dr. J. Scheuermann, Prof. G. Schneider, Prof. D. Neri
Department of Chemistry and Applied Biosciences
Institute of Pharmaceutical Sciences
Results and Discussion
Library design and construction: In our library construction
protocol, 100 different N-protected amino acids were cou-
pled to 100 distinct amino-tagged oligonucleotides by stand-
ard amide bond forming reactions. The oligonucleotides dif-
fered from each other in a central sequence of 9 nucleotides,
which served as code for the identification of the chemical
moiety linked to the DNA fragment (yellow “code A”). The
general structures of the 100 amino acids used for library
Wolfgang Pauli-Str. 10, 8093 Zꢁrich (Switzerland)
Fax : (+41)44-633-13-58
E-mail: Dario.Neri@pharma.ethz.ch
[b] Dr. L. Mannocci
Philochem AG, Libernstrasse 3
8112 Otelfingen (Switzerland)
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.201200952.
Chem. Eur. J. 2012, 18, 7729 – 7737
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
7729

Scheme 1. Library construction strategy. a) Schematic representation of the synthesis and encoding of a 30000 compound library. i) Coupling of 100 dif-
ferent N-protected amino acids (N-Fmoc- or N-NVOC-protection, Table S1 in the Supporting Information) to 100 distinct 5’-amino-modified oligonu-
cleotides by amide bond formation. The reaction was mediated by EDC and S-NHS in DMSO/H2O, TEA-Cl (75 mm), pH 10 (308C, 16 h) and quenched
with Tris-HCl (30 mm, pH 8; 308C, 1 h). ii) Equimolar pooling of HPLC-purified oligonucleotide–compound conjugates (Table S2 in the Supporting In-
formation) and N-NVOC deprotection by UV (l=366 nm, 48C, 90 min) and N-Fmoc deprotection by piperidine (500 mm, 48C, 2 h). iii) Equimolar split
of the resulting sublibrary. iv) Coupling of 300 methylene carboxylic acids (Table S3 in the Supporting Information) on solid-phase by amide bond forma-
tion (EDC-HCl, HOAt, DMF, 258C, 3 h, on DEAE-resin for DNA immobilization by ionic interactions). v) Partial annealing of 300 different corre-
sponding oligonucleotides and enzymatic encoding by Klenow polymerization. vi) Pooling in equimolar amounts to yield the final 30000-membered
DNA-encoded chemical library (DEL30000 library). b) Structural scaffolds of chiral library building blocks (numbers in brackets) employed in the li-
brary construction. Alkynylbenzenes were obtained by Sonogashira coupling, phenyl ethers by Mitsunobu reaction, biphenyls by Suzuki coupling, and
the triazole derivatives by azide–alkyne cycloaddition by using Sharpless conditions. The two sets of building blocks were completed by a reaction with-
out building block in order to monitor the performance of the single fragments during selection experiments.
construction (synthesized by using Suzuki, Sonogashira, Mit-
sunobu reactions or azide alkyne cycloadditions under
Sharpless conditions) are illustrated in Scheme 1b and
Table S1 in the Supporting Information. The 100 oligonu-
cleotide–compound conjugates (Table S2 in the Supporting
Information) were individually HPLC purified, pooled in
equimolar amounts, deprotected and transferred to aliquots
in 300 different vessels for a second reaction step, involving
300 different carboxylic acids (Table S3 in the Supporting
Information). These reagents were chosen out of more than
600 methylene carboxylic acids, which have been shown in
model reactions on solid-phase to give rise to the desired
amides with sufficiently high yields.
(Scheme 1). Analysis of the physicochemical properties of
the library compounds showed that most of the molecules
(78%) had molecular weight <500 Da (and 93% were
<550 Da). Approximately 77% of the compounds had
clogP<5, H-bond acceptors <10 (90%), H-bond donors
<5 (99%) and polar surface area <140 ꢀ2 (92%), and
thereby fulfilled the generally accepted criteria for drug-
likeness (Figure S1 in the Supporting Information). Com-
pounds were pooled at a final total concentration of 360 nm
(12 pm for each compound) and stored in frozen aliquots,
which were subsequently thawed and used for selections
against target proteins of pharmaceutical relevance.
For 270 out of 300 compounds, conversion yields were
>70%, whereas for 30 compounds the amide bond was
formed with a yield between 40–70% (Table S3 in the Sup-
porting Information). After amide bond formation, the iden-
tity of the carboxylic acids was encoded, as previously de-
scribed,[20] by using specific oligonucleotides differing from
each other in a central sequence of 8 nucleotides (green
“code B”). Finally, a polymerization reaction catalyzed by
Klenow polymerase yielded the final set of 30000 com-
pounds covalently attached to the cognate double-stranded
DNA fragments, serving as identification barcodes
Optimization of affinity selection protocols: As little is
known about the impact of experimental conditions on af-
finity selection results, we performed a systematic parameter
analysis (e.g., protein amount, washing stringency, detergent
concentration, and protein immobilization strategy) using
human CA IX as protein antigen (Figure 1 and Figure S3 in
the Supporting Information). CA IX is a protein of consider-
able pharmaceutical interest,[21–23] as it is strongly over-ex-
pressed in hypoxic tissues and in several tumor types, where-
as its expression in normal tissues is mainly confined to the
gastrointestinal mucosa. Pharmacological inhibition of
7730
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 7729 – 7737

Small-Molecule Interleukin-2 Inhibitors
FULL PAPER
Figure 1. Influence of varying experimental conditions on readouts in affinity selections against human CA IX. Selections were performed by using poly-
histidine-tagged CA IX immobilized on magnetic cobalt beads (Dynabeads). Histograms plots represent the 30000 library members in the x–y plane and
the corresponding normalized relative frequencies (rf) after selection on the z axis. The expected value for normalized relative frequencies of library
members before selection equals 1. The numbers of analyzed sequences (n) are indicated. Selection results by using: a) 40 mg, b) 8 mg, or c) 2 mg human
CA IX show an increasing discrimination between binders (i.e., phenyl sulfonamides A1–100B151 and A1–100B170) and nonbinders. This effect was further en-
hanced by using higher detergent concentration (0.01% (a, b) or 0.1% (c) Tween-20). The washing stringency in the illustrated selections (a–c) remained
constant (i.e., six washing rounds with 100 mL PBS). d) Summary of the impact of protein amount and detergent concentration on selection results. De-
tailed analyses of the systematic investigation of selection conditions (i.e., protein amount, detergent concentration, protein immobilization strategy, and
washing stringency) are illustrated in Figure S3 in the Supporting Information.
CA IX activity has been reported to reduce tumor growth
rate and metastatic spread in animal models of cancer.[24–26]
We found that modifications in the biopanning protocol
critically influence selection readouts. Figure 1 illustrates the
impact of target protein amount (ranging from 40 to 0.2 mg)
and detergent concentration (ranging from 0.01 to 0.1%
Tween-20) on selection fingerprints. The individual library
compounds were unambiguously identified as elements in
the x–y plane by the values of code A and B (corresponding
to the identity of chemical building blocks), whereas their
relative frequencies (normalized counts of the correspond-
ing DNA sequence) are displayed on the z axis. The use of
higher detergent concentrations and of lower protein
amounts substantially improved signal-to-noise ratios in bio-
panning reactions, indicated by an enhanced enrichment of
phenyl sulfonamide compounds with code B=151 and 170
(Figure 1a–c). Very stringent washing protocols led to slight-
ly worse discrimination between binders and nonbinders
(Figure S3 in the Supporting Information). Importantly, all
200 phenyl sulfonamide derivatives within the DEL30000 li-
brary were enriched as a result of the selection; this reflects
the covalent coupling of the phenyl sulfonamide groups to
the complete set of first chemical building blocks.
Ligand discovery: The findings obtained in the systematic
investigation of selection protocols were implemented in an
affinity screening campaign against several commercially
available proteins. Selections against streptavidin, human
CA IX and human IL-2 allowed both a robust process vali-
dation and comparative studies of selection readouts and
target affinities including target specificity.
After selection, eluted DNA–compound conjugates were
subjected to a second round of selection on immobilized
streptavidin. The resulting decoding profile exhibited an en-
hanced discrimination among compounds with code B=96,
compared to the first selection round (Figure S2 in the Sup-
porting Information). The preferential loss of compounds
with code A chemical moieties corresponding to proline de-
rivatives obtained by azide–alkyne cycloaddition, reflected a
consistent behavior of structurally related compounds in se-
lection experiments. The decoding profiles for panning reac-
tions by using quantities of immobilized streptavidin ranging
from 1 mg to 1 ng were very similar; this provides additional
evidence for the robustness of the selection results.
Figure 2c and d show decoding profiles of selections
against human CA IX and human IL-2. The latter is a repre-
sentative target involved in high-affinity protein–protein in-
teraction with its cognate receptor and is a pro-inflammato-
Chem. Eur. J. 2012, 18, 7729 – 7737
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7731

D. Neri et al.
Figure 2. High-throughput sequencing results before and after selection against three different targets. Histograms plots represent the 30000 library
members in the x–y plane and the corresponding normalized relative frequencies (rf) after selection on the z axis. The expected value for normalized rel-
ative frequencies of library members before selection equals 1. The numbers of analyzed sequences (n) are indicated. a) Library sequencing before selec-
tion showed equally distributed frequencies of library members. b) Streptavidin selection results exhibited a strong enrichment of all 100 d-desthiobiotin
derivatives (code B=96, rf values >44-times expected value). c) Selection results against human CA IX showed an enrichment for all 200 phenyl sulfo-
namide derivatives in the DEL30000 library (code B=151 or 170, rf values >8-times expected value). d) Selection results against human IL-2 revealed
an N-substituted 2-methyl-1H-indole moiety (code B=284, rf values >14-times expected value) being strongly overrepresented.
ry cytokine of considerable pharmaceutical interest.[27–29] IL-
2 selection results reveal a striking preferential enrichment
of compounds with code B=284, corresponding to an N-
substituted 2-methyl-1H-indole moiety.
tion of building block A (Figure 3c). The fluorescein deriva-
tives of B151 and B170 did not exhibit any affinity against
streptavidin or IL-2 (Figure 3b and d).
Affinities against human IL-2 (Proleukinꢃ) for two of the
most enriched compounds (A15B284 and A17B284) were meas-
ured by fluorescence polarization (Figure 3d). The ligands
Hit validation and specificity analysis: Some of the most
abundant molecules after affinity selections were chosen for
hit validation and specificity analyses (Figure 3). The affini-
ties of two representative hits from streptavidin selections
were confirmed by fluorescence polarization (Figure 3b)
and tested for specificity in CA IX inhibition and IL-2 bind-
ing assays (Figure 3c and d).
exhibited Kd values of 4.2
N
U
cellent specificity (Figure 3b and c). A fluorescein-derivative
of the 2-methyl-1H-indole moiety exhibited a Kd value of
>60 mm, confirming that building block A contributed to IL-
2 recognition. Importantly, the selected 2-methyl-1H-indole
moiety was not found to be enriched in any of the other 170
selections performed with the DEL30000 library against a
set of target proteins under various experimental conditions
(Figure 4a). Furthermore, compounds A15B284 and A17B284
exhibited a complete and selective inhibition of IL-2 activity
in a T-cell proliferation assay (Figure 4b) and exhibited IC50
values of 71(Æ9) and 32(Æ3) mm. The same compounds did
not exhibit any cytotoxicity to primary fibroblast cell cul-
tures up to 128 mm concentration (Table S5 in the Supporting
Information).
Two strongly enriched phenyl sulfonamides in CA IX se-
lections (A69B151 and A6B170) inhibited CA IX activity with
Ki values of 76
N
U
hibitory constants of selected compounds were confirmed by
competition assays by using fluorescence polarization
(Table S4 in the Supporting Information). Phenyl sulfona-
mides have previously been reported to bind various car-
bonic anhydrases with dissociation constants in the nanomo-
lar range.[21,22] However, our findings suggest that omission
of building block A (e.g., compounds B151 or B170) led to
worse inhibitory constants of Ki =560
ACHTUNGTRENNUNG
(Æ130) and 230-
Molecular docking: In order to gain insight about the bind-
ing mode on the IL-2 surface, all hundred 2-methyl-1H-
7732
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 7729 – 7737

Small-Molecule Interleukin-2 Inhibitors
FULL PAPER
Figure 3. Hit validation and specificity analysis of chemical compounds isolated from the DEL30000 library in affinity capture experiments. a) Chemical
structures of test compounds. b), d) Dissociation constants against streptavidin or human IL-2 (Proleukinꢃ) were determined by fluorescence polarization.
Individual test compounds were synthesized as fluorescein conjugates and incubated with different protein concentrations. c) Inhibitory constants against
human CA IX were measured by monitoring the hydrolytic activity of CA IX by using a colorimetric CA IX assay. The test compounds were measured
in three different in vitro assays in order to investigate the efficacy and specificity of binding or inhibition.
Figure 4. Specificity analysis and T-cell growth rate inhibition of revealed IL-2 ligands. a) Histogram plot representing all 100 compounds with code B=
284 (2-methyl-1H-indole derivatives, A1–100B284) in the DEL30000 library on the x axis, 170 different selection experiments on the y axis, and the corre-
sponding normalized relative frequencies (rf) on the z axis. The 170 selections were performed against 36 different proteins under varying selection con-
ditions. The 2-methyl-1H-indole fragments were recovered with high specificity in selection #114 against human IL-2. b) IL-2-mediated (Proleukinꢃ) T-
cell proliferation assay. The 2-methyl-1H-indole derivatives depicted in Figure 4a (A15B284 and A17B284) inhibited IL-2-mediated T-cell proliferation with
IC50 values of 71(Æ9) or 32(Æ3) mm, whereas the other test compounds showed no or very moderate inhibition (i.e., IC50 >400 mm for A53B96) of T-cell
proliferation.
indole derivatives enriched in the corresponding library se-
lection were submitted to a computational docking proce-
dure, based on the experimental crystal structure of IL-2
(PDB ID: 1M48)[30,31] and on molecular docking by using
GOLD (Version 5.1).[32] The 2-methyl-1H-indole moieties of
67 out of 100 compounds were positioned in a hydrophobic
Chem. Eur. J. 2012, 18, 7729 – 7737
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7733

D. Neri et al.
Figure 5. Molecular docking of A17B284 against human IL-2. a) Predicted docking pose of compound A17B284 at the IL-2 surface. Protein surface represen-
tation of IL-2 (PDB ID: 1M48).[30] Interface binding pockets are represented by blue shaded dots (darker color indicates greater buriedness). The ligand
docking pose occupies both pockets. b) Specific interactions between A17B284 and IL-2. Orange: hydrogen bond; blue: cation–p interaction; yellow: CH–
p interaction. The interactions were determined by using MOE Version 2010.10 and depicted by PyMOL Version 1.2r2 Mac. The corresponding figure
for compound A15B284 is provided as Figure S4 in the Supporting Information.
groove between Arg38 and Leu72 (Figure 5 and Figure S4
in the Supporting Information), an IL-2 site with a strong
propensity for binding aromatic or fused ring structures.[30]
Interestingly, Wells and co-workers had previously identi-
fied a structurally different indole glyoxylate substructure
from a tethering fragment library of approximately 7000
compounds and solved the X-ray structure of IL-2 with the
tethered fragment.[30] The N-substituted 2-methyl-1H-indole
ring system of A15B284 and A17B284 is positioned in the same
hydrophobic groove by molecular docking (Figure 5 and
Figure S4 in the Supporting Information) and forms a
cation–p interaction with Arg38 and three CH–p interac-
tions with Leu72. The ethylene spacer between the 2-
methyl-1H-indole moiety and the peptide backbone of the
ligands allowed hydrogen bond interactions to be assigned
to both carbonyl groups with Arg38 as well as between the
amide nitrogen and the hydroxyl side chain of Thr41. The li-
gands were docked around Phe42 and Leu72, a hydrophobic
binding hot-spot in the contact patch of IL-2 and its a recep-
tor unit.[31] For only ten compounds the docking algorithm
positioned the 2-methyl-1H-indole moiety in a pocket be-
tween Phe42 and Tyr45.
Superimposed docking poses of 57 out of 67 positioned 2-
methyl-1H-indole derivatives show the respective indole
moieties being oriented likewise with the methyl group
pointing towards the IL-2 surface (Figure S5 in the Support-
ing Information). Consistently, the ethylamine linker repre-
senting the attachment site for the DNA barcode or the flu-
orescein label projects towards the solvent (Figure 5b and
Figure S4 in the Supporting Information).
Specificity analyses of 164 selections revealed two struc-
turally closely related building blocks B being mostly en-
riched besides building block B284 in the presented IL-2 se-
lection, that is, an N-substituted 2-methyl-1H-benzo[d]imi-
dazole moiety (B277) and a 2-methyl-benzene moiety (B213),
both linked by an ethylene spacer to the peptide backbone
(Table S6 in the Supporting Information).
Conclusion
Little is known about how experimental conditions influence
the performance of DNA-encoded chemical libraries in bio-
panning reactions. We observed favorable signal-to-noise
ratios using submicrogram amounts of immobilized protein
and high detergent concentrations in affinity capture experi-
ments. The systematic investigation of selection protocols
led to substantial methodological progress by using DNA-
encoded chemical libraries as a versatile tool for ligand dis-
covery.
The implementation of optimized selection protocols and
the library design employed in our study allowed the discov-
ery of readily synthesizable drug-like ligands. For instance,
compound A17B284 represents a previously unknown and bio-
logically active binder against human IL-2, and thereby pro-
vides an attractive starting point for hit-to-lead develop-
ment. Furthermore, the exquisite antigen specificity of en-
riched library members in selection readouts was confirmed
in four different hit validation assays.
In our experiments, the number of sequence tags read for
each selection ranged between 105 and 106 (Figures 1 and 2).
In principle, the same library construction strategy could be
performed with a larger number of building blocks, reaching
approximately 106 compounds (i.e., a library size that is
compatible with suitable sampling of DNA codes, by using
high-throughput sequencing technologies).
The stepwise coupling of more than two sets of building
blocks can yield larger libraries.[7,33,34] However, almost in-
evitably, such a procedure is associated with the synthesis of
more complex library members and thus could violate gen-
erally accepted rules for drug-likeness,[7,17,33,35] which in turn
entails increased predicted risks of developability and their
compatibility with chemical optimization procedures.[19] As
it is difficult to ensure quality (e.g., purity, reaction yields,
equimolarity of compounds) on very large combinatorial li-
braries assembled by multiple chemical reactions, we antici-
7734
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 7729 – 7737

Small-Molecule Interleukin-2 Inhibitors
FULL PAPER
adding 1m HCl (110 mL), filtered over Millexꢃ-FG 0.2 mm, purified by
HPLC and the desired fractions were evaporated to dryness.
pate that DNA-encoded chemistry will increasingly focus on
the synthesis of compounds based on two sets of building
blocks,[6,35] as illustrated in the examples presented in this
study.
Oligonucleotide-compound coupling reactions: The respective NVOC- or
Fmoc-protected amino acid (1.25 mmol, 1.0 equiv, dissolved in DMSO)
was diluted with DMSO to a reaction volume of 230 mL and activated
with S-NHS (10 mL of a 330 mm stock solution in DMSO/H2O 2:1,
3.3 mmol, 2.7 equiv) and EDC (12 mL of a 100 mm stock solution in
DMSO, 1.20 mmol, 0.96 equiv) for 20 min at 308C. Afterwards, 500 mm
aqueous TEA/HCl buffer (50 mL, pH 10.0) and 15 nmol of the corre-
sponding 5’-amino-modified oligonucleotide (5’-H2N-(CH2)12-GGAGCT
TGTGAA TTCTGG NNNNNNNNN GGACGT GTGTGA ATTGTC-
3’, where NNNNNNNNN unambiguously identifies the corresponding N-
protected amino acid compound; 30 mL of a 500 mm stock solution) were
added to the reaction mixture and stirred, overnight, at 308C. Residual
activated species were quenched by adding 500 mm aqueous Tris-HCl
(pH 8.0, 20 mL) and stirring the reaction mixture for 1 h at 308C. Reac-
tions were subsequently purified by HPLC and product-containing frac-
tions were dried under reduced pressure, redissolved in H2O (100 mL),
quantified by UV, and characterized by oligonucleotide-LC-ESI-MS.
Experimental Section
Resin preparation: H-b-Ala-2-ClTrt resin (800 mg, 0.58 mmol substitution
capacity, 1.0 equiv) was added to a solution of the respective N-protected
amino acid (NVOC-l-Tyr-OH or NVOC-d-Tyr-OH, 370 mg, 0.87 mmol,
1.5 equiv) or NVOC-p-I-l-Phe-OH or NVOC-p-I-d-Phe-OH, 460 mg,
0.87 mmol, 1.5 equiv) in dry DMF (20 mL) and slowly stirred for 30 min
at 258C. Afterwards, HBTU (660 mg, 1.74 mmol, 3.0 equiv) was dissolved
in dry DMF (20 mL) and was added to the first solution. Then, DIPEA
(1.0 mL, 5.8 mmol, 10 equiv) was added to the reaction vessel and slowly
stirred, overnight, at 258C. The reaction mixture was transferred into a
polypropylene filter syringe, washed five times with DMF, twice with
CH2Cl2, and dried in an exsiccator.
Coupling of the second chemical building blocks on solid-phase: DEAE
Sepharose resin (100 mL) was transferred to a SpinX column and washed
with H2O (0.8 mL) and HOAc (0.8 mL, 10 mm, pH 5). The resin was re-
suspended in HOAc (200 mL, 10 mm, pH 5) and an aliquot of the DNA
conjugate sublibrary (360 pmol) was then added and immobilized on the
resin by incubation for 15 min. Afterwards, the slurry was transferred to
an empty Glen Research column and washed with HOAc (1 mL, 10 mm)
and MeOH (2 mL).
Aryl alkyl ether formation (Mitsunobu reaction): PPh3 (66 mg, 250 mmol,
1.0 equiv) was dissolved in dry THF (500 mL) and was added to resin-
bound NVOC-l-Tyr-OH or NVOC-d-Tyr-OH (50 mg, 35 mmol,
0.14 equiv) and stirred for 30 min at 258C under an argon atmosphere
and light protection. The respective primary or secondary alcohol
(500 mmol, 2.0 equiv) was dissolved in dry THF (500 mL), added to the re-
action mixture and stirred for 30 min at 258C. Afterwards, DEAD
(46 mL, 50 mg, 290 mmol, 1.2 equiv) was added to the reaction vessel and
stirred, overnight, at 258C. Then, the reaction mixture was transferred
into a polypropylene filter syringe, washed five times with DMF, twice
with CH2Cl2 and released by resin-cleavage by using TFA (2%) in
CH2Cl2 for 1 h 258C. The crude product was filtered over Millexꢃ-FG
0.2 mm, purified by HPLC and the desired fractions were evaporated to
dryness.
Afterwards, resin-bound DNA conjugates were exposed to a freshly pre-
pared solution of 50 mm EDC-HCl (500 mL, 1.0 equiv), 5 mm HOAt
(0.1 equiv), and 50 mm of the respective methylene carboxylic acid
(1.0 equiv) in DMF/MeOH 3:2 for 20 min at a flow-rate of 25 mLminÀ1
.
In total the methylene carboxylic acid was six times freshly activated
with EDC-HCl and therefore exposed in total for 120 min at a flow-rate
of 25 mLminÀ1 to the resin-bound DNA conjugates.
Then, the columns were washed with MeOH (3 mL) and HOAc (1 mL,
10 mm). The DEAE-resin slurry was resuspended in HOAc (400 mL,
10 mm), transferred to a SpinX column and centrifuged for 1 min at
3400 g. The DEAE-resin slurry was resuspended in NaOAc buffer
(400 mL, 1m, pH 8.7), incubated for 2 min and briefly centrifuged at
16100 g; this yielded an elution fraction usually containing 5% of the
DNA conjugates. Afterwards, the DEAE-resin slurry was resuspended in
NaOAc/HOAc buffer (200 mL, 3m, pH 4.7), incubated for 2 min and cen-
trifuged for 1 min at 13200 rpm; this yielded an elution fraction usually
containing 50% of the DNA conjugates. The second elution fraction was
subjected to EtOH precipitation.
Palladium/copper-catalyzed Sonogashira reactions: Resin-bound NVOC-
p-I-l-Phe-OH or NVOC-p-I-d-Phe-OH (50 mg, 35 mmol, 1.0 equiv) were
suspended in dry dioxane (600 mL) and stirred for 30 min at 258C under
an argon atmosphere and light protection. The respective alkyne
(210 mmol, 6.0 equiv) was dissolved in dry dioxane (300 mL) and added to
the reaction vessel. After 15 min of stirring, CuI (1.3 mg, 7.0 mmol,
0.2 equiv), PdACHTUNGTRENNUNG(PPh3)2Cl2 (2.6 mg, 3.5 mmol, 0.1 equiv) and TEA (300 mL)
were added to the reaction mixture and stirred, overnight, at 258C. Then,
the reaction mixture was transferred into a polypropylene filter syringe,
washed five times with DMF, twice with CH2Cl2 and released by resin-
cleavage by using TFA (2%) in CH2Cl2 for 1 h 258C. The crude product
was filtered over Millexꢃ-FG 0.2 mm, purified by HPLC and the desired
fractions were evaporated to dryness.
Klenow polymerase encoding of the second chemical building blocks: An
aliquot of the modified DNA conjugate sublibrary (16 pmol, 1.0 equiv)
and of the respective partially complementary 44-mer oligonucleotide
(5’-GTAGTC GGATCC GACCAC NNNNNNNN GACAAT TCACAC
ACGTCC-3’, where NNNNNNNN unambiguously identifies the corre-
sponding methylene carboxylic acid; 24 pmol, 1.5 equiv) were dissolved
in Klenow polymerase buffer to a final volume of 44 mL. The reaction
mixture was incubated for 10 min at 508C and allowed to cool to room
temperature for the partial hybridization of the respective DNA stretch-
es. Afterwards, dNTP (33 mm, 5 mL) and Klenow polymerase enzyme
(5 U) were added and the reaction mixture was incubated for 30 min at
258C. Reactions were purified on ion-exchange cartridges and eluted
with QE buffer (50 mL). Equimolar pooling yielded the DEL30000 li-
brary.
Palladium-catalyzed cross-coupling reactions (Suzuki coupling): NVOC-
p-I-l-Phe-OH or NVOC-p-I-d-Phe-OH (10 mg, 24 mmol, 1.0 equiv) was
dissolved in THF (500 mL). The respective phenyl boronic acid (40 mmol,
1.7 equiv) was dissolved in THF (500 mL) and added to the first solution.
Afterwards a catalytic amount of polymer-bound palladium and 2m
K2CO3 (90 mL, 180 mmol, 7.5 equiv) were added to the reaction vessel
and heated to 758C, overnight, under reflux. The reaction mixture was
centrifuged, the supernatant was filtered over Millexꢃ-FG 0.2 mm, puri-
fied by HPLC and the desired fractions were evaporated to dryness.
Azide–alkyne cycloaddition by using Sharpless conditions: Fmoc-cis-N3-
Pro-b-Ala or Fmoc-trans-N3-Pro-b-Ala (10 mg, 22 mmol, 1.0 equiv) was
dissolved in degassed tBuOH/H2O 2:1 (720 mL) and added to the respec-
tive alkyne (33 mmol, 1.5 equiv) under an argon atmosphere. Afterwards,
a solution of 45 mm TBTA (44 mL, 2.2 mmol, 0.1 equiv) in degassed
tBuOH/H2O 2:1 (720 mL) and 50 mm CuSO4 (44 mL, 2.0 mmol, 0.09 equiv)
was added to the reaction vessel. Afterwards, 50 mm TCEP (66 mL,
3.3 mmol, 0.15 equiv) in degassed H2O and 1m NaHCO3 (110 mL,
110 mmol, 5.0 equiv) in degassed H2O were added to the reaction mixture
and stirred, overnight, at 258C. The reaction mixture was neutralized by
Selections against human carbonic anhydrase IX (CA IX): Cobalt mag-
netic Dynabeadsꢃ His-tag isolation and pulldown (50 mL, 2 mg,
40 mgmLÀ1) were transferred to a tube, which was placed on a magnet
for supernatant removal by aspiration with a pipette. Afterwards, 36 mm
CA IX (20 mL, 720 pmol, 40 mg) was dissolved in B&W buffer B
(680 mL), mixed with the cobalt magnetic beads and incubated and peri-
odically mixed for 10 min at 258C. CA IX-coated beads were washed
four times with B&W buffer B (300 mL). CA IX-coated bead dilutions
were generated to a final volume of 100 mL ranging from a bead density
Chem. Eur. J. 2012, 18, 7729 – 7737
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7735

D. Neri et al.
of 20 mgmLÀ1 (corresponding to 40 mg immobilized human CA IX,
720 pmol, 7.2 mm, 720 equiv) to a bead density of 100 mgmLÀ1 (corre-
sponding to 200 ng immobilized human CA IX, 3.6 pmol, 36 nm,
3.6 equiv). Coating densities remained constant (20 mg CA IX per 1 mg
cobalt magnetic beads. Then, 360 nm DEL30000 library (2.8 mL, 1 pmol,
10 nm final concentration, 1.0 equiv) was incubated with CA IX-coated
beads and periodically mixed for 1 h at 258C.
function of the test compound concentration and fitted by using Kaleida-
Graph V4.0.
Other methods: Detailed library and compound characterization data, al-
gorithms for used sequence decoding, data processing, and molecular
docking, and detailed selection and hit validation protocols are provided
in the Supporting Information.
The beads were washed six times with PD buffer (100 mL) by placing the
tube on a magnet for 1 min and aspiration of the supernatant with a pip-
ette. Alternative washing protocols with higher detergent concentration
(six washing cycles with 100 mL PBS pH 7.4/0.1% Tween-20) and with
higher washing frequency (eight washing cycles with 200 mL PD buffer)
were also applied. After the last washing cycle, the beads were resus-
pended in HE buffer (100 mL) for CA IX delivery, including bound DNA
conjugates. The tubes were placed on a magnet and the supernatant was
collected by aspiration with a pipette. The coding DNA of oligonucleo-
tide–compound conjugates was amplified by PCR after selection experi-
ments by using the eluate (1 mL) as DNA template. After PCR, corre-
sponding samples were pooled and purified by ion-exchange cartridges
and eluted with QE buffer (50 mL).
Acknowledgements
Financial support from ETH Zꢁrich, Swiss National Science Foundation,
Philochem AG, KTI (Grant Nr.: 8868.1 PFDS-LS), Swiss Bridge Founda-
tion, Stammbach Foundation and the Gebert Rꢁf Foundation. Instant
JVhem was used for structure and database management, search and pre-
diction, JChem 5.7.0, 2011 (http://www.chemaxon.com).
[1] B. A. Posner, Curr. Opin. Drug Discov. Devel. 2005, 8, 487.
[2] P. J. Hajduk, J. Greer, Nat. Rev. Drug Discovery 2007, 6, 211.
[3] L. M. Mayr, D. Bojanic, Curr. Opin. Pharmacol. 2009, 9, 580.
[4] D. R. Halpin, P. B. Harbury, PLoS Biol. 2004, 2, e174.
[5] Z. J. Gartner, B. N. Tse, R. Grubina, J. B. Doyon, T. M. Snyder,
D. R. Liu, Science 2004, 305, 1601.
CA IX inhibition assays: The potency of CA IX inhibitors was deter-
mined by a colorimetric kinetic CA IX esterase activity assay by using
pNPA as a chromogenic substrate. CA IX catalyzes the substrate hydrol-
ysis to HOAc and pNP, which was measured by absorbance at l=
360 nm. Experiments were performed in a final volume of 60 mL per well
by using transparent 384-well plates. Dilution series of inhibitor mole-
cules with final concentrations ranging from 100 mm to 100 pm (dissolved
in 50 mm Tris, 1 mm ZnSO4, pH 7.4) were tested for inhibition against a
final concentration of 100 nm CA IX.
[6] S. Melkko, J. Scheuermann, C. E. Dumelin, D. Neri, Nat. Biotechnol.
2004, 22, 568.
[7] M. A. Clark, R. A. Acharya, C. C. Arico-Muendel, S. L. Belyan-
skaya, D. R. Benjamin, N. R. Carlson, P. A. Centrella, C. H. Chiu,
S. P. Creaser, J. W. Cuozzo, C. P. Davie, Y. Ding, G. J. Franklin, K. D.
Franzen, M. L. Gefter, S. P. Hale, N. J. Hansen, D. I. Israel, J. Jiang,
M. J. Kavarana, M. S. Kelley, C. S. Kollmann, F. Li, K. Lind, S. Mat-
aruse, P. F. Medeiros, J. A. Messer, P. Myers, H. OꢄKeefe, M. C.
Oliff, C. E. Rise, A. L. Satz, S. R. Skinner, J. L. Svendsen, L. Tang,
K. van Vloten, R. W. Wagner, G. Yao, B. Zhao, B. A. Morgan, Nat.
Chem. Biol. 2009, 5, 647.
[8] M. Ciobanu, K. T. Huang, J. P. Daguer, S. Barluenga, O. Chaloin, E.
Schaeffer, C. G. Mueller, D. A. Mitchell, N. Winssinger, Chem.
Commun. 2011, 47, 9321.
[9] S. Brenner, R. A. Lerner, Proc. Natl. Acad. Sci. USA 1992, 89, 5381.
[10] G. Georghiou, R. E. Kleiner, M. Pulkoski-Gross, D. R. Liu, M. A.
Seeliger, Nat. Chem. Biol. 2012, 8, 366.
Respective sixfold stock solutions of the inhibitors (10 mL) and CA IX
(10 mL, 600 mm) were added to H2O (30 mL). The substrate pNPA
(10 mL, 6 mm stock solution in H2O/MeCN 15:1 resulting in a final pNPA
concentration of 1 mm) was added immediately prior to the measure-
ment. Importantly, background hydrolysis of the pNPA could be reduced
by using a final buffer consisting of Tris (17 mm), ZnSO4 (330 mm),
pH 7.4, MeCN (1%). Absorbance was measured in intervals of 45 s for
2.5 h. The vmax values were obtained by linear fitting of increasing OD360
signals over time. The background hydrolysis rate was subsequently sub-
tracted and normalized CA IX activities were plotted as a function of the
inhibitor concentration and fitted by using KaleidaGraph V4.0.
Fluorescence polarization assays: In a final volume of 30 mL per well,
constant final concentrations of fluorescein conjugates were incubated
with increasing final concentrations of streptavidin (from 3 nm to 50 mm),
CA IX (from 120 pm to 2 mm), or IL-2 (from 3 nm to 60 mm) in PBS,
[11] Y. He, D. R. Liu, J. Am. Chem. Soc. 2011, 133, 9972.
[12] J. P. Daguer, M. Ciobanu, S. Alvarez, S. Barluenga, N. Winssinger,
Chem. Sci. 2011, 2, 625.
[13] O. Ramstrom, J. M. Lehn, Nat. Rev. Drug Discovery 2002, 1, 26.
[14] G. P. Smith, Science 1985, 228, 1315.
pH 7.4, for 15 min at 258C. Fluorescence polarization (lex =485 nm, lem
=
535 nm) was determined by the average of at least 20 consecutive meas-
urements over 30 min. Response signals were plotted as a function of the
protein concentration and fitted by using KaleidaGraph V4.0.
[15] D. R. Bentley, S. Balasubramanian, H. P. Swerdlow, G. P. Smith, J.
Milton, C. G. Brown, K. P. Hall, D. J. Evers, C. L. Barnes, H. R.
Bignell, J. M. Boutell, J. Bryant, R. J. Carter, R. K. Cheetham, A. J.
Cox, D. J. Ellis, M. R. Flatbush, N. A. Gormley, S. J. Humphray, L. J.
Irving, M. S. Karbelashvili, S. M. Kirk, H. Li, X. Liu, K. S. Maisinger
et al., Nature 2008, 456, 53.
[16] S. Fox, S. Filichkin, T. C. Mockler, Methods Mol. Biol. 2009, 553, 79.
[17] C. A. Lipinski, F. Lombardo, B. W. Dominy, P. J. Feeney, Adv. Drug
Delivery Rev. 2001, 46, 3.
[18] M. M. Hann, T. I. Oprea, Curr. Opin. Chem. Biol. 2004, 8, 255.
[19] P. D. Leeson, B. Springthorpe, Nat. Rev. Drug Discovery 2007, 6,
881.
[20] L. Mannocci, Y. Zhang, J. Scheuermann, M. Leimbacher, G. De Bel-
lis, E. Rizzi, C. Dumelin, S. Melkko, D. Neri, Proc. Natl. Acad. Sci.
USA 2008, 105, 17670.
[21] C. T. Supuran, Nat. Rev. Drug Discovery 2008, 7, 168.
[22] D. Neri, C. T. Supuran, Nat. Rev. Drug Discovery 2011, 10, 767.
[23] A. Weber, M. Bohm, C. T. Supuran, A. Scozzafava, C. A. Sotriffer,
G. Klebe, J. Chem. Inf. Model. 2006, 46, 2737.
[24] J. Pouyssꢅgur, F. Dayan, N. M. Mazure, Nature 2006, 441, 437.
[25] J. K. Ahlskog, C. E. Dumelin, S. Trussel, J. Marlind, D. Neri, Bioorg.
Med. Chem. Lett. 2009, 19, 4851.
T-cell proliferation assays: Murine CTLL-2 cells were harvested and
washed three times with RPMI 1640 (20 mL) complete medium. The
number of viable cells was determined by adding 0.1% trypan blue
(50 mL) to the cell suspension (50 mL) for staining. The cell density is ad-
justed to 500000 cells per mL in RPMI 1640 complete medium.
Proleukinꢃ and selected test compounds were dissolved in RPMI 1640
complete medium and serially diluted with RPMI 1640 complete
medium/2% DMSO (resulting in a final concentration of 0.5% DMSO
in each well). Experiments were performed in a final volume of 100 mL
per well by using transparent 96-well plates. The test compounds (25 mL)
with final concentrations ranging from 512 to 1 mm were incubated with
Proleukinꢃ (25 mL) with a final concentration of 2.5 IUmLÀ1 (10 pm) for
15 min at 258C. Afterwards, freshly washed CTLL-2 cells (50 mL, 500000
cells per mL) were added to each well (resulting in an initial cell number
of 25000 cells per well). The plate was incubated at 378C and 5% CO2
for 48 h. Afterwards, MTS (20 mL) was added to each well and incubated
at 378C and 5% CO2 for 4 h. The signal response was obtained as OD490
values (absorbance at l=490 nm) and normalized in order to obtain rela-
tive T-cell growth rate. The normalized OD490 values were plotted as a
7736
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 7729 – 7737

Small-Molecule Interleukin-2 Inhibitors
FULL PAPER
[26] Y. Lou, P. C. McDonald, A. Oloumi, S. Chia, C. Ostlund, A.
Ahmadi, A. Kyle, U. Auf dem Keller, S. Leung, D. Huntsman, B.
Clarke, B. W. Sutherland, D. Waterhouse, M. Bally, C. Roskelley,
C. M. Overall, A. Minchinton, F. Pacchiano, F. Carta, A. Scozzafava,
N. Touisni, J. Y. Winum, C. T. Supuran, S. Dedhar, Cancer Res. 2011,
71, 3364.
[31] M. Rickert, X. Wang, M. J. Boulanger, N. Goriatcheva, K. C. Garcia,
Science 2005, 308, 1477.
[32] G. Jones, P. Willett, R. C. Glen, J. Mol. Biol. 1995, 245, 43.
[33] F. Buller, M. Steiner, K. Frey, D. Mircsof, J. Scheuermann, M. Ka-
lisch, P. Buhlmann, C. T. Supuran, D. Neri, ACS Chem. Biol. 2011, 6,
336.
[27] M. R. Arkin, J. A. Wells, Nat. Rev. Drug Discovery 2004, 3, 301.
[28] J. A. Wells, C. L. McClendon, Nature 2007, 450, 1001.
[29] J. W. Tilley, L. Chen, D. C. Fry, S. D. Emerson, G. D. Powers, D.
Biondi, T. Varnell, R. Trilles, R. Guthrie, F. Mennona, G. Kaplan,
R. A. LeMahieu, M. Carson, R. J. Han, C. M. Liu, R. Palermo, G.
Ju, J. Am. Chem. Soc. 1997, 119, 7589.
[34] F. Buller, L. Mannocci, J. Scheuermann, D. Neri, Bioconjugate
Chem. 2010, 21, 1571.
[35] L. Mannocci, M. Leimbacher, M. Wichert, J. Scheuermann, D. Neri,
Chem. Commun. 2011, 47, 12747.
[30] M. R. Arkin, M. Randal, W. L. DeLano, J. Hyde, T. N. Luong, J. D.
Oslob, D. R. Raphael, L. Taylor, J. Wang, R. S. McDowell, J. A.
Wells, A. C. Braisted, Proc. Natl. Acad. Sci. USA 2003, 100, 1603.
Received: March 20, 2012
Published online: May 15, 2012
Chem. Eur. J. 2012, 18, 7729 – 7737
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7737