Discovery of potent thrombin inhibitors from a protease-focused DNA-encoded chemical library

Article abstract of DOI:10.1073/pnas.2005447117

DNA-encoded chemical libraries are collections of compounds individually coupled to unique DNA tags serving as amplifiable identification barcodes. By bridging split-and-pool combinatorial synthesis with the ligation of unique encoding DNA oligomers, million- to billion-member libraries can be synthesized for use in hundreds of healthcare target screens. Although structural diversity and desirable molecular property ranges generally guide DNA-encoded chemical library design, recent reports have highlighted the utility of focused DNA-encoded chemical libraries that are structurally biased for a class of protein targets. Herein, a protease-focused DNA-encoded chemical library was designed that utilizes chemotypes known to engage conserved catalytic protease residues. The three-cycle library features functional moieties such as guanidine, which interacts strongly with aspartate of the protease catalytic triad, as well as mild electrophiles such as sulfonamide, urea, and carbamate. We developed a DNA-compatible method for guanidinylation of amines and reduction of nitriles. Employing these optimized reactions, we constructed a 9.8-million-membered DNA-encoded chemical library. Affinity selection of the library with thrombin, a common protease, revealed a number of enriched features which ultimately led to the discovery of a 1 nM inhibitor of thrombin. Thus, structurally focused DNA-encoded chemical libraries have tremendous potential to find clinically useful high-affinity hits for the rapid discovery of drugs for targets (e.g., proteases) with essential functions in infectious diseases (e.g., severe acute respiratory syndrome coronavirus 2) and relevant healthcare conditions (e.g., male contraception).

Full text of DOI:10.1073/pnas.2005447117

Discovery of potent thrombin inhibitors from a
protease-focused DNA-encoded chemical library
Surendra Dawadia,1, Nicholas Simmonsa, Gabriella Miklossya, Kurt M. Bohrena, John C. Favera,
Melek Nihan Ucisika, Pranavanand Nyshadhama, Zhifeng Yua, and Martin M. Matzuka,1
aCenter for Drug Discovery, Department of Pathology & Immunology, Baylor College of Medicine, Houston, TX 77030
Contributed by Martin M. Matzuk, May 27, 2020 (sent for review March 24, 2020; reviewed by Philip E. Dawson and Stanton F. McHardy)
DNA-encoded chemical libraries are collections of compounds in-
dividually coupled to unique DNA tags serving as amplifiable iden-
tification barcodes. By bridging split-and-pool combinatorial
synthesis with the ligation of unique encoding DNA oligomers,
million- to billion-member libraries can be synthesized for use in
hundreds of healthcare target screens. Although structural diver-
sity and desirable molecular property ranges generally guide DNA-
encoded chemical library design, recent reports have highlighted
the utility of focused DNA-encoded chemical libraries that are
proteases generally fall into five classes: cysteine proteases, serine
proteases, metalloproteases, threonine proteases, and aspartic
proteases (10). After activation of the amide with key residues,
three of the classes—cysteine, serine, and threonine—utilize the
namesake residue to attack the amide carbonyl group, where
metalloproteases and aspartic proteases use an activated water.
Strategies to directly target proteases with small molecules have
generally focused on engaging the active-site catalytic triad,
leading to peptidomimetic inhibitors that either react with a nu-
cleophilic residue in a covalent fashion or interact with the critical
catalytic residues (11). This strategy has been particularly suc-
cessful for inhibitors of serine proteases, which are designed to
reveal a mild electrophile such as sulfonamide to interact with the
serine hydroxyl group and a positively charged guanidine/benza-
midine to interact with a nearby aspartic acid carboxylate (12).
Clinically, there have been numerous successes including angiotensin-
converting enzyme inhibitors for cardiovascular disorders (13),
thrombin inhibitors for thromboembolism and bleeding disorders
(14, 15), and HIV protease inhibitors in the treatment of HIV and
AIDS (16), among others (17, 18). Recently, viral main protease
(Mpro, a cysteine protease) has presented as an attractive drug
target for the development of novel therapeutics in the COVID-
19 (coronavirus disease 2019) pandemic caused by severe acute
respiratory syndrome coronavirus 2 (SARS-CoV-2), and a number
of broad-spectrum inhibitors have been developed that target this
main protease (19). Another target in the COVID-19 pandemic is
human transmembrane serine protease 2 (TMPRSS2), which
functions in the processing of the SARS-CoV-2 spike (S) protein
structurally biased for
a class of protein targets. Herein, a
protease-focused DNA-encoded chemical library was designed
that utilizes chemotypes known to engage conserved catalytic
protease residues. The three-cycle library features functional moi-
eties such as guanidine, which interacts strongly with aspartate of
the protease catalytic triad, as well as mild electrophiles such as
sulfonamide, urea, and carbamate. We developed
a DNA-
compatible method for guanidinylation of amines and reduction
of nitriles. Employing these optimized reactions, we constructed a
9.8-million-membered DNA-encoded chemical library. Affinity se-
lection of the library with thrombin, a common protease, revealed
a number of enriched features which ultimately led to the discov-
ery of a 1 nM inhibitor of thrombin. Thus, structurally focused
DNA-encoded chemical libraries have tremendous potential to find
clinically useful high-affinity hits for the rapid discovery of drugs
for targets (e.g., proteases) with essential functions in infectious
diseases (e.g., severe acute respiratory syndrome coronavirus 2)
and relevant healthcare conditions (e.g., male contraception).
DNA-encoded chemical library focused DECL protease inhibitor
|
|
|
COVID-19 SARS-CoV-2
|
Significance
NA-encoded chemical libraries (DECLs) are a powerful
To rapidly identify small-molecule lead compounds to target
D
tool for target-directed hit discovery (1–4). By bridging
healthcare-associated proteases, we constructed
a unique
split-and-pool combinatorial synthesis approaches with the liga-
tion of unique encoding DNA oligomers, million- to billion-
member libraries may be synthesized at micromole scales for
use in hundreds of target screens. Although DECLs are often
built as a general screening resource and the significant cost and
reagent savings theoretically enable routine screens of an orga-
nization’s entire DECL collection for any compatible target, in
certain cases it is desirable to design and use focused DECLs for
a specific target. These focused libraries are particularly suitable
for targets in which mechanistic details, structural information,
or relevant chemical matter are well-known (5, 6). Use of fo-
cused DECLs may bolster chances for robust hit generation as
well as enable deeper and broader exploration of structure–
activity features if combined with curated building block selec-
tion and deep sequencing. Focused DECLs may also empower
designs with motifs not normally included in general DECLs
such as DECLs with covalent modifiers (7) and DECLs in-
corporating macrocycles (8) and complex natural products (9).
Proteases, enzymes which catalyze the hydrolysis of peptidic
bonds in proteins, have been an intensely studied class of targets
for drug discovery due to their involvement in various patho-
physiological processes (e.g., protein degradation, protein pro-
cessing, and signal transduction pathways). Mechanistically,
9.8-million-membered protease-focused DNA-encoded chem-
ical library. Affinity selection of this library with a healthcare-
relevant protease (i.e., thrombin, a key protein necessary for
blood coagulation) revealed potent inhibitors in the first
screening attempt. Our results emphasize the utility of a
structurally focused DNA-encoded chemical library approach to
rapidly uncover hits for healthcare targets (e.g., proteases)
where no drug exists (e.g., male contraception) and for
emerging diseases (e.g., coronavirus disease 2019).
Author contributions: S.D., N.S., and M.M.M. designed research; S.D., G.M., K.M.B., J.C.F.,
M.N.U., P.N., and Z.Y. performed research; S.D., N.S., and J.C.F. analyzed data; and S.D.,
N.S., and M.M.M. wrote the paper.
Reviewers: P.E.D., The Scripps Research Institute; and S.F.M., The University of Texas at
San Antonio.
The authors declare no competing interest.
This open access article is distributed under Creative Commons Attribution License 4.0
(CC BY).
1To whom correspondence may be addressed. Email: dawadi@bcm.edu or mmatzuk@
bcm.edu.
This article contains supporting information online at https://www.pnas.org/lookup/suppl/
doi:10.1073/pnas.2005447117/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.2005447117
PNAS Latest Articles
|
1 of 8

(20). Proteases also play important roles in reproduction, and
several serine proteases and serine protease-like proteins are
essential for normal fertility and therefore could be unique
targets for nonhormonal male contraception (i.e., a male pill).
For example, among the proteins essential for male fertility are
the following: TMPRSS12, a testis-specific transmembrane
serine protease required for sperm motility and uterotubal
junction migration (21) and structurally similar to TMPRSS2;
serine protease 55 (PRSS55), a glycosylphosphatidylinositol-
anchored testis-specific serine protease that is necessary for
sperm uterotubal junction migration and sperm–oocyte binding
(22, 23); serine protease 37 (PRSS37), a testis-specific inactive
serine protease-like protein required for sperm uterotubal
junction migration and fertilization (24); and ovochymase 2
(OVCH2), an epididymis-specific secreted serine protease re-
quired for maturation of sperm (25). Thus, there are multiple
healthcare areas that could benefit from advances in the pro-
duction of unique protease-biased chemical matter for inhibit-
ing this important target class.
Here, we report on a protease-focused DECL and DECL
screening strategies designed to engage the protease catalytic
triad. Since many of the known protease inhibitors contain
guanidine, sulfonamide, urea, and carbamate moieties, we in-
corporated these functional groups into the design of this library
(26, 27). We utilized the structure of argatroban (Fig. 1A), an
arginine containing peptidomimetic thrombin inhibitor, as in-
spiration for the design of this library since it bears a positively
charged guanidinium group flanked by two hydrophobic groups
(28). Retrosynthetically, argatroban could be portioned into
sections that could be connected in a three-cycle DECL. The
piperidine carboxylic acid can be viewed as building block 1
(BB1) which can be covalently linked to DNA via an amide
linker. Interestingly, this carboxylate is situated in the direction
to open space of the thrombin binding pocket (29), which makes
it correctly positioned for a DNA linkage site. The arginine
moiety of argatroban can be visualized as BB2 and the aryl sul-
fonamide as BB3. Thus, the proposed protease DECL (Fig. 1B)
could feature a BB1 set of various –NBoc amino acids and –NBoc
diamines connected by amide bonds to a DNA-attached amino or
DNA attachment site
A
B
BB1
BB2
BB3
C
Fig. 1. Protease-focused DECL design. (A) Structure of argatroban, a thrombin inhibitor in clinical use. Argatroban served as a prototype for the design of the
protease-focused DECL described in this study. (B) Design of a guanidine containing three-cycle protease-focused DECL. The library features guanidine, as well
as a sulfonamide, urea, and carbamate moiety, which are commonly found in protease inhibitors. (C) Examples of various types of building blocks in each
cycle of the library build and a list of reactions required to accomplish the synthesis of this library.
2 of 8
|
www.pnas.org/cgi/doi/10.1073/pnas.2005447117
Dawadi et al.

carboxyl terminated linker, respectively. After –NBoc depro-
tection, a variety of different trifunctional cores containing a free
carboxyl group, primary –NHBoc amine and orthogonally masked
amine (–NFmoc, nitro, or nitrile group) could be connected by
amidation to serve as BB2. After deprotection of the orthogonally
masked Fmoc group, and reduction of nitro or nitrile groups and
subsequent guanidinylation, the remaining –NHBoc amine could
be deprotected and functionalized with BB3 through sulfonyla-
tion, carbamolyation, and carbamylation (Fig. 1C). Critical to this
moiety through multistep DECL synthesis. However, to the best of
our knowledge there have been no reports of guanidines within
published DECLs, although
a benzamidine-containing dual
pharmacophore DECL has been reported (30).
Results and Discussion
To evaluate the synthetic feasibility of this DECL design, we
looked at the availability of DNA-compatible chemical reactions
for the construction of this library (Fig. 1C). The acylation re-
design is both the synthesis and maintenance of the guanidine action, where an amine functionalized DNA is coupled with a
Substrate
Reagent
Condition
Conversion (%)
pH 9.5 borate, rt, 16 h
pH 9.5 borate, 45 °C, 16 h
pH 7.4 MOPS, rt, 16 h
pH 7.4 MOPS, 45 °C, 16 h
pH 5.6 MES, rt, 16 h
65
90
<5
30
<5
<5
1a
A (200 equiv.)
pH 5.6 MES, 45 °C, 16 h
pH 9.5 borate, rt, 16 h
pH 9.5 borate, 45 °C, 16 h
<5
<5
1a
1a
1a
B (200 equiv.)
C (200 equiv.)
D (200 equiv.)
pH 9.5 borate, rt, 16 h
pH 9.5 borate, 45 °C, 16 h
90
>95
pH 9.5 borate, rt, 16 h
pH 9.5 borate, 45 °C, 16 h
10
70
A (200 equiv.)
A (200 equiv.)
A (50 equiv.)
A (100 equiv.)
pH 9.5 borate, 45 °C, 3 h
pH 9.5 borate, 45 °C, 8 h
pH 9.5 borate, 45 °C, 16 h
pH 9.5 borate, 45 °C, 16 h
15
35
30
60
1a
1b
A (200 equiv.)
C (200 equiv.)
D (200 equiv.)
pH 9.5 borate, 45 °C, 16 h
pH 9.5 borate, 45 °C, 16 h
pH 9.5 borate, 45 °C, 16 h
>95
>95
>95
Fig. 2. Optimization of reaction conditions for guanidinylation of amines.
Dawadi et al.
PNAS Latest Articles
|
3 of 8

carboxylic acid building block, or vice versa, forming an amide
bond is one of most common reactions in DECL technology.
Similarly, reports of deprotection conditions of –NBoc and –NFmoc
groups are also common. The reactions of an amine with elec-
trophiles such as sulfonyl chlorides and isocyanates to produce
sulfonamides and ureas, respectively, are also known (31). Sev-
eral conditions for the reduction of aromatic nitro groups to the
amine have also been reported (32, 33) Although the conversion
of an amine to an isocyanate followed by reaction with nucleo-
philes such as amine and phenol to form urea and carbamate
(34) is known in the presence of DNA, it has not been utilized in
the construction of a reported DECL. Overall two of our plan-
ned reactions, the conversion of amine to guanidine and the
reduction of nitriles to amines, would require optimization to
find a general, DNA-compatible process.
(Fig. 2) (35). The DNA linked primary and secondary amine
substrates 1a and 1b were prepared from the DNA starting
material DNA headpiece (DNA-HP 3; see SI Appendix, Scheme
S1) by acylation with corresponding –NHBoc amino acid fol-
lowed by the –NBoc deprotection (SI Appendix, Fig. S2). When
1a and 1b were treated with reagents A through D in various
reaction conditions (Fig. 2 and see SI Appendix for detailed
optimization conditions), we observed formation of guanidine
products 2a and 2b. Among different buffer systems (pH range
from 5.6 to 9.9) studied, pH 9.5 borate buffer gave better con-
versions. Warming the reaction to 45 °C gave higher conversions
in a shorter time. Overall, we determined that 45 °C for 16 h
(overnight) in pH 9.5 borate buffer (25% vol/vol MeCN) with
either reagent A or C gave excellent conversions with substrate
1a, whereas substrate 1b provided excellent conversions with
reagents A, C, and D. We employed these optimal conditions to
study the substrate scope of guanidinylation. The amines 1c–n
and anilines 1o–u are obtained from acylation of DNA-HP 3 with
commercially available –NBoc amino acids followed by –NBoc
deprotection (SI Appendix, Fig. S2). We performed the guanidinylation
Optimization of DNA-Compatible Guanidinylation of Amines and
Reduction of Nitriles. To explore the DNA-compatible guanidine
formation, we studied four commercially available 1H-pyrazole-
1-carboxamidine–based guanidinylation reagents A through D
Fig. 3. Scope of DNA-compatible guanidinylation of amines and anilines.
4 of 8
|
www.pnas.org/cgi/doi/10.1073/pnas.2005447117
Dawadi et al.

of these substrates and the products 2a–u were obtained in
various conversions (Fig. 3). Primary and unhindered secondary
amines readily undergo guanidinylation, whereas anilines have a
wide range of reactivity depending upon electronics and sterics.
With some secondary amines such as 1k and most anilines, re-
agent D gave significantly better conversions. Regardless of the
reagent and conditions used, the electron-deficient anilines such
as 1q, 1r, and 1u are poor substrates for guanidine formation,
which is also observed with the secondary aniline 2t. Importantly,
in the library synthesis setting, we avoided using such building
blocks containing electron-deficient anilines in cycle-2 chemistry.
Overall, this result suggested that we should use reagent D for
guanidinylation of anilines and some secondary amines whereas
reagents A and C are more preferable for the guanidinylation of
reactive amines.
Next, we studied the reduction of nitriles (see SI Appendix for
detail optimization condition). Among several conditions stud-
ied, the best outcomes were observed while utilizing NaBH4 as
hydride donor in the presence of following additives: Raney Ni
and NiCl2 (SI Appendix, Table S2). A mixture of Raney Ni and
NaBH4 in 40% MeOH in pH 5.6 MES [2-(N-morpholino)etha-
nesulfonic acid] buffer reduced both aromatic and aliphatic ni-
trile groups with excellent conversions within 1 h at room
temperature. Also, Raney Ni is known to be compatible with
DNA and has been previously employed for nitro reduction (34).
One important consideration provided while using Raney Ni was
the removal of metal particles by centrifugation before EtOH
precipitation of DNA material to prevent DNA damage in
subsequent step especially if it is the –NBoc deprotection.
representative of the three kinds of trifunctional groups (i.e., two
nitros, two nitriles, and two –NFmoc groups) to afford 44a–f.
The subsequent nitro and nitrile reduction and Fmoc depro-
tection proceeded with excellent conversion to provide 45a–f.
The guanidinylation of newly formed amines and anilines went
smoothly with reagent A or D and the subsequent –NBoc
deprotection provided compounds 47a–f. The newly exposed
primary amine underwent N-capping reactions with a sulfonyl
chloride, isocyanate, or anhydride cycle-3 BB to produce a sul-
fonamide, urea, or amide bond. In addition to those reactions,
the primary amine also underwent isocyanate formation using
dipyridin-2-yl carbonate followed by reaction with amines and
phenols to produce the corresponding urea and carbamate.
These reactions collectively afforded the final products 48a–l.
The chemistry validation also confirmed that the guanidine has
no reactivity toward cycle-3 BBs. Additionally, we confirmed that
the reaction conditions have negligible impact on DNA recovery
and integrity, which were evaluated with optical density mea-
surement (ultraviolet absorbance at 260 nm) and ligation
efficiency.
After establishing all optimized reactions work well in a DECL
synthesis setting, the next step was to select the building blocks.
Due to two –NBoc deprotection reactions that require heating of
the DNA material at 80 °C for >24 h, there are some limitations
on cycle-1 BB selection. If we use alpha –NHBoc amino acids,
the amide bond is cleaved by BB2 amino group during the
–NBoc deprotection during cycle 2 (SI Appendix, Fig. S4A).
Similarly, if we use primary –NHBoc amino acids, then the amide
–NH reacts with isocyanate intramolecularly during cycle 3 to
produce a stable hydantoin (SI Appendix, Fig. S4B). These results
suggest only nonalpha secondary –NBoc amino acids can be used
in cycle 1. For cycle 2, the BB has to be a primary –NHBoc
amino acid, preferably containing a nitro, nitrile, or –NFmoc
group. We also decided to include BBs that do not have this third
functional group and do not undergo guanidine formation but
preferably have some guanidine bioisosteres. For cycle-3
N-capping reactions, a huge collection of BBs is available and
they were selected based on diversity calculations and reactivity.
We proceeded to construct the library (Fig. 4 and see SI Ap-
pendix for detailed methods). The synthesis began with two DNA
starting materials, amine 3 and carboxylic acid 4. They were split
Chemistry Validation and Library Construction. After optimization
of the guanidinylation and nitrile reduction, we performed the
validation of these two new reactions by the synthesis and
analysis of three authentic on-DNA compounds 41–43 (SI Ap-
pendix, Scheme S2). A more detailed three-cycle chemistry val-
idation process that mimics the library design was performed to
prove the feasibility and versatility of our synthetic methods (SI
Appendix, Scheme S3). Importantly, we need to confirm that the
guanidine moiety formed in cycle-2 does not impact cycle-3
amine capping reactions. The validation process started with
compound 1b, which was acylated with six cycle-2 BBs that were
Fig. 4. Construction of a 9.8-million-member protease-focused DECL.
Dawadi et al.
PNAS Latest Articles
|
5 of 8

into a desired number of wells, ligated with cycle-1 codons, and
reacted with –NBoc amino acids and –NBoc diamines. Both in-
dividual ligations and acylations were monitored by liquid
chromatography–mass spectrometry (LC-MS). After the com-
pletion of cycle 1, the DNA materials were pooled and –NBoc
deprotection was performed. The DNA material was split based
on the number of cycle-2 BBs and acylation reactions were
carried out. Based on the nature of the BBs, nitro reduction,
nitrile reduction, Fmoc deprotection, or guanidine formation
reactions were performed. After the completion of these cycle-2
reactions, we performed codon-2 ligations which were monitored
by gel electrophoresis. After the completion of ligation, DNA
materials were pooled and –NBoc deprotection was performed.
The DNA material was split and codon-3 ligations were carried
out followed by the N-capping reactions with cycle-3 BBs. Im-
portantly, each cycle pooled material is used as a blank in the
next cycle to encode the unreacted materials in those cycles.
After completion of the main library build, the entire library
material was ligated with a duplexed pair of 12-mer oligonucleo-
tides to encode the library structure/design. Small portions of this
tagged library material were further ligated with two oligonucleo-
tides called closing primers to enable later amplification. At this
point, the library material was ready for selection experiments.
A
60
40
20
0
Count
181
Count
329
AC_zscore_n
33.79
AC_zscore_n
61.13
Count
314
AC_zscore_n
58.36
Count
152
AC_zscore_n
28.43
0
10
20
30
40
50
60
Enrichment for Thrombin
B
Affinity Selection with Thrombin. After completion of the build of
this 9.8-million-member library and prior to the selection ex-
periment with proteins, the quality of the DECL was evaluated
with a naïve sequencing codon distribution report (SI Appendix,
Fig. S8A). Further analysis of the relative codon distributions
between each of the cycle-2 subpools confirmed that the codons
representing all building blocks were observed (SI Appendix, Fig.
S8B). The NFmoc codon distribution served as a control for
DNA integrity when compared to the nitrile reduction method.
The observed codon counts were comparable between the
building blocks undergoing mild NFmoc deprotection and the
ones undergoing nitrile reduction, which was additional evidence
of DNA stability under newly developed reaction conditions.
After these evaluations of DNA integrity, we performed a se-
lection experiment to find small-molecule binders of thrombin
within the library. Selection assays were performed on
streptavidin-coated magnetic beads, which were pretreated with
biotinylated human thrombin. The binders were then eluted and
the DNA was amplified by PCR and analyzed by next-generation
Illumina sequencing in comparison with no-target control
(NTC). The enrichment of n-synthons (i.e., the encoded associ-
ation of a chemical building block and unique DNA sequence)
for thrombin in comparison with NTC were plotted (Fig. 5A)
(36). The enrichment of each n-synthon was measured using a
normalized z-score metric. The resulting enrichment values
(AC-zscore-n) were compared by plotting the enrichment in the
target sample (x axis) against enrichment in the NTC sample (y
axis). We observed a strong enrichment of a number of three-
cycle trisynthons displaying structure–enrichment relationships.
The same trisynthons encoded by different DNA tags were
similarly enriched, which confirmed the robustness of the se-
lection result. The enriched features enabled us to identify the
structure of the best binders 5 and 6 (Fig. 5B). In the library,
these molecules are covalently attached to DNA at the methyl
amide site.
5
6
Fig. 5. Discovery of thrombin inhibitors from the protease-focused DECL.
(A) Analysis of enrichment of building blocks in the selection of thrombin
with the library. The resulting enrichment values (AC-zscore-n) (36) were
compared by plotting the enrichment in the target sample (x axis) against
enrichment in the NTC sample (y axis). The higher the number in the x axis, the
stronger the enrichment of synthons. (B) Structure of the top two thrombin hits
5 and 6 identified from the selection experiment. In the library pool, these
molecules are attached to the DNA via the linker at the methyl amide site.
deprotection condition afforded 8. We anticipate the formation
of two diastereomers of 8. Compound 8 underwent amidation
with 9 to provide 10. Nitrile reduction of 10 afforded 11 which
underwent guanidinylation followed by –NBoc deprotection to
provide 12. The primary amine 12 produced an isocyanate in-
termediate in the presence of the reagent dipyridin-2-yl car-
bonate, which was treated with two amines 13 and 14 to afford
urea DNA-5 and DNA-6, respectively. We did not observe for-
mation of any significant by-products along the synthesis and the
final products were purified by high-performance LC (HPLC).
We studied the thrombin inhibitory potency of DNA-5 and
DNA-6 (Fig. 6C) and observed that DNA-5 (58 nM), which is
the top hit, is about 10-fold more potent than DNA-6 (589 nM)
(Fig. 6E).
Our analysis of the on-DNA potency of DNA-5 and DNA-6
led to the decision to synthesize isomerically pure 5 off-DNA
(Fig. 6B). The homochiral 3S,4R-enantiomer of 7 was the most
enriched cycle-1 building block in the selection experiment and
we desired to use this pure enantiomer for off-DNA hit resyn-
thesis. The starting material 7 was commercially available as a
racemic mixture and the enantiomers of 7 were separated by
chiral HPLC with retention times of 9 and 11 min. The absolute
configuration of each enantiomer was determined by chemical
derivatization to a known compound with reported optical ro-
tation (37). Once we established that the enantiomer with re-
tention time of 9 min was the desired 3S,4R isomer of 7, this
enantiomer was coupled with methylamine using HATU to
Hit Resynthesis and Evaluation of Thrombin Inhibition. The identi-
fied hits were first synthesized on-DNA. The successful com-
pletion of on-DNA synthesis confirmed that the hits are indeed
produced in the library. Also, by performing this we were able to
identify any byproducts that might have formed during the li-
brary build. Thus, we prepared DNA-5 and DNA-6 on DNA-HP
using the synthetic sequence exactly as in the library synthesis
(Fig. 6A). DNA-HP 3 on acylation with ( )-7 followed by –NBoc
6 of 8
|
www.pnas.org/cgi/doi/10.1073/pnas.2005447117
Dawadi et al.

A
B
C
D
E
1 . 0
0 . 8
0 . 6
0 . 4
0 . 2
0 . 0
1
0.8
0.6
0.4
0.2
0
thrombin
compound
Kiapp (nM)
AAr gragtar otrboa bn an
argatroban
9.0 0.5
C o m p o u n d
5
Compound 5
5
1.0 0.03
DNA-5
DNA-6
58
7
589 188
0
5 0
1 0 0
1 5 0
2 0 0
2 5 0
I n h i b i t o r ( n M
Inhibitor (nM)
)
Fig. 6. Synthesis and biological evaluation of thrombin hits. (A) On-DNA resynthesis of two hits to obtain small-molecule DNA-conjugates DNA-5 and DNA-6.
(B) Off-DNA resynthesis of the top hit 5. (C) Inhibition of thrombin by DNA-5 and DNA-6 compared with argatroban. (D) Dose–response curves of thrombin
inhibition by compound 5 and argatroban. (E) Inhibition constant (apparent Ki) values of synthesized hits. DNA-5 is 10-fold more potent than DNA-6.
Compound 5 is 10-fold more potent than argatroban against thrombin.
provide methyl amide, which under –NBoc deprotection condi-
tion provided isomerically pure (3S,4R)-15. The synthesis of the
other part of the molecule began with compound 9, which when
treated with iodomethane and K2CO3 afforded the methyl ester
which in turn underwent –NBoc deprotection to provide free
amine 16. When 16 was treated with dipyridin-2-yl carbonate at
0 °C, the amine was converted to isocyanate, which when treated
with freshly prepared secondary amine 13 afforded the urea 17.
The primary amine 16 is prone to react with the isocyanate
obtained from itself and produce dimeric urea in 50% yield.
Unfortunately, various purification methods were unable to
separate the desired product 17 from this by-product. Both 17
and the by-product were taken forward for ester hydrolysis with
LiOH at 0 °C. Purification by silica gel chromatography afforded
the acid 18 in 42% yield in three steps. We confirmed that
compound 18 is a single stereoisomer by optical rotation and
NMR, which suggested racemization did not occur during ester
hydrolysis. The acid 18 was coupled with amine 15 using HATU
to afford 19, which upon nitrile reduction and guanidinylation
with reagent A provided the final product 5. With purified
compound 5 in hand, we examined the inhibitory potency with
thrombin. Compound 5 inhibited thrombin very strongly in a
dose–response manner (Fig. 6D) with an inhibition constant of
1 nM. Compound 5 is about 10-fold more potent thrombin in-
hibitor than the clinical drug argatroban (9 nM) (Fig. 6E).
In summary, starting with known protease pharmacophores,
we designed and constructed a protease-focused DECL. Selec-
tion of this library with thrombin revealed a number of potent
thrombin binders. Our best thrombin inhibitor, urea- and
guanidine-functionalized compound 5, showed a remarkable in-
hibitory potency (Ki = 1 nM). The discovery of 5 has led to the
initiation of a medicinal chemistry project to further optimize
this small molecule thrombin inhibitor with a goal of identifying
novel anticoagulants for the treatment of blood clotting com-
plications. In vitro metabolism studies as well as in vivo clotting
time studies such as diluted thrombin time and activated partial
thromboplastin time are currently ongoing. Additionally, a
number of viral, bacterial, and contraception-relevant proteases
are undergoing affinity selection with our validated DECL. We
anticipate that this approach of utilizing structurally focused
DECLs for protease drug discovery can be applied to a wide
range of protein targets of therapeutic interest.
Materials and Methods
Procedure for reactions on DNA, protocols of library synthesis, protocol of
affinity selection, procedures of hit resynthesis, ultra-performance LC–MS
chromatograms of on-DNA compounds, and NMR characterization data of
synthetic compounds are available in SI Appendix.
Data Availability. The paper and SI Appendix contain all datasets generated
during this study.
Dawadi et al.
PNAS Latest Articles
|
7 of 8

ACKNOWLEDGMENTS. This work is supported by the Eunice Kennedy
Shriver National Institute of Child Health and Human Development
(P01HD087157), the Bill and Melinda Gates Foundation (INV-001902), the
Welch Foundation (H-Q-0042), a Core Facility Support Award from the
Cancer Prevention Research Institute of Texas (RP160805), and a COVID-19-
relevant exploratory grant from Baylor College of Medicine.
1. R. M. Franzini, D. Neri, J. Scheuermann, DNA-encoded chemical libraries: Advancing
beyond conventional small-molecule libraries. Acc. Chem. Res. 47, 1247–1255 (2014).
2. D. Neri, Twenty‐five years of DNA‐encoded chemical libraries. ChemBioChem 18,
827–828 (2017).
20. M. Hoffmann et al., SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is
blocked by a clinically proven protease inhibitor. Cell 181, 271–280.e8 (2020).
21. T. Larasati et al., TMPRSS12 is required for sperm motility and uterotubal junction
migration in mice. Biol. Reprod., 10.1093/biolre/ioaa060 (2020).
3. R. A. Lerner, S. Brenner, DNA-encoded compound libraries as open source: A powerful
pathway to new drugs. Angew. Chem. Int. Ed. Engl. 56, 1164–1165 (2017).
4. R. A. Goodnow Jr., C. E. Dumelin, A. D. Keefe, DNA-encoded chemistry: Enabling the
deeper sampling of chemical space. Nat. Rev. Drug Discov. 16, 131–147 (2017).
5. L. H. Yuen et al., A focused DNA-encoded chemical library for the discovery of in-
hibitors of NAD+-dependent enzymes. J. Am. Chem. Soc. 141, 5169–5181 (2019).
6. S. Wang et al., Optimization of ligands using focused DNA-encoded libraries to de-
velop a selective, cell-permeable CBX8 chromodomain inhibitor. ACS Chem. Biol. 15,
112–131 (2020).
7. C. Zambaldo, J.-P. Daguer, J. Saarbach, S. Barluengaa, N. Winssinger, Screening for
covalent inhibitors using DNA-display of small molecule libraries functionalized with
cysteine reactive moieties. MedChemComm 7, 1340–1351 (2016).
8. C. J. Stress, B. Sauter, L. A. Schneider, T. Sharpe, D. Gillingham, A DNA-encoded
chemical library incorporating elements of natural macrocycles. Angew. Chem. Int.
Ed. Engl. 58, 9570–9574 (2019).
22. K. Kobayashi et al., Prss55 but not Prss51 is required for male fertility in mice. Biol.
Reprod., 10.1093/biolre/ioaa041 (2020).
23. X. Shang et al., Serine protease PRSS55 is crucial for male mouse fertility via affecting
sperm migration and sperm-egg binding. Cell. Mol. Life Sci. 75, 4371–4384 (2018).
24. C. Shen et al., Prss37 is required for male fertility in the mouse. Biol. Reprod. 88, 123
(2013).
25. D. Kiyozumi et al., NELL2-mediated lumicrine signaling through OVCH2 is required
for male fertility. Science 368, 1132–1135 (2020).
26. D. A. Bachovchin, B. F. Cravatt, The pharmacological landscape and therapeutic po-
tential of serine hydrolases. Nat. Rev. Drug Discov. 11, 52–68 (2012).
27. J. Wagner, J. Kallen, C. Ehrhardt, J.-P. Evenou, D. Wagner, Rational design, synthesis,
and X-ray structure of selective noncovalent thrombin inhibitors. J. Med. Chem. 41,
3664–3674 (1998).
28. S. Okamoto et al., Potent inhibition of thrombin by the newly synthesized arginine
derivative No. 805. The importance of stereo-structure of its hydrophobic carbox-
amide portion. Biochem. Biophys. Res. Commun. 101, 440–446 (1981).
29. R. Kikumoto et al., Selective inhibition of thrombin by (2R,4R)-4-methyl-1-[N2-
[1,2,3,4-tetrahydro-8-quinolinyl)sulfonyl]-L-arginyl]-2-piperidinecarboxylic acid. Bio-
chemistry 23, 85–90 (1984).
9. P. Ma et al., Functionality-independent DNA encoding of complex natural products.
Angew. Chem. Int. Ed. Engl. 58, 9254–9261 (2019).
10. T. Klein, U. Eckhard, A. Dufour, N. Solis, C. M. Overall, Proteolytic cleavage—
Mechanisms, function, and “omic” approaches for
a near-ubiquitous post-
translational modification. Chem. Rev. 118, 1137–1168 (2018).
11. G. Fear, S. Komarnytsky, I. Raskin, Protease inhibitors and their peptidomimetic de-
rivatives as potential drugs. Pharmacol. Ther. 113, 354–368 (2007).
12. F. M. Franco et al., Structure-based discovery of small molecule hepsin and HGFA
protease inhibitors: Evaluation of potency and selectivity derived from distinct
binding pockets. Bioorg. Med. Chem. 23, 2328–2343 (2015).
30. L. Mannocci et al., Isolation of potent and specific trypsin inhibitors from a DNA-
encoded chemical library. Bioconjug. Chem. 21, 1836–1841 (2010).
31. H. Deng et al., Discovery of highly potent and selective small molecule ADAMTS-5
inhibitors that inhibit human cartilage degradation via encoded library technology
(ELT). J. Med. Chem. 55, 7061–7079 (2012).
13. K. R. Acharya, E. D. Sturrock, J. F. Riordan, M. R. W. Ehlers, Ace revisited: A new target
for structure-based drug design. Nat. Rev. Drug Discov. 2, 891–902 (2003).
14. A. Straub, S. Roehrig, A. Hillisch, Oral, direct thrombin and factor Xa inhibitors: The
replacement for warfarin, leeches, and pig intestines? Angew. Chem. Int. Ed. Engl. 50,
4574–4590 (2011).
15. E. A. Nutescu, A. K. Wittkowsky, Direct thrombin inhibitors for anticoagulation. Ann.
Pharmacother. 38, 99–109 (2004).
16. S. G. Deeks, M. Smith, M. Holodniy, J. O. Kahn, HIV-1 protease inhibitors. A review for
clinicians. JAMA 277, 145–153 (1997).
17. A. A. Agbowuro, W. M. Huston, A. B. Gamble, J. D. A. Tyndall, Proteases and protease
inhibitors in infectious diseases. Med. Res. Rev. 38, 1295–1331 (2018).
18. A. Eatemadi et al., Role of protease and protease inhibitors in cancer pathogenesis
and treatment. Biomed. Pharmacother. 86, 221–231 (2017).
19. L. Zhang et al., Crystal structure of SARS-CoV-2 main protease provides a basis for
design of improved α-ketoamide inhibitors. Science 368, 409–412 (2020).
32. H.-C. Du et al., DNA-compatible nitro reduction using B2(OH)4. Org. Lett. 21,
2194–2199 (2019).
33. H.-C. Du, H. Huang, DNA-compatible nitro reduction and synthesis of benzimidazoles.
Bioconjug. Chem. 28, 2575–2580 (2017).
34. A. L. Satz et al., DNA compatible multistep synthesis and applications to dna encoded
libraries. Bioconjug. Chem. 26, 1623–1632 (2015).
35. M. S. Bernatowicz, Y. Wu, G. R. Matsueda, 1H-Pyrazole-1-carboxamidine hydrochlo-
ride an attractive reagent for guanylation of amines and its application to peptide
synthesis. J. Org. Chem. 57, 2497–2502 (1992).
36. J. C. Faver et al., Quantitative comparison of enrichment from DNA-encoded chemical
library selections. ACS Comb. Sci. 21, 75–82 (2019).
37. D. Antermite, D. P. Affron, J. A. Bull, Regio- and stereoselective palladium-catalyzed
C(sp3)–H arylation of pyrrolidines and piperidines with C(3) directing groups. Org.
Lett. 20, 3948–3952 (2018).
8 of 8
|
www.pnas.org/cgi/doi/10.1073/pnas.2005447117
Dawadi et al.