Bioorganic & Medicinal Chemistry Letters 25 (2015) 438–443

Contents lists available at ScienceDirect

Bioorganic & Medicinal Chemistry Letters

journal homepage: www.elsevier.com/locate/bmcl

Structure-based design and optimization of potent inhibitors

of the adenoviral protease
Philipp Grosche a, Finton Sirockin a, Aengus Mac Sweeney a, Paul Ramage a, Paul Erbel a, Samu Melkko a,
Anna Bernardi a, Nicola Hughes a, David Ellis b, Keith D. Combrink b, Nadine Jarousse c, Eva Altmann a,⇑
a
Novartis Institute for Biomedical Research, Novartis Campus, CH-4002 Basel, Switzerland
b
Novartis Institute for Biomedical Research, 6201 South Freeway, Fort Worth, TX 76134-2099, United States
c
Novartis Institute for Biomedical Research, 4560 Horton Street, Emeryville, CA 94608-2916, United States

a r t i c l e i n f o a b s t r a c t

Article history: Adenoviral infections are associated with a wide range of acute diseases, among which ocular viral con-
Received 14 November 2014 junctivitis (EKC) and disseminated disease in immunocompromised patients. To date, no approved spe-
Revised 15 December 2014 cific anti-adenoviral drug is available, but there is a growing need for an effective treatment of such
Accepted 16 December 2014
infections. The adenoviral protease, adenain, plays a crucial role for the viral lifecycle and thus represents
Available online 24 December 2014
an attractive therapeutic target. Structure-guided design with the objective to depeptidize tetrapeptide
nitrile 1 led to the novel chemotype 2. Optimization of scaffold 2 resulted in picomolar adenain inhibitors
Keywords:
3a and 3b. In addition, a complementary series of irreversible vinyl sulfone containing inhibitors were
Adenoviral protease inhibitor
Irreversible inhibitor
rationally designed, prepared and evaluated against adenoviral protease. High resolution X-ray co-crystal
Peptidomimetic structures of representatives of each series proves the successful design of these inhibitors and provides
X-ray co-crystal structure an excellent basis for future medicinal chemistry optimization of these compounds.
Epidemic keratoconjunctivitis (EKC) Ó 2014 Elsevier Ltd. All rights reserved.

Adenoviruses are non-enveloped, double-stranded DNA viruses
with more than 55 serotypes identified that have been classified
into seven subgroups (A–G).1 Specific serotypes are linked to a
number of acute, highly contagious infections such as gastrointes-
tinal illness, disseminated adenovirus disease in immunocompro-
mised patients and ocular conjunctivitis such as epidemic
keratoconjunctivitis (EKC).2 The occurrence of severe adenoviral
infections is increasing due to the growing numbers of transplan-
tations. To date, there are no approved anti-adenoviral therapeu-
tics available, which contrasts with a growing unmet medical
need for specific anti-adenoviral drugs, in particular for immuno-
suppressed patients or for an effective treatment of EKC.3–5
Adenoviruses, as all viruses, encode their own protease and
depend on its activity for the development of infectivity. The pro-
tease of adenovirus, the cysteine protease adenain6,7 plays a key
role in several steps of the viral life cycle.8 Adenain is involved in
the uncoating of virus particles during viral entry,9 it is essential
for the cleavage of seven core precursor proteins that are required
for the formation of mature virions10 and it is also implicated in the
cleavage of cytokeratins, which leads to host cell lysis.11,12 In addi-
tion, adenain represents an attractive therapeutic target as it lacks
⇑ Corresponding author. Tel.: +41 (0)61 696 1227.
E-mail address: eva.altmann@novartis.com (E. Altmann).
human homologues. Thus, selective inhibitors of the adenoviral
protease may offer efficacious treatments for adenoviral infections.
The increasing need for specific and effective anti-adenoviral
therapies prompted us to engage in the development of inhibitors
of adenain, in particular for the treatment of EKC caused by sero-
types 8, 37, and 64. We recently disclosed our initial efforts in this
area, which led to the discovery and structure-based optimization
of glycine- and pyrimidine nitrile inhibitors of the adenoviral pro-
tease. Among other structures our previous hit discovery campaign
identified tetrapeptide nitrile 1 as a potent inhibitor of adenain.13
As part of our ongoing efforts to optimize this compound we aimed
at further reducing the peptidic character of 1 (Fig. 1) by designing
a novel replacement of the P4–P3 amide bond and by investigating
irreversible inhibitors.
While tetrapeptide nitrile 1 proved to be a low nM inhibitor of
adenain in vitro, it was inactive in a viral replication assay. It was
thus the primary objective of our work to develop compounds with
increased permeability and, hence, improved cellular potency.
Guided by the X-ray co-crystal structure of adenain and bound 1
we designed a novel peptidomimetic scaffold, where the NH–
CH(CH2Ph) unit of the central Phe residue is replaced by a meta-
substituted phenyl ring. Molecular modelling suggested that the
original P4–P3 binding mode could be conserved by introducing
an ethanone linker between this phenyl moiety and the P4 phenyl
group together with an ortho-substituent on the central phenyl

http://dx.doi.org/10.1016/j.bmcl.2014.12.057
0960-894X/Ó 2014 Elsevier Ltd. All rights reserved.
 P. Grosche et al. / Bioorg. Med. Chem. Lett. 25 (2015) 438–443 439

P4 P3 P2 P1 addition, the modelling also suggested that for proper positioning

IC50 AVP8 = 40 nM of the 3,5-dichlorophenyl moiety into the S4 subsite, gem-
dimethyl substitution at the benzylic position would be highly
IC50 AVP5 = 30 nM
advantageous. The 3,5-dichloro substitution at the P4 phenyl ring
O O Log PAMPA = -5.6 enhances hydrophobic interactions in S4, as we have already
O H H
N N shown in our previous work.13
S N N 2
H H N PSA = 156 Å Figure 3 illustrates the binding model of our newly designed
O O
MW = 505 peptide mimetic 2 to adenain and its overlay with the X-ray co-
Cl crystal structure of 1.
The SAR established for this new template 2 is summarized in
Figure 1. Structure of tetrapeptide nitrile inhibitor 1 and profiling data. Table 1. Compound 2a, which does not bear a substituent at the
benzylic position of the P4 group (R1, R2 = H), is only a 3 lM inhib-
itor of adenain 8 (AVP8). Attachment of one methyl group at this
P4 P3 P2 P1 position gave derivative 2b, which displayed 3-fold improved
R3 potency. Importantly, and quite intriguingly, disubstitution at the
O benzylic position as in compound 2c resulted in an IC50 of 40 nM,
R2 O which makes this analog equipotent with the initial peptide nitrile
R1 H
Cl N 1. This outcome is in accordance with the original modelling pre-
N
H N dictions; most likely, the gem-dimethyl substitution locks inhibitor
O O
2c in the preferred conformation for binding to the enzyme. Com-
Cl pound 2c combined high potency with good permeability (Table 1),
2
but it exhibited rather low solubility. Inspection of the structure
Figure 2. New designed scaffold 2. indicated that larger alkoxy groups than methoxy should be toler-
ated at the central phenyl ring, thus offering an option to introduce
solubilizing groups. This idea resulted in the design of compound
2d, which maintained the potency of 2c and displayed greatly
improved solubility.
Unfortunately, when tested in our anti-viral cytopathic effect
(CPE) assay both compounds 2c and 2d exhibited only weak
activity.
The X-ray co-crystal structure of inhibitor 2c bound in the
active site of adenoviral protease 8 (AVP8) is depicted in Figure 4
and provides experimental proof for the predicted binding mode.
The inhibitor is covalently bound to the enzyme by thioimidate
formation between the cyano group and the active site cysteine
122. The thioimidate NH forms a hydrogen bond with the carbonyl
oxygen of the side chain of glutamine 115. The new scaffold links
Figure 3. Superposition of the putative docking pose of 2c (blue) with the X-ray co- positions P2 through P4 in the non-primed part of the binding cleft
crystal structure of 1 (green, PDB code 4PIE). The picture was produced using
and displays four hydrogen bonding interactions with the active
PyMol.
site of adenain.
To further depeptidize scaffold 2 we replaced the P2–P1 unit by
ring. This led to structures of type 2 (Fig. 2) as targets for synthesis a 4-aminomethyl-2-cyano-pyrimidine, following a similar logic as
and biological evaluation. The keto group of the ethanone linker for the optimization of a different lead series.13 The resulting achi-
was predicted to form a hydrogen bond with the amide NH of ral peptidomimetics 3a and 3b are both exceptionally potent
Glu5, while the ortho-substituent would force the carbonyl group inhibitors of adenain (Table 2) only incorporating a single amide
out of the plane of the phenyl ring for optimal orientation. In bond. Figure 5 captures the X-ray co-crystal structure of inhibitor

Table 1
SAR and profiling data for inhibitors 2a–d

R3
O
R2 O
R1 H
Cl N
N
H N
O O

Cl
1 2 3
Compd R R R IC50 AVP8a [lM] IC50 AVP5a [lM] Log PAMPA [10 6
cm s 1
] Solb [mM] PSA [Å2]
c c
2a H H CH3 2.74 1.1 nd nd 108
2b H CH3 CH3 0.95 0.31 ndc ndc 108
2c CH3 CH3 CH3 0.04 0.03 5.0 0.05 108
2d CH3 CH3 (CH3)2N(CH3)2 0.01 0.007 5.5 >1 111
a
Inhibition of AVP8 and AVP514 in an FLT assay using Ac-WLRGG*ARC(PT14)-NH2 as substrate. Data represent mean of at least 2 experiments performed in duplicate.
Individual data points in each experiment were within 2-fold range of each other.
b
High throughput equilibrium solubility at pH 6.8.
c
Not determined.
440 P. Grosche et al. / Bioorg. Med. Chem. Lett. 25 (2015) 438–443

Ser3 Ser 3
Glu5 N-term Glu 5
N-term

Gln 115
Cys122
Arg 48
Arg48 Asn44
Asn 44

Figure 5. X-ray co-crystal structure of pyrimidine nitrile 3a in AVP8 (PDB code

Figure 4. X-ray co-crystal structure of inhibitor 2c in AVP8 (PDB code 4WX7). The
4WX4). The picture was produced using PyMol.
picture was produced using PyMol.

docking utilizing the same adenain X-ray structure. After inspection

3a bound in the active site of AVP8; the compound exhibits an
of the poses generated by GOLD, the vinyl sulfone derivative was
identical binding mode as the glycine nitrile derivative 2a.
predicted to offer optimal positioning of the polar groups and acco-
For compound 3b the favorable combination of high permeabil-
modation of the ligand in the protease.
ity and good solubility led to an EC50 of 3.9 lM in the CPE assay,
The profiling data for a selection of the resulting vinyl sulfone-
which was an encouraging finding. Nevertheless, a significant
containing analogs of peptide nitrile 1 (compd 4–7) are summarized
potency shift remained between biochemical and cellular assays
in Table 3. Out of this group of compounds, analog 4 is least peptidic
and closing this gap still posed a challenge even for the less pep-
in character (smallest number of amide bonds). This compound
tidic scaffold. Protein binding was not the cause for the observed
proved to be a 1 nM inhibitor of adenain and it showed good perme-
potency shift in the cellular assay, as enzyme inhibition by 3a or
ability in the Pampa assay, but unfortunately, it also displayed sig-
3b in the biochemical assay was reduced no more than 10-fold
nificant cytotoxicity in HEK293 cells. For those analogs based on a
in the presence of 10% serum proteins.
contiguous peptide backbone,13 compound 5 exhibited an EC50 of
The new series of highly potent adenain inhibitors described in
4.5 lM in the CPE assay. Irreversible binding of 5 was demonstrated
Tables 1 and 2 exhibit an exquisite specificity profile (>1700-fold
by a rapid dilution experiment17 in which inhibitor 5 was preincu-
for AVP8 over a panel of five human cathepsins). These data are
bated with adenain at high concentration to ensure complete loss of
summarized in the Supplementary data.
activity. After dialysis and dilution into assay buffer containing sub-
Lack of potent cellular activity might be associated with insuffi-
strate, the protease activity was not restored (Figure 6).
cient duration of inhibition of the target. Thus, one of our strategies
An attempt to increase cellular activity by attaching an amino-
to enhance cellular potency was to pursue irreversible inhibition of
ethyl substituent at the benzylic position of the P4 group (compd
adenain. Once again these efforts were guided by in silico studies,
6) was unsuccessful, in spite of an improvement of the IC50 in
which served to evaluate a set of potential irreversible active prin-
the biochemical assay. Most likely, this finding is due to the
ciples. These warheads were attached to a simplified version of
increased PSA of 6 compared to 5. With the aim to significantly
compound 1 containing neither the P4 sulfonamide capping group
reduce PSA we incorporated a P3–P2 lactam moiety and removed
nor the benzyl side chain of the P3 residue. An acyloxymethyl
the terminal N-atom in 6, leading to inhibitor 7. While the latter
ketone, an a-halomethyl ketone, an epoxide, a vinyl sulfone or an
exhibited single digit nanomolar inhibition of AVP8 and AVP5 in
acrylic ester group were attached to this simplified scaffold as
the biochemical assay, it was only a micromolar inhibitor in the
replacements of the electrophilic nitrile moiety. After minimization
CPE assay.
of these potential covalent inactivators in the active site of adenain
Despite compounds 4–7 incorporate an irreversible active prin-
(PDB code 4PIE) using ICM software,15 the presence of key hydrogen
ciple they all showed excellent specificity toward a panel of five
bonds and evaluation of the ligand internal constraints were used
human cysteine proteases; these results are summarized in the
as a selection criterion for further evaluation. A subset of these
Supplementary data.
virtual compounds was selected for submission to GOLD16 covalent

Table 2
SAR and profiling data for pyrimidine nitrile inhibitors 3a–b

R
O
N
H
Cl N
N
N
O O

Cl
a
Compd R IC50AVP8 [lM] IC50 AVP5a [lM] EC50 CPEb [lM] Log PAMPA [10 6
cm s 1
] Solc [mM] PSA [Å2]
d d
3a CH3 0.0006 0.0002 >50 nd nd 105
3b (CH3)2N(CH3)2 0.0001 0.0001 3.9 4.2 0.49 108
a
Inhibition of AVP8 and AVP5 in an FLT assay using Ac-WLRGG*ARC(PT14)-NH2 as substrate. Data represent mean of at least 2 experiments performed in duplicate.
Individual data points in each experiment were within 2-fold range of each other.
b
HEK293 cells.
c
High throughput equilibrium solubility at pH 6.8.
d
Not determined.
 P. Grosche et al. / Bioorg. Med. Chem. Lett. 25 (2015) 438–443 441

Table 3
SAR and profiling data for vinyl sulfone inhibitors 4–7

Compd Structures IC50 AVP8a [lM] IC50 AVP5a [lM] EC50 CPEb [lM] Log PAMPA [10 6
cm s 1
] PSA [Å2]

Cl
O O H
4 N S O 0.001 0.001 ndc 4.1 118
Cl N
H O
O
O

O H O
N O
N N S
5 H H O 0.01 0.01 4.5 5.6 121
O

Cl Cl

H O H O
N N O
N N S
6 H H O <0.001 <0.001 31 ndd 133
O

Cl Cl
O O
N O
N N S
7 H O H O 0.004 0.002 8.9 5.5 112

Cl Cl
a
Inhibition of AVP8 and AVP5 in an FLT assay using Ac-WLRGG*ARC(PT14)-NH2 as substrate. Compounds were pre-incubated with AVP8 and AVP5, respectively, for 1 h at
23 °C, then the reaction was started by the addition of substrate solution and the IC50 values were determined after 1 h reaction time. Complete experimental details are given
in the Supplementary material. Data represent mean of at least 2 experiments performed in duplicate. Individual data points in each experiment were within 2-fold range of
each other.
b
HEK293 cells.
c
Compound displayed cytotoxicity in HEK293 cells.
d
Not determined.

9500 Ser4
substrate control Ser3
enzyme control
Glu5
Fluorescence life time [ps]

9000 compound 5

Cys122
8500
Gln115
Arg 48
8000

Gly51 Val53
7500

7000
0 50 100
time (min)
Figure 6. Rapid dilution assay. Time course of substrate cleavage: compound 5
(blue triangles, n = 4); substrate control (black circles, n = 16). Enzyme control
shows the uninhibited reaction (red squares, n = 16). Error bars indicate standard
error of the mean.
The X-ray co-crystal structure of 6 captured in Figure 7 confirms
the covalent nature of the complex (PDB code 4WX6).
The active site Cys122 is irreversibly tethered via a thioether to
the carbon distal to the SO2 group of the vinyl sulfone moiety. In
addition, the inhibitor has extensive non-covalent interactions
with the substrate binding site. Specifically, both oxygen atoms
of the SO2 group are involved in hydrogen bonding interactions
with the side chain of Gln115 and with the phenolic hydroxyl
group of Tyr 84.
The preparation of compounds of type 2 and 3 is exemplified in
Scheme 1 for derivative 2c.
A Claisen condensation and in situ decarboxylation reaction of
methyl 5-cyano-2-methoxy-benzoate 8 with 3,5-dichlorophenyl-
acetic acid 9 in the presence of NaHDMS led to aryl ketone 10.18
The geminal dimethylation to obtain compound 11 was achieved
Tyr84
150 200
Figure 7. X-ray co-crystal structure of vinyl sulfone inhibitor 6 in AVP8 (PDB code
4WX6). The picture was produced using PyMol.
with CH3I and KOtBu. Acid hydrolysis of the nitrile moiety (HCl)
and subsequent TBTU-mediated coupling of the resulting acid 12
with 2-amino-N-(cyanomethyl)-acetamide yielded the target mol-
ecule 2c.
The synthesis of the vinyl sulfone-based inhibitors listed in
Table 3 followed the route exemplified in Scheme 2 for compound 5.
Starting from commercially available Boc-L-a-tert-butyl-glycine
13 and glycine methyl ester 14 we prepared intermediate 15 by
HBTU-mediated coupling. Treatment of 15 with 4 M HCl in dioxane
afforded amine 16. Subsequent HBTU-mediated coupling of 3,5-
dichlorophenylacetic acid 17 and amine 16 yielded intermediate
18. Methyl ester hydrolysis with LiOH followed by coupling of
the resulting acid with amine 20 then gave target molecule 5.
In conclusion, we have developed a novel class of highly potent
and specific non-peptidic inhibitors of the adenoviral protease by
depeptidization of tetrapeptide nitrile 1. Guided by the X-ray
co-crystal structure of adenain and 1, molecular modelling studies
led to the design of a novel peptidomimetic scaffold. Further
442 P. Grosche et al. / Bioorg. Med. Chem. Lett. 25 (2015) 438–443

O
O Cl OH
a Cl b
O
+ O CN
CN O
Cl
O
8 9 Cl 10

O O

Cl c Cl OH d
CN
O O O

Cl Cl
11 12

O
O
H
Cl N
N CN
H
O O

Cl
2c

Scheme 1. Reagents and conditions: (a) NaHDMS, DMF, toluene, 20 ? 10 °C, 62%; (b) CH3I, KOtBu, THF, rt; (c) 6 M aq HCl, AcOH, 100 °C, 27%; (d) 2-amino-N-
(cyanomethyl)-acetamide, TBTU, DIEA, DMF, rt, 75%.

O O O O
a H b
O + H N N
O N 2 O N O
O
H H
OH O
13 14 15

O O O
H
O OH N
H c N O d
N + H
H 2N O O
O
Cl Cl Cl Cl
16 17 18

O O O O
H H O
N O N
N OH e N N S
H + H 2N S H H
O O O
O

Cl Cl Cl Cl
19 20 5

Scheme 2. Reagents and conditions: (a) HBTU, DIEA, CH2Cl2, rt, 81%; (b) 4 M HCl in dioxane, rt, 99%; (c) HBTU, DIEA, CH2Cl2, rt, 56%; (d) LiOH
H2O, THF/H2O (6:1) rt, 98%, (e)
HBTU, DIEA, CH2Cl2, rt, 67%.

optimization of this template led to picomolar achiral inhibitors, Supplementary data

which incorporate only one amide bond and displayed improved
permeability compared to the peptidic starting point 1. X-ray co- Supplementary data associated with this article can be found, in
crystal structures of inhibitors 2c and 3a bound to AVP8 were in the online version, at http://dx.doi.org/10.1016/j.bmcl.2014.12.
excellent agreement with the predicted binding mode. Finally, 057.
we have successfully incorporated a vinyl sulfone moiety as reac-
tive principle in our inhibitor templates. The information provided References and notes
by the X-ray co-crystal structures for both inhibitor series have
important implications for further medicinal chemistry optimiza- 1. http://www.vmri.hu/~harrach/AdVtaxlong.htm.
2. Lenaerts, L.; De Clercq, E.; Naesens, L. Rev. Med. Virol. 2008, 18, 357.
tion of inhibitors specific for adenain. 3. Kaufman, H. E. Curr. Opin. Ophthalmol. 2011, 22, 290.
4. Zhou, X.; Robinson, C. M.; Rajaiya, J.; Seto, D.; Jones, M. S.; Dyer, D. W.;
Acknowledgments Chodosh, J. Invest. Ophthalmol. Vis. Sci. 2012, 53, 2804.
5. Pihos, A. M. J. Optometry 2013, 6, 69.
6. Weber, J. M. Acta Microbiol. Immunol. Hung. 2003, 50, 95.
The authors thank Estelle Ngo, Raphael Gattlen, Simone Schlee- 7. McGrath, W. J.; Baniecki, M. L.; Peters, E.; Green, D. T.; Mangel, W. F.
ger and Mina Aikawa for excellent technical assistance. Biochemistry 2001, 40, 14468.
 P. Grosche et al. / Bioorg. Med. Chem. Lett. 25 (2015) 438–443
8. Greber, U. F. Rev. Med. Virol. 1998, 8, 213.
9. Greber, U. F.; Webster, P.; Weber, J.; Helenius, A. EMBO J. 1996, 15, 1766.
10. Pérez-Berná, A. J.; Mangel, W. F.; McGrath, W. J.; Graziano, V.; Flint, J.; San
Martín, C. J. Virol. 2014, 88, 1513.
11. Chen, P. H.; Ornelles, D. A.; Shenk, T. J. Virol. 1993, 67, 3507.
12. Mangel, W. F.; Banjecki, M. L.; McGrath, W. J. Cell. Mol. Life Sci. 2003, 60,
2347.
13. Mac Sweeney, A.; Grosche, P.; Ellis, D.; Combrink, K.; Erbel, P.; Hughes, N.;
Sirockin, F.; Melkko, S.; Bernardi, A.; Jarousse, N.; Altmann, E. ACS Med. Chem.
Lett. 2014, 5, 937.
443
14. These subtypes were selected for testing, because AVP8 is the relevant protease
for EKC, while inhibition of AVP5 is required for assessment of in vivo activity
in a rabbit based disease model. Rabbits are not infected by AVP8, but only by
AVP5. AVP8 and AVP5 have 76% sequence identity and 88% sequence similarity
(BLAST).
15. Abagyan, R.; Totrov, M.; Kuznetsov, D. J. Comp. Chem. 1994, 15, 488.
16. Jones, G.; Willett, R.; Glen, R. C. J. Mol. Biol. 1995, 245, 43.
17. Copeland, R. A. In Evaluation of Enzyme Inhibitors in Drug Discovery; Copeland, R.
A., Ed.; John Wiley and Sons: Hoboken, NJ, 2005; p 125.
18. Wu, G.; Yin, W.; Hong, C.; Shen, H. C.; Huang, Y. Green Chem. 2012, 41, 580.